No OneTemporary
Actions

Size

6 MB

Referenced Files

None

Subscribers

None

View Options

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

	diff --git a/cddl/lib/libspl/Makefile b/cddl/lib/libspl/Makefile
	index e432b313d9d0..6c43e1750c3c 100644
	--- a/cddl/lib/libspl/Makefile
	+++ b/cddl/lib/libspl/Makefile
	@@ -1,49 +1,50 @@

	.include <bsd.init.mk>
	.include <bsd.compiler.mk>

	.PATH: ${SRCTOP}/sys/contrib/openzfs/lib/libspl
	.PATH: ${SRCTOP}/sys/contrib/openzfs/include

	LIB= spl
	LIBADD=
	PACKAGE= zfs

	SRCS = \
	assert.c \
	+ backtrace.c \
	list.c \
	mkdirp.c \
	os/freebsd/zone.c \
	page.c \
	timestamp.c \
	include/sys/list.h \
	include/sys/list_impl.h

	# These functions are not required when bootstrapping and the atomic code
	# will not compile when building on macOS.
	.if !defined(BOOTSTRAPPING)
	SRCS += \
	atomic.c \
	getexecname.c \
	os/freebsd/getexecname.c \
	os/freebsd/gethostid.c \
	os/freebsd/getmntany.c \
	os/freebsd/mnttab.c
	.endif

	WARNS?= 2
	CSTD= c99
	CFLAGS+= -DIN_BASE
	CFLAGS+= -I${SRCTOP}/sys/contrib/openzfs/include
	CFLAGS+= -I${SRCTOP}/sys/contrib/openzfs/lib/libspl/include/
	CFLAGS+= -I${SRCTOP}/sys/contrib/openzfs/lib/libspl/include/os/freebsd
	CFLAGS+= -I${SRCTOP}/cddl/compat/opensolaris/include
	CFLAGS+= -I${SRCTOP}/sys/contrib/openzfs/module/icp/include
	CFLAGS+= -include ${SRCTOP}/sys/contrib/openzfs/include/os/freebsd/spl/sys/ccompile.h
	CFLAGS+= -DHAVE_ISSETUGID
	CFLAGS+= -include ${SRCTOP}/sys/modules/zfs/zfs_config.h
	.if ${COMPILER_TYPE} == "clang"
	CFLAGS.atomic.c+= -Wno-error=atomic-alignment
	.endif

	.include <bsd.lib.mk>
	diff --git a/sys/contrib/openzfs/cmd/zdb/zdb.c b/sys/contrib/openzfs/cmd/zdb/zdb.c
	index 449b6bf2ccb3..704fcf4422d4 100644
	--- a/sys/contrib/openzfs/cmd/zdb/zdb.c
	+++ b/sys/contrib/openzfs/cmd/zdb/zdb.c
	@@ -1,9474 +1,9544 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2019 by Delphix. All rights reserved.
	* Copyright (c) 2014 Integros [integros.com]
	* Copyright 2016 Nexenta Systems, Inc.
	* Copyright (c) 2017, 2018 Lawrence Livermore National Security, LLC.
	* Copyright (c) 2015, 2017, Intel Corporation.
	* Copyright (c) 2020 Datto Inc.
	* Copyright (c) 2020, The FreeBSD Foundation [1]
	*
	* [1] Portions of this software were developed by Allan Jude
	* under sponsorship from the FreeBSD Foundation.
	* Copyright (c) 2021 Allan Jude
	* Copyright (c) 2021 Toomas Soome <tsoome@me.com>
	* Copyright (c) 2023, Klara Inc.
	* Copyright (c) 2023, Rob Norris <robn@despairlabs.com>
	*/

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

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

	#include <libzdb.h>

	#include "zdb.h"


	extern int reference_tracking_enable;
	extern int zfs_recover;
	extern uint_t zfs_vdev_async_read_max_active;
	extern boolean_t spa_load_verify_dryrun;
	extern boolean_t spa_mode_readable_spacemaps;
	extern uint_t zfs_reconstruct_indirect_combinations_max;
	extern uint_t zfs_btree_verify_intensity;

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

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

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


	static zopt_object_range_t *zopt_object_ranges = NULL;
	static unsigned zopt_object_args = 0;

	static int flagbits[256];


	static uint64_t max_inflight_bytes = 256 * 1024 * 1024; /* 256MB */
	static int leaked_objects = 0;
	static range_tree_t *mos_refd_objs;
	+static spa_t *spa;
	+static objset_t *os;
	+static boolean_t kernel_init_done;

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



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

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

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

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

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

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

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

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

	return (0);
	}

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

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

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

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

	return (err);
	}

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

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

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

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

	return (0);
	}

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

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

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

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

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

	typedef void ll_iter_t(dsl_deadlist_t ll, void arg);

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

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

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

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

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

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

	static void
	verify_livelist_allocs(metaslab_verify_t *mv, uint64_t txg,
	uint64_t offset, uint64_t size)
	{
	sublivelist_verify_block_t svb = {{{0}}};
	DVA_SET_VDEV(&svb.svb_dva, mv->mv_vdid);
	DVA_SET_OFFSET(&svb.svb_dva, offset);
	DVA_SET_ASIZE(&svb.svb_dva, size);
	zfs_btree_index_t where;
	uint64_t end_offset = offset + size;

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

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

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

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

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

	vdev_t *vd = vdev_lookup_top(spa, vdev_id);

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

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

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

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


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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

	if (!vdev_is_concrete(vd))
	continue;

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

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

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

	mv_populate_livelist_allocs(&mv, &sv);

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

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

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

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

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

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

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

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

	static void
	dump_debug_buffer(void)
	{
	- if (dump_opt['G']) {
	- (void) printf("\n");
	- (void) fflush(stdout);
	- zfs_dbgmsg_print("zdb");
	- }
	+ ssize_t ret __attribute__((unused));
	+
	+ if (!dump_opt['G'])
	+ return;
	+ /*
	+ * We use write() instead of printf() so that this function
	+ * is safe to call from a signal handler.
	+ */
	+ ret = write(STDERR_FILENO, "\n", 1);
	+ zfs_dbgmsg_print(STDERR_FILENO, "zdb");
	+}
	+
	+static void sig_handler(int signo)
	+{
	+ struct sigaction action;
	+
	+ libspl_backtrace(STDERR_FILENO);
	+ dump_debug_buffer();
	+
	+ /*
	+ * Restore default action and re-raise signal so SIGSEGV and
	+ * SIGABRT can trigger a core dump.
	+ */
	+ action.sa_handler = SIG_DFL;
	+ sigemptyset(&action.sa_mask);
	+ action.sa_flags = 0;
	+ (void) sigaction(signo, &action, NULL);
	+ raise(signo);
	}

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

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

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

	dump_debug_buffer();

	- exit(1);
	+ zdb_exit(1);
	}

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

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

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

	umem_free(packed, nvsize);

	dump_nvlist(nv, 8);

	nvlist_free(nv);
	}

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

	if (shp == NULL)
	return;

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

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

	static void
	zdb_nicebytes(uint64_t bytes, char *buf, size_t buflen)
	{
	if (dump_opt['P'])
	(void) snprintf(buf, buflen, "%llu", (longlong_t)bytes);
	else
	zfs_nicebytes(bytes, buf, buflen);
	}

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

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

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

	if (max < histo_width)
	max = histo_width;

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

	if (data == NULL) {
	dmu_object_info_t doi;

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

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

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

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

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

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

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

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

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

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

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

	if (bpop == NULL)
	return;

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

	return (refcount);
	}

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

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

	return (refcount);
	}

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

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

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

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

	return (refcount);
	}

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

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

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

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

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

	if (sm == NULL)
	return;

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

	metaslab_class_histogram_verify(mc);

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

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

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

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

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

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

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

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

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

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

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

	uint32_t *counts = vdev_indirect_mapping_load_obsolete_counts(vim);

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

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

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

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

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

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

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

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

	print_vdev_indirect(vd);

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

	ddt_object_name(ddt, type, class, name);

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

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

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

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

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

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

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

	ASSERT3U(error, ==, ENOENT);

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

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

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

	ddt_get_dedup_stats(spa, &dds_total);

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

	(void) printf("\n");

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

	dump_dedup_ratio(&dds_total);
	}

	static void
	dump_brt(spa_t *spa)
	{
	if (!spa_feature_is_enabled(spa, SPA_FEATURE_BLOCK_CLONING)) {
	printf("BRT: unsupported on this pool\n");
	return;
	}

	if (!spa_feature_is_active(spa, SPA_FEATURE_BLOCK_CLONING)) {
	printf("BRT: empty\n");
	return;
	}

	brt_t *brt = spa->spa_brt;
	VERIFY(brt);

	char count[32], used[32], saved[32];
	zdb_nicebytes(brt_get_used(spa), used, sizeof (used));
	zdb_nicebytes(brt_get_saved(spa), saved, sizeof (saved));
	uint64_t ratio = brt_get_ratio(spa);
	printf("BRT: used %s; saved %s; ratio %llu.%02llux\n", used, saved,
	(u_longlong_t)(ratio / 100), (u_longlong_t)(ratio % 100));

	if (dump_opt['T'] < 2)
	return;

	for (uint64_t vdevid = 0; vdevid < brt->brt_nvdevs; vdevid++) {
	brt_vdev_t *brtvd = &brt->brt_vdevs[vdevid];
	if (brtvd == NULL)
	continue;

	if (!brtvd->bv_initiated) {
	printf("BRT: vdev %" PRIu64 ": empty\n", vdevid);
	continue;
	}

	zdb_nicenum(brtvd->bv_totalcount, count, sizeof (count));
	zdb_nicebytes(brtvd->bv_usedspace, used, sizeof (used));
	zdb_nicebytes(brtvd->bv_savedspace, saved, sizeof (saved));
	printf("BRT: vdev %" PRIu64 ": refcnt %s; used %s; saved %s\n",
	vdevid, count, used, saved);
	}

	if (dump_opt['T'] < 3)
	return;

	char dva[64];
	printf("\n%-16s %-10s\n", "DVA", "REFCNT");

	for (uint64_t vdevid = 0; vdevid < brt->brt_nvdevs; vdevid++) {
	brt_vdev_t *brtvd = &brt->brt_vdevs[vdevid];
	if (brtvd == NULL \|\| !brtvd->bv_initiated)
	continue;

	zap_cursor_t zc;
	zap_attribute_t za;
	for (zap_cursor_init(&zc, brt->brt_mos, brtvd->bv_mos_entries);
	zap_cursor_retrieve(&zc, &za) == 0;
	zap_cursor_advance(&zc)) {
	uint64_t offset = (uint64_t )za.za_name;
	uint64_t refcnt = za.za_first_integer;

	snprintf(dva, sizeof (dva), "%" PRIu64 ":%llx", vdevid,
	(u_longlong_t)offset);
	printf("%-16s %-10llu\n", dva, (u_longlong_t)refcnt);
	}
	zap_cursor_fini(&zc);
	}
	}

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

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

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

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

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

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

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

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

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

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

	do {
	len = SPA_OLD_MAXBLOCKSIZE;

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

	ASSERT(zb->zb_level >= 0);

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

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

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

	if (BP_IS_HOLE(bp))
	return;

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

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

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

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

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

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

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

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

	blkbuf[0] = '\0';

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

	if (BP_IS_HOLE(bp)) {
	(void) snprintf(blkbuf + strlen(blkbuf),
	buflen - strlen(blkbuf),
	"%llxL B=%llu",
	(u_longlong_t)BP_GET_LSIZE(bp),
	(u_longlong_t)BP_GET_LOGICAL_BIRTH(bp));
	} else {
	(void) snprintf(blkbuf + strlen(blkbuf),
	buflen - strlen(blkbuf),
	"%llxL/%llxP F=%llu B=%llu/%llu",
	(u_longlong_t)BP_GET_LSIZE(bp),
	(u_longlong_t)BP_GET_PSIZE(bp),
	(u_longlong_t)BP_GET_FILL(bp),
	(u_longlong_t)BP_GET_LOGICAL_BIRTH(bp),
	(u_longlong_t)BP_GET_BIRTH(bp));
	if (bp_freed)
	(void) snprintf(blkbuf + strlen(blkbuf),
	buflen - strlen(blkbuf), " %s", "FREE");
	(void) snprintf(blkbuf + strlen(blkbuf),
	buflen - strlen(blkbuf),
	" cksum=%016llx:%016llx:%016llx:%016llx",
	(u_longlong_t)bp->blk_cksum.zc_word[0],
	(u_longlong_t)bp->blk_cksum.zc_word[1],
	(u_longlong_t)bp->blk_cksum.zc_word[2],
	(u_longlong_t)bp->blk_cksum.zc_word[3]);
	}
	}

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

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

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

	ASSERT(zb->zb_level >= 0);

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

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

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

	if (BP_GET_LOGICAL_BIRTH(bp) == 0)
	return (0);

	print_indirect(spa, bp, zb, dnp);

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

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

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

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

	return (err);
	}

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

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

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

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

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

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

	if (dd == NULL)
	return;

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

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

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

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

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

	if (ds == NULL)
	return;

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

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

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

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

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

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

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

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

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

	(void) printf("\n");

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

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

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

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

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

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

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

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

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


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

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

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

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

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

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

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

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

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

	redact_block_phys_t *rbp_buf;
	uint64_t size;
	dmu_object_info_t doi;

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

	(void) putchar('\n');

	dsl_deadlist_iterate(dl, dsl_deadlist_entry_dump, NULL);
	}

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

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

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

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

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

	static char *key_material = NULL;

	static boolean_t
	zdb_derive_key(dsl_dir_t dd, uint8_t key_out)
	{
	uint64_t keyformat, salt, iters;
	int i;
	unsigned char c;

	VERIFY0(zap_lookup(dd->dd_pool->dp_meta_objset, dd->dd_crypto_obj,
	zfs_prop_to_name(ZFS_PROP_KEYFORMAT), sizeof (uint64_t),
	1, &keyformat));

	switch (keyformat) {
	case ZFS_KEYFORMAT_HEX:
	for (i = 0; i < WRAPPING_KEY_LEN * 2; i += 2) {
	if (!isxdigit(key_material[i]) \|\|
	!isxdigit(key_material[i+1]))
	return (B_FALSE);
	if (sscanf(&key_material[i], "%02hhx", &c) != 1)
	return (B_FALSE);
	key_out[i / 2] = c;
	}
	break;

	case ZFS_KEYFORMAT_PASSPHRASE:
	VERIFY0(zap_lookup(dd->dd_pool->dp_meta_objset,
	dd->dd_crypto_obj, zfs_prop_to_name(ZFS_PROP_PBKDF2_SALT),
	sizeof (uint64_t), 1, &salt));
	VERIFY0(zap_lookup(dd->dd_pool->dp_meta_objset,
	dd->dd_crypto_obj, zfs_prop_to_name(ZFS_PROP_PBKDF2_ITERS),
	sizeof (uint64_t), 1, &iters));

	if (PKCS5_PBKDF2_HMAC_SHA1(key_material, strlen(key_material),
	((uint8_t *)&salt), sizeof (uint64_t), iters,
	WRAPPING_KEY_LEN, key_out) != 1)
	return (B_FALSE);

	break;

	default:
	fatal("no support for key format %u\n",
	(unsigned int) keyformat);
	}

	return (B_TRUE);
	}

	static char encroot[ZFS_MAX_DATASET_NAME_LEN];
	static boolean_t key_loaded = B_FALSE;

	static void
	zdb_load_key(objset_t *os)
	{
	dsl_pool_t *dp;
	dsl_dir_t dd, rdd;
	uint8_t key[WRAPPING_KEY_LEN];
	uint64_t rddobj;
	int err;

	dp = spa_get_dsl(os->os_spa);
	dd = os->os_dsl_dataset->ds_dir;

	dsl_pool_config_enter(dp, FTAG);
	VERIFY0(zap_lookup(dd->dd_pool->dp_meta_objset, dd->dd_crypto_obj,
	DSL_CRYPTO_KEY_ROOT_DDOBJ, sizeof (uint64_t), 1, &rddobj));
	VERIFY0(dsl_dir_hold_obj(dd->dd_pool, rddobj, NULL, FTAG, &rdd));
	dsl_dir_name(rdd, encroot);
	dsl_dir_rele(rdd, FTAG);

	if (!zdb_derive_key(dd, key))
	fatal("couldn't derive encryption key");

	dsl_pool_config_exit(dp, FTAG);

	ASSERT3U(dsl_dataset_get_keystatus(dd), ==, ZFS_KEYSTATUS_UNAVAILABLE);

	dsl_crypto_params_t *dcp;
	nvlist_t *crypto_args;

	crypto_args = fnvlist_alloc();
	fnvlist_add_uint8_array(crypto_args, "wkeydata",
	(uint8_t *)key, WRAPPING_KEY_LEN);
	VERIFY0(dsl_crypto_params_create_nvlist(DCP_CMD_NONE,
	NULL, crypto_args, &dcp));
	err = spa_keystore_load_wkey(encroot, dcp, B_FALSE);

	dsl_crypto_params_free(dcp, (err != 0));
	fnvlist_free(crypto_args);

	if (err != 0)
	fatal(
	"couldn't load encryption key for %s: %s",
	encroot, err == ZFS_ERR_CRYPTO_NOTSUP ?
	"crypto params not supported" : strerror(err));

	ASSERT3U(dsl_dataset_get_keystatus(dd), ==, ZFS_KEYSTATUS_AVAILABLE);

	printf("Unlocked encryption root: %s\n", encroot);
	key_loaded = B_TRUE;
	}

	static void
	zdb_unload_key(void)
	{
	if (!key_loaded)
	return;

	VERIFY0(spa_keystore_unload_wkey(encroot));
	key_loaded = B_FALSE;
	}

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

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

	VERIFY3P(sa_os, ==, NULL);

	/*
	* We can't own an objset if it's redacted. Therefore, we do this
	* dance: hold the objset, then acquire a long hold on its dataset, then
	* release the pool (which is held as part of holding the objset).
	*/

	if (dump_opt['K']) {
	/* decryption requested, try to load keys */
	err = dmu_objset_hold(path, tag, osp);
	if (err != 0) {
	(void) fprintf(stderr, "failed to hold dataset "
	"'%s': %s\n",
	path, strerror(err));
	return (err);
	}
	dsl_dataset_long_hold(dmu_objset_ds(*osp), tag);
	dsl_pool_rele(dmu_objset_pool(*osp), tag);

	/* succeeds or dies */
	zdb_load_key(*osp);

	/* release it all */
	dsl_dataset_long_rele(dmu_objset_ds(*osp), tag);
	dsl_dataset_rele(dmu_objset_ds(*osp), tag);
	}

	int ds_hold_flags = key_loaded ? DS_HOLD_FLAG_DECRYPT : 0;

	err = dmu_objset_hold_flags(path, ds_hold_flags, tag, osp);
	if (err != 0) {
	(void) fprintf(stderr, "failed to hold dataset '%s': %s\n",
	path, strerror(err));
	return (err);
	}
	dsl_dataset_long_hold(dmu_objset_ds(*osp), tag);
	dsl_pool_rele(dmu_objset_pool(*osp), tag);

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

	return (err);
	}

	static void
	close_objset(objset_t os, const void tag)
	{
	VERIFY3P(os, ==, sa_os);
	if (os->os_sa != NULL)
	sa_tear_down(os);
	dsl_dataset_long_rele(dmu_objset_ds(os), tag);
	dsl_dataset_rele_flags(dmu_objset_ds(os),
	key_loaded ? DS_HOLD_FLAG_DECRYPT : 0, tag);
	sa_attr_table = NULL;
	sa_os = NULL;

	zdb_unload_key();
	}

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

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

	}

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

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

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

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

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

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

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

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

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

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

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

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

	(void) printf("\t\t%s = ", nvpair_name(elem));
	nvpair_value_byte_array(elem, &value, &cnt);

	for (idx = 0; idx < cnt; ++idx) {
	if (!isprint(value[idx])) {
	can_print = B_FALSE;
	break;
	}
	}

	for (idx = 0; idx < cnt; ++idx) {
	if (can_print)
	(void) putchar(value[idx]);
	else
	(void) printf("\\%3.3o", value[idx]);
	}
	(void) putchar('\n');
	}

	nvlist_free(sa_xattr);
	free(sa_xattr_packed);
	}

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

	break;
	}

	return (match);
	}

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

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

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

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

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

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

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

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

	aux[0] = '\0';

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

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

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

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

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

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

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

	*print_header = B_TRUE;
	}

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

	print_header = B_TRUE;

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

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

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

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

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

	dmu_objset_name(os, osname);

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

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

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

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

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

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

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

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

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

	if (verbosity < 2)
	return;

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

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

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

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

	(void) printf("\n");

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

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

	ASSERT3U(object_count, ==, usedobjs);

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

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

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

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

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

	(void) printf("\traidz_reflow state=%u off=%llu\n",
	(int)RRSS_GET_STATE(ub),
	(u_longlong_t)RRSS_GET_OFFSET(ub));

	(void) printf("%s", footer ? footer : "");
	}

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


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

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

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

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

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

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

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

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

	(void) close(fd);

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

	free(buf);

	dump_nvlist(config, 0);

	nvlist_free(config);
	}

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

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

	stats->zns_list_count++;

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

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

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

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

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

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

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

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

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

	collect_nvlist_stats(nvl, &stats);

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

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

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

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

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

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

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

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

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

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

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

	return (difference);
	}

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

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

	return (rec);
	}

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

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

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

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

	return (rec);
	}

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

	return (-1);
	}

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

	#define MAX_UBERBLOCK_COUNT (VDEV_UBERBLOCK_RING >> UBERBLOCK_SHIFT)

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

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

	if (dump_opt['q'])
	return;

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

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

	label->header_printed = B_TRUE;
	}

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

	return (0);
	}

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

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

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

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

	#define ZDB_MAX_UB_HEADER_SIZE 32

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

	vdev_t vd;
	char header[ZDB_MAX_UB_HEADER_SIZE];

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

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

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

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

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

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

	static char curpath[PATH_MAX];

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

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

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

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

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

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

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

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

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

	return (EINVAL);
	}

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

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

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

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

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

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

	static int
	dump_backup_bytes(objset_t os, void buf, int len, void *arg)
	{
	const char p = (const char )buf;
	ssize_t nwritten;

	(void) os;
	(void) arg;

	/* Write the data out, handling short writes and signals. */
	while ((nwritten = write(STDOUT_FILENO, p, len)) < len) {
	if (nwritten < 0) {
	if (errno == EINTR)
	continue;
	return (errno);
	}
	p += nwritten;
	len -= nwritten;
	}

	return (0);
	}

	static void
	dump_backup(const char pool, uint64_t objset_id, const char flagstr)
	{
	boolean_t embed = B_FALSE;
	boolean_t large_block = B_FALSE;
	boolean_t compress = B_FALSE;
	boolean_t raw = B_FALSE;

	const char *c;
	for (c = flagstr; c != NULL && *c != '\0'; c++) {
	switch (*c) {
	case 'e':
	embed = B_TRUE;
	break;
	case 'L':
	large_block = B_TRUE;
	break;
	case 'c':
	compress = B_TRUE;
	break;
	case 'w':
	raw = B_TRUE;
	break;
	default:
	fprintf(stderr, "dump_backup: invalid flag "
	"'%c'\n", *c);
	return;
	}
	}

	if (isatty(STDOUT_FILENO)) {
	fprintf(stderr, "dump_backup: stream cannot be written "
	"to a terminal\n");
	return;
	}

	offset_t off = 0;
	dmu_send_outparams_t out = {
	.dso_outfunc = dump_backup_bytes,
	.dso_dryrun = B_FALSE,
	};

	int err = dmu_send_obj(pool, objset_id, /* fromsnap */0, embed,
	large_block, compress, raw, /* saved */ B_FALSE, STDOUT_FILENO,
	&off, &out);
	if (err != 0) {
	fprintf(stderr, "dump_backup: dmu_send_obj: %s\n",
	strerror(err));
	return;
	}
	}

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

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

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

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

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

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

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

	offset += readsize;
	}

	(void) close(fd);

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

	return (err);
	}

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

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

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

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

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

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

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

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

	return (B_FALSE);
	}

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

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

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

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

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

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

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

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

	psize = statbuf.st_size;
	- psize = P2ALIGN(psize, (uint64_t)sizeof (vdev_label_t));
	+ psize = P2ALIGN_TYPED(psize, sizeof (vdev_label_t), uint64_t);
	ashift = SPA_MINBLOCKSHIFT;

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

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

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

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

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

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

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

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

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

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

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

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

	if (uberblock_verify(ub))
	continue;

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

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

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

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

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

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

	nvlist_free(label->config_nv);
	}

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

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

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

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

	(void) close(fd);

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

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

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

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

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

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

	for (dsl_bookmark_node_t *dbn =
	avl_first(&dmu_objset_ds(os)->ds_bookmarks); dbn != NULL;
	dbn = AVL_NEXT(&dmu_objset_ds(os)->ds_bookmarks, dbn)) {
	mos_obj_refd(dbn->dbn_phys.zbm_redaction_obj);
	if (dbn->dbn_phys.zbm_redaction_obj != 0) {
	global_feature_count[
	SPA_FEATURE_REDACTION_BOOKMARKS]++;
	objset_t *mos = os->os_spa->spa_meta_objset;
	dnode_t *rl;
	VERIFY0(dnode_hold(mos,
	dbn->dbn_phys.zbm_redaction_obj, FTAG, &rl));
	if (rl->dn_have_spill) {
	global_feature_count[
	SPA_FEATURE_REDACTION_LIST_SPILL]++;
	}
	}
	if (dbn->dbn_phys.zbm_flags & ZBM_FLAG_HAS_FBN)
	global_feature_count[SPA_FEATURE_BOOKMARK_WRITTEN]++;
	}

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

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

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

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

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

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

	typedef struct zdb_brt_entry {
	dva_t zbre_dva;
	uint64_t zbre_refcount;
	avl_node_t zbre_node;
	} zdb_brt_entry_t;

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

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

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

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

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

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

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

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

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

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

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


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

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

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

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

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

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

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

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

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

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

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

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

	ASSERT(type < ZDB_OT_TOTAL);

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

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

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

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

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

	zb->zb_gangs += BP_COUNT_GANG(bp);

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

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

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

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

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

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

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

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

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

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

	if (zcb->zcb_brt_is_active && brt_maybe_exists(zcb->zcb_spa, bp)) {
	/*
	* Cloned blocks are special. We need to count them, so we can
	* later uncount them when reporting leaked space, and we must
	* only claim them them once.
	*
	* To do this, we keep our own in-memory BRT. For each block
	* we haven't seen before, we look it up in the real BRT and
	* if its there, we note it and its refcount then proceed as
	* normal. If we see the block again, we count it as a clone
	* and then give it no further consideration.
	*/
	zdb_brt_entry_t zbre_search, *zbre;
	avl_index_t where;

	zbre_search.zbre_dva = bp->blk_dva[0];
	zbre = avl_find(&zcb->zcb_brt, &zbre_search, &where);
	if (zbre != NULL) {
	zcb->zcb_clone_asize += BP_GET_ASIZE(bp);
	zcb->zcb_clone_blocks++;

	zbre->zbre_refcount--;
	if (zbre->zbre_refcount == 0) {
	avl_remove(&zcb->zcb_brt, zbre);
	umem_free(zbre, sizeof (zdb_brt_entry_t));
	}
	return;
	}

	uint64_t crefcnt = brt_entry_get_refcount(zcb->zcb_spa, bp);
	if (crefcnt > 0) {
	zbre = umem_zalloc(sizeof (zdb_brt_entry_t),
	UMEM_NOFAIL);
	zbre->zbre_dva = bp->blk_dva[0];
	zbre->zbre_refcount = crefcnt;
	avl_insert(&zcb->zcb_brt, zbre, where);
	}
	}

	if (dump_opt['L'])
	return;

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

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

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

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

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

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

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

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

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

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

	abd_free(zio->io_abd);
	}

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

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

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

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

	type = BP_GET_TYPE(bp);

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

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

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

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

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

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

	zcb->zcb_readfails = 0;

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

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

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

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

	zcb->zcb_lastprint = now;
	}

	return (0);
	}

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

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

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

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

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

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

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

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

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

	return (0);
	}

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

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

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

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

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

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

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

	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);

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

	ASSERT0(range_tree_space(svr->svr_allocd_segs));

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

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

	iterate_through_spacemap_logs(spa, load_unflushed_svr_segs_cb, svr);

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

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

	spa_config_exit(spa, SCL_CONFIG, FTAG);
	}

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

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

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

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

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

	return (0);
	}

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

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

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

	ASSERT(!dump_opt['L']);

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

	if (ddb.ddb_class == DDT_CLASS_UNIQUE)
	return;

	ASSERT(ddt_phys_total_refcnt(&dde) > 1);
	ddt_t *ddt = spa->spa_ddt[ddb.ddb_checksum];
	VERIFY(ddt);

	for (p = 0; p < DDT_PHYS_TYPES; p++, ddp++) {
	if (ddp->ddp_phys_birth == 0)
	continue;
	ddt_bp_create(ddb.ddb_checksum,
	&dde.dde_key, ddp, &blk);
	if (p == DDT_PHYS_DITTO) {
	zdb_count_block(zcb, NULL, &blk, ZDB_OT_DITTO);
	} else {
	zcb->zcb_dedup_asize +=
	BP_GET_ASIZE(&blk) * (ddp->ddp_refcnt - 1);
	zcb->zcb_dedup_blocks++;
	}
	}

	ddt_enter(ddt);
	VERIFY(ddt_lookup(ddt, &blk, B_TRUE) != NULL);
	ddt_exit(ddt);
	}

	ASSERT(error == ENOENT);
	}

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

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

	ASSERT(sme->sme_type == SM_FREE);

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

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

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

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

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

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

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

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

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

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

	zcb->zcb_checkpoint_size += cseea.cseea_checkpoint_size;
	}

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

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

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

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

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

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

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

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

	return (0);
	}

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

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

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

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

	vdev_t *vd = vdev_lookup_top(spa, vdev_id);

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

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

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

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

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

	return (0);
	}

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

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

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

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

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

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

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

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

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

	load_unflushed_to_ms_allocatables(spa, maptype);
	}

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

	if (dump_opt['L'])
	return;

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

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

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

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

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

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

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

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

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

	ASSERT(vim != NULL);

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

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

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

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

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

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

	return (leaks);
	}

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

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

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

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

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

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

	return (leaks);
	}

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

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

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

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

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

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

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

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

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

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

	static int
	zdb_brt_entry_compare(const void zcn1, const void zcn2)
	{
	const dva_t dva1 = &((const zdb_brt_entry_t )zcn1)->zbre_dva;
	const dva_t dva2 = &((const zdb_brt_entry_t )zcn2)->zbre_dva;
	int cmp;

	cmp = TREE_CMP(DVA_GET_VDEV(dva1), DVA_GET_VDEV(dva2));
	if (cmp == 0)
	cmp = TREE_CMP(DVA_GET_OFFSET(dva1), DVA_GET_OFFSET(dva2));

	return (cmp);
	}

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

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

	if (spa_feature_is_active(spa, SPA_FEATURE_BLOCK_CLONING)) {
	avl_create(&zcb->zcb_brt, zdb_brt_entry_compare,
	sizeof (zdb_brt_entry_t),
	offsetof(zdb_brt_entry_t, zbre_node));
	zcb->zcb_brt_is_active = B_TRUE;
	}

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

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

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

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

	zdb_claim_removing(spa, zcb);

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

	deleted_livelists_count_blocks(spa, zcb);

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

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

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

	/*
	* Done after zio_wait() since zcb_haderrors is modified in
	* zdb_blkptr_done()
	*/
	zcb->zcb_haderrors \|= err;

	if (zcb->zcb_haderrors) {
	(void) printf("\nError counts:\n\n");
	(void) printf("\t%5s %s\n", "errno", "count");
	for (e = 0; e < 256; e++) {
	if (zcb->zcb_errors[e] != 0) {
	(void) printf("\t%5d %llu\n",
	e, (u_longlong_t)zcb->zcb_errors[e]);
	}
	}
	}

	/*
	* Report any leaked segments.
	*/
	leaks \|= zdb_leak_fini(spa, zcb);

	tzb = &zcb->zcb_type[ZB_TOTAL][ZDB_OT_TOTAL];

	norm_alloc = metaslab_class_get_alloc(spa_normal_class(spa));
	norm_space = metaslab_class_get_space(spa_normal_class(spa));

	total_alloc = norm_alloc +
	metaslab_class_get_alloc(spa_log_class(spa)) +
	metaslab_class_get_alloc(spa_embedded_log_class(spa)) +
	metaslab_class_get_alloc(spa_special_class(spa)) +
	metaslab_class_get_alloc(spa_dedup_class(spa)) +
	get_unflushed_alloc_space(spa);
	total_found =
	tzb->zb_asize - zcb->zcb_dedup_asize - zcb->zcb_clone_asize +
	zcb->zcb_removing_size + zcb->zcb_checkpoint_size;

	if (total_found == total_alloc && !dump_opt['L']) {
	(void) printf("\n\tNo leaks (block sum matches space"
	" maps exactly)\n");
	} else if (!dump_opt['L']) {
	(void) printf("block traversal size %llu != alloc %llu "
	"(%s %lld)\n",
	(u_longlong_t)total_found,
	(u_longlong_t)total_alloc,
	(dump_opt['L']) ? "unreachable" : "leaked",
	(longlong_t)(total_alloc - total_found));
	leaks = B_TRUE;
	}

	if (tzb->zb_count == 0) {
	umem_free(zcb, sizeof (zdb_cb_t));
	return (2);
	}

	(void) printf("\n");
	(void) printf("\t%-16s %14llu\n", "bp count:",
	(u_longlong_t)tzb->zb_count);
	(void) printf("\t%-16s %14llu\n", "ganged count:",
	(longlong_t)tzb->zb_gangs);
	(void) printf("\t%-16s %14llu avg: %6llu\n", "bp logical:",
	(u_longlong_t)tzb->zb_lsize,
	(u_longlong_t)(tzb->zb_lsize / tzb->zb_count));
	(void) printf("\t%-16s %14llu avg: %6llu compression: %6.2f\n",
	"bp physical:", (u_longlong_t)tzb->zb_psize,
	(u_longlong_t)(tzb->zb_psize / tzb->zb_count),
	(double)tzb->zb_lsize / tzb->zb_psize);
	(void) printf("\t%-16s %14llu avg: %6llu compression: %6.2f\n",
	"bp allocated:", (u_longlong_t)tzb->zb_asize,
	(u_longlong_t)(tzb->zb_asize / tzb->zb_count),
	(double)tzb->zb_lsize / tzb->zb_asize);
	(void) printf("\t%-16s %14llu ref>1: %6llu deduplication: %6.2f\n",
	"bp deduped:", (u_longlong_t)zcb->zcb_dedup_asize,
	(u_longlong_t)zcb->zcb_dedup_blocks,
	(double)zcb->zcb_dedup_asize / tzb->zb_asize + 1.0);
	(void) printf("\t%-16s %14llu count: %6llu\n",
	"bp cloned:", (u_longlong_t)zcb->zcb_clone_asize,
	(u_longlong_t)zcb->zcb_clone_blocks);
	(void) printf("\t%-16s %14llu used: %5.2f%%\n", "Normal class:",
	(u_longlong_t)norm_alloc, 100.0 * norm_alloc / norm_space);

	if (spa_special_class(spa)->mc_allocator[0].mca_rotor != NULL) {
	uint64_t alloc = metaslab_class_get_alloc(
	spa_special_class(spa));
	uint64_t space = metaslab_class_get_space(
	spa_special_class(spa));

	(void) printf("\t%-16s %14llu used: %5.2f%%\n",
	"Special class", (u_longlong_t)alloc,
	100.0 * alloc / space);
	}

	if (spa_dedup_class(spa)->mc_allocator[0].mca_rotor != NULL) {
	uint64_t alloc = metaslab_class_get_alloc(
	spa_dedup_class(spa));
	uint64_t space = metaslab_class_get_space(
	spa_dedup_class(spa));

	(void) printf("\t%-16s %14llu used: %5.2f%%\n",
	"Dedup class", (u_longlong_t)alloc,
	100.0 * alloc / space);
	}

	if (spa_embedded_log_class(spa)->mc_allocator[0].mca_rotor != NULL) {
	uint64_t alloc = metaslab_class_get_alloc(
	spa_embedded_log_class(spa));
	uint64_t space = metaslab_class_get_space(
	spa_embedded_log_class(spa));

	(void) printf("\t%-16s %14llu used: %5.2f%%\n",
	"Embedded log class", (u_longlong_t)alloc,
	100.0 * alloc / space);
	}

	for (i = 0; i < NUM_BP_EMBEDDED_TYPES; i++) {
	if (zcb->zcb_embedded_blocks[i] == 0)
	continue;
	(void) printf("\n");
	(void) printf("\tadditional, non-pointer bps of type %u: "
	"%10llu\n",
	i, (u_longlong_t)zcb->zcb_embedded_blocks[i]);

	if (dump_opt['b'] >= 3) {
	(void) printf("\t number of (compressed) bytes: "
	"number of bps\n");
	dump_histogram(zcb->zcb_embedded_histogram[i],
	sizeof (zcb->zcb_embedded_histogram[i]) /
	sizeof (zcb->zcb_embedded_histogram[i][0]), 0);
	}
	}

	if (tzb->zb_ditto_samevdev != 0) {
	(void) printf("\tDittoed blocks on same vdev: %llu\n",
	(longlong_t)tzb->zb_ditto_samevdev);
	}
	if (tzb->zb_ditto_same_ms != 0) {
	(void) printf("\tDittoed blocks in same metaslab: %llu\n",
	(longlong_t)tzb->zb_ditto_same_ms);
	}

	for (uint64_t v = 0; v < spa->spa_root_vdev->vdev_children; v++) {
	vdev_t *vd = spa->spa_root_vdev->vdev_child[v];
	vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping;

	if (vim == NULL) {
	continue;
	}

	char mem[32];
	zdb_nicenum(vdev_indirect_mapping_num_entries(vim),
	mem, vdev_indirect_mapping_size(vim));

	(void) printf("\tindirect vdev id %llu has %llu segments "
	"(%s in memory)\n",
	(longlong_t)vd->vdev_id,
	(longlong_t)vdev_indirect_mapping_num_entries(vim), mem);
	}

	if (dump_opt['b'] >= 2) {
	int l, t, level;
	char csize[32], lsize[32], psize[32], asize[32];
	char avg[32], gang[32];
	(void) printf("\nBlocks\tLSIZE\tPSIZE\tASIZE"
	"\t avg\t comp\t%%Total\tType\n");

	zfs_blkstat_t *mdstats = umem_zalloc(sizeof (zfs_blkstat_t),
	UMEM_NOFAIL);

	for (t = 0; t <= ZDB_OT_TOTAL; t++) {
	const char *typename;

	/* make sure nicenum has enough space */
	_Static_assert(sizeof (csize) >= NN_NUMBUF_SZ,
	"csize truncated");
	_Static_assert(sizeof (lsize) >= NN_NUMBUF_SZ,
	"lsize truncated");
	_Static_assert(sizeof (psize) >= NN_NUMBUF_SZ,
	"psize truncated");
	_Static_assert(sizeof (asize) >= NN_NUMBUF_SZ,
	"asize truncated");
	_Static_assert(sizeof (avg) >= NN_NUMBUF_SZ,
	"avg truncated");
	_Static_assert(sizeof (gang) >= NN_NUMBUF_SZ,
	"gang truncated");

	if (t < DMU_OT_NUMTYPES)
	typename = dmu_ot[t].ot_name;
	else
	typename = zdb_ot_extname[t - DMU_OT_NUMTYPES];

	if (zcb->zcb_type[ZB_TOTAL][t].zb_asize == 0) {
	(void) printf("%6s\t%5s\t%5s\t%5s"
	"\t%5s\t%5s\t%6s\t%s\n",
	"-",
	"-",
	"-",
	"-",
	"-",
	"-",
	"-",
	typename);
	continue;
	}

	for (l = ZB_TOTAL - 1; l >= -1; l--) {
	level = (l == -1 ? ZB_TOTAL : l);
	zb = &zcb->zcb_type[level][t];

	if (zb->zb_asize == 0)
	continue;

	if (level != ZB_TOTAL && t < DMU_OT_NUMTYPES &&
	(level > 0 \|\| DMU_OT_IS_METADATA(t))) {
	mdstats->zb_count += zb->zb_count;
	mdstats->zb_lsize += zb->zb_lsize;
	mdstats->zb_psize += zb->zb_psize;
	mdstats->zb_asize += zb->zb_asize;
	mdstats->zb_gangs += zb->zb_gangs;
	}

	if (dump_opt['b'] < 3 && level != ZB_TOTAL)
	continue;

	if (level == 0 && zb->zb_asize ==
	zcb->zcb_type[ZB_TOTAL][t].zb_asize)
	continue;

	zdb_nicenum(zb->zb_count, csize,
	sizeof (csize));
	zdb_nicenum(zb->zb_lsize, lsize,
	sizeof (lsize));
	zdb_nicenum(zb->zb_psize, psize,
	sizeof (psize));
	zdb_nicenum(zb->zb_asize, asize,
	sizeof (asize));
	zdb_nicenum(zb->zb_asize / zb->zb_count, avg,
	sizeof (avg));
	zdb_nicenum(zb->zb_gangs, gang, sizeof (gang));

	(void) printf("%6s\t%5s\t%5s\t%5s\t%5s"
	"\t%5.2f\t%6.2f\t",
	csize, lsize, psize, asize, avg,
	(double)zb->zb_lsize / zb->zb_psize,
	100.0 * zb->zb_asize / tzb->zb_asize);

	if (level == ZB_TOTAL)
	(void) printf("%s\n", typename);
	else
	(void) printf(" L%d %s\n",
	level, typename);

	if (dump_opt['b'] >= 3 && zb->zb_gangs > 0) {
	(void) printf("\t number of ganged "
	"blocks: %s\n", gang);
	}

	if (dump_opt['b'] >= 4) {
	(void) printf("psize "
	"(in 512-byte sectors): "
	"number of blocks\n");
	dump_histogram(zb->zb_psize_histogram,
	PSIZE_HISTO_SIZE, 0);
	}
	}
	}
	zdb_nicenum(mdstats->zb_count, csize,
	sizeof (csize));
	zdb_nicenum(mdstats->zb_lsize, lsize,
	sizeof (lsize));
	zdb_nicenum(mdstats->zb_psize, psize,
	sizeof (psize));
	zdb_nicenum(mdstats->zb_asize, asize,
	sizeof (asize));
	zdb_nicenum(mdstats->zb_asize / mdstats->zb_count, avg,
	sizeof (avg));
	zdb_nicenum(mdstats->zb_gangs, gang, sizeof (gang));

	(void) printf("%6s\t%5s\t%5s\t%5s\t%5s"
	"\t%5.2f\t%6.2f\t",
	csize, lsize, psize, asize, avg,
	(double)mdstats->zb_lsize / mdstats->zb_psize,
	100.0 * mdstats->zb_asize / tzb->zb_asize);
	(void) printf("%s\n", "Metadata Total");

	/* Output a table summarizing block sizes in the pool */
	if (dump_opt['b'] >= 2) {
	dump_size_histograms(zcb);
	}

	umem_free(mdstats, sizeof (zfs_blkstat_t));
	}

	(void) printf("\n");

	if (leaks) {
	umem_free(zcb, sizeof (zdb_cb_t));
	return (2);
	}

	if (zcb->zcb_haderrors) {
	umem_free(zcb, sizeof (zdb_cb_t));
	return (3);
	}

	umem_free(zcb, sizeof (zdb_cb_t));
	return (0);
	}

	typedef struct zdb_ddt_entry {
	/* key must be first for ddt_key_compare */
	ddt_key_t zdde_key;
	uint64_t zdde_ref_blocks;
	uint64_t zdde_ref_lsize;
	uint64_t zdde_ref_psize;
	uint64_t zdde_ref_dsize;
	avl_node_t zdde_node;
	} zdb_ddt_entry_t;

	static int
	zdb_ddt_add_cb(spa_t spa, zilog_t zilog, const blkptr_t *bp,
	const zbookmark_phys_t zb, const dnode_phys_t dnp, void *arg)
	{
	(void) zilog, (void) dnp;
	avl_tree_t *t = arg;
	avl_index_t where;
	zdb_ddt_entry_t *zdde, zdde_search;

	if (zb->zb_level == ZB_DNODE_LEVEL \|\| BP_IS_HOLE(bp) \|\|
	BP_IS_EMBEDDED(bp))
	return (0);

	if (dump_opt['S'] > 1 && zb->zb_level == ZB_ROOT_LEVEL) {
	(void) printf("traversing objset %llu, %llu objects, "
	"%lu blocks so far\n",
	(u_longlong_t)zb->zb_objset,
	(u_longlong_t)BP_GET_FILL(bp),
	avl_numnodes(t));
	}

	if (BP_IS_HOLE(bp) \|\| BP_GET_CHECKSUM(bp) == ZIO_CHECKSUM_OFF \|\|
	BP_GET_LEVEL(bp) > 0 \|\| DMU_OT_IS_METADATA(BP_GET_TYPE(bp)))
	return (0);

	ddt_key_fill(&zdde_search.zdde_key, bp);

	zdde = avl_find(t, &zdde_search, &where);

	if (zdde == NULL) {
	zdde = umem_zalloc(sizeof (*zdde), UMEM_NOFAIL);
	zdde->zdde_key = zdde_search.zdde_key;
	avl_insert(t, zdde, where);
	}

	zdde->zdde_ref_blocks += 1;
	zdde->zdde_ref_lsize += BP_GET_LSIZE(bp);
	zdde->zdde_ref_psize += BP_GET_PSIZE(bp);
	zdde->zdde_ref_dsize += bp_get_dsize_sync(spa, bp);

	return (0);
	}

	static void
	dump_simulated_ddt(spa_t *spa)
	{
	avl_tree_t t;
	void *cookie = NULL;
	zdb_ddt_entry_t *zdde;
	ddt_histogram_t ddh_total = {{{0}}};
	ddt_stat_t dds_total = {0};

	avl_create(&t, ddt_key_compare,
	sizeof (zdb_ddt_entry_t), offsetof(zdb_ddt_entry_t, zdde_node));

	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);

	(void) traverse_pool(spa, 0, TRAVERSE_PRE \| TRAVERSE_PREFETCH_METADATA \|
	TRAVERSE_NO_DECRYPT, zdb_ddt_add_cb, &t);

	spa_config_exit(spa, SCL_CONFIG, FTAG);

	while ((zdde = avl_destroy_nodes(&t, &cookie)) != NULL) {
	ddt_stat_t dds;
	uint64_t refcnt = zdde->zdde_ref_blocks;
	ASSERT(refcnt != 0);

	dds.dds_blocks = zdde->zdde_ref_blocks / refcnt;
	dds.dds_lsize = zdde->zdde_ref_lsize / refcnt;
	dds.dds_psize = zdde->zdde_ref_psize / refcnt;
	dds.dds_dsize = zdde->zdde_ref_dsize / refcnt;

	dds.dds_ref_blocks = zdde->zdde_ref_blocks;
	dds.dds_ref_lsize = zdde->zdde_ref_lsize;
	dds.dds_ref_psize = zdde->zdde_ref_psize;
	dds.dds_ref_dsize = zdde->zdde_ref_dsize;

	ddt_stat_add(&ddh_total.ddh_stat[highbit64(refcnt) - 1],
	&dds, 0);

	umem_free(zdde, sizeof (*zdde));
	}

	avl_destroy(&t);

	ddt_histogram_stat(&dds_total, &ddh_total);

	(void) printf("Simulated DDT histogram:\n");

	zpool_dump_ddt(&dds_total, &ddh_total);

	dump_dedup_ratio(&dds_total);
	}

	static int
	verify_device_removal_feature_counts(spa_t *spa)
	{
	uint64_t dr_feature_refcount = 0;
	uint64_t oc_feature_refcount = 0;
	uint64_t indirect_vdev_count = 0;
	uint64_t precise_vdev_count = 0;
	uint64_t obsolete_counts_object_count = 0;
	uint64_t obsolete_sm_count = 0;
	uint64_t obsolete_counts_count = 0;
	uint64_t scip_count = 0;
	uint64_t obsolete_bpobj_count = 0;
	int ret = 0;

	spa_condensing_indirect_phys_t *scip =
	&spa->spa_condensing_indirect_phys;
	if (scip->scip_next_mapping_object != 0) {
	vdev_t *vd = spa->spa_root_vdev->vdev_child[scip->scip_vdev];
	ASSERT(scip->scip_prev_obsolete_sm_object != 0);
	ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);

	(void) printf("Condensing indirect vdev %llu: new mapping "
	"object %llu, prev obsolete sm %llu\n",
	(u_longlong_t)scip->scip_vdev,
	(u_longlong_t)scip->scip_next_mapping_object,
	(u_longlong_t)scip->scip_prev_obsolete_sm_object);
	if (scip->scip_prev_obsolete_sm_object != 0) {
	space_map_t *prev_obsolete_sm = NULL;
	VERIFY0(space_map_open(&prev_obsolete_sm,
	spa->spa_meta_objset,
	scip->scip_prev_obsolete_sm_object,
	0, vd->vdev_asize, 0));
	dump_spacemap(spa->spa_meta_objset, prev_obsolete_sm);
	(void) printf("\n");
	space_map_close(prev_obsolete_sm);
	}

	scip_count += 2;
	}

	for (uint64_t i = 0; i < spa->spa_root_vdev->vdev_children; i++) {
	vdev_t *vd = spa->spa_root_vdev->vdev_child[i];
	vdev_indirect_config_t *vic = &vd->vdev_indirect_config;

	if (vic->vic_mapping_object != 0) {
	ASSERT(vd->vdev_ops == &vdev_indirect_ops \|\|
	vd->vdev_removing);
	indirect_vdev_count++;

	if (vd->vdev_indirect_mapping->vim_havecounts) {
	obsolete_counts_count++;
	}
	}

	boolean_t are_precise;
	VERIFY0(vdev_obsolete_counts_are_precise(vd, &are_precise));
	if (are_precise) {
	ASSERT(vic->vic_mapping_object != 0);
	precise_vdev_count++;
	}

	uint64_t obsolete_sm_object;
	VERIFY0(vdev_obsolete_sm_object(vd, &obsolete_sm_object));
	if (obsolete_sm_object != 0) {
	ASSERT(vic->vic_mapping_object != 0);
	obsolete_sm_count++;
	}
	}

	(void) feature_get_refcount(spa,
	&spa_feature_table[SPA_FEATURE_DEVICE_REMOVAL],
	&dr_feature_refcount);
	(void) feature_get_refcount(spa,
	&spa_feature_table[SPA_FEATURE_OBSOLETE_COUNTS],
	&oc_feature_refcount);

	if (dr_feature_refcount != indirect_vdev_count) {
	ret = 1;
	(void) printf("Number of indirect vdevs (%llu) " \
	"does not match feature count (%llu)\n",
	(u_longlong_t)indirect_vdev_count,
	(u_longlong_t)dr_feature_refcount);
	} else {
	(void) printf("Verified device_removal feature refcount " \
	"of %llu is correct\n",
	(u_longlong_t)dr_feature_refcount);
	}

	if (zap_contains(spa_meta_objset(spa), DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_OBSOLETE_BPOBJ) == 0) {
	obsolete_bpobj_count++;
	}


	obsolete_counts_object_count = precise_vdev_count;
	obsolete_counts_object_count += obsolete_sm_count;
	obsolete_counts_object_count += obsolete_counts_count;
	obsolete_counts_object_count += scip_count;
	obsolete_counts_object_count += obsolete_bpobj_count;
	obsolete_counts_object_count += remap_deadlist_count;

	if (oc_feature_refcount != obsolete_counts_object_count) {
	ret = 1;
	(void) printf("Number of obsolete counts objects (%llu) " \
	"does not match feature count (%llu)\n",
	(u_longlong_t)obsolete_counts_object_count,
	(u_longlong_t)oc_feature_refcount);
	(void) printf("pv:%llu os:%llu oc:%llu sc:%llu "
	"ob:%llu rd:%llu\n",
	(u_longlong_t)precise_vdev_count,
	(u_longlong_t)obsolete_sm_count,
	(u_longlong_t)obsolete_counts_count,
	(u_longlong_t)scip_count,
	(u_longlong_t)obsolete_bpobj_count,
	(u_longlong_t)remap_deadlist_count);
	} else {
	(void) printf("Verified indirect_refcount feature refcount " \
	"of %llu is correct\n",
	(u_longlong_t)oc_feature_refcount);
	}
	return (ret);
	}

	static void
	zdb_set_skip_mmp(char *target)
	{
	spa_t *spa;

	/*
	* Disable the activity check to allow examination of
	* active pools.
	*/
	mutex_enter(&spa_namespace_lock);
	if ((spa = spa_lookup(target)) != NULL) {
	spa->spa_import_flags \|= ZFS_IMPORT_SKIP_MMP;
	}
	mutex_exit(&spa_namespace_lock);
	}

	#define BOGUS_SUFFIX "_CHECKPOINTED_UNIVERSE"
	/*
	* Import the checkpointed state of the pool specified by the target
	* parameter as readonly. The function also accepts a pool config
	* as an optional parameter, else it attempts to infer the config by
	* the name of the target pool.
	*
	* Note that the checkpointed state's pool name will be the name of
	* the original pool with the above suffix appended to it. In addition,
	* if the target is not a pool name (e.g. a path to a dataset) then
	* the new_path parameter is populated with the updated path to
	* reflect the fact that we are looking into the checkpointed state.
	*
	* The function returns a newly-allocated copy of the name of the
	* pool containing the checkpointed state. When this copy is no
	* longer needed it should be freed with free(3C). Same thing
	* applies to the new_path parameter if allocated.
	*/
	static char *
	import_checkpointed_state(char target, nvlist_t cfg, char **new_path)
	{
	int error = 0;
	char poolname, bogus_name = NULL;
	boolean_t freecfg = B_FALSE;

	/* If the target is not a pool, the extract the pool name */
	char *path_start = strchr(target, '/');
	if (path_start != NULL) {
	size_t poolname_len = path_start - target;
	poolname = strndup(target, poolname_len);
	} else {
	poolname = target;
	}

	if (cfg == NULL) {
	zdb_set_skip_mmp(poolname);
	error = spa_get_stats(poolname, &cfg, NULL, 0);
	if (error != 0) {
	fatal("Tried to read config of pool \"%s\" but "
	"spa_get_stats() failed with error %d\n",
	poolname, error);
	}
	freecfg = B_TRUE;
	}

	if (asprintf(&bogus_name, "%s%s", poolname, BOGUS_SUFFIX) == -1) {
	if (target != poolname)
	free(poolname);
	return (NULL);
	}
	fnvlist_add_string(cfg, ZPOOL_CONFIG_POOL_NAME, bogus_name);

	error = spa_import(bogus_name, cfg, NULL,
	ZFS_IMPORT_MISSING_LOG \| ZFS_IMPORT_CHECKPOINT \|
	ZFS_IMPORT_SKIP_MMP);
	if (freecfg)
	nvlist_free(cfg);
	if (error != 0) {
	fatal("Tried to import pool \"%s\" but spa_import() failed "
	"with error %d\n", bogus_name, error);
	}

	if (new_path != NULL && path_start != NULL) {
	if (asprintf(new_path, "%s%s", bogus_name, path_start) == -1) {
	free(bogus_name);
	if (path_start != NULL)
	free(poolname);
	return (NULL);
	}
	}

	if (target != poolname)
	free(poolname);

	return (bogus_name);
	}

	typedef struct verify_checkpoint_sm_entry_cb_arg {
	vdev_t *vcsec_vd;

	/* the following fields are only used for printing progress */
	uint64_t vcsec_entryid;
	uint64_t vcsec_num_entries;
	} verify_checkpoint_sm_entry_cb_arg_t;

	#define ENTRIES_PER_PROGRESS_UPDATE 10000

	static int
	verify_checkpoint_sm_entry_cb(space_map_entry_t sme, void arg)
	{
	verify_checkpoint_sm_entry_cb_arg_t *vcsec = arg;
	vdev_t *vd = vcsec->vcsec_vd;
	metaslab_t *ms = vd->vdev_ms[sme->sme_offset >> vd->vdev_ms_shift];
	uint64_t end = sme->sme_offset + sme->sme_run;

	ASSERT(sme->sme_type == SM_FREE);

	if ((vcsec->vcsec_entryid % ENTRIES_PER_PROGRESS_UPDATE) == 0) {
	(void) fprintf(stderr,
	"\rverifying vdev %llu, space map entry %llu of %llu ...",
	(longlong_t)vd->vdev_id,
	(longlong_t)vcsec->vcsec_entryid,
	(longlong_t)vcsec->vcsec_num_entries);
	}
	vcsec->vcsec_entryid++;

	/*
	* See comment in checkpoint_sm_exclude_entry_cb()
	*/
	VERIFY3U(sme->sme_offset, >=, ms->ms_start);
	VERIFY3U(end, <=, ms->ms_start + ms->ms_size);

	/*
	* The entries in the vdev_checkpoint_sm should be marked as
	* allocated in the checkpointed state of the pool, therefore
	* their respective ms_allocateable trees should not contain them.
	*/
	mutex_enter(&ms->ms_lock);
	range_tree_verify_not_present(ms->ms_allocatable,
	sme->sme_offset, sme->sme_run);
	mutex_exit(&ms->ms_lock);

	return (0);
	}

	/*
	* Verify that all segments in the vdev_checkpoint_sm are allocated
	* according to the checkpoint's ms_sm (i.e. are not in the checkpoint's
	* ms_allocatable).
	*
	* Do so by comparing the checkpoint space maps (vdev_checkpoint_sm) of
	* each vdev in the current state of the pool to the metaslab space maps
	* (ms_sm) of the checkpointed state of the pool.
	*
	* Note that the function changes the state of the ms_allocatable
	* trees of the current spa_t. The entries of these ms_allocatable
	* trees are cleared out and then repopulated from with the free
	* entries of their respective ms_sm space maps.
	*/
	static void
	verify_checkpoint_vdev_spacemaps(spa_t checkpoint, spa_t current)
	{
	vdev_t *ckpoint_rvd = checkpoint->spa_root_vdev;
	vdev_t *current_rvd = current->spa_root_vdev;

	load_concrete_ms_allocatable_trees(checkpoint, SM_FREE);

	for (uint64_t c = 0; c < ckpoint_rvd->vdev_children; c++) {
	vdev_t *ckpoint_vd = ckpoint_rvd->vdev_child[c];
	vdev_t *current_vd = current_rvd->vdev_child[c];

	space_map_t *checkpoint_sm = NULL;
	uint64_t checkpoint_sm_obj;

	if (ckpoint_vd->vdev_ops == &vdev_indirect_ops) {
	/*
	* Since we don't allow device removal in a pool
	* that has a checkpoint, we expect that all removed
	* vdevs were removed from the pool before the
	* checkpoint.
	*/
	ASSERT3P(current_vd->vdev_ops, ==, &vdev_indirect_ops);
	continue;
	}

	/*
	* If the checkpoint space map doesn't exist, then nothing
	* here is checkpointed so there's nothing to verify.
	*/
	if (current_vd->vdev_top_zap == 0 \|\|
	zap_contains(spa_meta_objset(current),
	current_vd->vdev_top_zap,
	VDEV_TOP_ZAP_POOL_CHECKPOINT_SM) != 0)
	continue;

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

	VERIFY0(space_map_open(&checkpoint_sm, spa_meta_objset(current),
	checkpoint_sm_obj, 0, current_vd->vdev_asize,
	current_vd->vdev_ashift));

	verify_checkpoint_sm_entry_cb_arg_t vcsec;
	vcsec.vcsec_vd = ckpoint_vd;
	vcsec.vcsec_entryid = 0;
	vcsec.vcsec_num_entries =
	space_map_length(checkpoint_sm) / sizeof (uint64_t);
	VERIFY0(space_map_iterate(checkpoint_sm,
	space_map_length(checkpoint_sm),
	verify_checkpoint_sm_entry_cb, &vcsec));
	if (dump_opt['m'] > 3)
	dump_spacemap(current->spa_meta_objset, checkpoint_sm);
	space_map_close(checkpoint_sm);
	}

	/*
	* If we've added vdevs since we took the checkpoint, ensure
	* that their checkpoint space maps are empty.
	*/
	if (ckpoint_rvd->vdev_children < current_rvd->vdev_children) {
	for (uint64_t c = ckpoint_rvd->vdev_children;
	c < current_rvd->vdev_children; c++) {
	vdev_t *current_vd = current_rvd->vdev_child[c];
	VERIFY3P(current_vd->vdev_checkpoint_sm, ==, NULL);
	}
	}

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

	/*
	* Verifies that all space that's allocated in the checkpoint is
	* still allocated in the current version, by checking that everything
	* in checkpoint's ms_allocatable (which is actually allocated, not
	* allocatable/free) is not present in current's ms_allocatable.
	*
	* Note that the function changes the state of the ms_allocatable
	* trees of both spas when called. The entries of all ms_allocatable
	* trees are cleared out and then repopulated from their respective
	* ms_sm space maps. In the checkpointed state we load the allocated
	* entries, and in the current state we load the free entries.
	*/
	static void
	verify_checkpoint_ms_spacemaps(spa_t checkpoint, spa_t current)
	{
	vdev_t *ckpoint_rvd = checkpoint->spa_root_vdev;
	vdev_t *current_rvd = current->spa_root_vdev;

	load_concrete_ms_allocatable_trees(checkpoint, SM_ALLOC);
	load_concrete_ms_allocatable_trees(current, SM_FREE);

	for (uint64_t i = 0; i < ckpoint_rvd->vdev_children; i++) {
	vdev_t *ckpoint_vd = ckpoint_rvd->vdev_child[i];
	vdev_t *current_vd = current_rvd->vdev_child[i];

	if (ckpoint_vd->vdev_ops == &vdev_indirect_ops) {
	/*
	* See comment in verify_checkpoint_vdev_spacemaps()
	*/
	ASSERT3P(current_vd->vdev_ops, ==, &vdev_indirect_ops);
	continue;
	}

	for (uint64_t m = 0; m < ckpoint_vd->vdev_ms_count; m++) {
	metaslab_t *ckpoint_msp = ckpoint_vd->vdev_ms[m];
	metaslab_t *current_msp = current_vd->vdev_ms[m];

	(void) fprintf(stderr,
	"\rverifying vdev %llu of %llu, "
	"metaslab %llu of %llu ...",
	(longlong_t)current_vd->vdev_id,
	(longlong_t)current_rvd->vdev_children,
	(longlong_t)current_vd->vdev_ms[m]->ms_id,
	(longlong_t)current_vd->vdev_ms_count);

	/*
	* We walk through the ms_allocatable trees that
	* are loaded with the allocated blocks from the
	* ms_sm spacemaps of the checkpoint. For each
	* one of these ranges we ensure that none of them
	* exists in the ms_allocatable trees of the
	* current state which are loaded with the ranges
	* that are currently free.
	*
	* This way we ensure that none of the blocks that
	* are part of the checkpoint were freed by mistake.
	*/
	range_tree_walk(ckpoint_msp->ms_allocatable,
	(range_tree_func_t *)range_tree_verify_not_present,
	current_msp->ms_allocatable);
	}
	}

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

	static void
	verify_checkpoint_blocks(spa_t *spa)
	{
	ASSERT(!dump_opt['L']);

	spa_t *checkpoint_spa;
	char *checkpoint_pool;
	int error = 0;

	/*
	* We import the checkpointed state of the pool (under a different
	* name) so we can do verification on it against the current state
	* of the pool.
	*/
	checkpoint_pool = import_checkpointed_state(spa->spa_name, NULL,
	NULL);
	ASSERT(strcmp(spa->spa_name, checkpoint_pool) != 0);

	error = spa_open(checkpoint_pool, &checkpoint_spa, FTAG);
	if (error != 0) {
	fatal("Tried to open pool \"%s\" but spa_open() failed with "
	"error %d\n", checkpoint_pool, error);
	}

	/*
	* Ensure that ranges in the checkpoint space maps of each vdev
	* are allocated according to the checkpointed state's metaslab
	* space maps.
	*/
	verify_checkpoint_vdev_spacemaps(checkpoint_spa, spa);

	/*
	* Ensure that allocated ranges in the checkpoint's metaslab
	* space maps remain allocated in the metaslab space maps of
	* the current state.
	*/
	verify_checkpoint_ms_spacemaps(checkpoint_spa, spa);

	/*
	* Once we are done, we get rid of the checkpointed state.
	*/
	spa_close(checkpoint_spa, FTAG);
	free(checkpoint_pool);
	}

	static void
	dump_leftover_checkpoint_blocks(spa_t *spa)
	{
	vdev_t *rvd = spa->spa_root_vdev;

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

	space_map_t *checkpoint_sm = NULL;
	uint64_t checkpoint_sm_obj;

	if (vd->vdev_top_zap == 0)
	continue;

	if (zap_contains(spa_meta_objset(spa), vd->vdev_top_zap,
	VDEV_TOP_ZAP_POOL_CHECKPOINT_SM) != 0)
	continue;

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

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

	static int
	verify_checkpoint(spa_t *spa)
	{
	uberblock_t checkpoint;
	int error;

	if (!spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT))
	return (0);

	error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_ZPOOL_CHECKPOINT, sizeof (uint64_t),
	sizeof (uberblock_t) / sizeof (uint64_t), &checkpoint);

	if (error == ENOENT && !dump_opt['L']) {
	/*
	* If the feature is active but the uberblock is missing
	* then we must be in the middle of discarding the
	* checkpoint.
	*/
	(void) printf("\nPartially discarded checkpoint "
	"state found:\n");
	if (dump_opt['m'] > 3)
	dump_leftover_checkpoint_blocks(spa);
	return (0);
	} else if (error != 0) {
	(void) printf("lookup error %d when looking for "
	"checkpointed uberblock in MOS\n", error);
	return (error);
	}
	dump_uberblock(&checkpoint, "\nCheckpointed uberblock found:\n", "\n");

	if (checkpoint.ub_checkpoint_txg == 0) {
	(void) printf("\nub_checkpoint_txg not set in checkpointed "
	"uberblock\n");
	error = 3;
	}

	if (error == 0 && !dump_opt['L'])
	verify_checkpoint_blocks(spa);

	return (error);
	}

	static void
	mos_leaks_cb(void *arg, uint64_t start, uint64_t size)
	{
	(void) arg;
	for (uint64_t i = start; i < size; i++) {
	(void) printf("MOS object %llu referenced but not allocated\n",
	(u_longlong_t)i);
	}
	}

	static void
	mos_obj_refd(uint64_t obj)
	{
	if (obj != 0 && mos_refd_objs != NULL)
	range_tree_add(mos_refd_objs, obj, 1);
	}

	/*
	* Call on a MOS object that may already have been referenced.
	*/
	static void
	mos_obj_refd_multiple(uint64_t obj)
	{
	if (obj != 0 && mos_refd_objs != NULL &&
	!range_tree_contains(mos_refd_objs, obj, 1))
	range_tree_add(mos_refd_objs, obj, 1);
	}

	static void
	mos_leak_vdev_top_zap(vdev_t *vd)
	{
	uint64_t ms_flush_data_obj;
	int error = zap_lookup(spa_meta_objset(vd->vdev_spa),
	vd->vdev_top_zap, VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS,
	sizeof (ms_flush_data_obj), 1, &ms_flush_data_obj);
	if (error == ENOENT)
	return;
	ASSERT0(error);

	mos_obj_refd(ms_flush_data_obj);
	}

	static void
	mos_leak_vdev(vdev_t *vd)
	{
	mos_obj_refd(vd->vdev_dtl_object);
	mos_obj_refd(vd->vdev_ms_array);
	mos_obj_refd(vd->vdev_indirect_config.vic_births_object);
	mos_obj_refd(vd->vdev_indirect_config.vic_mapping_object);
	mos_obj_refd(vd->vdev_leaf_zap);
	if (vd->vdev_checkpoint_sm != NULL)
	mos_obj_refd(vd->vdev_checkpoint_sm->sm_object);
	if (vd->vdev_indirect_mapping != NULL) {
	mos_obj_refd(vd->vdev_indirect_mapping->
	vim_phys->vimp_counts_object);
	}
	if (vd->vdev_obsolete_sm != NULL)
	mos_obj_refd(vd->vdev_obsolete_sm->sm_object);

	for (uint64_t m = 0; m < vd->vdev_ms_count; m++) {
	metaslab_t *ms = vd->vdev_ms[m];
	mos_obj_refd(space_map_object(ms->ms_sm));
	}

	if (vd->vdev_root_zap != 0)
	mos_obj_refd(vd->vdev_root_zap);

	if (vd->vdev_top_zap != 0) {
	mos_obj_refd(vd->vdev_top_zap);
	mos_leak_vdev_top_zap(vd);
	}

	for (uint64_t c = 0; c < vd->vdev_children; c++) {
	mos_leak_vdev(vd->vdev_child[c]);
	}
	}

	static void
	mos_leak_log_spacemaps(spa_t *spa)
	{
	uint64_t spacemap_zap;
	int error = zap_lookup(spa_meta_objset(spa),
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_LOG_SPACEMAP_ZAP,
	sizeof (spacemap_zap), 1, &spacemap_zap);
	if (error == ENOENT)
	return;
	ASSERT0(error);

	mos_obj_refd(spacemap_zap);
	for (spa_log_sm_t *sls = avl_first(&spa->spa_sm_logs_by_txg);
	sls; sls = AVL_NEXT(&spa->spa_sm_logs_by_txg, sls))
	mos_obj_refd(sls->sls_sm_obj);
	}

	static void
	errorlog_count_refd(objset_t *mos, uint64_t errlog)
	{
	zap_cursor_t zc;
	zap_attribute_t za;
	for (zap_cursor_init(&zc, mos, errlog);
	zap_cursor_retrieve(&zc, &za) == 0;
	zap_cursor_advance(&zc)) {
	mos_obj_refd(za.za_first_integer);
	}
	zap_cursor_fini(&zc);
	}

	static int
	dump_mos_leaks(spa_t *spa)
	{
	int rv = 0;
	objset_t *mos = spa->spa_meta_objset;
	dsl_pool_t *dp = spa->spa_dsl_pool;

	/* Visit and mark all referenced objects in the MOS */

	mos_obj_refd(DMU_POOL_DIRECTORY_OBJECT);
	mos_obj_refd(spa->spa_pool_props_object);
	mos_obj_refd(spa->spa_config_object);
	mos_obj_refd(spa->spa_ddt_stat_object);
	mos_obj_refd(spa->spa_feat_desc_obj);
	mos_obj_refd(spa->spa_feat_enabled_txg_obj);
	mos_obj_refd(spa->spa_feat_for_read_obj);
	mos_obj_refd(spa->spa_feat_for_write_obj);
	mos_obj_refd(spa->spa_history);
	mos_obj_refd(spa->spa_errlog_last);
	mos_obj_refd(spa->spa_errlog_scrub);

	if (spa_feature_is_enabled(spa, SPA_FEATURE_HEAD_ERRLOG)) {
	errorlog_count_refd(mos, spa->spa_errlog_last);
	errorlog_count_refd(mos, spa->spa_errlog_scrub);
	}

	mos_obj_refd(spa->spa_all_vdev_zaps);
	mos_obj_refd(spa->spa_dsl_pool->dp_bptree_obj);
	mos_obj_refd(spa->spa_dsl_pool->dp_tmp_userrefs_obj);
	mos_obj_refd(spa->spa_dsl_pool->dp_scan->scn_phys.scn_queue_obj);
	bpobj_count_refd(&spa->spa_deferred_bpobj);
	mos_obj_refd(dp->dp_empty_bpobj);
	bpobj_count_refd(&dp->dp_obsolete_bpobj);
	bpobj_count_refd(&dp->dp_free_bpobj);
	mos_obj_refd(spa->spa_l2cache.sav_object);
	mos_obj_refd(spa->spa_spares.sav_object);

	if (spa->spa_syncing_log_sm != NULL)
	mos_obj_refd(spa->spa_syncing_log_sm->sm_object);
	mos_leak_log_spacemaps(spa);

	mos_obj_refd(spa->spa_condensing_indirect_phys.
	scip_next_mapping_object);
	mos_obj_refd(spa->spa_condensing_indirect_phys.
	scip_prev_obsolete_sm_object);
	if (spa->spa_condensing_indirect_phys.scip_next_mapping_object != 0) {
	vdev_indirect_mapping_t *vim =
	vdev_indirect_mapping_open(mos,
	spa->spa_condensing_indirect_phys.scip_next_mapping_object);
	mos_obj_refd(vim->vim_phys->vimp_counts_object);
	vdev_indirect_mapping_close(vim);
	}
	deleted_livelists_dump_mos(spa);

	if (dp->dp_origin_snap != NULL) {
	dsl_dataset_t *ds;

	dsl_pool_config_enter(dp, FTAG);
	VERIFY0(dsl_dataset_hold_obj(dp,
	dsl_dataset_phys(dp->dp_origin_snap)->ds_next_snap_obj,
	FTAG, &ds));
	count_ds_mos_objects(ds);
	dump_blkptr_list(&ds->ds_deadlist, "Deadlist");
	dsl_dataset_rele(ds, FTAG);
	dsl_pool_config_exit(dp, FTAG);

	count_ds_mos_objects(dp->dp_origin_snap);
	dump_blkptr_list(&dp->dp_origin_snap->ds_deadlist, "Deadlist");
	}
	count_dir_mos_objects(dp->dp_mos_dir);
	if (dp->dp_free_dir != NULL)
	count_dir_mos_objects(dp->dp_free_dir);
	if (dp->dp_leak_dir != NULL)
	count_dir_mos_objects(dp->dp_leak_dir);

	mos_leak_vdev(spa->spa_root_vdev);

	for (uint64_t class = 0; class < DDT_CLASSES; class++) {
	for (uint64_t type = 0; type < DDT_TYPES; type++) {
	for (uint64_t cksum = 0;
	cksum < ZIO_CHECKSUM_FUNCTIONS; cksum++) {
	ddt_t *ddt = spa->spa_ddt[cksum];
	if (!ddt)
	continue;
	mos_obj_refd(ddt->ddt_object[type][class]);
	}
	}
	}

	if (spa->spa_brt != NULL) {
	brt_t *brt = spa->spa_brt;
	for (uint64_t vdevid = 0; vdevid < brt->brt_nvdevs; vdevid++) {
	brt_vdev_t *brtvd = &brt->brt_vdevs[vdevid];
	if (brtvd != NULL && brtvd->bv_initiated) {
	mos_obj_refd(brtvd->bv_mos_brtvdev);
	mos_obj_refd(brtvd->bv_mos_entries);
	}
	}
	}

	/*
	* Visit all allocated objects and make sure they are referenced.
	*/
	uint64_t object = 0;
	while (dmu_object_next(mos, &object, B_FALSE, 0) == 0) {
	if (range_tree_contains(mos_refd_objs, object, 1)) {
	range_tree_remove(mos_refd_objs, object, 1);
	} else {
	dmu_object_info_t doi;
	const char *name;
	VERIFY0(dmu_object_info(mos, object, &doi));
	if (doi.doi_type & DMU_OT_NEWTYPE) {
	dmu_object_byteswap_t bswap =
	DMU_OT_BYTESWAP(doi.doi_type);
	name = dmu_ot_byteswap[bswap].ob_name;
	} else {
	name = dmu_ot[doi.doi_type].ot_name;
	}

	(void) printf("MOS object %llu (%s) leaked\n",
	(u_longlong_t)object, name);
	rv = 2;
	}
	}
	(void) range_tree_walk(mos_refd_objs, mos_leaks_cb, NULL);
	if (!range_tree_is_empty(mos_refd_objs))
	rv = 2;
	range_tree_vacate(mos_refd_objs, NULL, NULL);
	range_tree_destroy(mos_refd_objs);
	return (rv);
	}

	typedef struct log_sm_obsolete_stats_arg {
	uint64_t lsos_current_txg;

	uint64_t lsos_total_entries;
	uint64_t lsos_valid_entries;

	uint64_t lsos_sm_entries;
	uint64_t lsos_valid_sm_entries;
	} log_sm_obsolete_stats_arg_t;

	static int
	log_spacemap_obsolete_stats_cb(spa_t spa, space_map_entry_t sme,
	uint64_t txg, void *arg)
	{
	log_sm_obsolete_stats_arg_t *lsos = arg;

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

	if (lsos->lsos_current_txg == 0) {
	/* this is the first log */
	lsos->lsos_current_txg = txg;
	} else if (lsos->lsos_current_txg < txg) {
	/* we just changed log - print stats and reset */
	(void) printf("%-8llu valid entries out of %-8llu - txg %llu\n",
	(u_longlong_t)lsos->lsos_valid_sm_entries,
	(u_longlong_t)lsos->lsos_sm_entries,
	(u_longlong_t)lsos->lsos_current_txg);
	lsos->lsos_valid_sm_entries = 0;
	lsos->lsos_sm_entries = 0;
	lsos->lsos_current_txg = txg;
	}
	ASSERT3U(lsos->lsos_current_txg, ==, txg);

	lsos->lsos_sm_entries++;
	lsos->lsos_total_entries++;

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

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

	if (txg < metaslab_unflushed_txg(ms))
	return (0);
	lsos->lsos_valid_sm_entries++;
	lsos->lsos_valid_entries++;
	return (0);
	}

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

	log_sm_obsolete_stats_arg_t lsos = {0};

	(void) printf("Log Space Map Obsolete Entry Statistics:\n");

	iterate_through_spacemap_logs(spa,
	log_spacemap_obsolete_stats_cb, &lsos);

	/* print stats for latest log */
	(void) printf("%-8llu valid entries out of %-8llu - txg %llu\n",
	(u_longlong_t)lsos.lsos_valid_sm_entries,
	(u_longlong_t)lsos.lsos_sm_entries,
	(u_longlong_t)lsos.lsos_current_txg);

	(void) printf("%-8llu valid entries out of %-8llu - total\n\n",
	(u_longlong_t)lsos.lsos_valid_entries,
	(u_longlong_t)lsos.lsos_total_entries);
	}

	static void
	dump_zpool(spa_t *spa)
	{
	dsl_pool_t *dp = spa_get_dsl(spa);
	int rc = 0;

	if (dump_opt['y']) {
	livelist_metaslab_validate(spa);
	}

	if (dump_opt['S']) {
	dump_simulated_ddt(spa);
	return;
	}

	if (!dump_opt['e'] && dump_opt['C'] > 1) {
	(void) printf("\nCached configuration:\n");
	dump_nvlist(spa->spa_config, 8);
	}

	if (dump_opt['C'])
	dump_config(spa);

	if (dump_opt['u'])
	dump_uberblock(&spa->spa_uberblock, "\nUberblock:\n", "\n");

	if (dump_opt['D'])
	dump_all_ddts(spa);

	if (dump_opt['T'])
	dump_brt(spa);

	if (dump_opt['d'] > 2 \|\| dump_opt['m'])
	dump_metaslabs(spa);
	if (dump_opt['M'])
	dump_metaslab_groups(spa, dump_opt['M'] > 1);
	if (dump_opt['d'] > 2 \|\| dump_opt['m']) {
	dump_log_spacemaps(spa);
	dump_log_spacemap_obsolete_stats(spa);
	}

	if (dump_opt['d'] \|\| dump_opt['i']) {
	spa_feature_t f;
	mos_refd_objs = range_tree_create(NULL, RANGE_SEG64, NULL, 0,
	0);
	dump_objset(dp->dp_meta_objset);

	if (dump_opt['d'] >= 3) {
	dsl_pool_t *dp = spa->spa_dsl_pool;
	dump_full_bpobj(&spa->spa_deferred_bpobj,
	"Deferred frees", 0);
	if (spa_version(spa) >= SPA_VERSION_DEADLISTS) {
	dump_full_bpobj(&dp->dp_free_bpobj,
	"Pool snapshot frees", 0);
	}
	if (bpobj_is_open(&dp->dp_obsolete_bpobj)) {
	ASSERT(spa_feature_is_enabled(spa,
	SPA_FEATURE_DEVICE_REMOVAL));
	dump_full_bpobj(&dp->dp_obsolete_bpobj,
	"Pool obsolete blocks", 0);
	}

	if (spa_feature_is_active(spa,
	SPA_FEATURE_ASYNC_DESTROY)) {
	dump_bptree(spa->spa_meta_objset,
	dp->dp_bptree_obj,
	"Pool dataset frees");
	}
	dump_dtl(spa->spa_root_vdev, 0);
	}

	for (spa_feature_t f = 0; f < SPA_FEATURES; f++)
	global_feature_count[f] = UINT64_MAX;
	global_feature_count[SPA_FEATURE_REDACTION_BOOKMARKS] = 0;
	global_feature_count[SPA_FEATURE_REDACTION_LIST_SPILL] = 0;
	global_feature_count[SPA_FEATURE_BOOKMARK_WRITTEN] = 0;
	global_feature_count[SPA_FEATURE_LIVELIST] = 0;

	(void) dmu_objset_find(spa_name(spa), dump_one_objset,
	NULL, DS_FIND_SNAPSHOTS \| DS_FIND_CHILDREN);

	if (rc == 0 && !dump_opt['L'])
	rc = dump_mos_leaks(spa);

	for (f = 0; f < SPA_FEATURES; f++) {
	uint64_t refcount;

	uint64_t *arr;
	if (!(spa_feature_table[f].fi_flags &
	ZFEATURE_FLAG_PER_DATASET)) {
	if (global_feature_count[f] == UINT64_MAX)
	continue;
	if (!spa_feature_is_enabled(spa, f)) {
	ASSERT0(global_feature_count[f]);
	continue;
	}
	arr = global_feature_count;
	} else {
	if (!spa_feature_is_enabled(spa, f)) {
	ASSERT0(dataset_feature_count[f]);
	continue;
	}
	arr = dataset_feature_count;
	}
	if (feature_get_refcount(spa, &spa_feature_table[f],
	&refcount) == ENOTSUP)
	continue;
	if (arr[f] != refcount) {
	(void) printf("%s feature refcount mismatch: "
	"%lld consumers != %lld refcount\n",
	spa_feature_table[f].fi_uname,
	(longlong_t)arr[f], (longlong_t)refcount);
	rc = 2;
	} else {
	(void) printf("Verified %s feature refcount "
	"of %llu is correct\n",
	spa_feature_table[f].fi_uname,
	(longlong_t)refcount);
	}
	}

	if (rc == 0)
	rc = verify_device_removal_feature_counts(spa);
	}

	if (rc == 0 && (dump_opt['b'] \|\| dump_opt['c']))
	rc = dump_block_stats(spa);

	if (rc == 0)
	rc = verify_spacemap_refcounts(spa);

	if (dump_opt['s'])
	show_pool_stats(spa);

	if (dump_opt['h'])
	dump_history(spa);

	if (rc == 0)
	rc = verify_checkpoint(spa);

	if (rc != 0) {
	dump_debug_buffer();
	- exit(rc);
	+ zdb_exit(rc);
	}
	}

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

	static int flagbits[256];
	static char flagbitstr[16];

	static void
	zdb_print_blkptr(const blkptr_t *bp, int flags)
	{
	char blkbuf[BP_SPRINTF_LEN];

	if (flags & ZDB_FLAG_BSWAP)
	byteswap_uint64_array((void *)bp, sizeof (blkptr_t));

	snprintf_blkptr(blkbuf, sizeof (blkbuf), bp);
	(void) printf("%s\n", blkbuf);
	}

	static void
	zdb_dump_indirect(blkptr_t *bp, int nbps, int flags)
	{
	int i;

	for (i = 0; i < nbps; i++)
	zdb_print_blkptr(&bp[i], flags);
	}

	static void
	zdb_dump_gbh(void *buf, int flags)
	{
	zdb_dump_indirect((blkptr_t *)buf, SPA_GBH_NBLKPTRS, flags);
	}

	static void
	zdb_dump_block_raw(void *buf, uint64_t size, int flags)
	{
	if (flags & ZDB_FLAG_BSWAP)
	byteswap_uint64_array(buf, size);
	VERIFY(write(fileno(stdout), buf, size) == size);
	}

	static void
	zdb_dump_block(char label, void buf, uint64_t size, int flags)
	{
	uint64_t d = (uint64_t )buf;
	unsigned nwords = size / sizeof (uint64_t);
	int do_bswap = !!(flags & ZDB_FLAG_BSWAP);
	unsigned i, j;
	const char *hdr;
	char *c;


	if (do_bswap)
	hdr = " 7 6 5 4 3 2 1 0 f e d c b a 9 8";
	else
	hdr = " 0 1 2 3 4 5 6 7 8 9 a b c d e f";

	(void) printf("\n%s\n%6s %s 0123456789abcdef\n", label, "", hdr);

	#ifdef _LITTLE_ENDIAN
	/* correct the endianness */
	do_bswap = !do_bswap;
	#endif
	for (i = 0; i < nwords; i += 2) {
	(void) printf("%06llx: %016llx %016llx ",
	(u_longlong_t)(i * sizeof (uint64_t)),
	(u_longlong_t)(do_bswap ? BSWAP_64(d[i]) : d[i]),
	(u_longlong_t)(do_bswap ? BSWAP_64(d[i + 1]) : d[i + 1]));

	c = (char *)&d[i];
	for (j = 0; j < 2 * sizeof (uint64_t); j++)
	(void) printf("%c", isprint(c[j]) ? c[j] : '.');
	(void) printf("\n");
	}
	}

	/*
	* There are two acceptable formats:
	* leaf_name - For example: c1t0d0 or /tmp/ztest.0a
	* child[.child]* - For example: 0.1.1
	*
	* The second form can be used to specify arbitrary vdevs anywhere
	* in the hierarchy. For example, in a pool with a mirror of
	* RAID-Zs, you can specify either RAID-Z vdev with 0.0 or 0.1 .
	*/
	static vdev_t *
	zdb_vdev_lookup(vdev_t vdev, const char path)
	{
	char s, p, *q;
	unsigned i;

	if (vdev == NULL)
	return (NULL);

	/* First, assume the x.x.x.x format */
	i = strtoul(path, &s, 10);
	if (s == path \|\| (s && s != '.' && s != '\0'))
	goto name;
	if (i >= vdev->vdev_children)
	return (NULL);

	vdev = vdev->vdev_child[i];
	if (s && *s == '\0')
	return (vdev);
	return (zdb_vdev_lookup(vdev, s+1));

	name:
	for (i = 0; i < vdev->vdev_children; i++) {
	vdev_t *vc = vdev->vdev_child[i];

	if (vc->vdev_path == NULL) {
	vc = zdb_vdev_lookup(vc, path);
	if (vc == NULL)
	continue;
	else
	return (vc);
	}

	p = strrchr(vc->vdev_path, '/');
	p = p ? p + 1 : vc->vdev_path;
	q = &vc->vdev_path[strlen(vc->vdev_path) - 2];

	if (strcmp(vc->vdev_path, path) == 0)
	return (vc);
	if (strcmp(p, path) == 0)
	return (vc);
	if (strcmp(q, "s0") == 0 && strncmp(p, path, q - p) == 0)
	return (vc);
	}

	return (NULL);
	}

	static int
	name_from_objset_id(spa_t spa, uint64_t objset_id, char outstr)
	{
	dsl_dataset_t *ds;

	dsl_pool_config_enter(spa->spa_dsl_pool, FTAG);
	int error = dsl_dataset_hold_obj(spa->spa_dsl_pool, objset_id,
	NULL, &ds);
	if (error != 0) {
	(void) fprintf(stderr, "failed to hold objset %llu: %s\n",
	(u_longlong_t)objset_id, strerror(error));
	dsl_pool_config_exit(spa->spa_dsl_pool, FTAG);
	return (error);
	}
	dsl_dataset_name(ds, outstr);
	dsl_dataset_rele(ds, NULL);
	dsl_pool_config_exit(spa->spa_dsl_pool, FTAG);
	return (0);
	}

	static boolean_t
	zdb_parse_block_sizes(char sizes, uint64_t lsize, uint64_t *psize)
	{
	char s0, s1, *tmp = NULL;

	if (sizes == NULL)
	return (B_FALSE);

	s0 = strtok_r(sizes, "/", &tmp);
	if (s0 == NULL)
	return (B_FALSE);
	s1 = strtok_r(NULL, "/", &tmp);
	*lsize = strtoull(s0, NULL, 16);
	psize = s1 ? strtoull(s1, NULL, 16) : lsize;
	return (lsize >= psize && *psize > 0);
	}

	#define ZIO_COMPRESS_MASK(alg) (1ULL << (ZIO_COMPRESS_##alg))

	static boolean_t
	try_decompress_block(abd_t *pabd, uint64_t lsize, uint64_t psize,
	int flags, int cfunc, void lbuf, void lbuf2)
	{
	if (flags & ZDB_FLAG_VERBOSE) {
	(void) fprintf(stderr,
	"Trying %05llx -> %05llx (%s)\n",
	(u_longlong_t)psize,
	(u_longlong_t)lsize,
	zio_compress_table[cfunc].ci_name);
	}

	/*
	* We set lbuf to all zeros and lbuf2 to all
	* ones, then decompress to both buffers and
	* compare their contents. This way we can
	* know if decompression filled exactly to
	* lsize or if it left some bytes unwritten.
	*/

	memset(lbuf, 0x00, lsize);
	memset(lbuf2, 0xff, lsize);

	if (zio_decompress_data(cfunc, pabd,
	lbuf, psize, lsize, NULL) == 0 &&
	zio_decompress_data(cfunc, pabd,
	lbuf2, psize, lsize, NULL) == 0 &&
	memcmp(lbuf, lbuf2, lsize) == 0)
	return (B_TRUE);
	return (B_FALSE);
	}

	static uint64_t
	zdb_decompress_block(abd_t pabd, void buf, void *lbuf, uint64_t lsize,
	uint64_t psize, int flags)
	{
	(void) buf;
	uint64_t orig_lsize = lsize;
	boolean_t tryzle = ((getenv("ZDB_NO_ZLE") == NULL));
	boolean_t found = B_FALSE;
	/*
	* We don't know how the data was compressed, so just try
	* every decompress function at every inflated blocksize.
	*/
	void *lbuf2 = umem_alloc(SPA_MAXBLOCKSIZE, UMEM_NOFAIL);
	int cfuncs[ZIO_COMPRESS_FUNCTIONS] = { 0 };
	int *cfuncp = cfuncs;
	uint64_t maxlsize = SPA_MAXBLOCKSIZE;
	uint64_t mask = ZIO_COMPRESS_MASK(ON) \| ZIO_COMPRESS_MASK(OFF) \|
	ZIO_COMPRESS_MASK(INHERIT) \| ZIO_COMPRESS_MASK(EMPTY) \|
	ZIO_COMPRESS_MASK(ZLE);
	*cfuncp++ = ZIO_COMPRESS_LZ4;
	*cfuncp++ = ZIO_COMPRESS_LZJB;
	mask \|= ZIO_COMPRESS_MASK(LZ4) \| ZIO_COMPRESS_MASK(LZJB);
	/*
	* Every gzip level has the same decompressor, no need to
	* run it 9 times per bruteforce attempt.
	*/
	mask \|= ZIO_COMPRESS_MASK(GZIP_2) \| ZIO_COMPRESS_MASK(GZIP_3);
	mask \|= ZIO_COMPRESS_MASK(GZIP_4) \| ZIO_COMPRESS_MASK(GZIP_5);
	mask \|= ZIO_COMPRESS_MASK(GZIP_6) \| ZIO_COMPRESS_MASK(GZIP_7);
	mask \|= ZIO_COMPRESS_MASK(GZIP_8) \| ZIO_COMPRESS_MASK(GZIP_9);
	for (int c = 0; c < ZIO_COMPRESS_FUNCTIONS; c++)
	if (((1ULL << c) & mask) == 0)
	*cfuncp++ = c;

	/*
	* On the one hand, with SPA_MAXBLOCKSIZE at 16MB, this
	* could take a while and we should let the user know
	* we are not stuck. On the other hand, printing progress
	* info gets old after a while. User can specify 'v' flag
	* to see the progression.
	*/
	if (lsize == psize)
	lsize += SPA_MINBLOCKSIZE;
	else
	maxlsize = lsize;

	for (; lsize <= maxlsize; lsize += SPA_MINBLOCKSIZE) {
	for (cfuncp = cfuncs; *cfuncp; cfuncp++) {
	if (try_decompress_block(pabd, lsize, psize, flags,
	*cfuncp, lbuf, lbuf2)) {
	found = B_TRUE;
	break;
	}
	}
	if (*cfuncp != 0)
	break;
	}
	if (!found && tryzle) {
	for (lsize = orig_lsize; lsize <= maxlsize;
	lsize += SPA_MINBLOCKSIZE) {
	if (try_decompress_block(pabd, lsize, psize, flags,
	ZIO_COMPRESS_ZLE, lbuf, lbuf2)) {
	*cfuncp = ZIO_COMPRESS_ZLE;
	found = B_TRUE;
	break;
	}
	}
	}
	umem_free(lbuf2, SPA_MAXBLOCKSIZE);

	if (*cfuncp == ZIO_COMPRESS_ZLE) {
	printf("\nZLE decompression was selected. If you "
	"suspect the results are wrong,\ntry avoiding ZLE "
	"by setting and exporting ZDB_NO_ZLE=\"true\"\n");
	}

	return (lsize > maxlsize ? -1 : lsize);
	}

	/*
	* Read a block from a pool and print it out. The syntax of the
	* block descriptor is:
	*
	* pool:vdev_specifier:offset:[lsize/]psize[:flags]
	*
	* pool - The name of the pool you wish to read from
	* vdev_specifier - Which vdev (see comment for zdb_vdev_lookup)
	* offset - offset, in hex, in bytes
	* size - Amount of data to read, in hex, in bytes
	* flags - A string of characters specifying options
	* b: Decode a blkptr at given offset within block
	* c: Calculate and display checksums
	* d: Decompress data before dumping
	* e: Byteswap data before dumping
	* g: Display data as a gang block header
	* i: Display as an indirect block
	* r: Dump raw data to stdout
	* v: Verbose
	*
	*/
	static void
	zdb_read_block(char thing, spa_t spa)
	{
	blkptr_t blk, *bp = &blk;
	dva_t *dva = bp->blk_dva;
	int flags = 0;
	uint64_t offset = 0, psize = 0, lsize = 0, blkptr_offset = 0;
	zio_t *zio;
	vdev_t *vd;
	abd_t *pabd;
	void lbuf, buf;
	char s, p, dup, flagstr, sizes, tmp = NULL;
	const char vdev, errmsg = NULL;
	int i, error;
	boolean_t borrowed = B_FALSE, found = B_FALSE;

	dup = strdup(thing);
	s = strtok_r(dup, ":", &tmp);
	vdev = s ?: "";
	s = strtok_r(NULL, ":", &tmp);
	offset = strtoull(s ? s : "", NULL, 16);
	sizes = strtok_r(NULL, ":", &tmp);
	s = strtok_r(NULL, ":", &tmp);
	flagstr = strdup(s ?: "");

	if (!zdb_parse_block_sizes(sizes, &lsize, &psize))
	errmsg = "invalid size(s)";
	if (!IS_P2ALIGNED(psize, DEV_BSIZE) \|\| !IS_P2ALIGNED(lsize, DEV_BSIZE))
	errmsg = "size must be a multiple of sector size";
	if (!IS_P2ALIGNED(offset, DEV_BSIZE))
	errmsg = "offset must be a multiple of sector size";
	if (errmsg) {
	(void) printf("Invalid block specifier: %s - %s\n",
	thing, errmsg);
	goto done;
	}

	tmp = NULL;
	for (s = strtok_r(flagstr, ":", &tmp);
	s != NULL;
	s = strtok_r(NULL, ":", &tmp)) {
	for (i = 0; i < strlen(flagstr); i++) {
	int bit = flagbits[(uchar_t)flagstr[i]];

	if (bit == 0) {
	(void) printf("***Ignoring flag: %c\n",
	(uchar_t)flagstr[i]);
	continue;
	}
	found = B_TRUE;
	flags \|= bit;

	p = &flagstr[i + 1];
	if (p != ':' && p != '\0') {
	int j = 0, nextbit = flagbits[(uchar_t)*p];
	char *end, offstr[8] = { 0 };
	if ((bit == ZDB_FLAG_PRINT_BLKPTR) &&
	(nextbit == 0)) {
	/* look ahead to isolate the offset */
	while (nextbit == 0 &&
	strchr(flagbitstr, *p) == NULL) {
	offstr[j] = *p;
	j++;
	if (i + j > strlen(flagstr))
	break;
	p++;
	nextbit = flagbits[(uchar_t)*p];
	}
	blkptr_offset = strtoull(offstr, &end,
	16);
	i += j;
	} else if (nextbit == 0) {
	(void) printf("***Ignoring flag arg:"
	" '%c'\n", (uchar_t)*p);
	}
	}
	}
	}
	if (blkptr_offset % sizeof (blkptr_t)) {
	printf("Block pointer offset 0x%llx "
	"must be divisible by 0x%x\n",
	(longlong_t)blkptr_offset, (int)sizeof (blkptr_t));
	goto done;
	}
	if (found == B_FALSE && strlen(flagstr) > 0) {
	printf("Invalid flag arg: '%s'\n", flagstr);
	goto done;
	}

	vd = zdb_vdev_lookup(spa->spa_root_vdev, vdev);
	if (vd == NULL) {
	(void) printf("***Invalid vdev: %s\n", vdev);
	goto done;
	} else {
	if (vd->vdev_path)
	(void) fprintf(stderr, "Found vdev: %s\n",
	vd->vdev_path);
	else
	(void) fprintf(stderr, "Found vdev type: %s\n",
	vd->vdev_ops->vdev_op_type);
	}

	pabd = abd_alloc_for_io(SPA_MAXBLOCKSIZE, B_FALSE);
	lbuf = umem_alloc(SPA_MAXBLOCKSIZE, UMEM_NOFAIL);

	BP_ZERO(bp);

	DVA_SET_VDEV(&dva[0], vd->vdev_id);
	DVA_SET_OFFSET(&dva[0], offset);
	DVA_SET_GANG(&dva[0], !!(flags & ZDB_FLAG_GBH));
	DVA_SET_ASIZE(&dva[0], vdev_psize_to_asize(vd, psize));

	BP_SET_BIRTH(bp, TXG_INITIAL, TXG_INITIAL);

	BP_SET_LSIZE(bp, lsize);
	BP_SET_PSIZE(bp, psize);
	BP_SET_COMPRESS(bp, ZIO_COMPRESS_OFF);
	BP_SET_CHECKSUM(bp, ZIO_CHECKSUM_OFF);
	BP_SET_TYPE(bp, DMU_OT_NONE);
	BP_SET_LEVEL(bp, 0);
	BP_SET_DEDUP(bp, 0);
	BP_SET_BYTEORDER(bp, ZFS_HOST_BYTEORDER);

	spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
	zio = zio_root(spa, NULL, NULL, 0);

	if (vd == vd->vdev_top) {
	/*
	* Treat this as a normal block read.
	*/
	zio_nowait(zio_read(zio, spa, bp, pabd, psize, NULL, NULL,
	ZIO_PRIORITY_SYNC_READ,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_RAW, NULL));
	} else {
	/*
	* Treat this as a vdev child I/O.
	*/
	zio_nowait(zio_vdev_child_io(zio, bp, vd, offset, pabd,
	psize, ZIO_TYPE_READ, ZIO_PRIORITY_SYNC_READ,
	ZIO_FLAG_DONT_PROPAGATE \| ZIO_FLAG_DONT_RETRY \|
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_RAW \| ZIO_FLAG_OPTIONAL,
	NULL, NULL));
	}

	error = zio_wait(zio);
	spa_config_exit(spa, SCL_STATE, FTAG);

	if (error) {
	(void) printf("Read of %s failed, error: %d\n", thing, error);
	goto out;
	}

	uint64_t orig_lsize = lsize;
	buf = lbuf;
	if (flags & ZDB_FLAG_DECOMPRESS) {
	lsize = zdb_decompress_block(pabd, buf, lbuf,
	lsize, psize, flags);
	if (lsize == -1) {
	(void) printf("Decompress of %s failed\n", thing);
	goto out;
	}
	} else {
	buf = abd_borrow_buf_copy(pabd, lsize);
	borrowed = B_TRUE;
	}
	/*
	* Try to detect invalid block pointer. If invalid, try
	* decompressing.
	*/
	if ((flags & ZDB_FLAG_PRINT_BLKPTR \|\| flags & ZDB_FLAG_INDIRECT) &&
	!(flags & ZDB_FLAG_DECOMPRESS)) {
	const blkptr_t b = (const blkptr_t )(void *)
	((uintptr_t)buf + (uintptr_t)blkptr_offset);
	if (zfs_blkptr_verify(spa, b,
	BLK_CONFIG_NEEDED, BLK_VERIFY_ONLY) == B_FALSE) {
	abd_return_buf_copy(pabd, buf, lsize);
	borrowed = B_FALSE;
	buf = lbuf;
	lsize = zdb_decompress_block(pabd, buf,
	lbuf, lsize, psize, flags);
	b = (const blkptr_t )(void )
	((uintptr_t)buf + (uintptr_t)blkptr_offset);
	if (lsize == -1 \|\| zfs_blkptr_verify(spa, b,
	BLK_CONFIG_NEEDED, BLK_VERIFY_LOG) == B_FALSE) {
	printf("invalid block pointer at this DVA\n");
	goto out;
	}
	}
	}

	if (flags & ZDB_FLAG_PRINT_BLKPTR)
	zdb_print_blkptr((blkptr_t )(void )
	((uintptr_t)buf + (uintptr_t)blkptr_offset), flags);
	else if (flags & ZDB_FLAG_RAW)
	zdb_dump_block_raw(buf, lsize, flags);
	else if (flags & ZDB_FLAG_INDIRECT)
	zdb_dump_indirect((blkptr_t *)buf,
	orig_lsize / sizeof (blkptr_t), flags);
	else if (flags & ZDB_FLAG_GBH)
	zdb_dump_gbh(buf, flags);
	else
	zdb_dump_block(thing, buf, lsize, flags);

	/*
	* If :c was specified, iterate through the checksum table to
	* calculate and display each checksum for our specified
	* DVA and length.
	*/
	if ((flags & ZDB_FLAG_CHECKSUM) && !(flags & ZDB_FLAG_RAW) &&
	!(flags & ZDB_FLAG_GBH)) {
	zio_t *czio;
	(void) printf("\n");
	for (enum zio_checksum ck = ZIO_CHECKSUM_LABEL;
	ck < ZIO_CHECKSUM_FUNCTIONS; ck++) {

	if ((zio_checksum_table[ck].ci_flags &
	ZCHECKSUM_FLAG_EMBEDDED) \|\|
	ck == ZIO_CHECKSUM_NOPARITY) {
	continue;
	}
	BP_SET_CHECKSUM(bp, ck);
	spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
	czio = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL);
	if (vd == vd->vdev_top) {
	zio_nowait(zio_read(czio, spa, bp, pabd, psize,
	NULL, NULL,
	ZIO_PRIORITY_SYNC_READ,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_RAW \|
	ZIO_FLAG_DONT_RETRY, NULL));
	} else {
	zio_nowait(zio_vdev_child_io(czio, bp, vd,
	offset, pabd, psize, ZIO_TYPE_READ,
	ZIO_PRIORITY_SYNC_READ,
	ZIO_FLAG_DONT_PROPAGATE \|
	ZIO_FLAG_DONT_RETRY \|
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_RAW \|
	ZIO_FLAG_SPECULATIVE \|
	ZIO_FLAG_OPTIONAL, NULL, NULL));
	}
	error = zio_wait(czio);
	if (error == 0 \|\| error == ECKSUM) {
	zio_t *ck_zio = zio_null(NULL, spa, NULL,
	NULL, NULL, 0);
	ck_zio->io_offset =
	DVA_GET_OFFSET(&bp->blk_dva[0]);
	ck_zio->io_bp = bp;
	zio_checksum_compute(ck_zio, ck, pabd, lsize);
	printf(
	"%12s\t"
	"cksum=%016llx:%016llx:%016llx:%016llx\n",
	zio_checksum_table[ck].ci_name,
	(u_longlong_t)bp->blk_cksum.zc_word[0],
	(u_longlong_t)bp->blk_cksum.zc_word[1],
	(u_longlong_t)bp->blk_cksum.zc_word[2],
	(u_longlong_t)bp->blk_cksum.zc_word[3]);
	zio_wait(ck_zio);
	} else {
	printf("error %d reading block\n", error);
	}
	spa_config_exit(spa, SCL_STATE, FTAG);
	}
	}

	if (borrowed)
	abd_return_buf_copy(pabd, buf, lsize);

	out:
	abd_free(pabd);
	umem_free(lbuf, SPA_MAXBLOCKSIZE);
	done:
	free(flagstr);
	free(dup);
	}

	static void
	zdb_embedded_block(char *thing)
	{
	blkptr_t bp = {{{{0}}}};
	unsigned long long words = (void )&bp;
	char *buf;
	int err;

	err = sscanf(thing, "%llx:%llx:%llx:%llx:%llx:%llx:%llx:%llx:"
	"%llx:%llx:%llx:%llx:%llx:%llx:%llx:%llx",
	words + 0, words + 1, words + 2, words + 3,
	words + 4, words + 5, words + 6, words + 7,
	words + 8, words + 9, words + 10, words + 11,
	words + 12, words + 13, words + 14, words + 15);
	if (err != 16) {
	(void) fprintf(stderr, "invalid input format\n");
	- exit(1);
	+ zdb_exit(1);
	}
	ASSERT3U(BPE_GET_LSIZE(&bp), <=, SPA_MAXBLOCKSIZE);
	buf = malloc(SPA_MAXBLOCKSIZE);
	if (buf == NULL) {
	(void) fprintf(stderr, "out of memory\n");
	- exit(1);
	+ zdb_exit(1);
	}
	err = decode_embedded_bp(&bp, buf, BPE_GET_LSIZE(&bp));
	if (err != 0) {
	(void) fprintf(stderr, "decode failed: %u\n", err);
	- exit(1);
	+ zdb_exit(1);
	}
	zdb_dump_block_raw(buf, BPE_GET_LSIZE(&bp), 0);
	free(buf);
	}

	/* check for valid hex or decimal numeric string */
	static boolean_t
	zdb_numeric(char *str)
	{
	int i = 0;

	if (strlen(str) == 0)
	return (B_FALSE);
	if (strncmp(str, "0x", 2) == 0 \|\| strncmp(str, "0X", 2) == 0)
	i = 2;
	for (; i < strlen(str); i++) {
	if (!isxdigit(str[i]))
	return (B_FALSE);
	}
	return (B_TRUE);
	}

	int
	main(int argc, char **argv)
	{
	int c;
	- spa_t *spa = NULL;
	- objset_t *os = NULL;
	int dump_all = 1;
	int verbose = 0;
	int error = 0;
	char **searchdirs = NULL;
	int nsearch = 0;
	char target, target_pool, dsname[ZFS_MAX_DATASET_NAME_LEN];
	nvlist_t *policy = NULL;
	uint64_t max_txg = UINT64_MAX;
	int64_t objset_id = -1;
	uint64_t object;
	int flags = ZFS_IMPORT_MISSING_LOG;
	int rewind = ZPOOL_NEVER_REWIND;
	char spa_config_path_env, objset_str;
	boolean_t target_is_spa = B_TRUE, dataset_lookup = B_FALSE;
	nvlist_t *cfg = NULL;
	+ struct sigaction action;

	dprintf_setup(&argc, argv);

	+ /*
	+ * Set up signal handlers, so if we crash due to bad on-disk data we
	+ * can get more info. Unlike ztest, we don't bail out if we can't set
	+ * up signal handlers, because zdb is very useful without them.
	+ */
	+ action.sa_handler = sig_handler;
	+ sigemptyset(&action.sa_mask);
	+ action.sa_flags = 0;
	+ if (sigaction(SIGSEGV, &action, NULL) < 0) {
	+ (void) fprintf(stderr, "zdb: cannot catch SIGSEGV: %s\n",
	+ strerror(errno));
	+ }
	+ if (sigaction(SIGABRT, &action, NULL) < 0) {
	+ (void) fprintf(stderr, "zdb: cannot catch SIGABRT: %s\n",
	+ strerror(errno));
	+ }
	+
	/*
	* If there is an environment variable SPA_CONFIG_PATH it overrides
	* default spa_config_path setting. If -U flag is specified it will
	* override this environment variable settings once again.
	*/
	spa_config_path_env = getenv("SPA_CONFIG_PATH");
	if (spa_config_path_env != NULL)
	spa_config_path = spa_config_path_env;

	/*
	* For performance reasons, we set this tunable down. We do so before
	* the arg parsing section so that the user can override this value if
	* they choose.
	*/
	zfs_btree_verify_intensity = 3;

	struct option long_options[] = {
	{"ignore-assertions", no_argument, NULL, 'A'},
	{"block-stats", no_argument, NULL, 'b'},
	{"backup", no_argument, NULL, 'B'},
	{"checksum", no_argument, NULL, 'c'},
	{"config", no_argument, NULL, 'C'},
	{"datasets", no_argument, NULL, 'd'},
	{"dedup-stats", no_argument, NULL, 'D'},
	{"exported", no_argument, NULL, 'e'},
	{"embedded-block-pointer", no_argument, NULL, 'E'},
	{"automatic-rewind", no_argument, NULL, 'F'},
	{"dump-debug-msg", no_argument, NULL, 'G'},
	{"history", no_argument, NULL, 'h'},
	{"intent-logs", no_argument, NULL, 'i'},
	{"inflight", required_argument, NULL, 'I'},
	{"checkpointed-state", no_argument, NULL, 'k'},
	{"key", required_argument, NULL, 'K'},
	{"label", no_argument, NULL, 'l'},
	{"disable-leak-tracking", no_argument, NULL, 'L'},
	{"metaslabs", no_argument, NULL, 'm'},
	{"metaslab-groups", no_argument, NULL, 'M'},
	{"numeric", no_argument, NULL, 'N'},
	{"option", required_argument, NULL, 'o'},
	{"object-lookups", no_argument, NULL, 'O'},
	{"path", required_argument, NULL, 'p'},
	{"parseable", no_argument, NULL, 'P'},
	{"skip-label", no_argument, NULL, 'q'},
	{"copy-object", no_argument, NULL, 'r'},
	{"read-block", no_argument, NULL, 'R'},
	{"io-stats", no_argument, NULL, 's'},
	{"simulate-dedup", no_argument, NULL, 'S'},
	{"txg", required_argument, NULL, 't'},
	{"brt-stats", no_argument, NULL, 'T'},
	{"uberblock", no_argument, NULL, 'u'},
	{"cachefile", required_argument, NULL, 'U'},
	{"verbose", no_argument, NULL, 'v'},
	{"verbatim", no_argument, NULL, 'V'},
	{"dump-blocks", required_argument, NULL, 'x'},
	{"extreme-rewind", no_argument, NULL, 'X'},
	{"all-reconstruction", no_argument, NULL, 'Y'},
	{"livelist", no_argument, NULL, 'y'},
	{"zstd-headers", no_argument, NULL, 'Z'},
	{0, 0, 0, 0}
	};

	while ((c = getopt_long(argc, argv,
	"AbBcCdDeEFGhiI:kK:lLmMNo:Op:PqrRsSt:TuU:vVx:XYyZ",
	long_options, NULL)) != -1) {
	switch (c) {
	case 'b':
	case 'B':
	case 'c':
	case 'C':
	case 'd':
	case 'D':
	case 'E':
	case 'G':
	case 'h':
	case 'i':
	case 'l':
	case 'm':
	case 'M':
	case 'N':
	case 'O':
	case 'r':
	case 'R':
	case 's':
	case 'S':
	case 'T':
	case 'u':
	case 'y':
	case 'Z':
	dump_opt[c]++;
	dump_all = 0;
	break;
	case 'A':
	case 'e':
	case 'F':
	case 'k':
	case 'L':
	case 'P':
	case 'q':
	case 'X':
	dump_opt[c]++;
	break;
	case 'Y':
	zfs_reconstruct_indirect_combinations_max = INT_MAX;
	zfs_deadman_enabled = 0;
	break;
	/* NB: Sort single match options below. */
	case 'I':
	max_inflight_bytes = strtoull(optarg, NULL, 0);
	if (max_inflight_bytes == 0) {
	(void) fprintf(stderr, "maximum number "
	"of inflight bytes must be greater "
	"than 0\n");
	usage();
	}
	break;
	case 'K':
	dump_opt[c]++;
	key_material = strdup(optarg);
	/* redact key material in process table */
	while (optarg != '\0') { optarg++ = '*'; }
	break;
	case 'o':
	error = set_global_var(optarg);
	if (error != 0)
	usage();
	break;
	case 'p':
	if (searchdirs == NULL) {
	searchdirs = umem_alloc(sizeof (char *),
	UMEM_NOFAIL);
	} else {
	char *tmp = umem_alloc((nsearch + 1)
	sizeof (char *), UMEM_NOFAIL);
	memcpy(tmp, searchdirs, nsearch *
	sizeof (char *));
	umem_free(searchdirs,
	nsearch * sizeof (char *));
	searchdirs = tmp;
	}
	searchdirs[nsearch++] = optarg;
	break;
	case 't':
	max_txg = strtoull(optarg, NULL, 0);
	if (max_txg < TXG_INITIAL) {
	(void) fprintf(stderr, "incorrect txg "
	"specified: %s\n", optarg);
	usage();
	}
	break;
	case 'U':
	spa_config_path = optarg;
	if (spa_config_path[0] != '/') {
	(void) fprintf(stderr,
	"cachefile must be an absolute path "
	"(i.e. start with a slash)\n");
	usage();
	}
	break;
	case 'v':
	verbose++;
	break;
	case 'V':
	flags = ZFS_IMPORT_VERBATIM;
	break;
	case 'x':
	vn_dumpdir = optarg;
	break;
	default:
	usage();
	break;
	}
	}

	if (!dump_opt['e'] && searchdirs != NULL) {
	(void) fprintf(stderr, "-p option requires use of -e\n");
	usage();
	}
	#if defined(_LP64)
	/*
	* ZDB does not typically re-read blocks; therefore limit the ARC
	* to 256 MB, which can be used entirely for metadata.
	*/
	zfs_arc_min = 2ULL << SPA_MAXBLOCKSHIFT;
	zfs_arc_max = 256 * 1024 * 1024;
	#endif

	/*
	* "zdb -c" uses checksum-verifying scrub i/os which are async reads.
	* "zdb -b" uses traversal prefetch which uses async reads.
	* For good performance, let several of them be active at once.
	*/
	zfs_vdev_async_read_max_active = 10;

	/*
	* Disable reference tracking for better performance.
	*/
	reference_tracking_enable = B_FALSE;

	/*
	* Do not fail spa_load when spa_load_verify fails. This is needed
	* to load non-idle pools.
	*/
	spa_load_verify_dryrun = B_TRUE;

	/*
	* ZDB should have ability to read spacemaps.
	*/
	spa_mode_readable_spacemaps = B_TRUE;

	kernel_init(SPA_MODE_READ);
	+ kernel_init_done = B_TRUE;

	if (dump_all)
	verbose = MAX(verbose, 1);

	for (c = 0; c < 256; c++) {
	if (dump_all && strchr("ABeEFkKlLNOPrRSXy", c) == NULL)
	dump_opt[c] = 1;
	if (dump_opt[c])
	dump_opt[c] += verbose;
	}

	libspl_set_assert_ok((dump_opt['A'] == 1) \|\| (dump_opt['A'] > 2));
	zfs_recover = (dump_opt['A'] > 1);

	argc -= optind;
	argv += optind;
	if (argc < 2 && dump_opt['R'])
	usage();

	if (dump_opt['E']) {
	if (argc != 1)
	usage();
	zdb_embedded_block(argv[0]);
	- return (0);
	+ error = 0;
	+ goto fini;
	}

	if (argc < 1) {
	if (!dump_opt['e'] && dump_opt['C']) {
	dump_cachefile(spa_config_path);
	- return (0);
	+ error = 0;
	+ goto fini;
	}
	usage();
	}

	- if (dump_opt['l'])
	- return (dump_label(argv[0]));
	+ if (dump_opt['l']) {
	+ error = dump_label(argv[0]);
	+ goto fini;
	+ }

	if (dump_opt['X'] \|\| dump_opt['F'])
	rewind = ZPOOL_DO_REWIND \|
	(dump_opt['X'] ? ZPOOL_EXTREME_REWIND : 0);

	/* -N implies -d */
	if (dump_opt['N'] && dump_opt['d'] == 0)
	dump_opt['d'] = dump_opt['N'];

	if (nvlist_alloc(&policy, NV_UNIQUE_NAME_TYPE, 0) != 0 \|\|
	nvlist_add_uint64(policy, ZPOOL_LOAD_REQUEST_TXG, max_txg) != 0 \|\|
	nvlist_add_uint32(policy, ZPOOL_LOAD_REWIND_POLICY, rewind) != 0)
	fatal("internal error: %s", strerror(ENOMEM));

	error = 0;
	target = argv[0];

	if (strpbrk(target, "/@") != NULL) {
	size_t targetlen;

	target_pool = strdup(target);
	*strpbrk(target_pool, "/@") = '\0';

	target_is_spa = B_FALSE;
	targetlen = strlen(target);
	if (targetlen && target[targetlen - 1] == '/')
	target[targetlen - 1] = '\0';

	/*
	* See if an objset ID was supplied (-d <pool>/<objset ID>).
	* To disambiguate tank/100, consider the 100 as objsetID
	* if -N was given, otherwise 100 is an objsetID iff
	* tank/100 as a named dataset fails on lookup.
	*/
	objset_str = strchr(target, '/');
	if (objset_str && strlen(objset_str) > 1 &&
	zdb_numeric(objset_str + 1)) {
	char *endptr;
	errno = 0;
	objset_str++;
	objset_id = strtoull(objset_str, &endptr, 0);
	/* dataset 0 is the same as opening the pool */
	if (errno == 0 && endptr != objset_str &&
	objset_id != 0) {
	if (dump_opt['N'])
	dataset_lookup = B_TRUE;
	}
	/* normal dataset name not an objset ID */
	if (endptr == objset_str) {
	objset_id = -1;
	}
	} else if (objset_str && !zdb_numeric(objset_str + 1) &&
	dump_opt['N']) {
	printf("Supply a numeric objset ID with -N\n");
	- exit(1);
	+ error = 1;
	+ goto fini;
	}
	} else {
	target_pool = target;
	}

	if (dump_opt['e']) {
	importargs_t args = { 0 };

	args.paths = nsearch;
	args.path = searchdirs;
	args.can_be_active = B_TRUE;

	libpc_handle_t lpch = {
	.lpc_lib_handle = NULL,
	.lpc_ops = &libzpool_config_ops,
	.lpc_printerr = B_TRUE
	};
	error = zpool_find_config(&lpch, target_pool, &cfg, &args);

	if (error == 0) {

	if (nvlist_add_nvlist(cfg,
	ZPOOL_LOAD_POLICY, policy) != 0) {
	fatal("can't open '%s': %s",
	target, strerror(ENOMEM));
	}

	if (dump_opt['C'] > 1) {
	(void) printf("\nConfiguration for import:\n");
	dump_nvlist(cfg, 8);
	}

	/*
	* Disable the activity check to allow examination of
	* active pools.
	*/
	error = spa_import(target_pool, cfg, NULL,
	flags \| ZFS_IMPORT_SKIP_MMP);
	}
	}

	if (searchdirs != NULL) {
	umem_free(searchdirs, nsearch * sizeof (char *));
	searchdirs = NULL;
	}

	/*
	* We need to make sure to process -O option or call
	* dump_path after the -e option has been processed,
	* which imports the pool to the namespace if it's
	* not in the cachefile.
	*/
	if (dump_opt['O']) {
	if (argc != 2)
	usage();
	dump_opt['v'] = verbose + 3;
	- return (dump_path(argv[0], argv[1], NULL));
	+ error = dump_path(argv[0], argv[1], NULL);
	+ goto fini;
	}

	if (dump_opt['r']) {
	target_is_spa = B_FALSE;
	if (argc != 3)
	usage();
	dump_opt['v'] = verbose;
	error = dump_path(argv[0], argv[1], &object);
	if (error != 0)
	fatal("internal error: %s", strerror(error));
	}

	/*
	* import_checkpointed_state makes the assumption that the
	* target pool that we pass it is already part of the spa
	* namespace. Because of that we need to make sure to call
	* it always after the -e option has been processed, which
	* imports the pool to the namespace if it's not in the
	* cachefile.
	*/
	char *checkpoint_pool = NULL;
	char *checkpoint_target = NULL;
	if (dump_opt['k']) {
	checkpoint_pool = import_checkpointed_state(target, cfg,
	&checkpoint_target);

	if (checkpoint_target != NULL)
	target = checkpoint_target;
	}

	if (cfg != NULL) {
	nvlist_free(cfg);
	cfg = NULL;
	}

	if (target_pool != target)
	free(target_pool);

	if (error == 0) {
	if (dump_opt['k'] && (target_is_spa \|\| dump_opt['R'])) {
	ASSERT(checkpoint_pool != NULL);
	ASSERT(checkpoint_target == NULL);

	error = spa_open(checkpoint_pool, &spa, FTAG);
	if (error != 0) {
	fatal("Tried to open pool \"%s\" but "
	"spa_open() failed with error %d\n",
	checkpoint_pool, error);
	}

	} else if (target_is_spa \|\| dump_opt['R'] \|\| dump_opt['B'] \|\|
	objset_id == 0) {
	zdb_set_skip_mmp(target);
	error = spa_open_rewind(target, &spa, FTAG, policy,
	NULL);
	if (error) {
	/*
	* If we're missing the log device then
	* try opening the pool after clearing the
	* log state.
	*/
	mutex_enter(&spa_namespace_lock);
	if ((spa = spa_lookup(target)) != NULL &&
	spa->spa_log_state == SPA_LOG_MISSING) {
	spa->spa_log_state = SPA_LOG_CLEAR;
	error = 0;
	}
	mutex_exit(&spa_namespace_lock);

	if (!error) {
	error = spa_open_rewind(target, &spa,
	FTAG, policy, NULL);
	}
	}
	} else if (strpbrk(target, "#") != NULL) {
	dsl_pool_t *dp;
	error = dsl_pool_hold(target, FTAG, &dp);
	if (error != 0) {
	fatal("can't dump '%s': %s", target,
	strerror(error));
	}
	error = dump_bookmark(dp, target, B_TRUE, verbose > 1);
	dsl_pool_rele(dp, FTAG);
	if (error != 0) {
	fatal("can't dump '%s': %s", target,
	strerror(error));
	}
	- return (error);
	+ goto fini;
	} else {
	target_pool = strdup(target);
	if (strpbrk(target, "/@") != NULL)
	*strpbrk(target_pool, "/@") = '\0';

	zdb_set_skip_mmp(target);
	/*
	* If -N was supplied, the user has indicated that
	* zdb -d <pool>/<objsetID> is in effect. Otherwise
	* we first assume that the dataset string is the
	* dataset name. If dmu_objset_hold fails with the
	* dataset string, and we have an objset_id, retry the
	* lookup with the objsetID.
	*/
	boolean_t retry = B_TRUE;
	retry_lookup:
	if (dataset_lookup == B_TRUE) {
	/*
	* Use the supplied id to get the name
	* for open_objset.
	*/
	error = spa_open(target_pool, &spa, FTAG);
	if (error == 0) {
	error = name_from_objset_id(spa,
	objset_id, dsname);
	spa_close(spa, FTAG);
	if (error == 0)
	target = dsname;
	}
	}
	if (error == 0) {
	if (objset_id > 0 && retry) {
	int err = dmu_objset_hold(target, FTAG,
	&os);
	if (err) {
	dataset_lookup = B_TRUE;
	retry = B_FALSE;
	goto retry_lookup;
	} else {
	dmu_objset_rele(os, FTAG);
	}
	}
	error = open_objset(target, FTAG, &os);
	}
	if (error == 0)
	spa = dmu_objset_spa(os);
	free(target_pool);
	}
	}
	nvlist_free(policy);

	if (error)
	fatal("can't open '%s': %s", target, strerror(error));

	/*
	* Set the pool failure mode to panic in order to prevent the pool
	* from suspending. A suspended I/O will have no way to resume and
	* can prevent the zdb(8) command from terminating as expected.
	*/
	if (spa != NULL)
	spa->spa_failmode = ZIO_FAILURE_MODE_PANIC;

	argv++;
	argc--;
	if (dump_opt['r']) {
	error = zdb_copy_object(os, object, argv[1]);
	} else if (!dump_opt['R']) {
	flagbits['d'] = ZOR_FLAG_DIRECTORY;
	flagbits['f'] = ZOR_FLAG_PLAIN_FILE;
	flagbits['m'] = ZOR_FLAG_SPACE_MAP;
	flagbits['z'] = ZOR_FLAG_ZAP;
	flagbits['A'] = ZOR_FLAG_ALL_TYPES;

	if (argc > 0 && dump_opt['d']) {
	zopt_object_args = argc;
	zopt_object_ranges = calloc(zopt_object_args,
	sizeof (zopt_object_range_t));
	for (unsigned i = 0; i < zopt_object_args; i++) {
	int err;
	const char *msg = NULL;

	err = parse_object_range(argv[i],
	&zopt_object_ranges[i], &msg);
	if (err != 0)
	fatal("Bad object or range: '%s': %s\n",
	argv[i], msg ?: "");
	}
	} else if (argc > 0 && dump_opt['m']) {
	zopt_metaslab_args = argc;
	zopt_metaslab = calloc(zopt_metaslab_args,
	sizeof (uint64_t));
	for (unsigned i = 0; i < zopt_metaslab_args; i++) {
	errno = 0;
	zopt_metaslab[i] = strtoull(argv[i], NULL, 0);
	if (zopt_metaslab[i] == 0 && errno != 0)
	fatal("bad number %s: %s", argv[i],
	strerror(errno));
	}
	}
	if (dump_opt['B']) {
	dump_backup(target, objset_id,
	argc > 0 ? argv[0] : NULL);
	} else if (os != NULL) {
	dump_objset(os);
	} else if (zopt_object_args > 0 && !dump_opt['m']) {
	dump_objset(spa->spa_meta_objset);
	} else {
	dump_zpool(spa);
	}
	} else {
	flagbits['b'] = ZDB_FLAG_PRINT_BLKPTR;
	flagbits['c'] = ZDB_FLAG_CHECKSUM;
	flagbits['d'] = ZDB_FLAG_DECOMPRESS;
	flagbits['e'] = ZDB_FLAG_BSWAP;
	flagbits['g'] = ZDB_FLAG_GBH;
	flagbits['i'] = ZDB_FLAG_INDIRECT;
	flagbits['r'] = ZDB_FLAG_RAW;
	flagbits['v'] = ZDB_FLAG_VERBOSE;

	for (int i = 0; i < argc; i++)
	zdb_read_block(argv[i], spa);
	}

	if (dump_opt['k']) {
	free(checkpoint_pool);
	if (!target_is_spa)
	free(checkpoint_target);
	}

	+fini:
	if (os != NULL) {
	close_objset(os, FTAG);
	- } else {
	+ } else if (spa != NULL) {
	spa_close(spa, FTAG);
	}

	fuid_table_destroy();

	dump_debug_buffer();

	- kernel_fini();
	+ if (kernel_init_done)
	+ kernel_fini();

	return (error);
	}
	diff --git a/sys/contrib/openzfs/cmd/zed/agents/zfs_mod.c b/sys/contrib/openzfs/cmd/zed/agents/zfs_mod.c
	index 69163b80bd5a..d0372608c723 100644
	--- a/sys/contrib/openzfs/cmd/zed/agents/zfs_mod.c
	+++ b/sys/contrib/openzfs/cmd/zed/agents/zfs_mod.c
	@@ -1,1370 +1,1370 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012 by Delphix. All rights reserved.
	* Copyright 2014 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2016, 2017, Intel Corporation.
	* Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
	* Copyright (c) 2023, Klara Inc.
	*/

	/*
	* ZFS syseventd module.
	*
	* file origin: openzfs/usr/src/cmd/syseventd/modules/zfs_mod/zfs_mod.c
	*
	* The purpose of this module is to identify when devices are added to the
	* system, and appropriately online or replace the affected vdevs.
	*
	* When a device is added to the system:
	*
	* 1. Search for any vdevs whose devid matches that of the newly added
	* device.
	*
	* 2. If no vdevs are found, then search for any vdevs whose udev path
	* matches that of the new device.
	*
	* 3. If no vdevs match by either method, then ignore the event.
	*
	* 4. Attempt to online the device with a flag to indicate that it should
	* be unspared when resilvering completes. If this succeeds, then the
	* same device was inserted and we should continue normally.
	*
	* 5. If the pool does not have the 'autoreplace' property set, attempt to
	* online the device again without the unspare flag, which will
	* generate a FMA fault.
	*
	* 6. If the pool has the 'autoreplace' property set, and the matching vdev
	* is a whole disk, then label the new disk and attempt a 'zpool
	* replace'.
	*
	* The module responds to EC_DEV_ADD events. The special ESC_ZFS_VDEV_CHECK
	* event indicates that a device failed to open during pool load, but the
	* autoreplace property was set. In this case, we deferred the associated
	* FMA fault until our module had a chance to process the autoreplace logic.
	* If the device could not be replaced, then the second online attempt will
	* trigger the FMA fault that we skipped earlier.
	*
	* On Linux udev provides a disk insert for both the disk and the partition.
	*/

	#include <ctype.h>
	#include <fcntl.h>
	#include <libnvpair.h>
	#include <libzfs.h>
	#include <libzutil.h>
	#include <limits.h>
	#include <stddef.h>
	#include <stdlib.h>
	#include <string.h>
	#include <syslog.h>
	#include <sys/list.h>
	#include <sys/sunddi.h>
	#include <sys/sysevent/eventdefs.h>
	#include <sys/sysevent/dev.h>
	#include <thread_pool.h>
	#include <pthread.h>
	#include <unistd.h>
	#include <errno.h>
	#include "zfs_agents.h"
	#include "../zed_log.h"

	#define DEV_BYID_PATH "/dev/disk/by-id/"
	#define DEV_BYPATH_PATH "/dev/disk/by-path/"
	#define DEV_BYVDEV_PATH "/dev/disk/by-vdev/"

	typedef void (zfs_process_func_t)(zpool_handle_t , nvlist_t *, boolean_t);

	libzfs_handle_t *g_zfshdl;
	list_t g_pool_list; /* list of unavailable pools at initialization */
	list_t g_device_list; /* list of disks with asynchronous label request */
	tpool_t *g_tpool;
	boolean_t g_enumeration_done;
	pthread_t g_zfs_tid; /* zfs_enum_pools() thread */

	typedef struct unavailpool {
	zpool_handle_t *uap_zhp;
	list_node_t uap_node;
	} unavailpool_t;

	typedef struct pendingdev {
	char pd_physpath[128];
	list_node_t pd_node;
	} pendingdev_t;

	static int
	zfs_toplevel_state(zpool_handle_t *zhp)
	{
	nvlist_t *nvroot;
	vdev_stat_t *vs;
	unsigned int c;

	verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
	ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
	verify(nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_VDEV_STATS,
	(uint64_t **)&vs, &c) == 0);
	return (vs->vs_state);
	}

	static int
	zfs_unavail_pool(zpool_handle_t zhp, void data)
	{
	zed_log_msg(LOG_INFO, "zfs_unavail_pool: examining '%s' (state %d)",
	zpool_get_name(zhp), (int)zfs_toplevel_state(zhp));

	if (zfs_toplevel_state(zhp) < VDEV_STATE_DEGRADED) {
	unavailpool_t *uap;
	uap = malloc(sizeof (unavailpool_t));
	if (uap == NULL) {
	perror("malloc");
	exit(EXIT_FAILURE);
	}

	uap->uap_zhp = zhp;
	list_insert_tail((list_t *)data, uap);
	} else {
	zpool_close(zhp);
	}
	return (0);
	}

	/*
	* Write an array of strings to the zed log
	*/
	static void lines_to_zed_log_msg(char **lines, int lines_cnt)
	{
	int i;
	for (i = 0; i < lines_cnt; i++) {
	zed_log_msg(LOG_INFO, "%s", lines[i]);
	}
	}

	/*
	* Two stage replace on Linux
	* since we get disk notifications
	* we can wait for partitioned disk slice to show up!
	*
	* First stage tags the disk, initiates async partitioning, and returns
	* Second stage finds the tag and proceeds to ZFS labeling/replace
	*
	* disk-add --> label-disk + tag-disk --> partition-add --> zpool_vdev_attach
	*
	* 1. physical match with no fs, no partition
	* tag it top, partition disk
	*
	* 2. physical match again, see partition and tag
	*
	*/

	/*
	* The device associated with the given vdev (either by devid or physical path)
	* has been added to the system. If 'isdisk' is set, then we only attempt a
	* replacement if it's a whole disk. This also implies that we should label the
	* disk first.
	*
	* First, we attempt to online the device (making sure to undo any spare
	* operation when finished). If this succeeds, then we're done. If it fails,
	* and the new state is VDEV_CANT_OPEN, it indicates that the device was opened,
	* but that the label was not what we expected. If the 'autoreplace' property
	* is enabled, then we relabel the disk (if specified), and attempt a 'zpool
	* replace'. If the online is successful, but the new state is something else
	* (REMOVED or FAULTED), it indicates that we're out of sync or in some sort of
	* race, and we should avoid attempting to relabel the disk.
	*
	* Also can arrive here from a ESC_ZFS_VDEV_CHECK event
	*/
	static void
	zfs_process_add(zpool_handle_t zhp, nvlist_t vdev, boolean_t labeled)
	{
	const char *path;
	vdev_state_t newstate;
	nvlist_t nvroot, newvd;
	pendingdev_t *device;
	uint64_t wholedisk = 0ULL;
	uint64_t offline = 0ULL, faulted = 0ULL;
	uint64_t guid = 0ULL;
	uint64_t is_spare = 0;
	const char physpath = NULL, new_devid = NULL, *enc_sysfs_path = NULL;
	char rawpath[PATH_MAX], fullpath[PATH_MAX];
	char pathbuf[PATH_MAX];
	int ret;
	int online_flag = ZFS_ONLINE_CHECKREMOVE \| ZFS_ONLINE_UNSPARE;
	boolean_t is_sd = B_FALSE;
	boolean_t is_mpath_wholedisk = B_FALSE;
	uint_t c;
	vdev_stat_t *vs;
	char **lines = NULL;
	int lines_cnt = 0;

	/*
	* Get the persistent path, typically under the '/dev/disk/by-id' or
	* '/dev/disk/by-vdev' directories. Note that this path can change
	* when a vdev is replaced with a new disk.
	*/
	if (nvlist_lookup_string(vdev, ZPOOL_CONFIG_PATH, &path) != 0)
	return;

	/* Skip healthy disks */
	verify(nvlist_lookup_uint64_array(vdev, ZPOOL_CONFIG_VDEV_STATS,
	(uint64_t **)&vs, &c) == 0);
	if (vs->vs_state == VDEV_STATE_HEALTHY) {
	zed_log_msg(LOG_INFO, "%s: %s is already healthy, skip it.",
	__func__, path);
	return;
	}

	(void) nvlist_lookup_string(vdev, ZPOOL_CONFIG_PHYS_PATH, &physpath);

	update_vdev_config_dev_sysfs_path(vdev, path,
	ZPOOL_CONFIG_VDEV_ENC_SYSFS_PATH);
	(void) nvlist_lookup_string(vdev, ZPOOL_CONFIG_VDEV_ENC_SYSFS_PATH,
	&enc_sysfs_path);

	(void) nvlist_lookup_uint64(vdev, ZPOOL_CONFIG_WHOLE_DISK, &wholedisk);
	(void) nvlist_lookup_uint64(vdev, ZPOOL_CONFIG_OFFLINE, &offline);
	(void) nvlist_lookup_uint64(vdev, ZPOOL_CONFIG_FAULTED, &faulted);

	(void) nvlist_lookup_uint64(vdev, ZPOOL_CONFIG_GUID, &guid);
	(void) nvlist_lookup_uint64(vdev, ZPOOL_CONFIG_IS_SPARE, &is_spare);

	/*
	* Special case:
	*
	* We've seen times where a disk won't have a ZPOOL_CONFIG_PHYS_PATH
	* entry in their config. For example, on this force-faulted disk:
	*
	* children[0]:
	* type: 'disk'
	* id: 0
	* guid: 14309659774640089719
	* path: '/dev/disk/by-vdev/L28'
	* whole_disk: 0
	* DTL: 654
	* create_txg: 4
	* com.delphix:vdev_zap_leaf: 1161
	* faulted: 1
	* aux_state: 'external'
	* children[1]:
	* type: 'disk'
	* id: 1
	* guid: 16002508084177980912
	* path: '/dev/disk/by-vdev/L29'
	* devid: 'dm-uuid-mpath-35000c500a61d68a3'
	* phys_path: 'L29'
	* vdev_enc_sysfs_path: '/sys/class/enclosure/0:0:1:0/SLOT 30 32'
	* whole_disk: 0
	* DTL: 1028
	* create_txg: 4
	* com.delphix:vdev_zap_leaf: 131
	*
	* If the disk's path is a /dev/disk/by-vdev/ path, then we can infer
	* the ZPOOL_CONFIG_PHYS_PATH from the by-vdev disk name.
	*/
	if (physpath == NULL && path != NULL) {
	/* If path begins with "/dev/disk/by-vdev/" ... */
	if (strncmp(path, DEV_BYVDEV_PATH,
	strlen(DEV_BYVDEV_PATH)) == 0) {
	/* Set physpath to the char after "/dev/disk/by-vdev" */
	physpath = &path[strlen(DEV_BYVDEV_PATH)];
	}
	}

	/*
	* We don't want to autoreplace offlined disks. However, we do want to
	* replace force-faulted disks (`zpool offline -f`). Force-faulted
	* disks have both offline=1 and faulted=1 in the nvlist.
	*/
	if (offline && !faulted) {
	zed_log_msg(LOG_INFO, "%s: %s is offline, skip autoreplace",
	__func__, path);
	return;
	}

	is_mpath_wholedisk = is_mpath_whole_disk(path);
	zed_log_msg(LOG_INFO, "zfs_process_add: pool '%s' vdev '%s', phys '%s'"
	" %s blank disk, %s mpath blank disk, %s labeled, enc sysfs '%s', "
	"(guid %llu)",
	zpool_get_name(zhp), path,
	physpath ? physpath : "NULL",
	wholedisk ? "is" : "not",
	is_mpath_wholedisk? "is" : "not",
	labeled ? "is" : "not",
	enc_sysfs_path,
	(long long unsigned int)guid);

	/*
	* The VDEV guid is preferred for identification (gets passed in path)
	*/
	if (guid != 0) {
	(void) snprintf(fullpath, sizeof (fullpath), "%llu",
	(long long unsigned int)guid);
	} else {
	/*
	* otherwise use path sans partition suffix for whole disks
	*/
	(void) strlcpy(fullpath, path, sizeof (fullpath));
	if (wholedisk) {
	char *spath = zfs_strip_partition(fullpath);
	if (!spath) {
	zed_log_msg(LOG_INFO, "%s: Can't alloc",
	__func__);
	return;
	}

	(void) strlcpy(fullpath, spath, sizeof (fullpath));
	free(spath);
	}
	}

	if (is_spare)
	online_flag \|= ZFS_ONLINE_SPARE;

	/*
	* Attempt to online the device.
	*/
	if (zpool_vdev_online(zhp, fullpath, online_flag, &newstate) == 0 &&
	(newstate == VDEV_STATE_HEALTHY \|\|
	newstate == VDEV_STATE_DEGRADED)) {
	zed_log_msg(LOG_INFO,
	" zpool_vdev_online: vdev '%s' ('%s') is "
	"%s", fullpath, physpath, (newstate == VDEV_STATE_HEALTHY) ?
	"HEALTHY" : "DEGRADED");
	return;
	}

	/*
	* vdev_id alias rule for using scsi_debug devices (FMA automated
	* testing)
	*/
	if (physpath != NULL && strcmp("scsidebug", physpath) == 0)
	is_sd = B_TRUE;

	/*
	* If the pool doesn't have the autoreplace property set, then use
	* vdev online to trigger a FMA fault by posting an ereport.
	*/
	if (!zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOREPLACE, NULL) \|\|
	!(wholedisk \|\| is_mpath_wholedisk) \|\| (physpath == NULL)) {
	(void) zpool_vdev_online(zhp, fullpath, ZFS_ONLINE_FORCEFAULT,
	&newstate);
	zed_log_msg(LOG_INFO, "Pool's autoreplace is not enabled or "
	"not a blank disk for '%s' ('%s')", fullpath,
	physpath);
	return;
	}

	/*
	* Convert physical path into its current device node. Rawpath
	* needs to be /dev/disk/by-vdev for a scsi_debug device since
	* /dev/disk/by-path will not be present.
	*/
	(void) snprintf(rawpath, sizeof (rawpath), "%s%s",
	is_sd ? DEV_BYVDEV_PATH : DEV_BYPATH_PATH, physpath);

	if (realpath(rawpath, pathbuf) == NULL && !is_mpath_wholedisk) {
	zed_log_msg(LOG_INFO, " realpath: %s failed (%s)",
	rawpath, strerror(errno));

	int err = zpool_vdev_online(zhp, fullpath,
	ZFS_ONLINE_FORCEFAULT, &newstate);

	zed_log_msg(LOG_INFO, " zpool_vdev_online: %s FORCEFAULT (%s) "
	"err %d, new state %d",
	fullpath, libzfs_error_description(g_zfshdl), err,
	err ? (int)newstate : 0);
	return;
	}

	/* Only autoreplace bad disks */
	if ((vs->vs_state != VDEV_STATE_DEGRADED) &&
	(vs->vs_state != VDEV_STATE_FAULTED) &&
	(vs->vs_state != VDEV_STATE_REMOVED) &&
	(vs->vs_state != VDEV_STATE_CANT_OPEN)) {
	zed_log_msg(LOG_INFO, " not autoreplacing since disk isn't in "
	"a bad state (currently %llu)", vs->vs_state);
	return;
	}

	nvlist_lookup_string(vdev, "new_devid", &new_devid);

	if (is_mpath_wholedisk) {
	/* Don't label device mapper or multipath disks. */
	zed_log_msg(LOG_INFO,
	" it's a multipath wholedisk, don't label");
	if (zpool_prepare_disk(zhp, vdev, "autoreplace", &lines,
	&lines_cnt) != 0) {
	zed_log_msg(LOG_INFO,
	" zpool_prepare_disk: could not "
	"prepare '%s' (%s)", fullpath,
	libzfs_error_description(g_zfshdl));
	if (lines_cnt > 0) {
	zed_log_msg(LOG_INFO,
	" zfs_prepare_disk output:");
	lines_to_zed_log_msg(lines, lines_cnt);
	}
	libzfs_free_str_array(lines, lines_cnt);
	return;
	}
	} else if (!labeled) {
	/*
	* we're auto-replacing a raw disk, so label it first
	*/
	char *leafname;

	/*
	* If this is a request to label a whole disk, then attempt to
	* write out the label. Before we can label the disk, we need
	* to map the physical string that was matched on to the under
	* lying device node.
	*
	* If any part of this process fails, then do a force online
	* to trigger a ZFS fault for the device (and any hot spare
	* replacement).
	*/
	leafname = strrchr(pathbuf, '/') + 1;

	/*
	* If this is a request to label a whole disk, then attempt to
	* write out the label.
	*/
	if (zpool_prepare_and_label_disk(g_zfshdl, zhp, leafname,
	vdev, "autoreplace", &lines, &lines_cnt) != 0) {
	zed_log_msg(LOG_WARNING,
	" zpool_prepare_and_label_disk: could not "
	"label '%s' (%s)", leafname,
	libzfs_error_description(g_zfshdl));
	if (lines_cnt > 0) {
	zed_log_msg(LOG_INFO,
	" zfs_prepare_disk output:");
	lines_to_zed_log_msg(lines, lines_cnt);
	}
	libzfs_free_str_array(lines, lines_cnt);

	(void) zpool_vdev_online(zhp, fullpath,
	ZFS_ONLINE_FORCEFAULT, &newstate);
	return;
	}

	/*
	* The disk labeling is asynchronous on Linux. Just record
	* this label request and return as there will be another
	* disk add event for the partition after the labeling is
	* completed.
	*/
	device = malloc(sizeof (pendingdev_t));
	if (device == NULL) {
	perror("malloc");
	exit(EXIT_FAILURE);
	}

	(void) strlcpy(device->pd_physpath, physpath,
	sizeof (device->pd_physpath));
	list_insert_tail(&g_device_list, device);

	zed_log_msg(LOG_NOTICE, " zpool_label_disk: async '%s' (%llu)",
	leafname, (u_longlong_t)guid);

	return; /* resumes at EC_DEV_ADD.ESC_DISK for partition */

	} else /* labeled */ {
	boolean_t found = B_FALSE;
	/*
	* match up with request above to label the disk
	*/
	for (device = list_head(&g_device_list); device != NULL;
	device = list_next(&g_device_list, device)) {
	if (strcmp(physpath, device->pd_physpath) == 0) {
	list_remove(&g_device_list, device);
	free(device);
	found = B_TRUE;
	break;
	}
	zed_log_msg(LOG_INFO, "zpool_label_disk: %s != %s",
	physpath, device->pd_physpath);
	}
	if (!found) {
	/* unexpected partition slice encountered */
	zed_log_msg(LOG_WARNING, "labeled disk %s was "
	"unexpected here", fullpath);
	(void) zpool_vdev_online(zhp, fullpath,
	ZFS_ONLINE_FORCEFAULT, &newstate);
	return;
	}

	zed_log_msg(LOG_INFO, " zpool_label_disk: resume '%s' (%llu)",
	physpath, (u_longlong_t)guid);

	/*
	* Paths that begin with '/dev/disk/by-id/' will change and so
	* they must be updated before calling zpool_vdev_attach().
	*/
	if (strncmp(path, DEV_BYID_PATH, strlen(DEV_BYID_PATH)) == 0) {
	(void) snprintf(pathbuf, sizeof (pathbuf), "%s%s",
	DEV_BYID_PATH, new_devid);
	zed_log_msg(LOG_INFO, " zpool_label_disk: path '%s' "
	"replaced by '%s'", path, pathbuf);
	path = pathbuf;
	}
	}

	libzfs_free_str_array(lines, lines_cnt);

	/*
	* Construct the root vdev to pass to zpool_vdev_attach(). While adding
	* the entire vdev structure is harmless, we construct a reduced set of
	* path/physpath/wholedisk to keep it simple.
	*/
	if (nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) != 0) {
	zed_log_msg(LOG_WARNING, "zfs_mod: nvlist_alloc out of memory");
	return;
	}
	if (nvlist_alloc(&newvd, NV_UNIQUE_NAME, 0) != 0) {
	zed_log_msg(LOG_WARNING, "zfs_mod: nvlist_alloc out of memory");
	nvlist_free(nvroot);
	return;
	}

	if (nvlist_add_string(newvd, ZPOOL_CONFIG_TYPE, VDEV_TYPE_DISK) != 0 \|\|
	nvlist_add_string(newvd, ZPOOL_CONFIG_PATH, path) != 0 \|\|
	nvlist_add_string(newvd, ZPOOL_CONFIG_DEVID, new_devid) != 0 \|\|
	(physpath != NULL && nvlist_add_string(newvd,
	ZPOOL_CONFIG_PHYS_PATH, physpath) != 0) \|\|
	(enc_sysfs_path != NULL && nvlist_add_string(newvd,
	ZPOOL_CONFIG_VDEV_ENC_SYSFS_PATH, enc_sysfs_path) != 0) \|\|
	nvlist_add_uint64(newvd, ZPOOL_CONFIG_WHOLE_DISK, wholedisk) != 0 \|\|
	nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT) != 0 \|\|
	nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
	(const nvlist_t **)&newvd, 1) != 0) {
	zed_log_msg(LOG_WARNING, "zfs_mod: unable to add nvlist pairs");
	nvlist_free(newvd);
	nvlist_free(nvroot);
	return;
	}

	nvlist_free(newvd);

	/*
	* Wait for udev to verify the links exist, then auto-replace
	* the leaf disk at same physical location.
	*/
	if (zpool_label_disk_wait(path, DISK_LABEL_WAIT) != 0) {
	zed_log_msg(LOG_WARNING, "zfs_mod: pool '%s', after labeling "
	"replacement disk, the expected disk partition link '%s' "
	"is missing after waiting %u ms",
	zpool_get_name(zhp), path, DISK_LABEL_WAIT);
	nvlist_free(nvroot);
	return;
	}

	/*
	* Prefer sequential resilvering when supported (mirrors and dRAID),
	* otherwise fallback to a traditional healing resilver.
	*/
	ret = zpool_vdev_attach(zhp, fullpath, path, nvroot, B_TRUE, B_TRUE);
	if (ret != 0) {
	ret = zpool_vdev_attach(zhp, fullpath, path, nvroot,
	B_TRUE, B_FALSE);
	}

	zed_log_msg(LOG_WARNING, " zpool_vdev_replace: %s with %s (%s)",
	fullpath, path, (ret == 0) ? "no errors" :
	libzfs_error_description(g_zfshdl));

	nvlist_free(nvroot);
	}

	/*
	* Utility functions to find a vdev matching given criteria.
	*/
	typedef struct dev_data {
	const char *dd_compare;
	const char *dd_prop;
	zfs_process_func_t dd_func;
	boolean_t dd_found;
	boolean_t dd_islabeled;
	uint64_t dd_pool_guid;
	uint64_t dd_vdev_guid;
	uint64_t dd_new_vdev_guid;
	const char *dd_new_devid;
	uint64_t dd_num_spares;
	} dev_data_t;

	static void
	zfs_iter_vdev(zpool_handle_t zhp, nvlist_t nvl, void *data)
	{
	dev_data_t *dp = data;
	const char *path = NULL;
	uint_t c, children;
	nvlist_t **child;
	uint64_t guid = 0;
	uint64_t isspare = 0;

	/*
	* First iterate over any children.
	*/
	if (nvlist_lookup_nvlist_array(nvl, ZPOOL_CONFIG_CHILDREN,
	&child, &children) == 0) {
	for (c = 0; c < children; c++)
	zfs_iter_vdev(zhp, child[c], data);
	}

	/*
	* Iterate over any spares and cache devices
	*/
	if (nvlist_lookup_nvlist_array(nvl, ZPOOL_CONFIG_SPARES,
	&child, &children) == 0) {
	for (c = 0; c < children; c++)
	zfs_iter_vdev(zhp, child[c], data);
	}
	if (nvlist_lookup_nvlist_array(nvl, ZPOOL_CONFIG_L2CACHE,
	&child, &children) == 0) {
	for (c = 0; c < children; c++)
	zfs_iter_vdev(zhp, child[c], data);
	}

	/* once a vdev was matched and processed there is nothing left to do */
	if (dp->dd_found && dp->dd_num_spares == 0)
	return;
	(void) nvlist_lookup_uint64(nvl, ZPOOL_CONFIG_GUID, &guid);

	/*
	* Match by GUID if available otherwise fallback to devid or physical
	*/
	if (dp->dd_vdev_guid != 0) {
	if (guid != dp->dd_vdev_guid)
	return;
	zed_log_msg(LOG_INFO, " zfs_iter_vdev: matched on %llu", guid);
	dp->dd_found = B_TRUE;

	} else if (dp->dd_compare != NULL) {
	/*
	* NOTE: On Linux there is an event for partition, so unlike
	* illumos, substring matching is not required to accommodate
	* the partition suffix. An exact match will be present in
	* the dp->dd_compare value.
	* If the attached disk already contains a vdev GUID, it means
	* the disk is not clean. In such a scenario, the physical path
	* would be a match that makes the disk faulted when trying to
	* online it. So, we would only want to proceed if either GUID
	* matches with the last attached disk or the disk is in clean
	* state.
	*/
	if (nvlist_lookup_string(nvl, dp->dd_prop, &path) != 0 \|\|
	strcmp(dp->dd_compare, path) != 0) {
	return;
	}
	if (dp->dd_new_vdev_guid != 0 && dp->dd_new_vdev_guid != guid) {
	zed_log_msg(LOG_INFO, " %s: no match (GUID:%llu"
	" != vdev GUID:%llu)", __func__,
	dp->dd_new_vdev_guid, guid);
	return;
	}

	zed_log_msg(LOG_INFO, " zfs_iter_vdev: matched %s on %s",
	dp->dd_prop, path);
	dp->dd_found = B_TRUE;

	/* pass the new devid for use by auto-replacing code */
	if (dp->dd_new_devid != NULL) {
	(void) nvlist_add_string(nvl, "new_devid",
	dp->dd_new_devid);
	}
	}

	if (dp->dd_found == B_TRUE && nvlist_lookup_uint64(nvl,
	ZPOOL_CONFIG_IS_SPARE, &isspare) == 0 && isspare)
	dp->dd_num_spares++;

	(dp->dd_func)(zhp, nvl, dp->dd_islabeled);
	}

	static void
	zfs_enable_ds(void *arg)
	{
	unavailpool_t pool = (unavailpool_t )arg;

	- (void) zpool_enable_datasets(pool->uap_zhp, NULL, 0);
	+ (void) zpool_enable_datasets(pool->uap_zhp, NULL, 0, 512);
	zpool_close(pool->uap_zhp);
	free(pool);
	}

	static int
	zfs_iter_pool(zpool_handle_t zhp, void data)
	{
	nvlist_t config, nvl;
	dev_data_t *dp = data;
	uint64_t pool_guid;
	unavailpool_t *pool;

	zed_log_msg(LOG_INFO, "zfs_iter_pool: evaluating vdevs on %s (by %s)",
	zpool_get_name(zhp), dp->dd_vdev_guid ? "GUID" : dp->dd_prop);

	/*
	* For each vdev in this pool, look for a match to apply dd_func
	*/
	if ((config = zpool_get_config(zhp, NULL)) != NULL) {
	if (dp->dd_pool_guid == 0 \|\|
	(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
	&pool_guid) == 0 && pool_guid == dp->dd_pool_guid)) {
	(void) nvlist_lookup_nvlist(config,
	ZPOOL_CONFIG_VDEV_TREE, &nvl);
	zfs_iter_vdev(zhp, nvl, data);
	}
	} else {
	zed_log_msg(LOG_INFO, "%s: no config\n", __func__);
	}

	/*
	* if this pool was originally unavailable,
	* then enable its datasets asynchronously
	*/
	if (g_enumeration_done) {
	for (pool = list_head(&g_pool_list); pool != NULL;
	pool = list_next(&g_pool_list, pool)) {

	if (strcmp(zpool_get_name(zhp),
	zpool_get_name(pool->uap_zhp)))
	continue;
	if (zfs_toplevel_state(zhp) >= VDEV_STATE_DEGRADED) {
	list_remove(&g_pool_list, pool);
	(void) tpool_dispatch(g_tpool, zfs_enable_ds,
	pool);
	break;
	}
	}
	}

	zpool_close(zhp);

	/* cease iteration after a match */
	return (dp->dd_found && dp->dd_num_spares == 0);
	}

	/*
	* Given a physical device location, iterate over all
	* (pool, vdev) pairs which correspond to that location.
	*/
	static boolean_t
	devphys_iter(const char physical, const char devid, zfs_process_func_t func,
	boolean_t is_slice, uint64_t new_vdev_guid)
	{
	dev_data_t data = { 0 };

	data.dd_compare = physical;
	data.dd_func = func;
	data.dd_prop = ZPOOL_CONFIG_PHYS_PATH;
	data.dd_found = B_FALSE;
	data.dd_islabeled = is_slice;
	data.dd_new_devid = devid; /* used by auto replace code */
	data.dd_new_vdev_guid = new_vdev_guid;

	(void) zpool_iter(g_zfshdl, zfs_iter_pool, &data);

	return (data.dd_found);
	}

	/*
	* Given a device identifier, find any vdevs with a matching by-vdev
	* path. Normally we shouldn't need this as the comparison would be
	* made earlier in the devphys_iter(). For example, if we were replacing
	* /dev/disk/by-vdev/L28, normally devphys_iter() would match the
	* ZPOOL_CONFIG_PHYS_PATH of "L28" from the old disk config to "L28"
	* of the new disk config. However, we've seen cases where
	* ZPOOL_CONFIG_PHYS_PATH was not in the config for the old disk. Here's
	* an example of a real 2-disk mirror pool where one disk was force
	* faulted:
	*
	* com.delphix:vdev_zap_top: 129
	* children[0]:
	* type: 'disk'
	* id: 0
	* guid: 14309659774640089719
	* path: '/dev/disk/by-vdev/L28'
	* whole_disk: 0
	* DTL: 654
	* create_txg: 4
	* com.delphix:vdev_zap_leaf: 1161
	* faulted: 1
	* aux_state: 'external'
	* children[1]:
	* type: 'disk'
	* id: 1
	* guid: 16002508084177980912
	* path: '/dev/disk/by-vdev/L29'
	* devid: 'dm-uuid-mpath-35000c500a61d68a3'
	* phys_path: 'L29'
	* vdev_enc_sysfs_path: '/sys/class/enclosure/0:0:1:0/SLOT 30 32'
	* whole_disk: 0
	* DTL: 1028
	* create_txg: 4
	* com.delphix:vdev_zap_leaf: 131
	*
	* So in the case above, the only thing we could compare is the path.
	*
	* We can do this because we assume by-vdev paths are authoritative as physical
	* paths. We could not assume this for normal paths like /dev/sda since the
	* physical location /dev/sda points to could change over time.
	*/
	static boolean_t
	by_vdev_path_iter(const char by_vdev_path, const char devid,
	zfs_process_func_t func, boolean_t is_slice)
	{
	dev_data_t data = { 0 };

	data.dd_compare = by_vdev_path;
	data.dd_func = func;
	data.dd_prop = ZPOOL_CONFIG_PATH;
	data.dd_found = B_FALSE;
	data.dd_islabeled = is_slice;
	data.dd_new_devid = devid;

	if (strncmp(by_vdev_path, DEV_BYVDEV_PATH,
	strlen(DEV_BYVDEV_PATH)) != 0) {
	/* by_vdev_path doesn't start with "/dev/disk/by-vdev/" */
	return (B_FALSE);
	}

	(void) zpool_iter(g_zfshdl, zfs_iter_pool, &data);

	return (data.dd_found);
	}

	/*
	* Given a device identifier, find any vdevs with a matching devid.
	* On Linux we can match devid directly which is always a whole disk.
	*/
	static boolean_t
	devid_iter(const char *devid, zfs_process_func_t func, boolean_t is_slice)
	{
	dev_data_t data = { 0 };

	data.dd_compare = devid;
	data.dd_func = func;
	data.dd_prop = ZPOOL_CONFIG_DEVID;
	data.dd_found = B_FALSE;
	data.dd_islabeled = is_slice;
	data.dd_new_devid = devid;

	(void) zpool_iter(g_zfshdl, zfs_iter_pool, &data);

	return (data.dd_found);
	}

	/*
	* Given a device guid, find any vdevs with a matching guid.
	*/
	static boolean_t
	guid_iter(uint64_t pool_guid, uint64_t vdev_guid, const char *devid,
	zfs_process_func_t func, boolean_t is_slice)
	{
	dev_data_t data = { 0 };

	data.dd_func = func;
	data.dd_found = B_FALSE;
	data.dd_pool_guid = pool_guid;
	data.dd_vdev_guid = vdev_guid;
	data.dd_islabeled = is_slice;
	data.dd_new_devid = devid;

	(void) zpool_iter(g_zfshdl, zfs_iter_pool, &data);

	return (data.dd_found);
	}

	/*
	* Handle a EC_DEV_ADD.ESC_DISK event.
	*
	* illumos
	* Expects: DEV_PHYS_PATH string in schema
	* Matches: vdev's ZPOOL_CONFIG_PHYS_PATH or ZPOOL_CONFIG_DEVID
	*
	* path: '/dev/dsk/c0t1d0s0' (persistent)
	* devid: 'id1,sd@SATA_____Hitachi_HDS72101______JP2940HZ3H74MC/a'
	* phys_path: '/pci@0,0/pci103c,1609@11/disk@1,0:a'
	*
	* linux
	* provides: DEV_PHYS_PATH and DEV_IDENTIFIER strings in schema
	* Matches: vdev's ZPOOL_CONFIG_PHYS_PATH or ZPOOL_CONFIG_DEVID
	*
	* path: '/dev/sdc1' (not persistent)
	* devid: 'ata-SAMSUNG_HD204UI_S2HGJD2Z805891-part1'
	* phys_path: 'pci-0000:04:00.0-sas-0x4433221106000000-lun-0'
	*/
	static int
	zfs_deliver_add(nvlist_t *nvl)
	{
	const char devpath = NULL, devid = NULL;
	uint64_t pool_guid = 0, vdev_guid = 0;
	boolean_t is_slice;

	/*
	* Expecting a devid string and an optional physical location and guid
	*/
	if (nvlist_lookup_string(nvl, DEV_IDENTIFIER, &devid) != 0) {
	zed_log_msg(LOG_INFO, "%s: no dev identifier\n", __func__);
	return (-1);
	}

	(void) nvlist_lookup_string(nvl, DEV_PHYS_PATH, &devpath);
	(void) nvlist_lookup_uint64(nvl, ZFS_EV_POOL_GUID, &pool_guid);
	(void) nvlist_lookup_uint64(nvl, ZFS_EV_VDEV_GUID, &vdev_guid);

	is_slice = (nvlist_lookup_boolean(nvl, DEV_IS_PART) == 0);

	zed_log_msg(LOG_INFO, "zfs_deliver_add: adding %s (%s) (is_slice %d)",
	devid, devpath ? devpath : "NULL", is_slice);

	/*
	* Iterate over all vdevs looking for a match in the following order:
	* 1. ZPOOL_CONFIG_DEVID (identifies the unique disk)
	* 2. ZPOOL_CONFIG_PHYS_PATH (identifies disk physical location).
	* 3. ZPOOL_CONFIG_GUID (identifies unique vdev).
	* 4. ZPOOL_CONFIG_PATH for /dev/disk/by-vdev devices only (since
	* by-vdev paths represent physical paths).
	*/
	if (devid_iter(devid, zfs_process_add, is_slice))
	return (0);
	if (devpath != NULL && devphys_iter(devpath, devid, zfs_process_add,
	is_slice, vdev_guid))
	return (0);
	if (vdev_guid != 0)
	(void) guid_iter(pool_guid, vdev_guid, devid, zfs_process_add,
	is_slice);

	if (devpath != NULL) {
	/* Can we match a /dev/disk/by-vdev/ path? */
	char by_vdev_path[MAXPATHLEN];
	snprintf(by_vdev_path, sizeof (by_vdev_path),
	"/dev/disk/by-vdev/%s", devpath);
	if (by_vdev_path_iter(by_vdev_path, devid, zfs_process_add,
	is_slice))
	return (0);
	}

	return (0);
	}

	/*
	* Called when we receive a VDEV_CHECK event, which indicates a device could not
	* be opened during initial pool open, but the autoreplace property was set on
	* the pool. In this case, we treat it as if it were an add event.
	*/
	static int
	zfs_deliver_check(nvlist_t *nvl)
	{
	dev_data_t data = { 0 };

	if (nvlist_lookup_uint64(nvl, ZFS_EV_POOL_GUID,
	&data.dd_pool_guid) != 0 \|\|
	nvlist_lookup_uint64(nvl, ZFS_EV_VDEV_GUID,
	&data.dd_vdev_guid) != 0 \|\|
	data.dd_vdev_guid == 0)
	return (0);

	zed_log_msg(LOG_INFO, "zfs_deliver_check: pool '%llu', vdev %llu",
	data.dd_pool_guid, data.dd_vdev_guid);

	data.dd_func = zfs_process_add;

	(void) zpool_iter(g_zfshdl, zfs_iter_pool, &data);

	return (0);
	}

	/*
	* Given a path to a vdev, lookup the vdev's physical size from its
	* config nvlist.
	*
	* Returns the vdev's physical size in bytes on success, 0 on error.
	*/
	static uint64_t
	vdev_size_from_config(zpool_handle_t zhp, const char vdev_path)
	{
	nvlist_t *nvl = NULL;
	boolean_t avail_spare, l2cache, log;
	vdev_stat_t *vs = NULL;
	uint_t c;

	nvl = zpool_find_vdev(zhp, vdev_path, &avail_spare, &l2cache, &log);
	if (!nvl)
	return (0);

	verify(nvlist_lookup_uint64_array(nvl, ZPOOL_CONFIG_VDEV_STATS,
	(uint64_t **)&vs, &c) == 0);
	if (!vs) {
	zed_log_msg(LOG_INFO, "%s: no nvlist for '%s'", __func__,
	vdev_path);
	return (0);
	}

	return (vs->vs_pspace);
	}

	/*
	* Given a path to a vdev, lookup if the vdev is a "whole disk" in the
	* config nvlist. "whole disk" means that ZFS was passed a whole disk
	* at pool creation time, which it partitioned up and has full control over.
	* Thus a partition with wholedisk=1 set tells us that zfs created the
	* partition at creation time. A partition without whole disk set would have
	* been created by externally (like with fdisk) and passed to ZFS.
	*
	* Returns the whole disk value (either 0 or 1).
	*/
	static uint64_t
	vdev_whole_disk_from_config(zpool_handle_t zhp, const char vdev_path)
	{
	nvlist_t *nvl = NULL;
	boolean_t avail_spare, l2cache, log;
	uint64_t wholedisk = 0;

	nvl = zpool_find_vdev(zhp, vdev_path, &avail_spare, &l2cache, &log);
	if (!nvl)
	return (0);

	(void) nvlist_lookup_uint64(nvl, ZPOOL_CONFIG_WHOLE_DISK, &wholedisk);

	return (wholedisk);
	}

	/*
	* If the device size grew more than 1% then return true.
	*/
	#define DEVICE_GREW(oldsize, newsize) \
	((newsize > oldsize) && \
	((newsize / (newsize - oldsize)) <= 100))

	static int
	zfsdle_vdev_online(zpool_handle_t zhp, void data)
	{
	boolean_t avail_spare, l2cache;
	nvlist_t *udev_nvl = data;
	nvlist_t *tgt;
	int error;

	const char *tmp_devname;
	char devname[MAXPATHLEN] = "";
	uint64_t guid;

	if (nvlist_lookup_uint64(udev_nvl, ZFS_EV_VDEV_GUID, &guid) == 0) {
	sprintf(devname, "%llu", (u_longlong_t)guid);
	} else if (nvlist_lookup_string(udev_nvl, DEV_PHYS_PATH,
	&tmp_devname) == 0) {
	strlcpy(devname, tmp_devname, MAXPATHLEN);
	zfs_append_partition(devname, MAXPATHLEN);
	} else {
	zed_log_msg(LOG_INFO, "%s: no guid or physpath", __func__);
	}

	zed_log_msg(LOG_INFO, "zfsdle_vdev_online: searching for '%s' in '%s'",
	devname, zpool_get_name(zhp));

	if ((tgt = zpool_find_vdev_by_physpath(zhp, devname,
	&avail_spare, &l2cache, NULL)) != NULL) {
	const char *path;
	char fullpath[MAXPATHLEN];
	uint64_t wholedisk = 0;

	error = nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH, &path);
	if (error) {
	zpool_close(zhp);
	return (0);
	}

	(void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
	&wholedisk);

	if (wholedisk) {
	char *tmp;
	path = strrchr(path, '/');
	if (path != NULL) {
	tmp = zfs_strip_partition(path + 1);
	if (tmp == NULL) {
	zpool_close(zhp);
	return (0);
	}
	} else {
	zpool_close(zhp);
	return (0);
	}

	(void) strlcpy(fullpath, tmp, sizeof (fullpath));
	free(tmp);

	/*
	* We need to reopen the pool associated with this
	* device so that the kernel can update the size of
	* the expanded device. When expanding there is no
	* need to restart the scrub from the beginning.
	*/
	boolean_t scrub_restart = B_FALSE;
	(void) zpool_reopen_one(zhp, &scrub_restart);
	} else {
	(void) strlcpy(fullpath, path, sizeof (fullpath));
	}

	if (zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) {
	vdev_state_t newstate;

	if (zpool_get_state(zhp) != POOL_STATE_UNAVAIL) {
	/*
	* If this disk size has not changed, then
	* there's no need to do an autoexpand. To
	* check we look at the disk's size in its
	* config, and compare it to the disk size
	* that udev is reporting.
	*/
	uint64_t udev_size = 0, conf_size = 0,
	wholedisk = 0, udev_parent_size = 0;

	/*
	* Get the size of our disk that udev is
	* reporting.
	*/
	if (nvlist_lookup_uint64(udev_nvl, DEV_SIZE,
	&udev_size) != 0) {
	udev_size = 0;
	}

	/*
	* Get the size of our disk's parent device
	* from udev (where sda1's parent is sda).
	*/
	if (nvlist_lookup_uint64(udev_nvl,
	DEV_PARENT_SIZE, &udev_parent_size) != 0) {
	udev_parent_size = 0;
	}

	conf_size = vdev_size_from_config(zhp,
	fullpath);

	wholedisk = vdev_whole_disk_from_config(zhp,
	fullpath);

	/*
	* Only attempt an autoexpand if the vdev size
	* changed. There are two different cases
	* to consider.
	*
	* 1. wholedisk=1
	* If you do a 'zpool create' on a whole disk
	* (like /dev/sda), then zfs will create
	* partitions on the disk (like /dev/sda1). In
	* that case, wholedisk=1 will be set in the
	* partition's nvlist config. So zed will need
	* to see if your parent device (/dev/sda)
	* expanded in size, and if so, then attempt
	* the autoexpand.
	*
	* 2. wholedisk=0
	* If you do a 'zpool create' on an existing
	* partition, or a device that doesn't allow
	* partitions, then wholedisk=0, and you will
	* simply need to check if the device itself
	* expanded in size.
	*/
	if (DEVICE_GREW(conf_size, udev_size) \|\|
	(wholedisk && DEVICE_GREW(conf_size,
	udev_parent_size))) {
	error = zpool_vdev_online(zhp, fullpath,
	0, &newstate);

	zed_log_msg(LOG_INFO,
	"%s: autoexpanding '%s' from %llu"
	" to %llu bytes in pool '%s': %d",
	__func__, fullpath, conf_size,
	MAX(udev_size, udev_parent_size),
	zpool_get_name(zhp), error);
	}
	}
	}
	zpool_close(zhp);
	return (1);
	}
	zpool_close(zhp);
	return (0);
	}

	/*
	* This function handles the ESC_DEV_DLE device change event. Use the
	* provided vdev guid when looking up a disk or partition, when the guid
	* is not present assume the entire disk is owned by ZFS and append the
	* expected -part1 partition information then lookup by physical path.
	*/
	static int
	zfs_deliver_dle(nvlist_t *nvl)
	{
	const char *devname;
	char name[MAXPATHLEN];
	uint64_t guid;

	if (nvlist_lookup_uint64(nvl, ZFS_EV_VDEV_GUID, &guid) == 0) {
	sprintf(name, "%llu", (u_longlong_t)guid);
	} else if (nvlist_lookup_string(nvl, DEV_PHYS_PATH, &devname) == 0) {
	strlcpy(name, devname, MAXPATHLEN);
	zfs_append_partition(name, MAXPATHLEN);
	} else {
	sprintf(name, "unknown");
	zed_log_msg(LOG_INFO, "zfs_deliver_dle: no guid or physpath");
	}

	if (zpool_iter(g_zfshdl, zfsdle_vdev_online, nvl) != 1) {
	zed_log_msg(LOG_INFO, "zfs_deliver_dle: device '%s' not "
	"found", name);
	return (1);
	}

	return (0);
	}

	/*
	* syseventd daemon module event handler
	*
	* Handles syseventd daemon zfs device related events:
	*
	* EC_DEV_ADD.ESC_DISK
	* EC_DEV_STATUS.ESC_DEV_DLE
	* EC_ZFS.ESC_ZFS_VDEV_CHECK
	*
	* Note: assumes only one thread active at a time (not thread safe)
	*/
	static int
	zfs_slm_deliver_event(const char class, const char subclass, nvlist_t *nvl)
	{
	int ret;
	boolean_t is_check = B_FALSE, is_dle = B_FALSE;

	if (strcmp(class, EC_DEV_ADD) == 0) {
	/*
	* We're mainly interested in disk additions, but we also listen
	* for new loop devices, to allow for simplified testing.
	*/
	if (strcmp(subclass, ESC_DISK) != 0 &&
	strcmp(subclass, ESC_LOFI) != 0)
	return (0);

	is_check = B_FALSE;
	} else if (strcmp(class, EC_ZFS) == 0 &&
	strcmp(subclass, ESC_ZFS_VDEV_CHECK) == 0) {
	/*
	* This event signifies that a device failed to open
	* during pool load, but the 'autoreplace' property was
	* set, so we should pretend it's just been added.
	*/
	is_check = B_TRUE;
	} else if (strcmp(class, EC_DEV_STATUS) == 0 &&
	strcmp(subclass, ESC_DEV_DLE) == 0) {
	is_dle = B_TRUE;
	} else {
	return (0);
	}

	if (is_dle)
	ret = zfs_deliver_dle(nvl);
	else if (is_check)
	ret = zfs_deliver_check(nvl);
	else
	ret = zfs_deliver_add(nvl);

	return (ret);
	}

	static void *
	zfs_enum_pools(void *arg)
	{
	(void) arg;

	(void) zpool_iter(g_zfshdl, zfs_unavail_pool, (void *)&g_pool_list);
	/*
	* Linux - instead of using a thread pool, each list entry
	* will spawn a thread when an unavailable pool transitions
	* to available. zfs_slm_fini will wait for these threads.
	*/
	g_enumeration_done = B_TRUE;
	return (NULL);
	}

	/*
	* called from zed daemon at startup
	*
	* sent messages from zevents or udev monitor
	*
	* For now, each agent has its own libzfs instance
	*/
	int
	zfs_slm_init(void)
	{
	if ((g_zfshdl = libzfs_init()) == NULL)
	return (-1);

	/*
	* collect a list of unavailable pools (asynchronously,
	* since this can take a while)
	*/
	list_create(&g_pool_list, sizeof (struct unavailpool),
	offsetof(struct unavailpool, uap_node));

	if (pthread_create(&g_zfs_tid, NULL, zfs_enum_pools, NULL) != 0) {
	list_destroy(&g_pool_list);
	libzfs_fini(g_zfshdl);
	return (-1);
	}

	pthread_setname_np(g_zfs_tid, "enum-pools");
	list_create(&g_device_list, sizeof (struct pendingdev),
	offsetof(struct pendingdev, pd_node));

	return (0);
	}

	void
	zfs_slm_fini(void)
	{
	unavailpool_t *pool;
	pendingdev_t *device;

	/* wait for zfs_enum_pools thread to complete */
	(void) pthread_join(g_zfs_tid, NULL);
	/* destroy the thread pool */
	if (g_tpool != NULL) {
	tpool_wait(g_tpool);
	tpool_destroy(g_tpool);
	}

	while ((pool = list_remove_head(&g_pool_list)) != NULL) {
	zpool_close(pool->uap_zhp);
	free(pool);
	}
	list_destroy(&g_pool_list);

	while ((device = list_remove_head(&g_device_list)) != NULL)
	free(device);
	list_destroy(&g_device_list);

	libzfs_fini(g_zfshdl);
	}

	void
	zfs_slm_event(const char class, const char subclass, nvlist_t *nvl)
	{
	zed_log_msg(LOG_INFO, "zfs_slm_event: %s.%s", class, subclass);
	(void) zfs_slm_deliver_event(class, subclass, nvl);
	}
	diff --git a/sys/contrib/openzfs/cmd/zed/zed.d/Makefile.am b/sys/contrib/openzfs/cmd/zed/zed.d/Makefile.am
	index 812558cf6d0f..093a04c4636a 100644
	--- a/sys/contrib/openzfs/cmd/zed/zed.d/Makefile.am
	+++ b/sys/contrib/openzfs/cmd/zed/zed.d/Makefile.am
	@@ -1,57 +1,59 @@
	zedconfdir = $(sysconfdir)/zfs/zed.d
	dist_zedconf_DATA = \
	%D%/zed-functions.sh \
	%D%/zed.rc

	zedexecdir = $(zfsexecdir)/zed.d

	dist_zedexec_SCRIPTS = \
	%D%/all-debug.sh \
	%D%/all-syslog.sh \
	%D%/data-notify.sh \
	+ %D%/deadman-slot_off.sh \
	%D%/generic-notify.sh \
	%D%/pool_import-led.sh \
	%D%/resilver_finish-notify.sh \
	%D%/resilver_finish-start-scrub.sh \
	%D%/scrub_finish-notify.sh \
	%D%/statechange-led.sh \
	%D%/statechange-notify.sh \
	%D%/statechange-slot_off.sh \
	%D%/trim_finish-notify.sh \
	%D%/vdev_attach-led.sh \
	%D%/vdev_clear-led.sh

	nodist_zedexec_SCRIPTS = \
	%D%/history_event-zfs-list-cacher.sh

	SUBSTFILES += $(nodist_zedexec_SCRIPTS)

	zedconfdefaults = \
	all-syslog.sh \
	data-notify.sh \
	+ deadman-slot_off.sh \
	history_event-zfs-list-cacher.sh \
	pool_import-led.sh \
	resilver_finish-notify.sh \
	resilver_finish-start-scrub.sh \
	scrub_finish-notify.sh \
	statechange-led.sh \
	statechange-notify.sh \
	statechange-slot_off.sh \
	vdev_attach-led.sh \
	vdev_clear-led.sh

	dist_noinst_DATA += %D%/README

	INSTALL_DATA_HOOKS += zed-install-data-hook
	zed-install-data-hook:
	$(MKDIR_P) "$(DESTDIR)$(zedconfdir)"
	set -x; for f in $(zedconfdefaults); do \
	[ -f "$(DESTDIR)$(zedconfdir)/$${f}" ] \|\|\
	[ -L "$(DESTDIR)$(zedconfdir)/$${f}" ] \|\| \
	$(LN_S) "$(zedexecdir)/$${f}" "$(DESTDIR)$(zedconfdir)"; \
	done

	SHELLCHECKSCRIPTS += $(dist_zedconf_DATA) $(dist_zedexec_SCRIPTS) $(nodist_zedexec_SCRIPTS)
	$(call SHELLCHECK_OPTS,$(dist_zedconf_DATA) $(dist_zedexec_SCRIPTS) $(nodist_zedexec_SCRIPTS)): SHELLCHECK_SHELL = sh
	# False positive: 1>&"${ZED_FLOCK_FD}" looks suspiciously similar to a >&filename bash extension
	$(call SHELLCHECK_OPTS,$(dist_zedconf_DATA) $(dist_zedexec_SCRIPTS) $(nodist_zedexec_SCRIPTS)): CHECKBASHISMS_IGNORE = -e 'should be >word 2>&1' -e '&"$${ZED_FLOCK_FD}"'
	diff --git a/sys/contrib/openzfs/cmd/zed/zed.d/deadman-slot_off.sh b/sys/contrib/openzfs/cmd/zed/zed.d/deadman-slot_off.sh
	new file mode 100755
	index 000000000000..7b339b3add01
	--- /dev/null
	+++ b/sys/contrib/openzfs/cmd/zed/zed.d/deadman-slot_off.sh
	@@ -0,0 +1,71 @@
	+#!/bin/sh
	+# shellcheck disable=SC3014,SC2154,SC2086,SC2034
	+#
	+# Turn off disk's enclosure slot if an I/O is hung triggering the deadman.
	+#
	+# It's possible for outstanding I/O to a misbehaving SCSI disk to neither
	+# promptly complete or return an error. This can occur due to retry and
	+# recovery actions taken by the SCSI layer, driver, or disk. When it occurs
	+# the pool will be unresponsive even though there may be sufficient redundancy
	+# configured to proceeded without this single disk.
	+#
	+# When a hung I/O is detected by the kmods it will be posted as a deadman
	+# event. By default an I/O is considered to be hung after 5 minutes. This
	+# value can be changed with the zfs_deadman_ziotime_ms module parameter.
	+# If ZED_POWER_OFF_ENCLOSURE_SLOT_ON_DEADMAN is set the disk's enclosure
	+# slot will be powered off causing the outstanding I/O to fail. The ZED
	+# will then handle this like a normal disk failure and FAULT the vdev.
	+#
	+# We assume the user will be responsible for turning the slot back on
	+# after replacing the disk.
	+#
	+# Note that this script requires that your enclosure be supported by the
	+# Linux SCSI Enclosure services (SES) driver. The script will do nothing
	+# if you have no enclosure, or if your enclosure isn't supported.
	+#
	+# Exit codes:
	+# 0: slot successfully powered off
	+# 1: enclosure not available
	+# 2: ZED_POWER_OFF_ENCLOSURE_SLOT_ON_DEADMAN disabled
	+# 3: System not configured to wait on deadman
	+# 4: The enclosure sysfs path passed from ZFS does not exist
	+# 5: Enclosure slot didn't actually turn off after we told it to
	+
	+[ -f "${ZED_ZEDLET_DIR}/zed.rc" ] && . "${ZED_ZEDLET_DIR}/zed.rc"
	+. "${ZED_ZEDLET_DIR}/zed-functions.sh"
	+
	+if [ ! -d /sys/class/enclosure ] ; then
	+ # No JBOD enclosure or NVMe slots
	+ exit 1
	+fi
	+
	+if [ "${ZED_POWER_OFF_ENCLOSURE_SLOT_ON_DEADMAN}" != "1" ] ; then
	+ exit 2
	+fi
	+
	+if [ "$ZEVENT_POOL_FAILMODE" != "wait" ] ; then
	+ exit 3
	+fi
	+
	+if [ ! -f "$ZEVENT_VDEV_ENC_SYSFS_PATH/power_status" ] ; then
	+ exit 4
	+fi
	+
	+# Turn off the slot and wait for sysfs to report that the slot is off.
	+# It can take ~400ms on some enclosures and multiple retries may be needed.
	+for i in $(seq 1 20) ; do
	+ echo "off" \| tee "$ZEVENT_VDEV_ENC_SYSFS_PATH/power_status"
	+
	+ for j in $(seq 1 5) ; do
	+ if [ "$(cat $ZEVENT_VDEV_ENC_SYSFS_PATH/power_status)" == "off" ] ; then
	+ break 2
	+ fi
	+ sleep 0.1
	+ done
	+done
	+
	+if [ "$(cat $ZEVENT_VDEV_ENC_SYSFS_PATH/power_status)" != "off" ] ; then
	+ exit 5
	+fi
	+
	+zed_log_msg "powered down slot $ZEVENT_VDEV_ENC_SYSFS_PATH for $ZEVENT_VDEV_PATH"
	diff --git a/sys/contrib/openzfs/cmd/zed/zed.d/zed.rc b/sys/contrib/openzfs/cmd/zed/zed.d/zed.rc
	index ec64ecfaa13c..af56147a969b 100644
	--- a/sys/contrib/openzfs/cmd/zed/zed.d/zed.rc
	+++ b/sys/contrib/openzfs/cmd/zed/zed.d/zed.rc
	@@ -1,192 +1,199 @@
	##
	# zed.rc – ZEDLET configuration.
	##
	# shellcheck disable=SC2034

	##
	# Absolute path to the debug output file.
	#
	#ZED_DEBUG_LOG="/tmp/zed.debug.log"

	##
	# Email address of the zpool administrator for receipt of notifications;
	# multiple addresses can be specified if they are delimited by whitespace.
	# Email will only be sent if ZED_EMAIL_ADDR is defined.
	# Enabled by default; comment to disable.
	#
	ZED_EMAIL_ADDR="root"

	##
	# Name or path of executable responsible for sending notifications via email;
	# the mail program must be capable of reading a message body from stdin.
	# Email will only be sent if ZED_EMAIL_ADDR is defined.
	#
	#ZED_EMAIL_PROG="mail"

	##
	# Command-line options for ZED_EMAIL_PROG.
	# The string @ADDRESS@ will be replaced with the recipient email address(es).
	# The string @SUBJECT@ will be replaced with the notification subject;
	# this should be protected with quotes to prevent word-splitting.
	# Email will only be sent if ZED_EMAIL_ADDR is defined.
	# If @SUBJECT@ was omited here, a "Subject: ..." header will be added to notification
	#
	#ZED_EMAIL_OPTS="-s '@SUBJECT@' @ADDRESS@"

	##
	# Default directory for zed lock files.
	#
	#ZED_LOCKDIR="/var/lock"

	##
	# Minimum number of seconds between notifications for a similar event.
	#
	#ZED_NOTIFY_INTERVAL_SECS=3600

	##
	# Notification verbosity.
	# If set to 0, suppress notification if the pool is healthy.
	# If set to 1, send notification regardless of pool health.
	#
	#ZED_NOTIFY_VERBOSE=0

	##
	# Send notifications for 'ereport.fs.zfs.data' events.
	# Disabled by default, any non-empty value will enable the feature.
	#
	#ZED_NOTIFY_DATA=

	##
	# Pushbullet access token.
	# This grants full access to your account -- protect it accordingly!
	# <https://www.pushbullet.com/get-started>
	# <https://www.pushbullet.com/account>
	# Disabled by default; uncomment to enable.
	#
	#ZED_PUSHBULLET_ACCESS_TOKEN=""

	##
	# Pushbullet channel tag for push notification feeds that can be subscribed to.
	# <https://www.pushbullet.com/my-channel>
	# If not defined, push notifications will instead be sent to all devices
	# associated with the account specified by the access token.
	# Disabled by default; uncomment to enable.
	#
	#ZED_PUSHBULLET_CHANNEL_TAG=""

	##
	# Slack Webhook URL.
	# This allows posting to the given channel and includes an access token.
	# <https://api.slack.com/incoming-webhooks>
	# Disabled by default; uncomment to enable.
	#
	#ZED_SLACK_WEBHOOK_URL=""

	##
	# Pushover token.
	# This defines the application from which the notification will be sent.
	# <https://pushover.net/api#registration>
	# Disabled by default; uncomment to enable.
	# ZED_PUSHOVER_USER, below, must also be configured.
	#
	#ZED_PUSHOVER_TOKEN=""

	##
	# Pushover user key.
	# This defines which user or group will receive Pushover notifications.
	# <https://pushover.net/api#identifiers>
	# Disabled by default; uncomment to enable.
	# ZED_PUSHOVER_TOKEN, above, must also be configured.
	#ZED_PUSHOVER_USER=""

	##
	# Default directory for zed state files.
	#
	#ZED_RUNDIR="/var/run"

	##
	# Turn on/off enclosure LEDs when drives get DEGRADED/FAULTED. This works for
	# device mapper and multipath devices as well. This works with JBOD enclosures
	# and NVMe PCI drives (assuming they're supported by Linux in sysfs).
	#
	ZED_USE_ENCLOSURE_LEDS=1

	##
	# Run a scrub after every resilver
	# Disabled by default, 1 to enable and 0 to disable.
	#ZED_SCRUB_AFTER_RESILVER=0

	##
	# The syslog priority (e.g., specified as a "facility.level" pair).
	#
	#ZED_SYSLOG_PRIORITY="daemon.notice"

	##
	# The syslog tag for marking zed events.
	#
	#ZED_SYSLOG_TAG="zed"

	##
	# Which set of event subclasses to log
	# By default, events from all subclasses are logged.
	# If ZED_SYSLOG_SUBCLASS_INCLUDE is set, only subclasses
	# matching the pattern are logged. Use the pipe symbol (\|)
	# or shell wildcards (*, ?) to match multiple subclasses.
	# Otherwise, if ZED_SYSLOG_SUBCLASS_EXCLUDE is set, the
	# matching subclasses are excluded from logging.
	#ZED_SYSLOG_SUBCLASS_INCLUDE="checksum\|scrub_\|vdev."
	ZED_SYSLOG_SUBCLASS_EXCLUDE="history_event"

	##
	# Use GUIDs instead of names when logging pool and vdevs
	# Disabled by default, 1 to enable and 0 to disable.
	#ZED_SYSLOG_DISPLAY_GUIDS=1

	##
	# Power off the drive's slot in the enclosure if it becomes FAULTED. This can
	# help silence misbehaving drives. This assumes your drive enclosure fully
	# supports slot power control via sysfs.
	#ZED_POWER_OFF_ENCLOSURE_SLOT_ON_FAULT=1

	+##
	+# Power off the drive's slot in the enclosure if there is a hung I/O which
	+# exceeds the deadman timeout. This can help prevent a single misbehaving
	+# drive from rendering a redundant pool unavailable. This assumes your drive
	+# enclosure fully supports slot power control via sysfs.
	+#ZED_POWER_OFF_ENCLOSURE_SLOT_ON_DEADMAN=1
	+
	##
	# Ntfy topic
	# This defines which topic will receive the ntfy notification.
	# <https://docs.ntfy.sh/publish/>
	# Disabled by default; uncomment to enable.
	#ZED_NTFY_TOPIC=""

	##
	# Ntfy access token (optional for public topics)
	# This defines an access token which can be used
	# to allow you to authenticate when sending to topics
	# <https://docs.ntfy.sh/publish/#access-tokens>
	# Disabled by default; uncomment to enable.
	#ZED_NTFY_ACCESS_TOKEN=""

	##
	# Ntfy Service URL
	# This defines which service the ntfy call will be directed toward
	# <https://docs.ntfy.sh/install/>
	# https://ntfy.sh by default; uncomment to enable an alternative service url.
	#ZED_NTFY_URL="https://ntfy.sh"

	##
	# Gotify server URL
	# This defines a URL that the Gotify call will be directed toward.
	# <https://gotify.net/docs/index>
	# Disabled by default; uncomment to enable.
	#ZED_GOTIFY_URL=""

	##
	# Gotify application token
	# This defines a Gotify application token which a message is associated with.
	# This token is generated when an application is created on the Gotify server.
	# Disabled by default; uncomment to enable.
	#ZED_GOTIFY_APPTOKEN=""

	##
	# Gotify priority (optional)
	# If defined, this overrides the default priority of the
	# Gotify application associated with ZED_GOTIFY_APPTOKEN.
	# Value is an integer 0 and up.
	#ZED_GOTIFY_PRIORITY=""
	diff --git a/sys/contrib/openzfs/cmd/zfs/zfs_main.c b/sys/contrib/openzfs/cmd/zfs/zfs_main.c
	index 3fc2d2810495..b77917764c86 100644
	--- a/sys/contrib/openzfs/cmd/zfs/zfs_main.c
	+++ b/sys/contrib/openzfs/cmd/zfs/zfs_main.c
	@@ -1,9058 +1,9060 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
	* Copyright 2012 Milan Jurik. All rights reserved.
	* Copyright (c) 2012, Joyent, Inc. All rights reserved.
	* Copyright (c) 2013 Steven Hartland. All rights reserved.
	* Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>.
	* Copyright 2016 Nexenta Systems, Inc.
	* Copyright (c) 2019 Datto Inc.
	* Copyright (c) 2019, loli10K <ezomori.nozomu@gmail.com>
	* Copyright 2019 Joyent, Inc.
	* Copyright (c) 2019, 2020 by Christian Schwarz. All rights reserved.
	*/

	#include <assert.h>
	#include <ctype.h>
	#include <sys/debug.h>
	#include <errno.h>
	#include <getopt.h>
	#include <libgen.h>
	#include <libintl.h>
	#include <libuutil.h>
	#include <libnvpair.h>
	#include <locale.h>
	#include <stddef.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <string.h>
	#include <unistd.h>
	#include <fcntl.h>
	#include <zone.h>
	#include <grp.h>
	#include <pwd.h>
	#include <umem.h>
	#include <pthread.h>
	#include <signal.h>
	#include <sys/list.h>
	#include <sys/mkdev.h>
	#include <sys/mntent.h>
	#include <sys/mnttab.h>
	#include <sys/mount.h>
	#include <sys/stat.h>
	#include <sys/fs/zfs.h>
	#include <sys/systeminfo.h>
	#include <sys/types.h>
	#include <time.h>
	#include <sys/zfs_project.h>

	#include <libzfs.h>
	#include <libzfs_core.h>
	#include <zfs_prop.h>
	#include <zfs_deleg.h>
	#include <libzutil.h>
	#ifdef HAVE_IDMAP
	#include <aclutils.h>
	#include <directory.h>
	#endif /* HAVE_IDMAP */

	#include "zfs_iter.h"
	#include "zfs_util.h"
	#include "zfs_comutil.h"
	#include "zfs_projectutil.h"

	libzfs_handle_t *g_zfs;

	static char history_str[HIS_MAX_RECORD_LEN];
	static boolean_t log_history = B_TRUE;

	static int zfs_do_clone(int argc, char **argv);
	static int zfs_do_create(int argc, char **argv);
	static int zfs_do_destroy(int argc, char **argv);
	static int zfs_do_get(int argc, char **argv);
	static int zfs_do_inherit(int argc, char **argv);
	static int zfs_do_list(int argc, char **argv);
	static int zfs_do_mount(int argc, char **argv);
	static int zfs_do_rename(int argc, char **argv);
	static int zfs_do_rollback(int argc, char **argv);
	static int zfs_do_set(int argc, char **argv);
	static int zfs_do_upgrade(int argc, char **argv);
	static int zfs_do_snapshot(int argc, char **argv);
	static int zfs_do_unmount(int argc, char **argv);
	static int zfs_do_share(int argc, char **argv);
	static int zfs_do_unshare(int argc, char **argv);
	static int zfs_do_send(int argc, char **argv);
	static int zfs_do_receive(int argc, char **argv);
	static int zfs_do_promote(int argc, char **argv);
	static int zfs_do_userspace(int argc, char **argv);
	static int zfs_do_allow(int argc, char **argv);
	static int zfs_do_unallow(int argc, char **argv);
	static int zfs_do_hold(int argc, char **argv);
	static int zfs_do_holds(int argc, char **argv);
	static int zfs_do_release(int argc, char **argv);
	static int zfs_do_diff(int argc, char **argv);
	static int zfs_do_bookmark(int argc, char **argv);
	static int zfs_do_channel_program(int argc, char **argv);
	static int zfs_do_load_key(int argc, char **argv);
	static int zfs_do_unload_key(int argc, char **argv);
	static int zfs_do_change_key(int argc, char **argv);
	static int zfs_do_project(int argc, char **argv);
	static int zfs_do_version(int argc, char **argv);
	static int zfs_do_redact(int argc, char **argv);
	static int zfs_do_wait(int argc, char **argv);

	#ifdef __FreeBSD__
	static int zfs_do_jail(int argc, char **argv);
	static int zfs_do_unjail(int argc, char **argv);
	#endif

	#ifdef __linux__
	static int zfs_do_zone(int argc, char **argv);
	static int zfs_do_unzone(int argc, char **argv);
	#endif

	static int zfs_do_help(int argc, char **argv);

	/*
	* Enable a reasonable set of defaults for libumem debugging on DEBUG builds.
	*/

	#ifdef DEBUG
	const char *
	_umem_debug_init(void)
	{
	return ("default,verbose"); /* $UMEM_DEBUG setting */
	}

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

	typedef enum {
	HELP_CLONE,
	HELP_CREATE,
	HELP_DESTROY,
	HELP_GET,
	HELP_INHERIT,
	HELP_UPGRADE,
	HELP_LIST,
	HELP_MOUNT,
	HELP_PROMOTE,
	HELP_RECEIVE,
	HELP_RENAME,
	HELP_ROLLBACK,
	HELP_SEND,
	HELP_SET,
	HELP_SHARE,
	HELP_SNAPSHOT,
	HELP_UNMOUNT,
	HELP_UNSHARE,
	HELP_ALLOW,
	HELP_UNALLOW,
	HELP_USERSPACE,
	HELP_GROUPSPACE,
	HELP_PROJECTSPACE,
	HELP_PROJECT,
	HELP_HOLD,
	HELP_HOLDS,
	HELP_RELEASE,
	HELP_DIFF,
	HELP_BOOKMARK,
	HELP_CHANNEL_PROGRAM,
	HELP_LOAD_KEY,
	HELP_UNLOAD_KEY,
	HELP_CHANGE_KEY,
	HELP_VERSION,
	HELP_REDACT,
	HELP_JAIL,
	HELP_UNJAIL,
	HELP_WAIT,
	HELP_ZONE,
	HELP_UNZONE,
	} zfs_help_t;

	typedef struct zfs_command {
	const char *name;
	int (func)(int argc, char *argv);
	zfs_help_t usage;
	} zfs_command_t;

	/*
	* Master command table. Each ZFS command has a name, associated function, and
	* usage message. The usage messages need to be internationalized, so we have
	* to have a function to return the usage message based on a command index.
	*
	* These commands are organized according to how they are displayed in the usage
	* message. An empty command (one with a NULL name) indicates an empty line in
	* the generic usage message.
	*/
	static zfs_command_t command_table[] = {
	{ "version", zfs_do_version, HELP_VERSION },
	{ NULL },
	{ "create", zfs_do_create, HELP_CREATE },
	{ "destroy", zfs_do_destroy, HELP_DESTROY },
	{ NULL },
	{ "snapshot", zfs_do_snapshot, HELP_SNAPSHOT },
	{ "rollback", zfs_do_rollback, HELP_ROLLBACK },
	{ "clone", zfs_do_clone, HELP_CLONE },
	{ "promote", zfs_do_promote, HELP_PROMOTE },
	{ "rename", zfs_do_rename, HELP_RENAME },
	{ "bookmark", zfs_do_bookmark, HELP_BOOKMARK },
	{ "program", zfs_do_channel_program, HELP_CHANNEL_PROGRAM },
	{ NULL },
	{ "list", zfs_do_list, HELP_LIST },
	{ NULL },
	{ "set", zfs_do_set, HELP_SET },
	{ "get", zfs_do_get, HELP_GET },
	{ "inherit", zfs_do_inherit, HELP_INHERIT },
	{ "upgrade", zfs_do_upgrade, HELP_UPGRADE },
	{ NULL },
	{ "userspace", zfs_do_userspace, HELP_USERSPACE },
	{ "groupspace", zfs_do_userspace, HELP_GROUPSPACE },
	{ "projectspace", zfs_do_userspace, HELP_PROJECTSPACE },
	{ NULL },
	{ "project", zfs_do_project, HELP_PROJECT },
	{ NULL },
	{ "mount", zfs_do_mount, HELP_MOUNT },
	{ "unmount", zfs_do_unmount, HELP_UNMOUNT },
	{ "share", zfs_do_share, HELP_SHARE },
	{ "unshare", zfs_do_unshare, HELP_UNSHARE },
	{ NULL },
	{ "send", zfs_do_send, HELP_SEND },
	{ "receive", zfs_do_receive, HELP_RECEIVE },
	{ NULL },
	{ "allow", zfs_do_allow, HELP_ALLOW },
	{ NULL },
	{ "unallow", zfs_do_unallow, HELP_UNALLOW },
	{ NULL },
	{ "hold", zfs_do_hold, HELP_HOLD },
	{ "holds", zfs_do_holds, HELP_HOLDS },
	{ "release", zfs_do_release, HELP_RELEASE },
	{ "diff", zfs_do_diff, HELP_DIFF },
	{ "load-key", zfs_do_load_key, HELP_LOAD_KEY },
	{ "unload-key", zfs_do_unload_key, HELP_UNLOAD_KEY },
	{ "change-key", zfs_do_change_key, HELP_CHANGE_KEY },
	{ "redact", zfs_do_redact, HELP_REDACT },
	{ "wait", zfs_do_wait, HELP_WAIT },

	#ifdef __FreeBSD__
	{ "jail", zfs_do_jail, HELP_JAIL },
	{ "unjail", zfs_do_unjail, HELP_UNJAIL },
	#endif

	#ifdef __linux__
	{ "zone", zfs_do_zone, HELP_ZONE },
	{ "unzone", zfs_do_unzone, HELP_UNZONE },
	#endif
	};

	#define NCOMMAND (sizeof (command_table) / sizeof (command_table[0]))

	zfs_command_t *current_command;

	static const char *
	get_usage(zfs_help_t idx)
	{
	switch (idx) {
	case HELP_CLONE:
	return (gettext("\tclone [-p] [-o property=value] ... "
	"<snapshot> <filesystem\|volume>\n"));
	case HELP_CREATE:
	return (gettext("\tcreate [-Pnpuv] [-o property=value] ... "
	"<filesystem>\n"
	"\tcreate [-Pnpsv] [-b blocksize] [-o property=value] ... "
	"-V <size> <volume>\n"));
	case HELP_DESTROY:
	return (gettext("\tdestroy [-fnpRrv] <filesystem\|volume>\n"
	"\tdestroy [-dnpRrv] "
	"<filesystem\|volume>@<snap>[%<snap>][,...]\n"
	"\tdestroy <filesystem\|volume>#<bookmark>\n"));
	case HELP_GET:
	return (gettext("\tget [-rHp] [-d max] "
	"[-o \"all\" \| field[,...]]\n"
	"\t [-t type[,...]] [-s source[,...]]\n"
	"\t <\"all\" \| property[,...]> "
	"[filesystem\|volume\|snapshot\|bookmark] ...\n"));
	case HELP_INHERIT:
	return (gettext("\tinherit [-rS] <property> "
	"<filesystem\|volume\|snapshot> ...\n"));
	case HELP_UPGRADE:
	return (gettext("\tupgrade [-v]\n"
	"\tupgrade [-r] [-V version] <-a \| filesystem ...>\n"));
	case HELP_LIST:
	return (gettext("\tlist [-Hp] [-r\|-d max] [-o property[,...]] "
	"[-s property]...\n\t [-S property]... [-t type[,...]] "
	"[filesystem\|volume\|snapshot] ...\n"));
	case HELP_MOUNT:
	return (gettext("\tmount\n"
	"\tmount [-flvO] [-o opts] <-a\|-R filesystem\|"
	"filesystem>\n"));
	case HELP_PROMOTE:
	return (gettext("\tpromote <clone-filesystem>\n"));
	case HELP_RECEIVE:
	return (gettext("\treceive [-vMnsFhu] "
	"[-o <property>=<value>] ... [-x <property>] ...\n"
	"\t <filesystem\|volume\|snapshot>\n"
	"\treceive [-vMnsFhu] [-o <property>=<value>] ... "
	"[-x <property>] ... \n"
	"\t [-d \| -e] <filesystem>\n"
	"\treceive -A <filesystem\|volume>\n"));
	case HELP_RENAME:
	return (gettext("\trename [-f] <filesystem\|volume\|snapshot> "
	"<filesystem\|volume\|snapshot>\n"
	"\trename -p [-f] <filesystem\|volume> <filesystem\|volume>\n"
	"\trename -u [-f] <filesystem> <filesystem>\n"
	"\trename -r <snapshot> <snapshot>\n"));
	case HELP_ROLLBACK:
	return (gettext("\trollback [-rRf] <snapshot>\n"));
	case HELP_SEND:
	return (gettext("\tsend [-DLPbcehnpsVvw] "
	"[-i\|-I snapshot]\n"
	"\t [-R [-X dataset[,dataset]...]] <snapshot>\n"
	"\tsend [-DnVvPLecw] [-i snapshot\|bookmark] "
	"<filesystem\|volume\|snapshot>\n"
	"\tsend [-DnPpVvLec] [-i bookmark\|snapshot] "
	"--redact <bookmark> <snapshot>\n"
	"\tsend [-nVvPe] -t <receive_resume_token>\n"
	"\tsend [-PnVv] --saved filesystem\n"));
	case HELP_SET:
	return (gettext("\tset [-u] <property=value> ... "
	"<filesystem\|volume\|snapshot> ...\n"));
	case HELP_SHARE:
	return (gettext("\tshare [-l] <-a [nfs\|smb] \| filesystem>\n"));
	case HELP_SNAPSHOT:
	return (gettext("\tsnapshot [-r] [-o property=value] ... "
	"<filesystem\|volume>@<snap> ...\n"));
	case HELP_UNMOUNT:
	return (gettext("\tunmount [-fu] "
	"<-a \| filesystem\|mountpoint>\n"));
	case HELP_UNSHARE:
	return (gettext("\tunshare "
	"<-a [nfs\|smb] \| filesystem\|mountpoint>\n"));
	case HELP_ALLOW:
	return (gettext("\tallow <filesystem\|volume>\n"
	"\tallow [-ldug] "
	"<\"everyone\"\|user\|group>[,...] <perm\|@setname>[,...]\n"
	"\t <filesystem\|volume>\n"
	"\tallow [-ld] -e <perm\|@setname>[,...] "
	"<filesystem\|volume>\n"
	"\tallow -c <perm\|@setname>[,...] <filesystem\|volume>\n"
	"\tallow -s @setname <perm\|@setname>[,...] "
	"<filesystem\|volume>\n"));
	case HELP_UNALLOW:
	return (gettext("\tunallow [-rldug] "
	"<\"everyone\"\|user\|group>[,...]\n"
	"\t [<perm\|@setname>[,...]] <filesystem\|volume>\n"
	"\tunallow [-rld] -e [<perm\|@setname>[,...]] "
	"<filesystem\|volume>\n"
	"\tunallow [-r] -c [<perm\|@setname>[,...]] "
	"<filesystem\|volume>\n"
	"\tunallow [-r] -s @setname [<perm\|@setname>[,...]] "
	"<filesystem\|volume>\n"));
	case HELP_USERSPACE:
	return (gettext("\tuserspace [-Hinp] [-o field[,...]] "
	"[-s field] ...\n"
	"\t [-S field] ... [-t type[,...]] "
	"<filesystem\|snapshot\|path>\n"));
	case HELP_GROUPSPACE:
	return (gettext("\tgroupspace [-Hinp] [-o field[,...]] "
	"[-s field] ...\n"
	"\t [-S field] ... [-t type[,...]] "
	"<filesystem\|snapshot\|path>\n"));
	case HELP_PROJECTSPACE:
	return (gettext("\tprojectspace [-Hp] [-o field[,...]] "
	"[-s field] ... \n"
	"\t [-S field] ... <filesystem\|snapshot\|path>\n"));
	case HELP_PROJECT:
	return (gettext("\tproject [-d\|-r] <directory\|file ...>\n"
	"\tproject -c [-0] [-d\|-r] [-p id] <directory\|file ...>\n"
	"\tproject -C [-k] [-r] <directory ...>\n"
	"\tproject [-p id] [-r] [-s] <directory ...>\n"));
	case HELP_HOLD:
	return (gettext("\thold [-r] <tag> <snapshot> ...\n"));
	case HELP_HOLDS:
	return (gettext("\tholds [-rHp] <snapshot> ...\n"));
	case HELP_RELEASE:
	return (gettext("\trelease [-r] <tag> <snapshot> ...\n"));
	case HELP_DIFF:
	return (gettext("\tdiff [-FHth] <snapshot> "
	"[snapshot\|filesystem]\n"));
	case HELP_BOOKMARK:
	return (gettext("\tbookmark <snapshot\|bookmark> "
	"<newbookmark>\n"));
	case HELP_CHANNEL_PROGRAM:
	return (gettext("\tprogram [-jn] [-t <instruction limit>] "
	"[-m <memory limit (b)>]\n"
	"\t <pool> <program file> [lua args...]\n"));
	case HELP_LOAD_KEY:
	return (gettext("\tload-key [-rn] [-L <keylocation>] "
	"<-a \| filesystem\|volume>\n"));
	case HELP_UNLOAD_KEY:
	return (gettext("\tunload-key [-r] "
	"<-a \| filesystem\|volume>\n"));
	case HELP_CHANGE_KEY:
	return (gettext("\tchange-key [-l] [-o keyformat=<value>]\n"
	"\t [-o keylocation=<value>] [-o pbkdf2iters=<value>]\n"
	"\t <filesystem\|volume>\n"
	"\tchange-key -i [-l] <filesystem\|volume>\n"));
	case HELP_VERSION:
	return (gettext("\tversion\n"));
	case HELP_REDACT:
	return (gettext("\tredact <snapshot> <bookmark> "
	"<redaction_snapshot> ...\n"));
	case HELP_JAIL:
	return (gettext("\tjail <jailid\|jailname> <filesystem>\n"));
	case HELP_UNJAIL:
	return (gettext("\tunjail <jailid\|jailname> <filesystem>\n"));
	case HELP_WAIT:
	return (gettext("\twait [-t <activity>] <filesystem>\n"));
	case HELP_ZONE:
	return (gettext("\tzone <nsfile> <filesystem>\n"));
	case HELP_UNZONE:
	return (gettext("\tunzone <nsfile> <filesystem>\n"));
	default:
	__builtin_unreachable();
	}
	}

	void
	nomem(void)
	{
	(void) fprintf(stderr, gettext("internal error: out of memory\n"));
	exit(1);
	}

	/*
	* Utility function to guarantee malloc() success.
	*/

	void *
	safe_malloc(size_t size)
	{
	void *data;

	if ((data = calloc(1, size)) == NULL)
	nomem();

	return (data);
	}

	static void *
	safe_realloc(void *data, size_t size)
	{
	void *newp;
	if ((newp = realloc(data, size)) == NULL) {
	free(data);
	nomem();
	}

	return (newp);
	}

	static char *
	safe_strdup(const char *str)
	{
	char *dupstr = strdup(str);

	if (dupstr == NULL)
	nomem();

	return (dupstr);
	}

	/*
	* Callback routine that will print out information for each of
	* the properties.
	*/
	static int
	usage_prop_cb(int prop, void *cb)
	{
	FILE *fp = cb;

	(void) fprintf(fp, "\t%-15s ", zfs_prop_to_name(prop));

	if (zfs_prop_readonly(prop))
	(void) fprintf(fp, " NO ");
	else
	(void) fprintf(fp, "YES ");

	if (zfs_prop_inheritable(prop))
	(void) fprintf(fp, " YES ");
	else
	(void) fprintf(fp, " NO ");

	(void) fprintf(fp, "%s\n", zfs_prop_values(prop) ?: "-");

	return (ZPROP_CONT);
	}

	/*
	* Display usage message. If we're inside a command, display only the usage for
	* that command. Otherwise, iterate over the entire command table and display
	* a complete usage message.
	*/
	static __attribute__((noreturn)) void
	usage(boolean_t requested)
	{
	int i;
	boolean_t show_properties = B_FALSE;
	FILE *fp = requested ? stdout : stderr;

	if (current_command == NULL) {

	(void) fprintf(fp, gettext("usage: zfs command args ...\n"));
	(void) fprintf(fp,
	gettext("where 'command' is one of the following:\n\n"));

	for (i = 0; i < NCOMMAND; i++) {
	if (command_table[i].name == NULL)
	(void) fprintf(fp, "\n");
	else
	(void) fprintf(fp, "%s",
	get_usage(command_table[i].usage));
	}

	(void) fprintf(fp, gettext("\nEach dataset is of the form: "
	"pool/[dataset/]*dataset[@name]\n"));
	} else {
	(void) fprintf(fp, gettext("usage:\n"));
	(void) fprintf(fp, "%s", get_usage(current_command->usage));
	}

	if (current_command != NULL &&
	(strcmp(current_command->name, "set") == 0 \|\|
	strcmp(current_command->name, "get") == 0 \|\|
	strcmp(current_command->name, "inherit") == 0 \|\|
	strcmp(current_command->name, "list") == 0))
	show_properties = B_TRUE;

	if (show_properties) {
	(void) fprintf(fp, "%s",
	gettext("\nThe following properties are supported:\n"));

	(void) fprintf(fp, "\n\t%-14s %s %s %s\n\n",
	"PROPERTY", "EDIT", "INHERIT", "VALUES");

	/* Iterate over all properties */
	(void) zprop_iter(usage_prop_cb, fp, B_FALSE, B_TRUE,
	ZFS_TYPE_DATASET);

	(void) fprintf(fp, "\t%-15s ", "userused@...");
	(void) fprintf(fp, " NO NO <size>\n");
	(void) fprintf(fp, "\t%-15s ", "groupused@...");
	(void) fprintf(fp, " NO NO <size>\n");
	(void) fprintf(fp, "\t%-15s ", "projectused@...");
	(void) fprintf(fp, " NO NO <size>\n");
	(void) fprintf(fp, "\t%-15s ", "userobjused@...");
	(void) fprintf(fp, " NO NO <size>\n");
	(void) fprintf(fp, "\t%-15s ", "groupobjused@...");
	(void) fprintf(fp, " NO NO <size>\n");
	(void) fprintf(fp, "\t%-15s ", "projectobjused@...");
	(void) fprintf(fp, " NO NO <size>\n");
	(void) fprintf(fp, "\t%-15s ", "userquota@...");
	(void) fprintf(fp, "YES NO <size> \| none\n");
	(void) fprintf(fp, "\t%-15s ", "groupquota@...");
	(void) fprintf(fp, "YES NO <size> \| none\n");
	(void) fprintf(fp, "\t%-15s ", "projectquota@...");
	(void) fprintf(fp, "YES NO <size> \| none\n");
	(void) fprintf(fp, "\t%-15s ", "userobjquota@...");
	(void) fprintf(fp, "YES NO <size> \| none\n");
	(void) fprintf(fp, "\t%-15s ", "groupobjquota@...");
	(void) fprintf(fp, "YES NO <size> \| none\n");
	(void) fprintf(fp, "\t%-15s ", "projectobjquota@...");
	(void) fprintf(fp, "YES NO <size> \| none\n");
	(void) fprintf(fp, "\t%-15s ", "written@<snap>");
	(void) fprintf(fp, " NO NO <size>\n");
	(void) fprintf(fp, "\t%-15s ", "written#<bookmark>");
	(void) fprintf(fp, " NO NO <size>\n");

	(void) fprintf(fp, gettext("\nSizes are specified in bytes "
	"with standard units such as K, M, G, etc.\n"));
	(void) fprintf(fp, "%s", gettext("\nUser-defined properties "
	"can be specified by using a name containing a colon "
	"(:).\n"));
	(void) fprintf(fp, gettext("\nThe {user\|group\|project}"
	"[obj]{used\|quota}@ properties must be appended with\n"
	"a user\|group\|project specifier of one of these forms:\n"
	" POSIX name (eg: \"matt\")\n"
	" POSIX id (eg: \"126829\")\n"
	" SMB name@domain (eg: \"matt@sun\")\n"
	" SMB SID (eg: \"S-1-234-567-89\")\n"));
	} else {
	(void) fprintf(fp,
	gettext("\nFor the property list, run: %s\n"),
	"zfs set\|get");
	(void) fprintf(fp,
	gettext("\nFor the delegated permission list, run: %s\n"),
	"zfs allow\|unallow");
	(void) fprintf(fp,
	gettext("\nFor further help on a command or topic, "
	"run: %s\n"), "zfs help [<topic>]");
	}

	/*
	* See comments at end of main().
	*/
	if (getenv("ZFS_ABORT") != NULL) {
	(void) printf("dumping core by request\n");
	abort();
	}

	exit(requested ? 0 : 2);
	}

	/*
	* Take a property=value argument string and add it to the given nvlist.
	* Modifies the argument inplace.
	*/
	static boolean_t
	parseprop(nvlist_t props, char propname)
	{
	char *propval;

	if ((propval = strchr(propname, '=')) == NULL) {
	(void) fprintf(stderr, gettext("missing "
	"'=' for property=value argument\n"));
	return (B_FALSE);
	}
	*propval = '\0';
	propval++;
	if (nvlist_exists(props, propname)) {
	(void) fprintf(stderr, gettext("property '%s' "
	"specified multiple times\n"), propname);
	return (B_FALSE);
	}
	if (nvlist_add_string(props, propname, propval) != 0)
	nomem();
	return (B_TRUE);
	}

	/*
	* Take a property name argument and add it to the given nvlist.
	* Modifies the argument inplace.
	*/
	static boolean_t
	parsepropname(nvlist_t props, char propname)
	{
	if (strchr(propname, '=') != NULL) {
	(void) fprintf(stderr, gettext("invalid character "
	"'=' in property argument\n"));
	return (B_FALSE);
	}
	if (nvlist_exists(props, propname)) {
	(void) fprintf(stderr, gettext("property '%s' "
	"specified multiple times\n"), propname);
	return (B_FALSE);
	}
	if (nvlist_add_boolean(props, propname) != 0)
	nomem();
	return (B_TRUE);
	}

	static int
	parse_depth(char opt, int flags)
	{
	char *tmp;
	int depth;

	depth = (int)strtol(opt, &tmp, 0);
	if (*tmp) {
	(void) fprintf(stderr,
	gettext("%s is not an integer\n"), optarg);
	usage(B_FALSE);
	}
	if (depth < 0) {
	(void) fprintf(stderr,
	gettext("Depth can not be negative.\n"));
	usage(B_FALSE);
	}
	*flags \|= (ZFS_ITER_DEPTH_LIMIT\|ZFS_ITER_RECURSE);
	return (depth);
	}

	#define PROGRESS_DELAY 2 /* seconds */

	static const char *pt_reverse =
	"\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b";
	static time_t pt_begin;
	static char *pt_header = NULL;
	static boolean_t pt_shown;

	static void
	start_progress_timer(void)
	{
	pt_begin = time(NULL) + PROGRESS_DELAY;
	pt_shown = B_FALSE;
	}

	static void
	set_progress_header(const char *header)
	{
	assert(pt_header == NULL);
	pt_header = safe_strdup(header);
	if (pt_shown) {
	(void) printf("%s: ", header);
	(void) fflush(stdout);
	}
	}

	static void
	update_progress(const char *update)
	{
	if (!pt_shown && time(NULL) > pt_begin) {
	int len = strlen(update);

	(void) printf("%s: %s%.s", pt_header, update, len, len,
	pt_reverse);
	(void) fflush(stdout);
	pt_shown = B_TRUE;
	} else if (pt_shown) {
	int len = strlen(update);

	(void) printf("%s%.s", update, len, len, pt_reverse);
	(void) fflush(stdout);
	}
	}

	static void
	finish_progress(const char *done)
	{
	if (pt_shown) {
	(void) puts(done);
	(void) fflush(stdout);
	}
	free(pt_header);
	pt_header = NULL;
	}

	static int
	zfs_mount_and_share(libzfs_handle_t hdl, const char dataset, zfs_type_t type)
	{
	zfs_handle_t *zhp = NULL;
	int ret = 0;

	zhp = zfs_open(hdl, dataset, type);
	if (zhp == NULL)
	return (1);

	/*
	* Volumes may neither be mounted or shared. Potentially in the
	* future filesystems detected on these volumes could be mounted.
	*/
	if (zfs_get_type(zhp) == ZFS_TYPE_VOLUME) {
	zfs_close(zhp);
	return (0);
	}

	/*
	* Mount and/or share the new filesystem as appropriate. We provide a
	* verbose error message to let the user know that their filesystem was
	* in fact created, even if we failed to mount or share it.
	*
	* If the user doesn't want the dataset automatically mounted, then
	* skip the mount/share step
	*/
	if (zfs_prop_valid_for_type(ZFS_PROP_CANMOUNT, type, B_FALSE) &&
	zfs_prop_get_int(zhp, ZFS_PROP_CANMOUNT) == ZFS_CANMOUNT_ON) {
	if (zfs_mount_delegation_check()) {
	(void) fprintf(stderr, gettext("filesystem "
	"successfully created, but it may only be "
	"mounted by root\n"));
	ret = 1;
	} else if (zfs_mount(zhp, NULL, 0) != 0) {
	(void) fprintf(stderr, gettext("filesystem "
	"successfully created, but not mounted\n"));
	ret = 1;
	} else if (zfs_share(zhp, NULL) != 0) {
	(void) fprintf(stderr, gettext("filesystem "
	"successfully created, but not shared\n"));
	ret = 1;
	}
	zfs_commit_shares(NULL);
	}

	zfs_close(zhp);

	return (ret);
	}

	/*
	* zfs clone [-p] [-o prop=value] ... <snap> <fs \| vol>
	*
	* Given an existing dataset, create a writable copy whose initial contents
	* are the same as the source. The newly created dataset maintains a
	* dependency on the original; the original cannot be destroyed so long as
	* the clone exists.
	*
	* The '-p' flag creates all the non-existing ancestors of the target first.
	*/
	static int
	zfs_do_clone(int argc, char **argv)
	{
	zfs_handle_t *zhp = NULL;
	boolean_t parents = B_FALSE;
	nvlist_t *props;
	int ret = 0;
	int c;

	if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0)
	nomem();

	/* check options */
	while ((c = getopt(argc, argv, "o:p")) != -1) {
	switch (c) {
	case 'o':
	if (!parseprop(props, optarg)) {
	nvlist_free(props);
	return (1);
	}
	break;
	case 'p':
	parents = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	goto usage;
	}
	}

	argc -= optind;
	argv += optind;

	/* check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing source dataset "
	"argument\n"));
	goto usage;
	}
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing target dataset "
	"argument\n"));
	goto usage;
	}
	if (argc > 2) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	goto usage;
	}

	/* open the source dataset */
	if ((zhp = zfs_open(g_zfs, argv[0], ZFS_TYPE_SNAPSHOT)) == NULL) {
	nvlist_free(props);
	return (1);
	}

	if (parents && zfs_name_valid(argv[1], ZFS_TYPE_FILESYSTEM \|
	ZFS_TYPE_VOLUME)) {
	/*
	* Now create the ancestors of the target dataset. If the
	* target already exists and '-p' option was used we should not
	* complain.
	*/
	if (zfs_dataset_exists(g_zfs, argv[1], ZFS_TYPE_FILESYSTEM \|
	ZFS_TYPE_VOLUME)) {
	zfs_close(zhp);
	nvlist_free(props);
	return (0);
	}
	if (zfs_create_ancestors(g_zfs, argv[1]) != 0) {
	zfs_close(zhp);
	nvlist_free(props);
	return (1);
	}
	}

	/* pass to libzfs */
	ret = zfs_clone(zhp, argv[1], props);

	/* create the mountpoint if necessary */
	if (ret == 0) {
	if (log_history) {
	(void) zpool_log_history(g_zfs, history_str);
	log_history = B_FALSE;
	}

	ret = zfs_mount_and_share(g_zfs, argv[1], ZFS_TYPE_DATASET);
	}

	zfs_close(zhp);
	nvlist_free(props);

	return (!!ret);

	usage:
	ASSERT3P(zhp, ==, NULL);
	nvlist_free(props);
	usage(B_FALSE);
	return (-1);
	}

	/*
	* Return a default volblocksize for the pool which always uses more than
	* half of the data sectors. This primarily applies to dRAID which always
	* writes full stripe widths.
	*/
	static uint64_t
	default_volblocksize(zpool_handle_t zhp, nvlist_t props)
	{
	uint64_t volblocksize, asize = SPA_MINBLOCKSIZE;
	nvlist_t tree, *vdevs;
	uint_t nvdevs;

	nvlist_t *config = zpool_get_config(zhp, NULL);

	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree) != 0 \|\|
	nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN,
	&vdevs, &nvdevs) != 0) {
	return (ZVOL_DEFAULT_BLOCKSIZE);
	}

	for (int i = 0; i < nvdevs; i++) {
	nvlist_t *nv = vdevs[i];
	uint64_t ashift, ndata, nparity;

	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ASHIFT, &ashift) != 0)
	continue;

	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_DRAID_NDATA,
	&ndata) == 0) {
	/* dRAID minimum allocation width */
	asize = MAX(asize, ndata * (1ULL << ashift));
	} else if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY,
	&nparity) == 0) {
	/* raidz minimum allocation width */
	if (nparity == 1)
	asize = MAX(asize, 2 * (1ULL << ashift));
	else
	asize = MAX(asize, 4 * (1ULL << ashift));
	} else {
	/* mirror or (non-redundant) leaf vdev */
	asize = MAX(asize, 1ULL << ashift);
	}
	}

	/*
	* Calculate the target volblocksize such that more than half
	* of the asize is used. The following table is for 4k sectors.
	*
	* n asize blksz used \| n asize blksz used
	* -------------------------+---------------------------------
	* 1 4,096 8,192 100% \| 9 36,864 32,768 88%
	* 2 8,192 8,192 100% \| 10 40,960 32,768 80%
	* 3 12,288 8,192 66% \| 11 45,056 32,768 72%
	* 4 16,384 16,384 100% \| 12 49,152 32,768 66%
	* 5 20,480 16,384 80% \| 13 53,248 32,768 61%
	* 6 24,576 16,384 66% \| 14 57,344 32,768 57%
	* 7 28,672 16,384 57% \| 15 61,440 32,768 53%
	* 8 32,768 32,768 100% \| 16 65,536 65,636 100%
	*
	* This is primarily a concern for dRAID which always allocates
	* a full stripe width. For dRAID the default stripe width is
	* n=8 in which case the volblocksize is set to 32k. Ignoring
	* compression there are no unused sectors. This same reasoning
	* applies to raidz[2,3] so target 4 sectors to minimize waste.
	*/
	uint64_t tgt_volblocksize = ZVOL_DEFAULT_BLOCKSIZE;
	while (tgt_volblocksize * 2 <= asize)
	tgt_volblocksize *= 2;

	const char *prop = zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE);
	if (nvlist_lookup_uint64(props, prop, &volblocksize) == 0) {

	/* Issue a warning when a non-optimal size is requested. */
	if (volblocksize < ZVOL_DEFAULT_BLOCKSIZE) {
	(void) fprintf(stderr, gettext("Warning: "
	"volblocksize (%llu) is less than the default "
	"minimum block size (%llu).\nTo reduce wasted "
	"space a volblocksize of %llu is recommended.\n"),
	(u_longlong_t)volblocksize,
	(u_longlong_t)ZVOL_DEFAULT_BLOCKSIZE,
	(u_longlong_t)tgt_volblocksize);
	} else if (volblocksize < tgt_volblocksize) {
	(void) fprintf(stderr, gettext("Warning: "
	"volblocksize (%llu) is much less than the "
	"minimum allocation\nunit (%llu), which wastes "
	"at least %llu%% of space. To reduce wasted "
	"space,\nuse a larger volblocksize (%llu is "
	"recommended), fewer dRAID data disks\n"
	"per group, or smaller sector size (ashift).\n"),
	(u_longlong_t)volblocksize, (u_longlong_t)asize,
	(u_longlong_t)((100 * (asize - volblocksize)) /
	asize), (u_longlong_t)tgt_volblocksize);
	}
	} else {
	volblocksize = tgt_volblocksize;
	fnvlist_add_uint64(props, prop, volblocksize);
	}

	return (volblocksize);
	}

	/*
	* zfs create [-Pnpv] [-o prop=value] ... fs
	* zfs create [-Pnpsv] [-b blocksize] [-o prop=value] ... -V vol size
	*
	* Create a new dataset. This command can be used to create filesystems
	* and volumes. Snapshot creation is handled by 'zfs snapshot'.
	* For volumes, the user must specify a size to be used.
	*
	* The '-s' flag applies only to volumes, and indicates that we should not try
	* to set the reservation for this volume. By default we set a reservation
	* equal to the size for any volume. For pools with SPA_VERSION >=
	* SPA_VERSION_REFRESERVATION, we set a refreservation instead.
	*
	* The '-p' flag creates all the non-existing ancestors of the target first.
	*
	* The '-n' flag is no-op (dry run) mode. This will perform a user-space sanity
	* check of arguments and properties, but does not check for permissions,
	* available space, etc.
	*
	* The '-u' flag prevents the newly created file system from being mounted.
	*
	* The '-v' flag is for verbose output.
	*
	* The '-P' flag is used for parseable output. It implies '-v'.
	*/
	static int
	zfs_do_create(int argc, char **argv)
	{
	zfs_type_t type = ZFS_TYPE_FILESYSTEM;
	zpool_handle_t *zpool_handle = NULL;
	nvlist_t *real_props = NULL;
	uint64_t volsize = 0;
	int c;
	boolean_t noreserve = B_FALSE;
	boolean_t bflag = B_FALSE;
	boolean_t parents = B_FALSE;
	boolean_t dryrun = B_FALSE;
	boolean_t nomount = B_FALSE;
	boolean_t verbose = B_FALSE;
	boolean_t parseable = B_FALSE;
	int ret = 1;
	nvlist_t *props;
	uint64_t intval;
	const char *strval;

	if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0)
	nomem();

	/* check options */
	while ((c = getopt(argc, argv, ":PV:b:nso:puv")) != -1) {
	switch (c) {
	case 'V':
	type = ZFS_TYPE_VOLUME;
	if (zfs_nicestrtonum(g_zfs, optarg, &intval) != 0) {
	(void) fprintf(stderr, gettext("bad volume "
	"size '%s': %s\n"), optarg,
	libzfs_error_description(g_zfs));
	goto error;
	}

	if (nvlist_add_uint64(props,
	zfs_prop_to_name(ZFS_PROP_VOLSIZE), intval) != 0)
	nomem();
	volsize = intval;
	break;
	case 'P':
	verbose = B_TRUE;
	parseable = B_TRUE;
	break;
	case 'p':
	parents = B_TRUE;
	break;
	case 'b':
	bflag = B_TRUE;
	if (zfs_nicestrtonum(g_zfs, optarg, &intval) != 0) {
	(void) fprintf(stderr, gettext("bad volume "
	"block size '%s': %s\n"), optarg,
	libzfs_error_description(g_zfs));
	goto error;
	}

	if (nvlist_add_uint64(props,
	zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
	intval) != 0)
	nomem();
	break;
	case 'n':
	dryrun = B_TRUE;
	break;
	case 'o':
	if (!parseprop(props, optarg))
	goto error;
	break;
	case 's':
	noreserve = B_TRUE;
	break;
	case 'u':
	nomount = B_TRUE;
	break;
	case 'v':
	verbose = B_TRUE;
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing size "
	"argument\n"));
	goto badusage;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	goto badusage;
	}
	}

	if ((bflag \|\| noreserve) && type != ZFS_TYPE_VOLUME) {
	(void) fprintf(stderr, gettext("'-s' and '-b' can only be "
	"used when creating a volume\n"));
	goto badusage;
	}
	if (nomount && type != ZFS_TYPE_FILESYSTEM) {
	(void) fprintf(stderr, gettext("'-u' can only be "
	"used when creating a filesystem\n"));
	goto badusage;
	}

	argc -= optind;
	argv += optind;

	/* check number of arguments */
	if (argc == 0) {
	(void) fprintf(stderr, gettext("missing %s argument\n"),
	zfs_type_to_name(type));
	goto badusage;
	}
	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	goto badusage;
	}

	if (dryrun \|\| type == ZFS_TYPE_VOLUME) {
	char msg[ZFS_MAX_DATASET_NAME_LEN * 2];
	char *p;

	if ((p = strchr(argv[0], '/')) != NULL)
	*p = '\0';
	zpool_handle = zpool_open(g_zfs, argv[0]);
	if (p != NULL)
	*p = '/';
	if (zpool_handle == NULL)
	goto error;

	(void) snprintf(msg, sizeof (msg),
	dryrun ? gettext("cannot verify '%s'") :
	gettext("cannot create '%s'"), argv[0]);
	if (props && (real_props = zfs_valid_proplist(g_zfs, type,
	props, 0, NULL, zpool_handle, B_TRUE, msg)) == NULL) {
	zpool_close(zpool_handle);
	goto error;
	}
	}

	if (type == ZFS_TYPE_VOLUME) {
	const char *prop = zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE);
	uint64_t volblocksize = default_volblocksize(zpool_handle,
	real_props);

	if (volblocksize != ZVOL_DEFAULT_BLOCKSIZE &&
	nvlist_lookup_string(props, prop, &strval) != 0) {
	char *tmp;
	if (asprintf(&tmp, "%llu",
	(u_longlong_t)volblocksize) == -1)
	nomem();
	nvlist_add_string(props, prop, tmp);
	free(tmp);
	}

	/*
	* If volsize is not a multiple of volblocksize, round it
	* up to the nearest multiple of the volblocksize.
	*/
	if (volsize % volblocksize) {
	volsize = P2ROUNDUP_TYPED(volsize, volblocksize,
	uint64_t);

	if (nvlist_add_uint64(props,
	zfs_prop_to_name(ZFS_PROP_VOLSIZE), volsize) != 0) {
	nvlist_free(props);
	nomem();
	}
	}
	}

	if (type == ZFS_TYPE_VOLUME && !noreserve) {
	uint64_t spa_version;
	zfs_prop_t resv_prop;

	spa_version = zpool_get_prop_int(zpool_handle,
	ZPOOL_PROP_VERSION, NULL);
	if (spa_version >= SPA_VERSION_REFRESERVATION)
	resv_prop = ZFS_PROP_REFRESERVATION;
	else
	resv_prop = ZFS_PROP_RESERVATION;

	volsize = zvol_volsize_to_reservation(zpool_handle, volsize,
	real_props);

	if (nvlist_lookup_string(props, zfs_prop_to_name(resv_prop),
	&strval) != 0) {
	if (nvlist_add_uint64(props,
	zfs_prop_to_name(resv_prop), volsize) != 0) {
	nvlist_free(props);
	nomem();
	}
	}
	}
	if (zpool_handle != NULL) {
	zpool_close(zpool_handle);
	nvlist_free(real_props);
	}

	if (parents && zfs_name_valid(argv[0], type)) {
	/*
	* Now create the ancestors of target dataset. If the target
	* already exists and '-p' option was used we should not
	* complain.
	*/
	if (zfs_dataset_exists(g_zfs, argv[0], type)) {
	ret = 0;
	goto error;
	}
	if (verbose) {
	(void) printf(parseable ? "create_ancestors\t%s\n" :
	dryrun ? "would create ancestors of %s\n" :
	"create ancestors of %s\n", argv[0]);
	}
	if (!dryrun) {
	if (zfs_create_ancestors(g_zfs, argv[0]) != 0) {
	goto error;
	}
	}
	}

	if (verbose) {
	nvpair_t *nvp = NULL;
	(void) printf(parseable ? "create\t%s\n" :
	dryrun ? "would create %s\n" : "create %s\n", argv[0]);
	while ((nvp = nvlist_next_nvpair(props, nvp)) != NULL) {
	uint64_t uval;
	const char *sval;

	switch (nvpair_type(nvp)) {
	case DATA_TYPE_UINT64:
	VERIFY0(nvpair_value_uint64(nvp, &uval));
	(void) printf(parseable ?
	"property\t%s\t%llu\n" : "\t%s=%llu\n",
	nvpair_name(nvp), (u_longlong_t)uval);
	break;
	case DATA_TYPE_STRING:
	VERIFY0(nvpair_value_string(nvp, &sval));
	(void) printf(parseable ?
	"property\t%s\t%s\n" : "\t%s=%s\n",
	nvpair_name(nvp), sval);
	break;
	default:
	(void) fprintf(stderr, "property '%s' "
	"has illegal type %d\n",
	nvpair_name(nvp), nvpair_type(nvp));
	abort();
	}
	}
	}
	if (dryrun) {
	ret = 0;
	goto error;
	}

	/* pass to libzfs */
	if (zfs_create(g_zfs, argv[0], type, props) != 0)
	goto error;

	if (log_history) {
	(void) zpool_log_history(g_zfs, history_str);
	log_history = B_FALSE;
	}

	if (nomount) {
	ret = 0;
	goto error;
	}

	ret = zfs_mount_and_share(g_zfs, argv[0], ZFS_TYPE_DATASET);
	error:
	nvlist_free(props);
	return (ret);
	badusage:
	nvlist_free(props);
	usage(B_FALSE);
	return (2);
	}

	/*
	* zfs destroy [-rRf] <fs, vol>
	* zfs destroy [-rRd] <snap>
	*
	* -r Recursively destroy all children
	* -R Recursively destroy all dependents, including clones
	* -f Force unmounting of any dependents
	* -d If we can't destroy now, mark for deferred destruction
	*
	* Destroys the given dataset. By default, it will unmount any filesystems,
	* and refuse to destroy a dataset that has any dependents. A dependent can
	* either be a child, or a clone of a child.
	*/
	typedef struct destroy_cbdata {
	boolean_t cb_first;
	boolean_t cb_force;
	boolean_t cb_recurse;
	boolean_t cb_error;
	boolean_t cb_doclones;
	zfs_handle_t *cb_target;
	boolean_t cb_defer_destroy;
	boolean_t cb_verbose;
	boolean_t cb_parsable;
	boolean_t cb_dryrun;
	nvlist_t *cb_nvl;
	nvlist_t *cb_batchedsnaps;

	/* first snap in contiguous run */
	char *cb_firstsnap;
	/* previous snap in contiguous run */
	char *cb_prevsnap;
	int64_t cb_snapused;
	char *cb_snapspec;
	char *cb_bookmark;
	uint64_t cb_snap_count;
	} destroy_cbdata_t;

	/*
	* Check for any dependents based on the '-r' or '-R' flags.
	*/
	static int
	destroy_check_dependent(zfs_handle_t zhp, void data)
	{
	destroy_cbdata_t *cbp = data;
	const char *tname = zfs_get_name(cbp->cb_target);
	const char *name = zfs_get_name(zhp);

	if (strncmp(tname, name, strlen(tname)) == 0 &&
	(name[strlen(tname)] == '/' \|\| name[strlen(tname)] == '@')) {
	/*
	* This is a direct descendant, not a clone somewhere else in
	* the hierarchy.
	*/
	if (cbp->cb_recurse)
	goto out;

	if (cbp->cb_first) {
	(void) fprintf(stderr, gettext("cannot destroy '%s': "
	"%s has children\n"),
	zfs_get_name(cbp->cb_target),
	zfs_type_to_name(zfs_get_type(cbp->cb_target)));
	(void) fprintf(stderr, gettext("use '-r' to destroy "
	"the following datasets:\n"));
	cbp->cb_first = B_FALSE;
	cbp->cb_error = B_TRUE;
	}

	(void) fprintf(stderr, "%s\n", zfs_get_name(zhp));
	} else {
	/*
	* This is a clone. We only want to report this if the '-r'
	* wasn't specified, or the target is a snapshot.
	*/
	if (!cbp->cb_recurse &&
	zfs_get_type(cbp->cb_target) != ZFS_TYPE_SNAPSHOT)
	goto out;

	if (cbp->cb_first) {
	(void) fprintf(stderr, gettext("cannot destroy '%s': "
	"%s has dependent clones\n"),
	zfs_get_name(cbp->cb_target),
	zfs_type_to_name(zfs_get_type(cbp->cb_target)));
	(void) fprintf(stderr, gettext("use '-R' to destroy "
	"the following datasets:\n"));
	cbp->cb_first = B_FALSE;
	cbp->cb_error = B_TRUE;
	cbp->cb_dryrun = B_TRUE;
	}

	(void) fprintf(stderr, "%s\n", zfs_get_name(zhp));
	}

	out:
	zfs_close(zhp);
	return (0);
	}

	static int
	destroy_batched(destroy_cbdata_t *cb)
	{
	int error = zfs_destroy_snaps_nvl(g_zfs,
	cb->cb_batchedsnaps, B_FALSE);
	fnvlist_free(cb->cb_batchedsnaps);
	cb->cb_batchedsnaps = fnvlist_alloc();
	return (error);
	}

	static int
	destroy_callback(zfs_handle_t zhp, void data)
	{
	destroy_cbdata_t *cb = data;
	const char *name = zfs_get_name(zhp);
	int error;

	if (cb->cb_verbose) {
	if (cb->cb_parsable) {
	(void) printf("destroy\t%s\n", name);
	} else if (cb->cb_dryrun) {
	(void) printf(gettext("would destroy %s\n"),
	name);
	} else {
	(void) printf(gettext("will destroy %s\n"),
	name);
	}
	}

	/*
	* Ignore pools (which we've already flagged as an error before getting
	* here).
	*/
	if (strchr(zfs_get_name(zhp), '/') == NULL &&
	zfs_get_type(zhp) == ZFS_TYPE_FILESYSTEM) {
	zfs_close(zhp);
	return (0);
	}
	if (cb->cb_dryrun) {
	zfs_close(zhp);
	return (0);
	}

	/*
	* We batch up all contiguous snapshots (even of different
	* filesystems) and destroy them with one ioctl. We can't
	* simply do all snap deletions and then all fs deletions,
	* because we must delete a clone before its origin.
	*/
	if (zfs_get_type(zhp) == ZFS_TYPE_SNAPSHOT) {
	cb->cb_snap_count++;
	fnvlist_add_boolean(cb->cb_batchedsnaps, name);
	if (cb->cb_snap_count % 10 == 0 && cb->cb_defer_destroy) {
	error = destroy_batched(cb);
	if (error != 0) {
	zfs_close(zhp);
	return (-1);
	}
	}
	} else {
	error = destroy_batched(cb);
	if (error != 0 \|\|
	zfs_unmount(zhp, NULL, cb->cb_force ? MS_FORCE : 0) != 0 \|\|
	zfs_destroy(zhp, cb->cb_defer_destroy) != 0) {
	zfs_close(zhp);
	/*
	* When performing a recursive destroy we ignore errors
	* so that the recursive destroy could continue
	* destroying past problem datasets
	*/
	if (cb->cb_recurse) {
	cb->cb_error = B_TRUE;
	return (0);
	}
	return (-1);
	}
	}

	zfs_close(zhp);
	return (0);
	}

	static int
	destroy_print_cb(zfs_handle_t zhp, void arg)
	{
	destroy_cbdata_t *cb = arg;
	const char *name = zfs_get_name(zhp);
	int err = 0;

	if (nvlist_exists(cb->cb_nvl, name)) {
	if (cb->cb_firstsnap == NULL)
	cb->cb_firstsnap = strdup(name);
	if (cb->cb_prevsnap != NULL)
	free(cb->cb_prevsnap);
	/* this snap continues the current range */
	cb->cb_prevsnap = strdup(name);
	if (cb->cb_firstsnap == NULL \|\| cb->cb_prevsnap == NULL)
	nomem();
	if (cb->cb_verbose) {
	if (cb->cb_parsable) {
	(void) printf("destroy\t%s\n", name);
	} else if (cb->cb_dryrun) {
	(void) printf(gettext("would destroy %s\n"),
	name);
	} else {
	(void) printf(gettext("will destroy %s\n"),
	name);
	}
	}
	} else if (cb->cb_firstsnap != NULL) {
	/* end of this range */
	uint64_t used = 0;
	err = lzc_snaprange_space(cb->cb_firstsnap,
	cb->cb_prevsnap, &used);
	cb->cb_snapused += used;
	free(cb->cb_firstsnap);
	cb->cb_firstsnap = NULL;
	free(cb->cb_prevsnap);
	cb->cb_prevsnap = NULL;
	}
	zfs_close(zhp);
	return (err);
	}

	static int
	destroy_print_snapshots(zfs_handle_t fs_zhp, destroy_cbdata_t cb)
	{
	int err;
	assert(cb->cb_firstsnap == NULL);
	assert(cb->cb_prevsnap == NULL);
	err = zfs_iter_snapshots_sorted_v2(fs_zhp, 0, destroy_print_cb, cb, 0,
	0);
	if (cb->cb_firstsnap != NULL) {
	uint64_t used = 0;
	if (err == 0) {
	err = lzc_snaprange_space(cb->cb_firstsnap,
	cb->cb_prevsnap, &used);
	}
	cb->cb_snapused += used;
	free(cb->cb_firstsnap);
	cb->cb_firstsnap = NULL;
	free(cb->cb_prevsnap);
	cb->cb_prevsnap = NULL;
	}
	return (err);
	}

	static int
	snapshot_to_nvl_cb(zfs_handle_t zhp, void arg)
	{
	destroy_cbdata_t *cb = arg;
	int err = 0;

	/* Check for clones. */
	if (!cb->cb_doclones && !cb->cb_defer_destroy) {
	cb->cb_target = zhp;
	cb->cb_first = B_TRUE;
	err = zfs_iter_dependents_v2(zhp, 0, B_TRUE,
	destroy_check_dependent, cb);
	}

	if (err == 0) {
	if (nvlist_add_boolean(cb->cb_nvl, zfs_get_name(zhp)))
	nomem();
	}
	zfs_close(zhp);
	return (err);
	}

	static int
	gather_snapshots(zfs_handle_t zhp, void arg)
	{
	destroy_cbdata_t *cb = arg;
	int err = 0;

	err = zfs_iter_snapspec_v2(zhp, 0, cb->cb_snapspec,
	snapshot_to_nvl_cb, cb);
	if (err == ENOENT)
	err = 0;
	if (err != 0)
	goto out;

	if (cb->cb_verbose) {
	err = destroy_print_snapshots(zhp, cb);
	if (err != 0)
	goto out;
	}

	if (cb->cb_recurse)
	err = zfs_iter_filesystems_v2(zhp, 0, gather_snapshots, cb);

	out:
	zfs_close(zhp);
	return (err);
	}

	static int
	destroy_clones(destroy_cbdata_t *cb)
	{
	nvpair_t *pair;
	for (pair = nvlist_next_nvpair(cb->cb_nvl, NULL);
	pair != NULL;
	pair = nvlist_next_nvpair(cb->cb_nvl, pair)) {
	zfs_handle_t *zhp = zfs_open(g_zfs, nvpair_name(pair),
	ZFS_TYPE_SNAPSHOT);
	if (zhp != NULL) {
	boolean_t defer = cb->cb_defer_destroy;
	int err;

	/*
	* We can't defer destroy non-snapshots, so set it to
	* false while destroying the clones.
	*/
	cb->cb_defer_destroy = B_FALSE;
	err = zfs_iter_dependents_v2(zhp, 0, B_FALSE,
	destroy_callback, cb);
	cb->cb_defer_destroy = defer;
	zfs_close(zhp);
	if (err != 0)
	return (err);
	}
	}
	return (0);
	}

	static int
	zfs_do_destroy(int argc, char **argv)
	{
	destroy_cbdata_t cb = { 0 };
	int rv = 0;
	int err = 0;
	int c;
	zfs_handle_t *zhp = NULL;
	char at, pound;
	zfs_type_t type = ZFS_TYPE_DATASET;

	/* check options */
	while ((c = getopt(argc, argv, "vpndfrR")) != -1) {
	switch (c) {
	case 'v':
	cb.cb_verbose = B_TRUE;
	break;
	case 'p':
	cb.cb_verbose = B_TRUE;
	cb.cb_parsable = B_TRUE;
	break;
	case 'n':
	cb.cb_dryrun = B_TRUE;
	break;
	case 'd':
	cb.cb_defer_destroy = B_TRUE;
	type = ZFS_TYPE_SNAPSHOT;
	break;
	case 'f':
	cb.cb_force = B_TRUE;
	break;
	case 'r':
	cb.cb_recurse = B_TRUE;
	break;
	case 'R':
	cb.cb_recurse = B_TRUE;
	cb.cb_doclones = B_TRUE;
	break;
	case '?':
	default:
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* check number of arguments */
	if (argc == 0) {
	(void) fprintf(stderr, gettext("missing dataset argument\n"));
	usage(B_FALSE);
	}
	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	at = strchr(argv[0], '@');
	pound = strchr(argv[0], '#');
	if (at != NULL) {

	/* Build the list of snaps to destroy in cb_nvl. */
	cb.cb_nvl = fnvlist_alloc();

	*at = '\0';
	zhp = zfs_open(g_zfs, argv[0],
	ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME);
	if (zhp == NULL) {
	nvlist_free(cb.cb_nvl);
	return (1);
	}

	cb.cb_snapspec = at + 1;
	if (gather_snapshots(zfs_handle_dup(zhp), &cb) != 0 \|\|
	cb.cb_error) {
	rv = 1;
	goto out;
	}

	if (nvlist_empty(cb.cb_nvl)) {
	(void) fprintf(stderr, gettext("could not find any "
	"snapshots to destroy; check snapshot names.\n"));
	rv = 1;
	goto out;
	}

	if (cb.cb_verbose) {
	char buf[16];
	zfs_nicebytes(cb.cb_snapused, buf, sizeof (buf));
	if (cb.cb_parsable) {
	(void) printf("reclaim\t%llu\n",
	(u_longlong_t)cb.cb_snapused);
	} else if (cb.cb_dryrun) {
	(void) printf(gettext("would reclaim %s\n"),
	buf);
	} else {
	(void) printf(gettext("will reclaim %s\n"),
	buf);
	}
	}

	if (!cb.cb_dryrun) {
	if (cb.cb_doclones) {
	cb.cb_batchedsnaps = fnvlist_alloc();
	err = destroy_clones(&cb);
	if (err == 0) {
	err = zfs_destroy_snaps_nvl(g_zfs,
	cb.cb_batchedsnaps, B_FALSE);
	}
	if (err != 0) {
	rv = 1;
	goto out;
	}
	}
	if (err == 0) {
	err = zfs_destroy_snaps_nvl(g_zfs, cb.cb_nvl,
	cb.cb_defer_destroy);
	}
	}

	if (err != 0)
	rv = 1;
	} else if (pound != NULL) {
	int err;
	nvlist_t *nvl;

	if (cb.cb_dryrun) {
	(void) fprintf(stderr,
	"dryrun is not supported with bookmark\n");
	return (-1);
	}

	if (cb.cb_defer_destroy) {
	(void) fprintf(stderr,
	"defer destroy is not supported with bookmark\n");
	return (-1);
	}

	if (cb.cb_recurse) {
	(void) fprintf(stderr,
	"recursive is not supported with bookmark\n");
	return (-1);
	}

	/*
	* Unfortunately, zfs_bookmark() doesn't honor the
	* casesensitivity setting. However, we can't simply
	* remove this check, because lzc_destroy_bookmarks()
	* ignores non-existent bookmarks, so this is necessary
	* to get a proper error message.
	*/
	if (!zfs_bookmark_exists(argv[0])) {
	(void) fprintf(stderr, gettext("bookmark '%s' "
	"does not exist.\n"), argv[0]);
	return (1);
	}

	nvl = fnvlist_alloc();
	fnvlist_add_boolean(nvl, argv[0]);

	err = lzc_destroy_bookmarks(nvl, NULL);
	if (err != 0) {
	(void) zfs_standard_error(g_zfs, err,
	"cannot destroy bookmark");
	}

	nvlist_free(nvl);

	return (err);
	} else {
	/* Open the given dataset */
	if ((zhp = zfs_open(g_zfs, argv[0], type)) == NULL)
	return (1);

	cb.cb_target = zhp;

	/*
	* Perform an explicit check for pools before going any further.
	*/
	if (!cb.cb_recurse && strchr(zfs_get_name(zhp), '/') == NULL &&
	zfs_get_type(zhp) == ZFS_TYPE_FILESYSTEM) {
	(void) fprintf(stderr, gettext("cannot destroy '%s': "
	"operation does not apply to pools\n"),
	zfs_get_name(zhp));
	(void) fprintf(stderr, gettext("use 'zfs destroy -r "
	"%s' to destroy all datasets in the pool\n"),
	zfs_get_name(zhp));
	(void) fprintf(stderr, gettext("use 'zpool destroy %s' "
	"to destroy the pool itself\n"), zfs_get_name(zhp));
	rv = 1;
	goto out;
	}

	/*
	* Check for any dependents and/or clones.
	*/
	cb.cb_first = B_TRUE;
	if (!cb.cb_doclones && zfs_iter_dependents_v2(zhp, 0, B_TRUE,
	destroy_check_dependent, &cb) != 0) {
	rv = 1;
	goto out;
	}

	if (cb.cb_error) {
	rv = 1;
	goto out;
	}
	cb.cb_batchedsnaps = fnvlist_alloc();
	if (zfs_iter_dependents_v2(zhp, 0, B_FALSE, destroy_callback,
	&cb) != 0) {
	rv = 1;
	goto out;
	}

	/*
	* Do the real thing. The callback will close the
	* handle regardless of whether it succeeds or not.
	*/
	err = destroy_callback(zhp, &cb);
	zhp = NULL;
	if (err == 0) {
	err = zfs_destroy_snaps_nvl(g_zfs,
	cb.cb_batchedsnaps, cb.cb_defer_destroy);
	}
	if (err != 0 \|\| cb.cb_error == B_TRUE)
	rv = 1;
	}

	out:
	fnvlist_free(cb.cb_batchedsnaps);
	fnvlist_free(cb.cb_nvl);
	if (zhp != NULL)
	zfs_close(zhp);
	return (rv);
	}

	static boolean_t
	is_recvd_column(zprop_get_cbdata_t *cbp)
	{
	int i;
	zfs_get_column_t col;

	for (i = 0; i < ZFS_GET_NCOLS &&
	(col = cbp->cb_columns[i]) != GET_COL_NONE; i++)
	if (col == GET_COL_RECVD)
	return (B_TRUE);
	return (B_FALSE);
	}

	/*
	* zfs get [-rHp] [-o all \| field[,field]...] [-s source[,source]...]
	* < all \| property[,property]... > < fs \| snap \| vol > ...
	*
	* -r recurse over any child datasets
	* -H scripted mode. Headers are stripped, and fields are separated
	* by tabs instead of spaces.
	* -o Set of fields to display. One of "name,property,value,
	* received,source". Default is "name,property,value,source".
	* "all" is an alias for all five.
	* -s Set of sources to allow. One of
	* "local,default,inherited,received,temporary,none". Default is
	* all six.
	* -p Display values in parsable (literal) format.
	*
	* Prints properties for the given datasets. The user can control which
	* columns to display as well as which property types to allow.
	*/

	/*
	* Invoked to display the properties for a single dataset.
	*/
	static int
	get_callback(zfs_handle_t zhp, void data)
	{
	char buf[ZFS_MAXPROPLEN];
	char rbuf[ZFS_MAXPROPLEN];
	zprop_source_t sourcetype;
	char source[ZFS_MAX_DATASET_NAME_LEN];
	zprop_get_cbdata_t *cbp = data;
	nvlist_t *user_props = zfs_get_user_props(zhp);
	zprop_list_t *pl = cbp->cb_proplist;
	nvlist_t *propval;
	const char *strval;
	const char *sourceval;
	boolean_t received = is_recvd_column(cbp);

	for (; pl != NULL; pl = pl->pl_next) {
	char *recvdval = NULL;
	/*
	* Skip the special fake placeholder. This will also skip over
	* the name property when 'all' is specified.
	*/
	if (pl->pl_prop == ZFS_PROP_NAME &&
	pl == cbp->cb_proplist)
	continue;

	if (pl->pl_prop != ZPROP_USERPROP) {
	if (zfs_prop_get(zhp, pl->pl_prop, buf,
	sizeof (buf), &sourcetype, source,
	sizeof (source),
	cbp->cb_literal) != 0) {
	if (pl->pl_all)
	continue;
	if (!zfs_prop_valid_for_type(pl->pl_prop,
	ZFS_TYPE_DATASET, B_FALSE)) {
	(void) fprintf(stderr,
	gettext("No such property '%s'\n"),
	zfs_prop_to_name(pl->pl_prop));
	continue;
	}
	sourcetype = ZPROP_SRC_NONE;
	(void) strlcpy(buf, "-", sizeof (buf));
	}

	if (received && (zfs_prop_get_recvd(zhp,
	zfs_prop_to_name(pl->pl_prop), rbuf, sizeof (rbuf),
	cbp->cb_literal) == 0))
	recvdval = rbuf;

	zprop_print_one_property(zfs_get_name(zhp), cbp,
	zfs_prop_to_name(pl->pl_prop),
	buf, sourcetype, source, recvdval);
	} else if (zfs_prop_userquota(pl->pl_user_prop)) {
	sourcetype = ZPROP_SRC_LOCAL;

	if (zfs_prop_get_userquota(zhp, pl->pl_user_prop,
	buf, sizeof (buf), cbp->cb_literal) != 0) {
	sourcetype = ZPROP_SRC_NONE;
	(void) strlcpy(buf, "-", sizeof (buf));
	}

	zprop_print_one_property(zfs_get_name(zhp), cbp,
	pl->pl_user_prop, buf, sourcetype, source, NULL);
	} else if (zfs_prop_written(pl->pl_user_prop)) {
	sourcetype = ZPROP_SRC_LOCAL;

	if (zfs_prop_get_written(zhp, pl->pl_user_prop,
	buf, sizeof (buf), cbp->cb_literal) != 0) {
	sourcetype = ZPROP_SRC_NONE;
	(void) strlcpy(buf, "-", sizeof (buf));
	}

	zprop_print_one_property(zfs_get_name(zhp), cbp,
	pl->pl_user_prop, buf, sourcetype, source, NULL);
	} else {
	if (nvlist_lookup_nvlist(user_props,
	pl->pl_user_prop, &propval) != 0) {
	if (pl->pl_all)
	continue;
	sourcetype = ZPROP_SRC_NONE;
	strval = "-";
	} else {
	strval = fnvlist_lookup_string(propval,
	ZPROP_VALUE);
	sourceval = fnvlist_lookup_string(propval,
	ZPROP_SOURCE);

	if (strcmp(sourceval,
	zfs_get_name(zhp)) == 0) {
	sourcetype = ZPROP_SRC_LOCAL;
	} else if (strcmp(sourceval,
	ZPROP_SOURCE_VAL_RECVD) == 0) {
	sourcetype = ZPROP_SRC_RECEIVED;
	} else {
	sourcetype = ZPROP_SRC_INHERITED;
	(void) strlcpy(source,
	sourceval, sizeof (source));
	}
	}

	if (received && (zfs_prop_get_recvd(zhp,
	pl->pl_user_prop, rbuf, sizeof (rbuf),
	cbp->cb_literal) == 0))
	recvdval = rbuf;

	zprop_print_one_property(zfs_get_name(zhp), cbp,
	pl->pl_user_prop, strval, sourcetype,
	source, recvdval);
	}
	}

	return (0);
	}

	static int
	zfs_do_get(int argc, char **argv)
	{
	zprop_get_cbdata_t cb = { 0 };
	int i, c, flags = ZFS_ITER_ARGS_CAN_BE_PATHS;
	int types = ZFS_TYPE_DATASET \| ZFS_TYPE_BOOKMARK;
	char *fields;
	int ret = 0;
	int limit = 0;
	zprop_list_t fake_name = { 0 };

	/*
	* Set up default columns and sources.
	*/
	cb.cb_sources = ZPROP_SRC_ALL;
	cb.cb_columns[0] = GET_COL_NAME;
	cb.cb_columns[1] = GET_COL_PROPERTY;
	cb.cb_columns[2] = GET_COL_VALUE;
	cb.cb_columns[3] = GET_COL_SOURCE;
	cb.cb_type = ZFS_TYPE_DATASET;

	/* check options */
	while ((c = getopt(argc, argv, ":d:o:s:rt:Hp")) != -1) {
	switch (c) {
	case 'p':
	cb.cb_literal = B_TRUE;
	break;
	case 'd':
	limit = parse_depth(optarg, &flags);
	break;
	case 'r':
	flags \|= ZFS_ITER_RECURSE;
	break;
	case 'H':
	cb.cb_scripted = B_TRUE;
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	usage(B_FALSE);
	break;
	case 'o':
	/*
	* Process the set of columns to display. We zero out
	* the structure to give us a blank slate.
	*/
	memset(&cb.cb_columns, 0, sizeof (cb.cb_columns));

	i = 0;
	for (char *tok; (tok = strsep(&optarg, ",")); ) {
	static const char *const col_subopts[] =
	{ "name", "property", "value",
	"received", "source", "all" };
	static const zfs_get_column_t col_subopt_col[] =
	{ GET_COL_NAME, GET_COL_PROPERTY, GET_COL_VALUE,
	GET_COL_RECVD, GET_COL_SOURCE };
	static const int col_subopt_flags[] =
	{ 0, 0, 0, ZFS_ITER_RECVD_PROPS, 0 };

	if (i == ZFS_GET_NCOLS) {
	(void) fprintf(stderr, gettext("too "
	"many fields given to -o "
	"option\n"));
	usage(B_FALSE);
	}

	for (c = 0; c < ARRAY_SIZE(col_subopts); ++c)
	if (strcmp(tok, col_subopts[c]) == 0)
	goto found;

	(void) fprintf(stderr,
	gettext("invalid column name '%s'\n"), tok);
	usage(B_FALSE);

	found:
	if (c >= 5) {
	if (i > 0) {
	(void) fprintf(stderr,
	gettext("\"all\" conflicts "
	"with specific fields "
	"given to -o option\n"));
	usage(B_FALSE);
	}

	memcpy(cb.cb_columns, col_subopt_col,
	sizeof (col_subopt_col));
	flags \|= ZFS_ITER_RECVD_PROPS;
	i = ZFS_GET_NCOLS;
	} else {
	cb.cb_columns[i++] = col_subopt_col[c];
	flags \|= col_subopt_flags[c];
	}
	}
	break;

	case 's':
	cb.cb_sources = 0;

	for (char *tok; (tok = strsep(&optarg, ",")); ) {
	static const char *const source_opt[] = {
	"local", "default",
	"inherited", "received",
	"temporary", "none" };
	static const int source_flg[] = {
	ZPROP_SRC_LOCAL, ZPROP_SRC_DEFAULT,
	ZPROP_SRC_INHERITED, ZPROP_SRC_RECEIVED,
	ZPROP_SRC_TEMPORARY, ZPROP_SRC_NONE };

	for (i = 0; i < ARRAY_SIZE(source_opt); ++i)
	if (strcmp(tok, source_opt[i]) == 0) {
	cb.cb_sources \|= source_flg[i];
	goto found2;
	}

	(void) fprintf(stderr,
	gettext("invalid source '%s'\n"), tok);
	usage(B_FALSE);
	found2:;
	}
	break;

	case 't':
	types = 0;
	flags &= ~ZFS_ITER_PROP_LISTSNAPS;

	for (char *tok; (tok = strsep(&optarg, ",")); ) {
	static const char *const type_opts[] = {
	"filesystem",
	"fs",
	"volume",
	"vol",
	"snapshot",
	"snap",
	"bookmark",
	"all"
	};
	static const int type_types[] = {
	ZFS_TYPE_FILESYSTEM,
	ZFS_TYPE_FILESYSTEM,
	ZFS_TYPE_VOLUME,
	ZFS_TYPE_VOLUME,
	ZFS_TYPE_SNAPSHOT,
	ZFS_TYPE_SNAPSHOT,
	ZFS_TYPE_BOOKMARK,
	ZFS_TYPE_DATASET \| ZFS_TYPE_BOOKMARK
	};

	for (i = 0; i < ARRAY_SIZE(type_opts); ++i)
	if (strcmp(tok, type_opts[i]) == 0) {
	types \|= type_types[i];
	goto found3;
	}

	(void) fprintf(stderr,
	gettext("invalid type '%s'\n"), tok);
	usage(B_FALSE);
	found3:;
	}
	break;

	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing property "
	"argument\n"));
	usage(B_FALSE);
	}

	fields = argv[0];

	/*
	* Handle users who want to get all snapshots or bookmarks
	* of a dataset (ex. 'zfs get -t snapshot refer <dataset>').
	*/
	if ((types == ZFS_TYPE_SNAPSHOT \|\| types == ZFS_TYPE_BOOKMARK) &&
	argc > 1 && (flags & ZFS_ITER_RECURSE) == 0 && limit == 0) {
	flags \|= (ZFS_ITER_DEPTH_LIMIT \| ZFS_ITER_RECURSE);
	limit = 1;
	}

	if (zprop_get_list(g_zfs, fields, &cb.cb_proplist, ZFS_TYPE_DATASET)
	!= 0)
	usage(B_FALSE);

	argc--;
	argv++;

	/*
	* As part of zfs_expand_proplist(), we keep track of the maximum column
	* width for each property. For the 'NAME' (and 'SOURCE') columns, we
	* need to know the maximum name length. However, the user likely did
	* not specify 'name' as one of the properties to fetch, so we need to
	* make sure we always include at least this property for
	* print_get_headers() to work properly.
	*/
	if (cb.cb_proplist != NULL) {
	fake_name.pl_prop = ZFS_PROP_NAME;
	fake_name.pl_width = strlen(gettext("NAME"));
	fake_name.pl_next = cb.cb_proplist;
	cb.cb_proplist = &fake_name;
	}

	cb.cb_first = B_TRUE;

	/* run for each object */
	ret = zfs_for_each(argc, argv, flags, types, NULL,
	&cb.cb_proplist, limit, get_callback, &cb);

	if (cb.cb_proplist == &fake_name)
	zprop_free_list(fake_name.pl_next);
	else
	zprop_free_list(cb.cb_proplist);

	return (ret);
	}

	/*
	* inherit [-rS] <property> <fs\|vol> ...
	*
	* -r Recurse over all children
	* -S Revert to received value, if any
	*
	* For each dataset specified on the command line, inherit the given property
	* from its parent. Inheriting a property at the pool level will cause it to
	* use the default value. The '-r' flag will recurse over all children, and is
	* useful for setting a property on a hierarchy-wide basis, regardless of any
	* local modifications for each dataset.
	*/

	typedef struct inherit_cbdata {
	const char *cb_propname;
	boolean_t cb_received;
	} inherit_cbdata_t;

	static int
	inherit_recurse_cb(zfs_handle_t zhp, void data)
	{
	inherit_cbdata_t *cb = data;
	zfs_prop_t prop = zfs_name_to_prop(cb->cb_propname);

	/*
	* If we're doing it recursively, then ignore properties that
	* are not valid for this type of dataset.
	*/
	if (prop != ZPROP_INVAL &&
	!zfs_prop_valid_for_type(prop, zfs_get_type(zhp), B_FALSE))
	return (0);

	return (zfs_prop_inherit(zhp, cb->cb_propname, cb->cb_received) != 0);
	}

	static int
	inherit_cb(zfs_handle_t zhp, void data)
	{
	inherit_cbdata_t *cb = data;

	return (zfs_prop_inherit(zhp, cb->cb_propname, cb->cb_received) != 0);
	}

	static int
	zfs_do_inherit(int argc, char **argv)
	{
	int c;
	zfs_prop_t prop;
	inherit_cbdata_t cb = { 0 };
	char *propname;
	int ret = 0;
	int flags = 0;
	boolean_t received = B_FALSE;

	/* check options */
	while ((c = getopt(argc, argv, "rS")) != -1) {
	switch (c) {
	case 'r':
	flags \|= ZFS_ITER_RECURSE;
	break;
	case 'S':
	received = B_TRUE;
	break;
	case '?':
	default:
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing property argument\n"));
	usage(B_FALSE);
	}
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing dataset argument\n"));
	usage(B_FALSE);
	}

	propname = argv[0];
	argc--;
	argv++;

	if ((prop = zfs_name_to_prop(propname)) != ZPROP_USERPROP) {
	if (zfs_prop_readonly(prop)) {
	(void) fprintf(stderr, gettext(
	"%s property is read-only\n"),
	propname);
	return (1);
	}
	if (!zfs_prop_inheritable(prop) && !received) {
	(void) fprintf(stderr, gettext("'%s' property cannot "
	"be inherited\n"), propname);
	if (prop == ZFS_PROP_QUOTA \|\|
	prop == ZFS_PROP_RESERVATION \|\|
	prop == ZFS_PROP_REFQUOTA \|\|
	prop == ZFS_PROP_REFRESERVATION) {
	(void) fprintf(stderr, gettext("use 'zfs set "
	"%s=none' to clear\n"), propname);
	(void) fprintf(stderr, gettext("use 'zfs "
	"inherit -S %s' to revert to received "
	"value\n"), propname);
	}
	return (1);
	}
	if (received && (prop == ZFS_PROP_VOLSIZE \|\|
	prop == ZFS_PROP_VERSION)) {
	(void) fprintf(stderr, gettext("'%s' property cannot "
	"be reverted to a received value\n"), propname);
	return (1);
	}
	} else if (!zfs_prop_user(propname)) {
	(void) fprintf(stderr, gettext("invalid property '%s'\n"),
	propname);
	usage(B_FALSE);
	}

	cb.cb_propname = propname;
	cb.cb_received = received;

	if (flags & ZFS_ITER_RECURSE) {
	ret = zfs_for_each(argc, argv, flags, ZFS_TYPE_DATASET,
	NULL, NULL, 0, inherit_recurse_cb, &cb);
	} else {
	ret = zfs_for_each(argc, argv, flags, ZFS_TYPE_DATASET,
	NULL, NULL, 0, inherit_cb, &cb);
	}

	return (ret);
	}

	typedef struct upgrade_cbdata {
	uint64_t cb_numupgraded;
	uint64_t cb_numsamegraded;
	uint64_t cb_numfailed;
	uint64_t cb_version;
	boolean_t cb_newer;
	boolean_t cb_foundone;
	char cb_lastfs[ZFS_MAX_DATASET_NAME_LEN];
	} upgrade_cbdata_t;

	static int
	same_pool(zfs_handle_t zhp, const char name)
	{
	int len1 = strcspn(name, "/@");
	const char *zhname = zfs_get_name(zhp);
	int len2 = strcspn(zhname, "/@");

	if (len1 != len2)
	return (B_FALSE);
	return (strncmp(name, zhname, len1) == 0);
	}

	static int
	upgrade_list_callback(zfs_handle_t zhp, void data)
	{
	upgrade_cbdata_t *cb = data;
	int version = zfs_prop_get_int(zhp, ZFS_PROP_VERSION);

	/* list if it's old/new */
	if ((!cb->cb_newer && version < ZPL_VERSION) \|\|
	(cb->cb_newer && version > ZPL_VERSION)) {
	char *str;
	if (cb->cb_newer) {
	str = gettext("The following filesystems are "
	"formatted using a newer software version and\n"
	"cannot be accessed on the current system.\n\n");
	} else {
	str = gettext("The following filesystems are "
	"out of date, and can be upgraded. After being\n"
	"upgraded, these filesystems (and any 'zfs send' "
	"streams generated from\n"
	"subsequent snapshots) will no longer be "
	"accessible by older software versions.\n\n");
	}

	if (!cb->cb_foundone) {
	(void) puts(str);
	(void) printf(gettext("VER FILESYSTEM\n"));
	(void) printf(gettext("--- ------------\n"));
	cb->cb_foundone = B_TRUE;
	}

	(void) printf("%2u %s\n", version, zfs_get_name(zhp));
	}

	return (0);
	}

	static int
	upgrade_set_callback(zfs_handle_t zhp, void data)
	{
	upgrade_cbdata_t *cb = data;
	int version = zfs_prop_get_int(zhp, ZFS_PROP_VERSION);
	int needed_spa_version;
	int spa_version;

	if (zfs_spa_version(zhp, &spa_version) < 0)
	return (-1);

	needed_spa_version = zfs_spa_version_map(cb->cb_version);

	if (needed_spa_version < 0)
	return (-1);

	if (spa_version < needed_spa_version) {
	/* can't upgrade */
	(void) printf(gettext("%s: can not be "
	"upgraded; the pool version needs to first "
	"be upgraded\nto version %d\n\n"),
	zfs_get_name(zhp), needed_spa_version);
	cb->cb_numfailed++;
	return (0);
	}

	/* upgrade */
	if (version < cb->cb_version) {
	char verstr[24];
	(void) snprintf(verstr, sizeof (verstr),
	"%llu", (u_longlong_t)cb->cb_version);
	if (cb->cb_lastfs[0] && !same_pool(zhp, cb->cb_lastfs)) {
	/*
	* If they did "zfs upgrade -a", then we could
	* be doing ioctls to different pools. We need
	* to log this history once to each pool, and bypass
	* the normal history logging that happens in main().
	*/
	(void) zpool_log_history(g_zfs, history_str);
	log_history = B_FALSE;
	}
	if (zfs_prop_set(zhp, "version", verstr) == 0)
	cb->cb_numupgraded++;
	else
	cb->cb_numfailed++;
	(void) strlcpy(cb->cb_lastfs, zfs_get_name(zhp),
	sizeof (cb->cb_lastfs));
	} else if (version > cb->cb_version) {
	/* can't downgrade */
	(void) printf(gettext("%s: can not be downgraded; "
	"it is already at version %u\n"),
	zfs_get_name(zhp), version);
	cb->cb_numfailed++;
	} else {
	cb->cb_numsamegraded++;
	}
	return (0);
	}

	/*
	* zfs upgrade
	* zfs upgrade -v
	* zfs upgrade [-r] [-V <version>] <-a \| filesystem>
	*/
	static int
	zfs_do_upgrade(int argc, char **argv)
	{
	boolean_t all = B_FALSE;
	boolean_t showversions = B_FALSE;
	int ret = 0;
	upgrade_cbdata_t cb = { 0 };
	int c;
	int flags = ZFS_ITER_ARGS_CAN_BE_PATHS;

	/* check options */
	while ((c = getopt(argc, argv, "rvV:a")) != -1) {
	switch (c) {
	case 'r':
	flags \|= ZFS_ITER_RECURSE;
	break;
	case 'v':
	showversions = B_TRUE;
	break;
	case 'V':
	if (zfs_prop_string_to_index(ZFS_PROP_VERSION,
	optarg, &cb.cb_version) != 0) {
	(void) fprintf(stderr,
	gettext("invalid version %s\n"), optarg);
	usage(B_FALSE);
	}
	break;
	case 'a':
	all = B_TRUE;
	break;
	case '?':
	default:
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	if ((!all && !argc) && ((flags & ZFS_ITER_RECURSE) \| cb.cb_version))
	usage(B_FALSE);
	if (showversions && (flags & ZFS_ITER_RECURSE \|\| all \|\|
	cb.cb_version \|\| argc))
	usage(B_FALSE);
	if ((all \|\| argc) && (showversions))
	usage(B_FALSE);
	if (all && argc)
	usage(B_FALSE);

	if (showversions) {
	/* Show info on available versions. */
	(void) printf(gettext("The following filesystem versions are "
	"supported:\n\n"));
	(void) printf(gettext("VER DESCRIPTION\n"));
	(void) printf("--- -----------------------------------------"
	"---------------\n");
	(void) printf(gettext(" 1 Initial ZFS filesystem version\n"));
	(void) printf(gettext(" 2 Enhanced directory entries\n"));
	(void) printf(gettext(" 3 Case insensitive and filesystem "
	"user identifier (FUID)\n"));
	(void) printf(gettext(" 4 userquota, groupquota "
	"properties\n"));
	(void) printf(gettext(" 5 System attributes\n"));
	(void) printf(gettext("\nFor more information on a particular "
	"version, including supported releases,\n"));
	(void) printf("see the ZFS Administration Guide.\n\n");
	ret = 0;
	} else if (argc \|\| all) {
	/* Upgrade filesystems */
	if (cb.cb_version == 0)
	cb.cb_version = ZPL_VERSION;
	ret = zfs_for_each(argc, argv, flags, ZFS_TYPE_FILESYSTEM,
	NULL, NULL, 0, upgrade_set_callback, &cb);
	(void) printf(gettext("%llu filesystems upgraded\n"),
	(u_longlong_t)cb.cb_numupgraded);
	if (cb.cb_numsamegraded) {
	(void) printf(gettext("%llu filesystems already at "
	"this version\n"),
	(u_longlong_t)cb.cb_numsamegraded);
	}
	if (cb.cb_numfailed != 0)
	ret = 1;
	} else {
	/* List old-version filesystems */
	boolean_t found;
	(void) printf(gettext("This system is currently running "
	"ZFS filesystem version %llu.\n\n"), ZPL_VERSION);

	flags \|= ZFS_ITER_RECURSE;
	ret = zfs_for_each(0, NULL, flags, ZFS_TYPE_FILESYSTEM,
	NULL, NULL, 0, upgrade_list_callback, &cb);

	found = cb.cb_foundone;
	cb.cb_foundone = B_FALSE;
	cb.cb_newer = B_TRUE;

	ret \|= zfs_for_each(0, NULL, flags, ZFS_TYPE_FILESYSTEM,
	NULL, NULL, 0, upgrade_list_callback, &cb);

	if (!cb.cb_foundone && !found) {
	(void) printf(gettext("All filesystems are "
	"formatted with the current version.\n"));
	}
	}

	return (ret);
	}

	/*
	* zfs userspace [-Hinp] [-o field[,...]] [-s field [-s field]...]
	* [-S field [-S field]...] [-t type[,...]]
	* filesystem \| snapshot \| path
	* zfs groupspace [-Hinp] [-o field[,...]] [-s field [-s field]...]
	* [-S field [-S field]...] [-t type[,...]]
	* filesystem \| snapshot \| path
	* zfs projectspace [-Hp] [-o field[,...]] [-s field [-s field]...]
	* [-S field [-S field]...] filesystem \| snapshot \| path
	*
	* -H Scripted mode; elide headers and separate columns by tabs.
	* -i Translate SID to POSIX ID.
	* -n Print numeric ID instead of user/group name.
	* -o Control which fields to display.
	* -p Use exact (parsable) numeric output.
	* -s Specify sort columns, descending order.
	* -S Specify sort columns, ascending order.
	* -t Control which object types to display.
	*
	* Displays space consumed by, and quotas on, each user in the specified
	* filesystem or snapshot.
	*/

	/* us_field_types, us_field_hdr and us_field_names should be kept in sync */
	enum us_field_types {
	USFIELD_TYPE,
	USFIELD_NAME,
	USFIELD_USED,
	USFIELD_QUOTA,
	USFIELD_OBJUSED,
	USFIELD_OBJQUOTA
	};
	static const char *const us_field_hdr[] = { "TYPE", "NAME", "USED", "QUOTA",
	"OBJUSED", "OBJQUOTA" };
	static const char *const us_field_names[] = { "type", "name", "used", "quota",
	"objused", "objquota" };
	#define USFIELD_LAST (sizeof (us_field_names) / sizeof (char *))

	#define USTYPE_PSX_GRP (1 << 0)
	#define USTYPE_PSX_USR (1 << 1)
	#define USTYPE_SMB_GRP (1 << 2)
	#define USTYPE_SMB_USR (1 << 3)
	#define USTYPE_PROJ (1 << 4)
	#define USTYPE_ALL \
	(USTYPE_PSX_GRP \| USTYPE_PSX_USR \| USTYPE_SMB_GRP \| USTYPE_SMB_USR \| \
	USTYPE_PROJ)

	static int us_type_bits[] = {
	USTYPE_PSX_GRP,
	USTYPE_PSX_USR,
	USTYPE_SMB_GRP,
	USTYPE_SMB_USR,
	USTYPE_ALL
	};
	static const char *const us_type_names[] = { "posixgroup", "posixuser",
	"smbgroup", "smbuser", "all" };

	typedef struct us_node {
	nvlist_t *usn_nvl;
	uu_avl_node_t usn_avlnode;
	uu_list_node_t usn_listnode;
	} us_node_t;

	typedef struct us_cbdata {
	nvlist_t **cb_nvlp;
	uu_avl_pool_t *cb_avl_pool;
	uu_avl_t *cb_avl;
	boolean_t cb_numname;
	boolean_t cb_nicenum;
	boolean_t cb_sid2posix;
	zfs_userquota_prop_t cb_prop;
	zfs_sort_column_t *cb_sortcol;
	size_t cb_width[USFIELD_LAST];
	} us_cbdata_t;

	static boolean_t us_populated = B_FALSE;

	typedef struct {
	zfs_sort_column_t *si_sortcol;
	boolean_t si_numname;
	} us_sort_info_t;

	static int
	us_field_index(const char *field)
	{
	for (int i = 0; i < USFIELD_LAST; i++) {
	if (strcmp(field, us_field_names[i]) == 0)
	return (i);
	}

	return (-1);
	}

	static int
	us_compare(const void larg, const void rarg, void *unused)
	{
	const us_node_t *l = larg;
	const us_node_t *r = rarg;
	us_sort_info_t si = (us_sort_info_t )unused;
	zfs_sort_column_t *sortcol = si->si_sortcol;
	boolean_t numname = si->si_numname;
	nvlist_t *lnvl = l->usn_nvl;
	nvlist_t *rnvl = r->usn_nvl;
	int rc = 0;
	boolean_t lvb, rvb;

	for (; sortcol != NULL; sortcol = sortcol->sc_next) {
	const char *lvstr = "";
	const char *rvstr = "";
	uint32_t lv32 = 0;
	uint32_t rv32 = 0;
	uint64_t lv64 = 0;
	uint64_t rv64 = 0;
	zfs_prop_t prop = sortcol->sc_prop;
	const char *propname = NULL;
	boolean_t reverse = sortcol->sc_reverse;

	switch (prop) {
	case ZFS_PROP_TYPE:
	propname = "type";
	(void) nvlist_lookup_uint32(lnvl, propname, &lv32);
	(void) nvlist_lookup_uint32(rnvl, propname, &rv32);
	if (rv32 != lv32)
	rc = (rv32 < lv32) ? 1 : -1;
	break;
	case ZFS_PROP_NAME:
	propname = "name";
	if (numname) {
	compare_nums:
	(void) nvlist_lookup_uint64(lnvl, propname,
	&lv64);
	(void) nvlist_lookup_uint64(rnvl, propname,
	&rv64);
	if (rv64 != lv64)
	rc = (rv64 < lv64) ? 1 : -1;
	} else {
	if ((nvlist_lookup_string(lnvl, propname,
	&lvstr) == ENOENT) \|\|
	(nvlist_lookup_string(rnvl, propname,
	&rvstr) == ENOENT)) {
	goto compare_nums;
	}
	rc = strcmp(lvstr, rvstr);
	}
	break;
	case ZFS_PROP_USED:
	case ZFS_PROP_QUOTA:
	if (!us_populated)
	break;
	if (prop == ZFS_PROP_USED)
	propname = "used";
	else
	propname = "quota";
	(void) nvlist_lookup_uint64(lnvl, propname, &lv64);
	(void) nvlist_lookup_uint64(rnvl, propname, &rv64);
	if (rv64 != lv64)
	rc = (rv64 < lv64) ? 1 : -1;
	break;

	default:
	break;
	}

	if (rc != 0) {
	if (rc < 0)
	return (reverse ? 1 : -1);
	else
	return (reverse ? -1 : 1);
	}
	}

	/*
	* If entries still seem to be the same, check if they are of the same
	* type (smbentity is added only if we are doing SID to POSIX ID
	* translation where we can have duplicate type/name combinations).
	*/
	if (nvlist_lookup_boolean_value(lnvl, "smbentity", &lvb) == 0 &&
	nvlist_lookup_boolean_value(rnvl, "smbentity", &rvb) == 0 &&
	lvb != rvb)
	return (lvb < rvb ? -1 : 1);

	return (0);
	}

	static boolean_t
	zfs_prop_is_user(unsigned p)
	{
	return (p == ZFS_PROP_USERUSED \|\| p == ZFS_PROP_USERQUOTA \|\|
	p == ZFS_PROP_USEROBJUSED \|\| p == ZFS_PROP_USEROBJQUOTA);
	}

	static boolean_t
	zfs_prop_is_group(unsigned p)
	{
	return (p == ZFS_PROP_GROUPUSED \|\| p == ZFS_PROP_GROUPQUOTA \|\|
	p == ZFS_PROP_GROUPOBJUSED \|\| p == ZFS_PROP_GROUPOBJQUOTA);
	}

	static boolean_t
	zfs_prop_is_project(unsigned p)
	{
	return (p == ZFS_PROP_PROJECTUSED \|\| p == ZFS_PROP_PROJECTQUOTA \|\|
	p == ZFS_PROP_PROJECTOBJUSED \|\| p == ZFS_PROP_PROJECTOBJQUOTA);
	}

	static inline const char *
	us_type2str(unsigned field_type)
	{
	switch (field_type) {
	case USTYPE_PSX_USR:
	return ("POSIX User");
	case USTYPE_PSX_GRP:
	return ("POSIX Group");
	case USTYPE_SMB_USR:
	return ("SMB User");
	case USTYPE_SMB_GRP:
	return ("SMB Group");
	case USTYPE_PROJ:
	return ("Project");
	default:
	return ("Undefined");
	}
	}

	static int
	userspace_cb(void arg, const char domain, uid_t rid, uint64_t space)
	{
	us_cbdata_t cb = (us_cbdata_t )arg;
	zfs_userquota_prop_t prop = cb->cb_prop;
	char *name = NULL;
	const char *propname;
	char sizebuf[32];
	us_node_t *node;
	uu_avl_pool_t *avl_pool = cb->cb_avl_pool;
	uu_avl_t *avl = cb->cb_avl;
	uu_avl_index_t idx;
	nvlist_t *props;
	us_node_t *n;
	zfs_sort_column_t *sortcol = cb->cb_sortcol;
	unsigned type = 0;
	const char *typestr;
	size_t namelen;
	size_t typelen;
	size_t sizelen;
	int typeidx, nameidx, sizeidx;
	us_sort_info_t sortinfo = { sortcol, cb->cb_numname };
	boolean_t smbentity = B_FALSE;

	if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0)
	nomem();
	node = safe_malloc(sizeof (us_node_t));
	uu_avl_node_init(node, &node->usn_avlnode, avl_pool);
	node->usn_nvl = props;

	if (domain != NULL && domain[0] != '\0') {
	#ifdef HAVE_IDMAP
	/* SMB */
	char sid[MAXNAMELEN + 32];
	uid_t id;
	uint64_t classes;
	int err;
	directory_error_t e;

	smbentity = B_TRUE;

	(void) snprintf(sid, sizeof (sid), "%s-%u", domain, rid);

	if (prop == ZFS_PROP_GROUPUSED \|\| prop == ZFS_PROP_GROUPQUOTA) {
	type = USTYPE_SMB_GRP;
	err = sid_to_id(sid, B_FALSE, &id);
	} else {
	type = USTYPE_SMB_USR;
	err = sid_to_id(sid, B_TRUE, &id);
	}

	if (err == 0) {
	rid = id;
	if (!cb->cb_sid2posix) {
	e = directory_name_from_sid(NULL, sid, &name,
	&classes);
	if (e != NULL)
	directory_error_free(e);
	if (name == NULL)
	name = sid;
	}
	}
	#else
	nvlist_free(props);
	free(node);

	return (-1);
	#endif /* HAVE_IDMAP */
	}

	if (cb->cb_sid2posix \|\| domain == NULL \|\| domain[0] == '\0') {
	/* POSIX or -i */
	if (zfs_prop_is_group(prop)) {
	type = USTYPE_PSX_GRP;
	if (!cb->cb_numname) {
	struct group *g;

	if ((g = getgrgid(rid)) != NULL)
	name = g->gr_name;
	}
	} else if (zfs_prop_is_user(prop)) {
	type = USTYPE_PSX_USR;
	if (!cb->cb_numname) {
	struct passwd *p;

	if ((p = getpwuid(rid)) != NULL)
	name = p->pw_name;
	}
	} else {
	type = USTYPE_PROJ;
	}
	}

	/*
	* Make sure that the type/name combination is unique when doing
	* SID to POSIX ID translation (hence changing the type from SMB to
	* POSIX).
	*/
	if (cb->cb_sid2posix &&
	nvlist_add_boolean_value(props, "smbentity", smbentity) != 0)
	nomem();

	/* Calculate/update width of TYPE field */
	typestr = us_type2str(type);
	typelen = strlen(gettext(typestr));
	typeidx = us_field_index("type");
	if (typelen > cb->cb_width[typeidx])
	cb->cb_width[typeidx] = typelen;
	if (nvlist_add_uint32(props, "type", type) != 0)
	nomem();

	/* Calculate/update width of NAME field */
	if ((cb->cb_numname && cb->cb_sid2posix) \|\| name == NULL) {
	if (nvlist_add_uint64(props, "name", rid) != 0)
	nomem();
	namelen = snprintf(NULL, 0, "%u", rid);
	} else {
	if (nvlist_add_string(props, "name", name) != 0)
	nomem();
	namelen = strlen(name);
	}
	nameidx = us_field_index("name");
	if (nameidx >= 0 && namelen > cb->cb_width[nameidx])
	cb->cb_width[nameidx] = namelen;

	/*
	* Check if this type/name combination is in the list and update it;
	* otherwise add new node to the list.
	*/
	if ((n = uu_avl_find(avl, node, &sortinfo, &idx)) == NULL) {
	uu_avl_insert(avl, node, idx);
	} else {
	nvlist_free(props);
	free(node);
	node = n;
	props = node->usn_nvl;
	}

	/* Calculate/update width of USED/QUOTA fields */
	if (cb->cb_nicenum) {
	if (prop == ZFS_PROP_USERUSED \|\| prop == ZFS_PROP_GROUPUSED \|\|
	prop == ZFS_PROP_USERQUOTA \|\| prop == ZFS_PROP_GROUPQUOTA \|\|
	prop == ZFS_PROP_PROJECTUSED \|\|
	prop == ZFS_PROP_PROJECTQUOTA) {
	zfs_nicebytes(space, sizebuf, sizeof (sizebuf));
	} else {
	zfs_nicenum(space, sizebuf, sizeof (sizebuf));
	}
	} else {
	(void) snprintf(sizebuf, sizeof (sizebuf), "%llu",
	(u_longlong_t)space);
	}
	sizelen = strlen(sizebuf);
	if (prop == ZFS_PROP_USERUSED \|\| prop == ZFS_PROP_GROUPUSED \|\|
	prop == ZFS_PROP_PROJECTUSED) {
	propname = "used";
	if (!nvlist_exists(props, "quota"))
	(void) nvlist_add_uint64(props, "quota", 0);
	} else if (prop == ZFS_PROP_USERQUOTA \|\| prop == ZFS_PROP_GROUPQUOTA \|\|
	prop == ZFS_PROP_PROJECTQUOTA) {
	propname = "quota";
	if (!nvlist_exists(props, "used"))
	(void) nvlist_add_uint64(props, "used", 0);
	} else if (prop == ZFS_PROP_USEROBJUSED \|\|
	prop == ZFS_PROP_GROUPOBJUSED \|\| prop == ZFS_PROP_PROJECTOBJUSED) {
	propname = "objused";
	if (!nvlist_exists(props, "objquota"))
	(void) nvlist_add_uint64(props, "objquota", 0);
	} else if (prop == ZFS_PROP_USEROBJQUOTA \|\|
	prop == ZFS_PROP_GROUPOBJQUOTA \|\|
	prop == ZFS_PROP_PROJECTOBJQUOTA) {
	propname = "objquota";
	if (!nvlist_exists(props, "objused"))
	(void) nvlist_add_uint64(props, "objused", 0);
	} else {
	return (-1);
	}
	sizeidx = us_field_index(propname);
	if (sizeidx >= 0 && sizelen > cb->cb_width[sizeidx])
	cb->cb_width[sizeidx] = sizelen;

	if (nvlist_add_uint64(props, propname, space) != 0)
	nomem();

	return (0);
	}

	static void
	print_us_node(boolean_t scripted, boolean_t parsable, int *fields, int types,
	size_t width, us_node_t node)
	{
	nvlist_t *nvl = node->usn_nvl;
	char valstr[MAXNAMELEN];
	boolean_t first = B_TRUE;
	int cfield = 0;
	int field;
	uint32_t ustype;

	/* Check type */
	(void) nvlist_lookup_uint32(nvl, "type", &ustype);
	if (!(ustype & types))
	return;

	while ((field = fields[cfield]) != USFIELD_LAST) {
	nvpair_t *nvp = NULL;
	data_type_t type;
	uint32_t val32 = -1;
	uint64_t val64 = -1;
	const char *strval = "-";

	while ((nvp = nvlist_next_nvpair(nvl, nvp)) != NULL)
	if (strcmp(nvpair_name(nvp),
	us_field_names[field]) == 0)
	break;

	type = nvp == NULL ? DATA_TYPE_UNKNOWN : nvpair_type(nvp);
	switch (type) {
	case DATA_TYPE_UINT32:
	val32 = fnvpair_value_uint32(nvp);
	break;
	case DATA_TYPE_UINT64:
	val64 = fnvpair_value_uint64(nvp);
	break;
	case DATA_TYPE_STRING:
	strval = fnvpair_value_string(nvp);
	break;
	case DATA_TYPE_UNKNOWN:
	break;
	default:
	(void) fprintf(stderr, "invalid data type\n");
	}

	switch (field) {
	case USFIELD_TYPE:
	if (type == DATA_TYPE_UINT32)
	strval = us_type2str(val32);
	break;
	case USFIELD_NAME:
	if (type == DATA_TYPE_UINT64) {
	(void) sprintf(valstr, "%llu",
	(u_longlong_t)val64);
	strval = valstr;
	}
	break;
	case USFIELD_USED:
	case USFIELD_QUOTA:
	if (type == DATA_TYPE_UINT64) {
	if (parsable) {
	(void) sprintf(valstr, "%llu",
	(u_longlong_t)val64);
	strval = valstr;
	} else if (field == USFIELD_QUOTA &&
	val64 == 0) {
	strval = "none";
	} else {
	zfs_nicebytes(val64, valstr,
	sizeof (valstr));
	strval = valstr;
	}
	}
	break;
	case USFIELD_OBJUSED:
	case USFIELD_OBJQUOTA:
	if (type == DATA_TYPE_UINT64) {
	if (parsable) {
	(void) sprintf(valstr, "%llu",
	(u_longlong_t)val64);
	strval = valstr;
	} else if (field == USFIELD_OBJQUOTA &&
	val64 == 0) {
	strval = "none";
	} else {
	zfs_nicenum(val64, valstr,
	sizeof (valstr));
	strval = valstr;
	}
	}
	break;
	}

	if (!first) {
	if (scripted)
	(void) putchar('\t');
	else
	(void) fputs(" ", stdout);
	}
	if (scripted)
	(void) fputs(strval, stdout);
	else if (field == USFIELD_TYPE \|\| field == USFIELD_NAME)
	(void) printf("%-*s", (int)width[field], strval);
	else
	(void) printf("%*s", (int)width[field], strval);

	first = B_FALSE;
	cfield++;
	}

	(void) putchar('\n');
	}

	static void
	print_us(boolean_t scripted, boolean_t parsable, int *fields, int types,
	size_t width, boolean_t rmnode, uu_avl_t avl)
	{
	us_node_t *node;
	const char *col;
	int cfield = 0;
	int field;

	if (!scripted) {
	boolean_t first = B_TRUE;

	while ((field = fields[cfield]) != USFIELD_LAST) {
	col = gettext(us_field_hdr[field]);
	if (field == USFIELD_TYPE \|\| field == USFIELD_NAME) {
	(void) printf(first ? "%-s" : " %-s",
	(int)width[field], col);
	} else {
	(void) printf(first ? "%s" : " %s",
	(int)width[field], col);
	}
	first = B_FALSE;
	cfield++;
	}
	(void) printf("\n");
	}

	for (node = uu_avl_first(avl); node; node = uu_avl_next(avl, node)) {
	print_us_node(scripted, parsable, fields, types, width, node);
	if (rmnode)
	nvlist_free(node->usn_nvl);
	}
	}

	static int
	zfs_do_userspace(int argc, char **argv)
	{
	zfs_handle_t *zhp;
	zfs_userquota_prop_t p;
	uu_avl_pool_t *avl_pool;
	uu_avl_t *avl_tree;
	uu_avl_walk_t *walk;
	char *delim;
	char deffields[] = "type,name,used,quota,objused,objquota";
	char *ofield = NULL;
	char *tfield = NULL;
	int cfield = 0;
	int fields[256];
	int i;
	boolean_t scripted = B_FALSE;
	boolean_t prtnum = B_FALSE;
	boolean_t parsable = B_FALSE;
	boolean_t sid2posix = B_FALSE;
	int ret = 0;
	int c;
	zfs_sort_column_t *sortcol = NULL;
	int types = USTYPE_PSX_USR \| USTYPE_SMB_USR;
	us_cbdata_t cb;
	us_node_t *node;
	us_node_t *rmnode;
	uu_list_pool_t *listpool;
	uu_list_t *list;
	uu_avl_index_t idx = 0;
	uu_list_index_t idx2 = 0;

	if (argc < 2)
	usage(B_FALSE);

	if (strcmp(argv[0], "groupspace") == 0) {
	/* Toggle default group types */
	types = USTYPE_PSX_GRP \| USTYPE_SMB_GRP;
	} else if (strcmp(argv[0], "projectspace") == 0) {
	types = USTYPE_PROJ;
	prtnum = B_TRUE;
	}

	while ((c = getopt(argc, argv, "nHpo:s:S:t:i")) != -1) {
	switch (c) {
	case 'n':
	if (types == USTYPE_PROJ) {
	(void) fprintf(stderr,
	gettext("invalid option 'n'\n"));
	usage(B_FALSE);
	}
	prtnum = B_TRUE;
	break;
	case 'H':
	scripted = B_TRUE;
	break;
	case 'p':
	parsable = B_TRUE;
	break;
	case 'o':
	ofield = optarg;
	break;
	case 's':
	case 'S':
	if (zfs_add_sort_column(&sortcol, optarg,
	c == 's' ? B_FALSE : B_TRUE) != 0) {
	(void) fprintf(stderr,
	gettext("invalid field '%s'\n"), optarg);
	usage(B_FALSE);
	}
	break;
	case 't':
	if (types == USTYPE_PROJ) {
	(void) fprintf(stderr,
	gettext("invalid option 't'\n"));
	usage(B_FALSE);
	}
	tfield = optarg;
	break;
	case 'i':
	if (types == USTYPE_PROJ) {
	(void) fprintf(stderr,
	gettext("invalid option 'i'\n"));
	usage(B_FALSE);
	}
	sid2posix = B_TRUE;
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	usage(B_FALSE);
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing dataset name\n"));
	usage(B_FALSE);
	}
	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	/* Use default output fields if not specified using -o */
	if (ofield == NULL)
	ofield = deffields;
	do {
	if ((delim = strchr(ofield, ',')) != NULL)
	*delim = '\0';
	if ((fields[cfield++] = us_field_index(ofield)) == -1) {
	(void) fprintf(stderr, gettext("invalid type '%s' "
	"for -o option\n"), ofield);
	return (-1);
	}
	if (delim != NULL)
	ofield = delim + 1;
	} while (delim != NULL);
	fields[cfield] = USFIELD_LAST;

	/* Override output types (-t option) */
	if (tfield != NULL) {
	types = 0;

	do {
	boolean_t found = B_FALSE;

	if ((delim = strchr(tfield, ',')) != NULL)
	*delim = '\0';
	for (i = 0; i < sizeof (us_type_bits) / sizeof (int);
	i++) {
	if (strcmp(tfield, us_type_names[i]) == 0) {
	found = B_TRUE;
	types \|= us_type_bits[i];
	break;
	}
	}
	if (!found) {
	(void) fprintf(stderr, gettext("invalid type "
	"'%s' for -t option\n"), tfield);
	return (-1);
	}
	if (delim != NULL)
	tfield = delim + 1;
	} while (delim != NULL);
	}

	if ((zhp = zfs_path_to_zhandle(g_zfs, argv[0], ZFS_TYPE_FILESYSTEM \|
	ZFS_TYPE_SNAPSHOT)) == NULL)
	return (1);
	if (zfs_get_underlying_type(zhp) != ZFS_TYPE_FILESYSTEM) {
	(void) fprintf(stderr, gettext("operation is only applicable "
	"to filesystems and their snapshots\n"));
	zfs_close(zhp);
	return (1);
	}

	if ((avl_pool = uu_avl_pool_create("us_avl_pool", sizeof (us_node_t),
	offsetof(us_node_t, usn_avlnode), us_compare, UU_DEFAULT)) == NULL)
	nomem();
	if ((avl_tree = uu_avl_create(avl_pool, NULL, UU_DEFAULT)) == NULL)
	nomem();

	/* Always add default sorting columns */
	(void) zfs_add_sort_column(&sortcol, "type", B_FALSE);
	(void) zfs_add_sort_column(&sortcol, "name", B_FALSE);

	cb.cb_sortcol = sortcol;
	cb.cb_numname = prtnum;
	cb.cb_nicenum = !parsable;
	cb.cb_avl_pool = avl_pool;
	cb.cb_avl = avl_tree;
	cb.cb_sid2posix = sid2posix;

	for (i = 0; i < USFIELD_LAST; i++)
	cb.cb_width[i] = strlen(gettext(us_field_hdr[i]));

	for (p = 0; p < ZFS_NUM_USERQUOTA_PROPS; p++) {
	if ((zfs_prop_is_user(p) &&
	!(types & (USTYPE_PSX_USR \| USTYPE_SMB_USR))) \|\|
	(zfs_prop_is_group(p) &&
	!(types & (USTYPE_PSX_GRP \| USTYPE_SMB_GRP))) \|\|
	(zfs_prop_is_project(p) && types != USTYPE_PROJ))
	continue;

	cb.cb_prop = p;
	if ((ret = zfs_userspace(zhp, p, userspace_cb, &cb)) != 0) {
	zfs_close(zhp);
	return (ret);
	}
	}
	zfs_close(zhp);

	/* Sort the list */
	if ((node = uu_avl_first(avl_tree)) == NULL)
	return (0);

	us_populated = B_TRUE;

	listpool = uu_list_pool_create("tmplist", sizeof (us_node_t),
	offsetof(us_node_t, usn_listnode), NULL, UU_DEFAULT);
	list = uu_list_create(listpool, NULL, UU_DEFAULT);
	uu_list_node_init(node, &node->usn_listnode, listpool);

	while (node != NULL) {
	rmnode = node;
	node = uu_avl_next(avl_tree, node);
	uu_avl_remove(avl_tree, rmnode);
	if (uu_list_find(list, rmnode, NULL, &idx2) == NULL)
	uu_list_insert(list, rmnode, idx2);
	}

	for (node = uu_list_first(list); node != NULL;
	node = uu_list_next(list, node)) {
	us_sort_info_t sortinfo = { sortcol, cb.cb_numname };

	if (uu_avl_find(avl_tree, node, &sortinfo, &idx) == NULL)
	uu_avl_insert(avl_tree, node, idx);
	}

	uu_list_destroy(list);
	uu_list_pool_destroy(listpool);

	/* Print and free node nvlist memory */
	print_us(scripted, parsable, fields, types, cb.cb_width, B_TRUE,
	cb.cb_avl);

	zfs_free_sort_columns(sortcol);

	/* Clean up the AVL tree */
	if ((walk = uu_avl_walk_start(cb.cb_avl, UU_WALK_ROBUST)) == NULL)
	nomem();

	while ((node = uu_avl_walk_next(walk)) != NULL) {
	uu_avl_remove(cb.cb_avl, node);
	free(node);
	}

	uu_avl_walk_end(walk);
	uu_avl_destroy(avl_tree);
	uu_avl_pool_destroy(avl_pool);

	return (ret);
	}

	/*
	* list [-Hp][-r\|-d max] [-o property[,...]] [-s property] ... [-S property]
	* [-t type[,...]] [filesystem\|volume\|snapshot] ...
	*
	* -H Scripted mode; elide headers and separate columns by tabs
	* -p Display values in parsable (literal) format.
	* -r Recurse over all children
	* -d Limit recursion by depth.
	* -o Control which fields to display.
	* -s Specify sort columns, descending order.
	* -S Specify sort columns, ascending order.
	* -t Control which object types to display.
	*
	* When given no arguments, list all filesystems in the system.
	* Otherwise, list the specified datasets, optionally recursing down them if
	* '-r' is specified.
	*/
	typedef struct list_cbdata {
	boolean_t cb_first;
	boolean_t cb_literal;
	boolean_t cb_scripted;
	zprop_list_t *cb_proplist;
	} list_cbdata_t;

	/*
	* Given a list of columns to display, output appropriate headers for each one.
	*/
	static void
	print_header(list_cbdata_t *cb)
	{
	zprop_list_t *pl = cb->cb_proplist;
	char headerbuf[ZFS_MAXPROPLEN];
	const char *header;
	int i;
	boolean_t first = B_TRUE;
	boolean_t right_justify;

	color_start(ANSI_BOLD);

	for (; pl != NULL; pl = pl->pl_next) {
	if (!first) {
	(void) printf(" ");
	} else {
	first = B_FALSE;
	}

	right_justify = B_FALSE;
	if (pl->pl_prop != ZPROP_USERPROP) {
	header = zfs_prop_column_name(pl->pl_prop);
	right_justify = zfs_prop_align_right(pl->pl_prop);
	} else {
	for (i = 0; pl->pl_user_prop[i] != '\0'; i++)
	headerbuf[i] = toupper(pl->pl_user_prop[i]);
	headerbuf[i] = '\0';
	header = headerbuf;
	}

	if (pl->pl_next == NULL && !right_justify)
	(void) printf("%s", header);
	else if (right_justify)
	(void) printf("%*s", (int)pl->pl_width, header);
	else
	(void) printf("%-*s", (int)pl->pl_width, header);
	}

	color_end();

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

	/*
	* Decides on the color that the avail value should be printed in.
	* > 80% used = yellow
	* > 90% used = red
	*/
	static const char *
	zfs_list_avail_color(zfs_handle_t *zhp)
	{
	uint64_t used = zfs_prop_get_int(zhp, ZFS_PROP_USED);
	uint64_t avail = zfs_prop_get_int(zhp, ZFS_PROP_AVAILABLE);
	int percentage = (int)((double)avail / MAX(avail + used, 1) * 100);

	if (percentage > 20)
	return (NULL);
	else if (percentage > 10)
	return (ANSI_YELLOW);
	else
	return (ANSI_RED);
	}

	/*
	* Given a dataset and a list of fields, print out all the properties according
	* to the described layout.
	*/
	static void
	print_dataset(zfs_handle_t zhp, list_cbdata_t cb)
	{
	zprop_list_t *pl = cb->cb_proplist;
	boolean_t first = B_TRUE;
	char property[ZFS_MAXPROPLEN];
	nvlist_t *userprops = zfs_get_user_props(zhp);
	nvlist_t *propval;
	const char *propstr;
	boolean_t right_justify;

	for (; pl != NULL; pl = pl->pl_next) {
	if (!first) {
	if (cb->cb_scripted)
	(void) putchar('\t');
	else
	(void) fputs(" ", stdout);
	} else {
	first = B_FALSE;
	}

	if (pl->pl_prop == ZFS_PROP_NAME) {
	(void) strlcpy(property, zfs_get_name(zhp),
	sizeof (property));
	propstr = property;
	right_justify = zfs_prop_align_right(pl->pl_prop);
	} else if (pl->pl_prop != ZPROP_USERPROP) {
	if (zfs_prop_get(zhp, pl->pl_prop, property,
	sizeof (property), NULL, NULL, 0,
	cb->cb_literal) != 0)
	propstr = "-";
	else
	propstr = property;
	right_justify = zfs_prop_align_right(pl->pl_prop);
	} else if (zfs_prop_userquota(pl->pl_user_prop)) {
	if (zfs_prop_get_userquota(zhp, pl->pl_user_prop,
	property, sizeof (property), cb->cb_literal) != 0)
	propstr = "-";
	else
	propstr = property;
	right_justify = B_TRUE;
	} else if (zfs_prop_written(pl->pl_user_prop)) {
	if (zfs_prop_get_written(zhp, pl->pl_user_prop,
	property, sizeof (property), cb->cb_literal) != 0)
	propstr = "-";
	else
	propstr = property;
	right_justify = B_TRUE;
	} else {
	if (nvlist_lookup_nvlist(userprops,
	pl->pl_user_prop, &propval) != 0)
	propstr = "-";
	else
	propstr = fnvlist_lookup_string(propval,
	ZPROP_VALUE);
	right_justify = B_FALSE;
	}

	/*
	* zfs_list_avail_color() needs ZFS_PROP_AVAILABLE + USED
	* - so we need another for() search for the USED part
	* - when no colors wanted, we can skip the whole thing
	*/
	if (use_color() && pl->pl_prop == ZFS_PROP_AVAILABLE) {
	zprop_list_t *pl2 = cb->cb_proplist;
	for (; pl2 != NULL; pl2 = pl2->pl_next) {
	if (pl2->pl_prop == ZFS_PROP_USED) {
	color_start(zfs_list_avail_color(zhp));
	/* found it, no need for more loops */
	break;
	}
	}
	}

	/*
	* If this is being called in scripted mode, or if this is the
	* last column and it is left-justified, don't include a width
	* format specifier.
	*/
	if (cb->cb_scripted \|\| (pl->pl_next == NULL && !right_justify))
	(void) fputs(propstr, stdout);
	else if (right_justify)
	(void) printf("%*s", (int)pl->pl_width, propstr);
	else
	(void) printf("%-*s", (int)pl->pl_width, propstr);

	if (pl->pl_prop == ZFS_PROP_AVAILABLE)
	color_end();
	}

	(void) putchar('\n');
	}

	/*
	* Generic callback function to list a dataset or snapshot.
	*/
	static int
	list_callback(zfs_handle_t zhp, void data)
	{
	list_cbdata_t *cbp = data;

	if (cbp->cb_first) {
	if (!cbp->cb_scripted)
	print_header(cbp);
	cbp->cb_first = B_FALSE;
	}

	print_dataset(zhp, cbp);

	return (0);
	}

	static int
	zfs_do_list(int argc, char **argv)
	{
	int c;
	char default_fields[] =
	"name,used,available,referenced,mountpoint";
	int types = ZFS_TYPE_DATASET;
	boolean_t types_specified = B_FALSE;
	char *fields = default_fields;
	list_cbdata_t cb = { 0 };
	int limit = 0;
	int ret = 0;
	zfs_sort_column_t *sortcol = NULL;
	int flags = ZFS_ITER_PROP_LISTSNAPS \| ZFS_ITER_ARGS_CAN_BE_PATHS;

	/* check options */
	while ((c = getopt(argc, argv, "HS:d:o:prs:t:")) != -1) {
	switch (c) {
	case 'o':
	fields = optarg;
	break;
	case 'p':
	cb.cb_literal = B_TRUE;
	flags \|= ZFS_ITER_LITERAL_PROPS;
	break;
	case 'd':
	limit = parse_depth(optarg, &flags);
	break;
	case 'r':
	flags \|= ZFS_ITER_RECURSE;
	break;
	case 'H':
	cb.cb_scripted = B_TRUE;
	break;
	case 's':
	if (zfs_add_sort_column(&sortcol, optarg,
	B_FALSE) != 0) {
	(void) fprintf(stderr,
	gettext("invalid property '%s'\n"), optarg);
	usage(B_FALSE);
	}
	break;
	case 'S':
	if (zfs_add_sort_column(&sortcol, optarg,
	B_TRUE) != 0) {
	(void) fprintf(stderr,
	gettext("invalid property '%s'\n"), optarg);
	usage(B_FALSE);
	}
	break;
	case 't':
	types = 0;
	types_specified = B_TRUE;
	flags &= ~ZFS_ITER_PROP_LISTSNAPS;

	for (char *tok; (tok = strsep(&optarg, ",")); ) {
	static const char *const type_subopts[] = {
	"filesystem",
	"fs",
	"volume",
	"vol",
	"snapshot",
	"snap",
	"bookmark",
	"all"
	};
	static const int type_types[] = {
	ZFS_TYPE_FILESYSTEM,
	ZFS_TYPE_FILESYSTEM,
	ZFS_TYPE_VOLUME,
	ZFS_TYPE_VOLUME,
	ZFS_TYPE_SNAPSHOT,
	ZFS_TYPE_SNAPSHOT,
	ZFS_TYPE_BOOKMARK,
	ZFS_TYPE_DATASET \| ZFS_TYPE_BOOKMARK
	};

	for (c = 0; c < ARRAY_SIZE(type_subopts); ++c)
	if (strcmp(tok, type_subopts[c]) == 0) {
	types \|= type_types[c];
	goto found3;
	}

	(void) fprintf(stderr,
	gettext("invalid type '%s'\n"), tok);
	usage(B_FALSE);
	found3:;
	}
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	usage(B_FALSE);
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/*
	* If "-o space" and no types were specified, don't display snapshots.
	*/
	if (strcmp(fields, "space") == 0 && types_specified == B_FALSE)
	types &= ~ZFS_TYPE_SNAPSHOT;

	/*
	* Handle users who want to list all snapshots or bookmarks
	* of the current dataset (ex. 'zfs list -t snapshot <dataset>').
	*/
	if ((types == ZFS_TYPE_SNAPSHOT \|\| types == ZFS_TYPE_BOOKMARK) &&
	argc > 0 && (flags & ZFS_ITER_RECURSE) == 0 && limit == 0) {
	flags \|= (ZFS_ITER_DEPTH_LIMIT \| ZFS_ITER_RECURSE);
	limit = 1;
	}

	/*
	* If the user specifies '-o all', the zprop_get_list() doesn't
	* normally include the name of the dataset. For 'zfs list', we always
	* want this property to be first.
	*/
	if (zprop_get_list(g_zfs, fields, &cb.cb_proplist, ZFS_TYPE_DATASET)
	!= 0)
	usage(B_FALSE);

	cb.cb_first = B_TRUE;

	/*
	* If we are only going to list and sort by properties that are "fast"
	* then we can use "simple" mode and avoid populating the properties
	* nvlist.
	*/
	if (zfs_list_only_by_fast(cb.cb_proplist) &&
	zfs_sort_only_by_fast(sortcol))
	flags \|= ZFS_ITER_SIMPLE;

	ret = zfs_for_each(argc, argv, flags, types, sortcol, &cb.cb_proplist,
	limit, list_callback, &cb);

	zprop_free_list(cb.cb_proplist);
	zfs_free_sort_columns(sortcol);

	if (ret == 0 && cb.cb_first && !cb.cb_scripted)
	(void) fprintf(stderr, gettext("no datasets available\n"));

	return (ret);
	}

	/*
	* zfs rename [-fu] <fs \| snap \| vol> <fs \| snap \| vol>
	* zfs rename [-f] -p <fs \| vol> <fs \| vol>
	* zfs rename [-u] -r <snap> <snap>
	*
	* Renames the given dataset to another of the same type.
	*
	* The '-p' flag creates all the non-existing ancestors of the target first.
	* The '-u' flag prevents file systems from being remounted during rename.
	*/
	static int
	zfs_do_rename(int argc, char **argv)
	{
	zfs_handle_t *zhp;
	renameflags_t flags = { 0 };
	int c;
	int ret = 0;
	int types;
	boolean_t parents = B_FALSE;

	/* check options */
	while ((c = getopt(argc, argv, "pruf")) != -1) {
	switch (c) {
	case 'p':
	parents = B_TRUE;
	break;
	case 'r':
	flags.recursive = B_TRUE;
	break;
	case 'u':
	flags.nounmount = B_TRUE;
	break;
	case 'f':
	flags.forceunmount = B_TRUE;
	break;
	case '?':
	default:
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing source dataset "
	"argument\n"));
	usage(B_FALSE);
	}
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing target dataset "
	"argument\n"));
	usage(B_FALSE);
	}
	if (argc > 2) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	if (flags.recursive && parents) {
	(void) fprintf(stderr, gettext("-p and -r options are mutually "
	"exclusive\n"));
	usage(B_FALSE);
	}

	if (flags.nounmount && parents) {
	(void) fprintf(stderr, gettext("-u and -p options are mutually "
	"exclusive\n"));
	usage(B_FALSE);
	}

	if (flags.recursive && strchr(argv[0], '@') == 0) {
	(void) fprintf(stderr, gettext("source dataset for recursive "
	"rename must be a snapshot\n"));
	usage(B_FALSE);
	}

	if (flags.nounmount)
	types = ZFS_TYPE_FILESYSTEM;
	else if (parents)
	types = ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME;
	else
	types = ZFS_TYPE_DATASET;

	if ((zhp = zfs_open(g_zfs, argv[0], types)) == NULL)
	return (1);

	/* If we were asked and the name looks good, try to create ancestors. */
	if (parents && zfs_name_valid(argv[1], zfs_get_type(zhp)) &&
	zfs_create_ancestors(g_zfs, argv[1]) != 0) {
	zfs_close(zhp);
	return (1);
	}

	ret = (zfs_rename(zhp, argv[1], flags) != 0);

	zfs_close(zhp);
	return (ret);
	}

	/*
	* zfs promote <fs>
	*
	* Promotes the given clone fs to be the parent
	*/
	static int
	zfs_do_promote(int argc, char **argv)
	{
	zfs_handle_t *zhp;
	int ret = 0;

	/* check options */
	if (argc > 1 && argv[1][0] == '-') {
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	argv[1][1]);
	usage(B_FALSE);
	}

	/* check number of arguments */
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing clone filesystem"
	" argument\n"));
	usage(B_FALSE);
	}
	if (argc > 2) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	zhp = zfs_open(g_zfs, argv[1], ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME);
	if (zhp == NULL)
	return (1);

	ret = (zfs_promote(zhp) != 0);


	zfs_close(zhp);
	return (ret);
	}

	static int
	zfs_do_redact(int argc, char **argv)
	{
	char *snap = NULL;
	char *bookname = NULL;
	char **rsnaps = NULL;
	int numrsnaps = 0;
	argv++;
	argc--;
	if (argc < 3) {
	(void) fprintf(stderr, gettext("too few arguments\n"));
	usage(B_FALSE);
	}

	snap = argv[0];
	bookname = argv[1];
	rsnaps = argv + 2;
	numrsnaps = argc - 2;

	nvlist_t *rsnapnv = fnvlist_alloc();

	for (int i = 0; i < numrsnaps; i++) {
	fnvlist_add_boolean(rsnapnv, rsnaps[i]);
	}

	int err = lzc_redact(snap, bookname, rsnapnv);
	fnvlist_free(rsnapnv);

	switch (err) {
	case 0:
	break;
	case ENOENT: {
	zfs_handle_t *zhp = zfs_open(g_zfs, snap, ZFS_TYPE_SNAPSHOT);
	if (zhp == NULL) {
	(void) fprintf(stderr, gettext("provided snapshot %s "
	"does not exist\n"), snap);
	} else {
	zfs_close(zhp);
	}
	for (int i = 0; i < numrsnaps; i++) {
	zhp = zfs_open(g_zfs, rsnaps[i], ZFS_TYPE_SNAPSHOT);
	if (zhp == NULL) {
	(void) fprintf(stderr, gettext("provided "
	"snapshot %s does not exist\n"), rsnaps[i]);
	} else {
	zfs_close(zhp);
	}
	}
	break;
	}
	case EEXIST:
	(void) fprintf(stderr, gettext("specified redaction bookmark "
	"(%s) provided already exists\n"), bookname);
	break;
	case ENAMETOOLONG:
	(void) fprintf(stderr, gettext("provided bookmark name cannot "
	"be used, final name would be too long\n"));
	break;
	case E2BIG:
	(void) fprintf(stderr, gettext("too many redaction snapshots "
	"specified\n"));
	break;
	case EINVAL:
	if (strchr(bookname, '#') != NULL)
	(void) fprintf(stderr, gettext(
	"redaction bookmark name must not contain '#'\n"));
	else
	(void) fprintf(stderr, gettext(
	"redaction snapshot must be descendent of "
	"snapshot being redacted\n"));
	break;
	case EALREADY:
	(void) fprintf(stderr, gettext("attempted to redact redacted "
	"dataset or with respect to redacted dataset\n"));
	break;
	case ENOTSUP:
	(void) fprintf(stderr, gettext("redaction bookmarks feature "
	"not enabled\n"));
	break;
	case EXDEV:
	(void) fprintf(stderr, gettext("potentially invalid redaction "
	"snapshot; full dataset names required\n"));
	break;
	case ESRCH:
	(void) fprintf(stderr, gettext("attempted to resume redaction "
	" with a mismatched redaction list\n"));
	break;
	default:
	(void) fprintf(stderr, gettext("internal error: %s\n"),
	strerror(errno));
	}

	return (err);
	}

	/*
	* zfs rollback [-rRf] <snapshot>
	*
	* -r Delete any intervening snapshots before doing rollback
	* -R Delete any snapshots and their clones
	* -f ignored for backwards compatibility
	*
	* Given a filesystem, rollback to a specific snapshot, discarding any changes
	* since then and making it the active dataset. If more recent snapshots exist,
	* the command will complain unless the '-r' flag is given.
	*/
	typedef struct rollback_cbdata {
	uint64_t cb_create;
	uint8_t cb_younger_ds_printed;
	boolean_t cb_first;
	int cb_doclones;
	char *cb_target;
	int cb_error;
	boolean_t cb_recurse;
	} rollback_cbdata_t;

	static int
	rollback_check_dependent(zfs_handle_t zhp, void data)
	{
	rollback_cbdata_t *cbp = data;

	if (cbp->cb_first && cbp->cb_recurse) {
	(void) fprintf(stderr, gettext("cannot rollback to "
	"'%s': clones of previous snapshots exist\n"),
	cbp->cb_target);
	(void) fprintf(stderr, gettext("use '-R' to "
	"force deletion of the following clones and "
	"dependents:\n"));
	cbp->cb_first = 0;
	cbp->cb_error = 1;
	}

	(void) fprintf(stderr, "%s\n", zfs_get_name(zhp));

	zfs_close(zhp);
	return (0);
	}


	/*
	* Report some snapshots/bookmarks more recent than the one specified.
	* Used when '-r' is not specified. We reuse this same callback for the
	* snapshot dependents - if 'cb_dependent' is set, then this is a
	* dependent and we should report it without checking the transaction group.
	*/
	static int
	rollback_check(zfs_handle_t zhp, void data)
	{
	rollback_cbdata_t *cbp = data;
	/*
	* Max number of younger snapshots and/or bookmarks to display before
	* we stop the iteration.
	*/
	const uint8_t max_younger = 32;

	if (cbp->cb_doclones) {
	zfs_close(zhp);
	return (0);
	}

	if (zfs_prop_get_int(zhp, ZFS_PROP_CREATETXG) > cbp->cb_create) {
	if (cbp->cb_first && !cbp->cb_recurse) {
	(void) fprintf(stderr, gettext("cannot "
	"rollback to '%s': more recent snapshots "
	"or bookmarks exist\n"),
	cbp->cb_target);
	(void) fprintf(stderr, gettext("use '-r' to "
	"force deletion of the following "
	"snapshots and bookmarks:\n"));
	cbp->cb_first = 0;
	cbp->cb_error = 1;
	}

	if (cbp->cb_recurse) {
	if (zfs_iter_dependents_v2(zhp, 0, B_TRUE,
	rollback_check_dependent, cbp) != 0) {
	zfs_close(zhp);
	return (-1);
	}
	} else {
	(void) fprintf(stderr, "%s\n",
	zfs_get_name(zhp));
	cbp->cb_younger_ds_printed++;
	}
	}
	zfs_close(zhp);

	if (cbp->cb_younger_ds_printed == max_younger) {
	/*
	* This non-recursive rollback is going to fail due to the
	* presence of snapshots and/or bookmarks that are younger than
	* the rollback target.
	* We printed some of the offending objects, now we stop
	* zfs_iter_snapshot/bookmark iteration so we can fail fast and
	* avoid iterating over the rest of the younger objects
	*/
	(void) fprintf(stderr, gettext("Output limited to %d "
	"snapshots/bookmarks\n"), max_younger);
	return (-1);
	}
	return (0);
	}

	static int
	zfs_do_rollback(int argc, char **argv)
	{
	int ret = 0;
	int c;
	boolean_t force = B_FALSE;
	rollback_cbdata_t cb = { 0 };
	zfs_handle_t zhp, snap;
	char parentname[ZFS_MAX_DATASET_NAME_LEN];
	char *delim;
	uint64_t min_txg = 0;

	/* check options */
	while ((c = getopt(argc, argv, "rRf")) != -1) {
	switch (c) {
	case 'r':
	cb.cb_recurse = 1;
	break;
	case 'R':
	cb.cb_recurse = 1;
	cb.cb_doclones = 1;
	break;
	case 'f':
	force = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing dataset argument\n"));
	usage(B_FALSE);
	}
	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	/* open the snapshot */
	if ((snap = zfs_open(g_zfs, argv[0], ZFS_TYPE_SNAPSHOT)) == NULL)
	return (1);

	/* open the parent dataset */
	(void) strlcpy(parentname, argv[0], sizeof (parentname));
	verify((delim = strrchr(parentname, '@')) != NULL);
	*delim = '\0';
	if ((zhp = zfs_open(g_zfs, parentname, ZFS_TYPE_DATASET)) == NULL) {
	zfs_close(snap);
	return (1);
	}

	/*
	* Check for more recent snapshots and/or clones based on the presence
	* of '-r' and '-R'.
	*/
	cb.cb_target = argv[0];
	cb.cb_create = zfs_prop_get_int(snap, ZFS_PROP_CREATETXG);
	cb.cb_first = B_TRUE;
	cb.cb_error = 0;

	if (cb.cb_create > 0)
	min_txg = cb.cb_create;

	if ((ret = zfs_iter_snapshots_v2(zhp, 0, rollback_check, &cb,
	min_txg, 0)) != 0)
	goto out;
	if ((ret = zfs_iter_bookmarks_v2(zhp, 0, rollback_check, &cb)) != 0)
	goto out;

	if ((ret = cb.cb_error) != 0)
	goto out;

	/*
	* Rollback parent to the given snapshot.
	*/
	ret = zfs_rollback(zhp, snap, force);

	out:
	zfs_close(snap);
	zfs_close(zhp);

	if (ret == 0)
	return (0);
	else
	return (1);
	}

	/*
	* zfs set property=value ... { fs \| snap \| vol } ...
	*
	* Sets the given properties for all datasets specified on the command line.
	*/

	static int
	set_callback(zfs_handle_t zhp, void data)
	{
	zprop_set_cbdata_t *cb = data;
	int ret = zfs_prop_set_list_flags(zhp, cb->cb_proplist, cb->cb_flags);

	if (ret != 0 \|\| libzfs_errno(g_zfs) != EZFS_SUCCESS) {
	switch (libzfs_errno(g_zfs)) {
	case EZFS_MOUNTFAILED:
	(void) fprintf(stderr, gettext("property may be set "
	"but unable to remount filesystem\n"));
	break;
	case EZFS_SHARENFSFAILED:
	(void) fprintf(stderr, gettext("property may be set "
	"but unable to reshare filesystem\n"));
	break;
	}
	}
	return (ret);
	}

	static int
	zfs_do_set(int argc, char **argv)
	{
	zprop_set_cbdata_t cb = { 0 };
	int ds_start = -1; /* argv idx of first dataset arg */
	int ret = 0;
	int i, c;

	/* check options */
	while ((c = getopt(argc, argv, "u")) != -1) {
	switch (c) {
	case 'u':
	cb.cb_flags \|= ZFS_SET_NOMOUNT;
	break;
	case '?':
	default:
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing arguments\n"));
	usage(B_FALSE);
	}
	if (argc < 2) {
	if (strchr(argv[0], '=') == NULL) {
	(void) fprintf(stderr, gettext("missing property=value "
	"argument(s)\n"));
	} else {
	(void) fprintf(stderr, gettext("missing dataset "
	"name(s)\n"));
	}
	usage(B_FALSE);
	}

	/* validate argument order: prop=val args followed by dataset args */
	for (i = 0; i < argc; i++) {
	if (strchr(argv[i], '=') != NULL) {
	if (ds_start > 0) {
	/* out-of-order prop=val argument */
	(void) fprintf(stderr, gettext("invalid "
	"argument order\n"));
	usage(B_FALSE);
	}
	} else if (ds_start < 0) {
	ds_start = i;
	}
	}
	if (ds_start < 0) {
	(void) fprintf(stderr, gettext("missing dataset name(s)\n"));
	usage(B_FALSE);
	}

	/* Populate a list of property settings */
	if (nvlist_alloc(&cb.cb_proplist, NV_UNIQUE_NAME, 0) != 0)
	nomem();
	for (i = 0; i < ds_start; i++) {
	if (!parseprop(cb.cb_proplist, argv[i])) {
	ret = -1;
	goto error;
	}
	}

	ret = zfs_for_each(argc - ds_start, argv + ds_start, 0,
	ZFS_TYPE_DATASET, NULL, NULL, 0, set_callback, &cb);

	error:
	nvlist_free(cb.cb_proplist);
	return (ret);
	}

	typedef struct snap_cbdata {
	nvlist_t *sd_nvl;
	boolean_t sd_recursive;
	const char *sd_snapname;
	} snap_cbdata_t;

	static int
	zfs_snapshot_cb(zfs_handle_t zhp, void arg)
	{
	snap_cbdata_t *sd = arg;
	char *name;
	int rv = 0;
	int error;

	if (sd->sd_recursive &&
	zfs_prop_get_int(zhp, ZFS_PROP_INCONSISTENT) != 0) {
	zfs_close(zhp);
	return (0);
	}

	error = asprintf(&name, "%s@%s", zfs_get_name(zhp), sd->sd_snapname);
	if (error == -1)
	nomem();
	fnvlist_add_boolean(sd->sd_nvl, name);
	free(name);

	if (sd->sd_recursive)
	rv = zfs_iter_filesystems_v2(zhp, 0, zfs_snapshot_cb, sd);
	zfs_close(zhp);
	return (rv);
	}

	/*
	* zfs snapshot [-r] [-o prop=value] ... <fs@snap>
	*
	* Creates a snapshot with the given name. While functionally equivalent to
	* 'zfs create', it is a separate command to differentiate intent.
	*/
	static int
	zfs_do_snapshot(int argc, char **argv)
	{
	int ret = 0;
	int c;
	nvlist_t *props;
	snap_cbdata_t sd = { 0 };
	boolean_t multiple_snaps = B_FALSE;

	if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0)
	nomem();
	if (nvlist_alloc(&sd.sd_nvl, NV_UNIQUE_NAME, 0) != 0)
	nomem();

	/* check options */
	while ((c = getopt(argc, argv, "ro:")) != -1) {
	switch (c) {
	case 'o':
	if (!parseprop(props, optarg)) {
	nvlist_free(sd.sd_nvl);
	nvlist_free(props);
	return (1);
	}
	break;
	case 'r':
	sd.sd_recursive = B_TRUE;
	multiple_snaps = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	goto usage;
	}
	}

	argc -= optind;
	argv += optind;

	/* check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing snapshot argument\n"));
	goto usage;
	}

	if (argc > 1)
	multiple_snaps = B_TRUE;
	for (; argc > 0; argc--, argv++) {
	char *atp;
	zfs_handle_t *zhp;

	atp = strchr(argv[0], '@');
	if (atp == NULL)
	goto usage;
	*atp = '\0';
	sd.sd_snapname = atp + 1;
	zhp = zfs_open(g_zfs, argv[0],
	ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME);
	if (zhp == NULL)
	goto usage;
	if (zfs_snapshot_cb(zhp, &sd) != 0)
	goto usage;
	}

	ret = zfs_snapshot_nvl(g_zfs, sd.sd_nvl, props);
	nvlist_free(sd.sd_nvl);
	nvlist_free(props);
	if (ret != 0 && multiple_snaps)
	(void) fprintf(stderr, gettext("no snapshots were created\n"));
	return (ret != 0);

	usage:
	nvlist_free(sd.sd_nvl);
	nvlist_free(props);
	usage(B_FALSE);
	return (-1);
	}

	/*
	* Array of prefixes to exclude –
	* a linear search, even if executed for each dataset,
	* is plenty good enough.
	*/
	typedef struct zfs_send_exclude_arg {
	size_t count;
	const char **list;
	} zfs_send_exclude_arg_t;

	static boolean_t
	zfs_do_send_exclude(zfs_handle_t zhp, void context)
	{
	zfs_send_exclude_arg_t *excludes = context;
	const char *name = zfs_get_name(zhp);

	for (size_t i = 0; i < excludes->count; ++i) {
	size_t len = strlen(excludes->list[i]);
	if (strncmp(name, excludes->list[i], len) == 0 &&
	memchr("/@", name[len], sizeof ("/@")))
	return (B_FALSE);
	}

	return (B_TRUE);
	}

	/*
	* Send a backup stream to stdout.
	*/
	static int
	zfs_do_send(int argc, char **argv)
	{
	char *fromname = NULL;
	char *toname = NULL;
	char *resume_token = NULL;
	char *cp;
	zfs_handle_t *zhp;
	sendflags_t flags = { 0 };
	int c, err;
	nvlist_t *dbgnv = NULL;
	char *redactbook = NULL;
	zfs_send_exclude_arg_t excludes = { 0 };

	struct option long_options[] = {
	{"replicate", no_argument, NULL, 'R'},
	{"skip-missing", no_argument, NULL, 's'},
	{"redact", required_argument, NULL, 'd'},
	{"props", no_argument, NULL, 'p'},
	{"parsable", no_argument, NULL, 'P'},
	{"dedup", no_argument, NULL, 'D'},
	{"proctitle", no_argument, NULL, 'V'},
	{"verbose", no_argument, NULL, 'v'},
	{"dryrun", no_argument, NULL, 'n'},
	{"large-block", no_argument, NULL, 'L'},
	{"embed", no_argument, NULL, 'e'},
	{"resume", required_argument, NULL, 't'},
	{"compressed", no_argument, NULL, 'c'},
	{"raw", no_argument, NULL, 'w'},
	{"backup", no_argument, NULL, 'b'},
	{"holds", no_argument, NULL, 'h'},
	{"saved", no_argument, NULL, 'S'},
	{"exclude", required_argument, NULL, 'X'},
	{0, 0, 0, 0}
	};

	/* check options */
	while ((c = getopt_long(argc, argv, ":i:I:RsDpVvnPLeht:cwbd:SX:",
	long_options, NULL)) != -1) {
	switch (c) {
	case 'X':
	for (char *ds; (ds = strsep(&optarg, ",")) != NULL; ) {
	if (!zfs_name_valid(ds, ZFS_TYPE_DATASET) \|\|
	strchr(ds, '/') == NULL) {
	(void) fprintf(stderr, gettext("-X %s: "
	"not a valid non-root dataset name"
	".\n"), ds);
	usage(B_FALSE);
	}
	excludes.list = safe_realloc(excludes.list,
	sizeof (char ) (excludes.count + 1));
	excludes.list[excludes.count++] = ds;
	}
	break;
	case 'i':
	if (fromname)
	usage(B_FALSE);
	fromname = optarg;
	break;
	case 'I':
	if (fromname)
	usage(B_FALSE);
	fromname = optarg;
	flags.doall = B_TRUE;
	break;
	case 'R':
	flags.replicate = B_TRUE;
	break;
	case 's':
	flags.skipmissing = B_TRUE;
	break;
	case 'd':
	redactbook = optarg;
	break;
	case 'p':
	flags.props = B_TRUE;
	break;
	case 'b':
	flags.backup = B_TRUE;
	break;
	case 'h':
	flags.holds = B_TRUE;
	break;
	case 'P':
	flags.parsable = B_TRUE;
	break;
	case 'V':
	flags.progressastitle = B_TRUE;
	break;
	case 'v':
	flags.verbosity++;
	flags.progress = B_TRUE;
	break;
	case 'D':
	(void) fprintf(stderr,
	gettext("WARNING: deduplicated send is no "
	"longer supported. A regular,\n"
	"non-deduplicated stream will be generated.\n\n"));
	break;
	case 'n':
	flags.dryrun = B_TRUE;
	break;
	case 'L':
	flags.largeblock = B_TRUE;
	break;
	case 'e':
	flags.embed_data = B_TRUE;
	break;
	case 't':
	resume_token = optarg;
	break;
	case 'c':
	flags.compress = B_TRUE;
	break;
	case 'w':
	flags.raw = B_TRUE;
	flags.compress = B_TRUE;
	flags.embed_data = B_TRUE;
	flags.largeblock = B_TRUE;
	break;
	case 'S':
	flags.saved = B_TRUE;
	break;
	case ':':
	/*
	* If a parameter was not passed, optopt contains the
	* value that would normally lead us into the
	* appropriate case statement. If it's > 256, then this
	* must be a longopt and we should look at argv to get
	* the string. Otherwise it's just the character, so we
	* should use it directly.
	*/
	if (optopt <= UINT8_MAX) {
	(void) fprintf(stderr,
	gettext("missing argument for '%c' "
	"option\n"), optopt);
	} else {
	(void) fprintf(stderr,
	gettext("missing argument for '%s' "
	"option\n"), argv[optind - 1]);
	}
	free(excludes.list);
	usage(B_FALSE);
	break;
	case '?':
	default:
	/*
	* If an invalid flag was passed, optopt contains the
	* character if it was a short flag, or 0 if it was a
	* longopt.
	*/
	if (optopt != 0) {
	(void) fprintf(stderr,
	gettext("invalid option '%c'\n"), optopt);
	} else {
	(void) fprintf(stderr,
	gettext("invalid option '%s'\n"),
	argv[optind - 1]);

	}
	free(excludes.list);
	usage(B_FALSE);
	}
	}

	if ((flags.parsable \|\| flags.progressastitle) && flags.verbosity == 0)
	flags.verbosity = 1;

	if (excludes.count > 0 && !flags.replicate) {
	free(excludes.list);
	(void) fprintf(stderr, gettext("Cannot specify "
	"dataset exclusion (-X) on a non-recursive "
	"send.\n"));
	return (1);
	}

	argc -= optind;
	argv += optind;

	if (resume_token != NULL) {
	if (fromname != NULL \|\| flags.replicate \|\| flags.props \|\|
	flags.backup \|\| flags.holds \|\|
	flags.saved \|\| redactbook != NULL) {
	free(excludes.list);
	(void) fprintf(stderr,
	gettext("invalid flags combined with -t\n"));
	usage(B_FALSE);
	}
	if (argc > 0) {
	free(excludes.list);
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}
	} else {
	if (argc < 1) {
	free(excludes.list);
	(void) fprintf(stderr,
	gettext("missing snapshot argument\n"));
	usage(B_FALSE);
	}
	if (argc > 1) {
	free(excludes.list);
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}
	}

	if (flags.saved) {
	if (fromname != NULL \|\| flags.replicate \|\| flags.props \|\|
	flags.doall \|\| flags.backup \|\|
	flags.holds \|\| flags.largeblock \|\| flags.embed_data \|\|
	flags.compress \|\| flags.raw \|\| redactbook != NULL) {
	free(excludes.list);

	(void) fprintf(stderr, gettext("incompatible flags "
	"combined with saved send flag\n"));
	usage(B_FALSE);
	}
	if (strchr(argv[0], '@') != NULL) {
	free(excludes.list);

	(void) fprintf(stderr, gettext("saved send must "
	"specify the dataset with partially-received "
	"state\n"));
	usage(B_FALSE);
	}
	}

	if (flags.raw && redactbook != NULL) {
	free(excludes.list);
	(void) fprintf(stderr,
	gettext("Error: raw sends may not be redacted.\n"));
	return (1);
	}

	if (!flags.dryrun && isatty(STDOUT_FILENO)) {
	free(excludes.list);
	(void) fprintf(stderr,
	gettext("Error: Stream can not be written to a terminal.\n"
	"You must redirect standard output.\n"));
	return (1);
	}

	if (flags.saved) {
	zhp = zfs_open(g_zfs, argv[0], ZFS_TYPE_DATASET);
	if (zhp == NULL) {
	free(excludes.list);
	return (1);
	}

	err = zfs_send_saved(zhp, &flags, STDOUT_FILENO,
	resume_token);
	free(excludes.list);
	zfs_close(zhp);
	return (err != 0);
	} else if (resume_token != NULL) {
	free(excludes.list);
	return (zfs_send_resume(g_zfs, &flags, STDOUT_FILENO,
	resume_token));
	}

	if (flags.skipmissing && !flags.replicate) {
	free(excludes.list);
	(void) fprintf(stderr,
	gettext("skip-missing flag can only be used in "
	"conjunction with replicate\n"));
	usage(B_FALSE);
	}

	/*
	* For everything except -R and -I, use the new, cleaner code path.
	*/
	if (!(flags.replicate \|\| flags.doall)) {
	char frombuf[ZFS_MAX_DATASET_NAME_LEN];

	if (fromname != NULL && (strchr(fromname, '#') == NULL &&
	strchr(fromname, '@') == NULL)) {
	/*
	* Neither bookmark or snapshot was specified. Print a
	* warning, and assume snapshot.
	*/
	(void) fprintf(stderr, "Warning: incremental source "
	"didn't specify type, assuming snapshot. Use '@' "
	"or '#' prefix to avoid ambiguity.\n");
	(void) snprintf(frombuf, sizeof (frombuf), "@%s",
	fromname);
	fromname = frombuf;
	}
	if (fromname != NULL &&
	(fromname[0] == '#' \|\| fromname[0] == '@')) {
	/*
	* Incremental source name begins with # or @.
	* Default to same fs as target.
	*/
	char tmpbuf[ZFS_MAX_DATASET_NAME_LEN];
	(void) strlcpy(tmpbuf, fromname, sizeof (tmpbuf));
	(void) strlcpy(frombuf, argv[0], sizeof (frombuf));
	cp = strchr(frombuf, '@');
	if (cp != NULL)
	*cp = '\0';
	(void) strlcat(frombuf, tmpbuf, sizeof (frombuf));
	fromname = frombuf;
	}

	zhp = zfs_open(g_zfs, argv[0], ZFS_TYPE_DATASET);
	if (zhp == NULL) {
	free(excludes.list);
	return (1);
	}
	err = zfs_send_one(zhp, fromname, STDOUT_FILENO, &flags,
	redactbook);

	free(excludes.list);
	zfs_close(zhp);
	return (err != 0);
	}

	if (fromname != NULL && strchr(fromname, '#')) {
	(void) fprintf(stderr,
	gettext("Error: multiple snapshots cannot be "
	"sent from a bookmark.\n"));
	free(excludes.list);
	return (1);
	}

	if (redactbook != NULL) {
	(void) fprintf(stderr, gettext("Error: multiple snapshots "
	"cannot be sent redacted.\n"));
	free(excludes.list);
	return (1);
	}

	if ((cp = strchr(argv[0], '@')) == NULL) {
	(void) fprintf(stderr, gettext("Error: "
	"Unsupported flag with filesystem or bookmark.\n"));
	free(excludes.list);
	return (1);
	}
	*cp = '\0';
	toname = cp + 1;
	zhp = zfs_open(g_zfs, argv[0], ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME);
	if (zhp == NULL) {
	free(excludes.list);
	return (1);
	}

	/*
	* If they specified the full path to the snapshot, chop off
	* everything except the short name of the snapshot, but special
	* case if they specify the origin.
	*/
	if (fromname && (cp = strchr(fromname, '@')) != NULL) {
	char origin[ZFS_MAX_DATASET_NAME_LEN];
	zprop_source_t src;

	(void) zfs_prop_get(zhp, ZFS_PROP_ORIGIN,
	origin, sizeof (origin), &src, NULL, 0, B_FALSE);

	if (strcmp(origin, fromname) == 0) {
	fromname = NULL;
	flags.fromorigin = B_TRUE;
	} else {
	*cp = '\0';
	if (cp != fromname && strcmp(argv[0], fromname)) {
	zfs_close(zhp);
	free(excludes.list);
	(void) fprintf(stderr,
	gettext("incremental source must be "
	"in same filesystem\n"));
	usage(B_FALSE);
	}
	fromname = cp + 1;
	if (strchr(fromname, '@') \|\| strchr(fromname, '/')) {
	zfs_close(zhp);
	free(excludes.list);
	(void) fprintf(stderr,
	gettext("invalid incremental source\n"));
	usage(B_FALSE);
	}
	}
	}

	if (flags.replicate && fromname == NULL)
	flags.doall = B_TRUE;

	err = zfs_send(zhp, fromname, toname, &flags, STDOUT_FILENO,
	excludes.count > 0 ? zfs_do_send_exclude : NULL,
	&excludes, flags.verbosity >= 3 ? &dbgnv : NULL);

	if (flags.verbosity >= 3 && dbgnv != NULL) {
	/*
	* dump_nvlist prints to stdout, but that's been
	* redirected to a file. Make it print to stderr
	* instead.
	*/
	(void) dup2(STDERR_FILENO, STDOUT_FILENO);
	dump_nvlist(dbgnv, 0);
	nvlist_free(dbgnv);
	}

	zfs_close(zhp);
	free(excludes.list);
	return (err != 0);
	}

	/*
	* Restore a backup stream from stdin.
	*/
	static int
	zfs_do_receive(int argc, char **argv)
	{
	int c, err = 0;
	recvflags_t flags = { 0 };
	boolean_t abort_resumable = B_FALSE;
	nvlist_t *props;

	if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0)
	nomem();

	/* check options */
	while ((c = getopt(argc, argv, ":o:x:dehMnuvFsAc")) != -1) {
	switch (c) {
	case 'o':
	if (!parseprop(props, optarg)) {
	nvlist_free(props);
	usage(B_FALSE);
	}
	break;
	case 'x':
	if (!parsepropname(props, optarg)) {
	nvlist_free(props);
	usage(B_FALSE);
	}
	break;
	case 'd':
	if (flags.istail) {
	(void) fprintf(stderr, gettext("invalid option "
	"combination: -d and -e are mutually "
	"exclusive\n"));
	usage(B_FALSE);
	}
	flags.isprefix = B_TRUE;
	break;
	case 'e':
	if (flags.isprefix) {
	(void) fprintf(stderr, gettext("invalid option "
	"combination: -d and -e are mutually "
	"exclusive\n"));
	usage(B_FALSE);
	}
	flags.istail = B_TRUE;
	break;
	case 'h':
	flags.skipholds = B_TRUE;
	break;
	case 'M':
	flags.forceunmount = B_TRUE;
	break;
	case 'n':
	flags.dryrun = B_TRUE;
	break;
	case 'u':
	flags.nomount = B_TRUE;
	break;
	case 'v':
	flags.verbose = B_TRUE;
	break;
	case 's':
	flags.resumable = B_TRUE;
	break;
	case 'F':
	flags.force = B_TRUE;
	break;
	case 'A':
	abort_resumable = B_TRUE;
	break;
	case 'c':
	flags.heal = B_TRUE;
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	usage(B_FALSE);
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* zfs recv -e (use "tail" name) implies -d (remove dataset "head") */
	if (flags.istail)
	flags.isprefix = B_TRUE;

	/* check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing snapshot argument\n"));
	usage(B_FALSE);
	}
	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	if (abort_resumable) {
	if (flags.isprefix \|\| flags.istail \|\| flags.dryrun \|\|
	flags.resumable \|\| flags.nomount) {
	(void) fprintf(stderr, gettext("invalid option\n"));
	usage(B_FALSE);
	}

	char namebuf[ZFS_MAX_DATASET_NAME_LEN];
	(void) snprintf(namebuf, sizeof (namebuf),
	"%s/%%recv", argv[0]);

	if (zfs_dataset_exists(g_zfs, namebuf,
	ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME)) {
	zfs_handle_t *zhp = zfs_open(g_zfs,
	namebuf, ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME);
	if (zhp == NULL) {
	nvlist_free(props);
	return (1);
	}
	err = zfs_destroy(zhp, B_FALSE);
	zfs_close(zhp);
	} else {
	zfs_handle_t *zhp = zfs_open(g_zfs,
	argv[0], ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME);
	if (zhp == NULL)
	usage(B_FALSE);
	if (!zfs_prop_get_int(zhp, ZFS_PROP_INCONSISTENT) \|\|
	zfs_prop_get(zhp, ZFS_PROP_RECEIVE_RESUME_TOKEN,
	NULL, 0, NULL, NULL, 0, B_TRUE) == -1) {
	(void) fprintf(stderr,
	gettext("'%s' does not have any "
	"resumable receive state to abort\n"),
	argv[0]);
	nvlist_free(props);
	zfs_close(zhp);
	return (1);
	}
	err = zfs_destroy(zhp, B_FALSE);
	zfs_close(zhp);
	}
	nvlist_free(props);
	return (err != 0);
	}

	if (isatty(STDIN_FILENO)) {
	(void) fprintf(stderr,
	gettext("Error: Backup stream can not be read "
	"from a terminal.\n"
	"You must redirect standard input.\n"));
	nvlist_free(props);
	return (1);
	}
	err = zfs_receive(g_zfs, argv[0], props, &flags, STDIN_FILENO, NULL);
	nvlist_free(props);

	return (err != 0);
	}

	/*
	* allow/unallow stuff
	*/
	/* copied from zfs/sys/dsl_deleg.h */
	#define ZFS_DELEG_PERM_CREATE "create"
	#define ZFS_DELEG_PERM_DESTROY "destroy"
	#define ZFS_DELEG_PERM_SNAPSHOT "snapshot"
	#define ZFS_DELEG_PERM_ROLLBACK "rollback"
	#define ZFS_DELEG_PERM_CLONE "clone"
	#define ZFS_DELEG_PERM_PROMOTE "promote"
	#define ZFS_DELEG_PERM_RENAME "rename"
	#define ZFS_DELEG_PERM_MOUNT "mount"
	#define ZFS_DELEG_PERM_SHARE "share"
	#define ZFS_DELEG_PERM_SEND "send"
	#define ZFS_DELEG_PERM_RECEIVE "receive"
	#define ZFS_DELEG_PERM_ALLOW "allow"
	#define ZFS_DELEG_PERM_USERPROP "userprop"
	#define ZFS_DELEG_PERM_VSCAN "vscan" /* ??? */
	#define ZFS_DELEG_PERM_USERQUOTA "userquota"
	#define ZFS_DELEG_PERM_GROUPQUOTA "groupquota"
	#define ZFS_DELEG_PERM_USERUSED "userused"
	#define ZFS_DELEG_PERM_GROUPUSED "groupused"
	#define ZFS_DELEG_PERM_USEROBJQUOTA "userobjquota"
	#define ZFS_DELEG_PERM_GROUPOBJQUOTA "groupobjquota"
	#define ZFS_DELEG_PERM_USEROBJUSED "userobjused"
	#define ZFS_DELEG_PERM_GROUPOBJUSED "groupobjused"

	#define ZFS_DELEG_PERM_HOLD "hold"
	#define ZFS_DELEG_PERM_RELEASE "release"
	#define ZFS_DELEG_PERM_DIFF "diff"
	#define ZFS_DELEG_PERM_BOOKMARK "bookmark"
	#define ZFS_DELEG_PERM_LOAD_KEY "load-key"
	#define ZFS_DELEG_PERM_CHANGE_KEY "change-key"

	#define ZFS_DELEG_PERM_PROJECTUSED "projectused"
	#define ZFS_DELEG_PERM_PROJECTQUOTA "projectquota"
	#define ZFS_DELEG_PERM_PROJECTOBJUSED "projectobjused"
	#define ZFS_DELEG_PERM_PROJECTOBJQUOTA "projectobjquota"

	#define ZFS_NUM_DELEG_NOTES ZFS_DELEG_NOTE_NONE

	static zfs_deleg_perm_tab_t zfs_deleg_perm_tbl[] = {
	{ ZFS_DELEG_PERM_ALLOW, ZFS_DELEG_NOTE_ALLOW },
	{ ZFS_DELEG_PERM_CLONE, ZFS_DELEG_NOTE_CLONE },
	{ ZFS_DELEG_PERM_CREATE, ZFS_DELEG_NOTE_CREATE },
	{ ZFS_DELEG_PERM_DESTROY, ZFS_DELEG_NOTE_DESTROY },
	{ ZFS_DELEG_PERM_DIFF, ZFS_DELEG_NOTE_DIFF},
	{ ZFS_DELEG_PERM_HOLD, ZFS_DELEG_NOTE_HOLD },
	{ ZFS_DELEG_PERM_MOUNT, ZFS_DELEG_NOTE_MOUNT },
	{ ZFS_DELEG_PERM_PROMOTE, ZFS_DELEG_NOTE_PROMOTE },
	{ ZFS_DELEG_PERM_RECEIVE, ZFS_DELEG_NOTE_RECEIVE },
	{ ZFS_DELEG_PERM_RELEASE, ZFS_DELEG_NOTE_RELEASE },
	{ ZFS_DELEG_PERM_RENAME, ZFS_DELEG_NOTE_RENAME },
	{ ZFS_DELEG_PERM_ROLLBACK, ZFS_DELEG_NOTE_ROLLBACK },
	{ ZFS_DELEG_PERM_SEND, ZFS_DELEG_NOTE_SEND },
	{ ZFS_DELEG_PERM_SHARE, ZFS_DELEG_NOTE_SHARE },
	{ ZFS_DELEG_PERM_SNAPSHOT, ZFS_DELEG_NOTE_SNAPSHOT },
	{ ZFS_DELEG_PERM_BOOKMARK, ZFS_DELEG_NOTE_BOOKMARK },
	{ ZFS_DELEG_PERM_LOAD_KEY, ZFS_DELEG_NOTE_LOAD_KEY },
	{ ZFS_DELEG_PERM_CHANGE_KEY, ZFS_DELEG_NOTE_CHANGE_KEY },

	{ ZFS_DELEG_PERM_GROUPQUOTA, ZFS_DELEG_NOTE_GROUPQUOTA },
	{ ZFS_DELEG_PERM_GROUPUSED, ZFS_DELEG_NOTE_GROUPUSED },
	{ ZFS_DELEG_PERM_USERPROP, ZFS_DELEG_NOTE_USERPROP },
	{ ZFS_DELEG_PERM_USERQUOTA, ZFS_DELEG_NOTE_USERQUOTA },
	{ ZFS_DELEG_PERM_USERUSED, ZFS_DELEG_NOTE_USERUSED },
	{ ZFS_DELEG_PERM_USEROBJQUOTA, ZFS_DELEG_NOTE_USEROBJQUOTA },
	{ ZFS_DELEG_PERM_USEROBJUSED, ZFS_DELEG_NOTE_USEROBJUSED },
	{ ZFS_DELEG_PERM_GROUPOBJQUOTA, ZFS_DELEG_NOTE_GROUPOBJQUOTA },
	{ ZFS_DELEG_PERM_GROUPOBJUSED, ZFS_DELEG_NOTE_GROUPOBJUSED },
	{ ZFS_DELEG_PERM_PROJECTUSED, ZFS_DELEG_NOTE_PROJECTUSED },
	{ ZFS_DELEG_PERM_PROJECTQUOTA, ZFS_DELEG_NOTE_PROJECTQUOTA },
	{ ZFS_DELEG_PERM_PROJECTOBJUSED, ZFS_DELEG_NOTE_PROJECTOBJUSED },
	{ ZFS_DELEG_PERM_PROJECTOBJQUOTA, ZFS_DELEG_NOTE_PROJECTOBJQUOTA },
	{ NULL, ZFS_DELEG_NOTE_NONE }
	};

	/* permission structure */
	typedef struct deleg_perm {
	zfs_deleg_who_type_t dp_who_type;
	const char *dp_name;
	boolean_t dp_local;
	boolean_t dp_descend;
	} deleg_perm_t;

	/* */
	typedef struct deleg_perm_node {
	deleg_perm_t dpn_perm;

	uu_avl_node_t dpn_avl_node;
	} deleg_perm_node_t;

	typedef struct fs_perm fs_perm_t;

	/* permissions set */
	typedef struct who_perm {
	zfs_deleg_who_type_t who_type;
	const char who_name; / id */
	char who_ug_name[256]; /* user/group name */
	fs_perm_t who_fsperm; / uplink */

	uu_avl_t who_deleg_perm_avl; / permissions */
	} who_perm_t;

	/* */
	typedef struct who_perm_node {
	who_perm_t who_perm;
	uu_avl_node_t who_avl_node;
	} who_perm_node_t;

	typedef struct fs_perm_set fs_perm_set_t;
	/* fs permissions */
	struct fs_perm {
	const char *fsp_name;

	uu_avl_t fsp_sc_avl; / sets,create */
	uu_avl_t fsp_uge_avl; / user,group,everyone */

	fs_perm_set_t fsp_set; / uplink */
	};

	/* */
	typedef struct fs_perm_node {
	fs_perm_t fspn_fsperm;
	uu_avl_t *fspn_avl;

	uu_list_node_t fspn_list_node;
	} fs_perm_node_t;

	/* top level structure */
	struct fs_perm_set {
	uu_list_pool_t *fsps_list_pool;
	uu_list_t fsps_list; / list of fs_perms */

	uu_avl_pool_t *fsps_named_set_avl_pool;
	uu_avl_pool_t *fsps_who_perm_avl_pool;
	uu_avl_pool_t *fsps_deleg_perm_avl_pool;
	};

	static inline const char *
	deleg_perm_type(zfs_deleg_note_t note)
	{
	/* subcommands */
	switch (note) {
	/* SUBCOMMANDS */
	/* OTHER */
	case ZFS_DELEG_NOTE_GROUPQUOTA:
	case ZFS_DELEG_NOTE_GROUPUSED:
	case ZFS_DELEG_NOTE_USERPROP:
	case ZFS_DELEG_NOTE_USERQUOTA:
	case ZFS_DELEG_NOTE_USERUSED:
	case ZFS_DELEG_NOTE_USEROBJQUOTA:
	case ZFS_DELEG_NOTE_USEROBJUSED:
	case ZFS_DELEG_NOTE_GROUPOBJQUOTA:
	case ZFS_DELEG_NOTE_GROUPOBJUSED:
	case ZFS_DELEG_NOTE_PROJECTUSED:
	case ZFS_DELEG_NOTE_PROJECTQUOTA:
	case ZFS_DELEG_NOTE_PROJECTOBJUSED:
	case ZFS_DELEG_NOTE_PROJECTOBJQUOTA:
	/* other */
	return (gettext("other"));
	default:
	return (gettext("subcommand"));
	}
	}

	static int
	who_type2weight(zfs_deleg_who_type_t who_type)
	{
	int res;
	switch (who_type) {
	case ZFS_DELEG_NAMED_SET_SETS:
	case ZFS_DELEG_NAMED_SET:
	res = 0;
	break;
	case ZFS_DELEG_CREATE_SETS:
	case ZFS_DELEG_CREATE:
	res = 1;
	break;
	case ZFS_DELEG_USER_SETS:
	case ZFS_DELEG_USER:
	res = 2;
	break;
	case ZFS_DELEG_GROUP_SETS:
	case ZFS_DELEG_GROUP:
	res = 3;
	break;
	case ZFS_DELEG_EVERYONE_SETS:
	case ZFS_DELEG_EVERYONE:
	res = 4;
	break;
	default:
	res = -1;
	}

	return (res);
	}

	static int
	who_perm_compare(const void larg, const void rarg, void *unused)
	{
	(void) unused;
	const who_perm_node_t *l = larg;
	const who_perm_node_t *r = rarg;
	zfs_deleg_who_type_t ltype = l->who_perm.who_type;
	zfs_deleg_who_type_t rtype = r->who_perm.who_type;
	int lweight = who_type2weight(ltype);
	int rweight = who_type2weight(rtype);
	int res = lweight - rweight;
	if (res == 0)
	res = strncmp(l->who_perm.who_name, r->who_perm.who_name,
	ZFS_MAX_DELEG_NAME-1);

	if (res == 0)
	return (0);
	if (res > 0)
	return (1);
	else
	return (-1);
	}

	static int
	deleg_perm_compare(const void larg, const void rarg, void *unused)
	{
	(void) unused;
	const deleg_perm_node_t *l = larg;
	const deleg_perm_node_t *r = rarg;
	int res = strncmp(l->dpn_perm.dp_name, r->dpn_perm.dp_name,
	ZFS_MAX_DELEG_NAME-1);

	if (res == 0)
	return (0);

	if (res > 0)
	return (1);
	else
	return (-1);
	}

	static inline void
	fs_perm_set_init(fs_perm_set_t *fspset)
	{
	memset(fspset, 0, sizeof (fs_perm_set_t));

	if ((fspset->fsps_list_pool = uu_list_pool_create("fsps_list_pool",
	sizeof (fs_perm_node_t), offsetof(fs_perm_node_t, fspn_list_node),
	NULL, UU_DEFAULT)) == NULL)
	nomem();
	if ((fspset->fsps_list = uu_list_create(fspset->fsps_list_pool, NULL,
	UU_DEFAULT)) == NULL)
	nomem();

	if ((fspset->fsps_named_set_avl_pool = uu_avl_pool_create(
	"named_set_avl_pool", sizeof (who_perm_node_t), offsetof(
	who_perm_node_t, who_avl_node), who_perm_compare,
	UU_DEFAULT)) == NULL)
	nomem();

	if ((fspset->fsps_who_perm_avl_pool = uu_avl_pool_create(
	"who_perm_avl_pool", sizeof (who_perm_node_t), offsetof(
	who_perm_node_t, who_avl_node), who_perm_compare,
	UU_DEFAULT)) == NULL)
	nomem();

	if ((fspset->fsps_deleg_perm_avl_pool = uu_avl_pool_create(
	"deleg_perm_avl_pool", sizeof (deleg_perm_node_t), offsetof(
	deleg_perm_node_t, dpn_avl_node), deleg_perm_compare, UU_DEFAULT))
	== NULL)
	nomem();
	}

	static inline void fs_perm_fini(fs_perm_t *);
	static inline void who_perm_fini(who_perm_t *);

	static inline void
	fs_perm_set_fini(fs_perm_set_t *fspset)
	{
	fs_perm_node_t *node = uu_list_first(fspset->fsps_list);

	while (node != NULL) {
	fs_perm_node_t *next_node =
	uu_list_next(fspset->fsps_list, node);
	fs_perm_t *fsperm = &node->fspn_fsperm;
	fs_perm_fini(fsperm);
	uu_list_remove(fspset->fsps_list, node);
	free(node);
	node = next_node;
	}

	uu_avl_pool_destroy(fspset->fsps_named_set_avl_pool);
	uu_avl_pool_destroy(fspset->fsps_who_perm_avl_pool);
	uu_avl_pool_destroy(fspset->fsps_deleg_perm_avl_pool);
	}

	static inline void
	deleg_perm_init(deleg_perm_t *deleg_perm, zfs_deleg_who_type_t type,
	const char *name)
	{
	deleg_perm->dp_who_type = type;
	deleg_perm->dp_name = name;
	}

	static inline void
	who_perm_init(who_perm_t who_perm, fs_perm_t fsperm,
	zfs_deleg_who_type_t type, const char *name)
	{
	uu_avl_pool_t *pool;
	pool = fsperm->fsp_set->fsps_deleg_perm_avl_pool;

	memset(who_perm, 0, sizeof (who_perm_t));

	if ((who_perm->who_deleg_perm_avl = uu_avl_create(pool, NULL,
	UU_DEFAULT)) == NULL)
	nomem();

	who_perm->who_type = type;
	who_perm->who_name = name;
	who_perm->who_fsperm = fsperm;
	}

	static inline void
	who_perm_fini(who_perm_t *who_perm)
	{
	deleg_perm_node_t *node = uu_avl_first(who_perm->who_deleg_perm_avl);

	while (node != NULL) {
	deleg_perm_node_t *next_node =
	uu_avl_next(who_perm->who_deleg_perm_avl, node);

	uu_avl_remove(who_perm->who_deleg_perm_avl, node);
	free(node);
	node = next_node;
	}

	uu_avl_destroy(who_perm->who_deleg_perm_avl);
	}

	static inline void
	fs_perm_init(fs_perm_t fsperm, fs_perm_set_t fspset, const char *fsname)
	{
	uu_avl_pool_t *nset_pool = fspset->fsps_named_set_avl_pool;
	uu_avl_pool_t *who_pool = fspset->fsps_who_perm_avl_pool;

	memset(fsperm, 0, sizeof (fs_perm_t));

	if ((fsperm->fsp_sc_avl = uu_avl_create(nset_pool, NULL, UU_DEFAULT))
	== NULL)
	nomem();

	if ((fsperm->fsp_uge_avl = uu_avl_create(who_pool, NULL, UU_DEFAULT))
	== NULL)
	nomem();

	fsperm->fsp_set = fspset;
	fsperm->fsp_name = fsname;
	}

	static inline void
	fs_perm_fini(fs_perm_t *fsperm)
	{
	who_perm_node_t *node = uu_avl_first(fsperm->fsp_sc_avl);
	while (node != NULL) {
	who_perm_node_t *next_node = uu_avl_next(fsperm->fsp_sc_avl,
	node);
	who_perm_t *who_perm = &node->who_perm;
	who_perm_fini(who_perm);
	uu_avl_remove(fsperm->fsp_sc_avl, node);
	free(node);
	node = next_node;
	}

	node = uu_avl_first(fsperm->fsp_uge_avl);
	while (node != NULL) {
	who_perm_node_t *next_node = uu_avl_next(fsperm->fsp_uge_avl,
	node);
	who_perm_t *who_perm = &node->who_perm;
	who_perm_fini(who_perm);
	uu_avl_remove(fsperm->fsp_uge_avl, node);
	free(node);
	node = next_node;
	}

	uu_avl_destroy(fsperm->fsp_sc_avl);
	uu_avl_destroy(fsperm->fsp_uge_avl);
	}

	static void
	set_deleg_perm_node(uu_avl_t avl, deleg_perm_node_t node,
	zfs_deleg_who_type_t who_type, const char *name, char locality)
	{
	uu_avl_index_t idx = 0;

	deleg_perm_node_t *found_node = NULL;
	deleg_perm_t *deleg_perm = &node->dpn_perm;

	deleg_perm_init(deleg_perm, who_type, name);

	if ((found_node = uu_avl_find(avl, node, NULL, &idx))
	== NULL)
	uu_avl_insert(avl, node, idx);
	else {
	node = found_node;
	deleg_perm = &node->dpn_perm;
	}


	switch (locality) {
	case ZFS_DELEG_LOCAL:
	deleg_perm->dp_local = B_TRUE;
	break;
	case ZFS_DELEG_DESCENDENT:
	deleg_perm->dp_descend = B_TRUE;
	break;
	case ZFS_DELEG_NA:
	break;
	default:
	assert(B_FALSE); /* invalid locality */
	}
	}

	static inline int
	parse_who_perm(who_perm_t who_perm, nvlist_t nvl, char locality)
	{
	nvpair_t *nvp = NULL;
	fs_perm_set_t *fspset = who_perm->who_fsperm->fsp_set;
	uu_avl_t *avl = who_perm->who_deleg_perm_avl;
	zfs_deleg_who_type_t who_type = who_perm->who_type;

	while ((nvp = nvlist_next_nvpair(nvl, nvp)) != NULL) {
	const char *name = nvpair_name(nvp);
	data_type_t type = nvpair_type(nvp);
	uu_avl_pool_t *avl_pool = fspset->fsps_deleg_perm_avl_pool;
	deleg_perm_node_t *node =
	safe_malloc(sizeof (deleg_perm_node_t));

	VERIFY(type == DATA_TYPE_BOOLEAN);

	uu_avl_node_init(node, &node->dpn_avl_node, avl_pool);
	set_deleg_perm_node(avl, node, who_type, name, locality);
	}

	return (0);
	}

	static inline int
	parse_fs_perm(fs_perm_t fsperm, nvlist_t nvl)
	{
	nvpair_t *nvp = NULL;
	fs_perm_set_t *fspset = fsperm->fsp_set;

	while ((nvp = nvlist_next_nvpair(nvl, nvp)) != NULL) {
	nvlist_t *nvl2 = NULL;
	const char *name = nvpair_name(nvp);
	uu_avl_t *avl = NULL;
	uu_avl_pool_t *avl_pool = NULL;
	zfs_deleg_who_type_t perm_type = name[0];
	char perm_locality = name[1];
	const char *perm_name = name + 3;
	who_perm_t *who_perm = NULL;

	assert('$' == name[2]);

	if (nvpair_value_nvlist(nvp, &nvl2) != 0)
	return (-1);

	switch (perm_type) {
	case ZFS_DELEG_CREATE:
	case ZFS_DELEG_CREATE_SETS:
	case ZFS_DELEG_NAMED_SET:
	case ZFS_DELEG_NAMED_SET_SETS:
	avl_pool = fspset->fsps_named_set_avl_pool;
	avl = fsperm->fsp_sc_avl;
	break;
	case ZFS_DELEG_USER:
	case ZFS_DELEG_USER_SETS:
	case ZFS_DELEG_GROUP:
	case ZFS_DELEG_GROUP_SETS:
	case ZFS_DELEG_EVERYONE:
	case ZFS_DELEG_EVERYONE_SETS:
	avl_pool = fspset->fsps_who_perm_avl_pool;
	avl = fsperm->fsp_uge_avl;
	break;

	default:
	assert(!"unhandled zfs_deleg_who_type_t");
	}

	who_perm_node_t *found_node = NULL;
	who_perm_node_t *node = safe_malloc(
	sizeof (who_perm_node_t));
	who_perm = &node->who_perm;
	uu_avl_index_t idx = 0;

	uu_avl_node_init(node, &node->who_avl_node, avl_pool);
	who_perm_init(who_perm, fsperm, perm_type, perm_name);

	if ((found_node = uu_avl_find(avl, node, NULL, &idx))
	== NULL) {
	if (avl == fsperm->fsp_uge_avl) {
	uid_t rid = 0;
	struct passwd *p = NULL;
	struct group *g = NULL;
	const char *nice_name = NULL;

	switch (perm_type) {
	case ZFS_DELEG_USER_SETS:
	case ZFS_DELEG_USER:
	rid = atoi(perm_name);
	p = getpwuid(rid);
	if (p)
	nice_name = p->pw_name;
	break;
	case ZFS_DELEG_GROUP_SETS:
	case ZFS_DELEG_GROUP:
	rid = atoi(perm_name);
	g = getgrgid(rid);
	if (g)
	nice_name = g->gr_name;
	break;

	default:
	break;
	}

	if (nice_name != NULL) {
	(void) strlcpy(
	node->who_perm.who_ug_name,
	nice_name, 256);
	} else {
	/* User or group unknown */
	(void) snprintf(
	node->who_perm.who_ug_name,
	sizeof (node->who_perm.who_ug_name),
	"(unknown: %d)", rid);
	}
	}

	uu_avl_insert(avl, node, idx);
	} else {
	node = found_node;
	who_perm = &node->who_perm;
	}

	assert(who_perm != NULL);
	(void) parse_who_perm(who_perm, nvl2, perm_locality);
	}

	return (0);
	}

	static inline int
	parse_fs_perm_set(fs_perm_set_t fspset, nvlist_t nvl)
	{
	nvpair_t *nvp = NULL;
	uu_avl_index_t idx = 0;

	while ((nvp = nvlist_next_nvpair(nvl, nvp)) != NULL) {
	nvlist_t *nvl2 = NULL;
	const char *fsname = nvpair_name(nvp);
	data_type_t type = nvpair_type(nvp);
	fs_perm_t *fsperm = NULL;
	fs_perm_node_t *node = safe_malloc(sizeof (fs_perm_node_t));

	fsperm = &node->fspn_fsperm;

	VERIFY(DATA_TYPE_NVLIST == type);

	uu_list_node_init(node, &node->fspn_list_node,
	fspset->fsps_list_pool);

	idx = uu_list_numnodes(fspset->fsps_list);
	fs_perm_init(fsperm, fspset, fsname);

	if (nvpair_value_nvlist(nvp, &nvl2) != 0)
	return (-1);

	(void) parse_fs_perm(fsperm, nvl2);

	uu_list_insert(fspset->fsps_list, node, idx);
	}

	return (0);
	}

	static inline const char *
	deleg_perm_comment(zfs_deleg_note_t note)
	{
	const char *str = "";

	/* subcommands */
	switch (note) {
	/* SUBCOMMANDS */
	case ZFS_DELEG_NOTE_ALLOW:
	str = gettext("Must also have the permission that is being"
	"\n\t\t\t\tallowed");
	break;
	case ZFS_DELEG_NOTE_CLONE:
	str = gettext("Must also have the 'create' ability and 'mount'"
	"\n\t\t\t\tability in the origin file system");
	break;
	case ZFS_DELEG_NOTE_CREATE:
	str = gettext("Must also have the 'mount' ability");
	break;
	case ZFS_DELEG_NOTE_DESTROY:
	str = gettext("Must also have the 'mount' ability");
	break;
	case ZFS_DELEG_NOTE_DIFF:
	str = gettext("Allows lookup of paths within a dataset;"
	"\n\t\t\t\tgiven an object number. Ordinary users need this"
	"\n\t\t\t\tin order to use zfs diff");
	break;
	case ZFS_DELEG_NOTE_HOLD:
	str = gettext("Allows adding a user hold to a snapshot");
	break;
	case ZFS_DELEG_NOTE_MOUNT:
	str = gettext("Allows mount/umount of ZFS datasets");
	break;
	case ZFS_DELEG_NOTE_PROMOTE:
	str = gettext("Must also have the 'mount'\n\t\t\t\tand"
	" 'promote' ability in the origin file system");
	break;
	case ZFS_DELEG_NOTE_RECEIVE:
	str = gettext("Must also have the 'mount' and 'create'"
	" ability");
	break;
	case ZFS_DELEG_NOTE_RELEASE:
	str = gettext("Allows releasing a user hold which\n\t\t\t\t"
	"might destroy the snapshot");
	break;
	case ZFS_DELEG_NOTE_RENAME:
	str = gettext("Must also have the 'mount' and 'create'"
	"\n\t\t\t\tability in the new parent");
	break;
	case ZFS_DELEG_NOTE_ROLLBACK:
	str = gettext("");
	break;
	case ZFS_DELEG_NOTE_SEND:
	str = gettext("");
	break;
	case ZFS_DELEG_NOTE_SHARE:
	str = gettext("Allows sharing file systems over NFS or SMB"
	"\n\t\t\t\tprotocols");
	break;
	case ZFS_DELEG_NOTE_SNAPSHOT:
	str = gettext("");
	break;
	case ZFS_DELEG_NOTE_LOAD_KEY:
	str = gettext("Allows loading or unloading an encryption key");
	break;
	case ZFS_DELEG_NOTE_CHANGE_KEY:
	str = gettext("Allows changing or adding an encryption key");
	break;
	/*
	* case ZFS_DELEG_NOTE_VSCAN:
	* str = gettext("");
	* break;
	*/
	/* OTHER */
	case ZFS_DELEG_NOTE_GROUPQUOTA:
	str = gettext("Allows accessing any groupquota@... property");
	break;
	case ZFS_DELEG_NOTE_GROUPUSED:
	str = gettext("Allows reading any groupused@... property");
	break;
	case ZFS_DELEG_NOTE_USERPROP:
	str = gettext("Allows changing any user property");
	break;
	case ZFS_DELEG_NOTE_USERQUOTA:
	str = gettext("Allows accessing any userquota@... property");
	break;
	case ZFS_DELEG_NOTE_USERUSED:
	str = gettext("Allows reading any userused@... property");
	break;
	case ZFS_DELEG_NOTE_USEROBJQUOTA:
	str = gettext("Allows accessing any userobjquota@... property");
	break;
	case ZFS_DELEG_NOTE_GROUPOBJQUOTA:
	str = gettext("Allows accessing any \n\t\t\t\t"
	"groupobjquota@... property");
	break;
	case ZFS_DELEG_NOTE_GROUPOBJUSED:
	str = gettext("Allows reading any groupobjused@... property");
	break;
	case ZFS_DELEG_NOTE_USEROBJUSED:
	str = gettext("Allows reading any userobjused@... property");
	break;
	case ZFS_DELEG_NOTE_PROJECTQUOTA:
	str = gettext("Allows accessing any projectquota@... property");
	break;
	case ZFS_DELEG_NOTE_PROJECTOBJQUOTA:
	str = gettext("Allows accessing any \n\t\t\t\t"
	"projectobjquota@... property");
	break;
	case ZFS_DELEG_NOTE_PROJECTUSED:
	str = gettext("Allows reading any projectused@... property");
	break;
	case ZFS_DELEG_NOTE_PROJECTOBJUSED:
	str = gettext("Allows accessing any \n\t\t\t\t"
	"projectobjused@... property");
	break;
	/* other */
	default:
	str = "";
	}

	return (str);
	}

	struct allow_opts {
	boolean_t local;
	boolean_t descend;
	boolean_t user;
	boolean_t group;
	boolean_t everyone;
	boolean_t create;
	boolean_t set;
	boolean_t recursive; /* unallow only */
	boolean_t prt_usage;

	boolean_t prt_perms;
	char *who;
	char *perms;
	const char *dataset;
	};

	static inline int
	prop_cmp(const void a, const void b)
	{
	const char str1 = (const char **)a;
	const char str2 = (const char **)b;
	return (strcmp(str1, str2));
	}

	static void
	allow_usage(boolean_t un, boolean_t requested, const char *msg)
	{
	const char *opt_desc[] = {
	"-h", gettext("show this help message and exit"),
	"-l", gettext("set permission locally"),
	"-d", gettext("set permission for descents"),
	"-u", gettext("set permission for user"),
	"-g", gettext("set permission for group"),
	"-e", gettext("set permission for everyone"),
	"-c", gettext("set create time permission"),
	"-s", gettext("define permission set"),
	/* unallow only */
	"-r", gettext("remove permissions recursively"),
	};
	size_t unallow_size = sizeof (opt_desc) / sizeof (char *);
	size_t allow_size = unallow_size - 2;
	const char *props[ZFS_NUM_PROPS];
	int i;
	size_t count = 0;
	FILE *fp = requested ? stdout : stderr;
	zprop_desc_t *pdtbl = zfs_prop_get_table();
	const char *fmt = gettext("%-16s %-14s\t%s\n");

	(void) fprintf(fp, gettext("Usage: %s\n"), get_usage(un ? HELP_UNALLOW :
	HELP_ALLOW));
	(void) fprintf(fp, gettext("Options:\n"));
	for (i = 0; i < (un ? unallow_size : allow_size); i += 2) {
	const char *opt = opt_desc[i];
	const char *optdsc = opt_desc[i + 1];
	(void) fprintf(fp, gettext(" %-10s %s\n"), opt, optdsc);
	}

	(void) fprintf(fp, gettext("\nThe following permissions are "
	"supported:\n\n"));
	(void) fprintf(fp, fmt, gettext("NAME"), gettext("TYPE"),
	gettext("NOTES"));
	for (i = 0; i < ZFS_NUM_DELEG_NOTES; i++) {
	const char *perm_name = zfs_deleg_perm_tbl[i].z_perm;
	zfs_deleg_note_t perm_note = zfs_deleg_perm_tbl[i].z_note;
	const char *perm_type = deleg_perm_type(perm_note);
	const char *perm_comment = deleg_perm_comment(perm_note);
	(void) fprintf(fp, fmt, perm_name, perm_type, perm_comment);
	}

	for (i = 0; i < ZFS_NUM_PROPS; i++) {
	zprop_desc_t *pd = &pdtbl[i];
	if (pd->pd_visible != B_TRUE)
	continue;

	if (pd->pd_attr == PROP_READONLY)
	continue;

	props[count++] = pd->pd_name;
	}
	props[count] = NULL;

	qsort(props, count, sizeof (char *), prop_cmp);

	for (i = 0; i < count; i++)
	(void) fprintf(fp, fmt, props[i], gettext("property"), "");

	if (msg != NULL)
	(void) fprintf(fp, gettext("\nzfs: error: %s"), msg);

	exit(requested ? 0 : 2);
	}

	static inline const char *
	munge_args(int argc, char **argv, boolean_t un, size_t expected_argc,
	char **permsp)
	{
	if (un && argc == expected_argc - 1)
	*permsp = NULL;
	else if (argc == expected_argc)
	*permsp = argv[argc - 2];
	else
	allow_usage(un, B_FALSE,
	gettext("wrong number of parameters\n"));

	return (argv[argc - 1]);
	}

	static void
	parse_allow_args(int argc, char *argv, boolean_t un, struct allow_opts opts)
	{
	int uge_sum = opts->user + opts->group + opts->everyone;
	int csuge_sum = opts->create + opts->set + uge_sum;
	int ldcsuge_sum = csuge_sum + opts->local + opts->descend;
	int all_sum = un ? ldcsuge_sum + opts->recursive : ldcsuge_sum;

	if (uge_sum > 1)
	allow_usage(un, B_FALSE,
	gettext("-u, -g, and -e are mutually exclusive\n"));

	if (opts->prt_usage) {
	if (argc == 0 && all_sum == 0)
	allow_usage(un, B_TRUE, NULL);
	else
	usage(B_FALSE);
	}

	if (opts->set) {
	if (csuge_sum > 1)
	allow_usage(un, B_FALSE,
	gettext("invalid options combined with -s\n"));

	opts->dataset = munge_args(argc, argv, un, 3, &opts->perms);
	if (argv[0][0] != '@')
	allow_usage(un, B_FALSE,
	gettext("invalid set name: missing '@' prefix\n"));
	opts->who = argv[0];
	} else if (opts->create) {
	if (ldcsuge_sum > 1)
	allow_usage(un, B_FALSE,
	gettext("invalid options combined with -c\n"));
	opts->dataset = munge_args(argc, argv, un, 2, &opts->perms);
	} else if (opts->everyone) {
	if (csuge_sum > 1)
	allow_usage(un, B_FALSE,
	gettext("invalid options combined with -e\n"));
	opts->dataset = munge_args(argc, argv, un, 2, &opts->perms);
	} else if (uge_sum == 0 && argc > 0 && strcmp(argv[0], "everyone")
	== 0) {
	opts->everyone = B_TRUE;
	argc--;
	argv++;
	opts->dataset = munge_args(argc, argv, un, 2, &opts->perms);
	} else if (argc == 1 && !un) {
	opts->prt_perms = B_TRUE;
	opts->dataset = argv[argc-1];
	} else {
	opts->dataset = munge_args(argc, argv, un, 3, &opts->perms);
	opts->who = argv[0];
	}

	if (!opts->local && !opts->descend) {
	opts->local = B_TRUE;
	opts->descend = B_TRUE;
	}
	}

	static void
	store_allow_perm(zfs_deleg_who_type_t type, boolean_t local, boolean_t descend,
	const char who, char perms, nvlist_t *top_nvl)
	{
	int i;
	char ld[2] = { '\0', '\0' };
	char who_buf[MAXNAMELEN + 32];
	char base_type = '\0';
	char set_type = '\0';
	nvlist_t *base_nvl = NULL;
	nvlist_t *set_nvl = NULL;
	nvlist_t *nvl;

	if (nvlist_alloc(&base_nvl, NV_UNIQUE_NAME, 0) != 0)
	nomem();
	if (nvlist_alloc(&set_nvl, NV_UNIQUE_NAME, 0) != 0)
	nomem();

	switch (type) {
	case ZFS_DELEG_NAMED_SET_SETS:
	case ZFS_DELEG_NAMED_SET:
	set_type = ZFS_DELEG_NAMED_SET_SETS;
	base_type = ZFS_DELEG_NAMED_SET;
	ld[0] = ZFS_DELEG_NA;
	break;
	case ZFS_DELEG_CREATE_SETS:
	case ZFS_DELEG_CREATE:
	set_type = ZFS_DELEG_CREATE_SETS;
	base_type = ZFS_DELEG_CREATE;
	ld[0] = ZFS_DELEG_NA;
	break;
	case ZFS_DELEG_USER_SETS:
	case ZFS_DELEG_USER:
	set_type = ZFS_DELEG_USER_SETS;
	base_type = ZFS_DELEG_USER;
	if (local)
	ld[0] = ZFS_DELEG_LOCAL;
	if (descend)
	ld[1] = ZFS_DELEG_DESCENDENT;
	break;
	case ZFS_DELEG_GROUP_SETS:
	case ZFS_DELEG_GROUP:
	set_type = ZFS_DELEG_GROUP_SETS;
	base_type = ZFS_DELEG_GROUP;
	if (local)
	ld[0] = ZFS_DELEG_LOCAL;
	if (descend)
	ld[1] = ZFS_DELEG_DESCENDENT;
	break;
	case ZFS_DELEG_EVERYONE_SETS:
	case ZFS_DELEG_EVERYONE:
	set_type = ZFS_DELEG_EVERYONE_SETS;
	base_type = ZFS_DELEG_EVERYONE;
	if (local)
	ld[0] = ZFS_DELEG_LOCAL;
	if (descend)
	ld[1] = ZFS_DELEG_DESCENDENT;
	break;

	default:
	assert(set_type != '\0' && base_type != '\0');
	}

	if (perms != NULL) {
	char *curr = perms;
	char *end = curr + strlen(perms);

	while (curr < end) {
	char *delim = strchr(curr, ',');
	if (delim == NULL)
	delim = end;
	else
	*delim = '\0';

	if (curr[0] == '@')
	nvl = set_nvl;
	else
	nvl = base_nvl;

	(void) nvlist_add_boolean(nvl, curr);
	if (delim != end)
	*delim = ',';
	curr = delim + 1;
	}

	for (i = 0; i < 2; i++) {
	char locality = ld[i];
	if (locality == 0)
	continue;

	if (!nvlist_empty(base_nvl)) {
	if (who != NULL)
	(void) snprintf(who_buf,
	sizeof (who_buf), "%c%c$%s",
	base_type, locality, who);
	else
	(void) snprintf(who_buf,
	sizeof (who_buf), "%c%c$",
	base_type, locality);

	(void) nvlist_add_nvlist(top_nvl, who_buf,
	base_nvl);
	}


	if (!nvlist_empty(set_nvl)) {
	if (who != NULL)
	(void) snprintf(who_buf,
	sizeof (who_buf), "%c%c$%s",
	set_type, locality, who);
	else
	(void) snprintf(who_buf,
	sizeof (who_buf), "%c%c$",
	set_type, locality);

	(void) nvlist_add_nvlist(top_nvl, who_buf,
	set_nvl);
	}
	}
	} else {
	for (i = 0; i < 2; i++) {
	char locality = ld[i];
	if (locality == 0)
	continue;

	if (who != NULL)
	(void) snprintf(who_buf, sizeof (who_buf),
	"%c%c$%s", base_type, locality, who);
	else
	(void) snprintf(who_buf, sizeof (who_buf),
	"%c%c$", base_type, locality);
	(void) nvlist_add_boolean(top_nvl, who_buf);

	if (who != NULL)
	(void) snprintf(who_buf, sizeof (who_buf),
	"%c%c$%s", set_type, locality, who);
	else
	(void) snprintf(who_buf, sizeof (who_buf),
	"%c%c$", set_type, locality);
	(void) nvlist_add_boolean(top_nvl, who_buf);
	}
	}
	}

	static int
	construct_fsacl_list(boolean_t un, struct allow_opts opts, nvlist_t *nvlp)
	{
	if (nvlist_alloc(nvlp, NV_UNIQUE_NAME, 0) != 0)
	nomem();

	if (opts->set) {
	store_allow_perm(ZFS_DELEG_NAMED_SET, opts->local,
	opts->descend, opts->who, opts->perms, *nvlp);
	} else if (opts->create) {
	store_allow_perm(ZFS_DELEG_CREATE, opts->local,
	opts->descend, NULL, opts->perms, *nvlp);
	} else if (opts->everyone) {
	store_allow_perm(ZFS_DELEG_EVERYONE, opts->local,
	opts->descend, NULL, opts->perms, *nvlp);
	} else {
	char *curr = opts->who;
	char *end = curr + strlen(curr);

	while (curr < end) {
	const char *who;
	zfs_deleg_who_type_t who_type = ZFS_DELEG_WHO_UNKNOWN;
	char *endch;
	char *delim = strchr(curr, ',');
	char errbuf[256];
	char id[64];
	struct passwd *p = NULL;
	struct group *g = NULL;

	uid_t rid;
	if (delim == NULL)
	delim = end;
	else
	*delim = '\0';

	rid = (uid_t)strtol(curr, &endch, 0);
	if (opts->user) {
	who_type = ZFS_DELEG_USER;
	if (*endch != '\0')
	p = getpwnam(curr);
	else
	p = getpwuid(rid);

	if (p != NULL)
	rid = p->pw_uid;
	else if (*endch != '\0') {
	(void) snprintf(errbuf, sizeof (errbuf),
	gettext("invalid user %s\n"), curr);
	allow_usage(un, B_TRUE, errbuf);
	}
	} else if (opts->group) {
	who_type = ZFS_DELEG_GROUP;
	if (*endch != '\0')
	g = getgrnam(curr);
	else
	g = getgrgid(rid);

	if (g != NULL)
	rid = g->gr_gid;
	else if (*endch != '\0') {
	(void) snprintf(errbuf, sizeof (errbuf),
	gettext("invalid group %s\n"),
	curr);
	allow_usage(un, B_TRUE, errbuf);
	}
	} else {
	if (*endch != '\0') {
	p = getpwnam(curr);
	} else {
	p = getpwuid(rid);
	}

	if (p == NULL) {
	if (*endch != '\0') {
	g = getgrnam(curr);
	} else {
	g = getgrgid(rid);
	}
	}

	if (p != NULL) {
	who_type = ZFS_DELEG_USER;
	rid = p->pw_uid;
	} else if (g != NULL) {
	who_type = ZFS_DELEG_GROUP;
	rid = g->gr_gid;
	} else {
	(void) snprintf(errbuf, sizeof (errbuf),
	gettext("invalid user/group %s\n"),
	curr);
	allow_usage(un, B_TRUE, errbuf);
	}
	}

	(void) sprintf(id, "%u", rid);
	who = id;

	store_allow_perm(who_type, opts->local,
	opts->descend, who, opts->perms, *nvlp);
	curr = delim + 1;
	}
	}

	return (0);
	}

	static void
	print_set_creat_perms(uu_avl_t *who_avl)
	{
	const char *sc_title[] = {
	gettext("Permission sets:\n"),
	gettext("Create time permissions:\n"),
	NULL
	};
	who_perm_node_t *who_node = NULL;
	int prev_weight = -1;

	for (who_node = uu_avl_first(who_avl); who_node != NULL;
	who_node = uu_avl_next(who_avl, who_node)) {
	uu_avl_t *avl = who_node->who_perm.who_deleg_perm_avl;
	zfs_deleg_who_type_t who_type = who_node->who_perm.who_type;
	const char *who_name = who_node->who_perm.who_name;
	int weight = who_type2weight(who_type);
	boolean_t first = B_TRUE;
	deleg_perm_node_t *deleg_node;

	if (prev_weight != weight) {
	(void) printf("%s", sc_title[weight]);
	prev_weight = weight;
	}

	if (who_name == NULL \|\| strnlen(who_name, 1) == 0)
	(void) printf("\t");
	else
	(void) printf("\t%s ", who_name);

	for (deleg_node = uu_avl_first(avl); deleg_node != NULL;
	deleg_node = uu_avl_next(avl, deleg_node)) {
	if (first) {
	(void) printf("%s",
	deleg_node->dpn_perm.dp_name);
	first = B_FALSE;
	} else
	(void) printf(",%s",
	deleg_node->dpn_perm.dp_name);
	}

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

	static void
	print_uge_deleg_perms(uu_avl_t *who_avl, boolean_t local, boolean_t descend,
	const char *title)
	{
	who_perm_node_t *who_node = NULL;
	boolean_t prt_title = B_TRUE;
	uu_avl_walk_t *walk;

	if ((walk = uu_avl_walk_start(who_avl, UU_WALK_ROBUST)) == NULL)
	nomem();

	while ((who_node = uu_avl_walk_next(walk)) != NULL) {
	const char *who_name = who_node->who_perm.who_name;
	const char *nice_who_name = who_node->who_perm.who_ug_name;
	uu_avl_t *avl = who_node->who_perm.who_deleg_perm_avl;
	zfs_deleg_who_type_t who_type = who_node->who_perm.who_type;
	char delim = ' ';
	deleg_perm_node_t *deleg_node;
	boolean_t prt_who = B_TRUE;

	for (deleg_node = uu_avl_first(avl);
	deleg_node != NULL;
	deleg_node = uu_avl_next(avl, deleg_node)) {
	if (local != deleg_node->dpn_perm.dp_local \|\|
	descend != deleg_node->dpn_perm.dp_descend)
	continue;

	if (prt_who) {
	const char *who = NULL;
	if (prt_title) {
	prt_title = B_FALSE;
	(void) printf("%s", title);
	}

	switch (who_type) {
	case ZFS_DELEG_USER_SETS:
	case ZFS_DELEG_USER:
	who = gettext("user");
	if (nice_who_name)
	who_name = nice_who_name;
	break;
	case ZFS_DELEG_GROUP_SETS:
	case ZFS_DELEG_GROUP:
	who = gettext("group");
	if (nice_who_name)
	who_name = nice_who_name;
	break;
	case ZFS_DELEG_EVERYONE_SETS:
	case ZFS_DELEG_EVERYONE:
	who = gettext("everyone");
	who_name = NULL;
	break;

	default:
	assert(who != NULL);
	}

	prt_who = B_FALSE;
	if (who_name == NULL)
	(void) printf("\t%s", who);
	else
	(void) printf("\t%s %s", who, who_name);
	}

	(void) printf("%c%s", delim,
	deleg_node->dpn_perm.dp_name);
	delim = ',';
	}

	if (!prt_who)
	(void) printf("\n");
	}

	uu_avl_walk_end(walk);
	}

	static void
	print_fs_perms(fs_perm_set_t *fspset)
	{
	fs_perm_node_t *node = NULL;
	char buf[MAXNAMELEN + 32];
	const char *dsname = buf;

	for (node = uu_list_first(fspset->fsps_list); node != NULL;
	node = uu_list_next(fspset->fsps_list, node)) {
	uu_avl_t *sc_avl = node->fspn_fsperm.fsp_sc_avl;
	uu_avl_t *uge_avl = node->fspn_fsperm.fsp_uge_avl;
	int left = 0;

	(void) snprintf(buf, sizeof (buf),
	gettext("---- Permissions on %s "),
	node->fspn_fsperm.fsp_name);
	(void) printf("%s", dsname);
	left = 70 - strlen(buf);
	while (left-- > 0)
	(void) printf("-");
	(void) printf("\n");

	print_set_creat_perms(sc_avl);
	print_uge_deleg_perms(uge_avl, B_TRUE, B_FALSE,
	gettext("Local permissions:\n"));
	print_uge_deleg_perms(uge_avl, B_FALSE, B_TRUE,
	gettext("Descendent permissions:\n"));
	print_uge_deleg_perms(uge_avl, B_TRUE, B_TRUE,
	gettext("Local+Descendent permissions:\n"));
	}
	}

	static fs_perm_set_t fs_perm_set = { NULL, NULL, NULL, NULL };

	struct deleg_perms {
	boolean_t un;
	nvlist_t *nvl;
	};

	static int
	set_deleg_perms(zfs_handle_t zhp, void data)
	{
	struct deleg_perms perms = (struct deleg_perms )data;
	zfs_type_t zfs_type = zfs_get_type(zhp);

	if (zfs_type != ZFS_TYPE_FILESYSTEM && zfs_type != ZFS_TYPE_VOLUME)
	return (0);

	return (zfs_set_fsacl(zhp, perms->un, perms->nvl));
	}

	static int
	zfs_do_allow_unallow_impl(int argc, char **argv, boolean_t un)
	{
	zfs_handle_t *zhp;
	nvlist_t *perm_nvl = NULL;
	nvlist_t *update_perm_nvl = NULL;
	int error = 1;
	int c;
	struct allow_opts opts = { 0 };

	const char *optstr = un ? "ldugecsrh" : "ldugecsh";

	/* check opts */
	while ((c = getopt(argc, argv, optstr)) != -1) {
	switch (c) {
	case 'l':
	opts.local = B_TRUE;
	break;
	case 'd':
	opts.descend = B_TRUE;
	break;
	case 'u':
	opts.user = B_TRUE;
	break;
	case 'g':
	opts.group = B_TRUE;
	break;
	case 'e':
	opts.everyone = B_TRUE;
	break;
	case 's':
	opts.set = B_TRUE;
	break;
	case 'c':
	opts.create = B_TRUE;
	break;
	case 'r':
	opts.recursive = B_TRUE;
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	usage(B_FALSE);
	break;
	case 'h':
	opts.prt_usage = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* check arguments */
	parse_allow_args(argc, argv, un, &opts);

	/* try to open the dataset */
	if ((zhp = zfs_open(g_zfs, opts.dataset, ZFS_TYPE_FILESYSTEM \|
	ZFS_TYPE_VOLUME)) == NULL) {
	(void) fprintf(stderr, "Failed to open dataset: %s\n",
	opts.dataset);
	return (-1);
	}

	if (zfs_get_fsacl(zhp, &perm_nvl) != 0)
	goto cleanup2;

	fs_perm_set_init(&fs_perm_set);
	if (parse_fs_perm_set(&fs_perm_set, perm_nvl) != 0) {
	(void) fprintf(stderr, "Failed to parse fsacl permissions\n");
	goto cleanup1;
	}

	if (opts.prt_perms)
	print_fs_perms(&fs_perm_set);
	else {
	(void) construct_fsacl_list(un, &opts, &update_perm_nvl);
	if (zfs_set_fsacl(zhp, un, update_perm_nvl) != 0)
	goto cleanup0;

	if (un && opts.recursive) {
	struct deleg_perms data = { un, update_perm_nvl };
	if (zfs_iter_filesystems_v2(zhp, 0, set_deleg_perms,
	&data) != 0)
	goto cleanup0;
	}
	}

	error = 0;

	cleanup0:
	nvlist_free(perm_nvl);
	nvlist_free(update_perm_nvl);
	cleanup1:
	fs_perm_set_fini(&fs_perm_set);
	cleanup2:
	zfs_close(zhp);

	return (error);
	}

	static int
	zfs_do_allow(int argc, char **argv)
	{
	return (zfs_do_allow_unallow_impl(argc, argv, B_FALSE));
	}

	static int
	zfs_do_unallow(int argc, char **argv)
	{
	return (zfs_do_allow_unallow_impl(argc, argv, B_TRUE));
	}

	static int
	zfs_do_hold_rele_impl(int argc, char **argv, boolean_t holding)
	{
	int errors = 0;
	int i;
	const char *tag;
	boolean_t recursive = B_FALSE;
	const char *opts = holding ? "rt" : "r";
	int c;

	/* check options */
	while ((c = getopt(argc, argv, opts)) != -1) {
	switch (c) {
	case 'r':
	recursive = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* check number of arguments */
	if (argc < 2)
	usage(B_FALSE);

	tag = argv[0];
	--argc;
	++argv;

	if (holding && tag[0] == '.') {
	/* tags starting with '.' are reserved for libzfs */
	(void) fprintf(stderr, gettext("tag may not start with '.'\n"));
	usage(B_FALSE);
	}

	for (i = 0; i < argc; ++i) {
	zfs_handle_t *zhp;
	char parent[ZFS_MAX_DATASET_NAME_LEN];
	const char *delim;
	char *path = argv[i];

	delim = strchr(path, '@');
	if (delim == NULL) {
	(void) fprintf(stderr,
	gettext("'%s' is not a snapshot\n"), path);
	++errors;
	continue;
	}
	(void) strlcpy(parent, path, MIN(sizeof (parent),
	delim - path + 1));

	zhp = zfs_open(g_zfs, parent,
	ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME);
	if (zhp == NULL) {
	++errors;
	continue;
	}
	if (holding) {
	if (zfs_hold(zhp, delim+1, tag, recursive, -1) != 0)
	++errors;
	} else {
	if (zfs_release(zhp, delim+1, tag, recursive) != 0)
	++errors;
	}
	zfs_close(zhp);
	}

	return (errors != 0);
	}

	/*
	* zfs hold [-r] [-t] <tag> <snap> ...
	*
	* -r Recursively hold
	*
	* Apply a user-hold with the given tag to the list of snapshots.
	*/
	static int
	zfs_do_hold(int argc, char **argv)
	{
	return (zfs_do_hold_rele_impl(argc, argv, B_TRUE));
	}

	/*
	* zfs release [-r] <tag> <snap> ...
	*
	* -r Recursively release
	*
	* Release a user-hold with the given tag from the list of snapshots.
	*/
	static int
	zfs_do_release(int argc, char **argv)
	{
	return (zfs_do_hold_rele_impl(argc, argv, B_FALSE));
	}

	typedef struct holds_cbdata {
	boolean_t cb_recursive;
	const char *cb_snapname;
	nvlist_t **cb_nvlp;
	size_t cb_max_namelen;
	size_t cb_max_taglen;
	} holds_cbdata_t;

	#define STRFTIME_FMT_STR "%a %b %e %H:%M %Y"
	#define DATETIME_BUF_LEN (32)
	/*
	*
	*/
	static void
	print_holds(boolean_t scripted, int nwidth, int tagwidth, nvlist_t *nvl,
	boolean_t parsable)
	{
	int i;
	nvpair_t *nvp = NULL;
	const char *const hdr_cols[] = { "NAME", "TAG", "TIMESTAMP" };
	const char *col;

	if (!scripted) {
	for (i = 0; i < 3; i++) {
	col = gettext(hdr_cols[i]);
	if (i < 2)
	(void) printf("%-*s ", i ? tagwidth : nwidth,
	col);
	else
	(void) printf("%s\n", col);
	}
	}

	while ((nvp = nvlist_next_nvpair(nvl, nvp)) != NULL) {
	const char *zname = nvpair_name(nvp);
	nvlist_t *nvl2;
	nvpair_t *nvp2 = NULL;
	(void) nvpair_value_nvlist(nvp, &nvl2);
	while ((nvp2 = nvlist_next_nvpair(nvl2, nvp2)) != NULL) {
	char tsbuf[DATETIME_BUF_LEN];
	const char *tagname = nvpair_name(nvp2);
	uint64_t val = 0;
	time_t time;
	struct tm t;

	(void) nvpair_value_uint64(nvp2, &val);
	time = (time_t)val;
	(void) localtime_r(&time, &t);
	(void) strftime(tsbuf, DATETIME_BUF_LEN,
	gettext(STRFTIME_FMT_STR), &t);

	if (scripted) {
	if (parsable) {
	(void) printf("%s\t%s\t%ld\n", zname,
	tagname, (unsigned long)time);
	} else {
	(void) printf("%s\t%s\t%s\n", zname,
	tagname, tsbuf);
	}
	} else {
	if (parsable) {
	(void) printf("%-s %-s %ld\n",
	nwidth, zname, tagwidth,
	tagname, (unsigned long)time);
	} else {
	(void) printf("%-s %-s %s\n",
	nwidth, zname, tagwidth,
	tagname, tsbuf);
	}
	}
	}
	}
	}

	/*
	* Generic callback function to list a dataset or snapshot.
	*/
	static int
	holds_callback(zfs_handle_t zhp, void data)
	{
	holds_cbdata_t *cbp = data;
	nvlist_t top_nvl = cbp->cb_nvlp;
	nvlist_t *nvl = NULL;
	nvpair_t *nvp = NULL;
	const char *zname = zfs_get_name(zhp);
	size_t znamelen = strlen(zname);

	if (cbp->cb_recursive) {
	const char *snapname;
	char *delim = strchr(zname, '@');
	if (delim == NULL)
	return (0);

	snapname = delim + 1;
	if (strcmp(cbp->cb_snapname, snapname))
	return (0);
	}

	if (zfs_get_holds(zhp, &nvl) != 0)
	return (-1);

	if (znamelen > cbp->cb_max_namelen)
	cbp->cb_max_namelen = znamelen;

	while ((nvp = nvlist_next_nvpair(nvl, nvp)) != NULL) {
	const char *tag = nvpair_name(nvp);
	size_t taglen = strlen(tag);
	if (taglen > cbp->cb_max_taglen)
	cbp->cb_max_taglen = taglen;
	}

	return (nvlist_add_nvlist(top_nvl, zname, nvl));
	}

	/*
	* zfs holds [-rHp] <snap> ...
	*
	* -r Lists holds that are set on the named snapshots recursively.
	* -H Scripted mode; elide headers and separate columns by tabs.
	* -p Display values in parsable (literal) format.
	*/
	static int
	zfs_do_holds(int argc, char **argv)
	{
	int c;
	boolean_t errors = B_FALSE;
	boolean_t scripted = B_FALSE;
	boolean_t recursive = B_FALSE;
	boolean_t parsable = B_FALSE;

	int types = ZFS_TYPE_SNAPSHOT;
	holds_cbdata_t cb = { 0 };

	int limit = 0;
	int ret = 0;
	int flags = 0;

	/* check options */
	while ((c = getopt(argc, argv, "rHp")) != -1) {
	switch (c) {
	case 'r':
	recursive = B_TRUE;
	break;
	case 'H':
	scripted = B_TRUE;
	break;
	case 'p':
	parsable = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	if (recursive) {
	types \|= ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME;
	flags \|= ZFS_ITER_RECURSE;
	}

	argc -= optind;
	argv += optind;

	/* check number of arguments */
	if (argc < 1)
	usage(B_FALSE);

	nvlist_t *nvl = fnvlist_alloc();

	for (int i = 0; i < argc; ++i) {
	char *snapshot = argv[i];
	const char *delim;
	const char *snapname;

	delim = strchr(snapshot, '@');
	if (delim == NULL) {
	(void) fprintf(stderr,
	gettext("'%s' is not a snapshot\n"), snapshot);
	errors = B_TRUE;
	continue;
	}
	snapname = delim + 1;
	if (recursive)
	snapshot[delim - snapshot] = '\0';

	cb.cb_recursive = recursive;
	cb.cb_snapname = snapname;
	cb.cb_nvlp = &nvl;

	/*
	* 1. collect holds data, set format options
	*/
	ret = zfs_for_each(1, argv + i, flags, types, NULL, NULL, limit,
	holds_callback, &cb);
	if (ret != 0)
	errors = B_TRUE;
	}

	/*
	* 2. print holds data
	*/
	print_holds(scripted, cb.cb_max_namelen, cb.cb_max_taglen, nvl,
	parsable);

	if (nvlist_empty(nvl))
	(void) fprintf(stderr, gettext("no datasets available\n"));

	nvlist_free(nvl);

	return (errors);
	}

	#define CHECK_SPINNER 30
	#define SPINNER_TIME 3 /* seconds */
	#define MOUNT_TIME 1 /* seconds */

	typedef struct get_all_state {
	char **ga_datasets;
	int ga_count;
	boolean_t ga_verbose;
	get_all_cb_t *ga_cbp;
	} get_all_state_t;

	static int
	get_one_dataset(zfs_handle_t zhp, void data)
	{
	static const char *const spin[] = { "-", "\\", "\|", "/" };
	static int spinval = 0;
	static int spincheck = 0;
	static time_t last_spin_time = (time_t)0;
	get_all_state_t *state = data;
	zfs_type_t type = zfs_get_type(zhp);

	if (state->ga_verbose) {
	if (--spincheck < 0) {
	time_t now = time(NULL);
	if (last_spin_time + SPINNER_TIME < now) {
	update_progress(spin[spinval++ % 4]);
	last_spin_time = now;
	}
	spincheck = CHECK_SPINNER;
	}
	}

	/*
	* Iterate over any nested datasets.
	*/
	if (zfs_iter_filesystems_v2(zhp, 0, get_one_dataset, data) != 0) {
	zfs_close(zhp);
	return (1);
	}

	/*
	* Skip any datasets whose type does not match.
	*/
	if ((type & ZFS_TYPE_FILESYSTEM) == 0) {
	zfs_close(zhp);
	return (0);
	}
	libzfs_add_handle(state->ga_cbp, zhp);
	assert(state->ga_cbp->cb_used <= state->ga_cbp->cb_alloc);

	return (0);
	}

	static int
	get_recursive_datasets(zfs_handle_t zhp, void data)
	{
	get_all_state_t *state = data;
	int len = strlen(zfs_get_name(zhp));
	for (int i = 0; i < state->ga_count; ++i) {
	if (strcmp(state->ga_datasets[i], zfs_get_name(zhp)) == 0)
	return (get_one_dataset(zhp, data));
	else if ((strncmp(state->ga_datasets[i], zfs_get_name(zhp),
	len) == 0) && state->ga_datasets[i][len] == '/') {
	(void) zfs_iter_filesystems_v2(zhp, 0,
	get_recursive_datasets, data);
	}
	}
	zfs_close(zhp);
	return (0);
	}

	static void
	get_all_datasets(get_all_state_t *state)
	{
	if (state->ga_verbose)
	set_progress_header(gettext("Reading ZFS config"));
	if (state->ga_datasets == NULL)
	(void) zfs_iter_root(g_zfs, get_one_dataset, state);
	else
	(void) zfs_iter_root(g_zfs, get_recursive_datasets, state);

	if (state->ga_verbose)
	finish_progress(gettext("done."));
	}

	/*
	* Generic callback for sharing or mounting filesystems. Because the code is so
	* similar, we have a common function with an extra parameter to determine which
	* mode we are using.
	*/
	typedef enum { OP_SHARE, OP_MOUNT } share_mount_op_t;

	typedef struct share_mount_state {
	share_mount_op_t sm_op;
	boolean_t sm_verbose;
	int sm_flags;
	char *sm_options;
	enum sa_protocol sm_proto; /* only valid for OP_SHARE */
	pthread_mutex_t sm_lock; /* protects the remaining fields */
	uint_t sm_total; /* number of filesystems to process */
	uint_t sm_done; /* number of filesystems processed */
	int sm_status; /* -1 if any of the share/mount operations failed */
	} share_mount_state_t;

	/*
	* Share or mount a dataset.
	*/
	static int
	share_mount_one(zfs_handle_t *zhp, int op, int flags, enum sa_protocol protocol,
	boolean_t explicit, const char *options)
	{
	char mountpoint[ZFS_MAXPROPLEN];
	char shareopts[ZFS_MAXPROPLEN];
	char smbshareopts[ZFS_MAXPROPLEN];
	const char *cmdname = op == OP_SHARE ? "share" : "mount";
	struct mnttab mnt;
	uint64_t zoned, canmount;
	boolean_t shared_nfs, shared_smb;

	assert(zfs_get_type(zhp) & ZFS_TYPE_FILESYSTEM);

	/*
	* Check to make sure we can mount/share this dataset. If we
	* are in the global zone and the filesystem is exported to a
	* local zone, or if we are in a local zone and the
	* filesystem is not exported, then it is an error.
	*/
	zoned = zfs_prop_get_int(zhp, ZFS_PROP_ZONED);

	if (zoned && getzoneid() == GLOBAL_ZONEID) {
	if (!explicit)
	return (0);

	(void) fprintf(stderr, gettext("cannot %s '%s': "
	"dataset is exported to a local zone\n"), cmdname,
	zfs_get_name(zhp));
	return (1);

	} else if (!zoned && getzoneid() != GLOBAL_ZONEID) {
	if (!explicit)
	return (0);

	(void) fprintf(stderr, gettext("cannot %s '%s': "
	"permission denied\n"), cmdname,
	zfs_get_name(zhp));
	return (1);
	}

	/*
	* Ignore any filesystems which don't apply to us. This
	* includes those with a legacy mountpoint, or those with
	* legacy share options.
	*/
	verify(zfs_prop_get(zhp, ZFS_PROP_MOUNTPOINT, mountpoint,
	sizeof (mountpoint), NULL, NULL, 0, B_FALSE) == 0);
	verify(zfs_prop_get(zhp, ZFS_PROP_SHARENFS, shareopts,
	sizeof (shareopts), NULL, NULL, 0, B_FALSE) == 0);
	verify(zfs_prop_get(zhp, ZFS_PROP_SHARESMB, smbshareopts,
	sizeof (smbshareopts), NULL, NULL, 0, B_FALSE) == 0);

	if (op == OP_SHARE && strcmp(shareopts, "off") == 0 &&
	strcmp(smbshareopts, "off") == 0) {
	if (!explicit)
	return (0);

	(void) fprintf(stderr, gettext("cannot share '%s': "
	"legacy share\n"), zfs_get_name(zhp));
	(void) fprintf(stderr, gettext("use exports(5) or "
	"smb.conf(5) to share this filesystem, or set "
	"the sharenfs or sharesmb property\n"));
	return (1);
	}

	/*
	* We cannot share or mount legacy filesystems. If the
	* shareopts is non-legacy but the mountpoint is legacy, we
	* treat it as a legacy share.
	*/
	if (strcmp(mountpoint, "legacy") == 0) {
	if (!explicit)
	return (0);

	(void) fprintf(stderr, gettext("cannot %s '%s': "
	"legacy mountpoint\n"), cmdname, zfs_get_name(zhp));
	(void) fprintf(stderr, gettext("use %s(8) to "
	"%s this filesystem\n"), cmdname, cmdname);
	return (1);
	}

	if (strcmp(mountpoint, "none") == 0) {
	if (!explicit)
	return (0);

	(void) fprintf(stderr, gettext("cannot %s '%s': no "
	"mountpoint set\n"), cmdname, zfs_get_name(zhp));
	return (1);
	}

	/*
	* canmount explicit outcome
	* on no pass through
	* on yes pass through
	* off no return 0
	* off yes display error, return 1
	* noauto no return 0
	* noauto yes pass through
	*/
	canmount = zfs_prop_get_int(zhp, ZFS_PROP_CANMOUNT);
	if (canmount == ZFS_CANMOUNT_OFF) {
	if (!explicit)
	return (0);

	(void) fprintf(stderr, gettext("cannot %s '%s': "
	"'canmount' property is set to 'off'\n"), cmdname,
	zfs_get_name(zhp));
	return (1);
	} else if (canmount == ZFS_CANMOUNT_NOAUTO && !explicit) {
	/*
	* When performing a 'zfs mount -a', we skip any mounts for
	* datasets that have 'noauto' set. Sharing a dataset with
	* 'noauto' set is only allowed if it's mounted.
	*/
	if (op == OP_MOUNT)
	return (0);
	if (op == OP_SHARE && !zfs_is_mounted(zhp, NULL)) {
	/* also purge it from existing exports */
	zfs_unshare(zhp, mountpoint, NULL);
	return (0);
	}
	}

	/*
	* If this filesystem is encrypted and does not have
	* a loaded key, we can not mount it.
	*/
	if ((flags & MS_CRYPT) == 0 &&
	zfs_prop_get_int(zhp, ZFS_PROP_ENCRYPTION) != ZIO_CRYPT_OFF &&
	zfs_prop_get_int(zhp, ZFS_PROP_KEYSTATUS) ==
	ZFS_KEYSTATUS_UNAVAILABLE) {
	if (!explicit)
	return (0);

	(void) fprintf(stderr, gettext("cannot %s '%s': "
	"encryption key not loaded\n"), cmdname, zfs_get_name(zhp));
	return (1);
	}

	/*
	* If this filesystem is inconsistent and has a receive resume
	* token, we can not mount it.
	*/
	if (zfs_prop_get_int(zhp, ZFS_PROP_INCONSISTENT) &&
	zfs_prop_get(zhp, ZFS_PROP_RECEIVE_RESUME_TOKEN,
	NULL, 0, NULL, NULL, 0, B_TRUE) == 0) {
	if (!explicit)
	return (0);

	(void) fprintf(stderr, gettext("cannot %s '%s': "
	"Contains partially-completed state from "
	"\"zfs receive -s\", which can be resumed with "
	"\"zfs send -t\"\n"),
	cmdname, zfs_get_name(zhp));
	return (1);
	}

	if (zfs_prop_get_int(zhp, ZFS_PROP_REDACTED) && !(flags & MS_FORCE)) {
	if (!explicit)
	return (0);

	(void) fprintf(stderr, gettext("cannot %s '%s': "
	"Dataset is not complete, was created by receiving "
	"a redacted zfs send stream.\n"), cmdname,
	zfs_get_name(zhp));
	return (1);
	}

	/*
	* At this point, we have verified that the mountpoint and/or
	* shareopts are appropriate for auto management. If the
	* filesystem is already mounted or shared, return (failing
	* for explicit requests); otherwise mount or share the
	* filesystem.
	*/
	switch (op) {
	case OP_SHARE: {
	enum sa_protocol prot[] = {SA_PROTOCOL_NFS, SA_NO_PROTOCOL};
	shared_nfs = zfs_is_shared(zhp, NULL, prot);
	*prot = SA_PROTOCOL_SMB;
	shared_smb = zfs_is_shared(zhp, NULL, prot);

	if ((shared_nfs && shared_smb) \|\|
	(shared_nfs && strcmp(shareopts, "on") == 0 &&
	strcmp(smbshareopts, "off") == 0) \|\|
	(shared_smb && strcmp(smbshareopts, "on") == 0 &&
	strcmp(shareopts, "off") == 0)) {
	if (!explicit)
	return (0);

	(void) fprintf(stderr, gettext("cannot share "
	"'%s': filesystem already shared\n"),
	zfs_get_name(zhp));
	return (1);
	}

	if (!zfs_is_mounted(zhp, NULL) &&
	zfs_mount(zhp, NULL, flags) != 0)
	return (1);

	*prot = protocol;
	if (zfs_share(zhp, protocol == SA_NO_PROTOCOL ? NULL : prot))
	return (1);

	}
	break;

	case OP_MOUNT:
	mnt.mnt_mntopts = (char *)(options ?: "");

	if (!hasmntopt(&mnt, MNTOPT_REMOUNT) &&
	zfs_is_mounted(zhp, NULL)) {
	if (!explicit)
	return (0);

	(void) fprintf(stderr, gettext("cannot mount "
	"'%s': filesystem already mounted\n"),
	zfs_get_name(zhp));
	return (1);
	}

	if (zfs_mount(zhp, options, flags) != 0)
	return (1);
	break;
	}

	return (0);
	}

	/*
	* Reports progress in the form "(current/total)". Not thread-safe.
	*/
	static void
	report_mount_progress(int current, int total)
	{
	static time_t last_progress_time = 0;
	time_t now = time(NULL);
	char info[32];

	/* display header if we're here for the first time */
	if (current == 1) {
	set_progress_header(gettext("Mounting ZFS filesystems"));
	} else if (current != total && last_progress_time + MOUNT_TIME >= now) {
	/* too soon to report again */
	return;
	}

	last_progress_time = now;

	(void) sprintf(info, "(%d/%d)", current, total);

	if (current == total)
	finish_progress(info);
	else
	update_progress(info);
	}

	/*
	* zfs_foreach_mountpoint() callback that mounts or shares one filesystem and
	* updates the progress meter.
	*/
	static int
	share_mount_one_cb(zfs_handle_t zhp, void arg)
	{
	share_mount_state_t *sms = arg;
	int ret;

	ret = share_mount_one(zhp, sms->sm_op, sms->sm_flags, sms->sm_proto,
	B_FALSE, sms->sm_options);

	pthread_mutex_lock(&sms->sm_lock);
	if (ret != 0)
	sms->sm_status = ret;
	sms->sm_done++;
	if (sms->sm_verbose)
	report_mount_progress(sms->sm_done, sms->sm_total);
	pthread_mutex_unlock(&sms->sm_lock);
	return (ret);
	}

	static void
	append_options(char mntopts, char newopts)
	{
	int len = strlen(mntopts);

	/* original length plus new string to append plus 1 for the comma */
	if (len + 1 + strlen(newopts) >= MNT_LINE_MAX) {
	(void) fprintf(stderr, gettext("the opts argument for "
	"'%s' option is too long (more than %d chars)\n"),
	"-o", MNT_LINE_MAX);
	usage(B_FALSE);
	}

	if (*mntopts)
	mntopts[len++] = ',';

	(void) strcpy(&mntopts[len], newopts);
	}

	static enum sa_protocol
	sa_protocol_decode(const char *protocol)
	{
	for (enum sa_protocol i = 0; i < ARRAY_SIZE(sa_protocol_names); ++i)
	if (strcmp(protocol, sa_protocol_names[i]) == 0)
	return (i);

	(void) fputs(gettext("share type must be one of: "), stderr);
	for (enum sa_protocol i = 0;
	i < ARRAY_SIZE(sa_protocol_names); ++i)
	(void) fprintf(stderr, "%s%s",
	i != 0 ? ", " : "", sa_protocol_names[i]);
	(void) fputc('\n', stderr);
	usage(B_FALSE);
	}

	static int
	share_mount(int op, int argc, char **argv)
	{
	int do_all = 0;
	int recursive = 0;
	boolean_t verbose = B_FALSE;
	int c, ret = 0;
	char *options = NULL;
	int flags = 0;
	+ const uint_t mount_nthr = 512;
	+ uint_t nthr;

	/* check options */
	while ((c = getopt(argc, argv, op == OP_MOUNT ? ":aRlvo:Of" : "al"))
	!= -1) {
	switch (c) {
	case 'a':
	do_all = 1;
	break;
	case 'R':
	recursive = 1;
	break;
	case 'v':
	verbose = B_TRUE;
	break;
	case 'l':
	flags \|= MS_CRYPT;
	break;
	case 'o':
	if (*optarg == '\0') {
	(void) fprintf(stderr, gettext("empty mount "
	"options (-o) specified\n"));
	usage(B_FALSE);
	}

	if (options == NULL)
	options = safe_malloc(MNT_LINE_MAX + 1);

	/* option validation is done later */
	append_options(options, optarg);
	break;
	case 'O':
	flags \|= MS_OVERLAY;
	break;
	case 'f':
	flags \|= MS_FORCE;
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	usage(B_FALSE);
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* check number of arguments */
	if (do_all \|\| recursive) {
	enum sa_protocol protocol = SA_NO_PROTOCOL;

	if (op == OP_SHARE && argc > 0) {
	protocol = sa_protocol_decode(argv[0]);
	argc--;
	argv++;
	}

	if (argc != 0 && do_all) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	if (argc == 0 && recursive) {
	(void) fprintf(stderr,
	gettext("no dataset provided\n"));
	usage(B_FALSE);
	}

	start_progress_timer();
	get_all_cb_t cb = { 0 };
	get_all_state_t state = { 0 };
	if (argc == 0) {
	state.ga_datasets = NULL;
	state.ga_count = -1;
	} else {
	zfs_handle_t *zhp;
	for (int i = 0; i < argc; i++) {
	zhp = zfs_open(g_zfs, argv[i],
	ZFS_TYPE_FILESYSTEM);
	if (zhp == NULL)
	usage(B_FALSE);
	zfs_close(zhp);
	}
	state.ga_datasets = argv;
	state.ga_count = argc;
	}
	state.ga_verbose = verbose;
	state.ga_cbp = &cb;
	get_all_datasets(&state);

	if (cb.cb_used == 0) {
	free(options);
	return (0);
	}

	share_mount_state_t share_mount_state = { 0 };
	share_mount_state.sm_op = op;
	share_mount_state.sm_verbose = verbose;
	share_mount_state.sm_flags = flags;
	share_mount_state.sm_options = options;
	share_mount_state.sm_proto = protocol;
	share_mount_state.sm_total = cb.cb_used;
	pthread_mutex_init(&share_mount_state.sm_lock, NULL);

	/* For a 'zfs share -a' operation start with a clean slate. */
	if (op == OP_SHARE)
	zfs_truncate_shares(NULL);

	/*
	* libshare isn't mt-safe, so only do the operation in parallel
	* if we're mounting. Additionally, the key-loading option must
	* be serialized so that we can prompt the user for their keys
	* in a consistent manner.
	*/
	+ nthr = op == OP_MOUNT && !(flags & MS_CRYPT) ? mount_nthr : 1;
	zfs_foreach_mountpoint(g_zfs, cb.cb_handles, cb.cb_used,
	- share_mount_one_cb, &share_mount_state,
	- op == OP_MOUNT && !(flags & MS_CRYPT));
	+ share_mount_one_cb, &share_mount_state, nthr);
	zfs_commit_shares(NULL);

	ret = share_mount_state.sm_status;

	for (int i = 0; i < cb.cb_used; i++)
	zfs_close(cb.cb_handles[i]);
	free(cb.cb_handles);
	} else if (argc == 0) {
	FILE *mnttab;
	struct mnttab entry;

	if ((op == OP_SHARE) \|\| (options != NULL)) {
	(void) fprintf(stderr, gettext("missing filesystem "
	"argument (specify -a for all)\n"));
	usage(B_FALSE);
	}

	/*
	* When mount is given no arguments, go through
	* /proc/self/mounts and display any active ZFS mounts.
	* We hide any snapshots, since they are controlled
	* automatically.
	*/

	if ((mnttab = fopen(MNTTAB, "re")) == NULL) {
	free(options);
	return (ENOENT);
	}

	while (getmntent(mnttab, &entry) == 0) {
	if (strcmp(entry.mnt_fstype, MNTTYPE_ZFS) != 0 \|\|
	strchr(entry.mnt_special, '@') != NULL)
	continue;

	(void) printf("%-30s %s\n", entry.mnt_special,
	entry.mnt_mountp);
	}

	(void) fclose(mnttab);
	} else {
	zfs_handle_t *zhp;

	if (argc > 1) {
	(void) fprintf(stderr,
	gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	if ((zhp = zfs_open(g_zfs, argv[0],
	ZFS_TYPE_FILESYSTEM)) == NULL) {
	ret = 1;
	} else {
	ret = share_mount_one(zhp, op, flags, SA_NO_PROTOCOL,
	B_TRUE, options);
	zfs_commit_shares(NULL);
	zfs_close(zhp);
	}
	}

	free(options);
	return (ret);
	}

	/*
	* zfs mount -a
	* zfs mount filesystem
	*
	* Mount all filesystems, or mount the given filesystem.
	*/
	static int
	zfs_do_mount(int argc, char **argv)
	{
	return (share_mount(OP_MOUNT, argc, argv));
	}

	/*
	* zfs share -a [nfs \| smb]
	* zfs share filesystem
	*
	* Share all filesystems, or share the given filesystem.
	*/
	static int
	zfs_do_share(int argc, char **argv)
	{
	return (share_mount(OP_SHARE, argc, argv));
	}

	typedef struct unshare_unmount_node {
	zfs_handle_t *un_zhp;
	char *un_mountp;
	uu_avl_node_t un_avlnode;
	} unshare_unmount_node_t;

	static int
	unshare_unmount_compare(const void larg, const void rarg, void *unused)
	{
	(void) unused;
	const unshare_unmount_node_t *l = larg;
	const unshare_unmount_node_t *r = rarg;

	return (strcmp(l->un_mountp, r->un_mountp));
	}

	/*
	* Convenience routine used by zfs_do_umount() and manual_unmount(). Given an
	* absolute path, find the entry /proc/self/mounts, verify that it's a
	* ZFS filesystem, and unmount it appropriately.
	*/
	static int
	unshare_unmount_path(int op, char *path, int flags, boolean_t is_manual)
	{
	zfs_handle_t *zhp;
	int ret = 0;
	struct stat64 statbuf;
	struct extmnttab entry;
	const char *cmdname = (op == OP_SHARE) ? "unshare" : "unmount";
	ino_t path_inode;

	/*
	* Search for the given (major,minor) pair in the mount table.
	*/

	if (getextmntent(path, &entry, &statbuf) != 0) {
	if (op == OP_SHARE) {
	(void) fprintf(stderr, gettext("cannot %s '%s': not "
	"currently mounted\n"), cmdname, path);
	return (1);
	}
	(void) fprintf(stderr, gettext("warning: %s not in"
	"/proc/self/mounts\n"), path);
	if ((ret = umount2(path, flags)) != 0)
	(void) fprintf(stderr, gettext("%s: %s\n"), path,
	strerror(errno));
	return (ret != 0);
	}
	path_inode = statbuf.st_ino;

	if (strcmp(entry.mnt_fstype, MNTTYPE_ZFS) != 0) {
	(void) fprintf(stderr, gettext("cannot %s '%s': not a ZFS "
	"filesystem\n"), cmdname, path);
	return (1);
	}

	if ((zhp = zfs_open(g_zfs, entry.mnt_special,
	ZFS_TYPE_FILESYSTEM)) == NULL)
	return (1);

	ret = 1;
	if (stat64(entry.mnt_mountp, &statbuf) != 0) {
	(void) fprintf(stderr, gettext("cannot %s '%s': %s\n"),
	cmdname, path, strerror(errno));
	goto out;
	} else if (statbuf.st_ino != path_inode) {
	(void) fprintf(stderr, gettext("cannot "
	"%s '%s': not a mountpoint\n"), cmdname, path);
	goto out;
	}

	if (op == OP_SHARE) {
	char nfs_mnt_prop[ZFS_MAXPROPLEN];
	char smbshare_prop[ZFS_MAXPROPLEN];

	verify(zfs_prop_get(zhp, ZFS_PROP_SHARENFS, nfs_mnt_prop,
	sizeof (nfs_mnt_prop), NULL, NULL, 0, B_FALSE) == 0);
	verify(zfs_prop_get(zhp, ZFS_PROP_SHARESMB, smbshare_prop,
	sizeof (smbshare_prop), NULL, NULL, 0, B_FALSE) == 0);

	if (strcmp(nfs_mnt_prop, "off") == 0 &&
	strcmp(smbshare_prop, "off") == 0) {
	(void) fprintf(stderr, gettext("cannot unshare "
	"'%s': legacy share\n"), path);
	(void) fprintf(stderr, gettext("use exportfs(8) "
	"or smbcontrol(1) to unshare this filesystem\n"));
	} else if (!zfs_is_shared(zhp, NULL, NULL)) {
	(void) fprintf(stderr, gettext("cannot unshare '%s': "
	"not currently shared\n"), path);
	} else {
	ret = zfs_unshare(zhp, path, NULL);
	zfs_commit_shares(NULL);
	}
	} else {
	char mtpt_prop[ZFS_MAXPROPLEN];

	verify(zfs_prop_get(zhp, ZFS_PROP_MOUNTPOINT, mtpt_prop,
	sizeof (mtpt_prop), NULL, NULL, 0, B_FALSE) == 0);

	if (is_manual) {
	ret = zfs_unmount(zhp, NULL, flags);
	} else if (strcmp(mtpt_prop, "legacy") == 0) {
	(void) fprintf(stderr, gettext("cannot unmount "
	"'%s': legacy mountpoint\n"),
	zfs_get_name(zhp));
	(void) fprintf(stderr, gettext("use umount(8) "
	"to unmount this filesystem\n"));
	} else {
	ret = zfs_unmountall(zhp, flags);
	}
	}

	out:
	zfs_close(zhp);

	return (ret != 0);
	}

	/*
	* Generic callback for unsharing or unmounting a filesystem.
	*/
	static int
	unshare_unmount(int op, int argc, char **argv)
	{
	int do_all = 0;
	int flags = 0;
	int ret = 0;
	int c;
	zfs_handle_t *zhp;
	char nfs_mnt_prop[ZFS_MAXPROPLEN];
	char sharesmb[ZFS_MAXPROPLEN];

	/* check options */
	while ((c = getopt(argc, argv, op == OP_SHARE ? ":a" : "afu")) != -1) {
	switch (c) {
	case 'a':
	do_all = 1;
	break;
	case 'f':
	flags \|= MS_FORCE;
	break;
	case 'u':
	flags \|= MS_CRYPT;
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	usage(B_FALSE);
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	if (do_all) {
	/*
	* We could make use of zfs_for_each() to walk all datasets in
	* the system, but this would be very inefficient, especially
	* since we would have to linearly search /proc/self/mounts for
	* each one. Instead, do one pass through /proc/self/mounts
	* looking for zfs entries and call zfs_unmount() for each one.
	*
	* Things get a little tricky if the administrator has created
	* mountpoints beneath other ZFS filesystems. In this case, we
	* have to unmount the deepest filesystems first. To accomplish
	* this, we place all the mountpoints in an AVL tree sorted by
	* the special type (dataset name), and walk the result in
	* reverse to make sure to get any snapshots first.
	*/
	FILE *mnttab;
	struct mnttab entry;
	uu_avl_pool_t *pool;
	uu_avl_t *tree = NULL;
	unshare_unmount_node_t *node;
	uu_avl_index_t idx;
	uu_avl_walk_t *walk;
	enum sa_protocol *protocol = NULL,
	single_protocol[] = {SA_NO_PROTOCOL, SA_NO_PROTOCOL};

	if (op == OP_SHARE && argc > 0) {
	*single_protocol = sa_protocol_decode(argv[0]);
	protocol = single_protocol;
	argc--;
	argv++;
	}

	if (argc != 0) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	if (((pool = uu_avl_pool_create("unmount_pool",
	sizeof (unshare_unmount_node_t),
	offsetof(unshare_unmount_node_t, un_avlnode),
	unshare_unmount_compare, UU_DEFAULT)) == NULL) \|\|
	((tree = uu_avl_create(pool, NULL, UU_DEFAULT)) == NULL))
	nomem();

	if ((mnttab = fopen(MNTTAB, "re")) == NULL) {
	uu_avl_destroy(tree);
	uu_avl_pool_destroy(pool);
	return (ENOENT);
	}

	while (getmntent(mnttab, &entry) == 0) {

	/* ignore non-ZFS entries */
	if (strcmp(entry.mnt_fstype, MNTTYPE_ZFS) != 0)
	continue;

	/* ignore snapshots */
	if (strchr(entry.mnt_special, '@') != NULL)
	continue;

	if ((zhp = zfs_open(g_zfs, entry.mnt_special,
	ZFS_TYPE_FILESYSTEM)) == NULL) {
	ret = 1;
	continue;
	}

	/*
	* Ignore datasets that are excluded/restricted by
	* parent pool name.
	*/
	if (zpool_skip_pool(zfs_get_pool_name(zhp))) {
	zfs_close(zhp);
	continue;
	}

	switch (op) {
	case OP_SHARE:
	verify(zfs_prop_get(zhp, ZFS_PROP_SHARENFS,
	nfs_mnt_prop,
	sizeof (nfs_mnt_prop),
	NULL, NULL, 0, B_FALSE) == 0);
	if (strcmp(nfs_mnt_prop, "off") != 0)
	break;
	verify(zfs_prop_get(zhp, ZFS_PROP_SHARESMB,
	nfs_mnt_prop,
	sizeof (nfs_mnt_prop),
	NULL, NULL, 0, B_FALSE) == 0);
	if (strcmp(nfs_mnt_prop, "off") == 0)
	continue;
	break;
	case OP_MOUNT:
	/* Ignore legacy mounts */
	verify(zfs_prop_get(zhp, ZFS_PROP_MOUNTPOINT,
	nfs_mnt_prop,
	sizeof (nfs_mnt_prop),
	NULL, NULL, 0, B_FALSE) == 0);
	if (strcmp(nfs_mnt_prop, "legacy") == 0)
	continue;
	/* Ignore canmount=noauto mounts */
	if (zfs_prop_get_int(zhp, ZFS_PROP_CANMOUNT) ==
	ZFS_CANMOUNT_NOAUTO)
	continue;
	break;
	default:
	break;
	}

	node = safe_malloc(sizeof (unshare_unmount_node_t));
	node->un_zhp = zhp;
	node->un_mountp = safe_strdup(entry.mnt_mountp);

	uu_avl_node_init(node, &node->un_avlnode, pool);

	if (uu_avl_find(tree, node, NULL, &idx) == NULL) {
	uu_avl_insert(tree, node, idx);
	} else {
	zfs_close(node->un_zhp);
	free(node->un_mountp);
	free(node);
	}
	}
	(void) fclose(mnttab);

	/*
	* Walk the AVL tree in reverse, unmounting each filesystem and
	* removing it from the AVL tree in the process.
	*/
	if ((walk = uu_avl_walk_start(tree,
	UU_WALK_REVERSE \| UU_WALK_ROBUST)) == NULL)
	nomem();

	while ((node = uu_avl_walk_next(walk)) != NULL) {
	const char *mntarg = NULL;

	uu_avl_remove(tree, node);
	switch (op) {
	case OP_SHARE:
	if (zfs_unshare(node->un_zhp,
	node->un_mountp, protocol) != 0)
	ret = 1;
	break;

	case OP_MOUNT:
	if (zfs_unmount(node->un_zhp,
	mntarg, flags) != 0)
	ret = 1;
	break;
	}

	zfs_close(node->un_zhp);
	free(node->un_mountp);
	free(node);
	}

	if (op == OP_SHARE)
	zfs_commit_shares(protocol);

	uu_avl_walk_end(walk);
	uu_avl_destroy(tree);
	uu_avl_pool_destroy(pool);

	} else {
	if (argc != 1) {
	if (argc == 0)
	(void) fprintf(stderr,
	gettext("missing filesystem argument\n"));
	else
	(void) fprintf(stderr,
	gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	/*
	* We have an argument, but it may be a full path or a ZFS
	* filesystem. Pass full paths off to unmount_path() (shared by
	* manual_unmount), otherwise open the filesystem and pass to
	* zfs_unmount().
	*/
	if (argv[0][0] == '/')
	return (unshare_unmount_path(op, argv[0],
	flags, B_FALSE));

	if ((zhp = zfs_open(g_zfs, argv[0],
	ZFS_TYPE_FILESYSTEM)) == NULL)
	return (1);

	verify(zfs_prop_get(zhp, op == OP_SHARE ?
	ZFS_PROP_SHARENFS : ZFS_PROP_MOUNTPOINT,
	nfs_mnt_prop, sizeof (nfs_mnt_prop), NULL,
	NULL, 0, B_FALSE) == 0);

	switch (op) {
	case OP_SHARE:
	verify(zfs_prop_get(zhp, ZFS_PROP_SHARENFS,
	nfs_mnt_prop,
	sizeof (nfs_mnt_prop),
	NULL, NULL, 0, B_FALSE) == 0);
	verify(zfs_prop_get(zhp, ZFS_PROP_SHARESMB,
	sharesmb, sizeof (sharesmb), NULL, NULL,
	0, B_FALSE) == 0);

	if (strcmp(nfs_mnt_prop, "off") == 0 &&
	strcmp(sharesmb, "off") == 0) {
	(void) fprintf(stderr, gettext("cannot "
	"unshare '%s': legacy share\n"),
	zfs_get_name(zhp));
	(void) fprintf(stderr, gettext("use "
	"exports(5) or smb.conf(5) to unshare "
	"this filesystem\n"));
	ret = 1;
	} else if (!zfs_is_shared(zhp, NULL, NULL)) {
	(void) fprintf(stderr, gettext("cannot "
	"unshare '%s': not currently "
	"shared\n"), zfs_get_name(zhp));
	ret = 1;
	} else if (zfs_unshareall(zhp, NULL) != 0) {
	ret = 1;
	}
	break;

	case OP_MOUNT:
	if (strcmp(nfs_mnt_prop, "legacy") == 0) {
	(void) fprintf(stderr, gettext("cannot "
	"unmount '%s': legacy "
	"mountpoint\n"), zfs_get_name(zhp));
	(void) fprintf(stderr, gettext("use "
	"umount(8) to unmount this "
	"filesystem\n"));
	ret = 1;
	} else if (!zfs_is_mounted(zhp, NULL)) {
	(void) fprintf(stderr, gettext("cannot "
	"unmount '%s': not currently "
	"mounted\n"),
	zfs_get_name(zhp));
	ret = 1;
	} else if (zfs_unmountall(zhp, flags) != 0) {
	ret = 1;
	}
	break;
	}

	zfs_close(zhp);
	}

	return (ret);
	}

	/*
	* zfs unmount [-fu] -a
	* zfs unmount [-fu] filesystem
	*
	* Unmount all filesystems, or a specific ZFS filesystem.
	*/
	static int
	zfs_do_unmount(int argc, char **argv)
	{
	return (unshare_unmount(OP_MOUNT, argc, argv));
	}

	/*
	* zfs unshare -a
	* zfs unshare filesystem
	*
	* Unshare all filesystems, or a specific ZFS filesystem.
	*/
	static int
	zfs_do_unshare(int argc, char **argv)
	{
	return (unshare_unmount(OP_SHARE, argc, argv));
	}

	static int
	find_command_idx(const char command, int idx)
	{
	int i;

	for (i = 0; i < NCOMMAND; i++) {
	if (command_table[i].name == NULL)
	continue;

	if (strcmp(command, command_table[i].name) == 0) {
	*idx = i;
	return (0);
	}
	}
	return (1);
	}

	static int
	zfs_do_diff(int argc, char **argv)
	{
	zfs_handle_t *zhp;
	int flags = 0;
	char *tosnap = NULL;
	char *fromsnap = NULL;
	char atp, copy;
	int err = 0;
	int c;
	struct sigaction sa;

	while ((c = getopt(argc, argv, "FHth")) != -1) {
	switch (c) {
	case 'F':
	flags \|= ZFS_DIFF_CLASSIFY;
	break;
	case 'H':
	flags \|= ZFS_DIFF_PARSEABLE;
	break;
	case 't':
	flags \|= ZFS_DIFF_TIMESTAMP;
	break;
	case 'h':
	flags \|= ZFS_DIFF_NO_MANGLE;
	break;
	default:
	(void) fprintf(stderr,
	gettext("invalid option '%c'\n"), optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr,
	gettext("must provide at least one snapshot name\n"));
	usage(B_FALSE);
	}

	if (argc > 2) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	fromsnap = argv[0];
	tosnap = (argc == 2) ? argv[1] : NULL;

	copy = NULL;
	if (*fromsnap != '@')
	copy = strdup(fromsnap);
	else if (tosnap)
	copy = strdup(tosnap);
	if (copy == NULL)
	usage(B_FALSE);

	if ((atp = strchr(copy, '@')) != NULL)
	*atp = '\0';

	if ((zhp = zfs_open(g_zfs, copy, ZFS_TYPE_FILESYSTEM)) == NULL) {
	free(copy);
	return (1);
	}
	free(copy);

	/*
	* Ignore SIGPIPE so that the library can give us
	* information on any failure
	*/
	if (sigemptyset(&sa.sa_mask) == -1) {
	err = errno;
	goto out;
	}
	sa.sa_flags = 0;
	sa.sa_handler = SIG_IGN;
	if (sigaction(SIGPIPE, &sa, NULL) == -1) {
	err = errno;
	goto out;
	}

	err = zfs_show_diffs(zhp, STDOUT_FILENO, fromsnap, tosnap, flags);
	out:
	zfs_close(zhp);

	return (err != 0);
	}

	/*
	* zfs bookmark <fs@source>\|<fs#source> <fs#bookmark>
	*
	* Creates a bookmark with the given name from the source snapshot
	* or creates a copy of an existing source bookmark.
	*/
	static int
	zfs_do_bookmark(int argc, char **argv)
	{
	char source, bookname;
	char expbuf[ZFS_MAX_DATASET_NAME_LEN];
	int source_type;
	nvlist_t *nvl;
	int ret = 0;
	int c;

	/* check options */
	while ((c = getopt(argc, argv, "")) != -1) {
	switch (c) {
	case '?':
	(void) fprintf(stderr,
	gettext("invalid option '%c'\n"), optopt);
	goto usage;
	}
	}

	argc -= optind;
	argv += optind;

	/* check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing source argument\n"));
	goto usage;
	}
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing bookmark argument\n"));
	goto usage;
	}

	source = argv[0];
	bookname = argv[1];

	if (strchr(source, '@') == NULL && strchr(source, '#') == NULL) {
	(void) fprintf(stderr,
	gettext("invalid source name '%s': "
	"must contain a '@' or '#'\n"), source);
	goto usage;
	}
	if (strchr(bookname, '#') == NULL) {
	(void) fprintf(stderr,
	gettext("invalid bookmark name '%s': "
	"must contain a '#'\n"), bookname);
	goto usage;
	}

	/*
	* expand source or bookname to full path:
	* one of them may be specified as short name
	*/
	{
	char **expand;
	char source_short, bookname_short;
	source_short = strpbrk(source, "@#");
	bookname_short = strpbrk(bookname, "#");
	if (source_short == source &&
	bookname_short == bookname) {
	(void) fprintf(stderr, gettext(
	"either source or bookmark must be specified as "
	"full dataset paths"));
	goto usage;
	} else if (source_short != source &&
	bookname_short != bookname) {
	expand = NULL;
	} else if (source_short != source) {
	strlcpy(expbuf, source, sizeof (expbuf));
	expand = &bookname;
	} else if (bookname_short != bookname) {
	strlcpy(expbuf, bookname, sizeof (expbuf));
	expand = &source;
	} else {
	abort();
	}
	if (expand != NULL) {
	strpbrk(expbuf, "@#") = '\0'; / dataset name in buf */
	(void) strlcat(expbuf, *expand, sizeof (expbuf));
	*expand = expbuf;
	}
	}

	/* determine source type */
	switch (*strpbrk(source, "@#")) {
	case '@': source_type = ZFS_TYPE_SNAPSHOT; break;
	case '#': source_type = ZFS_TYPE_BOOKMARK; break;
	default: abort();
	}

	/* test the source exists */
	zfs_handle_t *zhp;
	zhp = zfs_open(g_zfs, source, source_type);
	if (zhp == NULL)
	goto usage;
	zfs_close(zhp);

	nvl = fnvlist_alloc();
	fnvlist_add_string(nvl, bookname, source);
	ret = lzc_bookmark(nvl, NULL);
	fnvlist_free(nvl);

	if (ret != 0) {
	const char *err_msg = NULL;
	char errbuf[1024];

	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN,
	"cannot create bookmark '%s'"), bookname);

	switch (ret) {
	case EXDEV:
	err_msg = "bookmark is in a different pool";
	break;
	case ZFS_ERR_BOOKMARK_SOURCE_NOT_ANCESTOR:
	err_msg = "source is not an ancestor of the "
	"new bookmark's dataset";
	break;
	case EEXIST:
	err_msg = "bookmark exists";
	break;
	case EINVAL:
	err_msg = "invalid argument";
	break;
	case ENOTSUP:
	err_msg = "bookmark feature not enabled";
	break;
	case ENOSPC:
	err_msg = "out of space";
	break;
	case ENOENT:
	err_msg = "dataset does not exist";
	break;
	default:
	(void) zfs_standard_error(g_zfs, ret, errbuf);
	break;
	}
	if (err_msg != NULL) {
	(void) fprintf(stderr, "%s: %s\n", errbuf,
	dgettext(TEXT_DOMAIN, err_msg));
	}
	}

	return (ret != 0);

	usage:
	usage(B_FALSE);
	return (-1);
	}

	static int
	zfs_do_channel_program(int argc, char **argv)
	{
	int ret, fd, c;
	size_t progsize, progread;
	nvlist_t *outnvl = NULL;
	uint64_t instrlimit = ZCP_DEFAULT_INSTRLIMIT;
	uint64_t memlimit = ZCP_DEFAULT_MEMLIMIT;
	boolean_t sync_flag = B_TRUE, json_output = B_FALSE;
	zpool_handle_t *zhp;

	/* check options */
	while ((c = getopt(argc, argv, "nt:m:j")) != -1) {
	switch (c) {
	case 't':
	case 'm': {
	uint64_t arg;
	char *endp;

	errno = 0;
	arg = strtoull(optarg, &endp, 0);
	if (errno != 0 \|\| *endp != '\0') {
	(void) fprintf(stderr, gettext(
	"invalid argument "
	"'%s': expected integer\n"), optarg);
	goto usage;
	}

	if (c == 't') {
	instrlimit = arg;
	} else {
	ASSERT3U(c, ==, 'm');
	memlimit = arg;
	}
	break;
	}
	case 'n': {
	sync_flag = B_FALSE;
	break;
	}
	case 'j': {
	json_output = B_TRUE;
	break;
	}
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	goto usage;
	}
	}

	argc -= optind;
	argv += optind;

	if (argc < 2) {
	(void) fprintf(stderr,
	gettext("invalid number of arguments\n"));
	goto usage;
	}

	const char *poolname = argv[0];
	const char *filename = argv[1];
	if (strcmp(filename, "-") == 0) {
	fd = 0;
	filename = "standard input";
	} else if ((fd = open(filename, O_RDONLY)) < 0) {
	(void) fprintf(stderr, gettext("cannot open '%s': %s\n"),
	filename, strerror(errno));
	return (1);
	}

	if ((zhp = zpool_open(g_zfs, poolname)) == NULL) {
	(void) fprintf(stderr, gettext("cannot open pool '%s'\n"),
	poolname);
	if (fd != 0)
	(void) close(fd);
	return (1);
	}
	zpool_close(zhp);

	/*
	* Read in the channel program, expanding the program buffer as
	* necessary.
	*/
	progread = 0;
	progsize = 1024;
	char *progbuf = safe_malloc(progsize);
	do {
	ret = read(fd, progbuf + progread, progsize - progread);
	progread += ret;
	if (progread == progsize && ret > 0) {
	progsize *= 2;
	progbuf = safe_realloc(progbuf, progsize);
	}
	} while (ret > 0);

	if (fd != 0)
	(void) close(fd);
	if (ret < 0) {
	free(progbuf);
	(void) fprintf(stderr,
	gettext("cannot read '%s': %s\n"),
	filename, strerror(errno));
	return (1);
	}
	progbuf[progread] = '\0';

	/*
	* Any remaining arguments are passed as arguments to the lua script as
	* a string array:
	* {
	* "argv" -> [ "arg 1", ... "arg n" ],
	* }
	*/
	nvlist_t *argnvl = fnvlist_alloc();
	fnvlist_add_string_array(argnvl, ZCP_ARG_CLIARGV,
	(const char **)argv + 2, argc - 2);

	if (sync_flag) {
	ret = lzc_channel_program(poolname, progbuf,
	instrlimit, memlimit, argnvl, &outnvl);
	} else {
	ret = lzc_channel_program_nosync(poolname, progbuf,
	instrlimit, memlimit, argnvl, &outnvl);
	}

	if (ret != 0) {
	/*
	* On error, report the error message handed back by lua if one
	* exists. Otherwise, generate an appropriate error message,
	* falling back on strerror() for an unexpected return code.
	*/
	const char *errstring = NULL;
	const char *msg = gettext("Channel program execution failed");
	uint64_t instructions = 0;
	if (outnvl != NULL && nvlist_exists(outnvl, ZCP_RET_ERROR)) {
	const char *es = NULL;
	(void) nvlist_lookup_string(outnvl,
	ZCP_RET_ERROR, &es);
	if (es == NULL)
	errstring = strerror(ret);
	else
	errstring = es;
	if (ret == ETIME) {
	(void) nvlist_lookup_uint64(outnvl,
	ZCP_ARG_INSTRLIMIT, &instructions);
	}
	} else {
	switch (ret) {
	case EINVAL:
	errstring =
	"Invalid instruction or memory limit.";
	break;
	case ENOMEM:
	errstring = "Return value too large.";
	break;
	case ENOSPC:
	errstring = "Memory limit exhausted.";
	break;
	case ETIME:
	errstring = "Timed out.";
	break;
	case EPERM:
	errstring = "Permission denied. Channel "
	"programs must be run as root.";
	break;
	default:
	(void) zfs_standard_error(g_zfs, ret, msg);
	}
	}
	if (errstring != NULL)
	(void) fprintf(stderr, "%s:\n%s\n", msg, errstring);

	if (ret == ETIME && instructions != 0)
	(void) fprintf(stderr,
	gettext("%llu Lua instructions\n"),
	(u_longlong_t)instructions);
	} else {
	if (json_output) {
	(void) nvlist_print_json(stdout, outnvl);
	} else if (nvlist_empty(outnvl)) {
	(void) fprintf(stdout, gettext("Channel program fully "
	"executed and did not produce output.\n"));
	} else {
	(void) fprintf(stdout, gettext("Channel program fully "
	"executed and produced output:\n"));
	dump_nvlist(outnvl, 4);
	}
	}

	free(progbuf);
	fnvlist_free(outnvl);
	fnvlist_free(argnvl);
	return (ret != 0);

	usage:
	usage(B_FALSE);
	return (-1);
	}


	typedef struct loadkey_cbdata {
	boolean_t cb_loadkey;
	boolean_t cb_recursive;
	boolean_t cb_noop;
	char *cb_keylocation;
	uint64_t cb_numfailed;
	uint64_t cb_numattempted;
	} loadkey_cbdata_t;

	static int
	load_key_callback(zfs_handle_t zhp, void data)
	{
	int ret;
	boolean_t is_encroot;
	loadkey_cbdata_t *cb = data;
	uint64_t keystatus = zfs_prop_get_int(zhp, ZFS_PROP_KEYSTATUS);

	/*
	* If we are working recursively, we want to skip loading / unloading
	* keys for non-encryption roots and datasets whose keys are already
	* in the desired end-state.
	*/
	if (cb->cb_recursive) {
	ret = zfs_crypto_get_encryption_root(zhp, &is_encroot, NULL);
	if (ret != 0)
	return (ret);
	if (!is_encroot)
	return (0);

	if ((cb->cb_loadkey && keystatus == ZFS_KEYSTATUS_AVAILABLE) \|\|
	(!cb->cb_loadkey && keystatus == ZFS_KEYSTATUS_UNAVAILABLE))
	return (0);
	}

	cb->cb_numattempted++;

	if (cb->cb_loadkey)
	ret = zfs_crypto_load_key(zhp, cb->cb_noop, cb->cb_keylocation);
	else
	ret = zfs_crypto_unload_key(zhp);

	if (ret != 0) {
	cb->cb_numfailed++;
	return (ret);
	}

	return (0);
	}

	static int
	load_unload_keys(int argc, char **argv, boolean_t loadkey)
	{
	int c, ret = 0, flags = 0;
	boolean_t do_all = B_FALSE;
	loadkey_cbdata_t cb = { 0 };

	cb.cb_loadkey = loadkey;

	while ((c = getopt(argc, argv, "anrL:")) != -1) {
	/* noop and alternate keylocations only apply to zfs load-key */
	if (loadkey) {
	switch (c) {
	case 'n':
	cb.cb_noop = B_TRUE;
	continue;
	case 'L':
	cb.cb_keylocation = optarg;
	continue;
	default:
	break;
	}
	}

	switch (c) {
	case 'a':
	do_all = B_TRUE;
	cb.cb_recursive = B_TRUE;
	break;
	case 'r':
	flags \|= ZFS_ITER_RECURSE;
	cb.cb_recursive = B_TRUE;
	break;
	default:
	(void) fprintf(stderr,
	gettext("invalid option '%c'\n"), optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	if (!do_all && argc == 0) {
	(void) fprintf(stderr,
	gettext("Missing dataset argument or -a option\n"));
	usage(B_FALSE);
	}

	if (do_all && argc != 0) {
	(void) fprintf(stderr,
	gettext("Cannot specify dataset with -a option\n"));
	usage(B_FALSE);
	}

	if (cb.cb_recursive && cb.cb_keylocation != NULL &&
	strcmp(cb.cb_keylocation, "prompt") != 0) {
	(void) fprintf(stderr, gettext("alternate keylocation may only "
	"be 'prompt' with -r or -a\n"));
	usage(B_FALSE);
	}

	ret = zfs_for_each(argc, argv, flags,
	ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME, NULL, NULL, 0,
	load_key_callback, &cb);

	if (cb.cb_noop \|\| (cb.cb_recursive && cb.cb_numattempted != 0)) {
	(void) printf(gettext("%llu / %llu key(s) successfully %s\n"),
	(u_longlong_t)(cb.cb_numattempted - cb.cb_numfailed),
	(u_longlong_t)cb.cb_numattempted,
	loadkey ? (cb.cb_noop ? "verified" : "loaded") :
	"unloaded");
	}

	if (cb.cb_numfailed != 0)
	ret = -1;

	return (ret);
	}

	static int
	zfs_do_load_key(int argc, char **argv)
	{
	return (load_unload_keys(argc, argv, B_TRUE));
	}


	static int
	zfs_do_unload_key(int argc, char **argv)
	{
	return (load_unload_keys(argc, argv, B_FALSE));
	}

	static int
	zfs_do_change_key(int argc, char **argv)
	{
	int c, ret;
	uint64_t keystatus;
	boolean_t loadkey = B_FALSE, inheritkey = B_FALSE;
	zfs_handle_t *zhp = NULL;
	nvlist_t *props = fnvlist_alloc();

	while ((c = getopt(argc, argv, "lio:")) != -1) {
	switch (c) {
	case 'l':
	loadkey = B_TRUE;
	break;
	case 'i':
	inheritkey = B_TRUE;
	break;
	case 'o':
	if (!parseprop(props, optarg)) {
	nvlist_free(props);
	return (1);
	}
	break;
	default:
	(void) fprintf(stderr,
	gettext("invalid option '%c'\n"), optopt);
	usage(B_FALSE);
	}
	}

	if (inheritkey && !nvlist_empty(props)) {
	(void) fprintf(stderr,
	gettext("Properties not allowed for inheriting\n"));
	usage(B_FALSE);
	}

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("Missing dataset argument\n"));
	usage(B_FALSE);
	}

	if (argc > 1) {
	(void) fprintf(stderr, gettext("Too many arguments\n"));
	usage(B_FALSE);
	}

	zhp = zfs_open(g_zfs, argv[argc - 1],
	ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME);
	if (zhp == NULL)
	usage(B_FALSE);

	if (loadkey) {
	keystatus = zfs_prop_get_int(zhp, ZFS_PROP_KEYSTATUS);
	if (keystatus != ZFS_KEYSTATUS_AVAILABLE) {
	ret = zfs_crypto_load_key(zhp, B_FALSE, NULL);
	if (ret != 0) {
	nvlist_free(props);
	zfs_close(zhp);
	return (-1);
	}
	}

	/* refresh the properties so the new keystatus is visible */
	zfs_refresh_properties(zhp);
	}

	ret = zfs_crypto_rewrap(zhp, props, inheritkey);
	if (ret != 0) {
	nvlist_free(props);
	zfs_close(zhp);
	return (-1);
	}

	nvlist_free(props);
	zfs_close(zhp);
	return (0);
	}

	/*
	* 1) zfs project [-d\|-r] <file\|directory ...>
	* List project ID and inherit flag of file(s) or directories.
	* -d: List the directory itself, not its children.
	* -r: List subdirectories recursively.
	*
	* 2) zfs project -C [-k] [-r] <file\|directory ...>
	* Clear project inherit flag and/or ID on the file(s) or directories.
	* -k: Keep the project ID unchanged. If not specified, the project ID
	* will be reset as zero.
	* -r: Clear on subdirectories recursively.
	*
	* 3) zfs project -c [-0] [-d\|-r] [-p id] <file\|directory ...>
	* Check project ID and inherit flag on the file(s) or directories,
	* report the outliers.
	* -0: Print file name followed by a NUL instead of newline.
	* -d: Check the directory itself, not its children.
	* -p: Specify the referenced ID for comparing with the target file(s)
	* or directories' project IDs. If not specified, the target (top)
	* directory's project ID will be used as the referenced one.
	* -r: Check subdirectories recursively.
	*
	* 4) zfs project [-p id] [-r] [-s] <file\|directory ...>
	* Set project ID and/or inherit flag on the file(s) or directories.
	* -p: Set the project ID as the given id.
	* -r: Set on subdirectories recursively. If not specify "-p" option,
	* it will use top-level directory's project ID as the given id,
	* then set both project ID and inherit flag on all descendants
	* of the top-level directory.
	* -s: Set project inherit flag.
	*/
	static int
	zfs_do_project(int argc, char **argv)
	{
	zfs_project_control_t zpc = {
	.zpc_expected_projid = ZFS_INVALID_PROJID,
	.zpc_op = ZFS_PROJECT_OP_DEFAULT,
	.zpc_dironly = B_FALSE,
	.zpc_keep_projid = B_FALSE,
	.zpc_newline = B_TRUE,
	.zpc_recursive = B_FALSE,
	.zpc_set_flag = B_FALSE,
	};
	int ret = 0, c;

	if (argc < 2)
	usage(B_FALSE);

	while ((c = getopt(argc, argv, "0Ccdkp:rs")) != -1) {
	switch (c) {
	case '0':
	zpc.zpc_newline = B_FALSE;
	break;
	case 'C':
	if (zpc.zpc_op != ZFS_PROJECT_OP_DEFAULT) {
	(void) fprintf(stderr, gettext("cannot "
	"specify '-C' '-c' '-s' together\n"));
	usage(B_FALSE);
	}

	zpc.zpc_op = ZFS_PROJECT_OP_CLEAR;
	break;
	case 'c':
	if (zpc.zpc_op != ZFS_PROJECT_OP_DEFAULT) {
	(void) fprintf(stderr, gettext("cannot "
	"specify '-C' '-c' '-s' together\n"));
	usage(B_FALSE);
	}

	zpc.zpc_op = ZFS_PROJECT_OP_CHECK;
	break;
	case 'd':
	zpc.zpc_dironly = B_TRUE;
	/* overwrite "-r" option */
	zpc.zpc_recursive = B_FALSE;
	break;
	case 'k':
	zpc.zpc_keep_projid = B_TRUE;
	break;
	case 'p': {
	char *endptr;

	errno = 0;
	zpc.zpc_expected_projid = strtoull(optarg, &endptr, 0);
	if (errno != 0 \|\| *endptr != '\0') {
	(void) fprintf(stderr,
	gettext("project ID must be less than "
	"%u\n"), UINT32_MAX);
	usage(B_FALSE);
	}
	if (zpc.zpc_expected_projid >= UINT32_MAX) {
	(void) fprintf(stderr,
	gettext("invalid project ID\n"));
	usage(B_FALSE);
	}
	break;
	}
	case 'r':
	zpc.zpc_recursive = B_TRUE;
	/* overwrite "-d" option */
	zpc.zpc_dironly = B_FALSE;
	break;
	case 's':
	if (zpc.zpc_op != ZFS_PROJECT_OP_DEFAULT) {
	(void) fprintf(stderr, gettext("cannot "
	"specify '-C' '-c' '-s' together\n"));
	usage(B_FALSE);
	}

	zpc.zpc_set_flag = B_TRUE;
	zpc.zpc_op = ZFS_PROJECT_OP_SET;
	break;
	default:
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	if (zpc.zpc_op == ZFS_PROJECT_OP_DEFAULT) {
	if (zpc.zpc_expected_projid != ZFS_INVALID_PROJID)
	zpc.zpc_op = ZFS_PROJECT_OP_SET;
	else
	zpc.zpc_op = ZFS_PROJECT_OP_LIST;
	}

	switch (zpc.zpc_op) {
	case ZFS_PROJECT_OP_LIST:
	if (zpc.zpc_keep_projid) {
	(void) fprintf(stderr,
	gettext("'-k' is only valid together with '-C'\n"));
	usage(B_FALSE);
	}
	if (!zpc.zpc_newline) {
	(void) fprintf(stderr,
	gettext("'-0' is only valid together with '-c'\n"));
	usage(B_FALSE);
	}
	break;
	case ZFS_PROJECT_OP_CHECK:
	if (zpc.zpc_keep_projid) {
	(void) fprintf(stderr,
	gettext("'-k' is only valid together with '-C'\n"));
	usage(B_FALSE);
	}
	break;
	case ZFS_PROJECT_OP_CLEAR:
	if (zpc.zpc_dironly) {
	(void) fprintf(stderr,
	gettext("'-d' is useless together with '-C'\n"));
	usage(B_FALSE);
	}
	if (!zpc.zpc_newline) {
	(void) fprintf(stderr,
	gettext("'-0' is only valid together with '-c'\n"));
	usage(B_FALSE);
	}
	if (zpc.zpc_expected_projid != ZFS_INVALID_PROJID) {
	(void) fprintf(stderr,
	gettext("'-p' is useless together with '-C'\n"));
	usage(B_FALSE);
	}
	break;
	case ZFS_PROJECT_OP_SET:
	if (zpc.zpc_dironly) {
	(void) fprintf(stderr,
	gettext("'-d' is useless for set project ID and/or "
	"inherit flag\n"));
	usage(B_FALSE);
	}
	if (zpc.zpc_keep_projid) {
	(void) fprintf(stderr,
	gettext("'-k' is only valid together with '-C'\n"));
	usage(B_FALSE);
	}
	if (!zpc.zpc_newline) {
	(void) fprintf(stderr,
	gettext("'-0' is only valid together with '-c'\n"));
	usage(B_FALSE);
	}
	break;
	default:
	ASSERT(0);
	break;
	}

	argv += optind;
	argc -= optind;
	if (argc == 0) {
	(void) fprintf(stderr,
	gettext("missing file or directory target(s)\n"));
	usage(B_FALSE);
	}

	for (int i = 0; i < argc; i++) {
	int err;

	err = zfs_project_handle(argv[i], &zpc);
	if (err && !ret)
	ret = err;
	}

	return (ret);
	}

	static int
	zfs_do_wait(int argc, char **argv)
	{
	boolean_t enabled[ZFS_WAIT_NUM_ACTIVITIES];
	int error = 0, i;
	int c;

	/* By default, wait for all types of activity. */
	for (i = 0; i < ZFS_WAIT_NUM_ACTIVITIES; i++)
	enabled[i] = B_TRUE;

	while ((c = getopt(argc, argv, "t:")) != -1) {
	switch (c) {
	case 't':
	/* Reset activities array */
	memset(&enabled, 0, sizeof (enabled));

	for (char *tok; (tok = strsep(&optarg, ",")); ) {
	static const char *const col_subopts[
	ZFS_WAIT_NUM_ACTIVITIES] = { "deleteq" };

	for (i = 0; i < ARRAY_SIZE(col_subopts); ++i)
	if (strcmp(tok, col_subopts[i]) == 0) {
	enabled[i] = B_TRUE;
	goto found;
	}

	(void) fprintf(stderr,
	gettext("invalid activity '%s'\n"), tok);
	usage(B_FALSE);
	found:;
	}
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argv += optind;
	argc -= optind;
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing 'filesystem' "
	"argument\n"));
	usage(B_FALSE);
	}
	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	zfs_handle_t *zhp = zfs_open(g_zfs, argv[0], ZFS_TYPE_FILESYSTEM);
	if (zhp == NULL)
	return (1);

	for (;;) {
	boolean_t missing = B_FALSE;
	boolean_t any_waited = B_FALSE;

	for (int i = 0; i < ZFS_WAIT_NUM_ACTIVITIES; i++) {
	boolean_t waited;

	if (!enabled[i])
	continue;

	error = zfs_wait_status(zhp, i, &missing, &waited);
	if (error != 0 \|\| missing)
	break;

	any_waited = (any_waited \|\| waited);
	}

	if (error != 0 \|\| missing \|\| !any_waited)
	break;
	}

	zfs_close(zhp);

	return (error);
	}

	/*
	* Display version message
	*/
	static int
	zfs_do_version(int argc, char **argv)
	{
	(void) argc, (void) argv;
	return (zfs_version_print() != 0);
	}

	/* Display documentation */
	static int
	zfs_do_help(int argc, char **argv)
	{
	char page[MAXNAMELEN];
	if (argc < 3 \|\| strcmp(argv[2], "zfs") == 0)
	strcpy(page, "zfs");
	else if (strcmp(argv[2], "concepts") == 0 \|\|
	strcmp(argv[2], "props") == 0)
	snprintf(page, sizeof (page), "zfs%s", argv[2]);
	else
	snprintf(page, sizeof (page), "zfs-%s", argv[2]);

	execlp("man", "man", page, NULL);

	fprintf(stderr, "couldn't run man program: %s", strerror(errno));
	return (-1);
	}

	int
	main(int argc, char **argv)
	{
	int ret = 0;
	int i = 0;
	const char *cmdname;
	char **newargv;

	(void) setlocale(LC_ALL, "");
	(void) setlocale(LC_NUMERIC, "C");
	(void) textdomain(TEXT_DOMAIN);

	opterr = 0;

	/*
	* Make sure the user has specified some command.
	*/
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing command\n"));
	usage(B_FALSE);
	}

	cmdname = argv[1];

	/*
	* The 'umount' command is an alias for 'unmount'
	*/
	if (strcmp(cmdname, "umount") == 0)
	cmdname = "unmount";

	/*
	* The 'recv' command is an alias for 'receive'
	*/
	if (strcmp(cmdname, "recv") == 0)
	cmdname = "receive";

	/*
	* The 'snap' command is an alias for 'snapshot'
	*/
	if (strcmp(cmdname, "snap") == 0)
	cmdname = "snapshot";

	/*
	* Special case '-?'
	*/
	if ((strcmp(cmdname, "-?") == 0) \|\|
	(strcmp(cmdname, "--help") == 0))
	usage(B_TRUE);

	/*
	* Special case '-V\|--version'
	*/
	if ((strcmp(cmdname, "-V") == 0) \|\| (strcmp(cmdname, "--version") == 0))
	return (zfs_do_version(argc, argv));

	/*
	* Special case 'help'
	*/
	if (strcmp(cmdname, "help") == 0)
	return (zfs_do_help(argc, argv));

	if ((g_zfs = libzfs_init()) == NULL) {
	(void) fprintf(stderr, "%s\n", libzfs_error_init(errno));
	return (1);
	}

	zfs_save_arguments(argc, argv, history_str, sizeof (history_str));

	libzfs_print_on_error(g_zfs, B_TRUE);

	zfs_setproctitle_init(argc, argv, environ);

	/*
	* Many commands modify input strings for string parsing reasons.
	* We create a copy to protect the original argv.
	*/
	newargv = safe_malloc((argc + 1) * sizeof (newargv[0]));
	for (i = 0; i < argc; i++)
	newargv[i] = strdup(argv[i]);
	newargv[argc] = NULL;

	/*
	* Run the appropriate command.
	*/
	libzfs_mnttab_cache(g_zfs, B_TRUE);
	if (find_command_idx(cmdname, &i) == 0) {
	current_command = &command_table[i];
	ret = command_table[i].func(argc - 1, newargv + 1);
	} else if (strchr(cmdname, '=') != NULL) {
	verify(find_command_idx("set", &i) == 0);
	current_command = &command_table[i];
	ret = command_table[i].func(argc, newargv);
	} else {
	(void) fprintf(stderr, gettext("unrecognized "
	"command '%s'\n"), cmdname);
	usage(B_FALSE);
	ret = 1;
	}

	for (i = 0; i < argc; i++)
	free(newargv[i]);
	free(newargv);

	if (ret == 0 && log_history)
	(void) zpool_log_history(g_zfs, history_str);

	libzfs_fini(g_zfs);

	/*
	* The 'ZFS_ABORT' environment variable causes us to dump core on exit
	* for the purposes of running ::findleaks.
	*/
	if (getenv("ZFS_ABORT") != NULL) {
	(void) printf("dumping core by request\n");
	abort();
	}

	return (ret);
	}

	/*
	* zfs zone nsfile filesystem
	*
	* Add or delete the given dataset to/from the namespace.
	*/
	#ifdef __linux__
	static int
	zfs_do_zone_impl(int argc, char **argv, boolean_t attach)
	{
	zfs_handle_t *zhp;
	int ret;

	if (argc < 3) {
	(void) fprintf(stderr, gettext("missing argument(s)\n"));
	usage(B_FALSE);
	}
	if (argc > 3) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	zhp = zfs_open(g_zfs, argv[2], ZFS_TYPE_FILESYSTEM);
	if (zhp == NULL)
	return (1);

	ret = (zfs_userns(zhp, argv[1], attach) != 0);

	zfs_close(zhp);
	return (ret);
	}

	static int
	zfs_do_zone(int argc, char **argv)
	{
	return (zfs_do_zone_impl(argc, argv, B_TRUE));
	}

	static int
	zfs_do_unzone(int argc, char **argv)
	{
	return (zfs_do_zone_impl(argc, argv, B_FALSE));
	}
	#endif

	#ifdef __FreeBSD__
	#include <sys/jail.h>
	#include <jail.h>
	/*
	* Attach/detach the given dataset to/from the given jail
	*/
	static int
	zfs_do_jail_impl(int argc, char **argv, boolean_t attach)
	{
	zfs_handle_t *zhp;
	int jailid, ret;

	/* check number of arguments */
	if (argc < 3) {
	(void) fprintf(stderr, gettext("missing argument(s)\n"));
	usage(B_FALSE);
	}
	if (argc > 3) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	jailid = jail_getid(argv[1]);
	if (jailid < 0) {
	(void) fprintf(stderr, gettext("invalid jail id or name\n"));
	usage(B_FALSE);
	}

	zhp = zfs_open(g_zfs, argv[2], ZFS_TYPE_FILESYSTEM);
	if (zhp == NULL)
	return (1);

	ret = (zfs_jail(zhp, jailid, attach) != 0);

	zfs_close(zhp);
	return (ret);
	}

	/*
	* zfs jail jailid filesystem
	*
	* Attach the given dataset to the given jail
	*/
	static int
	zfs_do_jail(int argc, char **argv)
	{
	return (zfs_do_jail_impl(argc, argv, B_TRUE));
	}

	/*
	* zfs unjail jailid filesystem
	*
	* Detach the given dataset from the given jail
	*/
	static int
	zfs_do_unjail(int argc, char **argv)
	{
	return (zfs_do_jail_impl(argc, argv, B_FALSE));
	}
	#endif
	diff --git a/sys/contrib/openzfs/cmd/zpool/zpool_main.c b/sys/contrib/openzfs/cmd/zpool/zpool_main.c
	index 300b383af4f6..57170c8ae717 100644
	--- a/sys/contrib/openzfs/cmd/zpool/zpool_main.c
	+++ b/sys/contrib/openzfs/cmd/zpool/zpool_main.c
	@@ -1,11668 +1,11761 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2011, 2024 by Delphix. All rights reserved.
	* Copyright (c) 2012 by Frederik Wessels. All rights reserved.
	* Copyright (c) 2012 by Cyril Plisko. All rights reserved.
	* Copyright (c) 2013 by Prasad Joshi (sTec). All rights reserved.
	* Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>.
	* Copyright (c) 2017 Datto Inc.
	* Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
	* Copyright (c) 2017, Intel Corporation.
	* Copyright (c) 2019, loli10K <ezomori.nozomu@gmail.com>
	* Copyright (c) 2021, Colm Buckley <colm@tuatha.org>
	* Copyright (c) 2021, Klara Inc.
	* Copyright [2021] Hewlett Packard Enterprise Development LP
	*/

	#include <assert.h>
	#include <ctype.h>
	#include <dirent.h>
	#include <errno.h>
	#include <fcntl.h>
	#include <getopt.h>
	#include <libgen.h>
	#include <libintl.h>
	#include <libuutil.h>
	#include <locale.h>
	#include <pthread.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <string.h>
	#include <thread_pool.h>
	#include <time.h>
	#include <unistd.h>
	#include <pwd.h>
	#include <zone.h>
	#include <sys/wait.h>
	#include <zfs_prop.h>
	#include <sys/fs/zfs.h>
	#include <sys/stat.h>
	#include <sys/systeminfo.h>
	#include <sys/fm/fs/zfs.h>
	#include <sys/fm/util.h>
	#include <sys/fm/protocol.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/mount.h>
	#include <sys/sysmacros.h>

	#include <math.h>

	#include <libzfs.h>
	#include <libzutil.h>

	#include "zpool_util.h"
	#include "zfs_comutil.h"
	#include "zfeature_common.h"

	#include "statcommon.h"

	libzfs_handle_t *g_zfs;

	+static int mount_tp_nthr = 512; /* tpool threads for multi-threaded mounting */
	+
	static int zpool_do_create(int, char **);
	static int zpool_do_destroy(int, char **);

	static int zpool_do_add(int, char **);
	static int zpool_do_remove(int, char **);
	static int zpool_do_labelclear(int, char **);

	static int zpool_do_checkpoint(int, char **);

	static int zpool_do_list(int, char **);
	static int zpool_do_iostat(int, char **);
	static int zpool_do_status(int, char **);

	static int zpool_do_online(int, char **);
	static int zpool_do_offline(int, char **);
	static int zpool_do_clear(int, char **);
	static int zpool_do_reopen(int, char **);

	static int zpool_do_reguid(int, char **);

	static int zpool_do_attach(int, char **);
	static int zpool_do_detach(int, char **);
	static int zpool_do_replace(int, char **);
	static int zpool_do_split(int, char **);

	static int zpool_do_initialize(int, char **);
	static int zpool_do_scrub(int, char **);
	static int zpool_do_resilver(int, char **);
	static int zpool_do_trim(int, char **);

	static int zpool_do_import(int, char **);
	static int zpool_do_export(int, char **);

	static int zpool_do_upgrade(int, char **);

	static int zpool_do_history(int, char **);
	static int zpool_do_events(int, char **);

	static int zpool_do_get(int, char **);
	static int zpool_do_set(int, char **);

	static int zpool_do_sync(int, char **);

	static int zpool_do_version(int, char **);

	static int zpool_do_wait(int, char **);

	static int zpool_do_help(int argc, char **argv);

	static zpool_compat_status_t zpool_do_load_compat(
	const char , boolean_t );

	enum zpool_options {
	ZPOOL_OPTION_POWER = 1024,
	ZPOOL_OPTION_ALLOW_INUSE,
	ZPOOL_OPTION_ALLOW_REPLICATION_MISMATCH,
	ZPOOL_OPTION_ALLOW_ASHIFT_MISMATCH
	};

	/*
	* These libumem hooks provide a reasonable set of defaults for the allocator's
	* debugging facilities.
	*/

	#ifdef DEBUG
	const char *
	_umem_debug_init(void)
	{
	return ("default,verbose"); /* $UMEM_DEBUG setting */
	}

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

	typedef enum {
	HELP_ADD,
	HELP_ATTACH,
	HELP_CLEAR,
	HELP_CREATE,
	HELP_CHECKPOINT,
	HELP_DESTROY,
	HELP_DETACH,
	HELP_EXPORT,
	HELP_HISTORY,
	HELP_IMPORT,
	HELP_IOSTAT,
	HELP_LABELCLEAR,
	HELP_LIST,
	HELP_OFFLINE,
	HELP_ONLINE,
	HELP_REPLACE,
	HELP_REMOVE,
	HELP_INITIALIZE,
	HELP_SCRUB,
	HELP_RESILVER,
	HELP_TRIM,
	HELP_STATUS,
	HELP_UPGRADE,
	HELP_EVENTS,
	HELP_GET,
	HELP_SET,
	HELP_SPLIT,
	HELP_SYNC,
	HELP_REGUID,
	HELP_REOPEN,
	HELP_VERSION,
	HELP_WAIT
	} zpool_help_t;


	/*
	* Flags for stats to display with "zpool iostats"
	*/
	enum iostat_type {
	IOS_DEFAULT = 0,
	IOS_LATENCY = 1,
	IOS_QUEUES = 2,
	IOS_L_HISTO = 3,
	IOS_RQ_HISTO = 4,
	IOS_COUNT, /* always last element */
	};

	/* iostat_type entries as bitmasks */
	#define IOS_DEFAULT_M (1ULL << IOS_DEFAULT)
	#define IOS_LATENCY_M (1ULL << IOS_LATENCY)
	#define IOS_QUEUES_M (1ULL << IOS_QUEUES)
	#define IOS_L_HISTO_M (1ULL << IOS_L_HISTO)
	#define IOS_RQ_HISTO_M (1ULL << IOS_RQ_HISTO)

	/* Mask of all the histo bits */
	#define IOS_ANYHISTO_M (IOS_L_HISTO_M \| IOS_RQ_HISTO_M)

	/*
	* Lookup table for iostat flags to nvlist names. Basically a list
	* of all the nvlists a flag requires. Also specifies the order in
	* which data gets printed in zpool iostat.
	*/
	static const char *vsx_type_to_nvlist[IOS_COUNT][15] = {
	[IOS_L_HISTO] = {
	ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_SYNC_R_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_SYNC_W_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_ASYNC_R_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_ASYNC_W_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_SCRUB_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_TRIM_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_REBUILD_LAT_HISTO,
	NULL},
	[IOS_LATENCY] = {
	ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_TRIM_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_REBUILD_LAT_HISTO,
	NULL},
	[IOS_QUEUES] = {
	ZPOOL_CONFIG_VDEV_SYNC_R_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_SYNC_W_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_ASYNC_R_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_ASYNC_W_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_SCRUB_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_TRIM_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_REBUILD_ACTIVE_QUEUE,
	NULL},
	[IOS_RQ_HISTO] = {
	ZPOOL_CONFIG_VDEV_SYNC_IND_R_HISTO,
	ZPOOL_CONFIG_VDEV_SYNC_AGG_R_HISTO,
	ZPOOL_CONFIG_VDEV_SYNC_IND_W_HISTO,
	ZPOOL_CONFIG_VDEV_SYNC_AGG_W_HISTO,
	ZPOOL_CONFIG_VDEV_ASYNC_IND_R_HISTO,
	ZPOOL_CONFIG_VDEV_ASYNC_AGG_R_HISTO,
	ZPOOL_CONFIG_VDEV_ASYNC_IND_W_HISTO,
	ZPOOL_CONFIG_VDEV_ASYNC_AGG_W_HISTO,
	ZPOOL_CONFIG_VDEV_IND_SCRUB_HISTO,
	ZPOOL_CONFIG_VDEV_AGG_SCRUB_HISTO,
	ZPOOL_CONFIG_VDEV_IND_TRIM_HISTO,
	ZPOOL_CONFIG_VDEV_AGG_TRIM_HISTO,
	ZPOOL_CONFIG_VDEV_IND_REBUILD_HISTO,
	ZPOOL_CONFIG_VDEV_AGG_REBUILD_HISTO,
	NULL},
	};


	/*
	* Given a cb->cb_flags with a histogram bit set, return the iostat_type.
	* Right now, only one histo bit is ever set at one time, so we can
	* just do a highbit64(a)
	*/
	#define IOS_HISTO_IDX(a) (highbit64(a & IOS_ANYHISTO_M) - 1)

	typedef struct zpool_command {
	const char *name;
	int (func)(int, char *);
	zpool_help_t usage;
	} zpool_command_t;

	/*
	* Master command table. Each ZFS command has a name, associated function, and
	* usage message. The usage messages need to be internationalized, so we have
	* to have a function to return the usage message based on a command index.
	*
	* These commands are organized according to how they are displayed in the usage
	* message. An empty command (one with a NULL name) indicates an empty line in
	* the generic usage message.
	*/
	static zpool_command_t command_table[] = {
	{ "version", zpool_do_version, HELP_VERSION },
	{ NULL },
	{ "create", zpool_do_create, HELP_CREATE },
	{ "destroy", zpool_do_destroy, HELP_DESTROY },
	{ NULL },
	{ "add", zpool_do_add, HELP_ADD },
	{ "remove", zpool_do_remove, HELP_REMOVE },
	{ NULL },
	{ "labelclear", zpool_do_labelclear, HELP_LABELCLEAR },
	{ NULL },
	{ "checkpoint", zpool_do_checkpoint, HELP_CHECKPOINT },
	{ NULL },
	{ "list", zpool_do_list, HELP_LIST },
	{ "iostat", zpool_do_iostat, HELP_IOSTAT },
	{ "status", zpool_do_status, HELP_STATUS },
	{ NULL },
	{ "online", zpool_do_online, HELP_ONLINE },
	{ "offline", zpool_do_offline, HELP_OFFLINE },
	{ "clear", zpool_do_clear, HELP_CLEAR },
	{ "reopen", zpool_do_reopen, HELP_REOPEN },
	{ NULL },
	{ "attach", zpool_do_attach, HELP_ATTACH },
	{ "detach", zpool_do_detach, HELP_DETACH },
	{ "replace", zpool_do_replace, HELP_REPLACE },
	{ "split", zpool_do_split, HELP_SPLIT },
	{ NULL },
	{ "initialize", zpool_do_initialize, HELP_INITIALIZE },
	{ "resilver", zpool_do_resilver, HELP_RESILVER },
	{ "scrub", zpool_do_scrub, HELP_SCRUB },
	{ "trim", zpool_do_trim, HELP_TRIM },
	{ NULL },
	{ "import", zpool_do_import, HELP_IMPORT },
	{ "export", zpool_do_export, HELP_EXPORT },
	{ "upgrade", zpool_do_upgrade, HELP_UPGRADE },
	{ "reguid", zpool_do_reguid, HELP_REGUID },
	{ NULL },
	{ "history", zpool_do_history, HELP_HISTORY },
	{ "events", zpool_do_events, HELP_EVENTS },
	{ NULL },
	{ "get", zpool_do_get, HELP_GET },
	{ "set", zpool_do_set, HELP_SET },
	{ "sync", zpool_do_sync, HELP_SYNC },
	{ NULL },
	{ "wait", zpool_do_wait, HELP_WAIT },
	};

	#define NCOMMAND (ARRAY_SIZE(command_table))

	#define VDEV_ALLOC_CLASS_LOGS "logs"

	static zpool_command_t *current_command;
	static zfs_type_t current_prop_type = (ZFS_TYPE_POOL \| ZFS_TYPE_VDEV);
	static char history_str[HIS_MAX_RECORD_LEN];
	static boolean_t log_history = B_TRUE;
	static uint_t timestamp_fmt = NODATE;

	static const char *
	get_usage(zpool_help_t idx)
	{
	switch (idx) {
	case HELP_ADD:
	return (gettext("\tadd [-afgLnP] [-o property=value] "
	"<pool> <vdev> ...\n"));
	case HELP_ATTACH:
	return (gettext("\tattach [-fsw] [-o property=value] "
	"<pool> <device> <new-device>\n"));
	case HELP_CLEAR:
	return (gettext("\tclear [[--power]\|[-nF]] <pool> [device]\n"));
	case HELP_CREATE:
	return (gettext("\tcreate [-fnd] [-o property=value] ... \n"
	"\t [-O file-system-property=value] ... \n"
	"\t [-m mountpoint] [-R root] <pool> <vdev> ...\n"));
	case HELP_CHECKPOINT:
	return (gettext("\tcheckpoint [-d [-w]] <pool> ...\n"));
	case HELP_DESTROY:
	return (gettext("\tdestroy [-f] <pool>\n"));
	case HELP_DETACH:
	return (gettext("\tdetach <pool> <device>\n"));
	case HELP_EXPORT:
	return (gettext("\texport [-af] <pool> ...\n"));
	case HELP_HISTORY:
	return (gettext("\thistory [-il] [<pool>] ...\n"));
	case HELP_IMPORT:
	return (gettext("\timport [-d dir] [-D]\n"
	"\timport [-o mntopts] [-o property=value] ... \n"
	"\t [-d dir \| -c cachefile] [-D] [-l] [-f] [-m] [-N] "
	"[-R root] [-F [-n]] -a\n"
	"\timport [-o mntopts] [-o property=value] ... \n"
	"\t [-d dir \| -c cachefile] [-D] [-l] [-f] [-m] [-N] "
	"[-R root] [-F [-n]]\n"
	"\t [--rewind-to-checkpoint] <pool \| id> [newpool]\n"));
	case HELP_IOSTAT:
	return (gettext("\tiostat [[[-c [script1,script2,...]"
	"[-lq]]\|[-rw]] [-T d \| u] [-ghHLpPvy]\n"
	"\t [[pool ...]\|[pool vdev ...]\|[vdev ...]]"
	" [[-n] interval [count]]\n"));
	case HELP_LABELCLEAR:
	return (gettext("\tlabelclear [-f] <vdev>\n"));
	case HELP_LIST:
	return (gettext("\tlist [-gHLpPv] [-o property[,...]] "
	"[-T d\|u] [pool] ... \n"
	"\t [interval [count]]\n"));
	case HELP_OFFLINE:
	return (gettext("\toffline [--power]\|[[-f][-t]] <pool> "
	"<device> ...\n"));
	case HELP_ONLINE:
	return (gettext("\tonline [--power][-e] <pool> <device> "
	"...\n"));
	case HELP_REPLACE:
	return (gettext("\treplace [-fsw] [-o property=value] "
	"<pool> <device> [new-device]\n"));
	case HELP_REMOVE:
	return (gettext("\tremove [-npsw] <pool> <device> ...\n"));
	case HELP_REOPEN:
	return (gettext("\treopen [-n] <pool>\n"));
	case HELP_INITIALIZE:
	return (gettext("\tinitialize [-c \| -s \| -u] [-w] <pool> "
	"[<device> ...]\n"));
	case HELP_SCRUB:
	return (gettext("\tscrub [-s \| -p] [-w] [-e] <pool> ...\n"));
	case HELP_RESILVER:
	return (gettext("\tresilver <pool> ...\n"));
	case HELP_TRIM:
	return (gettext("\ttrim [-dw] [-r <rate>] [-c \| -s] <pool> "
	"[<device> ...]\n"));
	case HELP_STATUS:
	return (gettext("\tstatus [--power] [-c [script1,script2,...]] "
	"[-DegiLpPstvx] [-T d\|u] [pool] ...\n"
	"\t [interval [count]]\n"));
	case HELP_UPGRADE:
	return (gettext("\tupgrade\n"
	"\tupgrade -v\n"
	"\tupgrade [-V version] <-a \| pool ...>\n"));
	case HELP_EVENTS:
	return (gettext("\tevents [-vHf [pool] \| -c]\n"));
	case HELP_GET:
	return (gettext("\tget [-Hp] [-o \"all\" \| field[,...]] "
	"<\"all\" \| property[,...]> <pool> ...\n"));
	case HELP_SET:
	return (gettext("\tset <property=value> <pool>\n"
	"\tset <vdev_property=value> <pool> <vdev>\n"));
	case HELP_SPLIT:
	return (gettext("\tsplit [-gLnPl] [-R altroot] [-o mntopts]\n"
	"\t [-o property=value] <pool> <newpool> "
	"[<device> ...]\n"));
	case HELP_REGUID:
	return (gettext("\treguid <pool>\n"));
	case HELP_SYNC:
	return (gettext("\tsync [pool] ...\n"));
	case HELP_VERSION:
	return (gettext("\tversion\n"));
	case HELP_WAIT:
	return (gettext("\twait [-Hp] [-T d\|u] [-t <activity>[,...]] "
	"<pool> [interval]\n"));
	default:
	__builtin_unreachable();
	}
	}

	static void
	zpool_collect_leaves(zpool_handle_t zhp, nvlist_t nvroot, nvlist_t *res)
	{
	uint_t children = 0;
	nvlist_t **child;
	uint_t i;

	(void) nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
	&child, &children);

	if (children == 0) {
	char *path = zpool_vdev_name(g_zfs, zhp, nvroot,
	VDEV_NAME_PATH);

	if (strcmp(path, VDEV_TYPE_INDIRECT) != 0 &&
	strcmp(path, VDEV_TYPE_HOLE) != 0)
	fnvlist_add_boolean(res, path);

	free(path);
	return;
	}

	for (i = 0; i < children; i++) {
	zpool_collect_leaves(zhp, child[i], res);
	}
	}

	/*
	* Callback routine that will print out a pool property value.
	*/
	static int
	print_pool_prop_cb(int prop, void *cb)
	{
	FILE *fp = cb;

	(void) fprintf(fp, "\t%-19s ", zpool_prop_to_name(prop));

	if (zpool_prop_readonly(prop))
	(void) fprintf(fp, " NO ");
	else
	(void) fprintf(fp, " YES ");

	if (zpool_prop_values(prop) == NULL)
	(void) fprintf(fp, "-\n");
	else
	(void) fprintf(fp, "%s\n", zpool_prop_values(prop));

	return (ZPROP_CONT);
	}

	/*
	* Callback routine that will print out a vdev property value.
	*/
	static int
	print_vdev_prop_cb(int prop, void *cb)
	{
	FILE *fp = cb;

	(void) fprintf(fp, "\t%-19s ", vdev_prop_to_name(prop));

	if (vdev_prop_readonly(prop))
	(void) fprintf(fp, " NO ");
	else
	(void) fprintf(fp, " YES ");

	if (vdev_prop_values(prop) == NULL)
	(void) fprintf(fp, "-\n");
	else
	(void) fprintf(fp, "%s\n", vdev_prop_values(prop));

	return (ZPROP_CONT);
	}

	/*
	* Given a leaf vdev name like 'L5' return its VDEV_CONFIG_PATH like
	* '/dev/disk/by-vdev/L5'.
	*/
	static const char *
	vdev_name_to_path(zpool_handle_t zhp, char vdev)
	{
	nvlist_t *vdev_nv = zpool_find_vdev(zhp, vdev, NULL, NULL, NULL);
	if (vdev_nv == NULL) {
	return (NULL);
	}
	return (fnvlist_lookup_string(vdev_nv, ZPOOL_CONFIG_PATH));
	}

	static int
	zpool_power_on(zpool_handle_t zhp, char vdev)
	{
	return (zpool_power(zhp, vdev, B_TRUE));
	}

	static int
	zpool_power_on_and_disk_wait(zpool_handle_t zhp, char vdev)
	{
	int rc;

	rc = zpool_power_on(zhp, vdev);
	if (rc != 0)
	return (rc);

	zpool_disk_wait(vdev_name_to_path(zhp, vdev));

	return (0);
	}

	static int
	zpool_power_on_pool_and_wait_for_devices(zpool_handle_t *zhp)
	{
	nvlist_t *nv;
	const char *path = NULL;
	int rc;

	/* Power up all the devices first */
	FOR_EACH_REAL_LEAF_VDEV(zhp, nv) {
	path = fnvlist_lookup_string(nv, ZPOOL_CONFIG_PATH);
	if (path != NULL) {
	rc = zpool_power_on(zhp, (char *)path);
	if (rc != 0) {
	return (rc);
	}
	}
	}

	/*
	* Wait for their devices to show up. Since we powered them on
	* at roughly the same time, they should all come online around
	* the same time.
	*/
	FOR_EACH_REAL_LEAF_VDEV(zhp, nv) {
	path = fnvlist_lookup_string(nv, ZPOOL_CONFIG_PATH);
	zpool_disk_wait(path);
	}

	return (0);
	}

	static int
	zpool_power_off(zpool_handle_t zhp, char vdev)
	{
	return (zpool_power(zhp, vdev, B_FALSE));
	}

	/*
	* Display usage message. If we're inside a command, display only the usage for
	* that command. Otherwise, iterate over the entire command table and display
	* a complete usage message.
	*/
	static __attribute__((noreturn)) void
	usage(boolean_t requested)
	{
	FILE *fp = requested ? stdout : stderr;

	if (current_command == NULL) {
	int i;

	(void) fprintf(fp, gettext("usage: zpool command args ...\n"));
	(void) fprintf(fp,
	gettext("where 'command' is one of the following:\n\n"));

	for (i = 0; i < NCOMMAND; i++) {
	if (command_table[i].name == NULL)
	(void) fprintf(fp, "\n");
	else
	(void) fprintf(fp, "%s",
	get_usage(command_table[i].usage));
	}

	(void) fprintf(fp,
	gettext("\nFor further help on a command or topic, "
	"run: %s\n"), "zpool help [<topic>]");
	} else {
	(void) fprintf(fp, gettext("usage:\n"));
	(void) fprintf(fp, "%s", get_usage(current_command->usage));
	}

	if (current_command != NULL &&
	current_prop_type != (ZFS_TYPE_POOL \| ZFS_TYPE_VDEV) &&
	((strcmp(current_command->name, "set") == 0) \|\|
	(strcmp(current_command->name, "get") == 0) \|\|
	(strcmp(current_command->name, "list") == 0))) {

	(void) fprintf(fp, "%s",
	gettext("\nthe following properties are supported:\n"));

	(void) fprintf(fp, "\n\t%-19s %s %s\n\n",
	"PROPERTY", "EDIT", "VALUES");

	/* Iterate over all properties */
	if (current_prop_type == ZFS_TYPE_POOL) {
	(void) zprop_iter(print_pool_prop_cb, fp, B_FALSE,
	B_TRUE, current_prop_type);

	(void) fprintf(fp, "\t%-19s ", "feature@...");
	(void) fprintf(fp, "YES "
	"disabled \| enabled \| active\n");

	(void) fprintf(fp, gettext("\nThe feature@ properties "
	"must be appended with a feature name.\n"
	"See zpool-features(7).\n"));
	} else if (current_prop_type == ZFS_TYPE_VDEV) {
	(void) zprop_iter(print_vdev_prop_cb, fp, B_FALSE,
	B_TRUE, current_prop_type);
	}
	}

	/*
	* See comments at end of main().
	*/
	if (getenv("ZFS_ABORT") != NULL) {
	(void) printf("dumping core by request\n");
	abort();
	}

	exit(requested ? 0 : 2);
	}

	/*
	* zpool initialize [-c \| -s \| -u] [-w] <pool> [<vdev> ...]
	* Initialize all unused blocks in the specified vdevs, or all vdevs in the pool
	* if none specified.
	*
	* -c Cancel. Ends active initializing.
	* -s Suspend. Initializing can then be restarted with no flags.
	* -u Uninitialize. Clears initialization state.
	* -w Wait. Blocks until initializing has completed.
	*/
	int
	zpool_do_initialize(int argc, char **argv)
	{
	int c;
	char *poolname;
	zpool_handle_t *zhp;
	nvlist_t *vdevs;
	int err = 0;
	boolean_t wait = B_FALSE;

	struct option long_options[] = {
	{"cancel", no_argument, NULL, 'c'},
	{"suspend", no_argument, NULL, 's'},
	{"uninit", no_argument, NULL, 'u'},
	{"wait", no_argument, NULL, 'w'},
	{0, 0, 0, 0}
	};

	pool_initialize_func_t cmd_type = POOL_INITIALIZE_START;
	while ((c = getopt_long(argc, argv, "csuw", long_options,
	NULL)) != -1) {
	switch (c) {
	case 'c':
	if (cmd_type != POOL_INITIALIZE_START &&
	cmd_type != POOL_INITIALIZE_CANCEL) {
	(void) fprintf(stderr, gettext("-c cannot be "
	"combined with other options\n"));
	usage(B_FALSE);
	}
	cmd_type = POOL_INITIALIZE_CANCEL;
	break;
	case 's':
	if (cmd_type != POOL_INITIALIZE_START &&
	cmd_type != POOL_INITIALIZE_SUSPEND) {
	(void) fprintf(stderr, gettext("-s cannot be "
	"combined with other options\n"));
	usage(B_FALSE);
	}
	cmd_type = POOL_INITIALIZE_SUSPEND;
	break;
	case 'u':
	if (cmd_type != POOL_INITIALIZE_START &&
	cmd_type != POOL_INITIALIZE_UNINIT) {
	(void) fprintf(stderr, gettext("-u cannot be "
	"combined with other options\n"));
	usage(B_FALSE);
	}
	cmd_type = POOL_INITIALIZE_UNINIT;
	break;
	case 'w':
	wait = B_TRUE;
	break;
	case '?':
	if (optopt != 0) {
	(void) fprintf(stderr,
	gettext("invalid option '%c'\n"), optopt);
	} else {
	(void) fprintf(stderr,
	gettext("invalid option '%s'\n"),
	argv[optind - 1]);
	}
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name argument\n"));
	usage(B_FALSE);
	return (-1);
	}

	if (wait && (cmd_type != POOL_INITIALIZE_START)) {
	(void) fprintf(stderr, gettext("-w cannot be used with -c, -s"
	"or -u\n"));
	usage(B_FALSE);
	}

	poolname = argv[0];
	zhp = zpool_open(g_zfs, poolname);
	if (zhp == NULL)
	return (-1);

	vdevs = fnvlist_alloc();
	if (argc == 1) {
	/* no individual leaf vdevs specified, so add them all */
	nvlist_t *config = zpool_get_config(zhp, NULL);
	nvlist_t *nvroot = fnvlist_lookup_nvlist(config,
	ZPOOL_CONFIG_VDEV_TREE);
	zpool_collect_leaves(zhp, nvroot, vdevs);
	} else {
	for (int i = 1; i < argc; i++) {
	fnvlist_add_boolean(vdevs, argv[i]);
	}
	}

	if (wait)
	err = zpool_initialize_wait(zhp, cmd_type, vdevs);
	else
	err = zpool_initialize(zhp, cmd_type, vdevs);

	fnvlist_free(vdevs);
	zpool_close(zhp);

	return (err);
	}

	/*
	* print a pool vdev config for dry runs
	*/
	static void
	print_vdev_tree(zpool_handle_t zhp, const char name, nvlist_t *nv, int indent,
	const char *match, int name_flags)
	{
	nvlist_t **child;
	uint_t c, children;
	char *vname;
	boolean_t printed = B_FALSE;

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	&child, &children) != 0) {
	if (name != NULL)
	(void) printf("\t%*s%s\n", indent, "", name);
	return;
	}

	for (c = 0; c < children; c++) {
	uint64_t is_log = B_FALSE, is_hole = B_FALSE;
	const char *class = "";

	(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
	&is_hole);

	if (is_hole == B_TRUE) {
	continue;
	}

	(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
	&is_log);
	if (is_log)
	class = VDEV_ALLOC_BIAS_LOG;
	(void) nvlist_lookup_string(child[c],
	ZPOOL_CONFIG_ALLOCATION_BIAS, &class);
	if (strcmp(match, class) != 0)
	continue;

	if (!printed && name != NULL) {
	(void) printf("\t%*s%s\n", indent, "", name);
	printed = B_TRUE;
	}
	vname = zpool_vdev_name(g_zfs, zhp, child[c], name_flags);
	print_vdev_tree(zhp, vname, child[c], indent + 2, "",
	name_flags);
	free(vname);
	}
	}

	/*
	* Print the list of l2cache devices for dry runs.
	*/
	static void
	print_cache_list(nvlist_t *nv, int indent)
	{
	nvlist_t **child;
	uint_t c, children;

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
	&child, &children) == 0 && children > 0) {
	(void) printf("\t%*s%s\n", indent, "", "cache");
	} else {
	return;
	}
	for (c = 0; c < children; c++) {
	char *vname;

	vname = zpool_vdev_name(g_zfs, NULL, child[c], 0);
	(void) printf("\t%*s%s\n", indent + 2, "", vname);
	free(vname);
	}
	}

	/*
	* Print the list of spares for dry runs.
	*/
	static void
	print_spare_list(nvlist_t *nv, int indent)
	{
	nvlist_t **child;
	uint_t c, children;

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
	&child, &children) == 0 && children > 0) {
	(void) printf("\t%*s%s\n", indent, "", "spares");
	} else {
	return;
	}
	for (c = 0; c < children; c++) {
	char *vname;

	vname = zpool_vdev_name(g_zfs, NULL, child[c], 0);
	(void) printf("\t%*s%s\n", indent + 2, "", vname);
	free(vname);
	}
	}

	static boolean_t
	prop_list_contains_feature(nvlist_t *proplist)
	{
	nvpair_t *nvp;
	for (nvp = nvlist_next_nvpair(proplist, NULL); NULL != nvp;
	nvp = nvlist_next_nvpair(proplist, nvp)) {
	if (zpool_prop_feature(nvpair_name(nvp)))
	return (B_TRUE);
	}
	return (B_FALSE);
	}

	/*
	* Add a property pair (name, string-value) into a property nvlist.
	*/
	static int
	add_prop_list(const char propname, const char propval, nvlist_t **props,
	boolean_t poolprop)
	{
	zpool_prop_t prop = ZPOOL_PROP_INVAL;
	nvlist_t *proplist;
	const char *normnm;
	const char *strval;

	if (*props == NULL &&
	nvlist_alloc(props, NV_UNIQUE_NAME, 0) != 0) {
	(void) fprintf(stderr,
	gettext("internal error: out of memory\n"));
	return (1);
	}

	proplist = *props;

	if (poolprop) {
	const char *vname = zpool_prop_to_name(ZPOOL_PROP_VERSION);
	const char *cname =
	zpool_prop_to_name(ZPOOL_PROP_COMPATIBILITY);

	if ((prop = zpool_name_to_prop(propname)) == ZPOOL_PROP_INVAL &&
	(!zpool_prop_feature(propname) &&
	!zpool_prop_vdev(propname))) {
	(void) fprintf(stderr, gettext("property '%s' is "
	"not a valid pool or vdev property\n"), propname);
	return (2);
	}

	/*
	* feature@ properties and version should not be specified
	* at the same time.
	*/
	if ((prop == ZPOOL_PROP_INVAL && zpool_prop_feature(propname) &&
	nvlist_exists(proplist, vname)) \|\|
	(prop == ZPOOL_PROP_VERSION &&
	prop_list_contains_feature(proplist))) {
	(void) fprintf(stderr, gettext("'feature@' and "
	"'version' properties cannot be specified "
	"together\n"));
	return (2);
	}

	/*
	* if version is specified, only "legacy" compatibility
	* may be requested
	*/
	if ((prop == ZPOOL_PROP_COMPATIBILITY &&
	strcmp(propval, ZPOOL_COMPAT_LEGACY) != 0 &&
	nvlist_exists(proplist, vname)) \|\|
	(prop == ZPOOL_PROP_VERSION &&
	nvlist_exists(proplist, cname) &&
	strcmp(fnvlist_lookup_string(proplist, cname),
	ZPOOL_COMPAT_LEGACY) != 0)) {
	(void) fprintf(stderr, gettext("when 'version' is "
	"specified, the 'compatibility' feature may only "
	"be set to '" ZPOOL_COMPAT_LEGACY "'\n"));
	return (2);
	}

	if (zpool_prop_feature(propname) \|\| zpool_prop_vdev(propname))
	normnm = propname;
	else
	normnm = zpool_prop_to_name(prop);
	} else {
	zfs_prop_t fsprop = zfs_name_to_prop(propname);

	if (zfs_prop_valid_for_type(fsprop, ZFS_TYPE_FILESYSTEM,
	B_FALSE)) {
	normnm = zfs_prop_to_name(fsprop);
	} else if (zfs_prop_user(propname) \|\|
	zfs_prop_userquota(propname)) {
	normnm = propname;
	} else {
	(void) fprintf(stderr, gettext("property '%s' is "
	"not a valid filesystem property\n"), propname);
	return (2);
	}
	}

	if (nvlist_lookup_string(proplist, normnm, &strval) == 0 &&
	prop != ZPOOL_PROP_CACHEFILE) {
	(void) fprintf(stderr, gettext("property '%s' "
	"specified multiple times\n"), propname);
	return (2);
	}

	if (nvlist_add_string(proplist, normnm, propval) != 0) {
	(void) fprintf(stderr, gettext("internal "
	"error: out of memory\n"));
	return (1);
	}

	return (0);
	}

	/*
	* Set a default property pair (name, string-value) in a property nvlist
	*/
	static int
	add_prop_list_default(const char propname, const char propval,
	nvlist_t **props)
	{
	const char *pval;

	if (nvlist_lookup_string(*props, propname, &pval) == 0)
	return (0);

	return (add_prop_list(propname, propval, props, B_TRUE));
	}

	/*
	* zpool add [-afgLnP] [-o property=value] <pool> <vdev> ...
	*
	* -a Disable the ashift validation checks
	* -f Force addition of devices, even if they appear in use
	* -g Display guid for individual vdev name.
	* -L Follow links when resolving vdev path name.
	* -n Do not add the devices, but display the resulting layout if
	* they were to be added.
	* -o Set property=value.
	* -P Display full path for vdev name.
	*
	* Adds the given vdevs to 'pool'. As with create, the bulk of this work is
	* handled by make_root_vdev(), which constructs the nvlist needed to pass to
	* libzfs.
	*/
	int
	zpool_do_add(int argc, char **argv)
	{
	boolean_t check_replication = B_TRUE;
	boolean_t check_inuse = B_TRUE;
	boolean_t dryrun = B_FALSE;
	boolean_t check_ashift = B_TRUE;
	boolean_t force = B_FALSE;
	int name_flags = 0;
	int c;
	nvlist_t *nvroot;
	char *poolname;
	int ret;
	zpool_handle_t *zhp;
	nvlist_t *config;
	nvlist_t *props = NULL;
	char *propval;

	struct option long_options[] = {
	{"allow-in-use", no_argument, NULL, ZPOOL_OPTION_ALLOW_INUSE},
	{"allow-replication-mismatch", no_argument, NULL,
	ZPOOL_OPTION_ALLOW_REPLICATION_MISMATCH},
	{"allow-ashift-mismatch", no_argument, NULL,
	ZPOOL_OPTION_ALLOW_ASHIFT_MISMATCH},
	{0, 0, 0, 0}
	};

	/* check options */
	while ((c = getopt_long(argc, argv, "fgLno:P", long_options, NULL))
	!= -1) {
	switch (c) {
	case 'f':
	force = B_TRUE;
	break;
	case 'g':
	name_flags \|= VDEV_NAME_GUID;
	break;
	case 'L':
	name_flags \|= VDEV_NAME_FOLLOW_LINKS;
	break;
	case 'n':
	dryrun = B_TRUE;
	break;
	case 'o':
	if ((propval = strchr(optarg, '=')) == NULL) {
	(void) fprintf(stderr, gettext("missing "
	"'=' for -o option\n"));
	usage(B_FALSE);
	}
	*propval = '\0';
	propval++;

	if ((strcmp(optarg, ZPOOL_CONFIG_ASHIFT) != 0) \|\|
	(add_prop_list(optarg, propval, &props, B_TRUE)))
	usage(B_FALSE);
	break;
	case 'P':
	name_flags \|= VDEV_NAME_PATH;
	break;
	case ZPOOL_OPTION_ALLOW_INUSE:
	check_inuse = B_FALSE;
	break;
	case ZPOOL_OPTION_ALLOW_REPLICATION_MISMATCH:
	check_replication = B_FALSE;
	break;
	case ZPOOL_OPTION_ALLOW_ASHIFT_MISMATCH:
	check_ashift = B_FALSE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* get pool name and check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name argument\n"));
	usage(B_FALSE);
	}
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing vdev specification\n"));
	usage(B_FALSE);
	}

	if (force) {
	if (!check_inuse \|\| !check_replication \|\| !check_ashift) {
	(void) fprintf(stderr, gettext("'-f' option is not "
	"allowed with '--allow-replication-mismatch', "
	"'--allow-ashift-mismatch', or "
	"'--allow-in-use'\n"));
	usage(B_FALSE);
	}
	check_inuse = B_FALSE;
	check_replication = B_FALSE;
	check_ashift = B_FALSE;
	}

	poolname = argv[0];

	argc--;
	argv++;

	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
	return (1);

	if ((config = zpool_get_config(zhp, NULL)) == NULL) {
	(void) fprintf(stderr, gettext("pool '%s' is unavailable\n"),
	poolname);
	zpool_close(zhp);
	return (1);
	}

	/* unless manually specified use "ashift" pool property (if set) */
	if (!nvlist_exists(props, ZPOOL_CONFIG_ASHIFT)) {
	int intval;
	zprop_source_t src;
	char strval[ZPOOL_MAXPROPLEN];

	intval = zpool_get_prop_int(zhp, ZPOOL_PROP_ASHIFT, &src);
	if (src != ZPROP_SRC_DEFAULT) {
	(void) sprintf(strval, "%" PRId32, intval);
	verify(add_prop_list(ZPOOL_CONFIG_ASHIFT, strval,
	&props, B_TRUE) == 0);
	}
	}

	/* pass off to make_root_vdev for processing */
	nvroot = make_root_vdev(zhp, props, !check_inuse,
	check_replication, B_FALSE, dryrun, argc, argv);
	if (nvroot == NULL) {
	zpool_close(zhp);
	return (1);
	}

	if (dryrun) {
	nvlist_t *poolnvroot;
	nvlist_t l2child, sparechild;
	uint_t l2children, sparechildren, c;
	char *vname;
	boolean_t hadcache = B_FALSE, hadspare = B_FALSE;

	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
	&poolnvroot) == 0);

	(void) printf(gettext("would update '%s' to the following "
	"configuration:\n\n"), zpool_get_name(zhp));

	/* print original main pool and new tree */
	print_vdev_tree(zhp, poolname, poolnvroot, 0, "",
	name_flags \| VDEV_NAME_TYPE_ID);
	print_vdev_tree(zhp, NULL, nvroot, 0, "", name_flags);

	/* print other classes: 'dedup', 'special', and 'log' */
	if (zfs_special_devs(poolnvroot, VDEV_ALLOC_BIAS_DEDUP)) {
	print_vdev_tree(zhp, "dedup", poolnvroot, 0,
	VDEV_ALLOC_BIAS_DEDUP, name_flags);
	print_vdev_tree(zhp, NULL, nvroot, 0,
	VDEV_ALLOC_BIAS_DEDUP, name_flags);
	} else if (zfs_special_devs(nvroot, VDEV_ALLOC_BIAS_DEDUP)) {
	print_vdev_tree(zhp, "dedup", nvroot, 0,
	VDEV_ALLOC_BIAS_DEDUP, name_flags);
	}

	if (zfs_special_devs(poolnvroot, VDEV_ALLOC_BIAS_SPECIAL)) {
	print_vdev_tree(zhp, "special", poolnvroot, 0,
	VDEV_ALLOC_BIAS_SPECIAL, name_flags);
	print_vdev_tree(zhp, NULL, nvroot, 0,
	VDEV_ALLOC_BIAS_SPECIAL, name_flags);
	} else if (zfs_special_devs(nvroot, VDEV_ALLOC_BIAS_SPECIAL)) {
	print_vdev_tree(zhp, "special", nvroot, 0,
	VDEV_ALLOC_BIAS_SPECIAL, name_flags);
	}

	if (num_logs(poolnvroot) > 0) {
	print_vdev_tree(zhp, "logs", poolnvroot, 0,
	VDEV_ALLOC_BIAS_LOG, name_flags);
	print_vdev_tree(zhp, NULL, nvroot, 0,
	VDEV_ALLOC_BIAS_LOG, name_flags);
	} else if (num_logs(nvroot) > 0) {
	print_vdev_tree(zhp, "logs", nvroot, 0,
	VDEV_ALLOC_BIAS_LOG, name_flags);
	}

	/* Do the same for the caches */
	if (nvlist_lookup_nvlist_array(poolnvroot, ZPOOL_CONFIG_L2CACHE,
	&l2child, &l2children) == 0 && l2children) {
	hadcache = B_TRUE;
	(void) printf(gettext("\tcache\n"));
	for (c = 0; c < l2children; c++) {
	vname = zpool_vdev_name(g_zfs, NULL,
	l2child[c], name_flags);
	(void) printf("\t %s\n", vname);
	free(vname);
	}
	}
	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
	&l2child, &l2children) == 0 && l2children) {
	if (!hadcache)
	(void) printf(gettext("\tcache\n"));
	for (c = 0; c < l2children; c++) {
	vname = zpool_vdev_name(g_zfs, NULL,
	l2child[c], name_flags);
	(void) printf("\t %s\n", vname);
	free(vname);
	}
	}
	/* And finally the spares */
	if (nvlist_lookup_nvlist_array(poolnvroot, ZPOOL_CONFIG_SPARES,
	&sparechild, &sparechildren) == 0 && sparechildren > 0) {
	hadspare = B_TRUE;
	(void) printf(gettext("\tspares\n"));
	for (c = 0; c < sparechildren; c++) {
	vname = zpool_vdev_name(g_zfs, NULL,
	sparechild[c], name_flags);
	(void) printf("\t %s\n", vname);
	free(vname);
	}
	}
	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
	&sparechild, &sparechildren) == 0 && sparechildren > 0) {
	if (!hadspare)
	(void) printf(gettext("\tspares\n"));
	for (c = 0; c < sparechildren; c++) {
	vname = zpool_vdev_name(g_zfs, NULL,
	sparechild[c], name_flags);
	(void) printf("\t %s\n", vname);
	free(vname);
	}
	}

	ret = 0;
	} else {
	ret = (zpool_add(zhp, nvroot, check_ashift) != 0);
	}

	nvlist_free(props);
	nvlist_free(nvroot);
	zpool_close(zhp);

	return (ret);
	}

	/*
	* zpool remove [-npsw] <pool> <vdev> ...
	*
	* Removes the given vdev from the pool.
	*/
	int
	zpool_do_remove(int argc, char **argv)
	{
	char *poolname;
	int i, ret = 0;
	zpool_handle_t *zhp = NULL;
	boolean_t stop = B_FALSE;
	int c;
	boolean_t noop = B_FALSE;
	boolean_t parsable = B_FALSE;
	boolean_t wait = B_FALSE;

	/* check options */
	while ((c = getopt(argc, argv, "npsw")) != -1) {
	switch (c) {
	case 'n':
	noop = B_TRUE;
	break;
	case 'p':
	parsable = B_TRUE;
	break;
	case 's':
	stop = B_TRUE;
	break;
	case 'w':
	wait = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* get pool name and check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name argument\n"));
	usage(B_FALSE);
	}

	poolname = argv[0];

	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
	return (1);

	if (stop && noop) {
	zpool_close(zhp);
	(void) fprintf(stderr, gettext("stop request ignored\n"));
	return (0);
	}

	if (stop) {
	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}
	if (zpool_vdev_remove_cancel(zhp) != 0)
	ret = 1;
	if (wait) {
	(void) fprintf(stderr, gettext("invalid option "
	"combination: -w cannot be used with -s\n"));
	usage(B_FALSE);
	}
	} else {
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing device\n"));
	usage(B_FALSE);
	}

	for (i = 1; i < argc; i++) {
	if (noop) {
	uint64_t size;

	if (zpool_vdev_indirect_size(zhp, argv[i],
	&size) != 0) {
	ret = 1;
	break;
	}
	if (parsable) {
	(void) printf("%s %llu\n",
	argv[i], (unsigned long long)size);
	} else {
	char valstr[32];
	zfs_nicenum(size, valstr,
	sizeof (valstr));
	(void) printf("Memory that will be "
	"used after removing %s: %s\n",
	argv[i], valstr);
	}
	} else {
	if (zpool_vdev_remove(zhp, argv[i]) != 0)
	ret = 1;
	}
	}

	if (ret == 0 && wait)
	ret = zpool_wait(zhp, ZPOOL_WAIT_REMOVE);
	}
	zpool_close(zhp);

	return (ret);
	}

	/*
	* Return 1 if a vdev is active (being used in a pool)
	* Return 0 if a vdev is inactive (offlined or faulted, or not in active pool)
	*
	* This is useful for checking if a disk in an active pool is offlined or
	* faulted.
	*/
	static int
	vdev_is_active(char *vdev_path)
	{
	int fd;
	fd = open(vdev_path, O_EXCL);
	if (fd < 0) {
	return (1); /* cant open O_EXCL - disk is active */
	}

	close(fd);
	return (0); /* disk is inactive in the pool */
	}

	/*
	* zpool labelclear [-f] <vdev>
	*
	* -f Force clearing the label for the vdevs which are members of
	* the exported or foreign pools.
	*
	* Verifies that the vdev is not active and zeros out the label information
	* on the device.
	*/
	int
	zpool_do_labelclear(int argc, char **argv)
	{
	char vdev[MAXPATHLEN];
	char *name = NULL;
	int c, fd = -1, ret = 0;
	nvlist_t *config;
	pool_state_t state;
	boolean_t inuse = B_FALSE;
	boolean_t force = B_FALSE;

	/* check options */
	while ((c = getopt(argc, argv, "f")) != -1) {
	switch (c) {
	case 'f':
	force = B_TRUE;
	break;
	default:
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* get vdev name */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing vdev name\n"));
	usage(B_FALSE);
	}
	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	(void) strlcpy(vdev, argv[0], sizeof (vdev));

	/*
	* If we cannot open an absolute path, we quit.
	* Otherwise if the provided vdev name doesn't point to a file,
	* try prepending expected disk paths and partition numbers.
	*/
	if ((fd = open(vdev, O_RDWR)) < 0) {
	int error;
	if (vdev[0] == '/') {
	(void) fprintf(stderr, gettext("failed to open "
	"%s: %s\n"), vdev, strerror(errno));
	return (1);
	}

	error = zfs_resolve_shortname(argv[0], vdev, MAXPATHLEN);
	if (error == 0 && zfs_dev_is_whole_disk(vdev)) {
	if (zfs_append_partition(vdev, MAXPATHLEN) == -1)
	error = ENOENT;
	}

	if (error \|\| ((fd = open(vdev, O_RDWR)) < 0)) {
	if (errno == ENOENT) {
	(void) fprintf(stderr, gettext(
	"failed to find device %s, try "
	"specifying absolute path instead\n"),
	argv[0]);
	return (1);
	}

	(void) fprintf(stderr, gettext("failed to open %s:"
	" %s\n"), vdev, strerror(errno));
	return (1);
	}
	}

	/*
	* Flush all dirty pages for the block device. This should not be
	* fatal when the device does not support BLKFLSBUF as would be the
	* case for a file vdev.
	*/
	if ((zfs_dev_flush(fd) != 0) && (errno != ENOTTY))
	(void) fprintf(stderr, gettext("failed to invalidate "
	"cache for %s: %s\n"), vdev, strerror(errno));

	if (zpool_read_label(fd, &config, NULL) != 0) {
	(void) fprintf(stderr,
	gettext("failed to read label from %s\n"), vdev);
	ret = 1;
	goto errout;
	}
	nvlist_free(config);

	ret = zpool_in_use(g_zfs, fd, &state, &name, &inuse);
	if (ret != 0) {
	(void) fprintf(stderr,
	gettext("failed to check state for %s\n"), vdev);
	ret = 1;
	goto errout;
	}

	if (!inuse)
	goto wipe_label;

	switch (state) {
	default:
	case POOL_STATE_ACTIVE:
	case POOL_STATE_SPARE:
	case POOL_STATE_L2CACHE:
	/*
	* We allow the user to call 'zpool offline -f'
	* on an offlined disk in an active pool. We can check if
	* the disk is online by calling vdev_is_active().
	*/
	if (force && !vdev_is_active(vdev))
	break;

	(void) fprintf(stderr, gettext(
	"%s is a member (%s) of pool \"%s\""),
	vdev, zpool_pool_state_to_name(state), name);

	if (force) {
	(void) fprintf(stderr, gettext(
	". Offline the disk first to clear its label."));
	}
	printf("\n");
	ret = 1;
	goto errout;

	case POOL_STATE_EXPORTED:
	if (force)
	break;
	(void) fprintf(stderr, gettext(
	"use '-f' to override the following error:\n"
	"%s is a member of exported pool \"%s\"\n"),
	vdev, name);
	ret = 1;
	goto errout;

	case POOL_STATE_POTENTIALLY_ACTIVE:
	if (force)
	break;
	(void) fprintf(stderr, gettext(
	"use '-f' to override the following error:\n"
	"%s is a member of potentially active pool \"%s\"\n"),
	vdev, name);
	ret = 1;
	goto errout;

	case POOL_STATE_DESTROYED:
	/* inuse should never be set for a destroyed pool */
	assert(0);
	break;
	}

	wipe_label:
	ret = zpool_clear_label(fd);
	if (ret != 0) {
	(void) fprintf(stderr,
	gettext("failed to clear label for %s\n"), vdev);
	}

	errout:
	free(name);
	(void) close(fd);

	return (ret);
	}

	/*
	* zpool create [-fnd] [-o property=value] ...
	* [-O file-system-property=value] ...
	* [-R root] [-m mountpoint] <pool> <dev> ...
	*
	* -f Force creation, even if devices appear in use
	* -n Do not create the pool, but display the resulting layout if it
	* were to be created.
	* -R Create a pool under an alternate root
	* -m Set default mountpoint for the root dataset. By default it's
	* '/<pool>'
	* -o Set property=value.
	* -o Set feature@feature=enabled\|disabled.
	* -d Don't automatically enable all supported pool features
	* (individual features can be enabled with -o).
	* -O Set fsproperty=value in the pool's root file system
	*
	* Creates the named pool according to the given vdev specification. The
	* bulk of the vdev processing is done in make_root_vdev() in zpool_vdev.c.
	* Once we get the nvlist back from make_root_vdev(), we either print out the
	* contents (if '-n' was specified), or pass it to libzfs to do the creation.
	*/
	int
	zpool_do_create(int argc, char **argv)
	{
	boolean_t force = B_FALSE;
	boolean_t dryrun = B_FALSE;
	boolean_t enable_pool_features = B_TRUE;

	int c;
	nvlist_t *nvroot = NULL;
	char *poolname;
	char *tname = NULL;
	int ret = 1;
	char *altroot = NULL;
	char *compat = NULL;
	char *mountpoint = NULL;
	nvlist_t *fsprops = NULL;
	nvlist_t *props = NULL;
	char *propval;

	/* check options */
	while ((c = getopt(argc, argv, ":fndR:m:o:O:t:")) != -1) {
	switch (c) {
	case 'f':
	force = B_TRUE;
	break;
	case 'n':
	dryrun = B_TRUE;
	break;
	case 'd':
	enable_pool_features = B_FALSE;
	break;
	case 'R':
	altroot = optarg;
	if (add_prop_list(zpool_prop_to_name(
	ZPOOL_PROP_ALTROOT), optarg, &props, B_TRUE))
	goto errout;
	if (add_prop_list_default(zpool_prop_to_name(
	ZPOOL_PROP_CACHEFILE), "none", &props))
	goto errout;
	break;
	case 'm':
	/* Equivalent to -O mountpoint=optarg */
	mountpoint = optarg;
	break;
	case 'o':
	if ((propval = strchr(optarg, '=')) == NULL) {
	(void) fprintf(stderr, gettext("missing "
	"'=' for -o option\n"));
	goto errout;
	}
	*propval = '\0';
	propval++;

	if (add_prop_list(optarg, propval, &props, B_TRUE))
	goto errout;

	/*
	* If the user is creating a pool that doesn't support
	* feature flags, don't enable any features.
	*/
	if (zpool_name_to_prop(optarg) == ZPOOL_PROP_VERSION) {
	char *end;
	u_longlong_t ver;

	ver = strtoull(propval, &end, 10);
	if (*end == '\0' &&
	ver < SPA_VERSION_FEATURES) {
	enable_pool_features = B_FALSE;
	}
	}
	if (zpool_name_to_prop(optarg) == ZPOOL_PROP_ALTROOT)
	altroot = propval;
	if (zpool_name_to_prop(optarg) ==
	ZPOOL_PROP_COMPATIBILITY)
	compat = propval;
	break;
	case 'O':
	if ((propval = strchr(optarg, '=')) == NULL) {
	(void) fprintf(stderr, gettext("missing "
	"'=' for -O option\n"));
	goto errout;
	}
	*propval = '\0';
	propval++;

	/*
	* Mountpoints are checked and then added later.
	* Uniquely among properties, they can be specified
	* more than once, to avoid conflict with -m.
	*/
	if (0 == strcmp(optarg,
	zfs_prop_to_name(ZFS_PROP_MOUNTPOINT))) {
	mountpoint = propval;
	} else if (add_prop_list(optarg, propval, &fsprops,
	B_FALSE)) {
	goto errout;
	}
	break;
	case 't':
	/*
	* Sanity check temporary pool name.
	*/
	if (strchr(optarg, '/') != NULL) {
	(void) fprintf(stderr, gettext("cannot create "
	"'%s': invalid character '/' in temporary "
	"name\n"), optarg);
	(void) fprintf(stderr, gettext("use 'zfs "
	"create' to create a dataset\n"));
	goto errout;
	}

	if (add_prop_list(zpool_prop_to_name(
	ZPOOL_PROP_TNAME), optarg, &props, B_TRUE))
	goto errout;
	if (add_prop_list_default(zpool_prop_to_name(
	ZPOOL_PROP_CACHEFILE), "none", &props))
	goto errout;
	tname = optarg;
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	goto badusage;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	goto badusage;
	}
	}

	argc -= optind;
	argv += optind;

	/* get pool name and check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name argument\n"));
	goto badusage;
	}
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing vdev specification\n"));
	goto badusage;
	}

	poolname = argv[0];

	/*
	* As a special case, check for use of '/' in the name, and direct the
	* user to use 'zfs create' instead.
	*/
	if (strchr(poolname, '/') != NULL) {
	(void) fprintf(stderr, gettext("cannot create '%s': invalid "
	"character '/' in pool name\n"), poolname);
	(void) fprintf(stderr, gettext("use 'zfs create' to "
	"create a dataset\n"));
	goto errout;
	}

	/* pass off to make_root_vdev for bulk processing */
	nvroot = make_root_vdev(NULL, props, force, !force, B_FALSE, dryrun,
	argc - 1, argv + 1);
	if (nvroot == NULL)
	goto errout;

	/* make_root_vdev() allows 0 toplevel children if there are spares */
	if (!zfs_allocatable_devs(nvroot)) {
	(void) fprintf(stderr, gettext("invalid vdev "
	"specification: at least one toplevel vdev must be "
	"specified\n"));
	goto errout;
	}

	if (altroot != NULL && altroot[0] != '/') {
	(void) fprintf(stderr, gettext("invalid alternate root '%s': "
	"must be an absolute path\n"), altroot);
	goto errout;
	}

	/*
	* Check the validity of the mountpoint and direct the user to use the
	* '-m' mountpoint option if it looks like its in use.
	*/
	if (mountpoint == NULL \|\|
	(strcmp(mountpoint, ZFS_MOUNTPOINT_LEGACY) != 0 &&
	strcmp(mountpoint, ZFS_MOUNTPOINT_NONE) != 0)) {
	char buf[MAXPATHLEN];
	DIR *dirp;

	if (mountpoint && mountpoint[0] != '/') {
	(void) fprintf(stderr, gettext("invalid mountpoint "
	"'%s': must be an absolute path, 'legacy', or "
	"'none'\n"), mountpoint);
	goto errout;
	}

	if (mountpoint == NULL) {
	if (altroot != NULL)
	(void) snprintf(buf, sizeof (buf), "%s/%s",
	altroot, poolname);
	else
	(void) snprintf(buf, sizeof (buf), "/%s",
	poolname);
	} else {
	if (altroot != NULL)
	(void) snprintf(buf, sizeof (buf), "%s%s",
	altroot, mountpoint);
	else
	(void) snprintf(buf, sizeof (buf), "%s",
	mountpoint);
	}

	if ((dirp = opendir(buf)) == NULL && errno != ENOENT) {
	(void) fprintf(stderr, gettext("mountpoint '%s' : "
	"%s\n"), buf, strerror(errno));
	(void) fprintf(stderr, gettext("use '-m' "
	"option to provide a different default\n"));
	goto errout;
	} else if (dirp) {
	int count = 0;

	while (count < 3 && readdir(dirp) != NULL)
	count++;
	(void) closedir(dirp);

	if (count > 2) {
	(void) fprintf(stderr, gettext("mountpoint "
	"'%s' exists and is not empty\n"), buf);
	(void) fprintf(stderr, gettext("use '-m' "
	"option to provide a "
	"different default\n"));
	goto errout;
	}
	}
	}

	/*
	* Now that the mountpoint's validity has been checked, ensure that
	* the property is set appropriately prior to creating the pool.
	*/
	if (mountpoint != NULL) {
	ret = add_prop_list(zfs_prop_to_name(ZFS_PROP_MOUNTPOINT),
	mountpoint, &fsprops, B_FALSE);
	if (ret != 0)
	goto errout;
	}

	ret = 1;
	if (dryrun) {
	/*
	* For a dry run invocation, print out a basic message and run
	* through all the vdevs in the list and print out in an
	* appropriate hierarchy.
	*/
	(void) printf(gettext("would create '%s' with the "
	"following layout:\n\n"), poolname);

	print_vdev_tree(NULL, poolname, nvroot, 0, "", 0);
	print_vdev_tree(NULL, "dedup", nvroot, 0,
	VDEV_ALLOC_BIAS_DEDUP, 0);
	print_vdev_tree(NULL, "special", nvroot, 0,
	VDEV_ALLOC_BIAS_SPECIAL, 0);
	print_vdev_tree(NULL, "logs", nvroot, 0,
	VDEV_ALLOC_BIAS_LOG, 0);
	print_cache_list(nvroot, 0);
	print_spare_list(nvroot, 0);

	ret = 0;
	} else {
	/*
	* Load in feature set.
	* Note: if compatibility property not given, we'll have
	* NULL, which means 'all features'.
	*/
	boolean_t requested_features[SPA_FEATURES];
	if (zpool_do_load_compat(compat, requested_features) !=
	ZPOOL_COMPATIBILITY_OK)
	goto errout;

	/*
	* props contains list of features to enable.
	* For each feature:
	* - remove it if feature@name=disabled
	* - leave it there if feature@name=enabled
	* - add it if:
	* - enable_pool_features (ie: no '-d' or '-o version')
	* - it's supported by the kernel module
	* - it's in the requested feature set
	* - warn if it's enabled but not in compat
	*/
	for (spa_feature_t i = 0; i < SPA_FEATURES; i++) {
	char propname[MAXPATHLEN];
	const char *propval;
	zfeature_info_t *feat = &spa_feature_table[i];

	(void) snprintf(propname, sizeof (propname),
	"feature@%s", feat->fi_uname);

	if (!nvlist_lookup_string(props, propname, &propval)) {
	if (strcmp(propval,
	ZFS_FEATURE_DISABLED) == 0) {
	(void) nvlist_remove_all(props,
	propname);
	} else if (strcmp(propval,
	ZFS_FEATURE_ENABLED) == 0 &&
	!requested_features[i]) {
	(void) fprintf(stderr, gettext(
	"Warning: feature \"%s\" enabled "
	"but is not in specified "
	"'compatibility' feature set.\n"),
	feat->fi_uname);
	}
	} else if (
	enable_pool_features &&
	feat->fi_zfs_mod_supported &&
	requested_features[i]) {
	ret = add_prop_list(propname,
	ZFS_FEATURE_ENABLED, &props, B_TRUE);
	if (ret != 0)
	goto errout;
	}
	}

	ret = 1;
	if (zpool_create(g_zfs, poolname,
	nvroot, props, fsprops) == 0) {
	zfs_handle_t *pool = zfs_open(g_zfs,
	tname ? tname : poolname, ZFS_TYPE_FILESYSTEM);
	if (pool != NULL) {
	if (zfs_mount(pool, NULL, 0) == 0) {
	ret = zfs_share(pool, NULL);
	zfs_commit_shares(NULL);
	}
	zfs_close(pool);
	}
	} else if (libzfs_errno(g_zfs) == EZFS_INVALIDNAME) {
	(void) fprintf(stderr, gettext("pool name may have "
	"been omitted\n"));
	}
	}

	errout:
	nvlist_free(nvroot);
	nvlist_free(fsprops);
	nvlist_free(props);
	return (ret);
	badusage:
	nvlist_free(fsprops);
	nvlist_free(props);
	usage(B_FALSE);
	return (2);
	}

	/*
	* zpool destroy <pool>
	*
	* -f Forcefully unmount any datasets
	*
	* Destroy the given pool. Automatically unmounts any datasets in the pool.
	*/
	int
	zpool_do_destroy(int argc, char **argv)
	{
	boolean_t force = B_FALSE;
	int c;
	char *pool;
	zpool_handle_t *zhp;
	int ret;

	/* check options */
	while ((c = getopt(argc, argv, "f")) != -1) {
	switch (c) {
	case 'f':
	force = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* check arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool argument\n"));
	usage(B_FALSE);
	}
	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	pool = argv[0];

	if ((zhp = zpool_open_canfail(g_zfs, pool)) == NULL) {
	/*
	* As a special case, check for use of '/' in the name, and
	* direct the user to use 'zfs destroy' instead.
	*/
	if (strchr(pool, '/') != NULL)
	(void) fprintf(stderr, gettext("use 'zfs destroy' to "
	"destroy a dataset\n"));
	return (1);
	}

	if (zpool_disable_datasets(zhp, force) != 0) {
	(void) fprintf(stderr, gettext("could not destroy '%s': "
	"could not unmount datasets\n"), zpool_get_name(zhp));
	zpool_close(zhp);
	return (1);
	}

	/* The history must be logged as part of the export */
	log_history = B_FALSE;

	ret = (zpool_destroy(zhp, history_str) != 0);

	zpool_close(zhp);

	return (ret);
	}

	typedef struct export_cbdata {
	+ tpool_t *tpool;
	+ pthread_mutex_t mnttab_lock;
	boolean_t force;
	boolean_t hardforce;
	+ int retval;
	} export_cbdata_t;

	+
	+typedef struct {
	+ char *aea_poolname;
	+ export_cbdata_t *aea_cbdata;
	+} async_export_args_t;
	+
	/*
	* Export one pool
	*/
	static int
	zpool_export_one(zpool_handle_t zhp, void data)
	{
	export_cbdata_t *cb = data;

	- if (zpool_disable_datasets(zhp, cb->force) != 0)
	- return (1);
	+ /*
	+ * zpool_disable_datasets() is not thread-safe for mnttab access.
	+ * So we serialize access here for 'zpool export -a' parallel case.
	+ */
	+ if (cb->tpool != NULL)
	+ pthread_mutex_lock(&cb->mnttab_lock);

	- /* The history must be logged as part of the export */
	- log_history = B_FALSE;
	+ int retval = zpool_disable_datasets(zhp, cb->force);
	+
	+ if (cb->tpool != NULL)
	+ pthread_mutex_unlock(&cb->mnttab_lock);
	+
	+ if (retval)
	+ return (1);

	if (cb->hardforce) {
	if (zpool_export_force(zhp, history_str) != 0)
	return (1);
	} else if (zpool_export(zhp, cb->force, history_str) != 0) {
	return (1);
	}

	return (0);
	}

	+/*
	+ * Asynchronous export request
	+ */
	+static void
	+zpool_export_task(void *arg)
	+{
	+ async_export_args_t *aea = arg;
	+
	+ zpool_handle_t *zhp = zpool_open(g_zfs, aea->aea_poolname);
	+ if (zhp != NULL) {
	+ int ret = zpool_export_one(zhp, aea->aea_cbdata);
	+ if (ret != 0)
	+ aea->aea_cbdata->retval = ret;
	+ zpool_close(zhp);
	+ } else {
	+ aea->aea_cbdata->retval = 1;
	+ }
	+
	+ free(aea->aea_poolname);
	+ free(aea);
	+}
	+
	+/*
	+ * Process an export request in parallel
	+ */
	+static int
	+zpool_export_one_async(zpool_handle_t zhp, void data)
	+{
	+ tpool_t tpool = ((export_cbdata_t )data)->tpool;
	+ async_export_args_t *aea = safe_malloc(sizeof (async_export_args_t));
	+
	+ /* save pool name since zhp will go out of scope */
	+ aea->aea_poolname = strdup(zpool_get_name(zhp));
	+ aea->aea_cbdata = data;
	+
	+ /* ship off actual export to another thread */
	+ if (tpool_dispatch(tpool, zpool_export_task, (void *)aea) != 0)
	+ return (errno); /* unlikely */
	+ else
	+ return (0);
	+}
	+
	/*
	* zpool export [-f] <pool> ...
	*
	* -a Export all pools
	* -f Forcefully unmount datasets
	*
	* Export the given pools. By default, the command will attempt to cleanly
	* unmount any active datasets within the pool. If the '-f' flag is specified,
	* then the datasets will be forcefully unmounted.
	*/
	int
	zpool_do_export(int argc, char **argv)
	{
	export_cbdata_t cb;
	boolean_t do_all = B_FALSE;
	boolean_t force = B_FALSE;
	boolean_t hardforce = B_FALSE;
	int c, ret;

	/* check options */
	while ((c = getopt(argc, argv, "afF")) != -1) {
	switch (c) {
	case 'a':
	do_all = B_TRUE;
	break;
	case 'f':
	force = B_TRUE;
	break;
	case 'F':
	hardforce = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	cb.force = force;
	cb.hardforce = hardforce;
	+ cb.tpool = NULL;
	+ cb.retval = 0;
	argc -= optind;
	argv += optind;

	+ /* The history will be logged as part of the export itself */
	+ log_history = B_FALSE;
	+
	if (do_all) {
	if (argc != 0) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	- return (for_each_pool(argc, argv, B_TRUE, NULL,
	- ZFS_TYPE_POOL, B_FALSE, zpool_export_one, &cb));
	+ cb.tpool = tpool_create(1, 5 * sysconf(_SC_NPROCESSORS_ONLN),
	+ 0, NULL);
	+ pthread_mutex_init(&cb.mnttab_lock, NULL);
	+
	+ /* Asynchronously call zpool_export_one using thread pool */
	+ ret = for_each_pool(argc, argv, B_TRUE, NULL, ZFS_TYPE_POOL,
	+ B_FALSE, zpool_export_one_async, &cb);
	+
	+ tpool_wait(cb.tpool);
	+ tpool_destroy(cb.tpool);
	+ (void) pthread_mutex_destroy(&cb.mnttab_lock);
	+
	+ return (ret \| cb.retval);
	}

	/* check arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool argument\n"));
	usage(B_FALSE);
	}

	ret = for_each_pool(argc, argv, B_TRUE, NULL, ZFS_TYPE_POOL,
	B_FALSE, zpool_export_one, &cb);

	return (ret);
	}

	/*
	* Given a vdev configuration, determine the maximum width needed for the device
	* name column.
	*/
	static int
	max_width(zpool_handle_t zhp, nvlist_t nv, int depth, int max,
	int name_flags)
	{
	static const char *const subtypes[] =
	{ZPOOL_CONFIG_SPARES, ZPOOL_CONFIG_L2CACHE, ZPOOL_CONFIG_CHILDREN};

	char *name = zpool_vdev_name(g_zfs, zhp, nv, name_flags);
	max = MAX(strlen(name) + depth, max);
	free(name);

	nvlist_t **child;
	uint_t children;
	for (size_t i = 0; i < ARRAY_SIZE(subtypes); ++i)
	if (nvlist_lookup_nvlist_array(nv, subtypes[i],
	&child, &children) == 0)
	for (uint_t c = 0; c < children; ++c)
	max = MAX(max_width(zhp, child[c], depth + 2,
	max, name_flags), max);

	return (max);
	}

	typedef struct spare_cbdata {
	uint64_t cb_guid;
	zpool_handle_t *cb_zhp;
	} spare_cbdata_t;

	static boolean_t
	find_vdev(nvlist_t *nv, uint64_t search)
	{
	uint64_t guid;
	nvlist_t **child;
	uint_t c, children;

	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0 &&
	search == guid)
	return (B_TRUE);

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	&child, &children) == 0) {
	for (c = 0; c < children; c++)
	if (find_vdev(child[c], search))
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	static int
	find_spare(zpool_handle_t zhp, void data)
	{
	spare_cbdata_t *cbp = data;
	nvlist_t config, nvroot;

	config = zpool_get_config(zhp, NULL);
	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
	&nvroot) == 0);

	if (find_vdev(nvroot, cbp->cb_guid)) {
	cbp->cb_zhp = zhp;
	return (1);
	}

	zpool_close(zhp);
	return (0);
	}

	typedef struct status_cbdata {
	int cb_count;
	int cb_name_flags;
	int cb_namewidth;
	boolean_t cb_allpools;
	boolean_t cb_verbose;
	boolean_t cb_literal;
	boolean_t cb_explain;
	boolean_t cb_first;
	boolean_t cb_dedup_stats;
	boolean_t cb_print_unhealthy;
	boolean_t cb_print_status;
	boolean_t cb_print_slow_ios;
	boolean_t cb_print_vdev_init;
	boolean_t cb_print_vdev_trim;
	vdev_cmd_data_list_t *vcdl;
	boolean_t cb_print_power;
	} status_cbdata_t;

	/* Return 1 if string is NULL, empty, or whitespace; return 0 otherwise. */
	static boolean_t
	is_blank_str(const char *str)
	{
	for (; str != NULL && *str != '\0'; ++str)
	if (!isblank(*str))
	return (B_FALSE);
	return (B_TRUE);
	}

	/* Print command output lines for specific vdev in a specific pool */
	static void
	zpool_print_cmd(vdev_cmd_data_list_t vcdl, const char pool, const char *path)
	{
	vdev_cmd_data_t *data;
	int i, j;
	const char *val;

	for (i = 0; i < vcdl->count; i++) {
	if ((strcmp(vcdl->data[i].path, path) != 0) \|\|
	(strcmp(vcdl->data[i].pool, pool) != 0)) {
	/* Not the vdev we're looking for */
	continue;
	}

	data = &vcdl->data[i];
	/* Print out all the output values for this vdev */
	for (j = 0; j < vcdl->uniq_cols_cnt; j++) {
	val = NULL;
	/* Does this vdev have values for this column? */
	for (int k = 0; k < data->cols_cnt; k++) {
	if (strcmp(data->cols[k],
	vcdl->uniq_cols[j]) == 0) {
	/* yes it does, record the value */
	val = data->lines[k];
	break;
	}
	}
	/*
	* Mark empty values with dashes to make output
	* awk-able.
	*/
	if (val == NULL \|\| is_blank_str(val))
	val = "-";

	printf("%*s", vcdl->uniq_cols_width[j], val);
	if (j < vcdl->uniq_cols_cnt - 1)
	fputs(" ", stdout);
	}

	/* Print out any values that aren't in a column at the end */
	for (j = data->cols_cnt; j < data->lines_cnt; j++) {
	/* Did we have any columns? If so print a spacer. */
	if (vcdl->uniq_cols_cnt > 0)
	fputs(" ", stdout);

	val = data->lines[j];
	fputs(val ?: "", stdout);
	}
	break;
	}
	}

	/*
	* Print vdev initialization status for leaves
	*/
	static void
	print_status_initialize(vdev_stat_t *vs, boolean_t verbose)
	{
	if (verbose) {
	if ((vs->vs_initialize_state == VDEV_INITIALIZE_ACTIVE \|\|
	vs->vs_initialize_state == VDEV_INITIALIZE_SUSPENDED \|\|
	vs->vs_initialize_state == VDEV_INITIALIZE_COMPLETE) &&
	!vs->vs_scan_removing) {
	char zbuf[1024];
	char tbuf[256];

	time_t t = vs->vs_initialize_action_time;
	int initialize_pct = 100;
	if (vs->vs_initialize_state !=
	VDEV_INITIALIZE_COMPLETE) {
	initialize_pct = (vs->vs_initialize_bytes_done *
	100 / (vs->vs_initialize_bytes_est + 1));
	}

	(void) ctime_r(&t, tbuf);
	tbuf[24] = 0;

	switch (vs->vs_initialize_state) {
	case VDEV_INITIALIZE_SUSPENDED:
	(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
	gettext("suspended, started at"), tbuf);
	break;
	case VDEV_INITIALIZE_ACTIVE:
	(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
	gettext("started at"), tbuf);
	break;
	case VDEV_INITIALIZE_COMPLETE:
	(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
	gettext("completed at"), tbuf);
	break;
	}

	(void) printf(gettext(" (%d%% initialized%s)"),
	initialize_pct, zbuf);
	} else {
	(void) printf(gettext(" (uninitialized)"));
	}
	} else if (vs->vs_initialize_state == VDEV_INITIALIZE_ACTIVE) {
	(void) printf(gettext(" (initializing)"));
	}
	}

	/*
	* Print vdev TRIM status for leaves
	*/
	static void
	print_status_trim(vdev_stat_t *vs, boolean_t verbose)
	{
	if (verbose) {
	if ((vs->vs_trim_state == VDEV_TRIM_ACTIVE \|\|
	vs->vs_trim_state == VDEV_TRIM_SUSPENDED \|\|
	vs->vs_trim_state == VDEV_TRIM_COMPLETE) &&
	!vs->vs_scan_removing) {
	char zbuf[1024];
	char tbuf[256];

	time_t t = vs->vs_trim_action_time;
	int trim_pct = 100;
	if (vs->vs_trim_state != VDEV_TRIM_COMPLETE) {
	trim_pct = (vs->vs_trim_bytes_done *
	100 / (vs->vs_trim_bytes_est + 1));
	}

	(void) ctime_r(&t, tbuf);
	tbuf[24] = 0;

	switch (vs->vs_trim_state) {
	case VDEV_TRIM_SUSPENDED:
	(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
	gettext("suspended, started at"), tbuf);
	break;
	case VDEV_TRIM_ACTIVE:
	(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
	gettext("started at"), tbuf);
	break;
	case VDEV_TRIM_COMPLETE:
	(void) snprintf(zbuf, sizeof (zbuf), ", %s %s",
	gettext("completed at"), tbuf);
	break;
	}

	(void) printf(gettext(" (%d%% trimmed%s)"),
	trim_pct, zbuf);
	} else if (vs->vs_trim_notsup) {
	(void) printf(gettext(" (trim unsupported)"));
	} else {
	(void) printf(gettext(" (untrimmed)"));
	}
	} else if (vs->vs_trim_state == VDEV_TRIM_ACTIVE) {
	(void) printf(gettext(" (trimming)"));
	}
	}

	/*
	* Return the color associated with a health string. This includes returning
	* NULL for no color change.
	*/
	static const char *
	health_str_to_color(const char *health)
	{
	if (strcmp(health, gettext("FAULTED")) == 0 \|\|
	strcmp(health, gettext("SUSPENDED")) == 0 \|\|
	strcmp(health, gettext("UNAVAIL")) == 0) {
	return (ANSI_RED);
	}

	if (strcmp(health, gettext("OFFLINE")) == 0 \|\|
	strcmp(health, gettext("DEGRADED")) == 0 \|\|
	strcmp(health, gettext("REMOVED")) == 0) {
	return (ANSI_YELLOW);
	}

	return (NULL);
	}

	/*
	* Called for each leaf vdev. Returns 0 if the vdev is healthy.
	* A vdev is unhealthy if any of the following are true:
	* 1) there are read, write, or checksum errors,
	* 2) its state is not ONLINE, or
	* 3) slow IO reporting was requested (-s) and there are slow IOs.
	*/
	static int
	vdev_health_check_cb(void hdl_data, nvlist_t nv, void *data)
	{
	status_cbdata_t *cb = data;
	vdev_stat_t *vs;
	uint_t vsc;
	(void) hdl_data;

	if (nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
	(uint64_t **)&vs, &vsc) != 0)
	return (1);

	if (vs->vs_checksum_errors \|\| vs->vs_read_errors \|\|
	vs->vs_write_errors \|\| vs->vs_state != VDEV_STATE_HEALTHY)
	return (1);

	if (cb->cb_print_slow_ios && vs->vs_slow_ios)
	return (1);

	return (0);
	}

	/*
	* Print out configuration state as requested by status_callback.
	*/
	static void
	print_status_config(zpool_handle_t zhp, status_cbdata_t cb, const char *name,
	nvlist_t nv, int depth, boolean_t isspare, vdev_rebuild_stat_t vrs)
	{
	nvlist_t *child, root;
	uint_t c, i, vsc, children;
	pool_scan_stat_t *ps = NULL;
	vdev_stat_t *vs;
	char rbuf[6], wbuf[6], cbuf[6];
	char *vname;
	uint64_t notpresent;
	spare_cbdata_t spare_cb;
	const char *state;
	const char *type;
	const char *path = NULL;
	const char rcolor = NULL, wcolor = NULL, *ccolor = NULL,
	*scolor = NULL;

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	&child, &children) != 0)
	children = 0;

	verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
	(uint64_t **)&vs, &vsc) == 0);

	verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);

	if (strcmp(type, VDEV_TYPE_INDIRECT) == 0)
	return;

	state = zpool_state_to_name(vs->vs_state, vs->vs_aux);

	if (isspare) {
	/*
	* For hot spares, we use the terms 'INUSE' and 'AVAILABLE' for
	* online drives.
	*/
	if (vs->vs_aux == VDEV_AUX_SPARED)
	state = gettext("INUSE");
	else if (vs->vs_state == VDEV_STATE_HEALTHY)
	state = gettext("AVAIL");
	}

	/*
	* If '-e' is specified then top-level vdevs and their children
	* can be pruned if all of their leaves are healthy.
	*/
	if (cb->cb_print_unhealthy && depth > 0 &&
	for_each_vdev_in_nvlist(nv, vdev_health_check_cb, cb) == 0) {
	return;
	}

	printf_color(health_str_to_color(state),
	"\t%s%-s %-8s", depth, "", cb->cb_namewidth - depth,
	name, state);

	if (!isspare) {
	if (vs->vs_read_errors)
	rcolor = ANSI_RED;

	if (vs->vs_write_errors)
	wcolor = ANSI_RED;

	if (vs->vs_checksum_errors)
	ccolor = ANSI_RED;

	if (vs->vs_slow_ios)
	scolor = ANSI_BLUE;

	if (cb->cb_literal) {
	fputc(' ', stdout);
	printf_color(rcolor, "%5llu",
	(u_longlong_t)vs->vs_read_errors);
	fputc(' ', stdout);
	printf_color(wcolor, "%5llu",
	(u_longlong_t)vs->vs_write_errors);
	fputc(' ', stdout);
	printf_color(ccolor, "%5llu",
	(u_longlong_t)vs->vs_checksum_errors);
	} else {
	zfs_nicenum(vs->vs_read_errors, rbuf, sizeof (rbuf));
	zfs_nicenum(vs->vs_write_errors, wbuf, sizeof (wbuf));
	zfs_nicenum(vs->vs_checksum_errors, cbuf,
	sizeof (cbuf));
	fputc(' ', stdout);
	printf_color(rcolor, "%5s", rbuf);
	fputc(' ', stdout);
	printf_color(wcolor, "%5s", wbuf);
	fputc(' ', stdout);
	printf_color(ccolor, "%5s", cbuf);
	}
	if (cb->cb_print_slow_ios) {
	if (children == 0) {
	/* Only leafs vdevs have slow IOs */
	zfs_nicenum(vs->vs_slow_ios, rbuf,
	sizeof (rbuf));
	} else {
	snprintf(rbuf, sizeof (rbuf), "-");
	}

	if (cb->cb_literal)
	printf_color(scolor, " %5llu",
	(u_longlong_t)vs->vs_slow_ios);
	else
	printf_color(scolor, " %5s", rbuf);
	}
	if (cb->cb_print_power) {
	if (children == 0) {
	/* Only leaf vdevs have physical slots */
	switch (zpool_power_current_state(zhp, (char *)
	fnvlist_lookup_string(nv,
	ZPOOL_CONFIG_PATH))) {
	case 0:
	printf_color(ANSI_RED, " %5s",
	gettext("off"));
	break;
	case 1:
	printf(" %5s", gettext("on"));
	break;
	default:
	printf(" %5s", "-");
	}
	} else {
	printf(" %5s", "-");
	}
	}
	}

	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
	&notpresent) == 0) {
	verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0);
	(void) printf(" %s %s", gettext("was"), path);
	} else if (vs->vs_aux != 0) {
	(void) printf(" ");
	color_start(ANSI_RED);
	switch (vs->vs_aux) {
	case VDEV_AUX_OPEN_FAILED:
	(void) printf(gettext("cannot open"));
	break;

	case VDEV_AUX_BAD_GUID_SUM:
	(void) printf(gettext("missing device"));
	break;

	case VDEV_AUX_NO_REPLICAS:
	(void) printf(gettext("insufficient replicas"));
	break;

	case VDEV_AUX_VERSION_NEWER:
	(void) printf(gettext("newer version"));
	break;

	case VDEV_AUX_UNSUP_FEAT:
	(void) printf(gettext("unsupported feature(s)"));
	break;

	case VDEV_AUX_ASHIFT_TOO_BIG:
	(void) printf(gettext("unsupported minimum blocksize"));
	break;

	case VDEV_AUX_SPARED:
	verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
	&spare_cb.cb_guid) == 0);
	if (zpool_iter(g_zfs, find_spare, &spare_cb) == 1) {
	if (strcmp(zpool_get_name(spare_cb.cb_zhp),
	zpool_get_name(zhp)) == 0)
	(void) printf(gettext("currently in "
	"use"));
	else
	(void) printf(gettext("in use by "
	"pool '%s'"),
	zpool_get_name(spare_cb.cb_zhp));
	zpool_close(spare_cb.cb_zhp);
	} else {
	(void) printf(gettext("currently in use"));
	}
	break;

	case VDEV_AUX_ERR_EXCEEDED:
	if (vs->vs_read_errors + vs->vs_write_errors +
	vs->vs_checksum_errors == 0 && children == 0 &&
	vs->vs_slow_ios > 0) {
	(void) printf(gettext("too many slow I/Os"));
	} else {
	(void) printf(gettext("too many errors"));
	}
	break;

	case VDEV_AUX_IO_FAILURE:
	(void) printf(gettext("experienced I/O failures"));
	break;

	case VDEV_AUX_BAD_LOG:
	(void) printf(gettext("bad intent log"));
	break;

	case VDEV_AUX_EXTERNAL:
	(void) printf(gettext("external device fault"));
	break;

	case VDEV_AUX_SPLIT_POOL:
	(void) printf(gettext("split into new pool"));
	break;

	case VDEV_AUX_ACTIVE:
	(void) printf(gettext("currently in use"));
	break;

	case VDEV_AUX_CHILDREN_OFFLINE:
	(void) printf(gettext("all children offline"));
	break;

	case VDEV_AUX_BAD_LABEL:
	(void) printf(gettext("invalid label"));
	break;

	default:
	(void) printf(gettext("corrupted data"));
	break;
	}
	color_end();
	} else if (children == 0 && !isspare &&
	getenv("ZPOOL_STATUS_NON_NATIVE_ASHIFT_IGNORE") == NULL &&
	VDEV_STAT_VALID(vs_physical_ashift, vsc) &&
	vs->vs_configured_ashift < vs->vs_physical_ashift) {
	(void) printf(
	gettext(" block size: %dB configured, %dB native"),
	1 << vs->vs_configured_ashift, 1 << vs->vs_physical_ashift);
	}

	if (vs->vs_scan_removing != 0) {
	(void) printf(gettext(" (removing)"));
	} else if (VDEV_STAT_VALID(vs_noalloc, vsc) && vs->vs_noalloc != 0) {
	(void) printf(gettext(" (non-allocating)"));
	}

	/* The root vdev has the scrub/resilver stats */
	root = fnvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
	ZPOOL_CONFIG_VDEV_TREE);
	(void) nvlist_lookup_uint64_array(root, ZPOOL_CONFIG_SCAN_STATS,
	(uint64_t **)&ps, &c);

	/*
	* If you force fault a drive that's resilvering, its scan stats can
	* get frozen in time, giving the false impression that it's
	* being resilvered. That's why we check the state to see if the vdev
	* is healthy before reporting "resilvering" or "repairing".
	*/
	if (ps != NULL && ps->pss_state == DSS_SCANNING && children == 0 &&
	vs->vs_state == VDEV_STATE_HEALTHY) {
	if (vs->vs_scan_processed != 0) {
	(void) printf(gettext(" (%s)"),
	(ps->pss_func == POOL_SCAN_RESILVER) ?
	"resilvering" : "repairing");
	} else if (vs->vs_resilver_deferred) {
	(void) printf(gettext(" (awaiting resilver)"));
	}
	}

	/* The top-level vdevs have the rebuild stats */
	if (vrs != NULL && vrs->vrs_state == VDEV_REBUILD_ACTIVE &&
	children == 0 && vs->vs_state == VDEV_STATE_HEALTHY) {
	if (vs->vs_rebuild_processed != 0) {
	(void) printf(gettext(" (resilvering)"));
	}
	}

	if (cb->vcdl != NULL) {
	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
	printf(" ");
	zpool_print_cmd(cb->vcdl, zpool_get_name(zhp), path);
	}
	}

	/* Display vdev initialization and trim status for leaves. */
	if (children == 0) {
	print_status_initialize(vs, cb->cb_print_vdev_init);
	print_status_trim(vs, cb->cb_print_vdev_trim);
	}

	(void) printf("\n");

	for (c = 0; c < children; c++) {
	uint64_t islog = B_FALSE, ishole = B_FALSE;

	/* Don't print logs or holes here */
	(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
	&islog);
	(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
	&ishole);
	if (islog \|\| ishole)
	continue;
	/* Only print normal classes here */
	if (nvlist_exists(child[c], ZPOOL_CONFIG_ALLOCATION_BIAS))
	continue;

	/* Provide vdev_rebuild_stats to children if available */
	if (vrs == NULL) {
	(void) nvlist_lookup_uint64_array(nv,
	ZPOOL_CONFIG_REBUILD_STATS,
	(uint64_t **)&vrs, &i);
	}

	vname = zpool_vdev_name(g_zfs, zhp, child[c],
	cb->cb_name_flags \| VDEV_NAME_TYPE_ID);
	print_status_config(zhp, cb, vname, child[c], depth + 2,
	isspare, vrs);
	free(vname);
	}
	}

	/*
	* Print the configuration of an exported pool. Iterate over all vdevs in the
	* pool, printing out the name and status for each one.
	*/
	static void
	print_import_config(status_cbdata_t cb, const char name, nvlist_t *nv,
	int depth)
	{
	nvlist_t **child;
	uint_t c, children;
	vdev_stat_t *vs;
	const char *type;
	char *vname;

	verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);
	if (strcmp(type, VDEV_TYPE_MISSING) == 0 \|\|
	strcmp(type, VDEV_TYPE_HOLE) == 0)
	return;

	verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
	(uint64_t **)&vs, &c) == 0);

	(void) printf("\t%s%-s", depth, "", cb->cb_namewidth - depth, name);
	(void) printf(" %s", zpool_state_to_name(vs->vs_state, vs->vs_aux));

	if (vs->vs_aux != 0) {
	(void) printf(" ");

	switch (vs->vs_aux) {
	case VDEV_AUX_OPEN_FAILED:
	(void) printf(gettext("cannot open"));
	break;

	case VDEV_AUX_BAD_GUID_SUM:
	(void) printf(gettext("missing device"));
	break;

	case VDEV_AUX_NO_REPLICAS:
	(void) printf(gettext("insufficient replicas"));
	break;

	case VDEV_AUX_VERSION_NEWER:
	(void) printf(gettext("newer version"));
	break;

	case VDEV_AUX_UNSUP_FEAT:
	(void) printf(gettext("unsupported feature(s)"));
	break;

	case VDEV_AUX_ERR_EXCEEDED:
	(void) printf(gettext("too many errors"));
	break;

	case VDEV_AUX_ACTIVE:
	(void) printf(gettext("currently in use"));
	break;

	case VDEV_AUX_CHILDREN_OFFLINE:
	(void) printf(gettext("all children offline"));
	break;

	case VDEV_AUX_BAD_LABEL:
	(void) printf(gettext("invalid label"));
	break;

	default:
	(void) printf(gettext("corrupted data"));
	break;
	}
	}
	(void) printf("\n");

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	&child, &children) != 0)
	return;

	for (c = 0; c < children; c++) {
	uint64_t is_log = B_FALSE;

	(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
	&is_log);
	if (is_log)
	continue;
	if (nvlist_exists(child[c], ZPOOL_CONFIG_ALLOCATION_BIAS))
	continue;

	vname = zpool_vdev_name(g_zfs, NULL, child[c],
	cb->cb_name_flags \| VDEV_NAME_TYPE_ID);
	print_import_config(cb, vname, child[c], depth + 2);
	free(vname);
	}

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
	&child, &children) == 0) {
	(void) printf(gettext("\tcache\n"));
	for (c = 0; c < children; c++) {
	vname = zpool_vdev_name(g_zfs, NULL, child[c],
	cb->cb_name_flags);
	(void) printf("\t %s\n", vname);
	free(vname);
	}
	}

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
	&child, &children) == 0) {
	(void) printf(gettext("\tspares\n"));
	for (c = 0; c < children; c++) {
	vname = zpool_vdev_name(g_zfs, NULL, child[c],
	cb->cb_name_flags);
	(void) printf("\t %s\n", vname);
	free(vname);
	}
	}
	}

	/*
	* Print specialized class vdevs.
	*
	* These are recorded as top level vdevs in the main pool child array
	* but with "is_log" set to 1 or an "alloc_bias" string. We use either
	* print_status_config() or print_import_config() to print the top level
	* class vdevs then any of their children (eg mirrored slogs) are printed
	* recursively - which works because only the top level vdev is marked.
	*/
	static void
	print_class_vdevs(zpool_handle_t zhp, status_cbdata_t cb, nvlist_t *nv,
	const char *class)
	{
	uint_t c, children;
	nvlist_t **child;
	boolean_t printed = B_FALSE;

	assert(zhp != NULL \|\| !cb->cb_verbose);

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, &child,
	&children) != 0)
	return;

	for (c = 0; c < children; c++) {
	uint64_t is_log = B_FALSE;
	const char *bias = NULL;
	const char *type = NULL;

	(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
	&is_log);

	if (is_log) {
	bias = (char *)VDEV_ALLOC_CLASS_LOGS;
	} else {
	(void) nvlist_lookup_string(child[c],
	ZPOOL_CONFIG_ALLOCATION_BIAS, &bias);
	(void) nvlist_lookup_string(child[c],
	ZPOOL_CONFIG_TYPE, &type);
	}

	if (bias == NULL \|\| strcmp(bias, class) != 0)
	continue;
	if (!is_log && strcmp(type, VDEV_TYPE_INDIRECT) == 0)
	continue;

	if (!printed) {
	(void) printf("\t%s\t\n", gettext(class));
	printed = B_TRUE;
	}

	char *name = zpool_vdev_name(g_zfs, zhp, child[c],
	cb->cb_name_flags \| VDEV_NAME_TYPE_ID);
	if (cb->cb_print_status)
	print_status_config(zhp, cb, name, child[c], 2,
	B_FALSE, NULL);
	else
	print_import_config(cb, name, child[c], 2);
	free(name);
	}
	}

	/*
	* Display the status for the given pool.
	*/
	static int
	show_import(nvlist_t *config, boolean_t report_error)
	{
	uint64_t pool_state;
	vdev_stat_t *vs;
	const char *name;
	uint64_t guid;
	uint64_t hostid = 0;
	const char *msgid;
	const char *hostname = "unknown";
	nvlist_t nvroot, nvinfo;
	zpool_status_t reason;
	zpool_errata_t errata;
	const char *health;
	uint_t vsc;
	const char *comment;
	+ const char *indent;
	status_cbdata_t cb = { 0 };

	verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
	&name) == 0);
	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
	&guid) == 0);
	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
	&pool_state) == 0);
	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
	&nvroot) == 0);

	verify(nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_VDEV_STATS,
	(uint64_t **)&vs, &vsc) == 0);
	health = zpool_state_to_name(vs->vs_state, vs->vs_aux);

	reason = zpool_import_status(config, &msgid, &errata);

	/*
	* If we're importing using a cachefile, then we won't report any
	* errors unless we are in the scan phase of the import.
	*/
	if (reason != ZPOOL_STATUS_OK && !report_error)
	return (reason);

	- (void) printf(gettext(" pool: %s\n"), name);
	- (void) printf(gettext(" id: %llu\n"), (u_longlong_t)guid);
	- (void) printf(gettext(" state: %s"), health);
	+ if (nvlist_lookup_string(config, ZPOOL_CONFIG_COMMENT, &comment) == 0) {
	+ indent = " ";
	+ } else {
	+ comment = NULL;
	+ indent = "";
	+ }
	+
	+ (void) printf(gettext("%s pool: %s\n"), indent, name);
	+ (void) printf(gettext("%s id: %llu\n"), indent, (u_longlong_t)guid);
	+ (void) printf(gettext("%s state: %s"), indent, health);
	if (pool_state == POOL_STATE_DESTROYED)
	(void) printf(gettext(" (DESTROYED)"));
	(void) printf("\n");

	+ if (reason != ZPOOL_STATUS_OK) {
	+ (void) printf("%s", indent);
	+ printf_color(ANSI_BOLD, gettext("status: "));
	+ }
	switch (reason) {
	case ZPOOL_STATUS_MISSING_DEV_R:
	case ZPOOL_STATUS_MISSING_DEV_NR:
	case ZPOOL_STATUS_BAD_GUID_SUM:
	- printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices are "
	"missing from the system.\n"));
	break;

	case ZPOOL_STATUS_CORRUPT_LABEL_R:
	case ZPOOL_STATUS_CORRUPT_LABEL_NR:
	- printf_color(ANSI_BOLD, gettext("status: "));
	- printf_color(ANSI_YELLOW, gettext("One or more devices contains"
	- " corrupted data.\n"));
	+ printf_color(ANSI_YELLOW, gettext("One or more devices "
	+ "contains corrupted data.\n"));
	break;

	case ZPOOL_STATUS_CORRUPT_DATA:
	- (void) printf(
	- gettext(" status: The pool data is corrupted.\n"));
	+ printf_color(ANSI_YELLOW, gettext("The pool data is "
	+ "corrupted.\n"));
	break;

	case ZPOOL_STATUS_OFFLINE_DEV:
	- printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices "
	"are offlined.\n"));
	break;

	case ZPOOL_STATUS_CORRUPT_POOL:
	- printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool metadata is "
	"corrupted.\n"));
	break;

	case ZPOOL_STATUS_VERSION_OLDER:
	- printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool is formatted using "
	"a legacy on-disk version.\n"));
	break;

	case ZPOOL_STATUS_VERSION_NEWER:
	- printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool is formatted using "
	"an incompatible version.\n"));
	break;

	case ZPOOL_STATUS_FEAT_DISABLED:
	- printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("Some supported "
	- "features are not enabled on the pool.\n\t"
	- "(Note that they may be intentionally disabled "
	- "if the\n\t'compatibility' property is set.)\n"));
	+ "features are not enabled on the pool.\n"
	+ "\t%s(Note that they may be intentionally disabled if the\n"
	+ "\t%s'compatibility' property is set.)\n"), indent, indent);
	break;

	case ZPOOL_STATUS_COMPATIBILITY_ERR:
	- printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("Error reading or parsing "
	"the file(s) indicated by the 'compatibility'\n"
	- "property.\n"));
	+ "\t%sproperty.\n"), indent);
	break;

	case ZPOOL_STATUS_INCOMPATIBLE_FEAT:
	- printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more features "
	"are enabled on the pool despite not being\n"
	- "requested by the 'compatibility' property.\n"));
	+ "\t%srequested by the 'compatibility' property.\n"),
	+ indent);
	break;

	case ZPOOL_STATUS_UNSUP_FEAT_READ:
	- printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool uses the following "
	"feature(s) not supported on this system:\n"));
	color_start(ANSI_YELLOW);
	zpool_print_unsup_feat(config);
	color_end();
	break;

	case ZPOOL_STATUS_UNSUP_FEAT_WRITE:
	- printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool can only be "
	- "accessed in read-only mode on this system. It\n\tcannot be"
	- " accessed in read-write mode because it uses the "
	- "following\n\tfeature(s) not supported on this system:\n"));
	+ "accessed in read-only mode on this system. It\n"
	+ "\t%scannot be accessed in read-write mode because it uses "
	+ "the following\n"
	+ "\t%sfeature(s) not supported on this system:\n"),
	+ indent, indent);
	color_start(ANSI_YELLOW);
	zpool_print_unsup_feat(config);
	color_end();
	break;

	case ZPOOL_STATUS_HOSTID_ACTIVE:
	- printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool is currently "
	"imported by another system.\n"));
	break;

	case ZPOOL_STATUS_HOSTID_REQUIRED:
	- printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool has the "
	- "multihost property on. It cannot\n\tbe safely imported "
	- "when the system hostid is not set.\n"));
	+ "multihost property on. It cannot\n"
	+ "\t%sbe safely imported when the system hostid is not "
	+ "set.\n"), indent);
	break;

	case ZPOOL_STATUS_HOSTID_MISMATCH:
	- printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool was last accessed "
	"by another system.\n"));
	break;

	case ZPOOL_STATUS_FAULTED_DEV_R:
	case ZPOOL_STATUS_FAULTED_DEV_NR:
	- printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices are "
	"faulted.\n"));
	break;

	case ZPOOL_STATUS_BAD_LOG:
	- printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("An intent log record cannot "
	"be read.\n"));
	break;

	case ZPOOL_STATUS_RESILVERING:
	case ZPOOL_STATUS_REBUILDING:
	- printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices were "
	"being resilvered.\n"));
	break;

	case ZPOOL_STATUS_ERRATA:
	- printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("Errata #%d detected.\n"),
	errata);
	break;

	case ZPOOL_STATUS_NON_NATIVE_ASHIFT:
	- printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices are "
	"configured to use a non-native block size.\n"
	- "\tExpect reduced performance.\n"));
	+ "\t%sExpect reduced performance.\n"), indent);
	break;

	default:
	/*
	* No other status can be seen when importing pools.
	*/
	assert(reason == ZPOOL_STATUS_OK);
	}

	/*
	* Print out an action according to the overall state of the pool.
	*/
	+ if (vs->vs_state != VDEV_STATE_HEALTHY \|\|
	+ reason != ZPOOL_STATUS_ERRATA \|\| errata != ZPOOL_ERRATA_NONE) {
	+ (void) printf("%s", indent);
	+ (void) printf(gettext("action: "));
	+ }
	if (vs->vs_state == VDEV_STATE_HEALTHY) {
	if (reason == ZPOOL_STATUS_VERSION_OLDER \|\|
	reason == ZPOOL_STATUS_FEAT_DISABLED) {
	- (void) printf(gettext(" action: The pool can be "
	- "imported using its name or numeric identifier, "
	- "though\n\tsome features will not be available "
	- "without an explicit 'zpool upgrade'.\n"));
	+ (void) printf(gettext("The pool can be imported using "
	+ "its name or numeric identifier, though\n"
	+ "\t%ssome features will not be available without "
	+ "an explicit 'zpool upgrade'.\n"), indent);
	} else if (reason == ZPOOL_STATUS_COMPATIBILITY_ERR) {
	- (void) printf(gettext(" action: The pool can be "
	- "imported using its name or numeric\n\tidentifier, "
	- "though the file(s) indicated by its "
	- "'compatibility'\n\tproperty cannot be parsed at "
	- "this time.\n"));
	+ (void) printf(gettext("The pool can be imported using "
	+ "its name or numeric\n"
	+ "\t%sidentifier, though the file(s) indicated by "
	+ "its 'compatibility'\n"
	+ "\t%sproperty cannot be parsed at this time.\n"),
	+ indent, indent);
	} else if (reason == ZPOOL_STATUS_HOSTID_MISMATCH) {
	- (void) printf(gettext(" action: The pool can be "
	- "imported using its name or numeric "
	- "identifier and\n\tthe '-f' flag.\n"));
	+ (void) printf(gettext("The pool can be imported using "
	+ "its name or numeric identifier and\n"
	+ "\t%sthe '-f' flag.\n"), indent);
	} else if (reason == ZPOOL_STATUS_ERRATA) {
	switch (errata) {
	- case ZPOOL_ERRATA_NONE:
	- break;
	-
	case ZPOOL_ERRATA_ZOL_2094_SCRUB:
	- (void) printf(gettext(" action: The pool can "
	- "be imported using its name or numeric "
	- "identifier,\n\thowever there is a compat"
	- "ibility issue which should be corrected"
	- "\n\tby running 'zpool scrub'\n"));
	+ (void) printf(gettext("The pool can be "
	+ "imported using its name or numeric "
	+ "identifier,\n"
	+ "\t%showever there is a compatibility "
	+ "issue which should be corrected\n"
	+ "\t%sby running 'zpool scrub'\n"),
	+ indent, indent);
	break;

	case ZPOOL_ERRATA_ZOL_2094_ASYNC_DESTROY:
	- (void) printf(gettext(" action: The pool can"
	- "not be imported with this version of ZFS "
	- "due to\n\tan active asynchronous destroy. "
	- "Revert to an earlier version\n\tand "
	- "allow the destroy to complete before "
	- "updating.\n"));
	+ (void) printf(gettext("The pool cannot be "
	+ "imported with this version of ZFS due to\n"
	+ "\t%san active asynchronous destroy. "
	+ "Revert to an earlier version\n"
	+ "\t%sand allow the destroy to complete "
	+ "before updating.\n"), indent, indent);
	break;

	case ZPOOL_ERRATA_ZOL_6845_ENCRYPTION:
	- (void) printf(gettext(" action: Existing "
	- "encrypted datasets contain an on-disk "
	- "incompatibility, which\n\tneeds to be "
	- "corrected. Backup these datasets to new "
	- "encrypted datasets\n\tand destroy the "
	- "old ones.\n"));
	+ (void) printf(gettext("Existing encrypted "
	+ "datasets contain an on-disk "
	+ "incompatibility, which\n"
	+ "\t%sneeds to be corrected. Backup these "
	+ "datasets to new encrypted datasets\n"
	+ "\t%sand destroy the old ones.\n"),
	+ indent, indent);
	break;

	case ZPOOL_ERRATA_ZOL_8308_ENCRYPTION:
	- (void) printf(gettext(" action: Existing "
	- "encrypted snapshots and bookmarks contain "
	- "an on-disk\n\tincompatibility. This may "
	- "cause on-disk corruption if they are used"
	- "\n\twith 'zfs recv'. To correct the "
	- "issue, enable the bookmark_v2 feature.\n\t"
	- "No additional action is needed if there "
	- "are no encrypted snapshots or\n\t"
	- "bookmarks. If preserving the encrypted "
	- "snapshots and bookmarks is\n\trequired, "
	- "use a non-raw send to backup and restore "
	- "them. Alternately,\n\tthey may be removed"
	- " to resolve the incompatibility.\n"));
	+ (void) printf(gettext("Existing encrypted "
	+ "snapshots and bookmarks contain an "
	+ "on-disk\n"
	+ "\t%sincompatibility. This may cause "
	+ "on-disk corruption if they are used\n"
	+ "\t%swith 'zfs recv'. To correct the "
	+ "issue, enable the bookmark_v2 feature.\n"
	+ "\t%sNo additional action is needed if "
	+ "there are no encrypted snapshots or\n"
	+ "\t%sbookmarks. If preserving the "
	+ "encrypted snapshots and bookmarks is\n"
	+ "\t%srequired, use a non-raw send to "
	+ "backup and restore them. Alternately,\n"
	+ "\t%sthey may be removed to resolve the "
	+ "incompatibility.\n"), indent, indent,
	+ indent, indent, indent, indent);
	break;
	default:
	/*
	* All errata must contain an action message.
	*/
	- assert(0);
	+ assert(errata == ZPOOL_ERRATA_NONE);
	}
	} else {
	- (void) printf(gettext(" action: The pool can be "
	- "imported using its name or numeric "
	- "identifier.\n"));
	+ (void) printf(gettext("The pool can be imported using "
	+ "its name or numeric identifier.\n"));
	}
	} else if (vs->vs_state == VDEV_STATE_DEGRADED) {
	- (void) printf(gettext(" action: The pool can be imported "
	- "despite missing or damaged devices. The\n\tfault "
	- "tolerance of the pool may be compromised if imported.\n"));
	+ (void) printf(gettext("The pool can be imported despite "
	+ "missing or damaged devices. The\n"
	+ "\t%sfault tolerance of the pool may be compromised if "
	+ "imported.\n"), indent);
	} else {
	switch (reason) {
	case ZPOOL_STATUS_VERSION_NEWER:
	- (void) printf(gettext(" action: The pool cannot be "
	- "imported. Access the pool on a system running "
	- "newer\n\tsoftware, or recreate the pool from "
	- "backup.\n"));
	+ (void) printf(gettext("The pool cannot be imported. "
	+ "Access the pool on a system running newer\n"
	+ "\t%ssoftware, or recreate the pool from "
	+ "backup.\n"), indent);
	break;
	case ZPOOL_STATUS_UNSUP_FEAT_READ:
	- printf_color(ANSI_BOLD, gettext("action: "));
	- printf_color(ANSI_YELLOW, gettext("The pool cannot be "
	- "imported. Access the pool on a system that "
	- "supports\n\tthe required feature(s), or recreate "
	- "the pool from backup.\n"));
	+ (void) printf(gettext("The pool cannot be imported. "
	+ "Access the pool on a system that supports\n"
	+ "\t%sthe required feature(s), or recreate the pool "
	+ "from backup.\n"), indent);
	break;
	case ZPOOL_STATUS_UNSUP_FEAT_WRITE:
	- printf_color(ANSI_BOLD, gettext("action: "));
	- printf_color(ANSI_YELLOW, gettext("The pool cannot be "
	- "imported in read-write mode. Import the pool "
	- "with\n"
	- "\t\"-o readonly=on\", access the pool on a system "
	- "that supports the\n\trequired feature(s), or "
	- "recreate the pool from backup.\n"));
	+ (void) printf(gettext("The pool cannot be imported in "
	+ "read-write mode. Import the pool with\n"
	+ "\t%s'-o readonly=on', access the pool on a system "
	+ "that supports the\n"
	+ "\t%srequired feature(s), or recreate the pool "
	+ "from backup.\n"), indent, indent);
	break;
	case ZPOOL_STATUS_MISSING_DEV_R:
	case ZPOOL_STATUS_MISSING_DEV_NR:
	case ZPOOL_STATUS_BAD_GUID_SUM:
	- (void) printf(gettext(" action: The pool cannot be "
	- "imported. Attach the missing\n\tdevices and try "
	- "again.\n"));
	+ (void) printf(gettext("The pool cannot be imported. "
	+ "Attach the missing\n"
	+ "\t%sdevices and try again.\n"), indent);
	break;
	case ZPOOL_STATUS_HOSTID_ACTIVE:
	VERIFY0(nvlist_lookup_nvlist(config,
	ZPOOL_CONFIG_LOAD_INFO, &nvinfo));

	if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_HOSTNAME))
	hostname = fnvlist_lookup_string(nvinfo,
	ZPOOL_CONFIG_MMP_HOSTNAME);

	if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_HOSTID))
	hostid = fnvlist_lookup_uint64(nvinfo,
	ZPOOL_CONFIG_MMP_HOSTID);

	- (void) printf(gettext(" action: The pool must be "
	- "exported from %s (hostid=%"PRIx64")\n\tbefore it "
	- "can be safely imported.\n"), hostname, hostid);
	+ (void) printf(gettext("The pool must be exported from "
	+ "%s (hostid=%"PRIx64")\n"
	+ "\t%sbefore it can be safely imported.\n"),
	+ hostname, hostid, indent);
	break;
	case ZPOOL_STATUS_HOSTID_REQUIRED:
	- (void) printf(gettext(" action: Set a unique system "
	- "hostid with the zgenhostid(8) command.\n"));
	+ (void) printf(gettext("Set a unique system hostid with "
	+ "the zgenhostid(8) command.\n"));
	break;
	default:
	- (void) printf(gettext(" action: The pool cannot be "
	- "imported due to damaged devices or data.\n"));
	+ (void) printf(gettext("The pool cannot be imported due "
	+ "to damaged devices or data.\n"));
	}
	}

	/* Print the comment attached to the pool. */
	- if (nvlist_lookup_string(config, ZPOOL_CONFIG_COMMENT, &comment) == 0)
	+ if (comment != NULL)
	(void) printf(gettext("comment: %s\n"), comment);

	/*
	* If the state is "closed" or "can't open", and the aux state
	* is "corrupt data":
	*/
	- if (((vs->vs_state == VDEV_STATE_CLOSED) \|\|
	- (vs->vs_state == VDEV_STATE_CANT_OPEN)) &&
	- (vs->vs_aux == VDEV_AUX_CORRUPT_DATA)) {
	+ if ((vs->vs_state == VDEV_STATE_CLOSED \|\|
	+ vs->vs_state == VDEV_STATE_CANT_OPEN) &&
	+ vs->vs_aux == VDEV_AUX_CORRUPT_DATA) {
	if (pool_state == POOL_STATE_DESTROYED)
	- (void) printf(gettext("\tThe pool was destroyed, "
	- "but can be imported using the '-Df' flags.\n"));
	+ (void) printf(gettext("\t%sThe pool was destroyed, "
	+ "but can be imported using the '-Df' flags.\n"),
	+ indent);
	else if (pool_state != POOL_STATE_EXPORTED)
	- (void) printf(gettext("\tThe pool may be active on "
	- "another system, but can be imported using\n\t"
	- "the '-f' flag.\n"));
	+ (void) printf(gettext("\t%sThe pool may be active on "
	+ "another system, but can be imported using\n"
	+ "\t%sthe '-f' flag.\n"), indent, indent);
	}

	if (msgid != NULL) {
	- (void) printf(gettext(
	- " see: https://openzfs.github.io/openzfs-docs/msg/%s\n"),
	- msgid);
	+ (void) printf(gettext("%s see: "
	+ "https://openzfs.github.io/openzfs-docs/msg/%s\n"),
	+ indent, msgid);
	}

	- (void) printf(gettext(" config:\n\n"));
	+ (void) printf(gettext("%sconfig:\n\n"), indent);

	cb.cb_namewidth = max_width(NULL, nvroot, 0, strlen(name),
	VDEV_NAME_TYPE_ID);
	if (cb.cb_namewidth < 10)
	cb.cb_namewidth = 10;

	print_import_config(&cb, name, nvroot, 0);

	print_class_vdevs(NULL, &cb, nvroot, VDEV_ALLOC_BIAS_DEDUP);
	print_class_vdevs(NULL, &cb, nvroot, VDEV_ALLOC_BIAS_SPECIAL);
	print_class_vdevs(NULL, &cb, nvroot, VDEV_ALLOC_CLASS_LOGS);

	if (reason == ZPOOL_STATUS_BAD_GUID_SUM) {
	- (void) printf(gettext("\n\tAdditional devices are known to "
	- "be part of this pool, though their\n\texact "
	- "configuration cannot be determined.\n"));
	+ (void) printf(gettext("\n\t%sAdditional devices are known to "
	+ "be part of this pool, though their\n"
	+ "\t%sexact configuration cannot be determined.\n"),
	+ indent, indent);
	}
	return (0);
	}

	static boolean_t
	zfs_force_import_required(nvlist_t *config)
	{
	uint64_t state;
	uint64_t hostid = 0;
	nvlist_t *nvinfo;

	state = fnvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE);
	nvinfo = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO);

	/*
	* The hostid on LOAD_INFO comes from the MOS label via
	* spa_tryimport(). If its not there then we're likely talking to an
	* older kernel, so use the top one, which will be from the label
	* discovered in zpool_find_import(), or if a cachefile is in use, the
	* local hostid.
	*/
	if (nvlist_lookup_uint64(nvinfo, ZPOOL_CONFIG_HOSTID, &hostid) != 0)
	(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_HOSTID,
	&hostid);

	if (state != POOL_STATE_EXPORTED && hostid != get_system_hostid())
	return (B_TRUE);

	if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_STATE)) {
	mmp_state_t mmp_state = fnvlist_lookup_uint64(nvinfo,
	ZPOOL_CONFIG_MMP_STATE);

	if (mmp_state != MMP_STATE_INACTIVE)
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	/*
	* Perform the import for the given configuration. This passes the heavy
	* lifting off to zpool_import_props(), and then mounts the datasets contained
	* within the pool.
	*/
	static int
	do_import(nvlist_t config, const char newname, const char *mntopts,
	- nvlist_t *props, int flags)
	+ nvlist_t *props, int flags, uint_t mntthreads)
	{
	int ret = 0;
	int ms_status = 0;
	zpool_handle_t *zhp;
	const char *name;
	uint64_t version;

	name = fnvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME);
	version = fnvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION);

	if (!SPA_VERSION_IS_SUPPORTED(version)) {
	(void) fprintf(stderr, gettext("cannot import '%s': pool "
	"is formatted using an unsupported ZFS version\n"), name);
	return (1);
	} else if (zfs_force_import_required(config) &&
	!(flags & ZFS_IMPORT_ANY_HOST)) {
	mmp_state_t mmp_state = MMP_STATE_INACTIVE;
	nvlist_t *nvinfo;

	nvinfo = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO);
	if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_STATE))
	mmp_state = fnvlist_lookup_uint64(nvinfo,
	ZPOOL_CONFIG_MMP_STATE);

	if (mmp_state == MMP_STATE_ACTIVE) {
	const char *hostname = "<unknown>";
	uint64_t hostid = 0;

	if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_HOSTNAME))
	hostname = fnvlist_lookup_string(nvinfo,
	ZPOOL_CONFIG_MMP_HOSTNAME);

	if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_HOSTID))
	hostid = fnvlist_lookup_uint64(nvinfo,
	ZPOOL_CONFIG_MMP_HOSTID);

	(void) fprintf(stderr, gettext("cannot import '%s': "
	"pool is imported on %s (hostid: "
	"0x%"PRIx64")\nExport the pool on the other "
	"system, then run 'zpool import'.\n"),
	name, hostname, hostid);
	} else if (mmp_state == MMP_STATE_NO_HOSTID) {
	(void) fprintf(stderr, gettext("Cannot import '%s': "
	"pool has the multihost property on and the\n"
	"system's hostid is not set. Set a unique hostid "
	"with the zgenhostid(8) command.\n"), name);
	} else {
	const char *hostname = "<unknown>";
	time_t timestamp = 0;
	uint64_t hostid = 0;

	if (nvlist_exists(nvinfo, ZPOOL_CONFIG_HOSTNAME))
	hostname = fnvlist_lookup_string(nvinfo,
	ZPOOL_CONFIG_HOSTNAME);
	else if (nvlist_exists(config, ZPOOL_CONFIG_HOSTNAME))
	hostname = fnvlist_lookup_string(config,
	ZPOOL_CONFIG_HOSTNAME);

	if (nvlist_exists(config, ZPOOL_CONFIG_TIMESTAMP))
	timestamp = fnvlist_lookup_uint64(config,
	ZPOOL_CONFIG_TIMESTAMP);

	if (nvlist_exists(nvinfo, ZPOOL_CONFIG_HOSTID))
	hostid = fnvlist_lookup_uint64(nvinfo,
	ZPOOL_CONFIG_HOSTID);
	else if (nvlist_exists(config, ZPOOL_CONFIG_HOSTID))
	hostid = fnvlist_lookup_uint64(config,
	ZPOOL_CONFIG_HOSTID);

	(void) fprintf(stderr, gettext("cannot import '%s': "
	"pool was previously in use from another system.\n"
	"Last accessed by %s (hostid=%"PRIx64") at %s"
	"The pool can be imported, use 'zpool import -f' "
	"to import the pool.\n"), name, hostname,
	hostid, ctime(&timestamp));
	}

	return (1);
	}

	if (zpool_import_props(g_zfs, config, newname, props, flags) != 0)
	return (1);

	if (newname != NULL)
	name = newname;

	if ((zhp = zpool_open_canfail(g_zfs, name)) == NULL)
	return (1);

	/*
	* Loading keys is best effort. We don't want to return immediately
	* if it fails but we do want to give the error to the caller.
	*/
	if (flags & ZFS_IMPORT_LOAD_KEYS &&
	zfs_crypto_attempt_load_keys(g_zfs, name) != 0)
	ret = 1;

	if (zpool_get_state(zhp) != POOL_STATE_UNAVAIL &&
	!(flags & ZFS_IMPORT_ONLY)) {
	- ms_status = zpool_enable_datasets(zhp, mntopts, 0);
	+ ms_status = zpool_enable_datasets(zhp, mntopts, 0, mntthreads);
	if (ms_status == EZFS_SHAREFAILED) {
	(void) fprintf(stderr, gettext("Import was "
	"successful, but unable to share some datasets\n"));
	} else if (ms_status == EZFS_MOUNTFAILED) {
	(void) fprintf(stderr, gettext("Import was "
	"successful, but unable to mount some datasets\n"));
	}
	}

	zpool_close(zhp);
	return (ret);
	}

	typedef struct import_parameters {
	nvlist_t *ip_config;
	const char *ip_mntopts;
	nvlist_t *ip_props;
	int ip_flags;
	+ uint_t ip_mntthreads;
	int *ip_err;
	} import_parameters_t;

	static void
	do_import_task(void *arg)
	{
	import_parameters_t *ip = arg;
	*ip->ip_err \|= do_import(ip->ip_config, NULL, ip->ip_mntopts,
	- ip->ip_props, ip->ip_flags);
	+ ip->ip_props, ip->ip_flags, ip->ip_mntthreads);
	free(ip);
	}


	static int
	import_pools(nvlist_t pools, nvlist_t props, char *mntopts, int flags,
	char orig_name, char new_name, importargs_t *import)
	{
	nvlist_t *config = NULL;
	nvlist_t *found_config = NULL;
	uint64_t pool_state;
	boolean_t pool_specified = (import->poolname != NULL \|\|
	import->guid != 0);
	+ uint_t npools = 0;


	tpool_t *tp = NULL;
	if (import->do_all) {
	tp = tpool_create(1, 5 * sysconf(_SC_NPROCESSORS_ONLN),
	0, NULL);
	}

	/*
	* At this point we have a list of import candidate configs. Even if
	* we were searching by pool name or guid, we still need to
	* post-process the list to deal with pool state and possible
	* duplicate names.
	*/
	int err = 0;
	nvpair_t *elem = NULL;
	boolean_t first = B_TRUE;
	+ if (!pool_specified && import->do_all) {
	+ while ((elem = nvlist_next_nvpair(pools, elem)) != NULL)
	+ npools++;
	+ }
	while ((elem = nvlist_next_nvpair(pools, elem)) != NULL) {

	verify(nvpair_value_nvlist(elem, &config) == 0);

	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
	&pool_state) == 0);
	if (!import->do_destroyed &&
	pool_state == POOL_STATE_DESTROYED)
	continue;
	if (import->do_destroyed &&
	pool_state != POOL_STATE_DESTROYED)
	continue;

	verify(nvlist_add_nvlist(config, ZPOOL_LOAD_POLICY,
	import->policy) == 0);

	if (!pool_specified) {
	if (first)
	first = B_FALSE;
	else if (!import->do_all)
	(void) fputc('\n', stdout);

	if (import->do_all) {
	import_parameters_t *ip = safe_malloc(
	sizeof (import_parameters_t));

	ip->ip_config = config;
	ip->ip_mntopts = mntopts;
	ip->ip_props = props;
	ip->ip_flags = flags;
	+ ip->ip_mntthreads = mount_tp_nthr / npools;
	ip->ip_err = &err;

	(void) tpool_dispatch(tp, do_import_task,
	(void *)ip);
	} else {
	/*
	* If we're importing from cachefile, then
	* we don't want to report errors until we
	* are in the scan phase of the import. If
	* we get an error, then we return that error
	* to invoke the scan phase.
	*/
	if (import->cachefile && !import->scan)
	err = show_import(config, B_FALSE);
	else
	(void) show_import(config, B_TRUE);
	}
	} else if (import->poolname != NULL) {
	const char *name;

	/*
	* We are searching for a pool based on name.
	*/
	verify(nvlist_lookup_string(config,
	ZPOOL_CONFIG_POOL_NAME, &name) == 0);

	if (strcmp(name, import->poolname) == 0) {
	if (found_config != NULL) {
	(void) fprintf(stderr, gettext(
	"cannot import '%s': more than "
	"one matching pool\n"),
	import->poolname);
	(void) fprintf(stderr, gettext(
	"import by numeric ID instead\n"));
	err = B_TRUE;
	}
	found_config = config;
	}
	} else {
	uint64_t guid;

	/*
	* Search for a pool by guid.
	*/
	verify(nvlist_lookup_uint64(config,
	ZPOOL_CONFIG_POOL_GUID, &guid) == 0);

	if (guid == import->guid)
	found_config = config;
	}
	}
	if (import->do_all) {
	tpool_wait(tp);
	tpool_destroy(tp);
	}

	/*
	* If we were searching for a specific pool, verify that we found a
	* pool, and then do the import.
	*/
	if (pool_specified && err == 0) {
	if (found_config == NULL) {
	(void) fprintf(stderr, gettext("cannot import '%s': "
	"no such pool available\n"), orig_name);
	err = B_TRUE;
	} else {
	err \|= do_import(found_config, new_name,
	- mntopts, props, flags);
	+ mntopts, props, flags, mount_tp_nthr);
	}
	}

	/*
	* If we were just looking for pools, report an error if none were
	* found.
	*/
	if (!pool_specified && first)
	(void) fprintf(stderr,
	gettext("no pools available to import\n"));
	return (err);
	}

	typedef struct target_exists_args {
	const char *poolname;
	uint64_t poolguid;
	} target_exists_args_t;

	static int
	name_or_guid_exists(zpool_handle_t zhp, void data)
	{
	target_exists_args_t *args = data;
	nvlist_t *config = zpool_get_config(zhp, NULL);
	int found = 0;

	if (config == NULL)
	return (0);

	if (args->poolname != NULL) {
	const char *pool_name;

	verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
	&pool_name) == 0);
	if (strcmp(pool_name, args->poolname) == 0)
	found = 1;
	} else {
	uint64_t pool_guid;

	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
	&pool_guid) == 0);
	if (pool_guid == args->poolguid)
	found = 1;
	}
	zpool_close(zhp);

	return (found);
	}
	/*
	* zpool checkpoint <pool>
	* checkpoint --discard <pool>
	*
	* -d Discard the checkpoint from a checkpointed
	* --discard pool.
	*
	* -w Wait for discarding a checkpoint to complete.
	* --wait
	*
	* Checkpoints the specified pool, by taking a "snapshot" of its
	* current state. A pool can only have one checkpoint at a time.
	*/
	int
	zpool_do_checkpoint(int argc, char **argv)
	{
	boolean_t discard, wait;
	char *pool;
	zpool_handle_t *zhp;
	int c, err;

	struct option long_options[] = {
	{"discard", no_argument, NULL, 'd'},
	{"wait", no_argument, NULL, 'w'},
	{0, 0, 0, 0}
	};

	discard = B_FALSE;
	wait = B_FALSE;
	while ((c = getopt_long(argc, argv, ":dw", long_options, NULL)) != -1) {
	switch (c) {
	case 'd':
	discard = B_TRUE;
	break;
	case 'w':
	wait = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	if (wait && !discard) {
	(void) fprintf(stderr, gettext("--wait only valid when "
	"--discard also specified\n"));
	usage(B_FALSE);
	}

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool argument\n"));
	usage(B_FALSE);
	}

	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	pool = argv[0];

	if ((zhp = zpool_open(g_zfs, pool)) == NULL) {
	/* As a special case, check for use of '/' in the name */
	if (strchr(pool, '/') != NULL)
	(void) fprintf(stderr, gettext("'zpool checkpoint' "
	"doesn't work on datasets. To save the state "
	"of a dataset from a specific point in time "
	"please use 'zfs snapshot'\n"));
	return (1);
	}

	if (discard) {
	err = (zpool_discard_checkpoint(zhp) != 0);
	if (err == 0 && wait)
	err = zpool_wait(zhp, ZPOOL_WAIT_CKPT_DISCARD);
	} else {
	err = (zpool_checkpoint(zhp) != 0);
	}

	zpool_close(zhp);

	return (err);
	}

	#define CHECKPOINT_OPT 1024

	/*
	* zpool import [-d dir] [-D]
	* import [-o mntopts] [-o prop=value] ... [-R root] [-D] [-l]
	* [-d dir \| -c cachefile \| -s] [-f] -a
	* import [-o mntopts] [-o prop=value] ... [-R root] [-D] [-l]
	* [-d dir \| -c cachefile \| -s] [-f] [-n] [-F] <pool \| id>
	* [newpool]
	*
	* -c Read pool information from a cachefile instead of searching
	* devices. If importing from a cachefile config fails, then
	* fallback to searching for devices only in the directories that
	* exist in the cachefile.
	*
	* -d Scan in a specific directory, other than /dev/. More than
	* one directory can be specified using multiple '-d' options.
	*
	* -D Scan for previously destroyed pools or import all or only
	* specified destroyed pools.
	*
	* -R Temporarily import the pool, with all mountpoints relative to
	* the given root. The pool will remain exported when the machine
	* is rebooted.
	*
	* -V Import even in the presence of faulted vdevs. This is an
	* intentionally undocumented option for testing purposes, and
	* treats the pool configuration as complete, leaving any bad
	* vdevs in the FAULTED state. In other words, it does verbatim
	* import.
	*
	* -f Force import, even if it appears that the pool is active.
	*
	* -F Attempt rewind if necessary.
	*
	* -n See if rewind would work, but don't actually rewind.
	*
	* -N Import the pool but don't mount datasets.
	*
	* -T Specify a starting txg to use for import. This option is
	* intentionally undocumented option for testing purposes.
	*
	* -a Import all pools found.
	*
	* -l Load encryption keys while importing.
	*
	* -o Set property=value and/or temporary mount options (without '=').
	*
	* -s Scan using the default search path, the libblkid cache will
	* not be consulted.
	*
	* --rewind-to-checkpoint
	* Import the pool and revert back to the checkpoint.
	*
	* The import command scans for pools to import, and import pools based on pool
	* name and GUID. The pool can also be renamed as part of the import process.
	*/
	int
	zpool_do_import(int argc, char **argv)
	{
	char **searchdirs = NULL;
	char env, envdup = NULL;
	int nsearch = 0;
	int c;
	int err = 0;
	nvlist_t *pools = NULL;
	boolean_t do_all = B_FALSE;
	boolean_t do_destroyed = B_FALSE;
	char *mntopts = NULL;
	uint64_t searchguid = 0;
	char *searchname = NULL;
	char *propval;
	nvlist_t *policy = NULL;
	nvlist_t *props = NULL;
	int flags = ZFS_IMPORT_NORMAL;
	uint32_t rewind_policy = ZPOOL_NO_REWIND;
	boolean_t dryrun = B_FALSE;
	boolean_t do_rewind = B_FALSE;
	boolean_t xtreme_rewind = B_FALSE;
	boolean_t do_scan = B_FALSE;
	boolean_t pool_exists = B_FALSE;
	uint64_t txg = -1ULL;
	char *cachefile = NULL;
	importargs_t idata = { 0 };
	char *endptr;

	struct option long_options[] = {
	{"rewind-to-checkpoint", no_argument, NULL, CHECKPOINT_OPT},
	{0, 0, 0, 0}
	};

	/* check options */
	while ((c = getopt_long(argc, argv, ":aCc:d:DEfFlmnNo:R:stT:VX",
	long_options, NULL)) != -1) {
	switch (c) {
	case 'a':
	do_all = B_TRUE;
	break;
	case 'c':
	cachefile = optarg;
	break;
	case 'd':
	searchdirs = safe_realloc(searchdirs,
	(nsearch + 1) * sizeof (char *));
	searchdirs[nsearch++] = optarg;
	break;
	case 'D':
	do_destroyed = B_TRUE;
	break;
	case 'f':
	flags \|= ZFS_IMPORT_ANY_HOST;
	break;
	case 'F':
	do_rewind = B_TRUE;
	break;
	case 'l':
	flags \|= ZFS_IMPORT_LOAD_KEYS;
	break;
	case 'm':
	flags \|= ZFS_IMPORT_MISSING_LOG;
	break;
	case 'n':
	dryrun = B_TRUE;
	break;
	case 'N':
	flags \|= ZFS_IMPORT_ONLY;
	break;
	case 'o':
	if ((propval = strchr(optarg, '=')) != NULL) {
	*propval = '\0';
	propval++;
	if (add_prop_list(optarg, propval,
	&props, B_TRUE))
	goto error;
	} else {
	mntopts = optarg;
	}
	break;
	case 'R':
	if (add_prop_list(zpool_prop_to_name(
	ZPOOL_PROP_ALTROOT), optarg, &props, B_TRUE))
	goto error;
	if (add_prop_list_default(zpool_prop_to_name(
	ZPOOL_PROP_CACHEFILE), "none", &props))
	goto error;
	break;
	case 's':
	do_scan = B_TRUE;
	break;
	case 't':
	flags \|= ZFS_IMPORT_TEMP_NAME;
	if (add_prop_list_default(zpool_prop_to_name(
	ZPOOL_PROP_CACHEFILE), "none", &props))
	goto error;
	break;

	case 'T':
	errno = 0;
	txg = strtoull(optarg, &endptr, 0);
	if (errno != 0 \|\| *endptr != '\0') {
	(void) fprintf(stderr,
	gettext("invalid txg value\n"));
	usage(B_FALSE);
	}
	rewind_policy = ZPOOL_DO_REWIND \| ZPOOL_EXTREME_REWIND;
	break;
	case 'V':
	flags \|= ZFS_IMPORT_VERBATIM;
	break;
	case 'X':
	xtreme_rewind = B_TRUE;
	break;
	case CHECKPOINT_OPT:
	flags \|= ZFS_IMPORT_CHECKPOINT;
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	usage(B_FALSE);
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	if (cachefile && nsearch != 0) {
	(void) fprintf(stderr, gettext("-c is incompatible with -d\n"));
	usage(B_FALSE);
	}

	if (cachefile && do_scan) {
	(void) fprintf(stderr, gettext("-c is incompatible with -s\n"));
	usage(B_FALSE);
	}

	if ((flags & ZFS_IMPORT_LOAD_KEYS) && (flags & ZFS_IMPORT_ONLY)) {
	(void) fprintf(stderr, gettext("-l is incompatible with -N\n"));
	usage(B_FALSE);
	}

	if ((flags & ZFS_IMPORT_LOAD_KEYS) && !do_all && argc == 0) {
	(void) fprintf(stderr, gettext("-l is only meaningful during "
	"an import\n"));
	usage(B_FALSE);
	}

	if ((dryrun \|\| xtreme_rewind) && !do_rewind) {
	(void) fprintf(stderr,
	gettext("-n or -X only meaningful with -F\n"));
	usage(B_FALSE);
	}
	if (dryrun)
	rewind_policy = ZPOOL_TRY_REWIND;
	else if (do_rewind)
	rewind_policy = ZPOOL_DO_REWIND;
	if (xtreme_rewind)
	rewind_policy \|= ZPOOL_EXTREME_REWIND;

	/* In the future, we can capture further policy and include it here */
	if (nvlist_alloc(&policy, NV_UNIQUE_NAME, 0) != 0 \|\|
	nvlist_add_uint64(policy, ZPOOL_LOAD_REQUEST_TXG, txg) != 0 \|\|
	nvlist_add_uint32(policy, ZPOOL_LOAD_REWIND_POLICY,
	rewind_policy) != 0)
	goto error;

	/* check argument count */
	if (do_all) {
	if (argc != 0) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}
	} else {
	if (argc > 2) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}
	}

	/*
	* Check for the effective uid. We do this explicitly here because
	* otherwise any attempt to discover pools will silently fail.
	*/
	if (argc == 0 && geteuid() != 0) {
	(void) fprintf(stderr, gettext("cannot "
	"discover pools: permission denied\n"));

	free(searchdirs);
	nvlist_free(props);
	nvlist_free(policy);
	return (1);
	}

	/*
	* Depending on the arguments given, we do one of the following:
	*
	* <none> Iterate through all pools and display information about
	* each one.
	*
	* -a Iterate through all pools and try to import each one.
	*
	* <id> Find the pool that corresponds to the given GUID/pool
	* name and import that one.
	*
	* -D Above options applies only to destroyed pools.
	*/
	if (argc != 0) {
	char *endptr;

	errno = 0;
	searchguid = strtoull(argv[0], &endptr, 10);
	if (errno != 0 \|\| *endptr != '\0') {
	searchname = argv[0];
	searchguid = 0;
	}

	/*
	* User specified a name or guid. Ensure it's unique.
	*/
	target_exists_args_t search = {searchname, searchguid};
	pool_exists = zpool_iter(g_zfs, name_or_guid_exists, &search);
	}

	/*
	* Check the environment for the preferred search path.
	*/
	if ((searchdirs == NULL) && (env = getenv("ZPOOL_IMPORT_PATH"))) {
	char dir, tmp = NULL;

	envdup = strdup(env);

	for (dir = strtok_r(envdup, ":", &tmp);
	dir != NULL;
	dir = strtok_r(NULL, ":", &tmp)) {
	searchdirs = safe_realloc(searchdirs,
	(nsearch + 1) * sizeof (char *));
	searchdirs[nsearch++] = dir;
	}
	}

	idata.path = searchdirs;
	idata.paths = nsearch;
	idata.poolname = searchname;
	idata.guid = searchguid;
	idata.cachefile = cachefile;
	idata.scan = do_scan;
	idata.policy = policy;
	idata.do_destroyed = do_destroyed;
	idata.do_all = do_all;

	libpc_handle_t lpch = {
	.lpc_lib_handle = g_zfs,
	.lpc_ops = &libzfs_config_ops,
	.lpc_printerr = B_TRUE
	};
	pools = zpool_search_import(&lpch, &idata);

	if (pools != NULL && pool_exists &&
	(argc == 1 \|\| strcmp(argv[0], argv[1]) == 0)) {
	(void) fprintf(stderr, gettext("cannot import '%s': "
	"a pool with that name already exists\n"),
	argv[0]);
	(void) fprintf(stderr, gettext("use the form '%s "
	"<pool \| id> <newpool>' to give it a new name\n"),
	"zpool import");
	err = 1;
	} else if (pools == NULL && pool_exists) {
	(void) fprintf(stderr, gettext("cannot import '%s': "
	"a pool with that name is already created/imported,\n"),
	argv[0]);
	(void) fprintf(stderr, gettext("and no additional pools "
	"with that name were found\n"));
	err = 1;
	} else if (pools == NULL) {
	if (argc != 0) {
	(void) fprintf(stderr, gettext("cannot import '%s': "
	"no such pool available\n"), argv[0]);
	}
	err = 1;
	}

	if (err == 1) {
	free(searchdirs);
	free(envdup);
	nvlist_free(policy);
	nvlist_free(pools);
	nvlist_free(props);
	return (1);
	}

	err = import_pools(pools, props, mntopts, flags,
	argc >= 1 ? argv[0] : NULL, argc >= 2 ? argv[1] : NULL, &idata);

	/*
	* If we're using the cachefile and we failed to import, then
	* fallback to scanning the directory for pools that match
	* those in the cachefile.
	*/
	if (err != 0 && cachefile != NULL) {
	(void) printf(gettext("cachefile import failed, retrying\n"));

	/*
	* We use the scan flag to gather the directories that exist
	* in the cachefile. If we need to fallback to searching for
	* the pool config, we will only search devices in these
	* directories.
	*/
	idata.scan = B_TRUE;
	nvlist_free(pools);
	pools = zpool_search_import(&lpch, &idata);

	err = import_pools(pools, props, mntopts, flags,
	argc >= 1 ? argv[0] : NULL, argc >= 2 ? argv[1] : NULL,
	&idata);
	}

	error:
	nvlist_free(props);
	nvlist_free(pools);
	nvlist_free(policy);
	free(searchdirs);
	free(envdup);

	return (err ? 1 : 0);
	}

	/*
	* zpool sync [-f] [pool] ...
	*
	* -f (undocumented) force uberblock (and config including zpool cache file)
	* update.
	*
	* Sync the specified pool(s).
	* Without arguments "zpool sync" will sync all pools.
	* This command initiates TXG sync(s) and will return after the TXG(s) commit.
	*
	*/
	static int
	zpool_do_sync(int argc, char **argv)
	{
	int ret;
	boolean_t force = B_FALSE;

	/* check options */
	while ((ret = getopt(argc, argv, "f")) != -1) {
	switch (ret) {
	case 'f':
	force = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* if argc == 0 we will execute zpool_sync_one on all pools */
	ret = for_each_pool(argc, argv, B_FALSE, NULL, ZFS_TYPE_POOL,
	B_FALSE, zpool_sync_one, &force);

	return (ret);
	}

	typedef struct iostat_cbdata {
	uint64_t cb_flags;
	int cb_namewidth;
	int cb_iteration;
	boolean_t cb_verbose;
	boolean_t cb_literal;
	boolean_t cb_scripted;
	zpool_list_t *cb_list;
	vdev_cmd_data_list_t *vcdl;
	vdev_cbdata_t cb_vdevs;
	} iostat_cbdata_t;

	/* iostat labels */
	typedef struct name_and_columns {
	const char name; / Column name */
	unsigned int columns; /* Center name to this number of columns */
	} name_and_columns_t;

	#define IOSTAT_MAX_LABELS 15 /* Max number of labels on one line */

	static const name_and_columns_t iostat_top_labels[][IOSTAT_MAX_LABELS] =
	{
	[IOS_DEFAULT] = {{"capacity", 2}, {"operations", 2}, {"bandwidth", 2},
	{NULL}},
	[IOS_LATENCY] = {{"total_wait", 2}, {"disk_wait", 2}, {"syncq_wait", 2},
	{"asyncq_wait", 2}, {"scrub", 1}, {"trim", 1}, {"rebuild", 1},
	{NULL}},
	[IOS_QUEUES] = {{"syncq_read", 2}, {"syncq_write", 2},
	{"asyncq_read", 2}, {"asyncq_write", 2}, {"scrubq_read", 2},
	{"trimq_write", 2}, {"rebuildq_write", 2}, {NULL}},
	[IOS_L_HISTO] = {{"total_wait", 2}, {"disk_wait", 2}, {"syncq_wait", 2},
	{"asyncq_wait", 2}, {NULL}},
	[IOS_RQ_HISTO] = {{"sync_read", 2}, {"sync_write", 2},
	{"async_read", 2}, {"async_write", 2}, {"scrub", 2},
	{"trim", 2}, {"rebuild", 2}, {NULL}},
	};

	/* Shorthand - if "columns" field not set, default to 1 column */
	static const name_and_columns_t iostat_bottom_labels[][IOSTAT_MAX_LABELS] =
	{
	[IOS_DEFAULT] = {{"alloc"}, {"free"}, {"read"}, {"write"}, {"read"},
	{"write"}, {NULL}},
	[IOS_LATENCY] = {{"read"}, {"write"}, {"read"}, {"write"}, {"read"},
	{"write"}, {"read"}, {"write"}, {"wait"}, {"wait"}, {"wait"},
	{NULL}},
	[IOS_QUEUES] = {{"pend"}, {"activ"}, {"pend"}, {"activ"}, {"pend"},
	{"activ"}, {"pend"}, {"activ"}, {"pend"}, {"activ"},
	{"pend"}, {"activ"}, {"pend"}, {"activ"}, {NULL}},
	[IOS_L_HISTO] = {{"read"}, {"write"}, {"read"}, {"write"}, {"read"},
	{"write"}, {"read"}, {"write"}, {"scrub"}, {"trim"}, {"rebuild"},
	{NULL}},
	[IOS_RQ_HISTO] = {{"ind"}, {"agg"}, {"ind"}, {"agg"}, {"ind"}, {"agg"},
	{"ind"}, {"agg"}, {"ind"}, {"agg"}, {"ind"}, {"agg"},
	{"ind"}, {"agg"}, {NULL}},
	};

	static const char *histo_to_title[] = {
	[IOS_L_HISTO] = "latency",
	[IOS_RQ_HISTO] = "req_size",
	};

	/*
	* Return the number of labels in a null-terminated name_and_columns_t
	* array.
	*
	*/
	static unsigned int
	label_array_len(const name_and_columns_t *labels)
	{
	int i = 0;

	while (labels[i].name)
	i++;

	return (i);
	}

	/*
	* Return the number of strings in a null-terminated string array.
	* For example:
	*
	* const char foo[] = {"bar", "baz", NULL}
	*
	* returns 2
	*/
	static uint64_t
	str_array_len(const char *array[])
	{
	uint64_t i = 0;
	while (array[i])
	i++;

	return (i);
	}


	/*
	* Return a default column width for default/latency/queue columns. This does
	* not include histograms, which have their columns autosized.
	*/
	static unsigned int
	default_column_width(iostat_cbdata_t *cb, enum iostat_type type)
	{
	unsigned long column_width = 5; /* Normal niceprint */
	static unsigned long widths[] = {
	/*
	* Choose some sane default column sizes for printing the
	* raw numbers.
	*/
	[IOS_DEFAULT] = 15, /* 1PB capacity */
	[IOS_LATENCY] = 10, /* 1B ns = 10sec */
	[IOS_QUEUES] = 6, /* 1M queue entries */
	[IOS_L_HISTO] = 10, /* 1B ns = 10sec */
	[IOS_RQ_HISTO] = 6, /* 1M queue entries */
	};

	if (cb->cb_literal)
	column_width = widths[type];

	return (column_width);
	}

	/*
	* Print the column labels, i.e:
	*
	* capacity operations bandwidth
	* alloc free read write read write ...
	*
	* If force_column_width is set, use it for the column width. If not set, use
	* the default column width.
	*/
	static void
	print_iostat_labels(iostat_cbdata_t *cb, unsigned int force_column_width,
	const name_and_columns_t labels[][IOSTAT_MAX_LABELS])
	{
	int i, idx, s;
	int text_start, rw_column_width, spaces_to_end;
	uint64_t flags = cb->cb_flags;
	uint64_t f;
	unsigned int column_width = force_column_width;

	/* For each bit set in flags */
	for (f = flags; f; f &= ~(1ULL << idx)) {
	idx = lowbit64(f) - 1;
	if (!force_column_width)
	column_width = default_column_width(cb, idx);
	/* Print our top labels centered over "read write" label. */
	for (i = 0; i < label_array_len(labels[idx]); i++) {
	const char *name = labels[idx][i].name;
	/*
	* We treat labels[][].columns == 0 as shorthand
	* for one column. It makes writing out the label
	* tables more concise.
	*/
	unsigned int columns = MAX(1, labels[idx][i].columns);
	unsigned int slen = strlen(name);

	rw_column_width = (column_width * columns) +
	(2 * (columns - 1));

	text_start = (int)((rw_column_width) / columns -
	slen / columns);
	if (text_start < 0)
	text_start = 0;

	printf(" "); /* Two spaces between columns */

	/* Space from beginning of column to label */
	for (s = 0; s < text_start; s++)
	printf(" ");

	printf("%s", name);

	/* Print space after label to end of column */
	spaces_to_end = rw_column_width - text_start - slen;
	if (spaces_to_end < 0)
	spaces_to_end = 0;

	for (s = 0; s < spaces_to_end; s++)
	printf(" ");
	}
	}
	}


	/*
	* print_cmd_columns - Print custom column titles from -c
	*
	* If the user specified the "zpool status\|iostat -c" then print their custom
	* column titles in the header. For example, print_cmd_columns() would print
	* the " col1 col2" part of this:
	*
	* $ zpool iostat -vc 'echo col1=val1; echo col2=val2'
	* ...
	* capacity operations bandwidth
	* pool alloc free read write read write col1 col2
	* ---------- ----- ----- ----- ----- ----- ----- ---- ----
	* mypool 269K 1008M 0 0 107 946
	* mirror 269K 1008M 0 0 107 946
	* sdb - - 0 0 102 473 val1 val2
	* sdc - - 0 0 5 473 val1 val2
	* ---------- ----- ----- ----- ----- ----- ----- ---- ----
	*/
	static void
	print_cmd_columns(vdev_cmd_data_list_t *vcdl, int use_dashes)
	{
	int i, j;
	vdev_cmd_data_t *data = &vcdl->data[0];

	if (vcdl->count == 0 \|\| data == NULL)
	return;

	/*
	* Each vdev cmd should have the same column names unless the user did
	* something weird with their cmd. Just take the column names from the
	* first vdev and assume it works for all of them.
	*/
	for (i = 0; i < vcdl->uniq_cols_cnt; i++) {
	printf(" ");
	if (use_dashes) {
	for (j = 0; j < vcdl->uniq_cols_width[i]; j++)
	printf("-");
	} else {
	printf_color(ANSI_BOLD, "%*s", vcdl->uniq_cols_width[i],
	vcdl->uniq_cols[i]);
	}
	}
	}


	/*
	* Utility function to print out a line of dashes like:
	*
	* -------------------------------- ----- ----- ----- ----- -----
	*
	* ...or a dashed named-row line like:
	*
	* logs - - - - -
	*
	* @cb: iostat data
	*
	* @force_column_width If non-zero, use the value as the column width.
	* Otherwise use the default column widths.
	*
	* @name: Print a dashed named-row line starting
	* with @name. Otherwise, print a regular
	* dashed line.
	*/
	static void
	print_iostat_dashes(iostat_cbdata_t *cb, unsigned int force_column_width,
	const char *name)
	{
	int i;
	unsigned int namewidth;
	uint64_t flags = cb->cb_flags;
	uint64_t f;
	int idx;
	const name_and_columns_t *labels;
	const char *title;


	if (cb->cb_flags & IOS_ANYHISTO_M) {
	title = histo_to_title[IOS_HISTO_IDX(cb->cb_flags)];
	} else if (cb->cb_vdevs.cb_names_count) {
	title = "vdev";
	} else {
	title = "pool";
	}

	namewidth = MAX(MAX(strlen(title), cb->cb_namewidth),
	name ? strlen(name) : 0);


	if (name) {
	printf("%-*s", namewidth, name);
	} else {
	for (i = 0; i < namewidth; i++)
	(void) printf("-");
	}

	/* For each bit in flags */
	for (f = flags; f; f &= ~(1ULL << idx)) {
	unsigned int column_width;
	idx = lowbit64(f) - 1;
	if (force_column_width)
	column_width = force_column_width;
	else
	column_width = default_column_width(cb, idx);

	labels = iostat_bottom_labels[idx];
	for (i = 0; i < label_array_len(labels); i++) {
	if (name)
	printf(" %*s-", column_width - 1, " ");
	else
	printf(" %.*s", column_width,
	"--------------------");
	}
	}
	}


	static void
	print_iostat_separator_impl(iostat_cbdata_t *cb,
	unsigned int force_column_width)
	{
	print_iostat_dashes(cb, force_column_width, NULL);
	}

	static void
	print_iostat_separator(iostat_cbdata_t *cb)
	{
	print_iostat_separator_impl(cb, 0);
	}

	static void
	print_iostat_header_impl(iostat_cbdata_t *cb, unsigned int force_column_width,
	const char *histo_vdev_name)
	{
	unsigned int namewidth;
	const char *title;

	color_start(ANSI_BOLD);

	if (cb->cb_flags & IOS_ANYHISTO_M) {
	title = histo_to_title[IOS_HISTO_IDX(cb->cb_flags)];
	} else if (cb->cb_vdevs.cb_names_count) {
	title = "vdev";
	} else {
	title = "pool";
	}

	namewidth = MAX(MAX(strlen(title), cb->cb_namewidth),
	histo_vdev_name ? strlen(histo_vdev_name) : 0);

	if (histo_vdev_name)
	printf("%-*s", namewidth, histo_vdev_name);
	else
	printf("%*s", namewidth, "");


	print_iostat_labels(cb, force_column_width, iostat_top_labels);
	printf("\n");

	printf("%-*s", namewidth, title);

	print_iostat_labels(cb, force_column_width, iostat_bottom_labels);
	if (cb->vcdl != NULL)
	print_cmd_columns(cb->vcdl, 0);

	printf("\n");

	print_iostat_separator_impl(cb, force_column_width);

	if (cb->vcdl != NULL)
	print_cmd_columns(cb->vcdl, 1);

	color_end();

	printf("\n");
	}

	static void
	print_iostat_header(iostat_cbdata_t *cb)
	{
	print_iostat_header_impl(cb, 0, NULL);
	}

	/*
	* Prints a size string (i.e. 120M) with the suffix ("M") colored
	* by order of magnitude. Uses column_size to add padding.
	*/
	static void
	print_stat_color(const char *statbuf, unsigned int column_size)
	{
	fputs(" ", stdout);
	size_t len = strlen(statbuf);
	while (len < column_size) {
	fputc(' ', stdout);
	column_size--;
	}
	if (*statbuf == '0') {
	color_start(ANSI_GRAY);
	fputc('0', stdout);
	} else {
	for (; *statbuf; statbuf++) {
	if (*statbuf == 'K') color_start(ANSI_GREEN);
	else if (*statbuf == 'M') color_start(ANSI_YELLOW);
	else if (*statbuf == 'G') color_start(ANSI_RED);
	else if (*statbuf == 'T') color_start(ANSI_BOLD_BLUE);
	else if (*statbuf == 'P') color_start(ANSI_MAGENTA);
	else if (*statbuf == 'E') color_start(ANSI_CYAN);
	fputc(*statbuf, stdout);
	if (--column_size <= 0)
	break;
	}
	}
	color_end();
	}

	/*
	* Display a single statistic.
	*/
	static void
	print_one_stat(uint64_t value, enum zfs_nicenum_format format,
	unsigned int column_size, boolean_t scripted)
	{
	char buf[64];

	zfs_nicenum_format(value, buf, sizeof (buf), format);

	if (scripted)
	printf("\t%s", buf);
	else
	print_stat_color(buf, column_size);
	}

	/*
	* Calculate the default vdev stats
	*
	* Subtract oldvs from newvs, apply a scaling factor, and save the resulting
	* stats into calcvs.
	*/
	static void
	calc_default_iostats(vdev_stat_t oldvs, vdev_stat_t newvs,
	vdev_stat_t *calcvs)
	{
	int i;

	memcpy(calcvs, newvs, sizeof (*calcvs));
	for (i = 0; i < ARRAY_SIZE(calcvs->vs_ops); i++)
	calcvs->vs_ops[i] = (newvs->vs_ops[i] - oldvs->vs_ops[i]);

	for (i = 0; i < ARRAY_SIZE(calcvs->vs_bytes); i++)
	calcvs->vs_bytes[i] = (newvs->vs_bytes[i] - oldvs->vs_bytes[i]);
	}

	/*
	* Internal representation of the extended iostats data.
	*
	* The extended iostat stats are exported in nvlists as either uint64_t arrays
	* or single uint64_t's. We make both look like arrays to make them easier
	* to process. In order to make single uint64_t's look like arrays, we set
	* __data to the stat data, and then set *data = &__data with count = 1. Then,
	* we can just use *data and count.
	*/
	struct stat_array {
	uint64_t *data;
	uint_t count; /* Number of entries in data[] */
	uint64_t __data; /* Only used when data is a single uint64_t */
	};

	static uint64_t
	stat_histo_max(struct stat_array *nva, unsigned int len)
	{
	uint64_t max = 0;
	int i;
	for (i = 0; i < len; i++)
	max = MAX(max, array64_max(nva[i].data, nva[i].count));

	return (max);
	}

	/*
	* Helper function to lookup a uint64_t array or uint64_t value and store its
	* data as a stat_array. If the nvpair is a single uint64_t value, then we make
	* it look like a one element array to make it easier to process.
	*/
	static int
	nvpair64_to_stat_array(nvlist_t nvl, const char name,
	struct stat_array *nva)
	{
	nvpair_t *tmp;
	int ret;

	verify(nvlist_lookup_nvpair(nvl, name, &tmp) == 0);
	switch (nvpair_type(tmp)) {
	case DATA_TYPE_UINT64_ARRAY:
	ret = nvpair_value_uint64_array(tmp, &nva->data, &nva->count);
	break;
	case DATA_TYPE_UINT64:
	ret = nvpair_value_uint64(tmp, &nva->__data);
	nva->data = &nva->__data;
	nva->count = 1;
	break;
	default:
	/* Not a uint64_t */
	ret = EINVAL;
	break;
	}

	return (ret);
	}

	/*
	* Given a list of nvlist names, look up the extended stats in newnv and oldnv,
	* subtract them, and return the results in a newly allocated stat_array.
	* You must free the returned array after you are done with it with
	* free_calc_stats().
	*
	* Additionally, you can set "oldnv" to NULL if you simply want the newnv
	* values.
	*/
	static struct stat_array *
	calc_and_alloc_stats_ex(const char *names, unsigned int len, nvlist_t oldnv,
	nvlist_t *newnv)
	{
	nvlist_t oldnvx = NULL, newnvx;
	struct stat_array oldnva, newnva, *calcnva;
	int i, j;
	unsigned int alloc_size = (sizeof (struct stat_array)) * len;

	/* Extract our extended stats nvlist from the main list */
	verify(nvlist_lookup_nvlist(newnv, ZPOOL_CONFIG_VDEV_STATS_EX,
	&newnvx) == 0);
	if (oldnv) {
	verify(nvlist_lookup_nvlist(oldnv, ZPOOL_CONFIG_VDEV_STATS_EX,
	&oldnvx) == 0);
	}

	newnva = safe_malloc(alloc_size);
	oldnva = safe_malloc(alloc_size);
	calcnva = safe_malloc(alloc_size);

	for (j = 0; j < len; j++) {
	verify(nvpair64_to_stat_array(newnvx, names[j],
	&newnva[j]) == 0);
	calcnva[j].count = newnva[j].count;
	alloc_size = calcnva[j].count * sizeof (calcnva[j].data[0]);
	calcnva[j].data = safe_malloc(alloc_size);
	memcpy(calcnva[j].data, newnva[j].data, alloc_size);

	if (oldnvx) {
	verify(nvpair64_to_stat_array(oldnvx, names[j],
	&oldnva[j]) == 0);
	for (i = 0; i < oldnva[j].count; i++)
	calcnva[j].data[i] -= oldnva[j].data[i];
	}
	}
	free(newnva);
	free(oldnva);
	return (calcnva);
	}

	static void
	free_calc_stats(struct stat_array *nva, unsigned int len)
	{
	int i;
	for (i = 0; i < len; i++)
	free(nva[i].data);

	free(nva);
	}

	static void
	print_iostat_histo(struct stat_array *nva, unsigned int len,
	iostat_cbdata_t *cb, unsigned int column_width, unsigned int namewidth,
	double scale)
	{
	int i, j;
	char buf[6];
	uint64_t val;
	enum zfs_nicenum_format format;
	unsigned int buckets;
	unsigned int start_bucket;

	if (cb->cb_literal)
	format = ZFS_NICENUM_RAW;
	else
	format = ZFS_NICENUM_1024;

	/* All these histos are the same size, so just use nva[0].count */
	buckets = nva[0].count;

	if (cb->cb_flags & IOS_RQ_HISTO_M) {
	/* Start at 512 - req size should never be lower than this */
	start_bucket = 9;
	} else {
	start_bucket = 0;
	}

	for (j = start_bucket; j < buckets; j++) {
	/* Print histogram bucket label */
	if (cb->cb_flags & IOS_L_HISTO_M) {
	/* Ending range of this bucket */
	val = (1UL << (j + 1)) - 1;
	zfs_nicetime(val, buf, sizeof (buf));
	} else {
	/* Request size (starting range of bucket) */
	val = (1UL << j);
	zfs_nicenum(val, buf, sizeof (buf));
	}

	if (cb->cb_scripted)
	printf("%llu", (u_longlong_t)val);
	else
	printf("%-*s", namewidth, buf);

	/* Print the values on the line */
	for (i = 0; i < len; i++) {
	print_one_stat(nva[i].data[j] * scale, format,
	column_width, cb->cb_scripted);
	}
	printf("\n");
	}
	}

	static void
	print_solid_separator(unsigned int length)
	{
	while (length--)
	printf("-");
	printf("\n");
	}

	static void
	print_iostat_histos(iostat_cbdata_t cb, nvlist_t oldnv,
	nvlist_t newnv, double scale, const char name)
	{
	unsigned int column_width;
	unsigned int namewidth;
	unsigned int entire_width;
	enum iostat_type type;
	struct stat_array *nva;
	const char **names;
	unsigned int names_len;

	/* What type of histo are we? */
	type = IOS_HISTO_IDX(cb->cb_flags);

	/* Get NULL-terminated array of nvlist names for our histo */
	names = vsx_type_to_nvlist[type];
	names_len = str_array_len(names); /* num of names */

	nva = calc_and_alloc_stats_ex(names, names_len, oldnv, newnv);

	if (cb->cb_literal) {
	column_width = MAX(5,
	(unsigned int) log10(stat_histo_max(nva, names_len)) + 1);
	} else {
	column_width = 5;
	}

	namewidth = MAX(cb->cb_namewidth,
	strlen(histo_to_title[IOS_HISTO_IDX(cb->cb_flags)]));

	/*
	* Calculate the entire line width of what we're printing. The
	* +2 is for the two spaces between columns:
	*/
	/* read write */
	/* ----- ----- */
	/* \|___\| <---------- column_width */
	/* */
	/* \|__________\| <--- entire_width */
	/* */
	entire_width = namewidth + (column_width + 2) *
	label_array_len(iostat_bottom_labels[type]);

	if (cb->cb_scripted)
	printf("%s\n", name);
	else
	print_iostat_header_impl(cb, column_width, name);

	print_iostat_histo(nva, names_len, cb, column_width,
	namewidth, scale);

	free_calc_stats(nva, names_len);
	if (!cb->cb_scripted)
	print_solid_separator(entire_width);
	}

	/*
	* Calculate the average latency of a power-of-two latency histogram
	*/
	static uint64_t
	single_histo_average(uint64_t *histo, unsigned int buckets)
	{
	int i;
	uint64_t count = 0, total = 0;

	for (i = 0; i < buckets; i++) {
	/*
	* Our buckets are power-of-two latency ranges. Use the
	* midpoint latency of each bucket to calculate the average.
	* For example:
	*
	* Bucket Midpoint
	* 8ns-15ns: 12ns
	* 16ns-31ns: 24ns
	* ...
	*/
	if (histo[i] != 0) {
	total += histo[i] * (((1UL << i) + ((1UL << i)/2)));
	count += histo[i];
	}
	}

	/* Prevent divide by zero */
	return (count == 0 ? 0 : total / count);
	}

	static void
	print_iostat_queues(iostat_cbdata_t cb, nvlist_t newnv)
	{
	const char *names[] = {
	ZPOOL_CONFIG_VDEV_SYNC_R_PEND_QUEUE,
	ZPOOL_CONFIG_VDEV_SYNC_R_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_SYNC_W_PEND_QUEUE,
	ZPOOL_CONFIG_VDEV_SYNC_W_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_ASYNC_R_PEND_QUEUE,
	ZPOOL_CONFIG_VDEV_ASYNC_R_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_ASYNC_W_PEND_QUEUE,
	ZPOOL_CONFIG_VDEV_ASYNC_W_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_SCRUB_PEND_QUEUE,
	ZPOOL_CONFIG_VDEV_SCRUB_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_TRIM_PEND_QUEUE,
	ZPOOL_CONFIG_VDEV_TRIM_ACTIVE_QUEUE,
	ZPOOL_CONFIG_VDEV_REBUILD_PEND_QUEUE,
	ZPOOL_CONFIG_VDEV_REBUILD_ACTIVE_QUEUE,
	};

	struct stat_array *nva;

	unsigned int column_width = default_column_width(cb, IOS_QUEUES);
	enum zfs_nicenum_format format;

	nva = calc_and_alloc_stats_ex(names, ARRAY_SIZE(names), NULL, newnv);

	if (cb->cb_literal)
	format = ZFS_NICENUM_RAW;
	else
	format = ZFS_NICENUM_1024;

	for (int i = 0; i < ARRAY_SIZE(names); i++) {
	uint64_t val = nva[i].data[0];
	print_one_stat(val, format, column_width, cb->cb_scripted);
	}

	free_calc_stats(nva, ARRAY_SIZE(names));
	}

	static void
	print_iostat_latency(iostat_cbdata_t cb, nvlist_t oldnv,
	nvlist_t *newnv)
	{
	int i;
	uint64_t val;
	const char *names[] = {
	ZPOOL_CONFIG_VDEV_TOT_R_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_TOT_W_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_DISK_R_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_DISK_W_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_SYNC_R_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_SYNC_W_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_ASYNC_R_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_ASYNC_W_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_SCRUB_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_TRIM_LAT_HISTO,
	ZPOOL_CONFIG_VDEV_REBUILD_LAT_HISTO,
	};
	struct stat_array *nva;

	unsigned int column_width = default_column_width(cb, IOS_LATENCY);
	enum zfs_nicenum_format format;

	nva = calc_and_alloc_stats_ex(names, ARRAY_SIZE(names), oldnv, newnv);

	if (cb->cb_literal)
	format = ZFS_NICENUM_RAWTIME;
	else
	format = ZFS_NICENUM_TIME;

	/* Print our avg latencies on the line */
	for (i = 0; i < ARRAY_SIZE(names); i++) {
	/* Compute average latency for a latency histo */
	val = single_histo_average(nva[i].data, nva[i].count);
	print_one_stat(val, format, column_width, cb->cb_scripted);
	}
	free_calc_stats(nva, ARRAY_SIZE(names));
	}

	/*
	* Print default statistics (capacity/operations/bandwidth)
	*/
	static void
	print_iostat_default(vdev_stat_t vs, iostat_cbdata_t cb, double scale)
	{
	unsigned int column_width = default_column_width(cb, IOS_DEFAULT);
	enum zfs_nicenum_format format;
	char na; /* char to print for "not applicable" values */

	if (cb->cb_literal) {
	format = ZFS_NICENUM_RAW;
	na = '0';
	} else {
	format = ZFS_NICENUM_1024;
	na = '-';
	}

	/* only toplevel vdevs have capacity stats */
	if (vs->vs_space == 0) {
	if (cb->cb_scripted)
	printf("\t%c\t%c", na, na);
	else
	printf(" %c %c", column_width, na, column_width,
	na);
	} else {
	print_one_stat(vs->vs_alloc, format, column_width,
	cb->cb_scripted);
	print_one_stat(vs->vs_space - vs->vs_alloc, format,
	column_width, cb->cb_scripted);
	}

	print_one_stat((uint64_t)(vs->vs_ops[ZIO_TYPE_READ] * scale),
	format, column_width, cb->cb_scripted);
	print_one_stat((uint64_t)(vs->vs_ops[ZIO_TYPE_WRITE] * scale),
	format, column_width, cb->cb_scripted);
	print_one_stat((uint64_t)(vs->vs_bytes[ZIO_TYPE_READ] * scale),
	format, column_width, cb->cb_scripted);
	print_one_stat((uint64_t)(vs->vs_bytes[ZIO_TYPE_WRITE] * scale),
	format, column_width, cb->cb_scripted);
	}

	static const char *const class_name[] = {
	VDEV_ALLOC_BIAS_DEDUP,
	VDEV_ALLOC_BIAS_SPECIAL,
	VDEV_ALLOC_CLASS_LOGS
	};

	/*
	* Print out all the statistics for the given vdev. This can either be the
	* toplevel configuration, or called recursively. If 'name' is NULL, then this
	* is a verbose output, and we don't want to display the toplevel pool stats.
	*
	* Returns the number of stat lines printed.
	*/
	static unsigned int
	print_vdev_stats(zpool_handle_t zhp, const char name, nvlist_t *oldnv,
	nvlist_t newnv, iostat_cbdata_t cb, int depth)
	{
	nvlist_t oldchild, newchild;
	uint_t c, children, oldchildren;
	vdev_stat_t oldvs, newvs, *calcvs;
	vdev_stat_t zerovs = { 0 };
	char *vname;
	int i;
	int ret = 0;
	uint64_t tdelta;
	double scale;

	if (strcmp(name, VDEV_TYPE_INDIRECT) == 0)
	return (ret);

	calcvs = safe_malloc(sizeof (*calcvs));

	if (oldnv != NULL) {
	verify(nvlist_lookup_uint64_array(oldnv,
	ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&oldvs, &c) == 0);
	} else {
	oldvs = &zerovs;
	}

	/* Do we only want to see a specific vdev? */
	for (i = 0; i < cb->cb_vdevs.cb_names_count; i++) {
	/* Yes we do. Is this the vdev? */
	if (strcmp(name, cb->cb_vdevs.cb_names[i]) == 0) {
	/*
	* This is our vdev. Since it is the only vdev we
	* will be displaying, make depth = 0 so that it
	* doesn't get indented.
	*/
	depth = 0;
	break;
	}
	}

	if (cb->cb_vdevs.cb_names_count && (i == cb->cb_vdevs.cb_names_count)) {
	/* Couldn't match the name */
	goto children;
	}


	verify(nvlist_lookup_uint64_array(newnv, ZPOOL_CONFIG_VDEV_STATS,
	(uint64_t **)&newvs, &c) == 0);

	/*
	* Print the vdev name unless it's is a histogram. Histograms
	* display the vdev name in the header itself.
	*/
	if (!(cb->cb_flags & IOS_ANYHISTO_M)) {
	if (cb->cb_scripted) {
	printf("%s", name);
	} else {
	if (strlen(name) + depth > cb->cb_namewidth)
	(void) printf("%*s%s", depth, "", name);
	else
	(void) printf("%s%s%s", depth, "", name,
	(int)(cb->cb_namewidth - strlen(name) -
	depth), "");
	}
	}

	/* Calculate our scaling factor */
	tdelta = newvs->vs_timestamp - oldvs->vs_timestamp;
	if ((oldvs->vs_timestamp == 0) && (cb->cb_flags & IOS_ANYHISTO_M)) {
	/*
	* If we specify printing histograms with no time interval, then
	* print the histogram numbers over the entire lifetime of the
	* vdev.
	*/
	scale = 1;
	} else {
	if (tdelta == 0)
	scale = 1.0;
	else
	scale = (double)NANOSEC / tdelta;
	}

	if (cb->cb_flags & IOS_DEFAULT_M) {
	calc_default_iostats(oldvs, newvs, calcvs);
	print_iostat_default(calcvs, cb, scale);
	}
	if (cb->cb_flags & IOS_LATENCY_M)
	print_iostat_latency(cb, oldnv, newnv);
	if (cb->cb_flags & IOS_QUEUES_M)
	print_iostat_queues(cb, newnv);
	if (cb->cb_flags & IOS_ANYHISTO_M) {
	printf("\n");
	print_iostat_histos(cb, oldnv, newnv, scale, name);
	}

	if (cb->vcdl != NULL) {
	const char *path;
	if (nvlist_lookup_string(newnv, ZPOOL_CONFIG_PATH,
	&path) == 0) {
	printf(" ");
	zpool_print_cmd(cb->vcdl, zpool_get_name(zhp), path);
	}
	}

	if (!(cb->cb_flags & IOS_ANYHISTO_M))
	printf("\n");

	ret++;

	children:

	free(calcvs);

	if (!cb->cb_verbose)
	return (ret);

	if (nvlist_lookup_nvlist_array(newnv, ZPOOL_CONFIG_CHILDREN,
	&newchild, &children) != 0)
	return (ret);

	if (oldnv) {
	if (nvlist_lookup_nvlist_array(oldnv, ZPOOL_CONFIG_CHILDREN,
	&oldchild, &oldchildren) != 0)
	return (ret);

	children = MIN(oldchildren, children);
	}

	/*
	* print normal top-level devices
	*/
	for (c = 0; c < children; c++) {
	uint64_t ishole = B_FALSE, islog = B_FALSE;

	(void) nvlist_lookup_uint64(newchild[c], ZPOOL_CONFIG_IS_HOLE,
	&ishole);

	(void) nvlist_lookup_uint64(newchild[c], ZPOOL_CONFIG_IS_LOG,
	&islog);

	if (ishole \|\| islog)
	continue;

	if (nvlist_exists(newchild[c], ZPOOL_CONFIG_ALLOCATION_BIAS))
	continue;

	vname = zpool_vdev_name(g_zfs, zhp, newchild[c],
	cb->cb_vdevs.cb_name_flags \| VDEV_NAME_TYPE_ID);
	ret += print_vdev_stats(zhp, vname, oldnv ? oldchild[c] : NULL,
	newchild[c], cb, depth + 2);
	free(vname);
	}

	/*
	* print all other top-level devices
	*/
	for (uint_t n = 0; n < ARRAY_SIZE(class_name); n++) {
	boolean_t printed = B_FALSE;

	for (c = 0; c < children; c++) {
	uint64_t islog = B_FALSE;
	const char *bias = NULL;
	const char *type = NULL;

	(void) nvlist_lookup_uint64(newchild[c],
	ZPOOL_CONFIG_IS_LOG, &islog);
	if (islog) {
	bias = VDEV_ALLOC_CLASS_LOGS;
	} else {
	(void) nvlist_lookup_string(newchild[c],
	ZPOOL_CONFIG_ALLOCATION_BIAS, &bias);
	(void) nvlist_lookup_string(newchild[c],
	ZPOOL_CONFIG_TYPE, &type);
	}
	if (bias == NULL \|\| strcmp(bias, class_name[n]) != 0)
	continue;
	if (!islog && strcmp(type, VDEV_TYPE_INDIRECT) == 0)
	continue;

	if (!printed) {
	if ((!(cb->cb_flags & IOS_ANYHISTO_M)) &&
	!cb->cb_scripted &&
	!cb->cb_vdevs.cb_names) {
	print_iostat_dashes(cb, 0,
	class_name[n]);
	}
	printf("\n");
	printed = B_TRUE;
	}

	vname = zpool_vdev_name(g_zfs, zhp, newchild[c],
	cb->cb_vdevs.cb_name_flags \| VDEV_NAME_TYPE_ID);
	ret += print_vdev_stats(zhp, vname, oldnv ?
	oldchild[c] : NULL, newchild[c], cb, depth + 2);
	free(vname);
	}
	}

	/*
	* Include level 2 ARC devices in iostat output
	*/
	if (nvlist_lookup_nvlist_array(newnv, ZPOOL_CONFIG_L2CACHE,
	&newchild, &children) != 0)
	return (ret);

	if (oldnv) {
	if (nvlist_lookup_nvlist_array(oldnv, ZPOOL_CONFIG_L2CACHE,
	&oldchild, &oldchildren) != 0)
	return (ret);

	children = MIN(oldchildren, children);
	}

	if (children > 0) {
	if ((!(cb->cb_flags & IOS_ANYHISTO_M)) && !cb->cb_scripted &&
	!cb->cb_vdevs.cb_names) {
	print_iostat_dashes(cb, 0, "cache");
	}
	printf("\n");

	for (c = 0; c < children; c++) {
	vname = zpool_vdev_name(g_zfs, zhp, newchild[c],
	cb->cb_vdevs.cb_name_flags);
	ret += print_vdev_stats(zhp, vname, oldnv ? oldchild[c]
	: NULL, newchild[c], cb, depth + 2);
	free(vname);
	}
	}

	return (ret);
	}

	static int
	refresh_iostat(zpool_handle_t zhp, void data)
	{
	iostat_cbdata_t *cb = data;
	boolean_t missing;

	/*
	* If the pool has disappeared, remove it from the list and continue.
	*/
	if (zpool_refresh_stats(zhp, &missing) != 0)
	return (-1);

	if (missing)
	pool_list_remove(cb->cb_list, zhp);

	return (0);
	}

	/*
	* Callback to print out the iostats for the given pool.
	*/
	static int
	print_iostat(zpool_handle_t zhp, void data)
	{
	iostat_cbdata_t *cb = data;
	nvlist_t oldconfig, newconfig;
	nvlist_t oldnvroot, newnvroot;
	int ret;

	newconfig = zpool_get_config(zhp, &oldconfig);

	if (cb->cb_iteration == 1)
	oldconfig = NULL;

	verify(nvlist_lookup_nvlist(newconfig, ZPOOL_CONFIG_VDEV_TREE,
	&newnvroot) == 0);

	if (oldconfig == NULL)
	oldnvroot = NULL;
	else
	verify(nvlist_lookup_nvlist(oldconfig, ZPOOL_CONFIG_VDEV_TREE,
	&oldnvroot) == 0);

	ret = print_vdev_stats(zhp, zpool_get_name(zhp), oldnvroot, newnvroot,
	cb, 0);
	if ((ret != 0) && !(cb->cb_flags & IOS_ANYHISTO_M) &&
	!cb->cb_scripted && cb->cb_verbose &&
	!cb->cb_vdevs.cb_names_count) {
	print_iostat_separator(cb);
	if (cb->vcdl != NULL) {
	print_cmd_columns(cb->vcdl, 1);
	}
	printf("\n");
	}

	return (ret);
	}

	static int
	get_columns(void)
	{
	struct winsize ws;
	int columns = 80;
	int error;

	if (isatty(STDOUT_FILENO)) {
	error = ioctl(STDOUT_FILENO, TIOCGWINSZ, &ws);
	if (error == 0)
	columns = ws.ws_col;
	} else {
	columns = 999;
	}

	return (columns);
	}

	/*
	* Return the required length of the pool/vdev name column. The minimum
	* allowed width and output formatting flags must be provided.
	*/
	static int
	get_namewidth(zpool_handle_t *zhp, int min_width, int flags, boolean_t verbose)
	{
	nvlist_t config, nvroot;
	int width = min_width;

	if ((config = zpool_get_config(zhp, NULL)) != NULL) {
	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
	&nvroot) == 0);
	size_t poolname_len = strlen(zpool_get_name(zhp));
	if (verbose == B_FALSE) {
	width = MAX(poolname_len, min_width);
	} else {
	width = MAX(poolname_len,
	max_width(zhp, nvroot, 0, min_width, flags));
	}
	}

	return (width);
	}

	/*
	* Parse the input string, get the 'interval' and 'count' value if there is one.
	*/
	static void
	get_interval_count(int argcp, char argv, float iv,
	unsigned long *cnt)
	{
	float interval = 0;
	unsigned long count = 0;
	int argc = *argcp;

	/*
	* Determine if the last argument is an integer or a pool name
	*/
	if (argc > 0 && zfs_isnumber(argv[argc - 1])) {
	char *end;

	errno = 0;
	interval = strtof(argv[argc - 1], &end);

	if (*end == '\0' && errno == 0) {
	if (interval == 0) {
	(void) fprintf(stderr, gettext(
	"interval cannot be zero\n"));
	usage(B_FALSE);
	}
	/*
	* Ignore the last parameter
	*/
	argc--;
	} else {
	/*
	* If this is not a valid number, just plow on. The
	* user will get a more informative error message later
	* on.
	*/
	interval = 0;
	}
	}

	/*
	* If the last argument is also an integer, then we have both a count
	* and an interval.
	*/
	if (argc > 0 && zfs_isnumber(argv[argc - 1])) {
	char *end;

	errno = 0;
	count = interval;
	interval = strtof(argv[argc - 1], &end);

	if (*end == '\0' && errno == 0) {
	if (interval == 0) {
	(void) fprintf(stderr, gettext(
	"interval cannot be zero\n"));
	usage(B_FALSE);
	}

	/*
	* Ignore the last parameter
	*/
	argc--;
	} else {
	interval = 0;
	}
	}

	*iv = interval;
	*cnt = count;
	*argcp = argc;
	}

	static void
	get_timestamp_arg(char c)
	{
	if (c == 'u')
	timestamp_fmt = UDATE;
	else if (c == 'd')
	timestamp_fmt = DDATE;
	else
	usage(B_FALSE);
	}

	/*
	* Return stat flags that are supported by all pools by both the module and
	* zpool iostat. "*data" should be initialized to all 0xFFs before running.
	* It will get ANDed down until only the flags that are supported on all pools
	* remain.
	*/
	static int
	get_stat_flags_cb(zpool_handle_t zhp, void data)
	{
	uint64_t *mask = data;
	nvlist_t config, nvroot, *nvx;
	uint64_t flags = 0;
	int i, j;

	config = zpool_get_config(zhp, NULL);
	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
	&nvroot) == 0);

	/* Default stats are always supported, but for completeness.. */
	if (nvlist_exists(nvroot, ZPOOL_CONFIG_VDEV_STATS))
	flags \|= IOS_DEFAULT_M;

	/* Get our extended stats nvlist from the main list */
	if (nvlist_lookup_nvlist(nvroot, ZPOOL_CONFIG_VDEV_STATS_EX,
	&nvx) != 0) {
	/*
	* No extended stats; they're probably running an older
	* module. No big deal, we support that too.
	*/
	goto end;
	}

	/* For each extended stat, make sure all its nvpairs are supported */
	for (j = 0; j < ARRAY_SIZE(vsx_type_to_nvlist); j++) {
	if (!vsx_type_to_nvlist[j][0])
	continue;

	/* Start off by assuming the flag is supported, then check */
	flags \|= (1ULL << j);
	for (i = 0; vsx_type_to_nvlist[j][i]; i++) {
	if (!nvlist_exists(nvx, vsx_type_to_nvlist[j][i])) {
	/* flag isn't supported */
	flags = flags & ~(1ULL << j);
	break;
	}
	}
	}
	end:
	mask = mask & flags;
	return (0);
	}

	/*
	* Return a bitmask of stats that are supported on all pools by both the module
	* and zpool iostat.
	*/
	static uint64_t
	get_stat_flags(zpool_list_t *list)
	{
	uint64_t mask = -1;

	/*
	* get_stat_flags_cb() will lop off bits from "mask" until only the
	* flags that are supported on all pools remain.
	*/
	pool_list_iter(list, B_FALSE, get_stat_flags_cb, &mask);
	return (mask);
	}

	/*
	* Return 1 if cb_data->cb_names[0] is this vdev's name, 0 otherwise.
	*/
	static int
	is_vdev_cb(void zhp_data, nvlist_t nv, void *cb_data)
	{
	uint64_t guid;
	vdev_cbdata_t *cb = cb_data;
	zpool_handle_t *zhp = zhp_data;

	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) != 0)
	return (0);

	return (guid == zpool_vdev_path_to_guid(zhp, cb->cb_names[0]));
	}

	/*
	* Returns 1 if cb_data->cb_names[0] is a vdev name, 0 otherwise.
	*/
	static int
	is_vdev(zpool_handle_t zhp, void cb_data)
	{
	return (for_each_vdev(zhp, is_vdev_cb, cb_data));
	}

	/*
	* Check if vdevs are in a pool
	*
	* Return 1 if all argv[] strings are vdev names in pool "pool_name". Otherwise
	* return 0. If pool_name is NULL, then search all pools.
	*/
	static int
	are_vdevs_in_pool(int argc, char *argv, char pool_name,
	vdev_cbdata_t *cb)
	{
	char **tmp_name;
	int ret = 0;
	int i;
	int pool_count = 0;

	if ((argc == 0) \|\| !*argv)
	return (0);

	if (pool_name)
	pool_count = 1;

	/* Temporarily hijack cb_names for a second... */
	tmp_name = cb->cb_names;

	/* Go though our list of prospective vdev names */
	for (i = 0; i < argc; i++) {
	cb->cb_names = argv + i;

	/* Is this name a vdev in our pools? */
	ret = for_each_pool(pool_count, &pool_name, B_TRUE, NULL,
	ZFS_TYPE_POOL, B_FALSE, is_vdev, cb);
	if (!ret) {
	/* No match */
	break;
	}
	}

	cb->cb_names = tmp_name;

	return (ret);
	}

	static int
	is_pool_cb(zpool_handle_t zhp, void data)
	{
	char *name = data;
	if (strcmp(name, zpool_get_name(zhp)) == 0)
	return (1);

	return (0);
	}

	/*
	* Do we have a pool named *name? If so, return 1, otherwise 0.
	*/
	static int
	is_pool(char *name)
	{
	return (for_each_pool(0, NULL, B_TRUE, NULL, ZFS_TYPE_POOL, B_FALSE,
	is_pool_cb, name));
	}

	/* Are all our argv[] strings pool names? If so return 1, 0 otherwise. */
	static int
	are_all_pools(int argc, char **argv)
	{
	if ((argc == 0) \|\| !*argv)
	return (0);

	while (--argc >= 0)
	if (!is_pool(argv[argc]))
	return (0);

	return (1);
	}

	/*
	* Helper function to print out vdev/pool names we can't resolve. Used for an
	* error message.
	*/
	static void
	error_list_unresolved_vdevs(int argc, char *argv, char pool_name,
	vdev_cbdata_t *cb)
	{
	int i;
	char *name;
	char *str;
	for (i = 0; i < argc; i++) {
	name = argv[i];

	if (is_pool(name))
	str = gettext("pool");
	else if (are_vdevs_in_pool(1, &name, pool_name, cb))
	str = gettext("vdev in this pool");
	else if (are_vdevs_in_pool(1, &name, NULL, cb))
	str = gettext("vdev in another pool");
	else
	str = gettext("unknown");

	fprintf(stderr, "\t%s (%s)\n", name, str);
	}
	}

	/*
	* Same as get_interval_count(), but with additional checks to not misinterpret
	* guids as interval/count values. Assumes VDEV_NAME_GUID is set in
	* cb.cb_vdevs.cb_name_flags.
	*/
	static void
	get_interval_count_filter_guids(int argc, char argv, float interval,
	unsigned long count, iostat_cbdata_t cb)
	{
	char **tmpargv = argv;
	int argc_for_interval = 0;

	/* Is the last arg an interval value? Or a guid? */
	if (argc >= 1 && !are_vdevs_in_pool(1, &argv[argc - 1], NULL,
	&cb->cb_vdevs)) {
	/*
	* The last arg is not a guid, so it's probably an
	* interval value.
	*/
	argc_for_interval++;

	if (*argc >= 2 &&
	!are_vdevs_in_pool(1, &argv[*argc - 2], NULL,
	&cb->cb_vdevs)) {
	/*
	* The 2nd to last arg is not a guid, so it's probably
	* an interval value.
	*/
	argc_for_interval++;
	}
	}

	/* Point to our list of possible intervals */
	tmpargv = &argv[*argc - argc_for_interval];

	argc = argc - argc_for_interval;
	get_interval_count(&argc_for_interval, tmpargv,
	interval, count);
	}

	/*
	* Terminal height, in rows. Returns -1 if stdout is not connected to a TTY or
	* if we were unable to determine its size.
	*/
	static int
	terminal_height(void)
	{
	struct winsize win;

	if (isatty(STDOUT_FILENO) == 0)
	return (-1);

	if (ioctl(STDOUT_FILENO, TIOCGWINSZ, &win) != -1 && win.ws_row > 0)
	return (win.ws_row);

	return (-1);
	}

	/*
	* Run one of the zpool status/iostat -c scripts with the help (-h) option and
	* print the result.
	*
	* name: Short name of the script ('iostat').
	* path: Full path to the script ('/usr/local/etc/zfs/zpool.d/iostat');
	*/
	static void
	print_zpool_script_help(char name, char path)
	{
	char argv[] = {path, (char )"-h", NULL};
	char **lines = NULL;
	int lines_cnt = 0;
	int rc;

	rc = libzfs_run_process_get_stdout_nopath(path, argv, NULL, &lines,
	&lines_cnt);
	if (rc != 0 \|\| lines == NULL \|\| lines_cnt <= 0) {
	if (lines != NULL)
	libzfs_free_str_array(lines, lines_cnt);
	return;
	}

	for (int i = 0; i < lines_cnt; i++)
	if (!is_blank_str(lines[i]))
	printf(" %-14s %s\n", name, lines[i]);

	libzfs_free_str_array(lines, lines_cnt);
	}

	/*
	* Go though the zpool status/iostat -c scripts in the user's path, run their
	* help option (-h), and print out the results.
	*/
	static void
	print_zpool_dir_scripts(char *dirpath)
	{
	DIR *dir;
	struct dirent *ent;
	char fullpath[MAXPATHLEN];
	struct stat dir_stat;

	if ((dir = opendir(dirpath)) != NULL) {
	/* print all the files and directories within directory */
	while ((ent = readdir(dir)) != NULL) {
	if (snprintf(fullpath, sizeof (fullpath), "%s/%s",
	dirpath, ent->d_name) >= sizeof (fullpath)) {
	(void) fprintf(stderr,
	gettext("internal error: "
	"ZPOOL_SCRIPTS_PATH too large.\n"));
	exit(1);
	}

	/* Print the scripts */
	if (stat(fullpath, &dir_stat) == 0)
	if (dir_stat.st_mode & S_IXUSR &&
	S_ISREG(dir_stat.st_mode))
	print_zpool_script_help(ent->d_name,
	fullpath);
	}
	closedir(dir);
	}
	}

	/*
	* Print out help text for all zpool status/iostat -c scripts.
	*/
	static void
	print_zpool_script_list(const char *subcommand)
	{
	char dir, sp, *tmp;

	printf(gettext("Available 'zpool %s -c' commands:\n"), subcommand);

	sp = zpool_get_cmd_search_path();
	if (sp == NULL)
	return;

	for (dir = strtok_r(sp, ":", &tmp);
	dir != NULL;
	dir = strtok_r(NULL, ":", &tmp))
	print_zpool_dir_scripts(dir);

	free(sp);
	}

	/*
	* Set the minimum pool/vdev name column width. The width must be at least 10,
	* but may be as large as the column width - 42 so it still fits on one line.
	* NOTE: 42 is the width of the default capacity/operations/bandwidth output
	*/
	static int
	get_namewidth_iostat(zpool_handle_t zhp, void data)
	{
	iostat_cbdata_t *cb = data;
	int width, available_width;

	/*
	* get_namewidth() returns the maximum width of any name in that column
	* for any pool/vdev/device line that will be output.
	*/
	width = get_namewidth(zhp, cb->cb_namewidth,
	cb->cb_vdevs.cb_name_flags \| VDEV_NAME_TYPE_ID, cb->cb_verbose);

	/*
	* The width we are calculating is the width of the header and also the
	* padding width for names that are less than maximum width. The stats
	* take up 42 characters, so the width available for names is:
	*/
	available_width = get_columns() - 42;

	/*
	* If the maximum width fits on a screen, then great! Make everything
	* line up by justifying all lines to the same width. If that max
	* width is larger than what's available, the name plus stats won't fit
	* on one line, and justifying to that width would cause every line to
	* wrap on the screen. We only want lines with long names to wrap.
	* Limit the padding to what won't wrap.
	*/
	if (width > available_width)
	width = available_width;

	/*
	* And regardless of whatever the screen width is (get_columns can
	* return 0 if the width is not known or less than 42 for a narrow
	* terminal) have the width be a minimum of 10.
	*/
	if (width < 10)
	width = 10;

	/* Save the calculated width */
	cb->cb_namewidth = width;

	return (0);
	}

	/*
	* zpool iostat [[-c [script1,script2,...]] [-lq]\|[-rw]] [-ghHLpPvy] [-n name]
	* [-T d\|u] [[ pool ...]\|[pool vdev ...]\|[vdev ...]]
	* [interval [count]]
	*
	* -c CMD For each vdev, run command CMD
	* -g Display guid for individual vdev name.
	* -L Follow links when resolving vdev path name.
	* -P Display full path for vdev name.
	* -v Display statistics for individual vdevs
	* -h Display help
	* -p Display values in parsable (exact) format.
	* -H Scripted mode. Don't display headers, and separate properties
	* by a single tab.
	* -l Display average latency
	* -q Display queue depths
	* -w Display latency histograms
	* -r Display request size histogram
	* -T Display a timestamp in date(1) or Unix format
	* -n Only print headers once
	*
	* This command can be tricky because we want to be able to deal with pool
	* creation/destruction as well as vdev configuration changes. The bulk of this
	* processing is handled by the pool_list_* routines in zpool_iter.c. We rely
	* on pool_list_update() to detect the addition of new pools. Configuration
	* changes are all handled within libzfs.
	*/
	int
	zpool_do_iostat(int argc, char **argv)
	{
	int c;
	int ret;
	int npools;
	float interval = 0;
	unsigned long count = 0;
	int winheight = 24;
	zpool_list_t *list;
	boolean_t verbose = B_FALSE;
	boolean_t latency = B_FALSE, l_histo = B_FALSE, rq_histo = B_FALSE;
	boolean_t queues = B_FALSE, parsable = B_FALSE, scripted = B_FALSE;
	boolean_t omit_since_boot = B_FALSE;
	boolean_t guid = B_FALSE;
	boolean_t follow_links = B_FALSE;
	boolean_t full_name = B_FALSE;
	boolean_t headers_once = B_FALSE;
	iostat_cbdata_t cb = { 0 };
	char *cmd = NULL;

	/* Used for printing error message */
	const char flag_to_arg[] = {[IOS_LATENCY] = 'l', [IOS_QUEUES] = 'q',
	[IOS_L_HISTO] = 'w', [IOS_RQ_HISTO] = 'r'};

	uint64_t unsupported_flags;

	/* check options */
	while ((c = getopt(argc, argv, "c:gLPT:vyhplqrwnH")) != -1) {
	switch (c) {
	case 'c':
	if (cmd != NULL) {
	fprintf(stderr,
	gettext("Can't set -c flag twice\n"));
	exit(1);
	}

	if (getenv("ZPOOL_SCRIPTS_ENABLED") != NULL &&
	!libzfs_envvar_is_set("ZPOOL_SCRIPTS_ENABLED")) {
	fprintf(stderr, gettext(
	"Can't run -c, disabled by "
	"ZPOOL_SCRIPTS_ENABLED.\n"));
	exit(1);
	}

	if ((getuid() <= 0 \|\| geteuid() <= 0) &&
	!libzfs_envvar_is_set("ZPOOL_SCRIPTS_AS_ROOT")) {
	fprintf(stderr, gettext(
	"Can't run -c with root privileges "
	"unless ZPOOL_SCRIPTS_AS_ROOT is set.\n"));
	exit(1);
	}
	cmd = optarg;
	verbose = B_TRUE;
	break;
	case 'g':
	guid = B_TRUE;
	break;
	case 'L':
	follow_links = B_TRUE;
	break;
	case 'P':
	full_name = B_TRUE;
	break;
	case 'T':
	get_timestamp_arg(*optarg);
	break;
	case 'v':
	verbose = B_TRUE;
	break;
	case 'p':
	parsable = B_TRUE;
	break;
	case 'l':
	latency = B_TRUE;
	break;
	case 'q':
	queues = B_TRUE;
	break;
	case 'H':
	scripted = B_TRUE;
	break;
	case 'w':
	l_histo = B_TRUE;
	break;
	case 'r':
	rq_histo = B_TRUE;
	break;
	case 'y':
	omit_since_boot = B_TRUE;
	break;
	case 'n':
	headers_once = B_TRUE;
	break;
	case 'h':
	usage(B_FALSE);
	break;
	case '?':
	if (optopt == 'c') {
	print_zpool_script_list("iostat");
	exit(0);
	} else {
	fprintf(stderr,
	gettext("invalid option '%c'\n"), optopt);
	}
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	cb.cb_literal = parsable;
	cb.cb_scripted = scripted;

	if (guid)
	cb.cb_vdevs.cb_name_flags \|= VDEV_NAME_GUID;
	if (follow_links)
	cb.cb_vdevs.cb_name_flags \|= VDEV_NAME_FOLLOW_LINKS;
	if (full_name)
	cb.cb_vdevs.cb_name_flags \|= VDEV_NAME_PATH;
	cb.cb_iteration = 0;
	cb.cb_namewidth = 0;
	cb.cb_verbose = verbose;

	/* Get our interval and count values (if any) */
	if (guid) {
	get_interval_count_filter_guids(&argc, argv, &interval,
	&count, &cb);
	} else {
	get_interval_count(&argc, argv, &interval, &count);
	}

	if (argc == 0) {
	/* No args, so just print the defaults. */
	} else if (are_all_pools(argc, argv)) {
	/* All the args are pool names */
	} else if (are_vdevs_in_pool(argc, argv, NULL, &cb.cb_vdevs)) {
	/* All the args are vdevs */
	cb.cb_vdevs.cb_names = argv;
	cb.cb_vdevs.cb_names_count = argc;
	argc = 0; /* No pools to process */
	} else if (are_all_pools(1, argv)) {
	/* The first arg is a pool name */
	if (are_vdevs_in_pool(argc - 1, argv + 1, argv[0],
	&cb.cb_vdevs)) {
	/* ...and the rest are vdev names */
	cb.cb_vdevs.cb_names = argv + 1;
	cb.cb_vdevs.cb_names_count = argc - 1;
	argc = 1; /* One pool to process */
	} else {
	fprintf(stderr, gettext("Expected either a list of "));
	fprintf(stderr, gettext("pools, or list of vdevs in"));
	fprintf(stderr, " \"%s\", ", argv[0]);
	fprintf(stderr, gettext("but got:\n"));
	error_list_unresolved_vdevs(argc - 1, argv + 1,
	argv[0], &cb.cb_vdevs);
	fprintf(stderr, "\n");
	usage(B_FALSE);
	return (1);
	}
	} else {
	/*
	* The args don't make sense. The first arg isn't a pool name,
	* nor are all the args vdevs.
	*/
	fprintf(stderr, gettext("Unable to parse pools/vdevs list.\n"));
	fprintf(stderr, "\n");
	return (1);
	}

	if (cb.cb_vdevs.cb_names_count != 0) {
	/*
	* If user specified vdevs, it implies verbose.
	*/
	cb.cb_verbose = B_TRUE;
	}

	/*
	* Construct the list of all interesting pools.
	*/
	ret = 0;
	if ((list = pool_list_get(argc, argv, NULL, ZFS_TYPE_POOL, parsable,
	&ret)) == NULL)
	return (1);

	if (pool_list_count(list) == 0 && argc != 0) {
	pool_list_free(list);
	return (1);
	}

	if (pool_list_count(list) == 0 && interval == 0) {
	pool_list_free(list);
	(void) fprintf(stderr, gettext("no pools available\n"));
	return (1);
	}

	if ((l_histo \|\| rq_histo) && (cmd != NULL \|\| latency \|\| queues)) {
	pool_list_free(list);
	(void) fprintf(stderr,
	gettext("[-r\|-w] isn't allowed with [-c\|-l\|-q]\n"));
	usage(B_FALSE);
	return (1);
	}

	if (l_histo && rq_histo) {
	pool_list_free(list);
	(void) fprintf(stderr,
	gettext("Only one of [-r\|-w] can be passed at a time\n"));
	usage(B_FALSE);
	return (1);
	}

	/*
	* Enter the main iostat loop.
	*/
	cb.cb_list = list;

	if (l_histo) {
	/*
	* Histograms tables look out of place when you try to display
	* them with the other stats, so make a rule that you can only
	* print histograms by themselves.
	*/
	cb.cb_flags = IOS_L_HISTO_M;
	} else if (rq_histo) {
	cb.cb_flags = IOS_RQ_HISTO_M;
	} else {
	cb.cb_flags = IOS_DEFAULT_M;
	if (latency)
	cb.cb_flags \|= IOS_LATENCY_M;
	if (queues)
	cb.cb_flags \|= IOS_QUEUES_M;
	}

	/*
	* See if the module supports all the stats we want to display.
	*/
	unsupported_flags = cb.cb_flags & ~get_stat_flags(list);
	if (unsupported_flags) {
	uint64_t f;
	int idx;
	fprintf(stderr,
	gettext("The loaded zfs module doesn't support:"));

	/* for each bit set in unsupported_flags */
	for (f = unsupported_flags; f; f &= ~(1ULL << idx)) {
	idx = lowbit64(f) - 1;
	fprintf(stderr, " -%c", flag_to_arg[idx]);
	}

	fprintf(stderr, ". Try running a newer module.\n");
	pool_list_free(list);

	return (1);
	}

	for (;;) {
	if ((npools = pool_list_count(list)) == 0)
	(void) fprintf(stderr, gettext("no pools available\n"));
	else {
	/*
	* If this is the first iteration and -y was supplied
	* we skip any printing.
	*/
	boolean_t skip = (omit_since_boot &&
	cb.cb_iteration == 0);

	/*
	* Refresh all statistics. This is done as an
	* explicit step before calculating the maximum name
	* width, so that any * configuration changes are
	* properly accounted for.
	*/
	(void) pool_list_iter(list, B_FALSE, refresh_iostat,
	&cb);

	/*
	* Iterate over all pools to determine the maximum width
	* for the pool / device name column across all pools.
	*/
	cb.cb_namewidth = 0;
	(void) pool_list_iter(list, B_FALSE,
	get_namewidth_iostat, &cb);

	if (timestamp_fmt != NODATE)
	print_timestamp(timestamp_fmt);

	if (cmd != NULL && cb.cb_verbose &&
	!(cb.cb_flags & IOS_ANYHISTO_M)) {
	cb.vcdl = all_pools_for_each_vdev_run(argc,
	argv, cmd, g_zfs, cb.cb_vdevs.cb_names,
	cb.cb_vdevs.cb_names_count,
	cb.cb_vdevs.cb_name_flags);
	} else {
	cb.vcdl = NULL;
	}


	/*
	* Check terminal size so we can print headers
	* even when terminal window has its height
	* changed.
	*/
	winheight = terminal_height();
	/*
	* Are we connected to TTY? If not, headers_once
	* should be true, to avoid breaking scripts.
	*/
	if (winheight < 0)
	headers_once = B_TRUE;

	/*
	* If it's the first time and we're not skipping it,
	* or either skip or verbose mode, print the header.
	*
	* The histogram code explicitly prints its header on
	* every vdev, so skip this for histograms.
	*/
	if (((++cb.cb_iteration == 1 && !skip) \|\|
	(skip != verbose) \|\|
	(!headers_once &&
	(cb.cb_iteration % winheight) == 0)) &&
	(!(cb.cb_flags & IOS_ANYHISTO_M)) &&
	!cb.cb_scripted)
	print_iostat_header(&cb);

	if (skip) {
	(void) fflush(stdout);
	(void) fsleep(interval);
	continue;
	}

	pool_list_iter(list, B_FALSE, print_iostat, &cb);

	/*
	* If there's more than one pool, and we're not in
	* verbose mode (which prints a separator for us),
	* then print a separator.
	*
	* In addition, if we're printing specific vdevs then
	* we also want an ending separator.
	*/
	if (((npools > 1 && !verbose &&
	!(cb.cb_flags & IOS_ANYHISTO_M)) \|\|
	(!(cb.cb_flags & IOS_ANYHISTO_M) &&
	cb.cb_vdevs.cb_names_count)) &&
	!cb.cb_scripted) {
	print_iostat_separator(&cb);
	if (cb.vcdl != NULL)
	print_cmd_columns(cb.vcdl, 1);
	printf("\n");
	}

	if (cb.vcdl != NULL)
	free_vdev_cmd_data_list(cb.vcdl);

	}

	if (interval == 0)
	break;

	if (count != 0 && --count == 0)
	break;

	(void) fflush(stdout);
	(void) fsleep(interval);
	}

	pool_list_free(list);

	return (ret);
	}

	typedef struct list_cbdata {
	boolean_t cb_verbose;
	int cb_name_flags;
	int cb_namewidth;
	boolean_t cb_scripted;
	zprop_list_t *cb_proplist;
	boolean_t cb_literal;
	} list_cbdata_t;


	/*
	* Given a list of columns to display, output appropriate headers for each one.
	*/
	static void
	print_header(list_cbdata_t *cb)
	{
	zprop_list_t *pl = cb->cb_proplist;
	char headerbuf[ZPOOL_MAXPROPLEN];
	const char *header;
	boolean_t first = B_TRUE;
	boolean_t right_justify;
	size_t width = 0;

	for (; pl != NULL; pl = pl->pl_next) {
	width = pl->pl_width;
	if (first && cb->cb_verbose) {
	/*
	* Reset the width to accommodate the verbose listing
	* of devices.
	*/
	width = cb->cb_namewidth;
	}

	if (!first)
	(void) fputs(" ", stdout);
	else
	first = B_FALSE;

	right_justify = B_FALSE;
	if (pl->pl_prop != ZPROP_USERPROP) {
	header = zpool_prop_column_name(pl->pl_prop);
	right_justify = zpool_prop_align_right(pl->pl_prop);
	} else {
	int i;

	for (i = 0; pl->pl_user_prop[i] != '\0'; i++)
	headerbuf[i] = toupper(pl->pl_user_prop[i]);
	headerbuf[i] = '\0';
	header = headerbuf;
	}

	if (pl->pl_next == NULL && !right_justify)
	(void) fputs(header, stdout);
	else if (right_justify)
	(void) printf("%*s", (int)width, header);
	else
	(void) printf("%-*s", (int)width, header);
	}

	(void) fputc('\n', stdout);
	}

	/*
	* Given a pool and a list of properties, print out all the properties according
	* to the described layout. Used by zpool_do_list().
	*/
	static void
	print_pool(zpool_handle_t zhp, list_cbdata_t cb)
	{
	zprop_list_t *pl = cb->cb_proplist;
	boolean_t first = B_TRUE;
	char property[ZPOOL_MAXPROPLEN];
	const char *propstr;
	boolean_t right_justify;
	size_t width;

	for (; pl != NULL; pl = pl->pl_next) {

	width = pl->pl_width;
	if (first && cb->cb_verbose) {
	/*
	* Reset the width to accommodate the verbose listing
	* of devices.
	*/
	width = cb->cb_namewidth;
	}

	if (!first) {
	if (cb->cb_scripted)
	(void) fputc('\t', stdout);
	else
	(void) fputs(" ", stdout);
	} else {
	first = B_FALSE;
	}

	right_justify = B_FALSE;
	if (pl->pl_prop != ZPROP_USERPROP) {
	if (zpool_get_prop(zhp, pl->pl_prop, property,
	sizeof (property), NULL, cb->cb_literal) != 0)
	propstr = "-";
	else
	propstr = property;

	right_justify = zpool_prop_align_right(pl->pl_prop);
	} else if ((zpool_prop_feature(pl->pl_user_prop) \|\|
	zpool_prop_unsupported(pl->pl_user_prop)) &&
	zpool_prop_get_feature(zhp, pl->pl_user_prop, property,
	sizeof (property)) == 0) {
	propstr = property;
	} else if (zfs_prop_user(pl->pl_user_prop) &&
	zpool_get_userprop(zhp, pl->pl_user_prop, property,
	sizeof (property), NULL) == 0) {
	propstr = property;
	} else {
	propstr = "-";
	}

	/*
	* If this is being called in scripted mode, or if this is the
	* last column and it is left-justified, don't include a width
	* format specifier.
	*/
	if (cb->cb_scripted \|\| (pl->pl_next == NULL && !right_justify))
	(void) fputs(propstr, stdout);
	else if (right_justify)
	(void) printf("%*s", (int)width, propstr);
	else
	(void) printf("%-*s", (int)width, propstr);
	}

	(void) fputc('\n', stdout);
	}

	static void
	print_one_column(zpool_prop_t prop, uint64_t value, const char *str,
	boolean_t scripted, boolean_t valid, enum zfs_nicenum_format format)
	{
	char propval[64];
	boolean_t fixed;
	size_t width = zprop_width(prop, &fixed, ZFS_TYPE_POOL);

	switch (prop) {
	case ZPOOL_PROP_SIZE:
	case ZPOOL_PROP_EXPANDSZ:
	case ZPOOL_PROP_CHECKPOINT:
	case ZPOOL_PROP_DEDUPRATIO:
	if (value == 0)
	(void) strlcpy(propval, "-", sizeof (propval));
	else
	zfs_nicenum_format(value, propval, sizeof (propval),
	format);
	break;
	case ZPOOL_PROP_FRAGMENTATION:
	if (value == ZFS_FRAG_INVALID) {
	(void) strlcpy(propval, "-", sizeof (propval));
	} else if (format == ZFS_NICENUM_RAW) {
	(void) snprintf(propval, sizeof (propval), "%llu",
	(unsigned long long)value);
	} else {
	(void) snprintf(propval, sizeof (propval), "%llu%%",
	(unsigned long long)value);
	}
	break;
	case ZPOOL_PROP_CAPACITY:
	/* capacity value is in parts-per-10,000 (aka permyriad) */
	if (format == ZFS_NICENUM_RAW)
	(void) snprintf(propval, sizeof (propval), "%llu",
	(unsigned long long)value / 100);
	else
	(void) snprintf(propval, sizeof (propval),
	value < 1000 ? "%1.2f%%" : value < 10000 ?
	"%2.1f%%" : "%3.0f%%", value / 100.0);
	break;
	case ZPOOL_PROP_HEALTH:
	width = 8;
	(void) strlcpy(propval, str, sizeof (propval));
	break;
	default:
	zfs_nicenum_format(value, propval, sizeof (propval), format);
	}

	if (!valid)
	(void) strlcpy(propval, "-", sizeof (propval));

	if (scripted)
	(void) printf("\t%s", propval);
	else
	(void) printf(" %*s", (int)width, propval);
	}

	/*
	* print static default line per vdev
	* not compatible with '-o' <proplist> option
	*/
	static void
	print_list_stats(zpool_handle_t zhp, const char name, nvlist_t *nv,
	list_cbdata_t *cb, int depth, boolean_t isspare)
	{
	nvlist_t **child;
	vdev_stat_t *vs;
	uint_t c, children;
	char *vname;
	boolean_t scripted = cb->cb_scripted;
	uint64_t islog = B_FALSE;
	const char dashes = "%-s - - - - "
	"- - - - -\n";

	verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
	(uint64_t **)&vs, &c) == 0);

	if (name != NULL) {
	boolean_t toplevel = (vs->vs_space != 0);
	uint64_t cap;
	enum zfs_nicenum_format format;
	const char *state;

	if (cb->cb_literal)
	format = ZFS_NICENUM_RAW;
	else
	format = ZFS_NICENUM_1024;

	if (strcmp(name, VDEV_TYPE_INDIRECT) == 0)
	return;

	if (scripted)
	(void) printf("\t%s", name);
	else if (strlen(name) + depth > cb->cb_namewidth)
	(void) printf("%*s%s", depth, "", name);
	else
	(void) printf("%s%s%s", depth, "", name,
	(int)(cb->cb_namewidth - strlen(name) - depth), "");

	/*
	* Print the properties for the individual vdevs. Some
	* properties are only applicable to toplevel vdevs. The
	* 'toplevel' boolean value is passed to the print_one_column()
	* to indicate that the value is valid.
	*/
	if (VDEV_STAT_VALID(vs_pspace, c) && vs->vs_pspace)
	print_one_column(ZPOOL_PROP_SIZE, vs->vs_pspace, NULL,
	scripted, B_TRUE, format);
	else
	print_one_column(ZPOOL_PROP_SIZE, vs->vs_space, NULL,
	scripted, toplevel, format);
	print_one_column(ZPOOL_PROP_ALLOCATED, vs->vs_alloc, NULL,
	scripted, toplevel, format);
	print_one_column(ZPOOL_PROP_FREE, vs->vs_space - vs->vs_alloc,
	NULL, scripted, toplevel, format);
	print_one_column(ZPOOL_PROP_CHECKPOINT,
	vs->vs_checkpoint_space, NULL, scripted, toplevel, format);
	print_one_column(ZPOOL_PROP_EXPANDSZ, vs->vs_esize, NULL,
	scripted, B_TRUE, format);
	print_one_column(ZPOOL_PROP_FRAGMENTATION,
	vs->vs_fragmentation, NULL, scripted,
	(vs->vs_fragmentation != ZFS_FRAG_INVALID && toplevel),
	format);
	cap = (vs->vs_space == 0) ? 0 :
	(vs->vs_alloc * 10000 / vs->vs_space);
	print_one_column(ZPOOL_PROP_CAPACITY, cap, NULL,
	scripted, toplevel, format);
	print_one_column(ZPOOL_PROP_DEDUPRATIO, 0, NULL,
	scripted, toplevel, format);
	state = zpool_state_to_name(vs->vs_state, vs->vs_aux);
	if (isspare) {
	if (vs->vs_aux == VDEV_AUX_SPARED)
	state = "INUSE";
	else if (vs->vs_state == VDEV_STATE_HEALTHY)
	state = "AVAIL";
	}
	print_one_column(ZPOOL_PROP_HEALTH, 0, state, scripted,
	B_TRUE, format);
	(void) fputc('\n', stdout);
	}

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	&child, &children) != 0)
	return;

	/* list the normal vdevs first */
	for (c = 0; c < children; c++) {
	uint64_t ishole = B_FALSE;

	if (nvlist_lookup_uint64(child[c],
	ZPOOL_CONFIG_IS_HOLE, &ishole) == 0 && ishole)
	continue;

	if (nvlist_lookup_uint64(child[c],
	ZPOOL_CONFIG_IS_LOG, &islog) == 0 && islog)
	continue;

	if (nvlist_exists(child[c], ZPOOL_CONFIG_ALLOCATION_BIAS))
	continue;

	vname = zpool_vdev_name(g_zfs, zhp, child[c],
	cb->cb_name_flags \| VDEV_NAME_TYPE_ID);
	print_list_stats(zhp, vname, child[c], cb, depth + 2, B_FALSE);
	free(vname);
	}

	/* list the classes: 'logs', 'dedup', and 'special' */
	for (uint_t n = 0; n < ARRAY_SIZE(class_name); n++) {
	boolean_t printed = B_FALSE;

	for (c = 0; c < children; c++) {
	const char *bias = NULL;
	const char *type = NULL;

	if (nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
	&islog) == 0 && islog) {
	bias = VDEV_ALLOC_CLASS_LOGS;
	} else {
	(void) nvlist_lookup_string(child[c],
	ZPOOL_CONFIG_ALLOCATION_BIAS, &bias);
	(void) nvlist_lookup_string(child[c],
	ZPOOL_CONFIG_TYPE, &type);
	}
	if (bias == NULL \|\| strcmp(bias, class_name[n]) != 0)
	continue;
	if (!islog && strcmp(type, VDEV_TYPE_INDIRECT) == 0)
	continue;

	if (!printed) {
	/* LINTED E_SEC_PRINTF_VAR_FMT */
	(void) printf(dashes, cb->cb_namewidth,
	class_name[n]);
	printed = B_TRUE;
	}
	vname = zpool_vdev_name(g_zfs, zhp, child[c],
	cb->cb_name_flags \| VDEV_NAME_TYPE_ID);
	print_list_stats(zhp, vname, child[c], cb, depth + 2,
	B_FALSE);
	free(vname);
	}
	}

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
	&child, &children) == 0 && children > 0) {
	/* LINTED E_SEC_PRINTF_VAR_FMT */
	(void) printf(dashes, cb->cb_namewidth, "cache");
	for (c = 0; c < children; c++) {
	vname = zpool_vdev_name(g_zfs, zhp, child[c],
	cb->cb_name_flags);
	print_list_stats(zhp, vname, child[c], cb, depth + 2,
	B_FALSE);
	free(vname);
	}
	}

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES, &child,
	&children) == 0 && children > 0) {
	/* LINTED E_SEC_PRINTF_VAR_FMT */
	(void) printf(dashes, cb->cb_namewidth, "spare");
	for (c = 0; c < children; c++) {
	vname = zpool_vdev_name(g_zfs, zhp, child[c],
	cb->cb_name_flags);
	print_list_stats(zhp, vname, child[c], cb, depth + 2,
	B_TRUE);
	free(vname);
	}
	}
	}

	/*
	* Generic callback function to list a pool.
	*/
	static int
	list_callback(zpool_handle_t zhp, void data)
	{
	list_cbdata_t *cbp = data;

	print_pool(zhp, cbp);

	if (cbp->cb_verbose) {
	nvlist_t config, nvroot;

	config = zpool_get_config(zhp, NULL);
	verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
	&nvroot) == 0);
	print_list_stats(zhp, NULL, nvroot, cbp, 0, B_FALSE);
	}

	return (0);
	}

	/*
	* Set the minimum pool/vdev name column width. The width must be at least 9,
	* but may be as large as needed.
	*/
	static int
	get_namewidth_list(zpool_handle_t zhp, void data)
	{
	list_cbdata_t *cb = data;
	int width;

	width = get_namewidth(zhp, cb->cb_namewidth,
	cb->cb_name_flags \| VDEV_NAME_TYPE_ID, cb->cb_verbose);

	if (width < 9)
	width = 9;

	cb->cb_namewidth = width;

	return (0);
	}

	/*
	* zpool list [-gHLpP] [-o prop[,prop]*] [-T d\|u] [pool] ... [interval [count]]
	*
	* -g Display guid for individual vdev name.
	* -H Scripted mode. Don't display headers, and separate properties
	* by a single tab.
	* -L Follow links when resolving vdev path name.
	* -o List of properties to display. Defaults to
	* "name,size,allocated,free,expandsize,fragmentation,capacity,"
	* "dedupratio,health,altroot"
	* -p Display values in parsable (exact) format.
	* -P Display full path for vdev name.
	* -T Display a timestamp in date(1) or Unix format
	*
	* List all pools in the system, whether or not they're healthy. Output space
	* statistics for each one, as well as health status summary.
	*/
	int
	zpool_do_list(int argc, char **argv)
	{
	int c;
	int ret = 0;
	list_cbdata_t cb = { 0 };
	static char default_props[] =
	"name,size,allocated,free,checkpoint,expandsize,fragmentation,"
	"capacity,dedupratio,health,altroot";
	char *props = default_props;
	float interval = 0;
	unsigned long count = 0;
	zpool_list_t *list;
	boolean_t first = B_TRUE;
	current_prop_type = ZFS_TYPE_POOL;

	/* check options */
	while ((c = getopt(argc, argv, ":gHLo:pPT:v")) != -1) {
	switch (c) {
	case 'g':
	cb.cb_name_flags \|= VDEV_NAME_GUID;
	break;
	case 'H':
	cb.cb_scripted = B_TRUE;
	break;
	case 'L':
	cb.cb_name_flags \|= VDEV_NAME_FOLLOW_LINKS;
	break;
	case 'o':
	props = optarg;
	break;
	case 'P':
	cb.cb_name_flags \|= VDEV_NAME_PATH;
	break;
	case 'p':
	cb.cb_literal = B_TRUE;
	break;
	case 'T':
	get_timestamp_arg(*optarg);
	break;
	case 'v':
	cb.cb_verbose = B_TRUE;
	cb.cb_namewidth = 8; /* 8 until precalc is avail */
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	usage(B_FALSE);
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	get_interval_count(&argc, argv, &interval, &count);

	if (zprop_get_list(g_zfs, props, &cb.cb_proplist, ZFS_TYPE_POOL) != 0)
	usage(B_FALSE);

	for (;;) {
	if ((list = pool_list_get(argc, argv, &cb.cb_proplist,
	ZFS_TYPE_POOL, cb.cb_literal, &ret)) == NULL)
	return (1);

	if (pool_list_count(list) == 0)
	break;

	cb.cb_namewidth = 0;
	(void) pool_list_iter(list, B_FALSE, get_namewidth_list, &cb);

	if (timestamp_fmt != NODATE)
	print_timestamp(timestamp_fmt);

	if (!cb.cb_scripted && (first \|\| cb.cb_verbose)) {
	print_header(&cb);
	first = B_FALSE;
	}
	ret = pool_list_iter(list, B_TRUE, list_callback, &cb);

	if (interval == 0)
	break;

	if (count != 0 && --count == 0)
	break;

	pool_list_free(list);

	(void) fflush(stdout);
	(void) fsleep(interval);
	}

	if (argc == 0 && !cb.cb_scripted && pool_list_count(list) == 0) {
	(void) printf(gettext("no pools available\n"));
	ret = 0;
	}

	pool_list_free(list);
	zprop_free_list(cb.cb_proplist);
	return (ret);
	}

	static int
	zpool_do_attach_or_replace(int argc, char **argv, int replacing)
	{
	boolean_t force = B_FALSE;
	boolean_t rebuild = B_FALSE;
	boolean_t wait = B_FALSE;
	int c;
	nvlist_t *nvroot;
	char poolname, old_disk, *new_disk;
	zpool_handle_t *zhp;
	nvlist_t *props = NULL;
	char *propval;
	int ret;

	/* check options */
	while ((c = getopt(argc, argv, "fo:sw")) != -1) {
	switch (c) {
	case 'f':
	force = B_TRUE;
	break;
	case 'o':
	if ((propval = strchr(optarg, '=')) == NULL) {
	(void) fprintf(stderr, gettext("missing "
	"'=' for -o option\n"));
	usage(B_FALSE);
	}
	*propval = '\0';
	propval++;

	if ((strcmp(optarg, ZPOOL_CONFIG_ASHIFT) != 0) \|\|
	(add_prop_list(optarg, propval, &props, B_TRUE)))
	usage(B_FALSE);
	break;
	case 's':
	rebuild = B_TRUE;
	break;
	case 'w':
	wait = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* get pool name and check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name argument\n"));
	usage(B_FALSE);
	}

	poolname = argv[0];

	if (argc < 2) {
	(void) fprintf(stderr,
	gettext("missing <device> specification\n"));
	usage(B_FALSE);
	}

	old_disk = argv[1];

	if (argc < 3) {
	if (!replacing) {
	(void) fprintf(stderr,
	gettext("missing <new_device> specification\n"));
	usage(B_FALSE);
	}
	new_disk = old_disk;
	argc -= 1;
	argv += 1;
	} else {
	new_disk = argv[2];
	argc -= 2;
	argv += 2;
	}

	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	if ((zhp = zpool_open(g_zfs, poolname)) == NULL) {
	nvlist_free(props);
	return (1);
	}

	if (zpool_get_config(zhp, NULL) == NULL) {
	(void) fprintf(stderr, gettext("pool '%s' is unavailable\n"),
	poolname);
	zpool_close(zhp);
	nvlist_free(props);
	return (1);
	}

	/* unless manually specified use "ashift" pool property (if set) */
	if (!nvlist_exists(props, ZPOOL_CONFIG_ASHIFT)) {
	int intval;
	zprop_source_t src;
	char strval[ZPOOL_MAXPROPLEN];

	intval = zpool_get_prop_int(zhp, ZPOOL_PROP_ASHIFT, &src);
	if (src != ZPROP_SRC_DEFAULT) {
	(void) sprintf(strval, "%" PRId32, intval);
	verify(add_prop_list(ZPOOL_CONFIG_ASHIFT, strval,
	&props, B_TRUE) == 0);
	}
	}

	nvroot = make_root_vdev(zhp, props, force, B_FALSE, replacing, B_FALSE,
	argc, argv);
	if (nvroot == NULL) {
	zpool_close(zhp);
	nvlist_free(props);
	return (1);
	}

	ret = zpool_vdev_attach(zhp, old_disk, new_disk, nvroot, replacing,
	rebuild);

	if (ret == 0 && wait) {
	zpool_wait_activity_t activity = ZPOOL_WAIT_RESILVER;
	char raidz_prefix[] = "raidz";
	if (replacing) {
	activity = ZPOOL_WAIT_REPLACE;
	} else if (strncmp(old_disk,
	raidz_prefix, strlen(raidz_prefix)) == 0) {
	activity = ZPOOL_WAIT_RAIDZ_EXPAND;
	}
	ret = zpool_wait(zhp, activity);
	}

	nvlist_free(props);
	nvlist_free(nvroot);
	zpool_close(zhp);

	return (ret);
	}

	/*
	* zpool replace [-fsw] [-o property=value] <pool> <device> <new_device>
	*
	* -f Force attach, even if <new_device> appears to be in use.
	* -s Use sequential instead of healing reconstruction for resilver.
	* -o Set property=value.
	* -w Wait for replacing to complete before returning
	*
	* Replace <device> with <new_device>.
	*/
	int
	zpool_do_replace(int argc, char **argv)
	{
	return (zpool_do_attach_or_replace(argc, argv, B_TRUE));
	}

	/*
	* zpool attach [-fsw] [-o property=value] <pool> <device>\|<vdev> <new_device>
	*
	* -f Force attach, even if <new_device> appears to be in use.
	* -s Use sequential instead of healing reconstruction for resilver.
	* -o Set property=value.
	* -w Wait for resilvering (mirror) or expansion (raidz) to complete
	* before returning.
	*
	* Attach <new_device> to a <device> or <vdev>, where the vdev can be of type
	* mirror or raidz. If <device> is not part of a mirror, then <device> will
	* be transformed into a mirror of <device> and <new_device>. When a mirror
	* is involved, <new_device> will begin life with a DTL of [0, now], and will
	* immediately begin to resilver itself. For the raidz case, a expansion will
	* commence and reflow the raidz data across all the disks including the
	* <new_device>.
	*/
	int
	zpool_do_attach(int argc, char **argv)
	{
	return (zpool_do_attach_or_replace(argc, argv, B_FALSE));
	}

	/*
	* zpool detach [-f] <pool> <device>
	*
	* -f Force detach of <device>, even if DTLs argue against it
	* (not supported yet)
	*
	* Detach a device from a mirror. The operation will be refused if <device>
	* is the last device in the mirror, or if the DTLs indicate that this device
	* has the only valid copy of some data.
	*/
	int
	zpool_do_detach(int argc, char **argv)
	{
	int c;
	char poolname, path;
	zpool_handle_t *zhp;
	int ret;

	/* check options */
	while ((c = getopt(argc, argv, "")) != -1) {
	switch (c) {
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* get pool name and check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name argument\n"));
	usage(B_FALSE);
	}

	if (argc < 2) {
	(void) fprintf(stderr,
	gettext("missing <device> specification\n"));
	usage(B_FALSE);
	}

	poolname = argv[0];
	path = argv[1];

	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
	return (1);

	ret = zpool_vdev_detach(zhp, path);

	zpool_close(zhp);

	return (ret);
	}

	/*
	* zpool split [-gLnP] [-o prop=val] ...
	* [-o mntopt] ...
	* [-R altroot] <pool> <newpool> [<device> ...]
	*
	* -g Display guid for individual vdev name.
	* -L Follow links when resolving vdev path name.
	* -n Do not split the pool, but display the resulting layout if
	* it were to be split.
	* -o Set property=value, or set mount options.
	* -P Display full path for vdev name.
	* -R Mount the split-off pool under an alternate root.
	* -l Load encryption keys while importing.
	*
	* Splits the named pool and gives it the new pool name. Devices to be split
	* off may be listed, provided that no more than one device is specified
	* per top-level vdev mirror. The newly split pool is left in an exported
	* state unless -R is specified.
	*
	* Restrictions: the top-level of the pool pool must only be made up of
	* mirrors; all devices in the pool must be healthy; no device may be
	* undergoing a resilvering operation.
	*/
	int
	zpool_do_split(int argc, char **argv)
	{
	char srcpool, newpool, *propval;
	char *mntopts = NULL;
	splitflags_t flags;
	int c, ret = 0;
	int ms_status = 0;
	boolean_t loadkeys = B_FALSE;
	zpool_handle_t *zhp;
	nvlist_t config, props = NULL;

	flags.dryrun = B_FALSE;
	flags.import = B_FALSE;
	flags.name_flags = 0;

	/* check options */
	while ((c = getopt(argc, argv, ":gLR:lno:P")) != -1) {
	switch (c) {
	case 'g':
	flags.name_flags \|= VDEV_NAME_GUID;
	break;
	case 'L':
	flags.name_flags \|= VDEV_NAME_FOLLOW_LINKS;
	break;
	case 'R':
	flags.import = B_TRUE;
	if (add_prop_list(
	zpool_prop_to_name(ZPOOL_PROP_ALTROOT), optarg,
	&props, B_TRUE) != 0) {
	nvlist_free(props);
	usage(B_FALSE);
	}
	break;
	case 'l':
	loadkeys = B_TRUE;
	break;
	case 'n':
	flags.dryrun = B_TRUE;
	break;
	case 'o':
	if ((propval = strchr(optarg, '=')) != NULL) {
	*propval = '\0';
	propval++;
	if (add_prop_list(optarg, propval,
	&props, B_TRUE) != 0) {
	nvlist_free(props);
	usage(B_FALSE);
	}
	} else {
	mntopts = optarg;
	}
	break;
	case 'P':
	flags.name_flags \|= VDEV_NAME_PATH;
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	usage(B_FALSE);
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	break;
	}
	}

	if (!flags.import && mntopts != NULL) {
	(void) fprintf(stderr, gettext("setting mntopts is only "
	"valid when importing the pool\n"));
	usage(B_FALSE);
	}

	if (!flags.import && loadkeys) {
	(void) fprintf(stderr, gettext("loading keys is only "
	"valid when importing the pool\n"));
	usage(B_FALSE);
	}

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("Missing pool name\n"));
	usage(B_FALSE);
	}
	if (argc < 2) {
	(void) fprintf(stderr, gettext("Missing new pool name\n"));
	usage(B_FALSE);
	}

	srcpool = argv[0];
	newpool = argv[1];

	argc -= 2;
	argv += 2;

	if ((zhp = zpool_open(g_zfs, srcpool)) == NULL) {
	nvlist_free(props);
	return (1);
	}

	config = split_mirror_vdev(zhp, newpool, props, flags, argc, argv);
	if (config == NULL) {
	ret = 1;
	} else {
	if (flags.dryrun) {
	(void) printf(gettext("would create '%s' with the "
	"following layout:\n\n"), newpool);
	print_vdev_tree(NULL, newpool, config, 0, "",
	flags.name_flags);
	print_vdev_tree(NULL, "dedup", config, 0,
	VDEV_ALLOC_BIAS_DEDUP, 0);
	print_vdev_tree(NULL, "special", config, 0,
	VDEV_ALLOC_BIAS_SPECIAL, 0);
	}
	}

	zpool_close(zhp);

	if (ret != 0 \|\| flags.dryrun \|\| !flags.import) {
	nvlist_free(config);
	nvlist_free(props);
	return (ret);
	}

	/*
	* The split was successful. Now we need to open the new
	* pool and import it.
	*/
	if ((zhp = zpool_open_canfail(g_zfs, newpool)) == NULL) {
	nvlist_free(config);
	nvlist_free(props);
	return (1);
	}

	if (loadkeys) {
	ret = zfs_crypto_attempt_load_keys(g_zfs, newpool);
	if (ret != 0)
	ret = 1;
	}

	if (zpool_get_state(zhp) != POOL_STATE_UNAVAIL) {
	- ms_status = zpool_enable_datasets(zhp, mntopts, 0);
	+ ms_status = zpool_enable_datasets(zhp, mntopts, 0,
	+ mount_tp_nthr);
	if (ms_status == EZFS_SHAREFAILED) {
	(void) fprintf(stderr, gettext("Split was successful, "
	"datasets are mounted but sharing of some datasets "
	"has failed\n"));
	} else if (ms_status == EZFS_MOUNTFAILED) {
	(void) fprintf(stderr, gettext("Split was successful"
	", but some datasets could not be mounted\n"));
	(void) fprintf(stderr, gettext("Try doing '%s' with a "
	"different altroot\n"), "zpool import");
	}
	}
	zpool_close(zhp);
	nvlist_free(config);
	nvlist_free(props);

	return (ret);
	}


	/*
	* zpool online [--power] <pool> <device> ...
	*
	* --power: Power on the enclosure slot to the drive (if possible)
	*/
	int
	zpool_do_online(int argc, char **argv)
	{
	int c, i;
	char *poolname;
	zpool_handle_t *zhp;
	int ret = 0;
	vdev_state_t newstate;
	int flags = 0;
	boolean_t is_power_on = B_FALSE;
	struct option long_options[] = {
	{"power", no_argument, NULL, ZPOOL_OPTION_POWER},
	{0, 0, 0, 0}
	};

	/* check options */
	while ((c = getopt_long(argc, argv, "e", long_options, NULL)) != -1) {
	switch (c) {
	case 'e':
	flags \|= ZFS_ONLINE_EXPAND;
	break;
	case ZPOOL_OPTION_POWER:
	is_power_on = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	if (libzfs_envvar_is_set("ZPOOL_AUTO_POWER_ON_SLOT"))
	is_power_on = B_TRUE;

	argc -= optind;
	argv += optind;

	/* get pool name and check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name\n"));
	usage(B_FALSE);
	}
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing device name\n"));
	usage(B_FALSE);
	}

	poolname = argv[0];

	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
	return (1);

	for (i = 1; i < argc; i++) {
	vdev_state_t oldstate;
	boolean_t avail_spare, l2cache;
	int rc;

	if (is_power_on) {
	rc = zpool_power_on_and_disk_wait(zhp, argv[i]);
	if (rc == ENOTSUP) {
	(void) fprintf(stderr,
	gettext("Power control not supported\n"));
	}
	if (rc != 0)
	return (rc);
	}

	nvlist_t *tgt = zpool_find_vdev(zhp, argv[i], &avail_spare,
	&l2cache, NULL);
	if (tgt == NULL) {
	ret = 1;
	continue;
	}
	uint_t vsc;
	oldstate = ((vdev_stat_t *)fnvlist_lookup_uint64_array(tgt,
	ZPOOL_CONFIG_VDEV_STATS, &vsc))->vs_state;
	if (zpool_vdev_online(zhp, argv[i], flags, &newstate) == 0) {
	if (newstate != VDEV_STATE_HEALTHY) {
	(void) printf(gettext("warning: device '%s' "
	"onlined, but remains in faulted state\n"),
	argv[i]);
	if (newstate == VDEV_STATE_FAULTED)
	(void) printf(gettext("use 'zpool "
	"clear' to restore a faulted "
	"device\n"));
	else
	(void) printf(gettext("use 'zpool "
	"replace' to replace devices "
	"that are no longer present\n"));
	if ((flags & ZFS_ONLINE_EXPAND)) {
	(void) printf(gettext("%s: failed "
	"to expand usable space on "
	"unhealthy device '%s'\n"),
	(oldstate >= VDEV_STATE_DEGRADED ?
	"error" : "warning"), argv[i]);
	if (oldstate >= VDEV_STATE_DEGRADED) {
	ret = 1;
	break;
	}
	}
	}
	} else {
	ret = 1;
	}
	}

	zpool_close(zhp);

	return (ret);
	}

	/*
	* zpool offline [-ft]\|[--power] <pool> <device> ...
	*
	*
	* -f Force the device into a faulted state.
	*
	* -t Only take the device off-line temporarily. The offline/faulted
	* state will not be persistent across reboots.
	*
	* --power Power off the enclosure slot to the drive (if possible)
	*/
	int
	zpool_do_offline(int argc, char **argv)
	{
	int c, i;
	char *poolname;
	zpool_handle_t *zhp;
	int ret = 0;
	boolean_t istmp = B_FALSE;
	boolean_t fault = B_FALSE;
	boolean_t is_power_off = B_FALSE;

	struct option long_options[] = {
	{"power", no_argument, NULL, ZPOOL_OPTION_POWER},
	{0, 0, 0, 0}
	};

	/* check options */
	while ((c = getopt_long(argc, argv, "ft", long_options, NULL)) != -1) {
	switch (c) {
	case 'f':
	fault = B_TRUE;
	break;
	case 't':
	istmp = B_TRUE;
	break;
	case ZPOOL_OPTION_POWER:
	is_power_off = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	if (is_power_off && fault) {
	(void) fprintf(stderr,
	gettext("-0 and -f cannot be used together\n"));
	usage(B_FALSE);
	return (1);
	}

	if (is_power_off && istmp) {
	(void) fprintf(stderr,
	gettext("-0 and -t cannot be used together\n"));
	usage(B_FALSE);
	return (1);
	}

	argc -= optind;
	argv += optind;

	/* get pool name and check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name\n"));
	usage(B_FALSE);
	}
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing device name\n"));
	usage(B_FALSE);
	}

	poolname = argv[0];

	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
	return (1);

	for (i = 1; i < argc; i++) {
	uint64_t guid = zpool_vdev_path_to_guid(zhp, argv[i]);
	if (is_power_off) {
	/*
	* Note: we have to power off first, then set REMOVED,
	* or else zpool_vdev_set_removed_state() returns
	* EAGAIN.
	*/
	ret = zpool_power_off(zhp, argv[i]);
	if (ret != 0) {
	(void) fprintf(stderr, "%s %s %d\n",
	gettext("unable to power off slot for"),
	argv[i], ret);
	}
	zpool_vdev_set_removed_state(zhp, guid, VDEV_AUX_NONE);

	} else if (fault) {
	vdev_aux_t aux;
	if (istmp == B_FALSE) {
	/* Force the fault to persist across imports */
	aux = VDEV_AUX_EXTERNAL_PERSIST;
	} else {
	aux = VDEV_AUX_EXTERNAL;
	}

	if (guid == 0 \|\| zpool_vdev_fault(zhp, guid, aux) != 0)
	ret = 1;
	} else {
	if (zpool_vdev_offline(zhp, argv[i], istmp) != 0)
	ret = 1;
	}
	}

	zpool_close(zhp);

	return (ret);
	}

	/*
	* zpool clear [-nF]\|[--power] <pool> [device]
	*
	* Clear all errors associated with a pool or a particular device.
	*/
	int
	zpool_do_clear(int argc, char **argv)
	{
	int c;
	int ret = 0;
	boolean_t dryrun = B_FALSE;
	boolean_t do_rewind = B_FALSE;
	boolean_t xtreme_rewind = B_FALSE;
	boolean_t is_power_on = B_FALSE;
	uint32_t rewind_policy = ZPOOL_NO_REWIND;
	nvlist_t *policy = NULL;
	zpool_handle_t *zhp;
	char pool, device;

	struct option long_options[] = {
	{"power", no_argument, NULL, ZPOOL_OPTION_POWER},
	{0, 0, 0, 0}
	};

	/* check options */
	while ((c = getopt_long(argc, argv, "FnX", long_options,
	NULL)) != -1) {
	switch (c) {
	case 'F':
	do_rewind = B_TRUE;
	break;
	case 'n':
	dryrun = B_TRUE;
	break;
	case 'X':
	xtreme_rewind = B_TRUE;
	break;
	case ZPOOL_OPTION_POWER:
	is_power_on = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	if (libzfs_envvar_is_set("ZPOOL_AUTO_POWER_ON_SLOT"))
	is_power_on = B_TRUE;

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name\n"));
	usage(B_FALSE);
	}

	if (argc > 2) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	if ((dryrun \|\| xtreme_rewind) && !do_rewind) {
	(void) fprintf(stderr,
	gettext("-n or -X only meaningful with -F\n"));
	usage(B_FALSE);
	}
	if (dryrun)
	rewind_policy = ZPOOL_TRY_REWIND;
	else if (do_rewind)
	rewind_policy = ZPOOL_DO_REWIND;
	if (xtreme_rewind)
	rewind_policy \|= ZPOOL_EXTREME_REWIND;

	/* In future, further rewind policy choices can be passed along here */
	if (nvlist_alloc(&policy, NV_UNIQUE_NAME, 0) != 0 \|\|
	nvlist_add_uint32(policy, ZPOOL_LOAD_REWIND_POLICY,
	rewind_policy) != 0) {
	return (1);
	}

	pool = argv[0];
	device = argc == 2 ? argv[1] : NULL;

	if ((zhp = zpool_open_canfail(g_zfs, pool)) == NULL) {
	nvlist_free(policy);
	return (1);
	}

	if (is_power_on) {
	if (device == NULL) {
	zpool_power_on_pool_and_wait_for_devices(zhp);
	} else {
	zpool_power_on_and_disk_wait(zhp, device);
	}
	}

	if (zpool_clear(zhp, device, policy) != 0)
	ret = 1;

	zpool_close(zhp);

	nvlist_free(policy);

	return (ret);
	}

	/*
	* zpool reguid <pool>
	*/
	int
	zpool_do_reguid(int argc, char **argv)
	{
	int c;
	char *poolname;
	zpool_handle_t *zhp;
	int ret = 0;

	/* check options */
	while ((c = getopt(argc, argv, "")) != -1) {
	switch (c) {
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* get pool name and check number of arguments */
	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name\n"));
	usage(B_FALSE);
	}

	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	poolname = argv[0];
	if ((zhp = zpool_open(g_zfs, poolname)) == NULL)
	return (1);

	ret = zpool_reguid(zhp);

	zpool_close(zhp);
	return (ret);
	}


	/*
	* zpool reopen <pool>
	*
	* Reopen the pool so that the kernel can update the sizes of all vdevs.
	*/
	int
	zpool_do_reopen(int argc, char **argv)
	{
	int c;
	int ret = 0;
	boolean_t scrub_restart = B_TRUE;

	/* check options */
	while ((c = getopt(argc, argv, "n")) != -1) {
	switch (c) {
	case 'n':
	scrub_restart = B_FALSE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	/* if argc == 0 we will execute zpool_reopen_one on all pools */
	ret = for_each_pool(argc, argv, B_TRUE, NULL, ZFS_TYPE_POOL,
	B_FALSE, zpool_reopen_one, &scrub_restart);

	return (ret);
	}

	typedef struct scrub_cbdata {
	int cb_type;
	pool_scrub_cmd_t cb_scrub_cmd;
	} scrub_cbdata_t;

	static boolean_t
	zpool_has_checkpoint(zpool_handle_t *zhp)
	{
	nvlist_t config, nvroot;

	config = zpool_get_config(zhp, NULL);

	if (config != NULL) {
	pool_checkpoint_stat_t *pcs = NULL;
	uint_t c;

	nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);
	(void) nvlist_lookup_uint64_array(nvroot,
	ZPOOL_CONFIG_CHECKPOINT_STATS, (uint64_t **)&pcs, &c);

	if (pcs == NULL \|\| pcs->pcs_state == CS_NONE)
	return (B_FALSE);

	assert(pcs->pcs_state == CS_CHECKPOINT_EXISTS \|\|
	pcs->pcs_state == CS_CHECKPOINT_DISCARDING);
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	static int
	scrub_callback(zpool_handle_t zhp, void data)
	{
	scrub_cbdata_t *cb = data;
	int err;

	/*
	* Ignore faulted pools.
	*/
	if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
	(void) fprintf(stderr, gettext("cannot scan '%s': pool is "
	"currently unavailable\n"), zpool_get_name(zhp));
	return (1);
	}

	err = zpool_scan(zhp, cb->cb_type, cb->cb_scrub_cmd);

	if (err == 0 && zpool_has_checkpoint(zhp) &&
	cb->cb_type == POOL_SCAN_SCRUB) {
	(void) printf(gettext("warning: will not scrub state that "
	"belongs to the checkpoint of pool '%s'\n"),
	zpool_get_name(zhp));
	}

	return (err != 0);
	}

	static int
	wait_callback(zpool_handle_t zhp, void data)
	{
	zpool_wait_activity_t *act = data;
	return (zpool_wait(zhp, *act));
	}

	/*
	* zpool scrub [-s \| -p] [-w] [-e] <pool> ...
	*
	* -e Only scrub blocks in the error log.
	* -s Stop. Stops any in-progress scrub.
	* -p Pause. Pause in-progress scrub.
	* -w Wait. Blocks until scrub has completed.
	*/
	int
	zpool_do_scrub(int argc, char **argv)
	{
	int c;
	scrub_cbdata_t cb;
	boolean_t wait = B_FALSE;
	int error;

	cb.cb_type = POOL_SCAN_SCRUB;
	cb.cb_scrub_cmd = POOL_SCRUB_NORMAL;

	boolean_t is_error_scrub = B_FALSE;
	boolean_t is_pause = B_FALSE;
	boolean_t is_stop = B_FALSE;

	/* check options */
	while ((c = getopt(argc, argv, "spwe")) != -1) {
	switch (c) {
	case 'e':
	is_error_scrub = B_TRUE;
	break;
	case 's':
	is_stop = B_TRUE;
	break;
	case 'p':
	is_pause = B_TRUE;
	break;
	case 'w':
	wait = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	if (is_pause && is_stop) {
	(void) fprintf(stderr, gettext("invalid option "
	"combination :-s and -p are mutually exclusive\n"));
	usage(B_FALSE);
	} else {
	if (is_error_scrub)
	cb.cb_type = POOL_SCAN_ERRORSCRUB;

	if (is_pause) {
	cb.cb_scrub_cmd = POOL_SCRUB_PAUSE;
	} else if (is_stop) {
	cb.cb_type = POOL_SCAN_NONE;
	} else {
	cb.cb_scrub_cmd = POOL_SCRUB_NORMAL;
	}
	}

	if (wait && (cb.cb_type == POOL_SCAN_NONE \|\|
	cb.cb_scrub_cmd == POOL_SCRUB_PAUSE)) {
	(void) fprintf(stderr, gettext("invalid option combination: "
	"-w cannot be used with -p or -s\n"));
	usage(B_FALSE);
	}

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name argument\n"));
	usage(B_FALSE);
	}

	error = for_each_pool(argc, argv, B_TRUE, NULL, ZFS_TYPE_POOL,
	B_FALSE, scrub_callback, &cb);

	if (wait && !error) {
	zpool_wait_activity_t act = ZPOOL_WAIT_SCRUB;
	error = for_each_pool(argc, argv, B_TRUE, NULL, ZFS_TYPE_POOL,
	B_FALSE, wait_callback, &act);
	}

	return (error);
	}

	/*
	* zpool resilver <pool> ...
	*
	* Restarts any in-progress resilver
	*/
	int
	zpool_do_resilver(int argc, char **argv)
	{
	int c;
	scrub_cbdata_t cb;

	cb.cb_type = POOL_SCAN_RESILVER;
	cb.cb_scrub_cmd = POOL_SCRUB_NORMAL;

	/* check options */
	while ((c = getopt(argc, argv, "")) != -1) {
	switch (c) {
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name argument\n"));
	usage(B_FALSE);
	}

	return (for_each_pool(argc, argv, B_TRUE, NULL, ZFS_TYPE_POOL,
	B_FALSE, scrub_callback, &cb));
	}

	/*
	* zpool trim [-d] [-r <rate>] [-c \| -s] <pool> [<device> ...]
	*
	* -c Cancel. Ends any in-progress trim.
	* -d Secure trim. Requires kernel and device support.
	* -r <rate> Sets the TRIM rate in bytes (per second). Supports
	* adding a multiplier suffix such as 'k' or 'm'.
	* -s Suspend. TRIM can then be restarted with no flags.
	* -w Wait. Blocks until trimming has completed.
	*/
	int
	zpool_do_trim(int argc, char **argv)
	{
	struct option long_options[] = {
	{"cancel", no_argument, NULL, 'c'},
	{"secure", no_argument, NULL, 'd'},
	{"rate", required_argument, NULL, 'r'},
	{"suspend", no_argument, NULL, 's'},
	{"wait", no_argument, NULL, 'w'},
	{0, 0, 0, 0}
	};

	pool_trim_func_t cmd_type = POOL_TRIM_START;
	uint64_t rate = 0;
	boolean_t secure = B_FALSE;
	boolean_t wait = B_FALSE;

	int c;
	while ((c = getopt_long(argc, argv, "cdr:sw", long_options, NULL))
	!= -1) {
	switch (c) {
	case 'c':
	if (cmd_type != POOL_TRIM_START &&
	cmd_type != POOL_TRIM_CANCEL) {
	(void) fprintf(stderr, gettext("-c cannot be "
	"combined with other options\n"));
	usage(B_FALSE);
	}
	cmd_type = POOL_TRIM_CANCEL;
	break;
	case 'd':
	if (cmd_type != POOL_TRIM_START) {
	(void) fprintf(stderr, gettext("-d cannot be "
	"combined with the -c or -s options\n"));
	usage(B_FALSE);
	}
	secure = B_TRUE;
	break;
	case 'r':
	if (cmd_type != POOL_TRIM_START) {
	(void) fprintf(stderr, gettext("-r cannot be "
	"combined with the -c or -s options\n"));
	usage(B_FALSE);
	}
	if (zfs_nicestrtonum(g_zfs, optarg, &rate) == -1) {
	(void) fprintf(stderr, "%s: %s\n",
	gettext("invalid value for rate"),
	libzfs_error_description(g_zfs));
	usage(B_FALSE);
	}
	break;
	case 's':
	if (cmd_type != POOL_TRIM_START &&
	cmd_type != POOL_TRIM_SUSPEND) {
	(void) fprintf(stderr, gettext("-s cannot be "
	"combined with other options\n"));
	usage(B_FALSE);
	}
	cmd_type = POOL_TRIM_SUSPEND;
	break;
	case 'w':
	wait = B_TRUE;
	break;
	case '?':
	if (optopt != 0) {
	(void) fprintf(stderr,
	gettext("invalid option '%c'\n"), optopt);
	} else {
	(void) fprintf(stderr,
	gettext("invalid option '%s'\n"),
	argv[optind - 1]);
	}
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing pool name argument\n"));
	usage(B_FALSE);
	return (-1);
	}

	if (wait && (cmd_type != POOL_TRIM_START)) {
	(void) fprintf(stderr, gettext("-w cannot be used with -c or "
	"-s\n"));
	usage(B_FALSE);
	}

	char *poolname = argv[0];
	zpool_handle_t *zhp = zpool_open(g_zfs, poolname);
	if (zhp == NULL)
	return (-1);

	trimflags_t trim_flags = {
	.secure = secure,
	.rate = rate,
	.wait = wait,
	};

	nvlist_t *vdevs = fnvlist_alloc();
	if (argc == 1) {
	/* no individual leaf vdevs specified, so add them all */
	nvlist_t *config = zpool_get_config(zhp, NULL);
	nvlist_t *nvroot = fnvlist_lookup_nvlist(config,
	ZPOOL_CONFIG_VDEV_TREE);
	zpool_collect_leaves(zhp, nvroot, vdevs);
	trim_flags.fullpool = B_TRUE;
	} else {
	trim_flags.fullpool = B_FALSE;
	for (int i = 1; i < argc; i++) {
	fnvlist_add_boolean(vdevs, argv[i]);
	}
	}

	int error = zpool_trim(zhp, cmd_type, vdevs, &trim_flags);

	fnvlist_free(vdevs);
	zpool_close(zhp);

	return (error);
	}

	/*
	* Converts a total number of seconds to a human readable string broken
	* down in to days/hours/minutes/seconds.
	*/
	static void
	secs_to_dhms(uint64_t total, char *buf)
	{
	uint64_t days = total / 60 / 60 / 24;
	uint64_t hours = (total / 60 / 60) % 24;
	uint64_t mins = (total / 60) % 60;
	uint64_t secs = (total % 60);

	if (days > 0) {
	(void) sprintf(buf, "%llu days %02llu:%02llu:%02llu",
	(u_longlong_t)days, (u_longlong_t)hours,
	(u_longlong_t)mins, (u_longlong_t)secs);
	} else {
	(void) sprintf(buf, "%02llu:%02llu:%02llu",
	(u_longlong_t)hours, (u_longlong_t)mins,
	(u_longlong_t)secs);
	}
	}

	/*
	* Print out detailed error scrub status.
	*/
	static void
	print_err_scrub_status(pool_scan_stat_t *ps)
	{
	time_t start, end, pause;
	uint64_t total_secs_left;
	uint64_t secs_left, mins_left, hours_left, days_left;
	uint64_t examined, to_be_examined;

	if (ps == NULL \|\| ps->pss_error_scrub_func != POOL_SCAN_ERRORSCRUB) {
	return;
	}

	(void) printf(gettext(" scrub: "));

	start = ps->pss_error_scrub_start;
	end = ps->pss_error_scrub_end;
	pause = ps->pss_pass_error_scrub_pause;
	examined = ps->pss_error_scrub_examined;
	to_be_examined = ps->pss_error_scrub_to_be_examined;

	assert(ps->pss_error_scrub_func == POOL_SCAN_ERRORSCRUB);

	if (ps->pss_error_scrub_state == DSS_FINISHED) {
	total_secs_left = end - start;
	days_left = total_secs_left / 60 / 60 / 24;
	hours_left = (total_secs_left / 60 / 60) % 24;
	mins_left = (total_secs_left / 60) % 60;
	secs_left = (total_secs_left % 60);

	(void) printf(gettext("scrubbed %llu error blocks in %llu days "
	"%02llu:%02llu:%02llu on %s"), (u_longlong_t)examined,
	(u_longlong_t)days_left, (u_longlong_t)hours_left,
	(u_longlong_t)mins_left, (u_longlong_t)secs_left,
	ctime(&end));

	return;
	} else if (ps->pss_error_scrub_state == DSS_CANCELED) {
	(void) printf(gettext("error scrub canceled on %s"),
	ctime(&end));
	return;
	}
	assert(ps->pss_error_scrub_state == DSS_ERRORSCRUBBING);

	/* Error scrub is in progress. */
	if (pause == 0) {
	(void) printf(gettext("error scrub in progress since %s"),
	ctime(&start));
	} else {
	(void) printf(gettext("error scrub paused since %s"),
	ctime(&pause));
	(void) printf(gettext("\terror scrub started on %s"),
	ctime(&start));
	}

	double fraction_done = (double)examined / (to_be_examined + examined);
	(void) printf(gettext("\t%.2f%% done, issued I/O for %llu error"
	" blocks"), 100 * fraction_done, (u_longlong_t)examined);

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

	/*
	* Print out detailed scrub status.
	*/
	static void
	print_scan_scrub_resilver_status(pool_scan_stat_t *ps)
	{
	time_t start, end, pause;
	uint64_t pass_scanned, scanned, pass_issued, issued, total_s, total_i;
	uint64_t elapsed, scan_rate, issue_rate;
	double fraction_done;
	char processed_buf[7], scanned_buf[7], issued_buf[7], total_s_buf[7];
	char total_i_buf[7], srate_buf[7], irate_buf[7], time_buf[32];

	printf(" ");
	printf_color(ANSI_BOLD, gettext("scan:"));
	printf(" ");

	/* If there's never been a scan, there's not much to say. */
	if (ps == NULL \|\| ps->pss_func == POOL_SCAN_NONE \|\|
	ps->pss_func >= POOL_SCAN_FUNCS) {
	(void) printf(gettext("none requested\n"));
	return;
	}

	start = ps->pss_start_time;
	end = ps->pss_end_time;
	pause = ps->pss_pass_scrub_pause;

	zfs_nicebytes(ps->pss_processed, processed_buf, sizeof (processed_buf));

	int is_resilver = ps->pss_func == POOL_SCAN_RESILVER;
	int is_scrub = ps->pss_func == POOL_SCAN_SCRUB;
	assert(is_resilver \|\| is_scrub);

	/* Scan is finished or canceled. */
	if (ps->pss_state == DSS_FINISHED) {
	secs_to_dhms(end - start, time_buf);

	if (is_scrub) {
	(void) printf(gettext("scrub repaired %s "
	"in %s with %llu errors on %s"), processed_buf,
	time_buf, (u_longlong_t)ps->pss_errors,
	ctime(&end));
	} else if (is_resilver) {
	(void) printf(gettext("resilvered %s "
	"in %s with %llu errors on %s"), processed_buf,
	time_buf, (u_longlong_t)ps->pss_errors,
	ctime(&end));
	}
	return;
	} else if (ps->pss_state == DSS_CANCELED) {
	if (is_scrub) {
	(void) printf(gettext("scrub canceled on %s"),
	ctime(&end));
	} else if (is_resilver) {
	(void) printf(gettext("resilver canceled on %s"),
	ctime(&end));
	}
	return;
	}

	assert(ps->pss_state == DSS_SCANNING);

	/* Scan is in progress. Resilvers can't be paused. */
	if (is_scrub) {
	if (pause == 0) {
	(void) printf(gettext("scrub in progress since %s"),
	ctime(&start));
	} else {
	(void) printf(gettext("scrub paused since %s"),
	ctime(&pause));
	(void) printf(gettext("\tscrub started on %s"),
	ctime(&start));
	}
	} else if (is_resilver) {
	(void) printf(gettext("resilver in progress since %s"),
	ctime(&start));
	}

	scanned = ps->pss_examined;
	pass_scanned = ps->pss_pass_exam;
	issued = ps->pss_issued;
	pass_issued = ps->pss_pass_issued;
	total_s = ps->pss_to_examine;
	total_i = ps->pss_to_examine - ps->pss_skipped;

	/* we are only done with a block once we have issued the IO for it */
	fraction_done = (double)issued / total_i;

	/* elapsed time for this pass, rounding up to 1 if it's 0 */
	elapsed = time(NULL) - ps->pss_pass_start;
	elapsed -= ps->pss_pass_scrub_spent_paused;
	elapsed = (elapsed != 0) ? elapsed : 1;

	scan_rate = pass_scanned / elapsed;
	issue_rate = pass_issued / elapsed;

	/* format all of the numbers we will be reporting */
	zfs_nicebytes(scanned, scanned_buf, sizeof (scanned_buf));
	zfs_nicebytes(issued, issued_buf, sizeof (issued_buf));
	zfs_nicebytes(total_s, total_s_buf, sizeof (total_s_buf));
	zfs_nicebytes(total_i, total_i_buf, sizeof (total_i_buf));

	/* do not print estimated time if we have a paused scrub */
	(void) printf(gettext("\t%s / %s scanned"), scanned_buf, total_s_buf);
	if (pause == 0 && scan_rate > 0) {
	zfs_nicebytes(scan_rate, srate_buf, sizeof (srate_buf));
	(void) printf(gettext(" at %s/s"), srate_buf);
	}
	(void) printf(gettext(", %s / %s issued"), issued_buf, total_i_buf);
	if (pause == 0 && issue_rate > 0) {
	zfs_nicebytes(issue_rate, irate_buf, sizeof (irate_buf));
	(void) printf(gettext(" at %s/s"), irate_buf);
	}
	(void) printf(gettext("\n"));

	if (is_resilver) {
	(void) printf(gettext("\t%s resilvered, %.2f%% done"),
	processed_buf, 100 * fraction_done);
	} else if (is_scrub) {
	(void) printf(gettext("\t%s repaired, %.2f%% done"),
	processed_buf, 100 * fraction_done);
	}

	if (pause == 0) {
	/*
	* Only provide an estimate iff:
	* 1) we haven't yet issued all we expected, and
	* 2) the issue rate exceeds 10 MB/s, and
	* 3) it's either:
	* a) a resilver which has started repairs, or
	* b) a scrub which has entered the issue phase.
	*/
	if (total_i >= issued && issue_rate >= 10 * 1024 * 1024 &&
	((is_resilver && ps->pss_processed > 0) \|\|
	(is_scrub && issued > 0))) {
	secs_to_dhms((total_i - issued) / issue_rate, time_buf);
	(void) printf(gettext(", %s to go\n"), time_buf);
	} else {
	(void) printf(gettext(", no estimated "
	"completion time\n"));
	}
	} else {
	(void) printf(gettext("\n"));
	}
	}

	static void
	print_rebuild_status_impl(vdev_rebuild_stat_t vrs, uint_t c, char vdev_name)
	{
	if (vrs == NULL \|\| vrs->vrs_state == VDEV_REBUILD_NONE)
	return;

	printf(" ");
	printf_color(ANSI_BOLD, gettext("scan:"));
	printf(" ");

	uint64_t bytes_scanned = vrs->vrs_bytes_scanned;
	uint64_t bytes_issued = vrs->vrs_bytes_issued;
	uint64_t bytes_rebuilt = vrs->vrs_bytes_rebuilt;
	uint64_t bytes_est_s = vrs->vrs_bytes_est;
	uint64_t bytes_est_i = vrs->vrs_bytes_est;
	if (c > offsetof(vdev_rebuild_stat_t, vrs_pass_bytes_skipped) / 8)
	bytes_est_i -= vrs->vrs_pass_bytes_skipped;
	uint64_t scan_rate = (vrs->vrs_pass_bytes_scanned /
	(vrs->vrs_pass_time_ms + 1)) * 1000;
	uint64_t issue_rate = (vrs->vrs_pass_bytes_issued /
	(vrs->vrs_pass_time_ms + 1)) * 1000;
	double scan_pct = MIN((double)bytes_scanned * 100 /
	(bytes_est_s + 1), 100);

	/* Format all of the numbers we will be reporting */
	char bytes_scanned_buf[7], bytes_issued_buf[7];
	char bytes_rebuilt_buf[7], bytes_est_s_buf[7], bytes_est_i_buf[7];
	char scan_rate_buf[7], issue_rate_buf[7], time_buf[32];
	zfs_nicebytes(bytes_scanned, bytes_scanned_buf,
	sizeof (bytes_scanned_buf));
	zfs_nicebytes(bytes_issued, bytes_issued_buf,
	sizeof (bytes_issued_buf));
	zfs_nicebytes(bytes_rebuilt, bytes_rebuilt_buf,
	sizeof (bytes_rebuilt_buf));
	zfs_nicebytes(bytes_est_s, bytes_est_s_buf, sizeof (bytes_est_s_buf));
	zfs_nicebytes(bytes_est_i, bytes_est_i_buf, sizeof (bytes_est_i_buf));

	time_t start = vrs->vrs_start_time;
	time_t end = vrs->vrs_end_time;

	/* Rebuild is finished or canceled. */
	if (vrs->vrs_state == VDEV_REBUILD_COMPLETE) {
	secs_to_dhms(vrs->vrs_scan_time_ms / 1000, time_buf);
	(void) printf(gettext("resilvered (%s) %s in %s "
	"with %llu errors on %s"), vdev_name, bytes_rebuilt_buf,
	time_buf, (u_longlong_t)vrs->vrs_errors, ctime(&end));
	return;
	} else if (vrs->vrs_state == VDEV_REBUILD_CANCELED) {
	(void) printf(gettext("resilver (%s) canceled on %s"),
	vdev_name, ctime(&end));
	return;
	} else if (vrs->vrs_state == VDEV_REBUILD_ACTIVE) {
	(void) printf(gettext("resilver (%s) in progress since %s"),
	vdev_name, ctime(&start));
	}

	assert(vrs->vrs_state == VDEV_REBUILD_ACTIVE);

	(void) printf(gettext("\t%s / %s scanned"), bytes_scanned_buf,
	bytes_est_s_buf);
	if (scan_rate > 0) {
	zfs_nicebytes(scan_rate, scan_rate_buf, sizeof (scan_rate_buf));
	(void) printf(gettext(" at %s/s"), scan_rate_buf);
	}
	(void) printf(gettext(", %s / %s issued"), bytes_issued_buf,
	bytes_est_i_buf);
	if (issue_rate > 0) {
	zfs_nicebytes(issue_rate, issue_rate_buf,
	sizeof (issue_rate_buf));
	(void) printf(gettext(" at %s/s"), issue_rate_buf);
	}
	(void) printf(gettext("\n"));

	(void) printf(gettext("\t%s resilvered, %.2f%% done"),
	bytes_rebuilt_buf, scan_pct);

	if (vrs->vrs_state == VDEV_REBUILD_ACTIVE) {
	if (bytes_est_s >= bytes_scanned &&
	scan_rate >= 10 * 1024 * 1024) {
	secs_to_dhms((bytes_est_s - bytes_scanned) / scan_rate,
	time_buf);
	(void) printf(gettext(", %s to go\n"), time_buf);
	} else {
	(void) printf(gettext(", no estimated "
	"completion time\n"));
	}
	} else {
	(void) printf(gettext("\n"));
	}
	}

	/*
	* Print rebuild status for top-level vdevs.
	*/
	static void
	print_rebuild_status(zpool_handle_t zhp, nvlist_t nvroot)
	{
	nvlist_t **child;
	uint_t children;

	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
	&child, &children) != 0)
	children = 0;

	for (uint_t c = 0; c < children; c++) {
	vdev_rebuild_stat_t *vrs;
	uint_t i;

	if (nvlist_lookup_uint64_array(child[c],
	ZPOOL_CONFIG_REBUILD_STATS, (uint64_t **)&vrs, &i) == 0) {
	char *name = zpool_vdev_name(g_zfs, zhp,
	child[c], VDEV_NAME_TYPE_ID);
	print_rebuild_status_impl(vrs, i, name);
	free(name);
	}
	}
	}

	/*
	* As we don't scrub checkpointed blocks, we want to warn the user that we
	* skipped scanning some blocks if a checkpoint exists or existed at any
	* time during the scan. If a sequential instead of healing reconstruction
	* was performed then the blocks were reconstructed. However, their checksums
	* have not been verified so we still print the warning.
	*/
	static void
	print_checkpoint_scan_warning(pool_scan_stat_t ps, pool_checkpoint_stat_t pcs)
	{
	if (ps == NULL \|\| pcs == NULL)
	return;

	if (pcs->pcs_state == CS_NONE \|\|
	pcs->pcs_state == CS_CHECKPOINT_DISCARDING)
	return;

	assert(pcs->pcs_state == CS_CHECKPOINT_EXISTS);

	if (ps->pss_state == DSS_NONE)
	return;

	if ((ps->pss_state == DSS_FINISHED \|\| ps->pss_state == DSS_CANCELED) &&
	ps->pss_end_time < pcs->pcs_start_time)
	return;

	if (ps->pss_state == DSS_FINISHED \|\| ps->pss_state == DSS_CANCELED) {
	(void) printf(gettext(" scan warning: skipped blocks "
	"that are only referenced by the checkpoint.\n"));
	} else {
	assert(ps->pss_state == DSS_SCANNING);
	(void) printf(gettext(" scan warning: skipping blocks "
	"that are only referenced by the checkpoint.\n"));
	}
	}

	/*
	* Returns B_TRUE if there is an active rebuild in progress. Otherwise,
	* B_FALSE is returned and 'rebuild_end_time' is set to the end time for
	* the last completed (or cancelled) rebuild.
	*/
	static boolean_t
	check_rebuilding(nvlist_t nvroot, uint64_t rebuild_end_time)
	{
	nvlist_t **child;
	uint_t children;
	boolean_t rebuilding = B_FALSE;
	uint64_t end_time = 0;

	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
	&child, &children) != 0)
	children = 0;

	for (uint_t c = 0; c < children; c++) {
	vdev_rebuild_stat_t *vrs;
	uint_t i;

	if (nvlist_lookup_uint64_array(child[c],
	ZPOOL_CONFIG_REBUILD_STATS, (uint64_t **)&vrs, &i) == 0) {

	if (vrs->vrs_end_time > end_time)
	end_time = vrs->vrs_end_time;

	if (vrs->vrs_state == VDEV_REBUILD_ACTIVE) {
	rebuilding = B_TRUE;
	end_time = 0;
	break;
	}
	}
	}

	if (rebuild_end_time != NULL)
	*rebuild_end_time = end_time;

	return (rebuilding);
	}

	/*
	* Print the scan status.
	*/
	static void
	print_scan_status(zpool_handle_t zhp, nvlist_t nvroot)
	{
	uint64_t rebuild_end_time = 0, resilver_end_time = 0;
	boolean_t have_resilver = B_FALSE, have_scrub = B_FALSE;
	boolean_t have_errorscrub = B_FALSE;
	boolean_t active_resilver = B_FALSE;
	pool_checkpoint_stat_t *pcs = NULL;
	pool_scan_stat_t *ps = NULL;
	uint_t c;
	time_t scrub_start = 0, errorscrub_start = 0;

	if (nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_SCAN_STATS,
	(uint64_t **)&ps, &c) == 0) {
	if (ps->pss_func == POOL_SCAN_RESILVER) {
	resilver_end_time = ps->pss_end_time;
	active_resilver = (ps->pss_state == DSS_SCANNING);
	}

	have_resilver = (ps->pss_func == POOL_SCAN_RESILVER);
	have_scrub = (ps->pss_func == POOL_SCAN_SCRUB);
	scrub_start = ps->pss_start_time;
	if (c > offsetof(pool_scan_stat_t,
	pss_pass_error_scrub_pause) / 8) {
	have_errorscrub = (ps->pss_error_scrub_func ==
	POOL_SCAN_ERRORSCRUB);
	errorscrub_start = ps->pss_error_scrub_start;
	}
	}

	boolean_t active_rebuild = check_rebuilding(nvroot, &rebuild_end_time);
	boolean_t have_rebuild = (active_rebuild \|\| (rebuild_end_time > 0));

	/* Always print the scrub status when available. */
	if (have_scrub && scrub_start > errorscrub_start)
	print_scan_scrub_resilver_status(ps);
	else if (have_errorscrub && errorscrub_start >= scrub_start)
	print_err_scrub_status(ps);

	/*
	* When there is an active resilver or rebuild print its status.
	* Otherwise print the status of the last resilver or rebuild.
	*/
	if (active_resilver \|\| (!active_rebuild && have_resilver &&
	resilver_end_time && resilver_end_time > rebuild_end_time)) {
	print_scan_scrub_resilver_status(ps);
	} else if (active_rebuild \|\| (!active_resilver && have_rebuild &&
	rebuild_end_time && rebuild_end_time > resilver_end_time)) {
	print_rebuild_status(zhp, nvroot);
	}

	(void) nvlist_lookup_uint64_array(nvroot,
	ZPOOL_CONFIG_CHECKPOINT_STATS, (uint64_t **)&pcs, &c);
	print_checkpoint_scan_warning(ps, pcs);
	}

	/*
	* Print out detailed removal status.
	*/
	static void
	print_removal_status(zpool_handle_t zhp, pool_removal_stat_t prs)
	{
	char copied_buf[7], examined_buf[7], total_buf[7], rate_buf[7];
	time_t start, end;
	nvlist_t config, nvroot;
	nvlist_t **child;
	uint_t children;
	char *vdev_name;

	if (prs == NULL \|\| prs->prs_state == DSS_NONE)
	return;

	/*
	* Determine name of vdev.
	*/
	config = zpool_get_config(zhp, NULL);
	nvroot = fnvlist_lookup_nvlist(config,
	ZPOOL_CONFIG_VDEV_TREE);
	verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
	&child, &children) == 0);
	assert(prs->prs_removing_vdev < children);
	vdev_name = zpool_vdev_name(g_zfs, zhp,
	child[prs->prs_removing_vdev], B_TRUE);

	printf_color(ANSI_BOLD, gettext("remove: "));

	start = prs->prs_start_time;
	end = prs->prs_end_time;
	zfs_nicenum(prs->prs_copied, copied_buf, sizeof (copied_buf));

	/*
	* Removal is finished or canceled.
	*/
	if (prs->prs_state == DSS_FINISHED) {
	uint64_t minutes_taken = (end - start) / 60;

	(void) printf(gettext("Removal of vdev %llu copied %s "
	"in %lluh%um, completed on %s"),
	(longlong_t)prs->prs_removing_vdev,
	copied_buf,
	(u_longlong_t)(minutes_taken / 60),
	(uint_t)(minutes_taken % 60),
	ctime((time_t *)&end));
	} else if (prs->prs_state == DSS_CANCELED) {
	(void) printf(gettext("Removal of %s canceled on %s"),
	vdev_name, ctime(&end));
	} else {
	uint64_t copied, total, elapsed, mins_left, hours_left;
	double fraction_done;
	uint_t rate;

	assert(prs->prs_state == DSS_SCANNING);

	/*
	* Removal is in progress.
	*/
	(void) printf(gettext(
	"Evacuation of %s in progress since %s"),
	vdev_name, ctime(&start));

	copied = prs->prs_copied > 0 ? prs->prs_copied : 1;
	total = prs->prs_to_copy;
	fraction_done = (double)copied / total;

	/* elapsed time for this pass */
	elapsed = time(NULL) - prs->prs_start_time;
	elapsed = elapsed > 0 ? elapsed : 1;
	rate = copied / elapsed;
	rate = rate > 0 ? rate : 1;
	mins_left = ((total - copied) / rate) / 60;
	hours_left = mins_left / 60;

	zfs_nicenum(copied, examined_buf, sizeof (examined_buf));
	zfs_nicenum(total, total_buf, sizeof (total_buf));
	zfs_nicenum(rate, rate_buf, sizeof (rate_buf));

	/*
	* do not print estimated time if hours_left is more than
	* 30 days
	*/
	(void) printf(gettext(
	"\t%s copied out of %s at %s/s, %.2f%% done"),
	examined_buf, total_buf, rate_buf, 100 * fraction_done);
	if (hours_left < (30 * 24)) {
	(void) printf(gettext(", %lluh%um to go\n"),
	(u_longlong_t)hours_left, (uint_t)(mins_left % 60));
	} else {
	(void) printf(gettext(
	", (copy is slow, no estimated time)\n"));
	}
	}
	free(vdev_name);

	if (prs->prs_mapping_memory > 0) {
	char mem_buf[7];
	zfs_nicenum(prs->prs_mapping_memory, mem_buf, sizeof (mem_buf));
	(void) printf(gettext(
	"\t%s memory used for removed device mappings\n"),
	mem_buf);
	}
	}

	/*
	* Print out detailed raidz expansion status.
	*/
	static void
	print_raidz_expand_status(zpool_handle_t zhp, pool_raidz_expand_stat_t pres)
	{
	char copied_buf[7];

	if (pres == NULL \|\| pres->pres_state == DSS_NONE)
	return;

	/*
	* Determine name of vdev.
	*/
	nvlist_t *config = zpool_get_config(zhp, NULL);
	nvlist_t *nvroot = fnvlist_lookup_nvlist(config,
	ZPOOL_CONFIG_VDEV_TREE);
	nvlist_t **child;
	uint_t children;
	verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
	&child, &children) == 0);
	assert(pres->pres_expanding_vdev < children);

	printf_color(ANSI_BOLD, gettext("expand: "));

	time_t start = pres->pres_start_time;
	time_t end = pres->pres_end_time;
	char *vname =
	zpool_vdev_name(g_zfs, zhp, child[pres->pres_expanding_vdev], 0);
	zfs_nicenum(pres->pres_reflowed, copied_buf, sizeof (copied_buf));

	/*
	* Expansion is finished or canceled.
	*/
	if (pres->pres_state == DSS_FINISHED) {
	char time_buf[32];
	secs_to_dhms(end - start, time_buf);

	(void) printf(gettext("expanded %s-%u copied %s in %s, "
	"on %s"), vname, (int)pres->pres_expanding_vdev,
	copied_buf, time_buf, ctime((time_t *)&end));
	} else {
	char examined_buf[7], total_buf[7], rate_buf[7];
	uint64_t copied, total, elapsed, secs_left;
	double fraction_done;
	uint_t rate;

	assert(pres->pres_state == DSS_SCANNING);

	/*
	* Expansion is in progress.
	*/
	(void) printf(gettext(
	"expansion of %s-%u in progress since %s"),
	vname, (int)pres->pres_expanding_vdev, ctime(&start));

	copied = pres->pres_reflowed > 0 ? pres->pres_reflowed : 1;
	total = pres->pres_to_reflow;
	fraction_done = (double)copied / total;

	/* elapsed time for this pass */
	elapsed = time(NULL) - pres->pres_start_time;
	elapsed = elapsed > 0 ? elapsed : 1;
	rate = copied / elapsed;
	rate = rate > 0 ? rate : 1;
	secs_left = (total - copied) / rate;

	zfs_nicenum(copied, examined_buf, sizeof (examined_buf));
	zfs_nicenum(total, total_buf, sizeof (total_buf));
	zfs_nicenum(rate, rate_buf, sizeof (rate_buf));

	/*
	* do not print estimated time if hours_left is more than
	* 30 days
	*/
	(void) printf(gettext("\t%s / %s copied at %s/s, %.2f%% done"),
	examined_buf, total_buf, rate_buf, 100 * fraction_done);
	if (pres->pres_waiting_for_resilver) {
	(void) printf(gettext(", paused for resilver or "
	"clear\n"));
	} else if (secs_left < (30 * 24 * 3600)) {
	char time_buf[32];
	secs_to_dhms(secs_left, time_buf);
	(void) printf(gettext(", %s to go\n"), time_buf);
	} else {
	(void) printf(gettext(
	", (copy is slow, no estimated time)\n"));
	}
	}
	free(vname);
	}
	static void
	print_checkpoint_status(pool_checkpoint_stat_t *pcs)
	{
	time_t start;
	char space_buf[7];

	if (pcs == NULL \|\| pcs->pcs_state == CS_NONE)
	return;

	(void) printf(gettext("checkpoint: "));

	start = pcs->pcs_start_time;
	zfs_nicenum(pcs->pcs_space, space_buf, sizeof (space_buf));

	if (pcs->pcs_state == CS_CHECKPOINT_EXISTS) {
	char *date = ctime(&start);

	/*
	* ctime() adds a newline at the end of the generated
	* string, thus the weird format specifier and the
	* strlen() call used to chop it off from the output.
	*/
	(void) printf(gettext("created %.*s, consumes %s\n"),
	(int)(strlen(date) - 1), date, space_buf);
	return;
	}

	assert(pcs->pcs_state == CS_CHECKPOINT_DISCARDING);

	(void) printf(gettext("discarding, %s remaining.\n"),
	space_buf);
	}

	static void
	print_error_log(zpool_handle_t *zhp)
	{
	nvlist_t *nverrlist = NULL;
	nvpair_t *elem;
	char *pathname;
	size_t len = MAXPATHLEN * 2;

	if (zpool_get_errlog(zhp, &nverrlist) != 0)
	return;

	(void) printf("errors: Permanent errors have been "
	"detected in the following files:\n\n");

	pathname = safe_malloc(len);
	elem = NULL;
	while ((elem = nvlist_next_nvpair(nverrlist, elem)) != NULL) {
	nvlist_t *nv;
	uint64_t dsobj, obj;

	verify(nvpair_value_nvlist(elem, &nv) == 0);
	verify(nvlist_lookup_uint64(nv, ZPOOL_ERR_DATASET,
	&dsobj) == 0);
	verify(nvlist_lookup_uint64(nv, ZPOOL_ERR_OBJECT,
	&obj) == 0);
	zpool_obj_to_path(zhp, dsobj, obj, pathname, len);
	(void) printf("%7s %s\n", "", pathname);
	}
	free(pathname);
	nvlist_free(nverrlist);
	}

	static void
	print_spares(zpool_handle_t zhp, status_cbdata_t cb, nvlist_t **spares,
	uint_t nspares)
	{
	uint_t i;
	char *name;

	if (nspares == 0)
	return;

	(void) printf(gettext("\tspares\n"));

	for (i = 0; i < nspares; i++) {
	name = zpool_vdev_name(g_zfs, zhp, spares[i],
	cb->cb_name_flags);
	print_status_config(zhp, cb, name, spares[i], 2, B_TRUE, NULL);
	free(name);
	}
	}

	static void
	print_l2cache(zpool_handle_t zhp, status_cbdata_t cb, nvlist_t **l2cache,
	uint_t nl2cache)
	{
	uint_t i;
	char *name;

	if (nl2cache == 0)
	return;

	(void) printf(gettext("\tcache\n"));

	for (i = 0; i < nl2cache; i++) {
	name = zpool_vdev_name(g_zfs, zhp, l2cache[i],
	cb->cb_name_flags);
	print_status_config(zhp, cb, name, l2cache[i], 2,
	B_FALSE, NULL);
	free(name);
	}
	}

	static void
	print_dedup_stats(nvlist_t *config)
	{
	ddt_histogram_t *ddh;
	ddt_stat_t *dds;
	ddt_object_t *ddo;
	uint_t c;
	char dspace[6], mspace[6];

	/*
	* If the pool was faulted then we may not have been able to
	* obtain the config. Otherwise, if we have anything in the dedup
	* table continue processing the stats.
	*/
	if (nvlist_lookup_uint64_array(config, ZPOOL_CONFIG_DDT_OBJ_STATS,
	(uint64_t **)&ddo, &c) != 0)
	return;

	(void) printf("\n");
	(void) printf(gettext(" dedup: "));
	if (ddo->ddo_count == 0) {
	(void) printf(gettext("no DDT entries\n"));
	return;
	}

	zfs_nicebytes(ddo->ddo_dspace, dspace, sizeof (dspace));
	zfs_nicebytes(ddo->ddo_mspace, mspace, sizeof (mspace));
	(void) printf("DDT entries %llu, size %s on disk, %s in core\n",
	(u_longlong_t)ddo->ddo_count,
	dspace,
	mspace);

	verify(nvlist_lookup_uint64_array(config, ZPOOL_CONFIG_DDT_STATS,
	(uint64_t **)&dds, &c) == 0);
	verify(nvlist_lookup_uint64_array(config, ZPOOL_CONFIG_DDT_HISTOGRAM,
	(uint64_t **)&ddh, &c) == 0);
	zpool_dump_ddt(dds, ddh);
	}

	/*
	* Display a summary of pool status. Displays a summary such as:
	*
	* pool: tank
	* status: DEGRADED
	* reason: One or more devices ...
	* see: https://openzfs.github.io/openzfs-docs/msg/ZFS-xxxx-01
	* config:
	* mirror DEGRADED
	* c1t0d0 OK
	* c2t0d0 UNAVAIL
	*
	* When given the '-v' option, we print out the complete config. If the '-e'
	* option is specified, then we print out error rate information as well.
	*/
	static int
	status_callback(zpool_handle_t zhp, void data)
	{
	status_cbdata_t *cbp = data;
	nvlist_t config, nvroot;
	const char *msgid;
	zpool_status_t reason;
	zpool_errata_t errata;
	const char *health;
	uint_t c;
	vdev_stat_t *vs;

	config = zpool_get_config(zhp, NULL);
	reason = zpool_get_status(zhp, &msgid, &errata);

	cbp->cb_count++;

	/*
	* If we were given 'zpool status -x', only report those pools with
	* problems.
	*/
	if (cbp->cb_explain &&
	(reason == ZPOOL_STATUS_OK \|\|
	reason == ZPOOL_STATUS_VERSION_OLDER \|\|
	reason == ZPOOL_STATUS_FEAT_DISABLED \|\|
	reason == ZPOOL_STATUS_COMPATIBILITY_ERR \|\|
	reason == ZPOOL_STATUS_INCOMPATIBLE_FEAT)) {
	if (!cbp->cb_allpools) {
	(void) printf(gettext("pool '%s' is healthy\n"),
	zpool_get_name(zhp));
	if (cbp->cb_first)
	cbp->cb_first = B_FALSE;
	}
	return (0);
	}

	if (cbp->cb_first)
	cbp->cb_first = B_FALSE;
	else
	(void) printf("\n");

	nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);
	verify(nvlist_lookup_uint64_array(nvroot, ZPOOL_CONFIG_VDEV_STATS,
	(uint64_t **)&vs, &c) == 0);

	health = zpool_get_state_str(zhp);

	printf(" ");
	printf_color(ANSI_BOLD, gettext("pool:"));
	printf(" %s\n", zpool_get_name(zhp));
	fputc(' ', stdout);
	printf_color(ANSI_BOLD, gettext("state: "));

	printf_color(health_str_to_color(health), "%s", health);

	fputc('\n', stdout);

	switch (reason) {
	case ZPOOL_STATUS_MISSING_DEV_R:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices could "
	"not be opened. Sufficient replicas exist for\n\tthe pool "
	"to continue functioning in a degraded state.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Attach the missing device "
	"and online it using 'zpool online'.\n"));
	break;

	case ZPOOL_STATUS_MISSING_DEV_NR:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices could "
	"not be opened. There are insufficient\n\treplicas for the"
	" pool to continue functioning.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Attach the missing device "
	"and online it using 'zpool online'.\n"));
	break;

	case ZPOOL_STATUS_CORRUPT_LABEL_R:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices could "
	"not be used because the label is missing or\n\tinvalid. "
	"Sufficient replicas exist for the pool to continue\n\t"
	"functioning in a degraded state.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Replace the device using "
	"'zpool replace'.\n"));
	break;

	case ZPOOL_STATUS_CORRUPT_LABEL_NR:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices could "
	"not be used because the label is missing \n\tor invalid. "
	"There are insufficient replicas for the pool to "
	"continue\n\tfunctioning.\n"));
	zpool_explain_recover(zpool_get_handle(zhp),
	zpool_get_name(zhp), reason, config);
	break;

	case ZPOOL_STATUS_FAILING_DEV:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices has "
	"experienced an unrecoverable error. An\n\tattempt was "
	"made to correct the error. Applications are "
	"unaffected.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Determine if the "
	"device needs to be replaced, and clear the errors\n\tusing"
	" 'zpool clear' or replace the device with 'zpool "
	"replace'.\n"));
	break;

	case ZPOOL_STATUS_OFFLINE_DEV:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices has "
	"been taken offline by the administrator.\n\tSufficient "
	"replicas exist for the pool to continue functioning in "
	"a\n\tdegraded state.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Online the device "
	"using 'zpool online' or replace the device with\n\t'zpool "
	"replace'.\n"));
	break;

	case ZPOOL_STATUS_REMOVED_DEV:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices has "
	"been removed by the administrator.\n\tSufficient "
	"replicas exist for the pool to continue functioning in "
	"a\n\tdegraded state.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Online the device "
	"using zpool online' or replace the device with\n\t'zpool "
	"replace'.\n"));
	break;

	case ZPOOL_STATUS_RESILVERING:
	case ZPOOL_STATUS_REBUILDING:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices is "
	"currently being resilvered. The pool will\n\tcontinue "
	"to function, possibly in a degraded state.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Wait for the resilver to "
	"complete.\n"));
	break;

	case ZPOOL_STATUS_REBUILD_SCRUB:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices have "
	"been sequentially resilvered, scrubbing\n\tthe pool "
	"is recommended.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Use 'zpool scrub' to "
	"verify all data checksums.\n"));
	break;

	case ZPOOL_STATUS_CORRUPT_DATA:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices has "
	"experienced an error resulting in data\n\tcorruption. "
	"Applications may be affected.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Restore the file in question"
	" if possible. Otherwise restore the\n\tentire pool from "
	"backup.\n"));
	break;

	case ZPOOL_STATUS_CORRUPT_POOL:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool metadata is "
	"corrupted and the pool cannot be opened.\n"));
	zpool_explain_recover(zpool_get_handle(zhp),
	zpool_get_name(zhp), reason, config);
	break;

	case ZPOOL_STATUS_VERSION_OLDER:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool is formatted using "
	"a legacy on-disk format. The pool can\n\tstill be used, "
	"but some features are unavailable.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Upgrade the pool using "
	"'zpool upgrade'. Once this is done, the\n\tpool will no "
	"longer be accessible on software that does not support\n\t"
	"feature flags.\n"));
	break;

	case ZPOOL_STATUS_VERSION_NEWER:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool has been upgraded "
	"to a newer, incompatible on-disk version.\n\tThe pool "
	"cannot be accessed on this system.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Access the pool from a "
	"system running more recent software, or\n\trestore the "
	"pool from backup.\n"));
	break;

	case ZPOOL_STATUS_FEAT_DISABLED:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("Some supported and "
	"requested features are not enabled on the pool.\n\t"
	"The pool can still be used, but some features are "
	"unavailable.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Enable all features using "
	"'zpool upgrade'. Once this is done,\n\tthe pool may no "
	"longer be accessible by software that does not support\n\t"
	"the features. See zpool-features(7) for details.\n"));
	break;

	case ZPOOL_STATUS_COMPATIBILITY_ERR:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("This pool has a "
	"compatibility list specified, but it could not be\n\t"
	"read/parsed at this time. The pool can still be used, "
	"but this\n\tshould be investigated.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Check the value of the "
	"'compatibility' property against the\n\t"
	"appropriate file in " ZPOOL_SYSCONF_COMPAT_D " or "
	ZPOOL_DATA_COMPAT_D ".\n"));
	break;

	case ZPOOL_STATUS_INCOMPATIBLE_FEAT:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more features "
	"are enabled on the pool despite not being\n\t"
	"requested by the 'compatibility' property.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Consider setting "
	"'compatibility' to an appropriate value, or\n\t"
	"adding needed features to the relevant file in\n\t"
	ZPOOL_SYSCONF_COMPAT_D " or " ZPOOL_DATA_COMPAT_D ".\n"));
	break;

	case ZPOOL_STATUS_UNSUP_FEAT_READ:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool cannot be accessed "
	"on this system because it uses the\n\tfollowing feature(s)"
	" not supported on this system:\n"));
	zpool_print_unsup_feat(config);
	(void) printf("\n");
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Access the pool from a "
	"system that supports the required feature(s),\n\tor "
	"restore the pool from backup.\n"));
	break;

	case ZPOOL_STATUS_UNSUP_FEAT_WRITE:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool can only be "
	"accessed in read-only mode on this system. It\n\tcannot be"
	" accessed in read-write mode because it uses the "
	"following\n\tfeature(s) not supported on this system:\n"));
	zpool_print_unsup_feat(config);
	(void) printf("\n");
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("The pool cannot be accessed "
	"in read-write mode. Import the pool with\n"
	"\t\"-o readonly=on\", access the pool from a system that "
	"supports the\n\trequired feature(s), or restore the "
	"pool from backup.\n"));
	break;

	case ZPOOL_STATUS_FAULTED_DEV_R:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices are "
	"faulted in response to persistent errors.\n\tSufficient "
	"replicas exist for the pool to continue functioning "
	"in a\n\tdegraded state.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Replace the faulted device, "
	"or use 'zpool clear' to mark the device\n\trepaired.\n"));
	break;

	case ZPOOL_STATUS_FAULTED_DEV_NR:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices are "
	"faulted in response to persistent errors. There are "
	"insufficient replicas for the pool to\n\tcontinue "
	"functioning.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Destroy and re-create the "
	"pool from a backup source. Manually marking the device\n"
	"\trepaired using 'zpool clear' may allow some data "
	"to be recovered.\n"));
	break;

	case ZPOOL_STATUS_IO_FAILURE_MMP:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("The pool is suspended "
	"because multihost writes failed or were delayed;\n\t"
	"another system could import the pool undetected.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Make sure the pool's devices"
	" are connected, then reboot your system and\n\timport the "
	"pool or run 'zpool clear' to resume the pool.\n"));
	break;

	case ZPOOL_STATUS_IO_FAILURE_WAIT:
	case ZPOOL_STATUS_IO_FAILURE_CONTINUE:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("One or more devices are "
	"faulted in response to IO failures.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Make sure the affected "
	"devices are connected, then run 'zpool clear'.\n"));
	break;

	case ZPOOL_STATUS_BAD_LOG:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("An intent log record "
	"could not be read.\n"
	"\tWaiting for administrator intervention to fix the "
	"faulted pool.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Either restore the affected "
	"device(s) and run 'zpool online',\n"
	"\tor ignore the intent log records by running "
	"'zpool clear'.\n"));
	break;

	case ZPOOL_STATUS_NON_NATIVE_ASHIFT:
	(void) printf(gettext("status: One or more devices are "
	"configured to use a non-native block size.\n"
	"\tExpect reduced performance.\n"));
	(void) printf(gettext("action: Replace affected devices with "
	"devices that support the\n\tconfigured block size, or "
	"migrate data to a properly configured\n\tpool.\n"));
	break;

	case ZPOOL_STATUS_HOSTID_MISMATCH:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("Mismatch between pool hostid"
	" and system hostid on imported pool.\n\tThis pool was "
	"previously imported into a system with a different "
	"hostid,\n\tand then was verbatim imported into this "
	"system.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("Export this pool on all "
	"systems on which it is imported.\n"
	"\tThen import it to correct the mismatch.\n"));
	break;

	case ZPOOL_STATUS_ERRATA:
	printf_color(ANSI_BOLD, gettext("status: "));
	printf_color(ANSI_YELLOW, gettext("Errata #%d detected.\n"),
	errata);

	switch (errata) {
	case ZPOOL_ERRATA_NONE:
	break;

	case ZPOOL_ERRATA_ZOL_2094_SCRUB:
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("To correct the issue"
	" run 'zpool scrub'.\n"));
	break;

	case ZPOOL_ERRATA_ZOL_6845_ENCRYPTION:
	(void) printf(gettext("\tExisting encrypted datasets "
	"contain an on-disk incompatibility\n\twhich "
	"needs to be corrected.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("To correct the issue"
	" backup existing encrypted datasets to new\n\t"
	"encrypted datasets and destroy the old ones. "
	"'zfs mount -o ro' can\n\tbe used to temporarily "
	"mount existing encrypted datasets readonly.\n"));
	break;

	case ZPOOL_ERRATA_ZOL_8308_ENCRYPTION:
	(void) printf(gettext("\tExisting encrypted snapshots "
	"and bookmarks contain an on-disk\n\tincompat"
	"ibility. This may cause on-disk corruption if "
	"they are used\n\twith 'zfs recv'.\n"));
	printf_color(ANSI_BOLD, gettext("action: "));
	printf_color(ANSI_YELLOW, gettext("To correct the"
	"issue, enable the bookmark_v2 feature. No "
	"additional\n\taction is needed if there are no "
	"encrypted snapshots or bookmarks.\n\tIf preserving"
	"the encrypted snapshots and bookmarks is required,"
	" use\n\ta non-raw send to backup and restore them."
	" Alternately, they may be\n\tremoved to resolve "
	"the incompatibility.\n"));
	break;

	default:
	/*
	* All errata which allow the pool to be imported
	* must contain an action message.
	*/
	assert(0);
	}
	break;

	default:
	/*
	* The remaining errors can't actually be generated, yet.
	*/
	assert(reason == ZPOOL_STATUS_OK);
	}

	if (msgid != NULL) {
	printf(" ");
	printf_color(ANSI_BOLD, gettext("see:"));
	printf(gettext(
	" https://openzfs.github.io/openzfs-docs/msg/%s\n"),
	msgid);
	}

	if (config != NULL) {
	uint64_t nerr;
	nvlist_t spares, l2cache;
	uint_t nspares, nl2cache;

	print_scan_status(zhp, nvroot);

	pool_removal_stat_t *prs = NULL;
	(void) nvlist_lookup_uint64_array(nvroot,
	ZPOOL_CONFIG_REMOVAL_STATS, (uint64_t **)&prs, &c);
	print_removal_status(zhp, prs);

	pool_checkpoint_stat_t *pcs = NULL;
	(void) nvlist_lookup_uint64_array(nvroot,
	ZPOOL_CONFIG_CHECKPOINT_STATS, (uint64_t **)&pcs, &c);
	print_checkpoint_status(pcs);

	pool_raidz_expand_stat_t *pres = NULL;
	(void) nvlist_lookup_uint64_array(nvroot,
	ZPOOL_CONFIG_RAIDZ_EXPAND_STATS, (uint64_t **)&pres, &c);
	print_raidz_expand_status(zhp, pres);

	cbp->cb_namewidth = max_width(zhp, nvroot, 0, 0,
	cbp->cb_name_flags \| VDEV_NAME_TYPE_ID);
	if (cbp->cb_namewidth < 10)
	cbp->cb_namewidth = 10;

	color_start(ANSI_BOLD);
	(void) printf(gettext("config:\n\n"));
	(void) printf(gettext("\t%-*s %-8s %5s %5s %5s"),
	cbp->cb_namewidth, "NAME", "STATE", "READ", "WRITE",
	"CKSUM");
	color_end();

	if (cbp->cb_print_slow_ios) {
	printf_color(ANSI_BOLD, " %5s", gettext("SLOW"));
	}

	if (cbp->cb_print_power) {
	printf_color(ANSI_BOLD, " %5s", gettext("POWER"));
	}

	if (cbp->vcdl != NULL)
	print_cmd_columns(cbp->vcdl, 0);

	printf("\n");

	print_status_config(zhp, cbp, zpool_get_name(zhp), nvroot, 0,
	B_FALSE, NULL);

	print_class_vdevs(zhp, cbp, nvroot, VDEV_ALLOC_BIAS_DEDUP);
	print_class_vdevs(zhp, cbp, nvroot, VDEV_ALLOC_BIAS_SPECIAL);
	print_class_vdevs(zhp, cbp, nvroot, VDEV_ALLOC_CLASS_LOGS);

	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
	&l2cache, &nl2cache) == 0)
	print_l2cache(zhp, cbp, l2cache, nl2cache);

	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
	&spares, &nspares) == 0)
	print_spares(zhp, cbp, spares, nspares);

	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_ERRCOUNT,
	&nerr) == 0) {
	(void) printf("\n");
	if (nerr == 0) {
	(void) printf(gettext(
	"errors: No known data errors\n"));
	} else if (!cbp->cb_verbose) {
	color_start(ANSI_RED);
	(void) printf(gettext("errors: %llu data "
	"errors, use '-v' for a list\n"),
	(u_longlong_t)nerr);
	color_end();
	} else {
	print_error_log(zhp);
	}
	}

	if (cbp->cb_dedup_stats)
	print_dedup_stats(config);
	} else {
	(void) printf(gettext("config: The configuration cannot be "
	"determined.\n"));
	}

	return (0);
	}

	/*
	* zpool status [-c [script1,script2,...]] [-DegiLpPstvx] [--power] [-T d\|u] ...
	* [pool] [interval [count]]
	*
	* -c CMD For each vdev, run command CMD
	* -D Display dedup status (undocumented)
	* -e Display only unhealthy vdevs
	* -g Display guid for individual vdev name.
	* -i Display vdev initialization status.
	* -L Follow links when resolving vdev path name.
	* -p Display values in parsable (exact) format.
	* -P Display full path for vdev name.
	* -s Display slow IOs column.
	* -t Display vdev TRIM status.
	* -T Display a timestamp in date(1) or Unix format
	* -v Display complete error logs
	* -x Display only pools with potential problems
	* --power Display vdev enclosure slot power status
	*
	* Describes the health status of all pools or some subset.
	*/
	int
	zpool_do_status(int argc, char **argv)
	{
	int c;
	int ret;
	float interval = 0;
	unsigned long count = 0;
	status_cbdata_t cb = { 0 };
	char *cmd = NULL;

	struct option long_options[] = {
	{"power", no_argument, NULL, ZPOOL_OPTION_POWER},
	{0, 0, 0, 0}
	};

	/* check options */
	while ((c = getopt_long(argc, argv, "c:DegiLpPstT:vx", long_options,
	NULL)) != -1) {
	switch (c) {
	case 'c':
	if (cmd != NULL) {
	fprintf(stderr,
	gettext("Can't set -c flag twice\n"));
	exit(1);
	}

	if (getenv("ZPOOL_SCRIPTS_ENABLED") != NULL &&
	!libzfs_envvar_is_set("ZPOOL_SCRIPTS_ENABLED")) {
	fprintf(stderr, gettext(
	"Can't run -c, disabled by "
	"ZPOOL_SCRIPTS_ENABLED.\n"));
	exit(1);
	}

	if ((getuid() <= 0 \|\| geteuid() <= 0) &&
	!libzfs_envvar_is_set("ZPOOL_SCRIPTS_AS_ROOT")) {
	fprintf(stderr, gettext(
	"Can't run -c with root privileges "
	"unless ZPOOL_SCRIPTS_AS_ROOT is set.\n"));
	exit(1);
	}
	cmd = optarg;
	break;
	case 'D':
	cb.cb_dedup_stats = B_TRUE;
	break;
	case 'e':
	cb.cb_print_unhealthy = B_TRUE;
	break;
	case 'g':
	cb.cb_name_flags \|= VDEV_NAME_GUID;
	break;
	case 'i':
	cb.cb_print_vdev_init = B_TRUE;
	break;
	case 'L':
	cb.cb_name_flags \|= VDEV_NAME_FOLLOW_LINKS;
	break;
	case 'p':
	cb.cb_literal = B_TRUE;
	break;
	case 'P':
	cb.cb_name_flags \|= VDEV_NAME_PATH;
	break;
	case 's':
	cb.cb_print_slow_ios = B_TRUE;
	break;
	case 't':
	cb.cb_print_vdev_trim = B_TRUE;
	break;
	case 'T':
	get_timestamp_arg(*optarg);
	break;
	case 'v':
	cb.cb_verbose = B_TRUE;
	break;
	case 'x':
	cb.cb_explain = B_TRUE;
	break;
	case ZPOOL_OPTION_POWER:
	cb.cb_print_power = B_TRUE;
	break;
	case '?':
	if (optopt == 'c') {
	print_zpool_script_list("status");
	exit(0);
	} else {
	fprintf(stderr,
	gettext("invalid option '%c'\n"), optopt);
	}
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	get_interval_count(&argc, argv, &interval, &count);

	if (argc == 0)
	cb.cb_allpools = B_TRUE;

	cb.cb_first = B_TRUE;
	cb.cb_print_status = B_TRUE;

	for (;;) {
	if (timestamp_fmt != NODATE)
	print_timestamp(timestamp_fmt);

	if (cmd != NULL)
	cb.vcdl = all_pools_for_each_vdev_run(argc, argv, cmd,
	NULL, NULL, 0, 0);

	ret = for_each_pool(argc, argv, B_TRUE, NULL, ZFS_TYPE_POOL,
	cb.cb_literal, status_callback, &cb);

	if (cb.vcdl != NULL)
	free_vdev_cmd_data_list(cb.vcdl);
	if (argc == 0 && cb.cb_count == 0)
	(void) fprintf(stderr, gettext("no pools available\n"));
	else if (cb.cb_explain && cb.cb_first && cb.cb_allpools)
	(void) printf(gettext("all pools are healthy\n"));

	if (ret != 0)
	return (ret);

	if (interval == 0)
	break;

	if (count != 0 && --count == 0)
	break;

	(void) fflush(stdout);
	(void) fsleep(interval);
	}

	return (0);
	}

	typedef struct upgrade_cbdata {
	int cb_first;
	int cb_argc;
	uint64_t cb_version;
	char **cb_argv;
	} upgrade_cbdata_t;

	static int
	check_unsupp_fs(zfs_handle_t zhp, void unsupp_fs)
	{
	int zfs_version = (int)zfs_prop_get_int(zhp, ZFS_PROP_VERSION);
	int count = (int )unsupp_fs;

	if (zfs_version > ZPL_VERSION) {
	(void) printf(gettext("%s (v%d) is not supported by this "
	"implementation of ZFS.\n"),
	zfs_get_name(zhp), zfs_version);
	(*count)++;
	}

	zfs_iter_filesystems_v2(zhp, 0, check_unsupp_fs, unsupp_fs);

	zfs_close(zhp);

	return (0);
	}

	static int
	upgrade_version(zpool_handle_t *zhp, uint64_t version)
	{
	int ret;
	nvlist_t *config;
	uint64_t oldversion;
	int unsupp_fs = 0;

	config = zpool_get_config(zhp, NULL);
	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
	&oldversion) == 0);

	char compat[ZFS_MAXPROPLEN];
	if (zpool_get_prop(zhp, ZPOOL_PROP_COMPATIBILITY, compat,
	ZFS_MAXPROPLEN, NULL, B_FALSE) != 0)
	compat[0] = '\0';

	assert(SPA_VERSION_IS_SUPPORTED(oldversion));
	assert(oldversion < version);

	ret = zfs_iter_root(zpool_get_handle(zhp), check_unsupp_fs, &unsupp_fs);
	if (ret != 0)
	return (ret);

	if (unsupp_fs) {
	(void) fprintf(stderr, gettext("Upgrade not performed due "
	"to %d unsupported filesystems (max v%d).\n"),
	unsupp_fs, (int)ZPL_VERSION);
	return (1);
	}

	if (strcmp(compat, ZPOOL_COMPAT_LEGACY) == 0) {
	(void) fprintf(stderr, gettext("Upgrade not performed because "
	"'compatibility' property set to '"
	ZPOOL_COMPAT_LEGACY "'.\n"));
	return (1);
	}

	ret = zpool_upgrade(zhp, version);
	if (ret != 0)
	return (ret);

	if (version >= SPA_VERSION_FEATURES) {
	(void) printf(gettext("Successfully upgraded "
	"'%s' from version %llu to feature flags.\n"),
	zpool_get_name(zhp), (u_longlong_t)oldversion);
	} else {
	(void) printf(gettext("Successfully upgraded "
	"'%s' from version %llu to version %llu.\n"),
	zpool_get_name(zhp), (u_longlong_t)oldversion,
	(u_longlong_t)version);
	}

	return (0);
	}

	static int
	upgrade_enable_all(zpool_handle_t zhp, int countp)
	{
	int i, ret, count;
	boolean_t firstff = B_TRUE;
	nvlist_t *enabled = zpool_get_features(zhp);

	char compat[ZFS_MAXPROPLEN];
	if (zpool_get_prop(zhp, ZPOOL_PROP_COMPATIBILITY, compat,
	ZFS_MAXPROPLEN, NULL, B_FALSE) != 0)
	compat[0] = '\0';

	boolean_t requested_features[SPA_FEATURES];
	if (zpool_do_load_compat(compat, requested_features) !=
	ZPOOL_COMPATIBILITY_OK)
	return (-1);

	count = 0;
	for (i = 0; i < SPA_FEATURES; i++) {
	const char *fname = spa_feature_table[i].fi_uname;
	const char *fguid = spa_feature_table[i].fi_guid;

	if (!spa_feature_table[i].fi_zfs_mod_supported)
	continue;

	if (!nvlist_exists(enabled, fguid) && requested_features[i]) {
	char *propname;
	verify(-1 != asprintf(&propname, "feature@%s", fname));
	ret = zpool_set_prop(zhp, propname,
	ZFS_FEATURE_ENABLED);
	if (ret != 0) {
	free(propname);
	return (ret);
	}
	count++;

	if (firstff) {
	(void) printf(gettext("Enabled the "
	"following features on '%s':\n"),
	zpool_get_name(zhp));
	firstff = B_FALSE;
	}
	(void) printf(gettext(" %s\n"), fname);
	free(propname);
	}
	}

	if (countp != NULL)
	*countp = count;
	return (0);
	}

	static int
	upgrade_cb(zpool_handle_t zhp, void arg)
	{
	upgrade_cbdata_t *cbp = arg;
	nvlist_t *config;
	uint64_t version;
	boolean_t modified_pool = B_FALSE;
	int ret;

	config = zpool_get_config(zhp, NULL);
	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
	&version) == 0);

	assert(SPA_VERSION_IS_SUPPORTED(version));

	if (version < cbp->cb_version) {
	cbp->cb_first = B_FALSE;
	ret = upgrade_version(zhp, cbp->cb_version);
	if (ret != 0)
	return (ret);
	modified_pool = B_TRUE;

	/*
	* If they did "zpool upgrade -a", then we could
	* be doing ioctls to different pools. We need
	* to log this history once to each pool, and bypass
	* the normal history logging that happens in main().
	*/
	(void) zpool_log_history(g_zfs, history_str);
	log_history = B_FALSE;
	}

	if (cbp->cb_version >= SPA_VERSION_FEATURES) {
	int count;
	ret = upgrade_enable_all(zhp, &count);
	if (ret != 0)
	return (ret);

	if (count > 0) {
	cbp->cb_first = B_FALSE;
	modified_pool = B_TRUE;
	}
	}

	if (modified_pool) {
	(void) printf("\n");
	(void) after_zpool_upgrade(zhp);
	}

	return (0);
	}

	static int
	upgrade_list_older_cb(zpool_handle_t zhp, void arg)
	{
	upgrade_cbdata_t *cbp = arg;
	nvlist_t *config;
	uint64_t version;

	config = zpool_get_config(zhp, NULL);
	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
	&version) == 0);

	assert(SPA_VERSION_IS_SUPPORTED(version));

	if (version < SPA_VERSION_FEATURES) {
	if (cbp->cb_first) {
	(void) printf(gettext("The following pools are "
	"formatted with legacy version numbers and can\n"
	"be upgraded to use feature flags. After "
	"being upgraded, these pools\nwill no "
	"longer be accessible by software that does not "
	"support feature\nflags.\n\n"
	"Note that setting a pool's 'compatibility' "
	"feature to '" ZPOOL_COMPAT_LEGACY "' will\n"
	"inhibit upgrades.\n\n"));
	(void) printf(gettext("VER POOL\n"));
	(void) printf(gettext("--- ------------\n"));
	cbp->cb_first = B_FALSE;
	}

	(void) printf("%2llu %s\n", (u_longlong_t)version,
	zpool_get_name(zhp));
	}

	return (0);
	}

	static int
	upgrade_list_disabled_cb(zpool_handle_t zhp, void arg)
	{
	upgrade_cbdata_t *cbp = arg;
	nvlist_t *config;
	uint64_t version;

	config = zpool_get_config(zhp, NULL);
	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
	&version) == 0);

	if (version >= SPA_VERSION_FEATURES) {
	int i;
	boolean_t poolfirst = B_TRUE;
	nvlist_t *enabled = zpool_get_features(zhp);

	for (i = 0; i < SPA_FEATURES; i++) {
	const char *fguid = spa_feature_table[i].fi_guid;
	const char *fname = spa_feature_table[i].fi_uname;

	if (!spa_feature_table[i].fi_zfs_mod_supported)
	continue;

	if (!nvlist_exists(enabled, fguid)) {
	if (cbp->cb_first) {
	(void) printf(gettext("\nSome "
	"supported features are not "
	"enabled on the following pools. "
	"Once a\nfeature is enabled the "
	"pool may become incompatible with "
	"software\nthat does not support "
	"the feature. See "
	"zpool-features(7) for "
	"details.\n\n"
	"Note that the pool "
	"'compatibility' feature can be "
	"used to inhibit\nfeature "
	"upgrades.\n\n"));
	(void) printf(gettext("POOL "
	"FEATURE\n"));
	(void) printf(gettext("------"
	"---------\n"));
	cbp->cb_first = B_FALSE;
	}

	if (poolfirst) {
	(void) printf(gettext("%s\n"),
	zpool_get_name(zhp));
	poolfirst = B_FALSE;
	}

	(void) printf(gettext(" %s\n"), fname);
	}
	/*
	* If they did "zpool upgrade -a", then we could
	* be doing ioctls to different pools. We need
	* to log this history once to each pool, and bypass
	* the normal history logging that happens in main().
	*/
	(void) zpool_log_history(g_zfs, history_str);
	log_history = B_FALSE;
	}
	}

	return (0);
	}

	static int
	upgrade_one(zpool_handle_t zhp, void data)
	{
	boolean_t modified_pool = B_FALSE;
	upgrade_cbdata_t *cbp = data;
	uint64_t cur_version;
	int ret;

	if (strcmp("log", zpool_get_name(zhp)) == 0) {
	(void) fprintf(stderr, gettext("'log' is now a reserved word\n"
	"Pool 'log' must be renamed using export and import"
	" to upgrade.\n"));
	return (1);
	}

	cur_version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
	if (cur_version > cbp->cb_version) {
	(void) printf(gettext("Pool '%s' is already formatted "
	"using more current version '%llu'.\n\n"),
	zpool_get_name(zhp), (u_longlong_t)cur_version);
	return (0);
	}

	if (cbp->cb_version != SPA_VERSION && cur_version == cbp->cb_version) {
	(void) printf(gettext("Pool '%s' is already formatted "
	"using version %llu.\n\n"), zpool_get_name(zhp),
	(u_longlong_t)cbp->cb_version);
	return (0);
	}

	if (cur_version != cbp->cb_version) {
	modified_pool = B_TRUE;
	ret = upgrade_version(zhp, cbp->cb_version);
	if (ret != 0)
	return (ret);
	}

	if (cbp->cb_version >= SPA_VERSION_FEATURES) {
	int count = 0;
	ret = upgrade_enable_all(zhp, &count);
	if (ret != 0)
	return (ret);

	if (count != 0) {
	modified_pool = B_TRUE;
	} else if (cur_version == SPA_VERSION) {
	(void) printf(gettext("Pool '%s' already has all "
	"supported and requested features enabled.\n"),
	zpool_get_name(zhp));
	}
	}

	if (modified_pool) {
	(void) printf("\n");
	(void) after_zpool_upgrade(zhp);
	}

	return (0);
	}

	/*
	* zpool upgrade
	* zpool upgrade -v
	* zpool upgrade [-V version] <-a \| pool ...>
	*
	* With no arguments, display downrev'd ZFS pool available for upgrade.
	* Individual pools can be upgraded by specifying the pool, and '-a' will
	* upgrade all pools.
	*/
	int
	zpool_do_upgrade(int argc, char **argv)
	{
	int c;
	upgrade_cbdata_t cb = { 0 };
	int ret = 0;
	boolean_t showversions = B_FALSE;
	boolean_t upgradeall = B_FALSE;
	char *end;


	/* check options */
	while ((c = getopt(argc, argv, ":avV:")) != -1) {
	switch (c) {
	case 'a':
	upgradeall = B_TRUE;
	break;
	case 'v':
	showversions = B_TRUE;
	break;
	case 'V':
	cb.cb_version = strtoll(optarg, &end, 10);
	if (*end != '\0' \|\|
	!SPA_VERSION_IS_SUPPORTED(cb.cb_version)) {
	(void) fprintf(stderr,
	gettext("invalid version '%s'\n"), optarg);
	usage(B_FALSE);
	}
	break;
	case ':':
	(void) fprintf(stderr, gettext("missing argument for "
	"'%c' option\n"), optopt);
	usage(B_FALSE);
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	cb.cb_argc = argc;
	cb.cb_argv = argv;
	argc -= optind;
	argv += optind;

	if (cb.cb_version == 0) {
	cb.cb_version = SPA_VERSION;
	} else if (!upgradeall && argc == 0) {
	(void) fprintf(stderr, gettext("-V option is "
	"incompatible with other arguments\n"));
	usage(B_FALSE);
	}

	if (showversions) {
	if (upgradeall \|\| argc != 0) {
	(void) fprintf(stderr, gettext("-v option is "
	"incompatible with other arguments\n"));
	usage(B_FALSE);
	}
	} else if (upgradeall) {
	if (argc != 0) {
	(void) fprintf(stderr, gettext("-a option should not "
	"be used along with a pool name\n"));
	usage(B_FALSE);
	}
	}

	(void) printf("%s", gettext("This system supports ZFS pool feature "
	"flags.\n\n"));
	if (showversions) {
	int i;

	(void) printf(gettext("The following features are "
	"supported:\n\n"));
	(void) printf(gettext("FEAT DESCRIPTION\n"));
	(void) printf("----------------------------------------------"
	"---------------\n");
	for (i = 0; i < SPA_FEATURES; i++) {
	zfeature_info_t *fi = &spa_feature_table[i];
	if (!fi->fi_zfs_mod_supported)
	continue;
	const char *ro =
	(fi->fi_flags & ZFEATURE_FLAG_READONLY_COMPAT) ?
	" (read-only compatible)" : "";

	(void) printf("%-37s%s\n", fi->fi_uname, ro);
	(void) printf(" %s\n", fi->fi_desc);
	}
	(void) printf("\n");

	(void) printf(gettext("The following legacy versions are also "
	"supported:\n\n"));
	(void) printf(gettext("VER DESCRIPTION\n"));
	(void) printf("--- -----------------------------------------"
	"---------------\n");
	(void) printf(gettext(" 1 Initial ZFS version\n"));
	(void) printf(gettext(" 2 Ditto blocks "
	"(replicated metadata)\n"));
	(void) printf(gettext(" 3 Hot spares and double parity "
	"RAID-Z\n"));
	(void) printf(gettext(" 4 zpool history\n"));
	(void) printf(gettext(" 5 Compression using the gzip "
	"algorithm\n"));
	(void) printf(gettext(" 6 bootfs pool property\n"));
	(void) printf(gettext(" 7 Separate intent log devices\n"));
	(void) printf(gettext(" 8 Delegated administration\n"));
	(void) printf(gettext(" 9 refquota and refreservation "
	"properties\n"));
	(void) printf(gettext(" 10 Cache devices\n"));
	(void) printf(gettext(" 11 Improved scrub performance\n"));
	(void) printf(gettext(" 12 Snapshot properties\n"));
	(void) printf(gettext(" 13 snapused property\n"));
	(void) printf(gettext(" 14 passthrough-x aclinherit\n"));
	(void) printf(gettext(" 15 user/group space accounting\n"));
	(void) printf(gettext(" 16 stmf property support\n"));
	(void) printf(gettext(" 17 Triple-parity RAID-Z\n"));
	(void) printf(gettext(" 18 Snapshot user holds\n"));
	(void) printf(gettext(" 19 Log device removal\n"));
	(void) printf(gettext(" 20 Compression using zle "
	"(zero-length encoding)\n"));
	(void) printf(gettext(" 21 Deduplication\n"));
	(void) printf(gettext(" 22 Received properties\n"));
	(void) printf(gettext(" 23 Slim ZIL\n"));
	(void) printf(gettext(" 24 System attributes\n"));
	(void) printf(gettext(" 25 Improved scrub stats\n"));
	(void) printf(gettext(" 26 Improved snapshot deletion "
	"performance\n"));
	(void) printf(gettext(" 27 Improved snapshot creation "
	"performance\n"));
	(void) printf(gettext(" 28 Multiple vdev replacements\n"));
	(void) printf(gettext("\nFor more information on a particular "
	"version, including supported releases,\n"));
	(void) printf(gettext("see the ZFS Administration Guide.\n\n"));
	} else if (argc == 0 && upgradeall) {
	cb.cb_first = B_TRUE;
	ret = zpool_iter(g_zfs, upgrade_cb, &cb);
	if (ret == 0 && cb.cb_first) {
	if (cb.cb_version == SPA_VERSION) {
	(void) printf(gettext("All pools are already "
	"formatted using feature flags.\n\n"));
	(void) printf(gettext("Every feature flags "
	"pool already has all supported and "
	"requested features enabled.\n"));
	} else {
	(void) printf(gettext("All pools are already "
	"formatted with version %llu or higher.\n"),
	(u_longlong_t)cb.cb_version);
	}
	}
	} else if (argc == 0) {
	cb.cb_first = B_TRUE;
	ret = zpool_iter(g_zfs, upgrade_list_older_cb, &cb);
	assert(ret == 0);

	if (cb.cb_first) {
	(void) printf(gettext("All pools are formatted "
	"using feature flags.\n\n"));
	} else {
	(void) printf(gettext("\nUse 'zpool upgrade -v' "
	"for a list of available legacy versions.\n"));
	}

	cb.cb_first = B_TRUE;
	ret = zpool_iter(g_zfs, upgrade_list_disabled_cb, &cb);
	assert(ret == 0);

	if (cb.cb_first) {
	(void) printf(gettext("Every feature flags pool has "
	"all supported and requested features enabled.\n"));
	} else {
	(void) printf(gettext("\n"));
	}
	} else {
	ret = for_each_pool(argc, argv, B_FALSE, NULL, ZFS_TYPE_POOL,
	B_FALSE, upgrade_one, &cb);
	}

	return (ret);
	}

	typedef struct hist_cbdata {
	boolean_t first;
	boolean_t longfmt;
	boolean_t internal;
	} hist_cbdata_t;

	static void
	print_history_records(nvlist_t nvhis, hist_cbdata_t cb)
	{
	nvlist_t **records;
	uint_t numrecords;
	int i;

	verify(nvlist_lookup_nvlist_array(nvhis, ZPOOL_HIST_RECORD,
	&records, &numrecords) == 0);
	for (i = 0; i < numrecords; i++) {
	nvlist_t *rec = records[i];
	char tbuf[64] = "";

	if (nvlist_exists(rec, ZPOOL_HIST_TIME)) {
	time_t tsec;
	struct tm t;

	tsec = fnvlist_lookup_uint64(records[i],
	ZPOOL_HIST_TIME);
	(void) localtime_r(&tsec, &t);
	(void) strftime(tbuf, sizeof (tbuf), "%F.%T", &t);
	}

	if (nvlist_exists(rec, ZPOOL_HIST_ELAPSED_NS)) {
	uint64_t elapsed_ns = fnvlist_lookup_int64(records[i],
	ZPOOL_HIST_ELAPSED_NS);
	(void) snprintf(tbuf + strlen(tbuf),
	sizeof (tbuf) - strlen(tbuf),
	" (%lldms)", (long long)elapsed_ns / 1000 / 1000);
	}

	if (nvlist_exists(rec, ZPOOL_HIST_CMD)) {
	(void) printf("%s %s", tbuf,
	fnvlist_lookup_string(rec, ZPOOL_HIST_CMD));
	} else if (nvlist_exists(rec, ZPOOL_HIST_INT_EVENT)) {
	int ievent =
	fnvlist_lookup_uint64(rec, ZPOOL_HIST_INT_EVENT);
	if (!cb->internal)
	continue;
	if (ievent >= ZFS_NUM_LEGACY_HISTORY_EVENTS) {
	(void) printf("%s unrecognized record:\n",
	tbuf);
	dump_nvlist(rec, 4);
	continue;
	}
	(void) printf("%s [internal %s txg:%lld] %s", tbuf,
	zfs_history_event_names[ievent],
	(longlong_t)fnvlist_lookup_uint64(
	rec, ZPOOL_HIST_TXG),
	fnvlist_lookup_string(rec, ZPOOL_HIST_INT_STR));
	} else if (nvlist_exists(rec, ZPOOL_HIST_INT_NAME)) {
	if (!cb->internal)
	continue;
	(void) printf("%s [txg:%lld] %s", tbuf,
	(longlong_t)fnvlist_lookup_uint64(
	rec, ZPOOL_HIST_TXG),
	fnvlist_lookup_string(rec, ZPOOL_HIST_INT_NAME));
	if (nvlist_exists(rec, ZPOOL_HIST_DSNAME)) {
	(void) printf(" %s (%llu)",
	fnvlist_lookup_string(rec,
	ZPOOL_HIST_DSNAME),
	(u_longlong_t)fnvlist_lookup_uint64(rec,
	ZPOOL_HIST_DSID));
	}
	(void) printf(" %s", fnvlist_lookup_string(rec,
	ZPOOL_HIST_INT_STR));
	} else if (nvlist_exists(rec, ZPOOL_HIST_IOCTL)) {
	if (!cb->internal)
	continue;
	(void) printf("%s ioctl %s\n", tbuf,
	fnvlist_lookup_string(rec, ZPOOL_HIST_IOCTL));
	if (nvlist_exists(rec, ZPOOL_HIST_INPUT_NVL)) {
	(void) printf(" input:\n");
	dump_nvlist(fnvlist_lookup_nvlist(rec,
	ZPOOL_HIST_INPUT_NVL), 8);
	}
	if (nvlist_exists(rec, ZPOOL_HIST_OUTPUT_NVL)) {
	(void) printf(" output:\n");
	dump_nvlist(fnvlist_lookup_nvlist(rec,
	ZPOOL_HIST_OUTPUT_NVL), 8);
	}
	if (nvlist_exists(rec, ZPOOL_HIST_OUTPUT_SIZE)) {
	(void) printf(" output nvlist omitted; "
	"original size: %lldKB\n",
	(longlong_t)fnvlist_lookup_int64(rec,
	ZPOOL_HIST_OUTPUT_SIZE) / 1024);
	}
	if (nvlist_exists(rec, ZPOOL_HIST_ERRNO)) {
	(void) printf(" errno: %lld\n",
	(longlong_t)fnvlist_lookup_int64(rec,
	ZPOOL_HIST_ERRNO));
	}
	} else {
	if (!cb->internal)
	continue;
	(void) printf("%s unrecognized record:\n", tbuf);
	dump_nvlist(rec, 4);
	}

	if (!cb->longfmt) {
	(void) printf("\n");
	continue;
	}
	(void) printf(" [");
	if (nvlist_exists(rec, ZPOOL_HIST_WHO)) {
	uid_t who = fnvlist_lookup_uint64(rec, ZPOOL_HIST_WHO);
	struct passwd *pwd = getpwuid(who);
	(void) printf("user %d ", (int)who);
	if (pwd != NULL)
	(void) printf("(%s) ", pwd->pw_name);
	}
	if (nvlist_exists(rec, ZPOOL_HIST_HOST)) {
	(void) printf("on %s",
	fnvlist_lookup_string(rec, ZPOOL_HIST_HOST));
	}
	if (nvlist_exists(rec, ZPOOL_HIST_ZONE)) {
	(void) printf(":%s",
	fnvlist_lookup_string(rec, ZPOOL_HIST_ZONE));
	}

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

	/*
	* Print out the command history for a specific pool.
	*/
	static int
	get_history_one(zpool_handle_t zhp, void data)
	{
	nvlist_t *nvhis;
	int ret;
	hist_cbdata_t cb = (hist_cbdata_t )data;
	uint64_t off = 0;
	boolean_t eof = B_FALSE;

	cb->first = B_FALSE;

	(void) printf(gettext("History for '%s':\n"), zpool_get_name(zhp));

	while (!eof) {
	if ((ret = zpool_get_history(zhp, &nvhis, &off, &eof)) != 0)
	return (ret);

	print_history_records(nvhis, cb);
	nvlist_free(nvhis);
	}
	(void) printf("\n");

	return (ret);
	}

	/*
	* zpool history <pool>
	*
	* Displays the history of commands that modified pools.
	*/
	int
	zpool_do_history(int argc, char **argv)
	{
	hist_cbdata_t cbdata = { 0 };
	int ret;
	int c;

	cbdata.first = B_TRUE;
	/* check options */
	while ((c = getopt(argc, argv, "li")) != -1) {
	switch (c) {
	case 'l':
	cbdata.longfmt = B_TRUE;
	break;
	case 'i':
	cbdata.internal = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}
	argc -= optind;
	argv += optind;

	ret = for_each_pool(argc, argv, B_FALSE, NULL, ZFS_TYPE_POOL,
	B_FALSE, get_history_one, &cbdata);

	if (argc == 0 && cbdata.first == B_TRUE) {
	(void) fprintf(stderr, gettext("no pools available\n"));
	return (0);
	}

	return (ret);
	}

	typedef struct ev_opts {
	int verbose;
	int scripted;
	int follow;
	int clear;
	char poolname[ZFS_MAX_DATASET_NAME_LEN];
	} ev_opts_t;

	static void
	zpool_do_events_short(nvlist_t nvl, ev_opts_t opts)
	{
	char ctime_str[26], str[32];
	const char *ptr;
	int64_t *tv;
	uint_t n;

	verify(nvlist_lookup_int64_array(nvl, FM_EREPORT_TIME, &tv, &n) == 0);
	memset(str, ' ', 32);
	(void) ctime_r((const time_t *)&tv[0], ctime_str);
	(void) memcpy(str, ctime_str+4, 6); /* 'Jun 30' */
	(void) memcpy(str+7, ctime_str+20, 4); /* '1993' */
	(void) memcpy(str+12, ctime_str+11, 8); /* '21:49:08' */
	(void) sprintf(str+20, ".%09lld", (longlong_t)tv[1]); /* '.123456789' */
	if (opts->scripted)
	(void) printf(gettext("%s\t"), str);
	else
	(void) printf(gettext("%s "), str);

	verify(nvlist_lookup_string(nvl, FM_CLASS, &ptr) == 0);
	(void) printf(gettext("%s\n"), ptr);
	}

	static void
	zpool_do_events_nvprint(nvlist_t *nvl, int depth)
	{
	nvpair_t *nvp;

	for (nvp = nvlist_next_nvpair(nvl, NULL);
	nvp != NULL; nvp = nvlist_next_nvpair(nvl, nvp)) {

	data_type_t type = nvpair_type(nvp);
	const char *name = nvpair_name(nvp);

	boolean_t b;
	uint8_t i8;
	uint16_t i16;
	uint32_t i32;
	uint64_t i64;
	const char *str;
	nvlist_t *cnv;

	printf(gettext("%*s%s = "), depth, "", name);

	switch (type) {
	case DATA_TYPE_BOOLEAN:
	printf(gettext("%s"), "1");
	break;

	case DATA_TYPE_BOOLEAN_VALUE:
	(void) nvpair_value_boolean_value(nvp, &b);
	printf(gettext("%s"), b ? "1" : "0");
	break;

	case DATA_TYPE_BYTE:
	(void) nvpair_value_byte(nvp, &i8);
	printf(gettext("0x%x"), i8);
	break;

	case DATA_TYPE_INT8:
	(void) nvpair_value_int8(nvp, (void *)&i8);
	printf(gettext("0x%x"), i8);
	break;

	case DATA_TYPE_UINT8:
	(void) nvpair_value_uint8(nvp, &i8);
	printf(gettext("0x%x"), i8);
	break;

	case DATA_TYPE_INT16:
	(void) nvpair_value_int16(nvp, (void *)&i16);
	printf(gettext("0x%x"), i16);
	break;

	case DATA_TYPE_UINT16:
	(void) nvpair_value_uint16(nvp, &i16);
	printf(gettext("0x%x"), i16);
	break;

	case DATA_TYPE_INT32:
	(void) nvpair_value_int32(nvp, (void *)&i32);
	printf(gettext("0x%x"), i32);
	break;

	case DATA_TYPE_UINT32:
	(void) nvpair_value_uint32(nvp, &i32);
	printf(gettext("0x%x"), i32);
	break;

	case DATA_TYPE_INT64:
	(void) nvpair_value_int64(nvp, (void *)&i64);
	printf(gettext("0x%llx"), (u_longlong_t)i64);
	break;

	case DATA_TYPE_UINT64:
	(void) nvpair_value_uint64(nvp, &i64);
	/*
	* translate vdev state values to readable
	* strings to aide zpool events consumers
	*/
	if (strcmp(name,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_STATE) == 0 \|\|
	strcmp(name,
	FM_EREPORT_PAYLOAD_ZFS_VDEV_LASTSTATE) == 0) {
	printf(gettext("\"%s\" (0x%llx)"),
	zpool_state_to_name(i64, VDEV_AUX_NONE),
	(u_longlong_t)i64);
	} else {
	printf(gettext("0x%llx"), (u_longlong_t)i64);
	}
	break;

	case DATA_TYPE_HRTIME:
	(void) nvpair_value_hrtime(nvp, (void *)&i64);
	printf(gettext("0x%llx"), (u_longlong_t)i64);
	break;

	case DATA_TYPE_STRING:
	(void) nvpair_value_string(nvp, &str);
	printf(gettext("\"%s\""), str ? str : "<NULL>");
	break;

	case DATA_TYPE_NVLIST:
	printf(gettext("(embedded nvlist)\n"));
	(void) nvpair_value_nvlist(nvp, &cnv);
	zpool_do_events_nvprint(cnv, depth + 8);
	printf(gettext("%*s(end %s)"), depth, "", name);
	break;

	case DATA_TYPE_NVLIST_ARRAY: {
	nvlist_t **val;
	uint_t i, nelem;

	(void) nvpair_value_nvlist_array(nvp, &val, &nelem);
	printf(gettext("(%d embedded nvlists)\n"), nelem);
	for (i = 0; i < nelem; i++) {
	printf(gettext("%*s%s[%d] = %s\n"),
	depth, "", name, i, "(embedded nvlist)");
	zpool_do_events_nvprint(val[i], depth + 8);
	printf(gettext("%*s(end %s[%i])\n"),
	depth, "", name, i);
	}
	printf(gettext("%*s(end %s)\n"), depth, "", name);
	}
	break;

	case DATA_TYPE_INT8_ARRAY: {
	int8_t *val;
	uint_t i, nelem;

	(void) nvpair_value_int8_array(nvp, &val, &nelem);
	for (i = 0; i < nelem; i++)
	printf(gettext("0x%x "), val[i]);

	break;
	}

	case DATA_TYPE_UINT8_ARRAY: {
	uint8_t *val;
	uint_t i, nelem;

	(void) nvpair_value_uint8_array(nvp, &val, &nelem);
	for (i = 0; i < nelem; i++)
	printf(gettext("0x%x "), val[i]);

	break;
	}

	case DATA_TYPE_INT16_ARRAY: {
	int16_t *val;
	uint_t i, nelem;

	(void) nvpair_value_int16_array(nvp, &val, &nelem);
	for (i = 0; i < nelem; i++)
	printf(gettext("0x%x "), val[i]);

	break;
	}

	case DATA_TYPE_UINT16_ARRAY: {
	uint16_t *val;
	uint_t i, nelem;

	(void) nvpair_value_uint16_array(nvp, &val, &nelem);
	for (i = 0; i < nelem; i++)
	printf(gettext("0x%x "), val[i]);

	break;
	}

	case DATA_TYPE_INT32_ARRAY: {
	int32_t *val;
	uint_t i, nelem;

	(void) nvpair_value_int32_array(nvp, &val, &nelem);
	for (i = 0; i < nelem; i++)
	printf(gettext("0x%x "), val[i]);

	break;
	}

	case DATA_TYPE_UINT32_ARRAY: {
	uint32_t *val;
	uint_t i, nelem;

	(void) nvpair_value_uint32_array(nvp, &val, &nelem);
	for (i = 0; i < nelem; i++)
	printf(gettext("0x%x "), val[i]);

	break;
	}

	case DATA_TYPE_INT64_ARRAY: {
	int64_t *val;
	uint_t i, nelem;

	(void) nvpair_value_int64_array(nvp, &val, &nelem);
	for (i = 0; i < nelem; i++)
	printf(gettext("0x%llx "),
	(u_longlong_t)val[i]);

	break;
	}

	case DATA_TYPE_UINT64_ARRAY: {
	uint64_t *val;
	uint_t i, nelem;

	(void) nvpair_value_uint64_array(nvp, &val, &nelem);
	for (i = 0; i < nelem; i++)
	printf(gettext("0x%llx "),
	(u_longlong_t)val[i]);

	break;
	}

	case DATA_TYPE_STRING_ARRAY: {
	const char **str;
	uint_t i, nelem;

	(void) nvpair_value_string_array(nvp, &str, &nelem);
	for (i = 0; i < nelem; i++)
	printf(gettext("\"%s\" "),
	str[i] ? str[i] : "<NULL>");

	break;
	}

	case DATA_TYPE_BOOLEAN_ARRAY:
	case DATA_TYPE_BYTE_ARRAY:
	case DATA_TYPE_DOUBLE:
	case DATA_TYPE_DONTCARE:
	case DATA_TYPE_UNKNOWN:
	printf(gettext("<unknown>"));
	break;
	}

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

	static int
	zpool_do_events_next(ev_opts_t *opts)
	{
	nvlist_t *nvl;
	int zevent_fd, ret, dropped;
	const char *pool;

	zevent_fd = open(ZFS_DEV, O_RDWR);
	VERIFY(zevent_fd >= 0);

	if (!opts->scripted)
	(void) printf(gettext("%-30s %s\n"), "TIME", "CLASS");

	while (1) {
	ret = zpool_events_next(g_zfs, &nvl, &dropped,
	(opts->follow ? ZEVENT_NONE : ZEVENT_NONBLOCK), zevent_fd);
	if (ret \|\| nvl == NULL)
	break;

	if (dropped > 0)
	(void) printf(gettext("dropped %d events\n"), dropped);

	if (strlen(opts->poolname) > 0 &&
	nvlist_lookup_string(nvl, FM_FMRI_ZFS_POOL, &pool) == 0 &&
	strcmp(opts->poolname, pool) != 0)
	continue;

	zpool_do_events_short(nvl, opts);

	if (opts->verbose) {
	zpool_do_events_nvprint(nvl, 8);
	printf(gettext("\n"));
	}
	(void) fflush(stdout);

	nvlist_free(nvl);
	}

	VERIFY(0 == close(zevent_fd));

	return (ret);
	}

	static int
	zpool_do_events_clear(void)
	{
	int count, ret;

	ret = zpool_events_clear(g_zfs, &count);
	if (!ret)
	(void) printf(gettext("cleared %d events\n"), count);

	return (ret);
	}

	/*
	* zpool events [-vHf [pool] \| -c]
	*
	* Displays events logs by ZFS.
	*/
	int
	zpool_do_events(int argc, char **argv)
	{
	ev_opts_t opts = { 0 };
	int ret;
	int c;

	/* check options */
	while ((c = getopt(argc, argv, "vHfc")) != -1) {
	switch (c) {
	case 'v':
	opts.verbose = 1;
	break;
	case 'H':
	opts.scripted = 1;
	break;
	case 'f':
	opts.follow = 1;
	break;
	case 'c':
	opts.clear = 1;
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}
	argc -= optind;
	argv += optind;

	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	} else if (argc == 1) {
	(void) strlcpy(opts.poolname, argv[0], sizeof (opts.poolname));
	if (!zfs_name_valid(opts.poolname, ZFS_TYPE_POOL)) {
	(void) fprintf(stderr,
	gettext("invalid pool name '%s'\n"), opts.poolname);
	usage(B_FALSE);
	}
	}

	if ((argc == 1 \|\| opts.verbose \|\| opts.scripted \|\| opts.follow) &&
	opts.clear) {
	(void) fprintf(stderr,
	gettext("invalid options combined with -c\n"));
	usage(B_FALSE);
	}

	if (opts.clear)
	ret = zpool_do_events_clear();
	else
	ret = zpool_do_events_next(&opts);

	return (ret);
	}

	static int
	get_callback_vdev(zpool_handle_t zhp, char vdevname, void *data)
	{
	zprop_get_cbdata_t cbp = (zprop_get_cbdata_t )data;
	char value[ZFS_MAXPROPLEN];
	zprop_source_t srctype;

	for (zprop_list_t *pl = cbp->cb_proplist; pl != NULL;
	pl = pl->pl_next) {
	char *prop_name;
	/*
	* If the first property is pool name, it is a special
	* placeholder that we can skip. This will also skip
	* over the name property when 'all' is specified.
	*/
	if (pl->pl_prop == ZPOOL_PROP_NAME &&
	pl == cbp->cb_proplist)
	continue;

	if (pl->pl_prop == ZPROP_INVAL) {
	prop_name = pl->pl_user_prop;
	} else {
	prop_name = (char *)vdev_prop_to_name(pl->pl_prop);
	}
	if (zpool_get_vdev_prop(zhp, vdevname, pl->pl_prop,
	prop_name, value, sizeof (value), &srctype,
	cbp->cb_literal) == 0) {
	zprop_print_one_property(vdevname, cbp, prop_name,
	value, srctype, NULL, NULL);
	}
	}

	return (0);
	}

	static int
	get_callback_vdev_cb(void zhp_data, nvlist_t nv, void *data)
	{
	zpool_handle_t *zhp = zhp_data;
	zprop_get_cbdata_t cbp = (zprop_get_cbdata_t )data;
	char *vdevname;
	const char *type;
	int ret;

	/*
	* zpool_vdev_name() transforms the root vdev name (i.e., root-0) to the
	* pool name for display purposes, which is not desired. Fallback to
	* zpool_vdev_name() when not dealing with the root vdev.
	*/
	type = fnvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE);
	if (zhp != NULL && strcmp(type, "root") == 0)
	vdevname = strdup("root-0");
	else
	vdevname = zpool_vdev_name(g_zfs, zhp, nv,
	cbp->cb_vdevs.cb_name_flags);

	(void) vdev_expand_proplist(zhp, vdevname, &cbp->cb_proplist);

	ret = get_callback_vdev(zhp, vdevname, data);

	free(vdevname);

	return (ret);
	}

	static int
	get_callback(zpool_handle_t zhp, void data)
	{
	zprop_get_cbdata_t cbp = (zprop_get_cbdata_t )data;
	char value[ZFS_MAXPROPLEN];
	zprop_source_t srctype;
	zprop_list_t *pl;
	int vid;

	if (cbp->cb_type == ZFS_TYPE_VDEV) {
	if (strcmp(cbp->cb_vdevs.cb_names[0], "all-vdevs") == 0) {
	for_each_vdev(zhp, get_callback_vdev_cb, data);
	} else {
	/* Adjust column widths for vdev properties */
	for (vid = 0; vid < cbp->cb_vdevs.cb_names_count;
	vid++) {
	vdev_expand_proplist(zhp,
	cbp->cb_vdevs.cb_names[vid],
	&cbp->cb_proplist);
	}
	/* Display the properties */
	for (vid = 0; vid < cbp->cb_vdevs.cb_names_count;
	vid++) {
	get_callback_vdev(zhp,
	cbp->cb_vdevs.cb_names[vid], data);
	}
	}
	} else {
	assert(cbp->cb_type == ZFS_TYPE_POOL);
	for (pl = cbp->cb_proplist; pl != NULL; pl = pl->pl_next) {
	/*
	* Skip the special fake placeholder. This will also
	* skip over the name property when 'all' is specified.
	*/
	if (pl->pl_prop == ZPOOL_PROP_NAME &&
	pl == cbp->cb_proplist)
	continue;

	if (pl->pl_prop == ZPROP_INVAL &&
	zfs_prop_user(pl->pl_user_prop)) {
	srctype = ZPROP_SRC_LOCAL;

	if (zpool_get_userprop(zhp, pl->pl_user_prop,
	value, sizeof (value), &srctype) != 0)
	continue;

	zprop_print_one_property(zpool_get_name(zhp),
	cbp, pl->pl_user_prop, value, srctype,
	NULL, NULL);
	} else if (pl->pl_prop == ZPROP_INVAL &&
	(zpool_prop_feature(pl->pl_user_prop) \|\|
	zpool_prop_unsupported(pl->pl_user_prop))) {
	srctype = ZPROP_SRC_LOCAL;

	if (zpool_prop_get_feature(zhp,
	pl->pl_user_prop, value,
	sizeof (value)) == 0) {
	zprop_print_one_property(
	zpool_get_name(zhp), cbp,
	pl->pl_user_prop, value, srctype,
	NULL, NULL);
	}
	} else {
	if (zpool_get_prop(zhp, pl->pl_prop, value,
	sizeof (value), &srctype,
	cbp->cb_literal) != 0)
	continue;

	zprop_print_one_property(zpool_get_name(zhp),
	cbp, zpool_prop_to_name(pl->pl_prop),
	value, srctype, NULL, NULL);
	}
	}
	}

	return (0);
	}

	/*
	* zpool get [-Hp] [-o "all" \| field[,...]] <"all" \| property[,...]> <pool> ...
	*
	* -H Scripted mode. Don't display headers, and separate properties
	* by a single tab.
	* -o List of columns to display. Defaults to
	* "name,property,value,source".
	* -p Display values in parsable (exact) format.
	*
	* Get properties of pools in the system. Output space statistics
	* for each one as well as other attributes.
	*/
	int
	zpool_do_get(int argc, char **argv)
	{
	zprop_get_cbdata_t cb = { 0 };
	zprop_list_t fake_name = { 0 };
	int ret;
	int c, i;
	char *propstr = NULL;
	char *vdev = NULL;

	cb.cb_first = B_TRUE;

	/*
	* Set up default columns and sources.
	*/
	cb.cb_sources = ZPROP_SRC_ALL;
	cb.cb_columns[0] = GET_COL_NAME;
	cb.cb_columns[1] = GET_COL_PROPERTY;
	cb.cb_columns[2] = GET_COL_VALUE;
	cb.cb_columns[3] = GET_COL_SOURCE;
	cb.cb_type = ZFS_TYPE_POOL;
	cb.cb_vdevs.cb_name_flags \|= VDEV_NAME_TYPE_ID;
	current_prop_type = cb.cb_type;

	/* check options */
	while ((c = getopt(argc, argv, ":Hpo:")) != -1) {
	switch (c) {
	case 'p':
	cb.cb_literal = B_TRUE;
	break;
	case 'H':
	cb.cb_scripted = B_TRUE;
	break;
	case 'o':
	memset(&cb.cb_columns, 0, sizeof (cb.cb_columns));
	i = 0;

	for (char *tok; (tok = strsep(&optarg, ",")); ) {
	static const char *const col_opts[] =
	{ "name", "property", "value", "source",
	"all" };
	static const zfs_get_column_t col_cols[] =
	{ GET_COL_NAME, GET_COL_PROPERTY, GET_COL_VALUE,
	GET_COL_SOURCE };

	if (i == ZFS_GET_NCOLS - 1) {
	(void) fprintf(stderr, gettext("too "
	"many fields given to -o "
	"option\n"));
	usage(B_FALSE);
	}

	for (c = 0; c < ARRAY_SIZE(col_opts); ++c)
	if (strcmp(tok, col_opts[c]) == 0)
	goto found;

	(void) fprintf(stderr,
	gettext("invalid column name '%s'\n"), tok);
	usage(B_FALSE);

	found:
	if (c >= 4) {
	if (i > 0) {
	(void) fprintf(stderr,
	gettext("\"all\" conflicts "
	"with specific fields "
	"given to -o option\n"));
	usage(B_FALSE);
	}

	memcpy(cb.cb_columns, col_cols,
	sizeof (col_cols));
	i = ZFS_GET_NCOLS - 1;
	} else
	cb.cb_columns[i++] = col_cols[c];
	}
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing property "
	"argument\n"));
	usage(B_FALSE);
	}

	/* Properties list is needed later by zprop_get_list() */
	propstr = argv[0];

	argc--;
	argv++;

	if (argc == 0) {
	/* No args, so just print the defaults. */
	} else if (are_all_pools(argc, argv)) {
	/* All the args are pool names */
	} else if (are_all_pools(1, argv)) {
	/* The first arg is a pool name */
	if ((argc == 2 && strcmp(argv[1], "all-vdevs") == 0) \|\|
	(argc == 2 && strcmp(argv[1], "root") == 0) \|\|
	are_vdevs_in_pool(argc - 1, argv + 1, argv[0],
	&cb.cb_vdevs)) {

	if (strcmp(argv[1], "root") == 0)
	vdev = strdup("root-0");
	else
	vdev = strdup(argv[1]);

	/* ... and the rest are vdev names */
	cb.cb_vdevs.cb_names = &vdev;
	cb.cb_vdevs.cb_names_count = argc - 1;
	cb.cb_type = ZFS_TYPE_VDEV;
	argc = 1; /* One pool to process */
	} else {
	fprintf(stderr, gettext("Expected a list of vdevs in"
	" \"%s\", but got:\n"), argv[0]);
	error_list_unresolved_vdevs(argc - 1, argv + 1,
	argv[0], &cb.cb_vdevs);
	fprintf(stderr, "\n");
	usage(B_FALSE);
	return (1);
	}
	} else {
	/*
	* The first arg isn't the name of a valid pool.
	*/
	fprintf(stderr, gettext("Cannot get properties of %s: "
	"no such pool available.\n"), argv[0]);
	return (1);
	}

	if (zprop_get_list(g_zfs, propstr, &cb.cb_proplist,
	cb.cb_type) != 0) {
	/* Use correct list of valid properties (pool or vdev) */
	current_prop_type = cb.cb_type;
	usage(B_FALSE);
	}

	if (cb.cb_proplist != NULL) {
	fake_name.pl_prop = ZPOOL_PROP_NAME;
	fake_name.pl_width = strlen(gettext("NAME"));
	fake_name.pl_next = cb.cb_proplist;
	cb.cb_proplist = &fake_name;
	}

	ret = for_each_pool(argc, argv, B_TRUE, &cb.cb_proplist, cb.cb_type,
	cb.cb_literal, get_callback, &cb);

	if (cb.cb_proplist == &fake_name)
	zprop_free_list(fake_name.pl_next);
	else
	zprop_free_list(cb.cb_proplist);

	if (vdev != NULL)
	free(vdev);

	return (ret);
	}

	typedef struct set_cbdata {
	char *cb_propname;
	char *cb_value;
	zfs_type_t cb_type;
	vdev_cbdata_t cb_vdevs;
	boolean_t cb_any_successful;
	} set_cbdata_t;

	static int
	set_pool_callback(zpool_handle_t zhp, set_cbdata_t cb)
	{
	int error;

	/* Check if we have out-of-bounds features */
	if (strcmp(cb->cb_propname, ZPOOL_CONFIG_COMPATIBILITY) == 0) {
	boolean_t features[SPA_FEATURES];
	if (zpool_do_load_compat(cb->cb_value, features) !=
	ZPOOL_COMPATIBILITY_OK)
	return (-1);

	nvlist_t *enabled = zpool_get_features(zhp);
	spa_feature_t i;
	for (i = 0; i < SPA_FEATURES; i++) {
	const char *fguid = spa_feature_table[i].fi_guid;
	if (nvlist_exists(enabled, fguid) && !features[i])
	break;
	}
	if (i < SPA_FEATURES)
	(void) fprintf(stderr, gettext("Warning: one or "
	"more features already enabled on pool '%s'\n"
	"are not present in this compatibility set.\n"),
	zpool_get_name(zhp));
	}

	/* if we're setting a feature, check it's in compatibility set */
	if (zpool_prop_feature(cb->cb_propname) &&
	strcmp(cb->cb_value, ZFS_FEATURE_ENABLED) == 0) {
	char *fname = strchr(cb->cb_propname, '@') + 1;
	spa_feature_t f;

	if (zfeature_lookup_name(fname, &f) == 0) {
	char compat[ZFS_MAXPROPLEN];
	if (zpool_get_prop(zhp, ZPOOL_PROP_COMPATIBILITY,
	compat, ZFS_MAXPROPLEN, NULL, B_FALSE) != 0)
	compat[0] = '\0';

	boolean_t features[SPA_FEATURES];
	if (zpool_do_load_compat(compat, features) !=
	ZPOOL_COMPATIBILITY_OK) {
	(void) fprintf(stderr, gettext("Error: "
	"cannot enable feature '%s' on pool '%s'\n"
	"because the pool's 'compatibility' "
	"property cannot be parsed.\n"),
	fname, zpool_get_name(zhp));
	return (-1);
	}

	if (!features[f]) {
	(void) fprintf(stderr, gettext("Error: "
	"cannot enable feature '%s' on pool '%s'\n"
	"as it is not specified in this pool's "
	"current compatibility set.\n"
	"Consider setting 'compatibility' to a "
	"less restrictive set, or to 'off'.\n"),
	fname, zpool_get_name(zhp));
	return (-1);
	}
	}
	}

	error = zpool_set_prop(zhp, cb->cb_propname, cb->cb_value);

	return (error);
	}

	static int
	set_callback(zpool_handle_t zhp, void data)
	{
	int error;
	set_cbdata_t cb = (set_cbdata_t )data;

	if (cb->cb_type == ZFS_TYPE_VDEV) {
	error = zpool_set_vdev_prop(zhp, *cb->cb_vdevs.cb_names,
	cb->cb_propname, cb->cb_value);
	} else {
	assert(cb->cb_type == ZFS_TYPE_POOL);
	error = set_pool_callback(zhp, cb);
	}

	cb->cb_any_successful = !error;
	return (error);
	}

	int
	zpool_do_set(int argc, char **argv)
	{
	set_cbdata_t cb = { 0 };
	int error;
	char *vdev = NULL;

	current_prop_type = ZFS_TYPE_POOL;
	if (argc > 1 && argv[1][0] == '-') {
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	argv[1][1]);
	usage(B_FALSE);
	}

	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing property=value "
	"argument\n"));
	usage(B_FALSE);
	}

	if (argc < 3) {
	(void) fprintf(stderr, gettext("missing pool name\n"));
	usage(B_FALSE);
	}

	if (argc > 4) {
	(void) fprintf(stderr, gettext("too many pool names\n"));
	usage(B_FALSE);
	}

	cb.cb_propname = argv[1];
	cb.cb_type = ZFS_TYPE_POOL;
	cb.cb_vdevs.cb_name_flags \|= VDEV_NAME_TYPE_ID;
	cb.cb_value = strchr(cb.cb_propname, '=');
	if (cb.cb_value == NULL) {
	(void) fprintf(stderr, gettext("missing value in "
	"property=value argument\n"));
	usage(B_FALSE);
	}

	*(cb.cb_value) = '\0';
	cb.cb_value++;
	argc -= 2;
	argv += 2;

	/* argv[0] is pool name */
	if (!is_pool(argv[0])) {
	(void) fprintf(stderr,
	gettext("cannot open '%s': is not a pool\n"), argv[0]);
	return (EINVAL);
	}

	/* argv[1], when supplied, is vdev name */
	if (argc == 2) {

	if (strcmp(argv[1], "root") == 0)
	vdev = strdup("root-0");
	else
	vdev = strdup(argv[1]);

	if (!are_vdevs_in_pool(1, &vdev, argv[0], &cb.cb_vdevs)) {
	(void) fprintf(stderr, gettext(
	"cannot find '%s' in '%s': device not in pool\n"),
	vdev, argv[0]);
	free(vdev);
	return (EINVAL);
	}
	cb.cb_vdevs.cb_names = &vdev;
	cb.cb_vdevs.cb_names_count = 1;
	cb.cb_type = ZFS_TYPE_VDEV;
	}

	error = for_each_pool(1, argv, B_TRUE, NULL, ZFS_TYPE_POOL,
	B_FALSE, set_callback, &cb);

	if (vdev != NULL)
	free(vdev);

	return (error);
	}

	/* Add up the total number of bytes left to initialize/trim across all vdevs */
	static uint64_t
	vdev_activity_remaining(nvlist_t *nv, zpool_wait_activity_t activity)
	{
	uint64_t bytes_remaining;
	nvlist_t **child;
	uint_t c, children;
	vdev_stat_t *vs;

	assert(activity == ZPOOL_WAIT_INITIALIZE \|\|
	activity == ZPOOL_WAIT_TRIM);

	verify(nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
	(uint64_t **)&vs, &c) == 0);

	if (activity == ZPOOL_WAIT_INITIALIZE &&
	vs->vs_initialize_state == VDEV_INITIALIZE_ACTIVE)
	bytes_remaining = vs->vs_initialize_bytes_est -
	vs->vs_initialize_bytes_done;
	else if (activity == ZPOOL_WAIT_TRIM &&
	vs->vs_trim_state == VDEV_TRIM_ACTIVE)
	bytes_remaining = vs->vs_trim_bytes_est -
	vs->vs_trim_bytes_done;
	else
	bytes_remaining = 0;

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	&child, &children) != 0)
	children = 0;

	for (c = 0; c < children; c++)
	bytes_remaining += vdev_activity_remaining(child[c], activity);

	return (bytes_remaining);
	}

	/* Add up the total number of bytes left to rebuild across top-level vdevs */
	static uint64_t
	vdev_activity_top_remaining(nvlist_t *nv)
	{
	uint64_t bytes_remaining = 0;
	nvlist_t **child;
	uint_t children;
	int error;

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	&child, &children) != 0)
	children = 0;

	for (uint_t c = 0; c < children; c++) {
	vdev_rebuild_stat_t *vrs;
	uint_t i;

	error = nvlist_lookup_uint64_array(child[c],
	ZPOOL_CONFIG_REBUILD_STATS, (uint64_t **)&vrs, &i);
	if (error == 0) {
	if (vrs->vrs_state == VDEV_REBUILD_ACTIVE) {
	bytes_remaining += (vrs->vrs_bytes_est -
	vrs->vrs_bytes_rebuilt);
	}
	}
	}

	return (bytes_remaining);
	}

	/* Whether any vdevs are 'spare' or 'replacing' vdevs */
	static boolean_t
	vdev_any_spare_replacing(nvlist_t *nv)
	{
	nvlist_t **child;
	uint_t c, children;
	const char *vdev_type;

	(void) nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &vdev_type);

	if (strcmp(vdev_type, VDEV_TYPE_REPLACING) == 0 \|\|
	strcmp(vdev_type, VDEV_TYPE_SPARE) == 0 \|\|
	strcmp(vdev_type, VDEV_TYPE_DRAID_SPARE) == 0) {
	return (B_TRUE);
	}

	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	&child, &children) != 0)
	children = 0;

	for (c = 0; c < children; c++) {
	if (vdev_any_spare_replacing(child[c]))
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	typedef struct wait_data {
	char *wd_poolname;
	boolean_t wd_scripted;
	boolean_t wd_exact;
	boolean_t wd_headers_once;
	boolean_t wd_should_exit;
	/* Which activities to wait for */
	boolean_t wd_enabled[ZPOOL_WAIT_NUM_ACTIVITIES];
	float wd_interval;
	pthread_cond_t wd_cv;
	pthread_mutex_t wd_mutex;
	} wait_data_t;

	/*
	* Print to stdout a single line, containing one column for each activity that
	* we are waiting for specifying how many bytes of work are left for that
	* activity.
	*/
	static void
	print_wait_status_row(wait_data_t wd, zpool_handle_t zhp, int row)
	{
	nvlist_t config, nvroot;
	uint_t c;
	int i;
	pool_checkpoint_stat_t *pcs = NULL;
	pool_scan_stat_t *pss = NULL;
	pool_removal_stat_t *prs = NULL;
	pool_raidz_expand_stat_t *pres = NULL;
	const char *const headers[] = {"DISCARD", "FREE", "INITIALIZE",
	"REPLACE", "REMOVE", "RESILVER", "SCRUB", "TRIM", "RAIDZ_EXPAND"};
	int col_widths[ZPOOL_WAIT_NUM_ACTIVITIES];

	/* Calculate the width of each column */
	for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++) {
	/*
	* Make sure we have enough space in the col for pretty-printed
	* numbers and for the column header, and then leave a couple
	* spaces between cols for readability.
	*/
	col_widths[i] = MAX(strlen(headers[i]), 6) + 2;
	}

	if (timestamp_fmt != NODATE)
	print_timestamp(timestamp_fmt);

	/* Print header if appropriate */
	int term_height = terminal_height();
	boolean_t reprint_header = (!wd->wd_headers_once && term_height > 0 &&
	row % (term_height-1) == 0);
	if (!wd->wd_scripted && (row == 0 \|\| reprint_header)) {
	for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++) {
	if (wd->wd_enabled[i])
	(void) printf("%*s", col_widths[i], headers[i]);
	}
	(void) fputc('\n', stdout);
	}

	/* Bytes of work remaining in each activity */
	int64_t bytes_rem[ZPOOL_WAIT_NUM_ACTIVITIES] = {0};

	bytes_rem[ZPOOL_WAIT_FREE] =
	zpool_get_prop_int(zhp, ZPOOL_PROP_FREEING, NULL);

	config = zpool_get_config(zhp, NULL);
	nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);

	(void) nvlist_lookup_uint64_array(nvroot,
	ZPOOL_CONFIG_CHECKPOINT_STATS, (uint64_t **)&pcs, &c);
	if (pcs != NULL && pcs->pcs_state == CS_CHECKPOINT_DISCARDING)
	bytes_rem[ZPOOL_WAIT_CKPT_DISCARD] = pcs->pcs_space;

	(void) nvlist_lookup_uint64_array(nvroot,
	ZPOOL_CONFIG_REMOVAL_STATS, (uint64_t **)&prs, &c);
	if (prs != NULL && prs->prs_state == DSS_SCANNING)
	bytes_rem[ZPOOL_WAIT_REMOVE] = prs->prs_to_copy -
	prs->prs_copied;

	(void) nvlist_lookup_uint64_array(nvroot,
	ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&pss, &c);
	if (pss != NULL && pss->pss_state == DSS_SCANNING &&
	pss->pss_pass_scrub_pause == 0) {
	int64_t rem = pss->pss_to_examine - pss->pss_issued;
	if (pss->pss_func == POOL_SCAN_SCRUB)
	bytes_rem[ZPOOL_WAIT_SCRUB] = rem;
	else
	bytes_rem[ZPOOL_WAIT_RESILVER] = rem;
	} else if (check_rebuilding(nvroot, NULL)) {
	bytes_rem[ZPOOL_WAIT_RESILVER] =
	vdev_activity_top_remaining(nvroot);
	}

	(void) nvlist_lookup_uint64_array(nvroot,
	ZPOOL_CONFIG_RAIDZ_EXPAND_STATS, (uint64_t **)&pres, &c);
	if (pres != NULL && pres->pres_state == DSS_SCANNING) {
	int64_t rem = pres->pres_to_reflow - pres->pres_reflowed;
	bytes_rem[ZPOOL_WAIT_RAIDZ_EXPAND] = rem;
	}

	bytes_rem[ZPOOL_WAIT_INITIALIZE] =
	vdev_activity_remaining(nvroot, ZPOOL_WAIT_INITIALIZE);
	bytes_rem[ZPOOL_WAIT_TRIM] =
	vdev_activity_remaining(nvroot, ZPOOL_WAIT_TRIM);

	/*
	* A replace finishes after resilvering finishes, so the amount of work
	* left for a replace is the same as for resilvering.
	*
	* It isn't quite correct to say that if we have any 'spare' or
	* 'replacing' vdevs and a resilver is happening, then a replace is in
	* progress, like we do here. When a hot spare is used, the faulted vdev
	* is not removed after the hot spare is resilvered, so parent 'spare'
	* vdev is not removed either. So we could have a 'spare' vdev, but be
	* resilvering for a different reason. However, we use it as a heuristic
	* because we don't have access to the DTLs, which could tell us whether
	* or not we have really finished resilvering a hot spare.
	*/
	if (vdev_any_spare_replacing(nvroot))
	bytes_rem[ZPOOL_WAIT_REPLACE] = bytes_rem[ZPOOL_WAIT_RESILVER];

	for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++) {
	char buf[64];
	if (!wd->wd_enabled[i])
	continue;

	if (wd->wd_exact) {
	(void) snprintf(buf, sizeof (buf), "%" PRIi64,
	bytes_rem[i]);
	} else {
	zfs_nicenum(bytes_rem[i], buf, sizeof (buf));
	}

	if (wd->wd_scripted)
	(void) printf(i == 0 ? "%s" : "\t%s", buf);
	else
	(void) printf(" %*s", col_widths[i] - 1, buf);
	}
	(void) printf("\n");
	(void) fflush(stdout);
	}

	static void *
	wait_status_thread(void *arg)
	{
	wait_data_t wd = (wait_data_t )arg;
	zpool_handle_t *zhp;

	if ((zhp = zpool_open(g_zfs, wd->wd_poolname)) == NULL)
	return (void *)(1);

	for (int row = 0; ; row++) {
	boolean_t missing;
	struct timespec timeout;
	int ret = 0;
	(void) clock_gettime(CLOCK_REALTIME, &timeout);

	if (zpool_refresh_stats(zhp, &missing) != 0 \|\| missing \|\|
	zpool_props_refresh(zhp) != 0) {
	zpool_close(zhp);
	return (void *)(uintptr_t)(missing ? 0 : 1);
	}

	print_wait_status_row(wd, zhp, row);

	timeout.tv_sec += floor(wd->wd_interval);
	long nanos = timeout.tv_nsec +
	(wd->wd_interval - floor(wd->wd_interval)) * NANOSEC;
	if (nanos >= NANOSEC) {
	timeout.tv_sec++;
	timeout.tv_nsec = nanos - NANOSEC;
	} else {
	timeout.tv_nsec = nanos;
	}
	pthread_mutex_lock(&wd->wd_mutex);
	if (!wd->wd_should_exit)
	ret = pthread_cond_timedwait(&wd->wd_cv, &wd->wd_mutex,
	&timeout);
	pthread_mutex_unlock(&wd->wd_mutex);
	if (ret == 0) {
	break; /* signaled by main thread */
	} else if (ret != ETIMEDOUT) {
	(void) fprintf(stderr, gettext("pthread_cond_timedwait "
	"failed: %s\n"), strerror(ret));
	zpool_close(zhp);
	return (void *)(uintptr_t)(1);
	}
	}

	zpool_close(zhp);
	return (void *)(0);
	}

	int
	zpool_do_wait(int argc, char **argv)
	{
	boolean_t verbose = B_FALSE;
	int c, i;
	unsigned long count;
	pthread_t status_thr;
	int error = 0;
	zpool_handle_t *zhp;

	wait_data_t wd;
	wd.wd_scripted = B_FALSE;
	wd.wd_exact = B_FALSE;
	wd.wd_headers_once = B_FALSE;
	wd.wd_should_exit = B_FALSE;

	pthread_mutex_init(&wd.wd_mutex, NULL);
	pthread_cond_init(&wd.wd_cv, NULL);

	/* By default, wait for all types of activity. */
	for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++)
	wd.wd_enabled[i] = B_TRUE;

	while ((c = getopt(argc, argv, "HpT:t:")) != -1) {
	switch (c) {
	case 'H':
	wd.wd_scripted = B_TRUE;
	break;
	case 'n':
	wd.wd_headers_once = B_TRUE;
	break;
	case 'p':
	wd.wd_exact = B_TRUE;
	break;
	case 'T':
	get_timestamp_arg(*optarg);
	break;
	case 't':
	/* Reset activities array */
	memset(&wd.wd_enabled, 0, sizeof (wd.wd_enabled));

	for (char *tok; (tok = strsep(&optarg, ",")); ) {
	static const char *const col_opts[] = {
	"discard", "free", "initialize", "replace",
	"remove", "resilver", "scrub", "trim",
	"raidz_expand" };

	for (i = 0; i < ARRAY_SIZE(col_opts); ++i)
	if (strcmp(tok, col_opts[i]) == 0) {
	wd.wd_enabled[i] = B_TRUE;
	goto found;
	}

	(void) fprintf(stderr,
	gettext("invalid activity '%s'\n"), tok);
	usage(B_FALSE);
	found:;
	}
	break;
	case '?':
	(void) fprintf(stderr, gettext("invalid option '%c'\n"),
	optopt);
	usage(B_FALSE);
	}
	}

	argc -= optind;
	argv += optind;

	get_interval_count(&argc, argv, &wd.wd_interval, &count);
	if (count != 0) {
	/* This subcmd only accepts an interval, not a count */
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	if (wd.wd_interval != 0)
	verbose = B_TRUE;

	if (argc < 1) {
	(void) fprintf(stderr, gettext("missing 'pool' argument\n"));
	usage(B_FALSE);
	}
	if (argc > 1) {
	(void) fprintf(stderr, gettext("too many arguments\n"));
	usage(B_FALSE);
	}

	wd.wd_poolname = argv[0];

	if ((zhp = zpool_open(g_zfs, wd.wd_poolname)) == NULL)
	return (1);

	if (verbose) {
	/*
	* We use a separate thread for printing status updates because
	* the main thread will call lzc_wait(), which blocks as long
	* as an activity is in progress, which can be a long time.
	*/
	if (pthread_create(&status_thr, NULL, wait_status_thread, &wd)
	!= 0) {
	(void) fprintf(stderr, gettext("failed to create status"
	"thread: %s\n"), strerror(errno));
	zpool_close(zhp);
	return (1);
	}
	}

	/*
	* Loop over all activities that we are supposed to wait for until none
	* of them are in progress. Note that this means we can end up waiting
	* for more activities to complete than just those that were in progress
	* when we began waiting; if an activity we are interested in begins
	* while we are waiting for another activity, we will wait for both to
	* complete before exiting.
	*/
	for (;;) {
	boolean_t missing = B_FALSE;
	boolean_t any_waited = B_FALSE;

	for (i = 0; i < ZPOOL_WAIT_NUM_ACTIVITIES; i++) {
	boolean_t waited;

	if (!wd.wd_enabled[i])
	continue;

	error = zpool_wait_status(zhp, i, &missing, &waited);
	if (error != 0 \|\| missing)
	break;

	any_waited = (any_waited \|\| waited);
	}

	if (error != 0 \|\| missing \|\| !any_waited)
	break;
	}

	zpool_close(zhp);

	if (verbose) {
	uintptr_t status;
	pthread_mutex_lock(&wd.wd_mutex);
	wd.wd_should_exit = B_TRUE;
	pthread_cond_signal(&wd.wd_cv);
	pthread_mutex_unlock(&wd.wd_mutex);
	(void) pthread_join(status_thr, (void *)&status);
	if (status != 0)
	error = status;
	}

	pthread_mutex_destroy(&wd.wd_mutex);
	pthread_cond_destroy(&wd.wd_cv);
	return (error);
	}

	static int
	find_command_idx(const char command, int idx)
	{
	for (int i = 0; i < NCOMMAND; ++i) {
	if (command_table[i].name == NULL)
	continue;

	if (strcmp(command, command_table[i].name) == 0) {
	*idx = i;
	return (0);
	}
	}
	return (1);
	}

	/*
	* Display version message
	*/
	static int
	zpool_do_version(int argc, char **argv)
	{
	(void) argc, (void) argv;
	return (zfs_version_print() != 0);
	}

	/* Display documentation */
	static int
	zpool_do_help(int argc, char **argv)
	{
	char page[MAXNAMELEN];
	if (argc < 3 \|\| strcmp(argv[2], "zpool") == 0)
	strcpy(page, "zpool");
	else if (strcmp(argv[2], "concepts") == 0 \|\|
	strcmp(argv[2], "props") == 0)
	snprintf(page, sizeof (page), "zpool%s", argv[2]);
	else
	snprintf(page, sizeof (page), "zpool-%s", argv[2]);

	execlp("man", "man", page, NULL);

	fprintf(stderr, "couldn't run man program: %s", strerror(errno));
	return (-1);
	}

	/*
	* Do zpool_load_compat() and print error message on failure
	*/
	static zpool_compat_status_t
	zpool_do_load_compat(const char compat, boolean_t list)
	{
	char report[1024];

	zpool_compat_status_t ret;

	ret = zpool_load_compat(compat, list, report, 1024);
	switch (ret) {

	case ZPOOL_COMPATIBILITY_OK:
	break;

	case ZPOOL_COMPATIBILITY_NOFILES:
	case ZPOOL_COMPATIBILITY_BADFILE:
	case ZPOOL_COMPATIBILITY_BADTOKEN:
	(void) fprintf(stderr, "Error: %s\n", report);
	break;

	case ZPOOL_COMPATIBILITY_WARNTOKEN:
	(void) fprintf(stderr, "Warning: %s\n", report);
	ret = ZPOOL_COMPATIBILITY_OK;
	break;
	}
	return (ret);
	}

	int
	main(int argc, char **argv)
	{
	int ret = 0;
	int i = 0;
	char *cmdname;
	char **newargv;

	(void) setlocale(LC_ALL, "");
	(void) setlocale(LC_NUMERIC, "C");
	(void) textdomain(TEXT_DOMAIN);
	srand(time(NULL));

	opterr = 0;

	/*
	* Make sure the user has specified some command.
	*/
	if (argc < 2) {
	(void) fprintf(stderr, gettext("missing command\n"));
	usage(B_FALSE);
	}

	cmdname = argv[1];

	/*
	* Special case '-?'
	*/
	if ((strcmp(cmdname, "-?") == 0) \|\| strcmp(cmdname, "--help") == 0)
	usage(B_TRUE);

	/*
	* Special case '-V\|--version'
	*/
	if ((strcmp(cmdname, "-V") == 0) \|\| (strcmp(cmdname, "--version") == 0))
	return (zpool_do_version(argc, argv));

	/*
	* Special case 'help'
	*/
	if (strcmp(cmdname, "help") == 0)
	return (zpool_do_help(argc, argv));

	if ((g_zfs = libzfs_init()) == NULL) {
	(void) fprintf(stderr, "%s\n", libzfs_error_init(errno));
	return (1);
	}

	libzfs_print_on_error(g_zfs, B_TRUE);

	zfs_save_arguments(argc, argv, history_str, sizeof (history_str));

	/*
	* Many commands modify input strings for string parsing reasons.
	* We create a copy to protect the original argv.
	*/
	newargv = safe_malloc((argc + 1) * sizeof (newargv[0]));
	for (i = 0; i < argc; i++)
	newargv[i] = strdup(argv[i]);
	newargv[argc] = NULL;

	/*
	* Run the appropriate command.
	*/
	if (find_command_idx(cmdname, &i) == 0) {
	current_command = &command_table[i];
	ret = command_table[i].func(argc - 1, newargv + 1);
	} else if (strchr(cmdname, '=')) {
	verify(find_command_idx("set", &i) == 0);
	current_command = &command_table[i];
	ret = command_table[i].func(argc, newargv);
	} else if (strcmp(cmdname, "freeze") == 0 && argc == 3) {
	/*
	* 'freeze' is a vile debugging abomination, so we treat
	* it as such.
	*/
	zfs_cmd_t zc = {"\0"};

	(void) strlcpy(zc.zc_name, argv[2], sizeof (zc.zc_name));
	ret = zfs_ioctl(g_zfs, ZFS_IOC_POOL_FREEZE, &zc);
	if (ret != 0) {
	(void) fprintf(stderr,
	gettext("failed to freeze pool: %d\n"), errno);
	ret = 1;
	}

	log_history = 0;
	} else {
	(void) fprintf(stderr, gettext("unrecognized "
	"command '%s'\n"), cmdname);
	usage(B_FALSE);
	ret = 1;
	}

	for (i = 0; i < argc; i++)
	free(newargv[i]);
	free(newargv);

	if (ret == 0 && log_history)
	(void) zpool_log_history(g_zfs, history_str);

	libzfs_fini(g_zfs);

	/*
	* The 'ZFS_ABORT' environment variable causes us to dump core on exit
	* for the purposes of running ::findleaks.
	*/
	if (getenv("ZFS_ABORT") != NULL) {
	(void) printf("dumping core by request\n");
	abort();
	}

	return (ret);
	}
	diff --git a/sys/contrib/openzfs/cmd/zstream/zstream_recompress.c b/sys/contrib/openzfs/cmd/zstream/zstream_recompress.c
	index 8392ef3de72f..f9e01d1aa4c4 100644
	--- a/sys/contrib/openzfs/cmd/zstream/zstream_recompress.c
	+++ b/sys/contrib/openzfs/cmd/zstream/zstream_recompress.c
	@@ -1,376 +1,376 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright 2022 Axcient. All rights reserved.
	* Use is subject to license terms.
	*/

	/*
	* Copyright (c) 2022 by Delphix. All rights reserved.
	*/

	#include <err.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <unistd.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/zio_checksum.h>
	#include <sys/zstd/zstd.h>
	#include "zfs_fletcher.h"
	#include "zstream.h"

	static int
	dump_record(dmu_replay_record_t drr, void payload, int payload_len,
	zio_cksum_t *zc, int outfd)
	{
	assert(offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum)
	== sizeof (dmu_replay_record_t) - sizeof (zio_cksum_t));
	fletcher_4_incremental_native(drr,
	offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum), zc);
	if (drr->drr_type != DRR_BEGIN) {
	assert(ZIO_CHECKSUM_IS_ZERO(&drr->drr_u.
	drr_checksum.drr_checksum));
	drr->drr_u.drr_checksum.drr_checksum = *zc;
	}
	fletcher_4_incremental_native(&drr->drr_u.drr_checksum.drr_checksum,
	sizeof (zio_cksum_t), zc);
	if (write(outfd, drr, sizeof (*drr)) == -1)
	return (errno);
	if (payload_len != 0) {
	fletcher_4_incremental_native(payload, payload_len, zc);
	if (write(outfd, payload, payload_len) == -1)
	return (errno);
	}
	return (0);
	}

	int
	zstream_do_recompress(int argc, char *argv[])
	{
	int bufsz = SPA_MAXBLOCKSIZE;
	char *buf = safe_malloc(bufsz);
	dmu_replay_record_t thedrr;
	dmu_replay_record_t *drr = &thedrr;
	zio_cksum_t stream_cksum;
	int c;
	int level = -1;

	while ((c = getopt(argc, argv, "l:")) != -1) {
	switch (c) {
	case 'l':
	- if (sscanf(optarg, "%d", &level) != 0) {
	+ if (sscanf(optarg, "%d", &level) != 1) {
	fprintf(stderr,
	"failed to parse level '%s'\n",
	optarg);
	zstream_usage();
	}
	break;
	case '?':
	(void) fprintf(stderr, "invalid option '%c'\n",
	optopt);
	zstream_usage();
	break;
	}
	}

	argc -= optind;
	argv += optind;

	if (argc != 1)
	zstream_usage();
	int type = 0;
	zio_compress_info_t *cinfo = NULL;
	if (0 == strcmp(argv[0], "off")) {
	type = ZIO_COMPRESS_OFF;
	cinfo = &zio_compress_table[type];
	} else if (0 == strcmp(argv[0], "inherit") \|\|
	0 == strcmp(argv[0], "empty") \|\|
	0 == strcmp(argv[0], "on")) {
	// Fall through to invalid compression type case
	} else {
	for (int i = 0; i < ZIO_COMPRESS_FUNCTIONS; i++) {
	if (0 == strcmp(zio_compress_table[i].ci_name,
	argv[0])) {
	cinfo = &zio_compress_table[i];
	type = i;
	break;
	}
	}
	}
	if (cinfo == NULL) {
	fprintf(stderr, "Invalid compression type %s.\n",
	argv[0]);
	exit(2);
	}

	if (cinfo->ci_compress == NULL) {
	type = 0;
	cinfo = &zio_compress_table[0];
	}

	if (isatty(STDIN_FILENO)) {
	(void) fprintf(stderr,
	"Error: The send stream is a binary format "
	"and can not be read from a\n"
	"terminal. Standard input must be redirected.\n");
	exit(1);
	}

	fletcher_4_init();
	zio_init();
	zstd_init();
	int begin = 0;
	boolean_t seen = B_FALSE;
	while (sfread(drr, sizeof (*drr), stdin) != 0) {
	struct drr_write *drrw;
	uint64_t payload_size = 0;

	/*
	* We need to regenerate the checksum.
	*/
	if (drr->drr_type != DRR_BEGIN) {
	memset(&drr->drr_u.drr_checksum.drr_checksum, 0,
	sizeof (drr->drr_u.drr_checksum.drr_checksum));
	}


	switch (drr->drr_type) {
	case DRR_BEGIN:
	{
	ZIO_SET_CHECKSUM(&stream_cksum, 0, 0, 0, 0);
	VERIFY0(begin++);
	seen = B_TRUE;

	uint32_t sz = drr->drr_payloadlen;

	VERIFY3U(sz, <=, 1U << 28);

	if (sz != 0) {
	if (sz > bufsz) {
	buf = realloc(buf, sz);
	if (buf == NULL)
	err(1, "realloc");
	bufsz = sz;
	}
	(void) sfread(buf, sz, stdin);
	}
	payload_size = sz;
	break;
	}
	case DRR_END:
	{
	struct drr_end *drre = &drr->drr_u.drr_end;
	/*
	* We would prefer to just check --begin == 0, but
	* replication streams have an end of stream END
	* record, so we must avoid tripping it.
	*/
	VERIFY3B(seen, ==, B_TRUE);
	begin--;
	/*
	* Use the recalculated checksum, unless this is
	* the END record of a stream package, which has
	* no checksum.
	*/
	if (!ZIO_CHECKSUM_IS_ZERO(&drre->drr_checksum))
	drre->drr_checksum = stream_cksum;
	break;
	}

	case DRR_OBJECT:
	{
	struct drr_object *drro = &drr->drr_u.drr_object;
	VERIFY3S(begin, ==, 1);

	if (drro->drr_bonuslen > 0) {
	payload_size = DRR_OBJECT_PAYLOAD_SIZE(drro);
	(void) sfread(buf, payload_size, stdin);
	}
	break;
	}

	case DRR_SPILL:
	{
	struct drr_spill *drrs = &drr->drr_u.drr_spill;
	VERIFY3S(begin, ==, 1);
	payload_size = DRR_SPILL_PAYLOAD_SIZE(drrs);
	(void) sfread(buf, payload_size, stdin);
	break;
	}

	case DRR_WRITE_BYREF:
	VERIFY3S(begin, ==, 1);
	fprintf(stderr,
	"Deduplicated streams are not supported\n");
	exit(1);
	break;

	case DRR_WRITE:
	{
	VERIFY3S(begin, ==, 1);
	drrw = &thedrr.drr_u.drr_write;
	payload_size = DRR_WRITE_PAYLOAD_SIZE(drrw);
	/*
	* In order to recompress an encrypted block, you have
	* to decrypt, decompress, recompress, and
	* re-encrypt. That can be a future enhancement (along
	* with decryption or re-encryption), but for now we
	* skip encrypted blocks.
	*/
	boolean_t encrypted = B_FALSE;
	for (int i = 0; i < ZIO_DATA_SALT_LEN; i++) {
	if (drrw->drr_salt[i] != 0) {
	encrypted = B_TRUE;
	break;
	}
	}
	if (encrypted) {
	(void) sfread(buf, payload_size, stdin);
	break;
	}
	if (drrw->drr_compressiontype >=
	ZIO_COMPRESS_FUNCTIONS) {
	fprintf(stderr, "Invalid compression type in "
	"stream: %d\n", drrw->drr_compressiontype);
	exit(3);
	}
	zio_compress_info_t *dinfo =
	&zio_compress_table[drrw->drr_compressiontype];

	/* Set up buffers to minimize memcpys */
	char cbuf, dbuf;
	if (cinfo->ci_compress == NULL)
	dbuf = buf;
	else
	dbuf = safe_calloc(bufsz);

	if (dinfo->ci_decompress == NULL)
	cbuf = dbuf;
	else
	cbuf = safe_calloc(payload_size);

	/* Read and decompress the payload */
	(void) sfread(cbuf, payload_size, stdin);
	if (dinfo->ci_decompress != NULL) {
	if (0 != dinfo->ci_decompress(cbuf, dbuf,
	payload_size, MIN(bufsz,
	drrw->drr_logical_size), dinfo->ci_level)) {
	warnx("decompression type %d failed "
	"for ino %llu offset %llu",
	type,
	(u_longlong_t)drrw->drr_object,
	(u_longlong_t)drrw->drr_offset);
	exit(4);
	}
	payload_size = drrw->drr_logical_size;
	free(cbuf);
	}

	/* Recompress the payload */
	if (cinfo->ci_compress != NULL) {
	payload_size = P2ROUNDUP(cinfo->ci_compress(
	dbuf, buf, drrw->drr_logical_size,
	MIN(payload_size, bufsz), (level == -1 ?
	cinfo->ci_level : level)),
	SPA_MINBLOCKSIZE);
	if (payload_size != drrw->drr_logical_size) {
	drrw->drr_compressiontype = type;
	drrw->drr_compressed_size =
	payload_size;
	} else {
	memcpy(buf, dbuf, payload_size);
	drrw->drr_compressiontype = 0;
	drrw->drr_compressed_size = 0;
	}
	free(dbuf);
	} else {
	drrw->drr_compressiontype = type;
	drrw->drr_compressed_size = 0;
	}
	break;
	}

	case DRR_WRITE_EMBEDDED:
	{
	struct drr_write_embedded *drrwe =
	&drr->drr_u.drr_write_embedded;
	VERIFY3S(begin, ==, 1);
	payload_size =
	P2ROUNDUP((uint64_t)drrwe->drr_psize, 8);
	(void) sfread(buf, payload_size, stdin);
	break;
	}

	case DRR_FREEOBJECTS:
	case DRR_FREE:
	case DRR_OBJECT_RANGE:
	VERIFY3S(begin, ==, 1);
	break;

	default:
	(void) fprintf(stderr, "INVALID record type 0x%x\n",
	drr->drr_type);
	/* should never happen, so assert */
	assert(B_FALSE);
	}

	if (feof(stdout)) {
	fprintf(stderr, "Error: unexpected end-of-file\n");
	exit(1);
	}
	if (ferror(stdout)) {
	fprintf(stderr, "Error while reading file: %s\n",
	strerror(errno));
	exit(1);
	}

	/*
	* We need to recalculate the checksum, and it needs to be
	* initially zero to do that. BEGIN records don't have
	* a checksum.
	*/
	if (drr->drr_type != DRR_BEGIN) {
	memset(&drr->drr_u.drr_checksum.drr_checksum, 0,
	sizeof (drr->drr_u.drr_checksum.drr_checksum));
	}
	if (dump_record(drr, buf, payload_size,
	&stream_cksum, STDOUT_FILENO) != 0)
	break;
	if (drr->drr_type == DRR_END) {
	/*
	* Typically the END record is either the last
	* thing in the stream, or it is followed
	* by a BEGIN record (which also zeros the checksum).
	* However, a stream package ends with two END
	* records. The last END record's checksum starts
	* from zero.
	*/
	ZIO_SET_CHECKSUM(&stream_cksum, 0, 0, 0, 0);
	}
	}
	free(buf);
	fletcher_4_fini();
	zio_fini();
	zstd_fini();

	return (0);
	}
	diff --git a/sys/contrib/openzfs/cmd/zstream/zstream_redup.c b/sys/contrib/openzfs/cmd/zstream/zstream_redup.c
	index c56a09cee75d..6866639fe465 100644
	--- a/sys/contrib/openzfs/cmd/zstream/zstream_redup.c
	+++ b/sys/contrib/openzfs/cmd/zstream/zstream_redup.c
	@@ -1,489 +1,491 @@
	/*
	* CDDL HEADER START
	*
	* This file and its contents are supplied under the terms of the
	* Common Development and Distribution License ("CDDL"), version 1.0.
	* You may only use this file in accordance with the terms of version
	* 1.0 of the CDDL.
	*
	* A full copy of the text of the CDDL should have accompanied this
	* source. A copy of the CDDL is also available via the Internet at
	* http://www.illumos.org/license/CDDL.
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2020 by Delphix. All rights reserved.
	*/

	#include <assert.h>
	#include <cityhash.h>
	#include <ctype.h>
	#include <errno.h>
	#include <fcntl.h>
	#include <libzfs.h>
	#include <libzutil.h>
	#include <stddef.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <string.h>
	#include <umem.h>
	#include <unistd.h>
	#include <sys/debug.h>
	#include <sys/stat.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/zio_checksum.h>
	#include "zfs_fletcher.h"
	#include "zstream.h"


	#define MAX_RDT_PHYSMEM_PERCENT 20
	#define SMALLEST_POSSIBLE_MAX_RDT_MB 128

	typedef struct redup_entry {
	struct redup_entry *rde_next;
	uint64_t rde_guid;
	uint64_t rde_object;
	uint64_t rde_offset;
	uint64_t rde_stream_offset;
	} redup_entry_t;

	typedef struct redup_table {
	redup_entry_t **redup_hash_array;
	umem_cache_t *ddecache;
	uint64_t ddt_count;
	int numhashbits;
	} redup_table_t;

	int
	highbit64(uint64_t i)
	{
	if (i == 0)
	return (0);

	return (NBBY * sizeof (uint64_t) - __builtin_clzll(i));
	}

	void *
	safe_calloc(size_t n)
	{
	void *rv = calloc(1, n);
	if (rv == NULL) {
	fprintf(stderr,
	"Error: could not allocate %u bytes of memory\n",
	(int)n);
	exit(1);
	}
	return (rv);
	}

	/*
	* Safe version of fread(), exits on error.
	*/
	int
	sfread(void buf, size_t size, FILE fp)
	{
	int rv = fread(buf, size, 1, fp);
	if (rv == 0 && ferror(fp)) {
	(void) fprintf(stderr, "Error while reading file: %s\n",
	strerror(errno));
	exit(1);
	}
	return (rv);
	}

	/*
	* Safe version of pread(), exits on error.
	*/
	static void
	spread(int fd, void *buf, size_t count, off_t offset)
	{
	ssize_t err = pread(fd, buf, count, offset);
	if (err == -1) {
	(void) fprintf(stderr,
	"Error while reading file: %s\n",
	strerror(errno));
	exit(1);
	} else if (err != count) {
	(void) fprintf(stderr,
	"Error while reading file: short read\n");
	exit(1);
	}
	}

	static int
	dump_record(dmu_replay_record_t drr, void payload, int payload_len,
	zio_cksum_t *zc, int outfd)
	{
	assert(offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum)
	== sizeof (dmu_replay_record_t) - sizeof (zio_cksum_t));
	fletcher_4_incremental_native(drr,
	offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum), zc);
	if (drr->drr_type != DRR_BEGIN) {
	assert(ZIO_CHECKSUM_IS_ZERO(&drr->drr_u.
	drr_checksum.drr_checksum));
	drr->drr_u.drr_checksum.drr_checksum = *zc;
	}
	fletcher_4_incremental_native(&drr->drr_u.drr_checksum.drr_checksum,
	sizeof (zio_cksum_t), zc);
	if (write(outfd, drr, sizeof (*drr)) == -1)
	return (errno);
	if (payload_len != 0) {
	fletcher_4_incremental_native(payload, payload_len, zc);
	if (write(outfd, payload, payload_len) == -1)
	return (errno);
	}
	return (0);
	}

	static void
	rdt_insert(redup_table_t *rdt,
	uint64_t guid, uint64_t object, uint64_t offset, uint64_t stream_offset)
	{
	uint64_t ch = cityhash4(guid, object, offset, 0);
	uint64_t hashcode = BF64_GET(ch, 0, rdt->numhashbits);
	redup_entry_t **rdepp;

	rdepp = &(rdt->redup_hash_array[hashcode]);
	redup_entry_t *rde = umem_cache_alloc(rdt->ddecache, UMEM_NOFAIL);
	rde->rde_next = *rdepp;
	rde->rde_guid = guid;
	rde->rde_object = object;
	rde->rde_offset = offset;
	rde->rde_stream_offset = stream_offset;
	*rdepp = rde;
	rdt->ddt_count++;
	}

	static void
	rdt_lookup(redup_table_t *rdt,
	uint64_t guid, uint64_t object, uint64_t offset,
	uint64_t *stream_offsetp)
	{
	uint64_t ch = cityhash4(guid, object, offset, 0);
	uint64_t hashcode = BF64_GET(ch, 0, rdt->numhashbits);

	for (redup_entry_t *rde = rdt->redup_hash_array[hashcode];
	rde != NULL; rde = rde->rde_next) {
	if (rde->rde_guid == guid &&
	rde->rde_object == object &&
	rde->rde_offset == offset) {
	*stream_offsetp = rde->rde_stream_offset;
	return;
	}
	}
	assert(!"could not find expected redup table entry");
	}

	/*
	* Convert a dedup stream (generated by "zfs send -D") to a
	* non-deduplicated stream. The entire infd will be converted, including
	* any substreams in a stream package (generated by "zfs send -RD"). The
	* infd must be seekable.
	*/
	static void
	zfs_redup_stream(int infd, int outfd, boolean_t verbose)
	{
	int bufsz = SPA_MAXBLOCKSIZE;
	- dmu_replay_record_t thedrr = { 0 };
	+ dmu_replay_record_t thedrr;
	dmu_replay_record_t *drr = &thedrr;
	redup_table_t rdt;
	zio_cksum_t stream_cksum;
	uint64_t numbuckets;
	uint64_t num_records = 0;
	uint64_t num_write_byref_records = 0;

	+ memset(&thedrr, 0, sizeof (dmu_replay_record_t));
	+
	#ifdef _ILP32
	uint64_t max_rde_size = SMALLEST_POSSIBLE_MAX_RDT_MB << 20;
	#else
	uint64_t physmem = sysconf(_SC_PHYS_PAGES) * sysconf(_SC_PAGESIZE);
	uint64_t max_rde_size =
	MAX((physmem * MAX_RDT_PHYSMEM_PERCENT) / 100,
	SMALLEST_POSSIBLE_MAX_RDT_MB << 20);
	#endif

	numbuckets = max_rde_size / (sizeof (redup_entry_t));

	/*
	* numbuckets must be a power of 2. Increase number to
	* a power of 2 if necessary.
	*/
	if (!ISP2(numbuckets))
	numbuckets = 1ULL << highbit64(numbuckets);

	rdt.redup_hash_array =
	safe_calloc(numbuckets * sizeof (redup_entry_t *));
	rdt.ddecache = umem_cache_create("rde", sizeof (redup_entry_t), 0,
	NULL, NULL, NULL, NULL, NULL, 0);
	rdt.numhashbits = highbit64(numbuckets) - 1;
	rdt.ddt_count = 0;

	char *buf = safe_calloc(bufsz);
	FILE *ofp = fdopen(infd, "r");
	long offset = ftell(ofp);
	int begin = 0;
	boolean_t seen = B_FALSE;
	while (sfread(drr, sizeof (*drr), ofp) != 0) {
	num_records++;

	/*
	* We need to regenerate the checksum.
	*/
	if (drr->drr_type != DRR_BEGIN) {
	memset(&drr->drr_u.drr_checksum.drr_checksum, 0,
	sizeof (drr->drr_u.drr_checksum.drr_checksum));
	}

	uint64_t payload_size = 0;
	switch (drr->drr_type) {
	case DRR_BEGIN:
	{
	struct drr_begin *drrb = &drr->drr_u.drr_begin;
	int fflags;
	ZIO_SET_CHECKSUM(&stream_cksum, 0, 0, 0, 0);
	VERIFY0(begin++);
	seen = B_TRUE;

	assert(drrb->drr_magic == DMU_BACKUP_MAGIC);

	/* clear the DEDUP feature flag for this stream */
	fflags = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo);
	fflags &= ~(DMU_BACKUP_FEATURE_DEDUP \|
	DMU_BACKUP_FEATURE_DEDUPPROPS);
	/* cppcheck-suppress syntaxError */
	DMU_SET_FEATUREFLAGS(drrb->drr_versioninfo, fflags);

	uint32_t sz = drr->drr_payloadlen;

	VERIFY3U(sz, <=, 1U << 28);

	if (sz != 0) {
	if (sz > bufsz) {
	free(buf);
	buf = safe_calloc(sz);
	bufsz = sz;
	}
	(void) sfread(buf, sz, ofp);
	}
	payload_size = sz;
	break;
	}

	case DRR_END:
	{
	struct drr_end *drre = &drr->drr_u.drr_end;
	/*
	* We would prefer to just check --begin == 0, but
	* replication streams have an end of stream END
	* record, so we must avoid tripping it.
	*/
	VERIFY3B(seen, ==, B_TRUE);
	begin--;
	/*
	* Use the recalculated checksum, unless this is
	* the END record of a stream package, which has
	* no checksum.
	*/
	if (!ZIO_CHECKSUM_IS_ZERO(&drre->drr_checksum))
	drre->drr_checksum = stream_cksum;
	break;
	}

	case DRR_OBJECT:
	{
	struct drr_object *drro = &drr->drr_u.drr_object;
	VERIFY3S(begin, ==, 1);

	if (drro->drr_bonuslen > 0) {
	payload_size = DRR_OBJECT_PAYLOAD_SIZE(drro);
	(void) sfread(buf, payload_size, ofp);
	}
	break;
	}

	case DRR_SPILL:
	{
	struct drr_spill *drrs = &drr->drr_u.drr_spill;
	VERIFY3S(begin, ==, 1);
	payload_size = DRR_SPILL_PAYLOAD_SIZE(drrs);
	(void) sfread(buf, payload_size, ofp);
	break;
	}

	case DRR_WRITE_BYREF:
	{
	struct drr_write_byref drrwb =
	drr->drr_u.drr_write_byref;
	VERIFY3S(begin, ==, 1);

	num_write_byref_records++;

	/*
	* Look up in hash table by drrwb->drr_refguid,
	* drr_refobject, drr_refoffset. Replace this
	* record with the found WRITE record, but with
	* drr_object,drr_offset,drr_toguid replaced with ours.
	*/
	uint64_t stream_offset = 0;
	rdt_lookup(&rdt, drrwb.drr_refguid,
	drrwb.drr_refobject, drrwb.drr_refoffset,
	&stream_offset);

	spread(infd, drr, sizeof (*drr), stream_offset);

	assert(drr->drr_type == DRR_WRITE);
	struct drr_write *drrw = &drr->drr_u.drr_write;
	assert(drrw->drr_toguid == drrwb.drr_refguid);
	assert(drrw->drr_object == drrwb.drr_refobject);
	assert(drrw->drr_offset == drrwb.drr_refoffset);

	payload_size = DRR_WRITE_PAYLOAD_SIZE(drrw);
	spread(infd, buf, payload_size,
	stream_offset + sizeof (*drr));

	drrw->drr_toguid = drrwb.drr_toguid;
	drrw->drr_object = drrwb.drr_object;
	drrw->drr_offset = drrwb.drr_offset;
	break;
	}

	case DRR_WRITE:
	{
	struct drr_write *drrw = &drr->drr_u.drr_write;
	VERIFY3S(begin, ==, 1);
	payload_size = DRR_WRITE_PAYLOAD_SIZE(drrw);
	(void) sfread(buf, payload_size, ofp);

	rdt_insert(&rdt, drrw->drr_toguid,
	drrw->drr_object, drrw->drr_offset, offset);
	break;
	}

	case DRR_WRITE_EMBEDDED:
	{
	struct drr_write_embedded *drrwe =
	&drr->drr_u.drr_write_embedded;
	VERIFY3S(begin, ==, 1);
	payload_size =
	P2ROUNDUP((uint64_t)drrwe->drr_psize, 8);
	(void) sfread(buf, payload_size, ofp);
	break;
	}

	case DRR_FREEOBJECTS:
	case DRR_FREE:
	case DRR_OBJECT_RANGE:
	VERIFY3S(begin, ==, 1);
	break;

	default:
	(void) fprintf(stderr, "INVALID record type 0x%x\n",
	drr->drr_type);
	/* should never happen, so assert */
	assert(B_FALSE);
	}

	if (feof(ofp)) {
	fprintf(stderr, "Error: unexpected end-of-file\n");
	exit(1);
	}
	if (ferror(ofp)) {
	fprintf(stderr, "Error while reading file: %s\n",
	strerror(errno));
	exit(1);
	}

	/*
	* We need to recalculate the checksum, and it needs to be
	* initially zero to do that. BEGIN records don't have
	* a checksum.
	*/
	if (drr->drr_type != DRR_BEGIN) {
	memset(&drr->drr_u.drr_checksum.drr_checksum, 0,
	sizeof (drr->drr_u.drr_checksum.drr_checksum));
	}
	if (dump_record(drr, buf, payload_size,
	&stream_cksum, outfd) != 0)
	break;
	if (drr->drr_type == DRR_END) {
	/*
	* Typically the END record is either the last
	* thing in the stream, or it is followed
	* by a BEGIN record (which also zeros the checksum).
	* However, a stream package ends with two END
	* records. The last END record's checksum starts
	* from zero.
	*/
	ZIO_SET_CHECKSUM(&stream_cksum, 0, 0, 0, 0);
	}
	offset = ftell(ofp);
	}

	if (verbose) {
	char mem_str[16];
	zfs_nicenum(rdt.ddt_count * sizeof (redup_entry_t),
	mem_str, sizeof (mem_str));
	fprintf(stderr, "converted stream with %llu total records, "
	"including %llu dedup records, using %sB memory.\n",
	(long long)num_records,
	(long long)num_write_byref_records,
	mem_str);
	}

	umem_cache_destroy(rdt.ddecache);
	free(rdt.redup_hash_array);
	free(buf);
	(void) fclose(ofp);
	}

	int
	zstream_do_redup(int argc, char *argv[])
	{
	boolean_t verbose = B_FALSE;
	int c;

	while ((c = getopt(argc, argv, "v")) != -1) {
	switch (c) {
	case 'v':
	verbose = B_TRUE;
	break;
	case '?':
	(void) fprintf(stderr, "invalid option '%c'\n",
	optopt);
	zstream_usage();
	break;
	}
	}

	argc -= optind;
	argv += optind;

	if (argc != 1)
	zstream_usage();

	const char *filename = argv[0];

	if (isatty(STDOUT_FILENO)) {
	(void) fprintf(stderr,
	"Error: Stream can not be written to a terminal.\n"
	"You must redirect standard output.\n");
	return (1);
	}

	int fd = open(filename, O_RDONLY);
	if (fd == -1) {
	(void) fprintf(stderr,
	"Error while opening file '%s': %s\n",
	filename, strerror(errno));
	exit(1);
	}

	fletcher_4_init();
	zfs_redup_stream(fd, STDOUT_FILENO, verbose);
	fletcher_4_fini();

	close(fd);

	return (0);
	}
	diff --git a/sys/contrib/openzfs/cmd/ztest.c b/sys/contrib/openzfs/cmd/ztest.c
	index b0fea8b3cfb4..f77a37c21545 100644
	--- a/sys/contrib/openzfs/cmd/ztest.c
	+++ b/sys/contrib/openzfs/cmd/ztest.c
	@@ -1,9040 +1,9034 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2024 by Delphix. All rights reserved.
	* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2013 Steven Hartland. All rights reserved.
	* Copyright (c) 2014 Integros [integros.com]
	* Copyright 2017 Joyent, Inc.
	* Copyright (c) 2017, Intel Corporation.
	*/

	/*
	* The objective of this program is to provide a DMU/ZAP/SPA stress test
	* that runs entirely in userland, is easy to use, and easy to extend.
	*
	* The overall design of the ztest program is as follows:
	*
	* (1) For each major functional area (e.g. adding vdevs to a pool,
	* creating and destroying datasets, reading and writing objects, etc)
	* we have a simple routine to test that functionality. These
	* individual routines do not have to do anything "stressful".
	*
	* (2) We turn these simple functionality tests into a stress test by
	* running them all in parallel, with as many threads as desired,
	* and spread across as many datasets, objects, and vdevs as desired.
	*
	* (3) While all this is happening, we inject faults into the pool to
	* verify that self-healing data really works.
	*
	* (4) Every time we open a dataset, we change its checksum and compression
	* functions. Thus even individual objects vary from block to block
	* in which checksum they use and whether they're compressed.
	*
	* (5) To verify that we never lose on-disk consistency after a crash,
	* we run the entire test in a child of the main process.
	* At random times, the child self-immolates with a SIGKILL.
	* This is the software equivalent of pulling the power cord.
	* The parent then runs the test again, using the existing
	* storage pool, as many times as desired. If backwards compatibility
	* testing is enabled ztest will sometimes run the "older" version
	* of ztest after a SIGKILL.
	*
	* (6) To verify that we don't have future leaks or temporal incursions,
	* many of the functional tests record the transaction group number
	* as part of their data. When reading old data, they verify that
	* the transaction group number is less than the current, open txg.
	* If you add a new test, please do this if applicable.
	*
	* (7) Threads are created with a reduced stack size, for sanity checking.
	* Therefore, it's important not to allocate huge buffers on the stack.
	*
	* When run with no arguments, ztest runs for about five minutes and
	* produces no output if successful. To get a little bit of information,
	* specify -V. To get more information, specify -VV, and so on.
	*
	* To turn this into an overnight stress test, use -T to specify run time.
	*
	* You can ask more vdevs [-v], datasets [-d], or threads [-t]
	* to increase the pool capacity, fanout, and overall stress level.
	*
	* Use the -k option to set the desired frequency of kills.
	*
	* When ztest invokes itself it passes all relevant information through a
	* temporary file which is mmap-ed in the child process. This allows shared
	* memory to survive the exec syscall. The ztest_shared_hdr_t struct is always
	* stored at offset 0 of this file and contains information on the size and
	* number of shared structures in the file. The information stored in this file
	* must remain backwards compatible with older versions of ztest so that
	* ztest can invoke them during backwards compatibility testing (-B).
	*/

	#include <sys/zfs_context.h>
	#include <sys/spa.h>
	#include <sys/dmu.h>
	#include <sys/txg.h>
	#include <sys/dbuf.h>
	#include <sys/zap.h>
	#include <sys/dmu_objset.h>
	#include <sys/poll.h>
	#include <sys/stat.h>
	#include <sys/time.h>
	#include <sys/wait.h>
	#include <sys/mman.h>
	#include <sys/resource.h>
	#include <sys/zio.h>
	#include <sys/zil.h>
	#include <sys/zil_impl.h>
	#include <sys/vdev_draid.h>
	#include <sys/vdev_impl.h>
	#include <sys/vdev_file.h>
	#include <sys/vdev_initialize.h>
	#include <sys/vdev_raidz.h>
	#include <sys/vdev_trim.h>
	#include <sys/spa_impl.h>
	#include <sys/metaslab_impl.h>
	#include <sys/dsl_prop.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_destroy.h>
	#include <sys/dsl_scan.h>
	#include <sys/zio_checksum.h>
	#include <sys/zfs_refcount.h>
	#include <sys/zfeature.h>
	#include <sys/dsl_userhold.h>
	#include <sys/abd.h>
	#include <sys/blake3.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <unistd.h>
	#include <getopt.h>
	#include <signal.h>
	#include <umem.h>
	#include <ctype.h>
	#include <math.h>
	#include <sys/fs/zfs.h>
	#include <zfs_fletcher.h>
	#include <libnvpair.h>
	#include <libzutil.h>
	#include <sys/crypto/icp.h>
	#include <sys/zfs_impl.h>
	-#if (__GLIBC__ && !__UCLIBC__)
	-#include <execinfo.h> /* for backtrace() */
	-#endif
	+#include <sys/backtrace.h>

	static int ztest_fd_data = -1;
	static int ztest_fd_rand = -1;

	typedef struct ztest_shared_hdr {
	uint64_t zh_hdr_size;
	uint64_t zh_opts_size;
	uint64_t zh_size;
	uint64_t zh_stats_size;
	uint64_t zh_stats_count;
	uint64_t zh_ds_size;
	uint64_t zh_ds_count;
	uint64_t zh_scratch_state_size;
	} ztest_shared_hdr_t;

	static ztest_shared_hdr_t *ztest_shared_hdr;

	enum ztest_class_state {
	ZTEST_VDEV_CLASS_OFF,
	ZTEST_VDEV_CLASS_ON,
	ZTEST_VDEV_CLASS_RND
	};

	/* Dedicated RAIDZ Expansion test states */
	typedef enum {
	RAIDZ_EXPAND_NONE, /* Default is none, must opt-in */
	RAIDZ_EXPAND_REQUESTED, /* The '-X' option was used */
	RAIDZ_EXPAND_STARTED, /* Testing has commenced */
	RAIDZ_EXPAND_KILLED, /* Reached the proccess kill */
	RAIDZ_EXPAND_CHECKED, /* Pool scrub verification done */
	} raidz_expand_test_state_t;


	#define ZO_GVARS_MAX_ARGLEN ((size_t)64)
	#define ZO_GVARS_MAX_COUNT ((size_t)10)

	typedef struct ztest_shared_opts {
	char zo_pool[ZFS_MAX_DATASET_NAME_LEN];
	char zo_dir[ZFS_MAX_DATASET_NAME_LEN];
	char zo_alt_ztest[MAXNAMELEN];
	char zo_alt_libpath[MAXNAMELEN];
	uint64_t zo_vdevs;
	uint64_t zo_vdevtime;
	size_t zo_vdev_size;
	int zo_ashift;
	int zo_mirrors;
	int zo_raid_do_expand;
	int zo_raid_children;
	int zo_raid_parity;
	char zo_raid_type[8];
	int zo_draid_data;
	int zo_draid_spares;
	int zo_datasets;
	int zo_threads;
	uint64_t zo_passtime;
	uint64_t zo_killrate;
	int zo_verbose;
	int zo_init;
	uint64_t zo_time;
	uint64_t zo_maxloops;
	uint64_t zo_metaslab_force_ganging;
	raidz_expand_test_state_t zo_raidz_expand_test;
	int zo_mmp_test;
	int zo_special_vdevs;
	int zo_dump_dbgmsg;
	int zo_gvars_count;
	char zo_gvars[ZO_GVARS_MAX_COUNT][ZO_GVARS_MAX_ARGLEN];
	} ztest_shared_opts_t;

	/* Default values for command line options. */
	#define DEFAULT_POOL "ztest"
	#define DEFAULT_VDEV_DIR "/tmp"
	#define DEFAULT_VDEV_COUNT 5
	#define DEFAULT_VDEV_SIZE (SPA_MINDEVSIZE * 4) /* 256m default size */
	#define DEFAULT_VDEV_SIZE_STR "256M"
	#define DEFAULT_ASHIFT SPA_MINBLOCKSHIFT
	#define DEFAULT_MIRRORS 2
	#define DEFAULT_RAID_CHILDREN 4
	#define DEFAULT_RAID_PARITY 1
	#define DEFAULT_DRAID_DATA 4
	#define DEFAULT_DRAID_SPARES 1
	#define DEFAULT_DATASETS_COUNT 7
	#define DEFAULT_THREADS 23
	#define DEFAULT_RUN_TIME 300 /* 300 seconds */
	#define DEFAULT_RUN_TIME_STR "300 sec"
	#define DEFAULT_PASS_TIME 60 /* 60 seconds */
	#define DEFAULT_PASS_TIME_STR "60 sec"
	#define DEFAULT_KILL_RATE 70 /* 70% kill rate */
	#define DEFAULT_KILLRATE_STR "70%"
	#define DEFAULT_INITS 1
	#define DEFAULT_MAX_LOOPS 50 /* 5 minutes */
	#define DEFAULT_FORCE_GANGING (64 << 10)
	#define DEFAULT_FORCE_GANGING_STR "64K"

	/* Simplifying assumption: -1 is not a valid default. */
	#define NO_DEFAULT -1

	static const ztest_shared_opts_t ztest_opts_defaults = {
	.zo_pool = DEFAULT_POOL,
	.zo_dir = DEFAULT_VDEV_DIR,
	.zo_alt_ztest = { '\0' },
	.zo_alt_libpath = { '\0' },
	.zo_vdevs = DEFAULT_VDEV_COUNT,
	.zo_ashift = DEFAULT_ASHIFT,
	.zo_mirrors = DEFAULT_MIRRORS,
	.zo_raid_children = DEFAULT_RAID_CHILDREN,
	.zo_raid_parity = DEFAULT_RAID_PARITY,
	.zo_raid_type = VDEV_TYPE_RAIDZ,
	.zo_vdev_size = DEFAULT_VDEV_SIZE,
	.zo_draid_data = DEFAULT_DRAID_DATA, /* data drives */
	.zo_draid_spares = DEFAULT_DRAID_SPARES, /* distributed spares */
	.zo_datasets = DEFAULT_DATASETS_COUNT,
	.zo_threads = DEFAULT_THREADS,
	.zo_passtime = DEFAULT_PASS_TIME,
	.zo_killrate = DEFAULT_KILL_RATE,
	.zo_verbose = 0,
	.zo_mmp_test = 0,
	.zo_init = DEFAULT_INITS,
	.zo_time = DEFAULT_RUN_TIME,
	.zo_maxloops = DEFAULT_MAX_LOOPS, /* max loops during spa_freeze() */
	.zo_metaslab_force_ganging = DEFAULT_FORCE_GANGING,
	.zo_special_vdevs = ZTEST_VDEV_CLASS_RND,
	.zo_gvars_count = 0,
	.zo_raidz_expand_test = RAIDZ_EXPAND_NONE,
	};

	extern uint64_t metaslab_force_ganging;
	extern uint64_t metaslab_df_alloc_threshold;
	extern uint64_t zfs_deadman_synctime_ms;
	extern uint_t metaslab_preload_limit;
	extern int zfs_compressed_arc_enabled;
	extern int zfs_abd_scatter_enabled;
	extern uint_t dmu_object_alloc_chunk_shift;
	extern boolean_t zfs_force_some_double_word_sm_entries;
	extern unsigned long zio_decompress_fail_fraction;
	extern unsigned long zfs_reconstruct_indirect_damage_fraction;
	extern uint64_t raidz_expand_max_reflow_bytes;
	extern uint_t raidz_expand_pause_point;


	static ztest_shared_opts_t *ztest_shared_opts;
	static ztest_shared_opts_t ztest_opts;
	static const char *const ztest_wkeydata = "abcdefghijklmnopqrstuvwxyz012345";

	typedef struct ztest_shared_ds {
	uint64_t zd_seq;
	} ztest_shared_ds_t;

	static ztest_shared_ds_t *ztest_shared_ds;
	#define ZTEST_GET_SHARED_DS(d) (&ztest_shared_ds[d])

	typedef struct ztest_scratch_state {
	uint64_t zs_raidz_scratch_verify_pause;
	} ztest_shared_scratch_state_t;

	static ztest_shared_scratch_state_t *ztest_scratch_state;

	#define BT_MAGIC 0x123456789abcdefULL
	#define MAXFAULTS(zs) \
	(MAX((zs)->zs_mirrors, 1) * (ztest_opts.zo_raid_parity + 1) - 1)

	enum ztest_io_type {
	ZTEST_IO_WRITE_TAG,
	ZTEST_IO_WRITE_PATTERN,
	ZTEST_IO_WRITE_ZEROES,
	ZTEST_IO_TRUNCATE,
	ZTEST_IO_SETATTR,
	ZTEST_IO_REWRITE,
	ZTEST_IO_TYPES
	};

	typedef struct ztest_block_tag {
	uint64_t bt_magic;
	uint64_t bt_objset;
	uint64_t bt_object;
	uint64_t bt_dnodesize;
	uint64_t bt_offset;
	uint64_t bt_gen;
	uint64_t bt_txg;
	uint64_t bt_crtxg;
	} ztest_block_tag_t;

	typedef struct bufwad {
	uint64_t bw_index;
	uint64_t bw_txg;
	uint64_t bw_data;
	} bufwad_t;

	/*
	* It would be better to use a rangelock_t per object. Unfortunately
	* the rangelock_t is not a drop-in replacement for rl_t, because we
	* still need to map from object ID to rangelock_t.
	*/
	typedef enum {
	ZTRL_READER,
	ZTRL_WRITER,
	ZTRL_APPEND
	} rl_type_t;

	typedef struct rll {
	void *rll_writer;
	int rll_readers;
	kmutex_t rll_lock;
	kcondvar_t rll_cv;
	} rll_t;

	typedef struct rl {
	uint64_t rl_object;
	uint64_t rl_offset;
	uint64_t rl_size;
	rll_t *rl_lock;
	} rl_t;

	#define ZTEST_RANGE_LOCKS 64
	#define ZTEST_OBJECT_LOCKS 64

	/*
	* Object descriptor. Used as a template for object lookup/create/remove.
	*/
	typedef struct ztest_od {
	uint64_t od_dir;
	uint64_t od_object;
	dmu_object_type_t od_type;
	dmu_object_type_t od_crtype;
	uint64_t od_blocksize;
	uint64_t od_crblocksize;
	uint64_t od_crdnodesize;
	uint64_t od_gen;
	uint64_t od_crgen;
	char od_name[ZFS_MAX_DATASET_NAME_LEN];
	} ztest_od_t;

	/*
	* Per-dataset state.
	*/
	typedef struct ztest_ds {
	ztest_shared_ds_t *zd_shared;
	objset_t *zd_os;
	pthread_rwlock_t zd_zilog_lock;
	zilog_t *zd_zilog;
	ztest_od_t zd_od; / debugging aid */
	char zd_name[ZFS_MAX_DATASET_NAME_LEN];
	kmutex_t zd_dirobj_lock;
	rll_t zd_object_lock[ZTEST_OBJECT_LOCKS];
	rll_t zd_range_lock[ZTEST_RANGE_LOCKS];
	} ztest_ds_t;

	/*
	* Per-iteration state.
	*/
	typedef void ztest_func_t(ztest_ds_t *zd, uint64_t id);

	typedef struct ztest_info {
	ztest_func_t zi_func; / test function */
	uint64_t zi_iters; /* iterations per execution */
	uint64_t zi_interval; / execute every <interval> seconds */
	const char zi_funcname; / name of test function */
	} ztest_info_t;

	typedef struct ztest_shared_callstate {
	uint64_t zc_count; /* per-pass count */
	uint64_t zc_time; /* per-pass time */
	uint64_t zc_next; /* next time to call this function */
	} ztest_shared_callstate_t;

	static ztest_shared_callstate_t *ztest_shared_callstate;
	#define ZTEST_GET_SHARED_CALLSTATE(c) (&ztest_shared_callstate[c])

	ztest_func_t ztest_dmu_read_write;
	ztest_func_t ztest_dmu_write_parallel;
	ztest_func_t ztest_dmu_object_alloc_free;
	ztest_func_t ztest_dmu_object_next_chunk;
	ztest_func_t ztest_dmu_commit_callbacks;
	ztest_func_t ztest_zap;
	ztest_func_t ztest_zap_parallel;
	ztest_func_t ztest_zil_commit;
	ztest_func_t ztest_zil_remount;
	ztest_func_t ztest_dmu_read_write_zcopy;
	ztest_func_t ztest_dmu_objset_create_destroy;
	ztest_func_t ztest_dmu_prealloc;
	ztest_func_t ztest_fzap;
	ztest_func_t ztest_dmu_snapshot_create_destroy;
	ztest_func_t ztest_dsl_prop_get_set;
	ztest_func_t ztest_spa_prop_get_set;
	ztest_func_t ztest_spa_create_destroy;
	ztest_func_t ztest_fault_inject;
	ztest_func_t ztest_dmu_snapshot_hold;
	ztest_func_t ztest_mmp_enable_disable;
	ztest_func_t ztest_scrub;
	ztest_func_t ztest_dsl_dataset_promote_busy;
	ztest_func_t ztest_vdev_attach_detach;
	ztest_func_t ztest_vdev_raidz_attach;
	ztest_func_t ztest_vdev_LUN_growth;
	ztest_func_t ztest_vdev_add_remove;
	ztest_func_t ztest_vdev_class_add;
	ztest_func_t ztest_vdev_aux_add_remove;
	ztest_func_t ztest_split_pool;
	ztest_func_t ztest_reguid;
	ztest_func_t ztest_spa_upgrade;
	ztest_func_t ztest_device_removal;
	ztest_func_t ztest_spa_checkpoint_create_discard;
	ztest_func_t ztest_initialize;
	ztest_func_t ztest_trim;
	ztest_func_t ztest_blake3;
	ztest_func_t ztest_fletcher;
	ztest_func_t ztest_fletcher_incr;
	ztest_func_t ztest_verify_dnode_bt;

	static uint64_t zopt_always = 0ULL * NANOSEC; /* all the time */
	static uint64_t zopt_incessant = 1ULL * NANOSEC / 10; /* every 1/10 second */
	static uint64_t zopt_often = 1ULL * NANOSEC; /* every second */
	static uint64_t zopt_sometimes = 10ULL * NANOSEC; /* every 10 seconds */
	static uint64_t zopt_rarely = 60ULL * NANOSEC; /* every 60 seconds */

	#define ZTI_INIT(func, iters, interval) \
	{ .zi_func = (func), \
	.zi_iters = (iters), \
	.zi_interval = (interval), \
	.zi_funcname = # func }

	static ztest_info_t ztest_info[] = {
	ZTI_INIT(ztest_dmu_read_write, 1, &zopt_always),
	ZTI_INIT(ztest_dmu_write_parallel, 10, &zopt_always),
	ZTI_INIT(ztest_dmu_object_alloc_free, 1, &zopt_always),
	ZTI_INIT(ztest_dmu_object_next_chunk, 1, &zopt_sometimes),
	ZTI_INIT(ztest_dmu_commit_callbacks, 1, &zopt_always),
	ZTI_INIT(ztest_zap, 30, &zopt_always),
	ZTI_INIT(ztest_zap_parallel, 100, &zopt_always),
	ZTI_INIT(ztest_split_pool, 1, &zopt_sometimes),
	ZTI_INIT(ztest_zil_commit, 1, &zopt_incessant),
	ZTI_INIT(ztest_zil_remount, 1, &zopt_sometimes),
	ZTI_INIT(ztest_dmu_read_write_zcopy, 1, &zopt_often),
	ZTI_INIT(ztest_dmu_objset_create_destroy, 1, &zopt_often),
	ZTI_INIT(ztest_dsl_prop_get_set, 1, &zopt_often),
	ZTI_INIT(ztest_spa_prop_get_set, 1, &zopt_sometimes),
	#if 0
	ZTI_INIT(ztest_dmu_prealloc, 1, &zopt_sometimes),
	#endif
	ZTI_INIT(ztest_fzap, 1, &zopt_sometimes),
	ZTI_INIT(ztest_dmu_snapshot_create_destroy, 1, &zopt_sometimes),
	ZTI_INIT(ztest_spa_create_destroy, 1, &zopt_sometimes),
	ZTI_INIT(ztest_fault_inject, 1, &zopt_sometimes),
	ZTI_INIT(ztest_dmu_snapshot_hold, 1, &zopt_sometimes),
	ZTI_INIT(ztest_mmp_enable_disable, 1, &zopt_sometimes),
	ZTI_INIT(ztest_reguid, 1, &zopt_rarely),
	ZTI_INIT(ztest_scrub, 1, &zopt_rarely),
	ZTI_INIT(ztest_spa_upgrade, 1, &zopt_rarely),
	ZTI_INIT(ztest_dsl_dataset_promote_busy, 1, &zopt_rarely),
	ZTI_INIT(ztest_vdev_attach_detach, 1, &zopt_sometimes),
	ZTI_INIT(ztest_vdev_raidz_attach, 1, &zopt_sometimes),
	ZTI_INIT(ztest_vdev_LUN_growth, 1, &zopt_rarely),
	ZTI_INIT(ztest_vdev_add_remove, 1, &ztest_opts.zo_vdevtime),
	ZTI_INIT(ztest_vdev_class_add, 1, &ztest_opts.zo_vdevtime),
	ZTI_INIT(ztest_vdev_aux_add_remove, 1, &ztest_opts.zo_vdevtime),
	ZTI_INIT(ztest_device_removal, 1, &zopt_sometimes),
	ZTI_INIT(ztest_spa_checkpoint_create_discard, 1, &zopt_rarely),
	ZTI_INIT(ztest_initialize, 1, &zopt_sometimes),
	ZTI_INIT(ztest_trim, 1, &zopt_sometimes),
	ZTI_INIT(ztest_blake3, 1, &zopt_rarely),
	ZTI_INIT(ztest_fletcher, 1, &zopt_rarely),
	ZTI_INIT(ztest_fletcher_incr, 1, &zopt_rarely),
	ZTI_INIT(ztest_verify_dnode_bt, 1, &zopt_sometimes),
	};

	#define ZTEST_FUNCS (sizeof (ztest_info) / sizeof (ztest_info_t))

	/*
	* The following struct is used to hold a list of uncalled commit callbacks.
	* The callbacks are ordered by txg number.
	*/
	typedef struct ztest_cb_list {
	kmutex_t zcl_callbacks_lock;
	list_t zcl_callbacks;
	} ztest_cb_list_t;

	/*
	* Stuff we need to share writably between parent and child.
	*/
	typedef struct ztest_shared {
	boolean_t zs_do_init;
	hrtime_t zs_proc_start;
	hrtime_t zs_proc_stop;
	hrtime_t zs_thread_start;
	hrtime_t zs_thread_stop;
	hrtime_t zs_thread_kill;
	uint64_t zs_enospc_count;
	uint64_t zs_vdev_next_leaf;
	uint64_t zs_vdev_aux;
	uint64_t zs_alloc;
	uint64_t zs_space;
	uint64_t zs_splits;
	uint64_t zs_mirrors;
	uint64_t zs_metaslab_sz;
	uint64_t zs_metaslab_df_alloc_threshold;
	uint64_t zs_guid;
	} ztest_shared_t;

	#define ID_PARALLEL -1ULL

	static char ztest_dev_template[] = "%s/%s.%llua";
	static char ztest_aux_template[] = "%s/%s.%s.%llu";
	static ztest_shared_t *ztest_shared;

	static spa_t *ztest_spa = NULL;
	static ztest_ds_t *ztest_ds;

	static kmutex_t ztest_vdev_lock;
	static boolean_t ztest_device_removal_active = B_FALSE;
	static boolean_t ztest_pool_scrubbed = B_FALSE;
	static kmutex_t ztest_checkpoint_lock;

	/*
	* The ztest_name_lock protects the pool and dataset namespace used by
	* the individual tests. To modify the namespace, consumers must grab
	* this lock as writer. Grabbing the lock as reader will ensure that the
	* namespace does not change while the lock is held.
	*/
	static pthread_rwlock_t ztest_name_lock;

	static boolean_t ztest_dump_core = B_TRUE;
	static boolean_t ztest_exiting;

	/* Global commit callback list */
	static ztest_cb_list_t zcl;
	/* Commit cb delay */
	static uint64_t zc_min_txg_delay = UINT64_MAX;
	static int zc_cb_counter = 0;

	/*
	* Minimum number of commit callbacks that need to be registered for us to check
	* whether the minimum txg delay is acceptable.
	*/
	#define ZTEST_COMMIT_CB_MIN_REG 100

	/*
	* If a number of txgs equal to this threshold have been created after a commit
	* callback has been registered but not called, then we assume there is an
	* implementation bug.
	*/
	#define ZTEST_COMMIT_CB_THRESH (TXG_CONCURRENT_STATES + 1000)

	enum ztest_object {
	ZTEST_META_DNODE = 0,
	ZTEST_DIROBJ,
	ZTEST_OBJECTS
	};

	static __attribute__((noreturn)) void usage(boolean_t requested);
	static int ztest_scrub_impl(spa_t *spa);

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

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

	static void
	dump_debug_buffer(void)
	{
	ssize_t ret __attribute__((unused));

	if (!ztest_opts.zo_dump_dbgmsg)
	return;

	/*
	* We use write() instead of printf() so that this function
	* is safe to call from a signal handler.
	*/
	- ret = write(STDOUT_FILENO, "\n", 1);
	- zfs_dbgmsg_print("ztest");
	+ ret = write(STDERR_FILENO, "\n", 1);
	+ zfs_dbgmsg_print(STDERR_FILENO, "ztest");
	}

	-#define BACKTRACE_SZ 100
	-
	static void sig_handler(int signo)
	{
	struct sigaction action;
	-#if (__GLIBC__ && !__UCLIBC__) /* backtrace() is a GNU extension */
	- int nptrs;
	- void *buffer[BACKTRACE_SZ];

	- nptrs = backtrace(buffer, BACKTRACE_SZ);
	- backtrace_symbols_fd(buffer, nptrs, STDERR_FILENO);
	-#endif
	+ libspl_backtrace(STDERR_FILENO);
	dump_debug_buffer();

	/*
	* Restore default action and re-raise signal so SIGSEGV and
	* SIGABRT can trigger a core dump.
	*/
	action.sa_handler = SIG_DFL;
	sigemptyset(&action.sa_mask);
	action.sa_flags = 0;
	(void) sigaction(signo, &action, NULL);
	raise(signo);
	}

	#define FATAL_MSG_SZ 1024

	static const char *fatal_msg;

	static __attribute__((format(printf, 2, 3))) __attribute__((noreturn)) void
	fatal(int do_perror, const char *message, ...)
	{
	va_list args;
	int save_errno = errno;
	char *buf;

	(void) fflush(stdout);
	buf = umem_alloc(FATAL_MSG_SZ, UMEM_NOFAIL);
	if (buf == NULL)
	goto out;

	va_start(args, message);
	(void) sprintf(buf, "ztest: ");
	/* LINTED */
	(void) vsprintf(buf + strlen(buf), message, args);
	va_end(args);
	if (do_perror) {
	(void) snprintf(buf + strlen(buf), FATAL_MSG_SZ - strlen(buf),
	": %s", strerror(save_errno));
	}
	(void) fprintf(stderr, "%s\n", buf);
	fatal_msg = buf; /* to ease debugging */

	out:
	if (ztest_dump_core)
	abort();
	else
	dump_debug_buffer();

	exit(3);
	}

	static int
	str2shift(const char *buf)
	{
	const char *ends = "BKMGTPEZ";
	int i;

	if (buf[0] == '\0')
	return (0);
	for (i = 0; i < strlen(ends); i++) {
	if (toupper(buf[0]) == ends[i])
	break;
	}
	if (i == strlen(ends)) {
	(void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n",
	buf);
	usage(B_FALSE);
	}
	if (buf[1] == '\0' \|\| (toupper(buf[1]) == 'B' && buf[2] == '\0')) {
	return (10*i);
	}
	(void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", buf);
	usage(B_FALSE);
	}

	static uint64_t
	nicenumtoull(const char *buf)
	{
	char *end;
	uint64_t val;

	val = strtoull(buf, &end, 0);
	if (end == buf) {
	(void) fprintf(stderr, "ztest: bad numeric value: %s\n", buf);
	usage(B_FALSE);
	} else if (end[0] == '.') {
	double fval = strtod(buf, &end);
	fval *= pow(2, str2shift(end));
	/*
	* UINT64_MAX is not exactly representable as a double.
	* The closest representation is UINT64_MAX + 1, so we
	* use a >= comparison instead of > for the bounds check.
	*/
	if (fval >= (double)UINT64_MAX) {
	(void) fprintf(stderr, "ztest: value too large: %s\n",
	buf);
	usage(B_FALSE);
	}
	val = (uint64_t)fval;
	} else {
	int shift = str2shift(end);
	if (shift >= 64 \|\| (val << shift) >> shift != val) {
	(void) fprintf(stderr, "ztest: value too large: %s\n",
	buf);
	usage(B_FALSE);
	}
	val <<= shift;
	}
	return (val);
	}

	typedef struct ztest_option {
	const char short_opt;
	const char *long_opt;
	const char *long_opt_param;
	const char *comment;
	unsigned int default_int;
	const char *default_str;
	} ztest_option_t;

	/*
	* The following option_table is used for generating the usage info as well as
	* the long and short option information for calling getopt_long().
	*/
	static ztest_option_t option_table[] = {
	{ 'v', "vdevs", "INTEGER", "Number of vdevs", DEFAULT_VDEV_COUNT,
	NULL},
	{ 's', "vdev-size", "INTEGER", "Size of each vdev",
	NO_DEFAULT, DEFAULT_VDEV_SIZE_STR},
	{ 'a', "alignment-shift", "INTEGER",
	"Alignment shift; use 0 for random", DEFAULT_ASHIFT, NULL},
	{ 'm', "mirror-copies", "INTEGER", "Number of mirror copies",
	DEFAULT_MIRRORS, NULL},
	{ 'r', "raid-disks", "INTEGER", "Number of raidz/draid disks",
	DEFAULT_RAID_CHILDREN, NULL},
	{ 'R', "raid-parity", "INTEGER", "Raid parity",
	DEFAULT_RAID_PARITY, NULL},
	{ 'K', "raid-kind", "raidz\|eraidz\|draid\|random", "Raid kind",
	NO_DEFAULT, "random"},
	{ 'D', "draid-data", "INTEGER", "Number of draid data drives",
	DEFAULT_DRAID_DATA, NULL},
	{ 'S', "draid-spares", "INTEGER", "Number of draid spares",
	DEFAULT_DRAID_SPARES, NULL},
	{ 'd', "datasets", "INTEGER", "Number of datasets",
	DEFAULT_DATASETS_COUNT, NULL},
	{ 't', "threads", "INTEGER", "Number of ztest threads",
	DEFAULT_THREADS, NULL},
	{ 'g', "gang-block-threshold", "INTEGER",
	"Metaslab gang block threshold",
	NO_DEFAULT, DEFAULT_FORCE_GANGING_STR},
	{ 'i', "init-count", "INTEGER", "Number of times to initialize pool",
	DEFAULT_INITS, NULL},
	{ 'k', "kill-percentage", "INTEGER", "Kill percentage",
	NO_DEFAULT, DEFAULT_KILLRATE_STR},
	{ 'p', "pool-name", "STRING", "Pool name",
	NO_DEFAULT, DEFAULT_POOL},
	{ 'f', "vdev-file-directory", "PATH", "File directory for vdev files",
	NO_DEFAULT, DEFAULT_VDEV_DIR},
	{ 'M', "multi-host", NULL,
	"Multi-host; simulate pool imported on remote host",
	NO_DEFAULT, NULL},
	{ 'E', "use-existing-pool", NULL,
	"Use existing pool instead of creating new one", NO_DEFAULT, NULL},
	{ 'T', "run-time", "INTEGER", "Total run time",
	NO_DEFAULT, DEFAULT_RUN_TIME_STR},
	{ 'P', "pass-time", "INTEGER", "Time per pass",
	NO_DEFAULT, DEFAULT_PASS_TIME_STR},
	{ 'F', "freeze-loops", "INTEGER", "Max loops in spa_freeze()",
	DEFAULT_MAX_LOOPS, NULL},
	{ 'B', "alt-ztest", "PATH", "Alternate ztest path",
	NO_DEFAULT, NULL},
	{ 'C', "vdev-class-state", "on\|off\|random", "vdev class state",
	NO_DEFAULT, "random"},
	{ 'X', "raidz-expansion", NULL,
	"Perform a dedicated raidz expansion test",
	NO_DEFAULT, NULL},
	{ 'o', "option", "\"OPTION=INTEGER\"",
	"Set global variable to an unsigned 32-bit integer value",
	NO_DEFAULT, NULL},
	{ 'G', "dump-debug-msg", NULL,
	"Dump zfs_dbgmsg buffer before exiting due to an error",
	NO_DEFAULT, NULL},
	{ 'V', "verbose", NULL,
	"Verbose (use multiple times for ever more verbosity)",
	NO_DEFAULT, NULL},
	{ 'h', "help", NULL, "Show this help",
	NO_DEFAULT, NULL},
	{0, 0, 0, 0, 0, 0}
	};

	static struct option *long_opts = NULL;
	static char *short_opts = NULL;

	static void
	init_options(void)
	{
	ASSERT3P(long_opts, ==, NULL);
	ASSERT3P(short_opts, ==, NULL);

	int count = sizeof (option_table) / sizeof (option_table[0]);
	long_opts = umem_alloc(sizeof (struct option) * count, UMEM_NOFAIL);

	short_opts = umem_alloc(sizeof (char) * 2 * count, UMEM_NOFAIL);
	int short_opt_index = 0;

	for (int i = 0; i < count; i++) {
	long_opts[i].val = option_table[i].short_opt;
	long_opts[i].name = option_table[i].long_opt;
	long_opts[i].has_arg = option_table[i].long_opt_param != NULL
	? required_argument : no_argument;
	long_opts[i].flag = NULL;
	short_opts[short_opt_index++] = option_table[i].short_opt;
	if (option_table[i].long_opt_param != NULL) {
	short_opts[short_opt_index++] = ':';
	}
	}
	}

	static void
	fini_options(void)
	{
	int count = sizeof (option_table) / sizeof (option_table[0]);

	umem_free(long_opts, sizeof (struct option) * count);
	umem_free(short_opts, sizeof (char) * 2 * count);

	long_opts = NULL;
	short_opts = NULL;
	}

	static __attribute__((noreturn)) void
	usage(boolean_t requested)
	{
	char option[80];
	FILE *fp = requested ? stdout : stderr;

	(void) fprintf(fp, "Usage: %s [OPTIONS...]\n", DEFAULT_POOL);
	for (int i = 0; option_table[i].short_opt != 0; i++) {
	if (option_table[i].long_opt_param != NULL) {
	(void) sprintf(option, " -%c --%s=%s",
	option_table[i].short_opt,
	option_table[i].long_opt,
	option_table[i].long_opt_param);
	} else {
	(void) sprintf(option, " -%c --%s",
	option_table[i].short_opt,
	option_table[i].long_opt);
	}
	(void) fprintf(fp, " %-43s%s", option,
	option_table[i].comment);

	if (option_table[i].long_opt_param != NULL) {
	if (option_table[i].default_str != NULL) {
	(void) fprintf(fp, " (default: %s)",
	option_table[i].default_str);
	} else if (option_table[i].default_int != NO_DEFAULT) {
	(void) fprintf(fp, " (default: %u)",
	option_table[i].default_int);
	}
	}
	(void) fprintf(fp, "\n");
	}
	exit(requested ? 0 : 1);
	}

	static uint64_t
	ztest_random(uint64_t range)
	{
	uint64_t r;

	ASSERT3S(ztest_fd_rand, >=, 0);

	if (range == 0)
	return (0);

	if (read(ztest_fd_rand, &r, sizeof (r)) != sizeof (r))
	fatal(B_TRUE, "short read from /dev/urandom");

	return (r % range);
	}

	static void
	ztest_parse_name_value(const char input, ztest_shared_opts_t zo)
	{
	char name[32];
	char *value;
	int state = ZTEST_VDEV_CLASS_RND;

	(void) strlcpy(name, input, sizeof (name));

	value = strchr(name, '=');
	if (value == NULL) {
	(void) fprintf(stderr, "missing value in property=value "
	"'-C' argument (%s)\n", input);
	usage(B_FALSE);
	}
	*(value) = '\0';
	value++;

	if (strcmp(value, "on") == 0) {
	state = ZTEST_VDEV_CLASS_ON;
	} else if (strcmp(value, "off") == 0) {
	state = ZTEST_VDEV_CLASS_OFF;
	} else if (strcmp(value, "random") == 0) {
	state = ZTEST_VDEV_CLASS_RND;
	} else {
	(void) fprintf(stderr, "invalid property value '%s'\n", value);
	usage(B_FALSE);
	}

	if (strcmp(name, "special") == 0) {
	zo->zo_special_vdevs = state;
	} else {
	(void) fprintf(stderr, "invalid property name '%s'\n", name);
	usage(B_FALSE);
	}
	if (zo->zo_verbose >= 3)
	(void) printf("%s vdev state is '%s'\n", name, value);
	}

	static void
	process_options(int argc, char **argv)
	{
	char *path;
	ztest_shared_opts_t *zo = &ztest_opts;

	int opt;
	uint64_t value;
	const char *raid_kind = "random";

	memcpy(zo, &ztest_opts_defaults, sizeof (*zo));

	init_options();

	while ((opt = getopt_long(argc, argv, short_opts, long_opts,
	NULL)) != EOF) {
	value = 0;
	switch (opt) {
	case 'v':
	case 's':
	case 'a':
	case 'm':
	case 'r':
	case 'R':
	case 'D':
	case 'S':
	case 'd':
	case 't':
	case 'g':
	case 'i':
	case 'k':
	case 'T':
	case 'P':
	case 'F':
	value = nicenumtoull(optarg);
	}
	switch (opt) {
	case 'v':
	zo->zo_vdevs = value;
	break;
	case 's':
	zo->zo_vdev_size = MAX(SPA_MINDEVSIZE, value);
	break;
	case 'a':
	zo->zo_ashift = value;
	break;
	case 'm':
	zo->zo_mirrors = value;
	break;
	case 'r':
	zo->zo_raid_children = MAX(1, value);
	break;
	case 'R':
	zo->zo_raid_parity = MIN(MAX(value, 1), 3);
	break;
	case 'K':
	raid_kind = optarg;
	break;
	case 'D':
	zo->zo_draid_data = MAX(1, value);
	break;
	case 'S':
	zo->zo_draid_spares = MAX(1, value);
	break;
	case 'd':
	zo->zo_datasets = MAX(1, value);
	break;
	case 't':
	zo->zo_threads = MAX(1, value);
	break;
	case 'g':
	zo->zo_metaslab_force_ganging =
	MAX(SPA_MINBLOCKSIZE << 1, value);
	break;
	case 'i':
	zo->zo_init = value;
	break;
	case 'k':
	zo->zo_killrate = value;
	break;
	case 'p':
	(void) strlcpy(zo->zo_pool, optarg,
	sizeof (zo->zo_pool));
	break;
	case 'f':
	path = realpath(optarg, NULL);
	if (path == NULL) {
	(void) fprintf(stderr, "error: %s: %s\n",
	optarg, strerror(errno));
	usage(B_FALSE);
	} else {
	(void) strlcpy(zo->zo_dir, path,
	sizeof (zo->zo_dir));
	free(path);
	}
	break;
	case 'M':
	zo->zo_mmp_test = 1;
	break;
	case 'V':
	zo->zo_verbose++;
	break;
	case 'X':
	zo->zo_raidz_expand_test = RAIDZ_EXPAND_REQUESTED;
	break;
	case 'E':
	zo->zo_init = 0;
	break;
	case 'T':
	zo->zo_time = value;
	break;
	case 'P':
	zo->zo_passtime = MAX(1, value);
	break;
	case 'F':
	zo->zo_maxloops = MAX(1, value);
	break;
	case 'B':
	(void) strlcpy(zo->zo_alt_ztest, optarg,
	sizeof (zo->zo_alt_ztest));
	break;
	case 'C':
	ztest_parse_name_value(optarg, zo);
	break;
	case 'o':
	if (zo->zo_gvars_count >= ZO_GVARS_MAX_COUNT) {
	(void) fprintf(stderr,
	"max global var count (%zu) exceeded\n",
	ZO_GVARS_MAX_COUNT);
	usage(B_FALSE);
	}
	char *v = zo->zo_gvars[zo->zo_gvars_count];
	if (strlcpy(v, optarg, ZO_GVARS_MAX_ARGLEN) >=
	ZO_GVARS_MAX_ARGLEN) {
	(void) fprintf(stderr,
	"global var option '%s' is too long\n",
	optarg);
	usage(B_FALSE);
	}
	zo->zo_gvars_count++;
	break;
	case 'G':
	zo->zo_dump_dbgmsg = 1;
	break;
	case 'h':
	usage(B_TRUE);
	break;
	case '?':
	default:
	usage(B_FALSE);
	break;
	}
	}

	fini_options();

	/* Force compatible options for raidz expansion run */
	if (zo->zo_raidz_expand_test == RAIDZ_EXPAND_REQUESTED) {
	zo->zo_mmp_test = 0;
	zo->zo_mirrors = 0;
	zo->zo_vdevs = 1;
	zo->zo_vdev_size = DEFAULT_VDEV_SIZE * 2;
	zo->zo_raid_do_expand = B_FALSE;
	raid_kind = "raidz";
	}

	if (strcmp(raid_kind, "random") == 0) {
	switch (ztest_random(3)) {
	case 0:
	raid_kind = "raidz";
	break;
	case 1:
	raid_kind = "eraidz";
	break;
	case 2:
	raid_kind = "draid";
	break;
	}

	if (ztest_opts.zo_verbose >= 3)
	(void) printf("choosing RAID type '%s'\n", raid_kind);
	}

	if (strcmp(raid_kind, "draid") == 0) {
	uint64_t min_devsize;

	/* With fewer disk use 256M, otherwise 128M is OK */
	min_devsize = (ztest_opts.zo_raid_children < 16) ?
	(256ULL << 20) : (128ULL << 20);

	/* No top-level mirrors with dRAID for now */
	zo->zo_mirrors = 0;

	/* Use more appropriate defaults for dRAID */
	if (zo->zo_vdevs == ztest_opts_defaults.zo_vdevs)
	zo->zo_vdevs = 1;
	if (zo->zo_raid_children ==
	ztest_opts_defaults.zo_raid_children)
	zo->zo_raid_children = 16;
	if (zo->zo_ashift < 12)
	zo->zo_ashift = 12;
	if (zo->zo_vdev_size < min_devsize)
	zo->zo_vdev_size = min_devsize;

	if (zo->zo_draid_data + zo->zo_raid_parity >
	zo->zo_raid_children - zo->zo_draid_spares) {
	(void) fprintf(stderr, "error: too few draid "
	"children (%d) for stripe width (%d)\n",
	zo->zo_raid_children,
	zo->zo_draid_data + zo->zo_raid_parity);
	usage(B_FALSE);
	}

	(void) strlcpy(zo->zo_raid_type, VDEV_TYPE_DRAID,
	sizeof (zo->zo_raid_type));

	} else if (strcmp(raid_kind, "eraidz") == 0) {
	/* using eraidz (expandable raidz) */
	zo->zo_raid_do_expand = B_TRUE;

	/* tests expect top-level to be raidz */
	zo->zo_mirrors = 0;
	zo->zo_vdevs = 1;

	/* Make sure parity is less than data columns */
	zo->zo_raid_parity = MIN(zo->zo_raid_parity,
	zo->zo_raid_children - 1);

	} else /* using raidz */ {
	ASSERT0(strcmp(raid_kind, "raidz"));

	zo->zo_raid_parity = MIN(zo->zo_raid_parity,
	zo->zo_raid_children - 1);
	}

	zo->zo_vdevtime =
	(zo->zo_vdevs > 0 ? zo->zo_time * NANOSEC / zo->zo_vdevs :
	UINT64_MAX >> 2);

	if (*zo->zo_alt_ztest) {
	const char *invalid_what = "ztest";
	char *val = zo->zo_alt_ztest;
	if (0 != access(val, X_OK) \|\|
	(strrchr(val, '/') == NULL && (errno == EINVAL)))
	goto invalid;

	int dirlen = strrchr(val, '/') - val;
	strlcpy(zo->zo_alt_libpath, val,
	MIN(sizeof (zo->zo_alt_libpath), dirlen + 1));
	invalid_what = "library path", val = zo->zo_alt_libpath;
	if (strrchr(val, '/') == NULL && (errno == EINVAL))
	goto invalid;
	*strrchr(val, '/') = '\0';
	strlcat(val, "/lib", sizeof (zo->zo_alt_libpath));

	if (0 != access(zo->zo_alt_libpath, X_OK))
	goto invalid;
	return;

	invalid:
	ztest_dump_core = B_FALSE;
	fatal(B_TRUE, "invalid alternate %s %s", invalid_what, val);
	}
	}

	static void
	ztest_kill(ztest_shared_t *zs)
	{
	zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(ztest_spa));
	zs->zs_space = metaslab_class_get_space(spa_normal_class(ztest_spa));

	/*
	* Before we kill ourselves, make sure that the config is updated.
	* See comment above spa_write_cachefile().
	*/
	if (raidz_expand_pause_point != RAIDZ_EXPAND_PAUSE_NONE) {
	if (mutex_tryenter(&spa_namespace_lock)) {
	spa_write_cachefile(ztest_spa, B_FALSE, B_FALSE,
	B_FALSE);
	mutex_exit(&spa_namespace_lock);

	ztest_scratch_state->zs_raidz_scratch_verify_pause =
	raidz_expand_pause_point;
	} else {
	/*
	* Do not verify scratch object in case if
	* spa_namespace_lock cannot be acquired,
	* it can cause deadlock in spa_config_update().
	*/
	raidz_expand_pause_point = RAIDZ_EXPAND_PAUSE_NONE;

	return;
	}
	} else {
	mutex_enter(&spa_namespace_lock);
	spa_write_cachefile(ztest_spa, B_FALSE, B_FALSE, B_FALSE);
	mutex_exit(&spa_namespace_lock);
	}

	(void) raise(SIGKILL);
	}

	static void
	ztest_record_enospc(const char *s)
	{
	(void) s;
	ztest_shared->zs_enospc_count++;
	}

	static uint64_t
	ztest_get_ashift(void)
	{
	if (ztest_opts.zo_ashift == 0)
	return (SPA_MINBLOCKSHIFT + ztest_random(5));
	return (ztest_opts.zo_ashift);
	}

	static boolean_t
	ztest_is_draid_spare(const char *name)
	{
	uint64_t spare_id = 0, parity = 0, vdev_id = 0;

	if (sscanf(name, VDEV_TYPE_DRAID "%"PRIu64"-%"PRIu64"-%"PRIu64"",
	&parity, &vdev_id, &spare_id) == 3) {
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	static nvlist_t *
	make_vdev_file(const char path, const char aux, const char *pool,
	size_t size, uint64_t ashift)
	{
	char *pathbuf = NULL;
	uint64_t vdev;
	nvlist_t *file;
	boolean_t draid_spare = B_FALSE;


	if (ashift == 0)
	ashift = ztest_get_ashift();

	if (path == NULL) {
	pathbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
	path = pathbuf;

	if (aux != NULL) {
	vdev = ztest_shared->zs_vdev_aux;
	(void) snprintf(pathbuf, MAXPATHLEN,
	ztest_aux_template, ztest_opts.zo_dir,
	pool == NULL ? ztest_opts.zo_pool : pool,
	aux, vdev);
	} else {
	vdev = ztest_shared->zs_vdev_next_leaf++;
	(void) snprintf(pathbuf, MAXPATHLEN,
	ztest_dev_template, ztest_opts.zo_dir,
	pool == NULL ? ztest_opts.zo_pool : pool, vdev);
	}
	} else {
	draid_spare = ztest_is_draid_spare(path);
	}

	if (size != 0 && !draid_spare) {
	int fd = open(path, O_RDWR \| O_CREAT \| O_TRUNC, 0666);
	if (fd == -1)
	fatal(B_TRUE, "can't open %s", path);
	if (ftruncate(fd, size) != 0)
	fatal(B_TRUE, "can't ftruncate %s", path);
	(void) close(fd);
	}

	file = fnvlist_alloc();
	fnvlist_add_string(file, ZPOOL_CONFIG_TYPE,
	draid_spare ? VDEV_TYPE_DRAID_SPARE : VDEV_TYPE_FILE);
	fnvlist_add_string(file, ZPOOL_CONFIG_PATH, path);
	fnvlist_add_uint64(file, ZPOOL_CONFIG_ASHIFT, ashift);
	umem_free(pathbuf, MAXPATHLEN);

	return (file);
	}

	static nvlist_t *
	make_vdev_raid(const char path, const char aux, const char *pool, size_t size,
	uint64_t ashift, int r)
	{
	nvlist_t raid, *child;
	int c;

	if (r < 2)
	return (make_vdev_file(path, aux, pool, size, ashift));
	child = umem_alloc(r * sizeof (nvlist_t *), UMEM_NOFAIL);

	for (c = 0; c < r; c++)
	child[c] = make_vdev_file(path, aux, pool, size, ashift);

	raid = fnvlist_alloc();
	fnvlist_add_string(raid, ZPOOL_CONFIG_TYPE,
	ztest_opts.zo_raid_type);
	fnvlist_add_uint64(raid, ZPOOL_CONFIG_NPARITY,
	ztest_opts.zo_raid_parity);
	fnvlist_add_nvlist_array(raid, ZPOOL_CONFIG_CHILDREN,
	(const nvlist_t **)child, r);

	if (strcmp(ztest_opts.zo_raid_type, VDEV_TYPE_DRAID) == 0) {
	uint64_t ndata = ztest_opts.zo_draid_data;
	uint64_t nparity = ztest_opts.zo_raid_parity;
	uint64_t nspares = ztest_opts.zo_draid_spares;
	uint64_t children = ztest_opts.zo_raid_children;
	uint64_t ngroups = 1;

	/*
	* Calculate the minimum number of groups required to fill a
	* slice. This is the LCM of the stripe width (data + parity)
	* and the number of data drives (children - spares).
	*/
	while (ngroups * (ndata + nparity) % (children - nspares) != 0)
	ngroups++;

	/* Store the basic dRAID configuration. */
	fnvlist_add_uint64(raid, ZPOOL_CONFIG_DRAID_NDATA, ndata);
	fnvlist_add_uint64(raid, ZPOOL_CONFIG_DRAID_NSPARES, nspares);
	fnvlist_add_uint64(raid, ZPOOL_CONFIG_DRAID_NGROUPS, ngroups);
	}

	for (c = 0; c < r; c++)
	fnvlist_free(child[c]);

	umem_free(child, r * sizeof (nvlist_t *));

	return (raid);
	}

	static nvlist_t *
	make_vdev_mirror(const char path, const char aux, const char *pool,
	size_t size, uint64_t ashift, int r, int m)
	{
	nvlist_t mirror, *child;
	int c;

	if (m < 1)
	return (make_vdev_raid(path, aux, pool, size, ashift, r));

	child = umem_alloc(m * sizeof (nvlist_t *), UMEM_NOFAIL);

	for (c = 0; c < m; c++)
	child[c] = make_vdev_raid(path, aux, pool, size, ashift, r);

	mirror = fnvlist_alloc();
	fnvlist_add_string(mirror, ZPOOL_CONFIG_TYPE, VDEV_TYPE_MIRROR);
	fnvlist_add_nvlist_array(mirror, ZPOOL_CONFIG_CHILDREN,
	(const nvlist_t **)child, m);

	for (c = 0; c < m; c++)
	fnvlist_free(child[c]);

	umem_free(child, m * sizeof (nvlist_t *));

	return (mirror);
	}

	static nvlist_t *
	make_vdev_root(const char path, const char aux, const char *pool, size_t size,
	uint64_t ashift, const char *class, int r, int m, int t)
	{
	nvlist_t root, *child;
	int c;
	boolean_t log;

	ASSERT3S(t, >, 0);

	log = (class != NULL && strcmp(class, "log") == 0);

	child = umem_alloc(t * sizeof (nvlist_t *), UMEM_NOFAIL);

	for (c = 0; c < t; c++) {
	child[c] = make_vdev_mirror(path, aux, pool, size, ashift,
	r, m);
	fnvlist_add_uint64(child[c], ZPOOL_CONFIG_IS_LOG, log);

	if (class != NULL && class[0] != '\0') {
	ASSERT(m > 1 \|\| log); /* expecting a mirror */
	fnvlist_add_string(child[c],
	ZPOOL_CONFIG_ALLOCATION_BIAS, class);
	}
	}

	root = fnvlist_alloc();
	fnvlist_add_string(root, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT);
	fnvlist_add_nvlist_array(root, aux ? aux : ZPOOL_CONFIG_CHILDREN,
	(const nvlist_t **)child, t);

	for (c = 0; c < t; c++)
	fnvlist_free(child[c]);

	umem_free(child, t * sizeof (nvlist_t *));

	return (root);
	}

	/*
	* Find a random spa version. Returns back a random spa version in the
	* range [initial_version, SPA_VERSION_FEATURES].
	*/
	static uint64_t
	ztest_random_spa_version(uint64_t initial_version)
	{
	uint64_t version = initial_version;

	if (version <= SPA_VERSION_BEFORE_FEATURES) {
	version = version +
	ztest_random(SPA_VERSION_BEFORE_FEATURES - version + 1);
	}

	if (version > SPA_VERSION_BEFORE_FEATURES)
	version = SPA_VERSION_FEATURES;

	ASSERT(SPA_VERSION_IS_SUPPORTED(version));
	return (version);
	}

	static int
	ztest_random_blocksize(void)
	{
	ASSERT3U(ztest_spa->spa_max_ashift, !=, 0);

	/*
	* Choose a block size >= the ashift.
	* If the SPA supports new MAXBLOCKSIZE, test up to 1MB blocks.
	*/
	int maxbs = SPA_OLD_MAXBLOCKSHIFT;
	if (spa_maxblocksize(ztest_spa) == SPA_MAXBLOCKSIZE)
	maxbs = 20;
	uint64_t block_shift =
	ztest_random(maxbs - ztest_spa->spa_max_ashift + 1);
	return (1 << (SPA_MINBLOCKSHIFT + block_shift));
	}

	static int
	ztest_random_dnodesize(void)
	{
	int slots;
	int max_slots = spa_maxdnodesize(ztest_spa) >> DNODE_SHIFT;

	if (max_slots == DNODE_MIN_SLOTS)
	return (DNODE_MIN_SIZE);

	/*
	* Weight the random distribution more heavily toward smaller
	* dnode sizes since that is more likely to reflect real-world
	* usage.
	*/
	ASSERT3U(max_slots, >, 4);
	switch (ztest_random(10)) {
	case 0:
	slots = 5 + ztest_random(max_slots - 4);
	break;
	case 1 ... 4:
	slots = 2 + ztest_random(3);
	break;
	default:
	slots = 1;
	break;
	}

	return (slots << DNODE_SHIFT);
	}

	static int
	ztest_random_ibshift(void)
	{
	return (DN_MIN_INDBLKSHIFT +
	ztest_random(DN_MAX_INDBLKSHIFT - DN_MIN_INDBLKSHIFT + 1));
	}

	static uint64_t
	ztest_random_vdev_top(spa_t *spa, boolean_t log_ok)
	{
	uint64_t top;
	vdev_t *rvd = spa->spa_root_vdev;
	vdev_t *tvd;

	ASSERT3U(spa_config_held(spa, SCL_ALL, RW_READER), !=, 0);

	do {
	top = ztest_random(rvd->vdev_children);
	tvd = rvd->vdev_child[top];
	} while (!vdev_is_concrete(tvd) \|\| (tvd->vdev_islog && !log_ok) \|\|
	tvd->vdev_mg == NULL \|\| tvd->vdev_mg->mg_class == NULL);

	return (top);
	}

	static uint64_t
	ztest_random_dsl_prop(zfs_prop_t prop)
	{
	uint64_t value;

	do {
	value = zfs_prop_random_value(prop, ztest_random(-1ULL));
	} while (prop == ZFS_PROP_CHECKSUM && value == ZIO_CHECKSUM_OFF);

	return (value);
	}

	static int
	ztest_dsl_prop_set_uint64(char *osname, zfs_prop_t prop, uint64_t value,
	boolean_t inherit)
	{
	const char *propname = zfs_prop_to_name(prop);
	const char *valname;
	char *setpoint;
	uint64_t curval;
	int error;

	error = dsl_prop_set_int(osname, propname,
	(inherit ? ZPROP_SRC_NONE : ZPROP_SRC_LOCAL), value);

	if (error == ENOSPC) {
	ztest_record_enospc(FTAG);
	return (error);
	}
	ASSERT0(error);

	setpoint = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
	VERIFY0(dsl_prop_get_integer(osname, propname, &curval, setpoint));

	if (ztest_opts.zo_verbose >= 6) {
	int err;

	err = zfs_prop_index_to_string(prop, curval, &valname);
	if (err)
	(void) printf("%s %s = %llu at '%s'\n", osname,
	propname, (unsigned long long)curval, setpoint);
	else
	(void) printf("%s %s = %s at '%s'\n",
	osname, propname, valname, setpoint);
	}
	umem_free(setpoint, MAXPATHLEN);

	return (error);
	}

	static int
	ztest_spa_prop_set_uint64(zpool_prop_t prop, uint64_t value)
	{
	spa_t *spa = ztest_spa;
	nvlist_t *props = NULL;
	int error;

	props = fnvlist_alloc();
	fnvlist_add_uint64(props, zpool_prop_to_name(prop), value);

	error = spa_prop_set(spa, props);

	fnvlist_free(props);

	if (error == ENOSPC) {
	ztest_record_enospc(FTAG);
	return (error);
	}
	ASSERT0(error);

	return (error);
	}

	static int
	ztest_dmu_objset_own(const char *name, dmu_objset_type_t type,
	boolean_t readonly, boolean_t decrypt, const void tag, objset_t *osp)
	{
	int err;
	char *cp = NULL;
	char ddname[ZFS_MAX_DATASET_NAME_LEN];

	strlcpy(ddname, name, sizeof (ddname));
	cp = strchr(ddname, '@');
	if (cp != NULL)
	*cp = '\0';

	err = dmu_objset_own(name, type, readonly, decrypt, tag, osp);
	while (decrypt && err == EACCES) {
	dsl_crypto_params_t *dcp;
	nvlist_t *crypto_args = fnvlist_alloc();

	fnvlist_add_uint8_array(crypto_args, "wkeydata",
	(uint8_t *)ztest_wkeydata, WRAPPING_KEY_LEN);
	VERIFY0(dsl_crypto_params_create_nvlist(DCP_CMD_NONE, NULL,
	crypto_args, &dcp));
	err = spa_keystore_load_wkey(ddname, dcp, B_FALSE);
	/*
	* Note: if there was an error loading, the wkey was not
	* consumed, and needs to be freed.
	*/
	dsl_crypto_params_free(dcp, (err != 0));
	fnvlist_free(crypto_args);

	if (err == EINVAL) {
	/*
	* We couldn't load a key for this dataset so try
	* the parent. This loop will eventually hit the
	* encryption root since ztest only makes clones
	* as children of their origin datasets.
	*/
	cp = strrchr(ddname, '/');
	if (cp == NULL)
	return (err);

	*cp = '\0';
	err = EACCES;
	continue;
	} else if (err != 0) {
	break;
	}

	err = dmu_objset_own(name, type, readonly, decrypt, tag, osp);
	break;
	}

	return (err);
	}

	static void
	ztest_rll_init(rll_t *rll)
	{
	rll->rll_writer = NULL;
	rll->rll_readers = 0;
	mutex_init(&rll->rll_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&rll->rll_cv, NULL, CV_DEFAULT, NULL);
	}

	static void
	ztest_rll_destroy(rll_t *rll)
	{
	ASSERT3P(rll->rll_writer, ==, NULL);
	ASSERT0(rll->rll_readers);
	mutex_destroy(&rll->rll_lock);
	cv_destroy(&rll->rll_cv);
	}

	static void
	ztest_rll_lock(rll_t *rll, rl_type_t type)
	{
	mutex_enter(&rll->rll_lock);

	if (type == ZTRL_READER) {
	while (rll->rll_writer != NULL)
	(void) cv_wait(&rll->rll_cv, &rll->rll_lock);
	rll->rll_readers++;
	} else {
	while (rll->rll_writer != NULL \|\| rll->rll_readers)
	(void) cv_wait(&rll->rll_cv, &rll->rll_lock);
	rll->rll_writer = curthread;
	}

	mutex_exit(&rll->rll_lock);
	}

	static void
	ztest_rll_unlock(rll_t *rll)
	{
	mutex_enter(&rll->rll_lock);

	if (rll->rll_writer) {
	ASSERT0(rll->rll_readers);
	rll->rll_writer = NULL;
	} else {
	ASSERT3S(rll->rll_readers, >, 0);
	ASSERT3P(rll->rll_writer, ==, NULL);
	rll->rll_readers--;
	}

	if (rll->rll_writer == NULL && rll->rll_readers == 0)
	cv_broadcast(&rll->rll_cv);

	mutex_exit(&rll->rll_lock);
	}

	static void
	ztest_object_lock(ztest_ds_t *zd, uint64_t object, rl_type_t type)
	{
	rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)];

	ztest_rll_lock(rll, type);
	}

	static void
	ztest_object_unlock(ztest_ds_t *zd, uint64_t object)
	{
	rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)];

	ztest_rll_unlock(rll);
	}

	static rl_t *
	ztest_range_lock(ztest_ds_t *zd, uint64_t object, uint64_t offset,
	uint64_t size, rl_type_t type)
	{
	uint64_t hash = object ^ (offset % (ZTEST_RANGE_LOCKS + 1));
	rll_t *rll = &zd->zd_range_lock[hash & (ZTEST_RANGE_LOCKS - 1)];
	rl_t *rl;

	rl = umem_alloc(sizeof (*rl), UMEM_NOFAIL);
	rl->rl_object = object;
	rl->rl_offset = offset;
	rl->rl_size = size;
	rl->rl_lock = rll;

	ztest_rll_lock(rll, type);

	return (rl);
	}

	static void
	ztest_range_unlock(rl_t *rl)
	{
	rll_t *rll = rl->rl_lock;

	ztest_rll_unlock(rll);

	umem_free(rl, sizeof (*rl));
	}

	static void
	ztest_zd_init(ztest_ds_t zd, ztest_shared_ds_t szd, objset_t *os)
	{
	zd->zd_os = os;
	zd->zd_zilog = dmu_objset_zil(os);
	zd->zd_shared = szd;
	dmu_objset_name(os, zd->zd_name);
	int l;

	if (zd->zd_shared != NULL)
	zd->zd_shared->zd_seq = 0;

	VERIFY0(pthread_rwlock_init(&zd->zd_zilog_lock, NULL));
	mutex_init(&zd->zd_dirobj_lock, NULL, MUTEX_DEFAULT, NULL);

	for (l = 0; l < ZTEST_OBJECT_LOCKS; l++)
	ztest_rll_init(&zd->zd_object_lock[l]);

	for (l = 0; l < ZTEST_RANGE_LOCKS; l++)
	ztest_rll_init(&zd->zd_range_lock[l]);
	}

	static void
	ztest_zd_fini(ztest_ds_t *zd)
	{
	int l;

	mutex_destroy(&zd->zd_dirobj_lock);
	(void) pthread_rwlock_destroy(&zd->zd_zilog_lock);

	for (l = 0; l < ZTEST_OBJECT_LOCKS; l++)
	ztest_rll_destroy(&zd->zd_object_lock[l]);

	for (l = 0; l < ZTEST_RANGE_LOCKS; l++)
	ztest_rll_destroy(&zd->zd_range_lock[l]);
	}

	#define TXG_MIGHTWAIT (ztest_random(10) == 0 ? TXG_NOWAIT : TXG_WAIT)

	static uint64_t
	ztest_tx_assign(dmu_tx_t tx, uint64_t txg_how, const char tag)
	{
	uint64_t txg;
	int error;

	/*
	* Attempt to assign tx to some transaction group.
	*/
	error = dmu_tx_assign(tx, txg_how);
	if (error) {
	if (error == ERESTART) {
	ASSERT3U(txg_how, ==, TXG_NOWAIT);
	dmu_tx_wait(tx);
	} else {
	ASSERT3U(error, ==, ENOSPC);
	ztest_record_enospc(tag);
	}
	dmu_tx_abort(tx);
	return (0);
	}
	txg = dmu_tx_get_txg(tx);
	ASSERT3U(txg, !=, 0);
	return (txg);
	}

	static void
	ztest_bt_generate(ztest_block_tag_t bt, objset_t os, uint64_t object,
	uint64_t dnodesize, uint64_t offset, uint64_t gen, uint64_t txg,
	uint64_t crtxg)
	{
	bt->bt_magic = BT_MAGIC;
	bt->bt_objset = dmu_objset_id(os);
	bt->bt_object = object;
	bt->bt_dnodesize = dnodesize;
	bt->bt_offset = offset;
	bt->bt_gen = gen;
	bt->bt_txg = txg;
	bt->bt_crtxg = crtxg;
	}

	static void
	ztest_bt_verify(ztest_block_tag_t bt, objset_t os, uint64_t object,
	uint64_t dnodesize, uint64_t offset, uint64_t gen, uint64_t txg,
	uint64_t crtxg)
	{
	ASSERT3U(bt->bt_magic, ==, BT_MAGIC);
	ASSERT3U(bt->bt_objset, ==, dmu_objset_id(os));
	ASSERT3U(bt->bt_object, ==, object);
	ASSERT3U(bt->bt_dnodesize, ==, dnodesize);
	ASSERT3U(bt->bt_offset, ==, offset);
	ASSERT3U(bt->bt_gen, <=, gen);
	ASSERT3U(bt->bt_txg, <=, txg);
	ASSERT3U(bt->bt_crtxg, ==, crtxg);
	}

	static ztest_block_tag_t *
	ztest_bt_bonus(dmu_buf_t *db)
	{
	dmu_object_info_t doi;
	ztest_block_tag_t *bt;

	dmu_object_info_from_db(db, &doi);
	ASSERT3U(doi.doi_bonus_size, <=, db->db_size);
	ASSERT3U(doi.doi_bonus_size, >=, sizeof (*bt));
	bt = (void )((char )db->db_data + doi.doi_bonus_size - sizeof (*bt));

	return (bt);
	}

	/*
	* Generate a token to fill up unused bonus buffer space. Try to make
	* it unique to the object, generation, and offset to verify that data
	* is not getting overwritten by data from other dnodes.
	*/
	#define ZTEST_BONUS_FILL_TOKEN(obj, ds, gen, offset) \
	(((ds) << 48) \| ((gen) << 32) \| ((obj) << 8) \| (offset))

	/*
	* Fill up the unused bonus buffer region before the block tag with a
	* verifiable pattern. Filling the whole bonus area with non-zero data
	* helps ensure that all dnode traversal code properly skips the
	* interior regions of large dnodes.
	*/
	static void
	ztest_fill_unused_bonus(dmu_buf_t db, void end, uint64_t obj,
	objset_t *os, uint64_t gen)
	{
	uint64_t *bonusp;

	ASSERT(IS_P2ALIGNED((char )end - (char )db->db_data, 8));

	for (bonusp = db->db_data; bonusp < (uint64_t *)end; bonusp++) {
	uint64_t token = ZTEST_BONUS_FILL_TOKEN(obj, dmu_objset_id(os),
	gen, bonusp - (uint64_t *)db->db_data);
	*bonusp = token;
	}
	}

	/*
	* Verify that the unused area of a bonus buffer is filled with the
	* expected tokens.
	*/
	static void
	ztest_verify_unused_bonus(dmu_buf_t db, void end, uint64_t obj,
	objset_t *os, uint64_t gen)
	{
	uint64_t *bonusp;

	for (bonusp = db->db_data; bonusp < (uint64_t *)end; bonusp++) {
	uint64_t token = ZTEST_BONUS_FILL_TOKEN(obj, dmu_objset_id(os),
	gen, bonusp - (uint64_t *)db->db_data);
	VERIFY3U(*bonusp, ==, token);
	}
	}

	/*
	* ZIL logging ops
	*/

	#define lrz_type lr_mode
	#define lrz_blocksize lr_uid
	#define lrz_ibshift lr_gid
	#define lrz_bonustype lr_rdev
	#define lrz_dnodesize lr_crtime[1]

	static void
	ztest_log_create(ztest_ds_t zd, dmu_tx_t tx, lr_create_t *lr)
	{
	char name = (void )(lr + 1); /* name follows lr */
	size_t namesize = strlen(name) + 1;
	itx_t *itx;

	if (zil_replaying(zd->zd_zilog, tx))
	return;

	itx = zil_itx_create(TX_CREATE, sizeof (*lr) + namesize);
	memcpy(&itx->itx_lr + 1, &lr->lr_common + 1,
	sizeof (*lr) + namesize - sizeof (lr_t));

	zil_itx_assign(zd->zd_zilog, itx, tx);
	}

	static void
	ztest_log_remove(ztest_ds_t zd, dmu_tx_t tx, lr_remove_t *lr, uint64_t object)
	{
	char name = (void )(lr + 1); /* name follows lr */
	size_t namesize = strlen(name) + 1;
	itx_t *itx;

	if (zil_replaying(zd->zd_zilog, tx))
	return;

	itx = zil_itx_create(TX_REMOVE, sizeof (*lr) + namesize);
	memcpy(&itx->itx_lr + 1, &lr->lr_common + 1,
	sizeof (*lr) + namesize - sizeof (lr_t));

	itx->itx_oid = object;
	zil_itx_assign(zd->zd_zilog, itx, tx);
	}

	static void
	ztest_log_write(ztest_ds_t zd, dmu_tx_t tx, lr_write_t *lr)
	{
	itx_t *itx;
	itx_wr_state_t write_state = ztest_random(WR_NUM_STATES);

	if (zil_replaying(zd->zd_zilog, tx))
	return;

	if (lr->lr_length > zil_max_log_data(zd->zd_zilog, sizeof (lr_write_t)))
	write_state = WR_INDIRECT;

	itx = zil_itx_create(TX_WRITE,
	sizeof (*lr) + (write_state == WR_COPIED ? lr->lr_length : 0));

	if (write_state == WR_COPIED &&
	dmu_read(zd->zd_os, lr->lr_foid, lr->lr_offset, lr->lr_length,
	((lr_write_t *)&itx->itx_lr) + 1, DMU_READ_NO_PREFETCH) != 0) {
	zil_itx_destroy(itx);
	itx = zil_itx_create(TX_WRITE, sizeof (*lr));
	write_state = WR_NEED_COPY;
	}
	itx->itx_private = zd;
	itx->itx_wr_state = write_state;
	itx->itx_sync = (ztest_random(8) == 0);

	memcpy(&itx->itx_lr + 1, &lr->lr_common + 1,
	sizeof (*lr) - sizeof (lr_t));

	zil_itx_assign(zd->zd_zilog, itx, tx);
	}

	static void
	ztest_log_truncate(ztest_ds_t zd, dmu_tx_t tx, lr_truncate_t *lr)
	{
	itx_t *itx;

	if (zil_replaying(zd->zd_zilog, tx))
	return;

	itx = zil_itx_create(TX_TRUNCATE, sizeof (*lr));
	memcpy(&itx->itx_lr + 1, &lr->lr_common + 1,
	sizeof (*lr) - sizeof (lr_t));

	itx->itx_sync = B_FALSE;
	zil_itx_assign(zd->zd_zilog, itx, tx);
	}

	static void
	ztest_log_setattr(ztest_ds_t zd, dmu_tx_t tx, lr_setattr_t *lr)
	{
	itx_t *itx;

	if (zil_replaying(zd->zd_zilog, tx))
	return;

	itx = zil_itx_create(TX_SETATTR, sizeof (*lr));
	memcpy(&itx->itx_lr + 1, &lr->lr_common + 1,
	sizeof (*lr) - sizeof (lr_t));

	itx->itx_sync = B_FALSE;
	zil_itx_assign(zd->zd_zilog, itx, tx);
	}

	/*
	* ZIL replay ops
	*/
	static int
	ztest_replay_create(void arg1, void arg2, boolean_t byteswap)
	{
	ztest_ds_t *zd = arg1;
	lr_create_t *lr = arg2;
	char name = (void )(lr + 1); /* name follows lr */
	objset_t *os = zd->zd_os;
	ztest_block_tag_t *bbt;
	dmu_buf_t *db;
	dmu_tx_t *tx;
	uint64_t txg;
	int error = 0;
	int bonuslen;

	if (byteswap)
	byteswap_uint64_array(lr, sizeof (*lr));

	ASSERT3U(lr->lr_doid, ==, ZTEST_DIROBJ);
	ASSERT3S(name[0], !=, '\0');

	tx = dmu_tx_create(os);

	dmu_tx_hold_zap(tx, lr->lr_doid, B_TRUE, name);

	if (lr->lrz_type == DMU_OT_ZAP_OTHER) {
	dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL);
	} else {
	dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
	}

	txg = ztest_tx_assign(tx, TXG_WAIT, FTAG);
	if (txg == 0)
	return (ENOSPC);

	ASSERT3U(dmu_objset_zil(os)->zl_replay, ==, !!lr->lr_foid);
	bonuslen = DN_BONUS_SIZE(lr->lrz_dnodesize);

	if (lr->lrz_type == DMU_OT_ZAP_OTHER) {
	if (lr->lr_foid == 0) {
	lr->lr_foid = zap_create_dnsize(os,
	lr->lrz_type, lr->lrz_bonustype,
	bonuslen, lr->lrz_dnodesize, tx);
	} else {
	error = zap_create_claim_dnsize(os, lr->lr_foid,
	lr->lrz_type, lr->lrz_bonustype,
	bonuslen, lr->lrz_dnodesize, tx);
	}
	} else {
	if (lr->lr_foid == 0) {
	lr->lr_foid = dmu_object_alloc_dnsize(os,
	lr->lrz_type, 0, lr->lrz_bonustype,
	bonuslen, lr->lrz_dnodesize, tx);
	} else {
	error = dmu_object_claim_dnsize(os, lr->lr_foid,
	lr->lrz_type, 0, lr->lrz_bonustype,
	bonuslen, lr->lrz_dnodesize, tx);
	}
	}

	if (error) {
	ASSERT3U(error, ==, EEXIST);
	ASSERT(zd->zd_zilog->zl_replay);
	dmu_tx_commit(tx);
	return (error);
	}

	ASSERT3U(lr->lr_foid, !=, 0);

	if (lr->lrz_type != DMU_OT_ZAP_OTHER)
	VERIFY0(dmu_object_set_blocksize(os, lr->lr_foid,
	lr->lrz_blocksize, lr->lrz_ibshift, tx));

	VERIFY0(dmu_bonus_hold(os, lr->lr_foid, FTAG, &db));
	bbt = ztest_bt_bonus(db);
	dmu_buf_will_dirty(db, tx);
	ztest_bt_generate(bbt, os, lr->lr_foid, lr->lrz_dnodesize, -1ULL,
	lr->lr_gen, txg, txg);
	ztest_fill_unused_bonus(db, bbt, lr->lr_foid, os, lr->lr_gen);
	dmu_buf_rele(db, FTAG);

	VERIFY0(zap_add(os, lr->lr_doid, name, sizeof (uint64_t), 1,
	&lr->lr_foid, tx));

	(void) ztest_log_create(zd, tx, lr);

	dmu_tx_commit(tx);

	return (0);
	}

	static int
	ztest_replay_remove(void arg1, void arg2, boolean_t byteswap)
	{
	ztest_ds_t *zd = arg1;
	lr_remove_t *lr = arg2;
	char name = (void )(lr + 1); /* name follows lr */
	objset_t *os = zd->zd_os;
	dmu_object_info_t doi;
	dmu_tx_t *tx;
	uint64_t object, txg;

	if (byteswap)
	byteswap_uint64_array(lr, sizeof (*lr));

	ASSERT3U(lr->lr_doid, ==, ZTEST_DIROBJ);
	ASSERT3S(name[0], !=, '\0');

	VERIFY0(
	zap_lookup(os, lr->lr_doid, name, sizeof (object), 1, &object));
	ASSERT3U(object, !=, 0);

	ztest_object_lock(zd, object, ZTRL_WRITER);

	VERIFY0(dmu_object_info(os, object, &doi));

	tx = dmu_tx_create(os);

	dmu_tx_hold_zap(tx, lr->lr_doid, B_FALSE, name);
	dmu_tx_hold_free(tx, object, 0, DMU_OBJECT_END);

	txg = ztest_tx_assign(tx, TXG_WAIT, FTAG);
	if (txg == 0) {
	ztest_object_unlock(zd, object);
	return (ENOSPC);
	}

	if (doi.doi_type == DMU_OT_ZAP_OTHER) {
	VERIFY0(zap_destroy(os, object, tx));
	} else {
	VERIFY0(dmu_object_free(os, object, tx));
	}

	VERIFY0(zap_remove(os, lr->lr_doid, name, tx));

	(void) ztest_log_remove(zd, tx, lr, object);

	dmu_tx_commit(tx);

	ztest_object_unlock(zd, object);

	return (0);
	}

	static int
	ztest_replay_write(void arg1, void arg2, boolean_t byteswap)
	{
	ztest_ds_t *zd = arg1;
	lr_write_t *lr = arg2;
	objset_t *os = zd->zd_os;
	void data = lr + 1; / data follows lr */
	uint64_t offset, length;
	ztest_block_tag_t *bt = data;
	ztest_block_tag_t *bbt;
	uint64_t gen, txg, lrtxg, crtxg;
	dmu_object_info_t doi;
	dmu_tx_t *tx;
	dmu_buf_t *db;
	arc_buf_t *abuf = NULL;
	rl_t *rl;

	if (byteswap)
	byteswap_uint64_array(lr, sizeof (*lr));

	offset = lr->lr_offset;
	length = lr->lr_length;

	/* If it's a dmu_sync() block, write the whole block */
	if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) {
	uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr);
	if (length < blocksize) {
	offset -= offset % blocksize;
	length = blocksize;
	}
	}

	if (bt->bt_magic == BSWAP_64(BT_MAGIC))
	byteswap_uint64_array(bt, sizeof (*bt));

	if (bt->bt_magic != BT_MAGIC)
	bt = NULL;

	ztest_object_lock(zd, lr->lr_foid, ZTRL_READER);
	rl = ztest_range_lock(zd, lr->lr_foid, offset, length, ZTRL_WRITER);

	VERIFY0(dmu_bonus_hold(os, lr->lr_foid, FTAG, &db));

	dmu_object_info_from_db(db, &doi);

	bbt = ztest_bt_bonus(db);
	ASSERT3U(bbt->bt_magic, ==, BT_MAGIC);
	gen = bbt->bt_gen;
	crtxg = bbt->bt_crtxg;
	lrtxg = lr->lr_common.lrc_txg;

	tx = dmu_tx_create(os);

	dmu_tx_hold_write(tx, lr->lr_foid, offset, length);

	if (ztest_random(8) == 0 && length == doi.doi_data_block_size &&
	P2PHASE(offset, length) == 0)
	abuf = dmu_request_arcbuf(db, length);

	txg = ztest_tx_assign(tx, TXG_WAIT, FTAG);
	if (txg == 0) {
	if (abuf != NULL)
	dmu_return_arcbuf(abuf);
	dmu_buf_rele(db, FTAG);
	ztest_range_unlock(rl);
	ztest_object_unlock(zd, lr->lr_foid);
	return (ENOSPC);
	}

	if (bt != NULL) {
	/*
	* Usually, verify the old data before writing new data --
	* but not always, because we also want to verify correct
	* behavior when the data was not recently read into cache.
	*/
	ASSERT(doi.doi_data_block_size);
	ASSERT0(offset % doi.doi_data_block_size);
	if (ztest_random(4) != 0) {
	int prefetch = ztest_random(2) ?
	DMU_READ_PREFETCH : DMU_READ_NO_PREFETCH;
	ztest_block_tag_t rbt;

	VERIFY(dmu_read(os, lr->lr_foid, offset,
	sizeof (rbt), &rbt, prefetch) == 0);
	if (rbt.bt_magic == BT_MAGIC) {
	ztest_bt_verify(&rbt, os, lr->lr_foid, 0,
	offset, gen, txg, crtxg);
	}
	}

	/*
	* Writes can appear to be newer than the bonus buffer because
	* the ztest_get_data() callback does a dmu_read() of the
	* open-context data, which may be different than the data
	* as it was when the write was generated.
	*/
	if (zd->zd_zilog->zl_replay) {
	ztest_bt_verify(bt, os, lr->lr_foid, 0, offset,
	MAX(gen, bt->bt_gen), MAX(txg, lrtxg),
	bt->bt_crtxg);
	}

	/*
	* Set the bt's gen/txg to the bonus buffer's gen/txg
	* so that all of the usual ASSERTs will work.
	*/
	ztest_bt_generate(bt, os, lr->lr_foid, 0, offset, gen, txg,
	crtxg);
	}

	if (abuf == NULL) {
	dmu_write(os, lr->lr_foid, offset, length, data, tx);
	} else {
	memcpy(abuf->b_data, data, length);
	VERIFY0(dmu_assign_arcbuf_by_dbuf(db, offset, abuf, tx));
	}

	(void) ztest_log_write(zd, tx, lr);

	dmu_buf_rele(db, FTAG);

	dmu_tx_commit(tx);

	ztest_range_unlock(rl);
	ztest_object_unlock(zd, lr->lr_foid);

	return (0);
	}

	static int
	ztest_replay_truncate(void arg1, void arg2, boolean_t byteswap)
	{
	ztest_ds_t *zd = arg1;
	lr_truncate_t *lr = arg2;
	objset_t *os = zd->zd_os;
	dmu_tx_t *tx;
	uint64_t txg;
	rl_t *rl;

	if (byteswap)
	byteswap_uint64_array(lr, sizeof (*lr));

	ztest_object_lock(zd, lr->lr_foid, ZTRL_READER);
	rl = ztest_range_lock(zd, lr->lr_foid, lr->lr_offset, lr->lr_length,
	ZTRL_WRITER);

	tx = dmu_tx_create(os);

	dmu_tx_hold_free(tx, lr->lr_foid, lr->lr_offset, lr->lr_length);

	txg = ztest_tx_assign(tx, TXG_WAIT, FTAG);
	if (txg == 0) {
	ztest_range_unlock(rl);
	ztest_object_unlock(zd, lr->lr_foid);
	return (ENOSPC);
	}

	VERIFY0(dmu_free_range(os, lr->lr_foid, lr->lr_offset,
	lr->lr_length, tx));

	(void) ztest_log_truncate(zd, tx, lr);

	dmu_tx_commit(tx);

	ztest_range_unlock(rl);
	ztest_object_unlock(zd, lr->lr_foid);

	return (0);
	}

	static int
	ztest_replay_setattr(void arg1, void arg2, boolean_t byteswap)
	{
	ztest_ds_t *zd = arg1;
	lr_setattr_t *lr = arg2;
	objset_t *os = zd->zd_os;
	dmu_tx_t *tx;
	dmu_buf_t *db;
	ztest_block_tag_t *bbt;
	uint64_t txg, lrtxg, crtxg, dnodesize;

	if (byteswap)
	byteswap_uint64_array(lr, sizeof (*lr));

	ztest_object_lock(zd, lr->lr_foid, ZTRL_WRITER);

	VERIFY0(dmu_bonus_hold(os, lr->lr_foid, FTAG, &db));

	tx = dmu_tx_create(os);
	dmu_tx_hold_bonus(tx, lr->lr_foid);

	txg = ztest_tx_assign(tx, TXG_WAIT, FTAG);
	if (txg == 0) {
	dmu_buf_rele(db, FTAG);
	ztest_object_unlock(zd, lr->lr_foid);
	return (ENOSPC);
	}

	bbt = ztest_bt_bonus(db);
	ASSERT3U(bbt->bt_magic, ==, BT_MAGIC);
	crtxg = bbt->bt_crtxg;
	lrtxg = lr->lr_common.lrc_txg;
	dnodesize = bbt->bt_dnodesize;

	if (zd->zd_zilog->zl_replay) {
	ASSERT3U(lr->lr_size, !=, 0);
	ASSERT3U(lr->lr_mode, !=, 0);
	ASSERT3U(lrtxg, !=, 0);
	} else {
	/*
	* Randomly change the size and increment the generation.
	*/
	lr->lr_size = (ztest_random(db->db_size / sizeof (bbt)) + 1)
	sizeof (*bbt);
	lr->lr_mode = bbt->bt_gen + 1;
	ASSERT0(lrtxg);
	}

	/*
	* Verify that the current bonus buffer is not newer than our txg.
	*/
	ztest_bt_verify(bbt, os, lr->lr_foid, dnodesize, -1ULL, lr->lr_mode,
	MAX(txg, lrtxg), crtxg);

	dmu_buf_will_dirty(db, tx);

	ASSERT3U(lr->lr_size, >=, sizeof (*bbt));
	ASSERT3U(lr->lr_size, <=, db->db_size);
	VERIFY0(dmu_set_bonus(db, lr->lr_size, tx));
	bbt = ztest_bt_bonus(db);

	ztest_bt_generate(bbt, os, lr->lr_foid, dnodesize, -1ULL, lr->lr_mode,
	txg, crtxg);
	ztest_fill_unused_bonus(db, bbt, lr->lr_foid, os, bbt->bt_gen);
	dmu_buf_rele(db, FTAG);

	(void) ztest_log_setattr(zd, tx, lr);

	dmu_tx_commit(tx);

	ztest_object_unlock(zd, lr->lr_foid);

	return (0);
	}

	static zil_replay_func_t *ztest_replay_vector[TX_MAX_TYPE] = {
	NULL, /* 0 no such transaction type */
	ztest_replay_create, /* TX_CREATE */
	NULL, /* TX_MKDIR */
	NULL, /* TX_MKXATTR */
	NULL, /* TX_SYMLINK */
	ztest_replay_remove, /* TX_REMOVE */
	NULL, /* TX_RMDIR */
	NULL, /* TX_LINK */
	NULL, /* TX_RENAME */
	ztest_replay_write, /* TX_WRITE */
	ztest_replay_truncate, /* TX_TRUNCATE */
	ztest_replay_setattr, /* TX_SETATTR */
	NULL, /* TX_ACL */
	NULL, /* TX_CREATE_ACL */
	NULL, /* TX_CREATE_ATTR */
	NULL, /* TX_CREATE_ACL_ATTR */
	NULL, /* TX_MKDIR_ACL */
	NULL, /* TX_MKDIR_ATTR */
	NULL, /* TX_MKDIR_ACL_ATTR */
	NULL, /* TX_WRITE2 */
	NULL, /* TX_SETSAXATTR */
	NULL, /* TX_RENAME_EXCHANGE */
	NULL, /* TX_RENAME_WHITEOUT */
	};

	/*
	* ZIL get_data callbacks
	*/

	static void
	ztest_get_done(zgd_t *zgd, int error)
	{
	(void) error;
	ztest_ds_t *zd = zgd->zgd_private;
	uint64_t object = ((rl_t *)zgd->zgd_lr)->rl_object;

	if (zgd->zgd_db)
	dmu_buf_rele(zgd->zgd_db, zgd);

	ztest_range_unlock((rl_t *)zgd->zgd_lr);
	ztest_object_unlock(zd, object);

	umem_free(zgd, sizeof (*zgd));
	}

	static int
	ztest_get_data(void arg, uint64_t arg2, lr_write_t lr, char *buf,
	struct lwb lwb, zio_t zio)
	{
	(void) arg2;
	ztest_ds_t *zd = arg;
	objset_t *os = zd->zd_os;
	uint64_t object = lr->lr_foid;
	uint64_t offset = lr->lr_offset;
	uint64_t size = lr->lr_length;
	uint64_t txg = lr->lr_common.lrc_txg;
	uint64_t crtxg;
	dmu_object_info_t doi;
	dmu_buf_t *db;
	zgd_t *zgd;
	int error;

	ASSERT3P(lwb, !=, NULL);
	ASSERT3U(size, !=, 0);

	ztest_object_lock(zd, object, ZTRL_READER);
	error = dmu_bonus_hold(os, object, FTAG, &db);
	if (error) {
	ztest_object_unlock(zd, object);
	return (error);
	}

	crtxg = ztest_bt_bonus(db)->bt_crtxg;

	if (crtxg == 0 \|\| crtxg > txg) {
	dmu_buf_rele(db, FTAG);
	ztest_object_unlock(zd, object);
	return (ENOENT);
	}

	dmu_object_info_from_db(db, &doi);
	dmu_buf_rele(db, FTAG);
	db = NULL;

	zgd = umem_zalloc(sizeof (*zgd), UMEM_NOFAIL);
	zgd->zgd_lwb = lwb;
	zgd->zgd_private = zd;

	if (buf != NULL) { /* immediate write */
	zgd->zgd_lr = (struct zfs_locked_range *)ztest_range_lock(zd,
	object, offset, size, ZTRL_READER);

	error = dmu_read(os, object, offset, size, buf,
	DMU_READ_NO_PREFETCH);
	ASSERT0(error);
	} else {
	ASSERT3P(zio, !=, NULL);
	size = doi.doi_data_block_size;
	if (ISP2(size)) {
	- offset = P2ALIGN(offset, size);
	+ offset = P2ALIGN_TYPED(offset, size, uint64_t);
	} else {
	ASSERT3U(offset, <, size);
	offset = 0;
	}

	zgd->zgd_lr = (struct zfs_locked_range *)ztest_range_lock(zd,
	object, offset, size, ZTRL_READER);

	error = dmu_buf_hold_noread(os, object, offset, zgd, &db);

	if (error == 0) {
	blkptr_t *bp = &lr->lr_blkptr;

	zgd->zgd_db = db;
	zgd->zgd_bp = bp;

	ASSERT3U(db->db_offset, ==, offset);
	ASSERT3U(db->db_size, ==, size);

	error = dmu_sync(zio, lr->lr_common.lrc_txg,
	ztest_get_done, zgd);

	if (error == 0)
	return (0);
	}
	}

	ztest_get_done(zgd, error);

	return (error);
	}

	static void *
	ztest_lr_alloc(size_t lrsize, char *name)
	{
	char *lr;
	size_t namesize = name ? strlen(name) + 1 : 0;

	lr = umem_zalloc(lrsize + namesize, UMEM_NOFAIL);

	if (name)
	memcpy(lr + lrsize, name, namesize);

	return (lr);
	}

	static void
	ztest_lr_free(void lr, size_t lrsize, char name)
	{
	size_t namesize = name ? strlen(name) + 1 : 0;

	umem_free(lr, lrsize + namesize);
	}

	/*
	* Lookup a bunch of objects. Returns the number of objects not found.
	*/
	static int
	ztest_lookup(ztest_ds_t zd, ztest_od_t od, int count)
	{
	int missing = 0;
	int error;
	int i;

	ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock));

	for (i = 0; i < count; i++, od++) {
	od->od_object = 0;
	error = zap_lookup(zd->zd_os, od->od_dir, od->od_name,
	sizeof (uint64_t), 1, &od->od_object);
	if (error) {
	ASSERT3S(error, ==, ENOENT);
	ASSERT0(od->od_object);
	missing++;
	} else {
	dmu_buf_t *db;
	ztest_block_tag_t *bbt;
	dmu_object_info_t doi;

	ASSERT3U(od->od_object, !=, 0);
	ASSERT0(missing); /* there should be no gaps */

	ztest_object_lock(zd, od->od_object, ZTRL_READER);
	VERIFY0(dmu_bonus_hold(zd->zd_os, od->od_object,
	FTAG, &db));
	dmu_object_info_from_db(db, &doi);
	bbt = ztest_bt_bonus(db);
	ASSERT3U(bbt->bt_magic, ==, BT_MAGIC);
	od->od_type = doi.doi_type;
	od->od_blocksize = doi.doi_data_block_size;
	od->od_gen = bbt->bt_gen;
	dmu_buf_rele(db, FTAG);
	ztest_object_unlock(zd, od->od_object);
	}
	}

	return (missing);
	}

	static int
	ztest_create(ztest_ds_t zd, ztest_od_t od, int count)
	{
	int missing = 0;
	int i;

	ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock));

	for (i = 0; i < count; i++, od++) {
	if (missing) {
	od->od_object = 0;
	missing++;
	continue;
	}

	lr_create_t lr = ztest_lr_alloc(sizeof (lr), od->od_name);

	lr->lr_doid = od->od_dir;
	lr->lr_foid = 0; /* 0 to allocate, > 0 to claim */
	lr->lrz_type = od->od_crtype;
	lr->lrz_blocksize = od->od_crblocksize;
	lr->lrz_ibshift = ztest_random_ibshift();
	lr->lrz_bonustype = DMU_OT_UINT64_OTHER;
	lr->lrz_dnodesize = od->od_crdnodesize;
	lr->lr_gen = od->od_crgen;
	lr->lr_crtime[0] = time(NULL);

	if (ztest_replay_create(zd, lr, B_FALSE) != 0) {
	ASSERT0(missing);
	od->od_object = 0;
	missing++;
	} else {
	od->od_object = lr->lr_foid;
	od->od_type = od->od_crtype;
	od->od_blocksize = od->od_crblocksize;
	od->od_gen = od->od_crgen;
	ASSERT3U(od->od_object, !=, 0);
	}

	ztest_lr_free(lr, sizeof (*lr), od->od_name);
	}

	return (missing);
	}

	static int
	ztest_remove(ztest_ds_t zd, ztest_od_t od, int count)
	{
	int missing = 0;
	int error;
	int i;

	ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock));

	od += count - 1;

	for (i = count - 1; i >= 0; i--, od--) {
	if (missing) {
	missing++;
	continue;
	}

	/*
	* No object was found.
	*/
	if (od->od_object == 0)
	continue;

	lr_remove_t lr = ztest_lr_alloc(sizeof (lr), od->od_name);

	lr->lr_doid = od->od_dir;

	if ((error = ztest_replay_remove(zd, lr, B_FALSE)) != 0) {
	ASSERT3U(error, ==, ENOSPC);
	missing++;
	} else {
	od->od_object = 0;
	}
	ztest_lr_free(lr, sizeof (*lr), od->od_name);
	}

	return (missing);
	}

	static int
	ztest_write(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size,
	const void *data)
	{
	lr_write_t *lr;
	int error;

	lr = ztest_lr_alloc(sizeof (*lr) + size, NULL);

	lr->lr_foid = object;
	lr->lr_offset = offset;
	lr->lr_length = size;
	lr->lr_blkoff = 0;
	BP_ZERO(&lr->lr_blkptr);

	memcpy(lr + 1, data, size);

	error = ztest_replay_write(zd, lr, B_FALSE);

	ztest_lr_free(lr, sizeof (*lr) + size, NULL);

	return (error);
	}

	static int
	ztest_truncate(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size)
	{
	lr_truncate_t *lr;
	int error;

	lr = ztest_lr_alloc(sizeof (*lr), NULL);

	lr->lr_foid = object;
	lr->lr_offset = offset;
	lr->lr_length = size;

	error = ztest_replay_truncate(zd, lr, B_FALSE);

	ztest_lr_free(lr, sizeof (*lr), NULL);

	return (error);
	}

	static int
	ztest_setattr(ztest_ds_t *zd, uint64_t object)
	{
	lr_setattr_t *lr;
	int error;

	lr = ztest_lr_alloc(sizeof (*lr), NULL);

	lr->lr_foid = object;
	lr->lr_size = 0;
	lr->lr_mode = 0;

	error = ztest_replay_setattr(zd, lr, B_FALSE);

	ztest_lr_free(lr, sizeof (*lr), NULL);

	return (error);
	}

	static void
	ztest_prealloc(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size)
	{
	objset_t *os = zd->zd_os;
	dmu_tx_t *tx;
	uint64_t txg;
	rl_t *rl;

	txg_wait_synced(dmu_objset_pool(os), 0);

	ztest_object_lock(zd, object, ZTRL_READER);
	rl = ztest_range_lock(zd, object, offset, size, ZTRL_WRITER);

	tx = dmu_tx_create(os);

	dmu_tx_hold_write(tx, object, offset, size);

	txg = ztest_tx_assign(tx, TXG_WAIT, FTAG);

	if (txg != 0) {
	dmu_prealloc(os, object, offset, size, tx);
	dmu_tx_commit(tx);
	txg_wait_synced(dmu_objset_pool(os), txg);
	} else {
	(void) dmu_free_long_range(os, object, offset, size);
	}

	ztest_range_unlock(rl);
	ztest_object_unlock(zd, object);
	}

	static void
	ztest_io(ztest_ds_t *zd, uint64_t object, uint64_t offset)
	{
	int err;
	ztest_block_tag_t wbt;
	dmu_object_info_t doi;
	enum ztest_io_type io_type;
	uint64_t blocksize;
	void *data;

	VERIFY0(dmu_object_info(zd->zd_os, object, &doi));
	blocksize = doi.doi_data_block_size;
	data = umem_alloc(blocksize, UMEM_NOFAIL);

	/*
	* Pick an i/o type at random, biased toward writing block tags.
	*/
	io_type = ztest_random(ZTEST_IO_TYPES);
	if (ztest_random(2) == 0)
	io_type = ZTEST_IO_WRITE_TAG;

	(void) pthread_rwlock_rdlock(&zd->zd_zilog_lock);

	switch (io_type) {

	case ZTEST_IO_WRITE_TAG:
	ztest_bt_generate(&wbt, zd->zd_os, object, doi.doi_dnodesize,
	offset, 0, 0, 0);
	(void) ztest_write(zd, object, offset, sizeof (wbt), &wbt);
	break;

	case ZTEST_IO_WRITE_PATTERN:
	(void) memset(data, 'a' + (object + offset) % 5, blocksize);
	if (ztest_random(2) == 0) {
	/*
	* Induce fletcher2 collisions to ensure that
	* zio_ddt_collision() detects and resolves them
	* when using fletcher2-verify for deduplication.
	*/
	((uint64_t *)data)[0] ^= 1ULL << 63;
	((uint64_t *)data)[4] ^= 1ULL << 63;
	}
	(void) ztest_write(zd, object, offset, blocksize, data);
	break;

	case ZTEST_IO_WRITE_ZEROES:
	memset(data, 0, blocksize);
	(void) ztest_write(zd, object, offset, blocksize, data);
	break;

	case ZTEST_IO_TRUNCATE:
	(void) ztest_truncate(zd, object, offset, blocksize);
	break;

	case ZTEST_IO_SETATTR:
	(void) ztest_setattr(zd, object);
	break;
	default:
	break;

	case ZTEST_IO_REWRITE:
	(void) pthread_rwlock_rdlock(&ztest_name_lock);
	err = ztest_dsl_prop_set_uint64(zd->zd_name,
	ZFS_PROP_CHECKSUM, spa_dedup_checksum(ztest_spa),
	B_FALSE);
	ASSERT(err == 0 \|\| err == ENOSPC);
	err = ztest_dsl_prop_set_uint64(zd->zd_name,
	ZFS_PROP_COMPRESSION,
	ztest_random_dsl_prop(ZFS_PROP_COMPRESSION),
	B_FALSE);
	ASSERT(err == 0 \|\| err == ENOSPC);
	(void) pthread_rwlock_unlock(&ztest_name_lock);

	VERIFY0(dmu_read(zd->zd_os, object, offset, blocksize, data,
	DMU_READ_NO_PREFETCH));

	(void) ztest_write(zd, object, offset, blocksize, data);
	break;
	}

	(void) pthread_rwlock_unlock(&zd->zd_zilog_lock);

	umem_free(data, blocksize);
	}

	/*
	* Initialize an object description template.
	*/
	static void
	ztest_od_init(ztest_od_t od, uint64_t id, const char tag, uint64_t index,
	dmu_object_type_t type, uint64_t blocksize, uint64_t dnodesize,
	uint64_t gen)
	{
	od->od_dir = ZTEST_DIROBJ;
	od->od_object = 0;

	od->od_crtype = type;
	od->od_crblocksize = blocksize ? blocksize : ztest_random_blocksize();
	od->od_crdnodesize = dnodesize ? dnodesize : ztest_random_dnodesize();
	od->od_crgen = gen;

	od->od_type = DMU_OT_NONE;
	od->od_blocksize = 0;
	od->od_gen = 0;

	(void) snprintf(od->od_name, sizeof (od->od_name),
	"%s(%"PRId64")[%"PRIu64"]",
	tag, id, index);
	}

	/*
	* Lookup or create the objects for a test using the od template.
	* If the objects do not all exist, or if 'remove' is specified,
	* remove any existing objects and create new ones. Otherwise,
	* use the existing objects.
	*/
	static int
	ztest_object_init(ztest_ds_t zd, ztest_od_t od, size_t size, boolean_t remove)
	{
	int count = size / sizeof (*od);
	int rv = 0;

	mutex_enter(&zd->zd_dirobj_lock);
	if ((ztest_lookup(zd, od, count) != 0 \|\| remove) &&
	(ztest_remove(zd, od, count) != 0 \|\|
	ztest_create(zd, od, count) != 0))
	rv = -1;
	zd->zd_od = od;
	mutex_exit(&zd->zd_dirobj_lock);

	return (rv);
	}

	void
	ztest_zil_commit(ztest_ds_t *zd, uint64_t id)
	{
	(void) id;
	zilog_t *zilog = zd->zd_zilog;

	(void) pthread_rwlock_rdlock(&zd->zd_zilog_lock);

	zil_commit(zilog, ztest_random(ZTEST_OBJECTS));

	/*
	* Remember the committed values in zd, which is in parent/child
	* shared memory. If we die, the next iteration of ztest_run()
	* will verify that the log really does contain this record.
	*/
	mutex_enter(&zilog->zl_lock);
	ASSERT3P(zd->zd_shared, !=, NULL);
	ASSERT3U(zd->zd_shared->zd_seq, <=, zilog->zl_commit_lr_seq);
	zd->zd_shared->zd_seq = zilog->zl_commit_lr_seq;
	mutex_exit(&zilog->zl_lock);

	(void) pthread_rwlock_unlock(&zd->zd_zilog_lock);
	}

	/*
	* This function is designed to simulate the operations that occur during a
	* mount/unmount operation. We hold the dataset across these operations in an
	* attempt to expose any implicit assumptions about ZIL management.
	*/
	void
	ztest_zil_remount(ztest_ds_t *zd, uint64_t id)
	{
	(void) id;
	objset_t *os = zd->zd_os;

	/*
	* We hold the ztest_vdev_lock so we don't cause problems with
	* other threads that wish to remove a log device, such as
	* ztest_device_removal().
	*/
	mutex_enter(&ztest_vdev_lock);

	/*
	* We grab the zd_dirobj_lock to ensure that no other thread is
	* updating the zil (i.e. adding in-memory log records) and the
	* zd_zilog_lock to block any I/O.
	*/
	mutex_enter(&zd->zd_dirobj_lock);
	(void) pthread_rwlock_wrlock(&zd->zd_zilog_lock);

	/* zfsvfs_teardown() */
	zil_close(zd->zd_zilog);

	/* zfsvfs_setup() */
	VERIFY3P(zil_open(os, ztest_get_data, NULL), ==, zd->zd_zilog);
	zil_replay(os, zd, ztest_replay_vector);

	(void) pthread_rwlock_unlock(&zd->zd_zilog_lock);
	mutex_exit(&zd->zd_dirobj_lock);
	mutex_exit(&ztest_vdev_lock);
	}

	/*
	* Verify that we can't destroy an active pool, create an existing pool,
	* or create a pool with a bad vdev spec.
	*/
	void
	ztest_spa_create_destroy(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	ztest_shared_opts_t *zo = &ztest_opts;
	spa_t *spa;
	nvlist_t *nvroot;

	if (zo->zo_mmp_test)
	return;

	/*
	* Attempt to create using a bad file.
	*/
	nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 0, 1);
	VERIFY3U(ENOENT, ==,
	spa_create("ztest_bad_file", nvroot, NULL, NULL, NULL));
	fnvlist_free(nvroot);

	/*
	* Attempt to create using a bad mirror.
	*/
	nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 2, 1);
	VERIFY3U(ENOENT, ==,
	spa_create("ztest_bad_mirror", nvroot, NULL, NULL, NULL));
	fnvlist_free(nvroot);

	/*
	* Attempt to create an existing pool. It shouldn't matter
	* what's in the nvroot; we should fail with EEXIST.
	*/
	(void) pthread_rwlock_rdlock(&ztest_name_lock);
	nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 0, 1);
	VERIFY3U(EEXIST, ==,
	spa_create(zo->zo_pool, nvroot, NULL, NULL, NULL));
	fnvlist_free(nvroot);

	/*
	* We open a reference to the spa and then we try to export it
	* expecting one of the following errors:
	*
	* EBUSY
	* Because of the reference we just opened.
	*
	* ZFS_ERR_EXPORT_IN_PROGRESS
	* For the case that there is another ztest thread doing
	* an export concurrently.
	*/
	VERIFY0(spa_open(zo->zo_pool, &spa, FTAG));
	int error = spa_destroy(zo->zo_pool);
	if (error != EBUSY && error != ZFS_ERR_EXPORT_IN_PROGRESS) {
	fatal(B_FALSE, "spa_destroy(%s) returned unexpected value %d",
	spa->spa_name, error);
	}
	spa_close(spa, FTAG);

	(void) pthread_rwlock_unlock(&ztest_name_lock);
	}

	/*
	* Start and then stop the MMP threads to ensure the startup and shutdown code
	* works properly. Actual protection and property-related code tested via ZTS.
	*/
	void
	ztest_mmp_enable_disable(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	ztest_shared_opts_t *zo = &ztest_opts;
	spa_t *spa = ztest_spa;

	if (zo->zo_mmp_test)
	return;

	/*
	* Since enabling MMP involves setting a property, it could not be done
	* while the pool is suspended.
	*/
	if (spa_suspended(spa))
	return;

	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	mutex_enter(&spa->spa_props_lock);

	zfs_multihost_fail_intervals = 0;

	if (!spa_multihost(spa)) {
	spa->spa_multihost = B_TRUE;
	mmp_thread_start(spa);
	}

	mutex_exit(&spa->spa_props_lock);
	spa_config_exit(spa, SCL_CONFIG, FTAG);

	txg_wait_synced(spa_get_dsl(spa), 0);
	mmp_signal_all_threads();
	txg_wait_synced(spa_get_dsl(spa), 0);

	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	mutex_enter(&spa->spa_props_lock);

	if (spa_multihost(spa)) {
	mmp_thread_stop(spa);
	spa->spa_multihost = B_FALSE;
	}

	mutex_exit(&spa->spa_props_lock);
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	}

	static int
	ztest_get_raidz_children(spa_t *spa)
	{
	(void) spa;
	vdev_t *raidvd;

	ASSERT(MUTEX_HELD(&ztest_vdev_lock));

	if (ztest_opts.zo_raid_do_expand) {
	raidvd = ztest_spa->spa_root_vdev->vdev_child[0];

	ASSERT(raidvd->vdev_ops == &vdev_raidz_ops);

	return (raidvd->vdev_children);
	}

	return (ztest_opts.zo_raid_children);
	}

	void
	ztest_spa_upgrade(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	spa_t *spa;
	uint64_t initial_version = SPA_VERSION_INITIAL;
	uint64_t raidz_children, version, newversion;
	nvlist_t nvroot, props;
	char *name;

	if (ztest_opts.zo_mmp_test)
	return;

	/* dRAID added after feature flags, skip upgrade test. */
	if (strcmp(ztest_opts.zo_raid_type, VDEV_TYPE_DRAID) == 0)
	return;

	mutex_enter(&ztest_vdev_lock);
	name = kmem_asprintf("%s_upgrade", ztest_opts.zo_pool);

	/*
	* Clean up from previous runs.
	*/
	(void) spa_destroy(name);

	raidz_children = ztest_get_raidz_children(ztest_spa);

	nvroot = make_vdev_root(NULL, NULL, name, ztest_opts.zo_vdev_size, 0,
	NULL, raidz_children, ztest_opts.zo_mirrors, 1);

	/*
	* If we're configuring a RAIDZ device then make sure that the
	* initial version is capable of supporting that feature.
	*/
	switch (ztest_opts.zo_raid_parity) {
	case 0:
	case 1:
	initial_version = SPA_VERSION_INITIAL;
	break;
	case 2:
	initial_version = SPA_VERSION_RAIDZ2;
	break;
	case 3:
	initial_version = SPA_VERSION_RAIDZ3;
	break;
	}

	/*
	* Create a pool with a spa version that can be upgraded. Pick
	* a value between initial_version and SPA_VERSION_BEFORE_FEATURES.
	*/
	do {
	version = ztest_random_spa_version(initial_version);
	} while (version > SPA_VERSION_BEFORE_FEATURES);

	props = fnvlist_alloc();
	fnvlist_add_uint64(props,
	zpool_prop_to_name(ZPOOL_PROP_VERSION), version);
	VERIFY0(spa_create(name, nvroot, props, NULL, NULL));
	fnvlist_free(nvroot);
	fnvlist_free(props);

	VERIFY0(spa_open(name, &spa, FTAG));
	VERIFY3U(spa_version(spa), ==, version);
	newversion = ztest_random_spa_version(version + 1);

	if (ztest_opts.zo_verbose >= 4) {
	(void) printf("upgrading spa version from "
	"%"PRIu64" to %"PRIu64"\n",
	version, newversion);
	}

	spa_upgrade(spa, newversion);
	VERIFY3U(spa_version(spa), >, version);
	VERIFY3U(spa_version(spa), ==, fnvlist_lookup_uint64(spa->spa_config,
	zpool_prop_to_name(ZPOOL_PROP_VERSION)));
	spa_close(spa, FTAG);

	kmem_strfree(name);
	mutex_exit(&ztest_vdev_lock);
	}

	static void
	ztest_spa_checkpoint(spa_t *spa)
	{
	ASSERT(MUTEX_HELD(&ztest_checkpoint_lock));

	int error = spa_checkpoint(spa->spa_name);

	switch (error) {
	case 0:
	case ZFS_ERR_DEVRM_IN_PROGRESS:
	case ZFS_ERR_DISCARDING_CHECKPOINT:
	case ZFS_ERR_CHECKPOINT_EXISTS:
	case ZFS_ERR_RAIDZ_EXPAND_IN_PROGRESS:
	break;
	case ENOSPC:
	ztest_record_enospc(FTAG);
	break;
	default:
	fatal(B_FALSE, "spa_checkpoint(%s) = %d", spa->spa_name, error);
	}
	}

	static void
	ztest_spa_discard_checkpoint(spa_t *spa)
	{
	ASSERT(MUTEX_HELD(&ztest_checkpoint_lock));

	int error = spa_checkpoint_discard(spa->spa_name);

	switch (error) {
	case 0:
	case ZFS_ERR_DISCARDING_CHECKPOINT:
	case ZFS_ERR_NO_CHECKPOINT:
	break;
	default:
	fatal(B_FALSE, "spa_discard_checkpoint(%s) = %d",
	spa->spa_name, error);
	}

	}

	void
	ztest_spa_checkpoint_create_discard(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	spa_t *spa = ztest_spa;

	mutex_enter(&ztest_checkpoint_lock);
	if (ztest_random(2) == 0) {
	ztest_spa_checkpoint(spa);
	} else {
	ztest_spa_discard_checkpoint(spa);
	}
	mutex_exit(&ztest_checkpoint_lock);
	}


	static vdev_t *
	vdev_lookup_by_path(vdev_t vd, const char path)
	{
	vdev_t *mvd;
	int c;

	if (vd->vdev_path != NULL && strcmp(path, vd->vdev_path) == 0)
	return (vd);

	for (c = 0; c < vd->vdev_children; c++)
	if ((mvd = vdev_lookup_by_path(vd->vdev_child[c], path)) !=
	NULL)
	return (mvd);

	return (NULL);
	}

	static int
	spa_num_top_vdevs(spa_t *spa)
	{
	vdev_t *rvd = spa->spa_root_vdev;
	ASSERT3U(spa_config_held(spa, SCL_VDEV, RW_READER), ==, SCL_VDEV);
	return (rvd->vdev_children);
	}

	/*
	* Verify that vdev_add() works as expected.
	*/
	void
	ztest_vdev_add_remove(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	ztest_shared_t *zs = ztest_shared;
	spa_t *spa = ztest_spa;
	uint64_t leaves;
	uint64_t guid;
	uint64_t raidz_children;

	nvlist_t *nvroot;
	int error;

	if (ztest_opts.zo_mmp_test)
	return;

	mutex_enter(&ztest_vdev_lock);
	raidz_children = ztest_get_raidz_children(spa);
	leaves = MAX(zs->zs_mirrors + zs->zs_splits, 1) * raidz_children;

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);

	ztest_shared->zs_vdev_next_leaf = spa_num_top_vdevs(spa) * leaves;

	/*
	* If we have slogs then remove them 1/4 of the time.
	*/
	if (spa_has_slogs(spa) && ztest_random(4) == 0) {
	metaslab_group_t *mg;

	/*
	* find the first real slog in log allocation class
	*/
	mg = spa_log_class(spa)->mc_allocator[0].mca_rotor;
	while (!mg->mg_vd->vdev_islog)
	mg = mg->mg_next;

	guid = mg->mg_vd->vdev_guid;

	spa_config_exit(spa, SCL_VDEV, FTAG);

	/*
	* We have to grab the zs_name_lock as writer to
	* prevent a race between removing a slog (dmu_objset_find)
	* and destroying a dataset. Removing the slog will
	* grab a reference on the dataset which may cause
	* dsl_destroy_head() to fail with EBUSY thus
	* leaving the dataset in an inconsistent state.
	*/
	pthread_rwlock_wrlock(&ztest_name_lock);
	error = spa_vdev_remove(spa, guid, B_FALSE);
	pthread_rwlock_unlock(&ztest_name_lock);

	switch (error) {
	case 0:
	case EEXIST: /* Generic zil_reset() error */
	case EBUSY: /* Replay required */
	case EACCES: /* Crypto key not loaded */
	case ZFS_ERR_CHECKPOINT_EXISTS:
	case ZFS_ERR_DISCARDING_CHECKPOINT:
	break;
	default:
	fatal(B_FALSE, "spa_vdev_remove() = %d", error);
	}
	} else {
	spa_config_exit(spa, SCL_VDEV, FTAG);

	/*
	* Make 1/4 of the devices be log devices
	*/
	nvroot = make_vdev_root(NULL, NULL, NULL,
	ztest_opts.zo_vdev_size, 0, (ztest_random(4) == 0) ?
	"log" : NULL, raidz_children, zs->zs_mirrors,
	1);

	error = spa_vdev_add(spa, nvroot, B_FALSE);
	fnvlist_free(nvroot);

	switch (error) {
	case 0:
	break;
	case ENOSPC:
	ztest_record_enospc("spa_vdev_add");
	break;
	default:
	fatal(B_FALSE, "spa_vdev_add() = %d", error);
	}
	}

	mutex_exit(&ztest_vdev_lock);
	}

	void
	ztest_vdev_class_add(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	ztest_shared_t *zs = ztest_shared;
	spa_t *spa = ztest_spa;
	uint64_t leaves;
	nvlist_t *nvroot;
	uint64_t raidz_children;
	const char *class = (ztest_random(2) == 0) ?
	VDEV_ALLOC_BIAS_SPECIAL : VDEV_ALLOC_BIAS_DEDUP;
	int error;

	/*
	* By default add a special vdev 50% of the time
	*/
	if ((ztest_opts.zo_special_vdevs == ZTEST_VDEV_CLASS_OFF) \|\|
	(ztest_opts.zo_special_vdevs == ZTEST_VDEV_CLASS_RND &&
	ztest_random(2) == 0)) {
	return;
	}

	mutex_enter(&ztest_vdev_lock);

	/* Only test with mirrors */
	if (zs->zs_mirrors < 2) {
	mutex_exit(&ztest_vdev_lock);
	return;
	}

	/* requires feature@allocation_classes */
	if (!spa_feature_is_enabled(spa, SPA_FEATURE_ALLOCATION_CLASSES)) {
	mutex_exit(&ztest_vdev_lock);
	return;
	}

	raidz_children = ztest_get_raidz_children(spa);
	leaves = MAX(zs->zs_mirrors + zs->zs_splits, 1) * raidz_children;

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
	ztest_shared->zs_vdev_next_leaf = spa_num_top_vdevs(spa) * leaves;
	spa_config_exit(spa, SCL_VDEV, FTAG);

	nvroot = make_vdev_root(NULL, NULL, NULL, ztest_opts.zo_vdev_size, 0,
	class, raidz_children, zs->zs_mirrors, 1);

	error = spa_vdev_add(spa, nvroot, B_FALSE);
	fnvlist_free(nvroot);

	if (error == ENOSPC)
	ztest_record_enospc("spa_vdev_add");
	else if (error != 0)
	fatal(B_FALSE, "spa_vdev_add() = %d", error);

	/*
	* 50% of the time allow small blocks in the special class
	*/
	if (error == 0 &&
	spa_special_class(spa)->mc_groups == 1 && ztest_random(2) == 0) {
	if (ztest_opts.zo_verbose >= 3)
	(void) printf("Enabling special VDEV small blocks\n");
	error = ztest_dsl_prop_set_uint64(zd->zd_name,
	ZFS_PROP_SPECIAL_SMALL_BLOCKS, 32768, B_FALSE);
	ASSERT(error == 0 \|\| error == ENOSPC);
	}

	mutex_exit(&ztest_vdev_lock);

	if (ztest_opts.zo_verbose >= 3) {
	metaslab_class_t *mc;

	if (strcmp(class, VDEV_ALLOC_BIAS_SPECIAL) == 0)
	mc = spa_special_class(spa);
	else
	mc = spa_dedup_class(spa);
	(void) printf("Added a %s mirrored vdev (of %d)\n",
	class, (int)mc->mc_groups);
	}
	}

	/*
	* Verify that adding/removing aux devices (l2arc, hot spare) works as expected.
	*/
	void
	ztest_vdev_aux_add_remove(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	ztest_shared_t *zs = ztest_shared;
	spa_t *spa = ztest_spa;
	vdev_t *rvd = spa->spa_root_vdev;
	spa_aux_vdev_t *sav;
	const char *aux;
	char *path;
	uint64_t guid = 0;
	int error, ignore_err = 0;

	if (ztest_opts.zo_mmp_test)
	return;

	path = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);

	if (ztest_random(2) == 0) {
	sav = &spa->spa_spares;
	aux = ZPOOL_CONFIG_SPARES;
	} else {
	sav = &spa->spa_l2cache;
	aux = ZPOOL_CONFIG_L2CACHE;
	}

	mutex_enter(&ztest_vdev_lock);

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);

	if (sav->sav_count != 0 && ztest_random(4) == 0) {
	/*
	* Pick a random device to remove.
	*/
	vdev_t *svd = sav->sav_vdevs[ztest_random(sav->sav_count)];

	/* dRAID spares cannot be removed; try anyways to see ENOTSUP */
	if (strstr(svd->vdev_path, VDEV_TYPE_DRAID) != NULL)
	ignore_err = ENOTSUP;

	guid = svd->vdev_guid;
	} else {
	/*
	* Find an unused device we can add.
	*/
	zs->zs_vdev_aux = 0;
	for (;;) {
	int c;
	(void) snprintf(path, MAXPATHLEN, ztest_aux_template,
	ztest_opts.zo_dir, ztest_opts.zo_pool, aux,
	zs->zs_vdev_aux);
	for (c = 0; c < sav->sav_count; c++)
	if (strcmp(sav->sav_vdevs[c]->vdev_path,
	path) == 0)
	break;
	if (c == sav->sav_count &&
	vdev_lookup_by_path(rvd, path) == NULL)
	break;
	zs->zs_vdev_aux++;
	}
	}

	spa_config_exit(spa, SCL_VDEV, FTAG);

	if (guid == 0) {
	/*
	* Add a new device.
	*/
	nvlist_t *nvroot = make_vdev_root(NULL, aux, NULL,
	(ztest_opts.zo_vdev_size * 5) / 4, 0, NULL, 0, 0, 1);
	error = spa_vdev_add(spa, nvroot, B_FALSE);

	switch (error) {
	case 0:
	break;
	default:
	fatal(B_FALSE, "spa_vdev_add(%p) = %d", nvroot, error);
	}
	fnvlist_free(nvroot);
	} else {
	/*
	* Remove an existing device. Sometimes, dirty its
	* vdev state first to make sure we handle removal
	* of devices that have pending state changes.
	*/
	if (ztest_random(2) == 0)
	(void) vdev_online(spa, guid, 0, NULL);

	error = spa_vdev_remove(spa, guid, B_FALSE);

	switch (error) {
	case 0:
	case EBUSY:
	case ZFS_ERR_CHECKPOINT_EXISTS:
	case ZFS_ERR_DISCARDING_CHECKPOINT:
	break;
	default:
	if (error != ignore_err)
	fatal(B_FALSE,
	"spa_vdev_remove(%"PRIu64") = %d",
	guid, error);
	}
	}

	mutex_exit(&ztest_vdev_lock);

	umem_free(path, MAXPATHLEN);
	}

	/*
	* split a pool if it has mirror tlvdevs
	*/
	void
	ztest_split_pool(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	ztest_shared_t *zs = ztest_shared;
	spa_t *spa = ztest_spa;
	vdev_t *rvd = spa->spa_root_vdev;
	nvlist_t tree, child, config, split, *schild;
	uint_t c, children, schildren = 0, lastlogid = 0;
	int error = 0;

	if (ztest_opts.zo_mmp_test)
	return;

	mutex_enter(&ztest_vdev_lock);

	/* ensure we have a usable config; mirrors of raidz aren't supported */
	if (zs->zs_mirrors < 3 \|\| ztest_opts.zo_raid_children > 1) {
	mutex_exit(&ztest_vdev_lock);
	return;
	}

	/* clean up the old pool, if any */
	(void) spa_destroy("splitp");

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);

	/* generate a config from the existing config */
	mutex_enter(&spa->spa_props_lock);
	tree = fnvlist_lookup_nvlist(spa->spa_config, ZPOOL_CONFIG_VDEV_TREE);
	mutex_exit(&spa->spa_props_lock);

	VERIFY0(nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN,
	&child, &children));

	schild = umem_alloc(rvd->vdev_children * sizeof (nvlist_t *),
	UMEM_NOFAIL);
	for (c = 0; c < children; c++) {
	vdev_t *tvd = rvd->vdev_child[c];
	nvlist_t **mchild;
	uint_t mchildren;

	if (tvd->vdev_islog \|\| tvd->vdev_ops == &vdev_hole_ops) {
	schild[schildren] = fnvlist_alloc();
	fnvlist_add_string(schild[schildren],
	ZPOOL_CONFIG_TYPE, VDEV_TYPE_HOLE);
	fnvlist_add_uint64(schild[schildren],
	ZPOOL_CONFIG_IS_HOLE, 1);
	if (lastlogid == 0)
	lastlogid = schildren;
	++schildren;
	continue;
	}
	lastlogid = 0;
	VERIFY0(nvlist_lookup_nvlist_array(child[c],
	ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren));
	schild[schildren++] = fnvlist_dup(mchild[0]);
	}

	/* OK, create a config that can be used to split */
	split = fnvlist_alloc();
	fnvlist_add_string(split, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT);
	fnvlist_add_nvlist_array(split, ZPOOL_CONFIG_CHILDREN,
	(const nvlist_t **)schild, lastlogid != 0 ? lastlogid : schildren);

	config = fnvlist_alloc();
	fnvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, split);

	for (c = 0; c < schildren; c++)
	fnvlist_free(schild[c]);
	umem_free(schild, rvd->vdev_children * sizeof (nvlist_t *));
	fnvlist_free(split);

	spa_config_exit(spa, SCL_VDEV, FTAG);

	(void) pthread_rwlock_wrlock(&ztest_name_lock);
	error = spa_vdev_split_mirror(spa, "splitp", config, NULL, B_FALSE);
	(void) pthread_rwlock_unlock(&ztest_name_lock);

	fnvlist_free(config);

	if (error == 0) {
	(void) printf("successful split - results:\n");
	mutex_enter(&spa_namespace_lock);
	show_pool_stats(spa);
	show_pool_stats(spa_lookup("splitp"));
	mutex_exit(&spa_namespace_lock);
	++zs->zs_splits;
	--zs->zs_mirrors;
	}
	mutex_exit(&ztest_vdev_lock);
	}

	/*
	* Verify that we can attach and detach devices.
	*/
	void
	ztest_vdev_attach_detach(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	ztest_shared_t *zs = ztest_shared;
	spa_t *spa = ztest_spa;
	spa_aux_vdev_t *sav = &spa->spa_spares;
	vdev_t *rvd = spa->spa_root_vdev;
	vdev_t oldvd, newvd, *pvd;
	nvlist_t *root;
	uint64_t leaves;
	uint64_t leaf, top;
	uint64_t ashift = ztest_get_ashift();
	uint64_t oldguid, pguid;
	uint64_t oldsize, newsize;
	uint64_t raidz_children;
	char oldpath, newpath;
	int replacing;
	int oldvd_has_siblings = B_FALSE;
	int newvd_is_spare = B_FALSE;
	int newvd_is_dspare = B_FALSE;
	int oldvd_is_log;
	int oldvd_is_special;
	int error, expected_error;

	if (ztest_opts.zo_mmp_test)
	return;

	oldpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
	newpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);

	mutex_enter(&ztest_vdev_lock);
	raidz_children = ztest_get_raidz_children(spa);
	leaves = MAX(zs->zs_mirrors, 1) * raidz_children;

	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);

	/*
	* If a vdev is in the process of being removed, its removal may
	* finish while we are in progress, leading to an unexpected error
	* value. Don't bother trying to attach while we are in the middle
	* of removal.
	*/
	if (ztest_device_removal_active) {
	spa_config_exit(spa, SCL_ALL, FTAG);
	goto out;
	}

	/*
	* RAIDZ leaf VDEV mirrors are not currently supported while a
	* RAIDZ expansion is in progress.
	*/
	if (ztest_opts.zo_raid_do_expand) {
	spa_config_exit(spa, SCL_ALL, FTAG);
	goto out;
	}

	/*
	* Decide whether to do an attach or a replace.
	*/
	replacing = ztest_random(2);

	/*
	* Pick a random top-level vdev.
	*/
	top = ztest_random_vdev_top(spa, B_TRUE);

	/*
	* Pick a random leaf within it.
	*/
	leaf = ztest_random(leaves);

	/*
	* Locate this vdev.
	*/
	oldvd = rvd->vdev_child[top];

	/* pick a child from the mirror */
	if (zs->zs_mirrors >= 1) {
	ASSERT3P(oldvd->vdev_ops, ==, &vdev_mirror_ops);
	ASSERT3U(oldvd->vdev_children, >=, zs->zs_mirrors);
	oldvd = oldvd->vdev_child[leaf / raidz_children];
	}

	/* pick a child out of the raidz group */
	if (ztest_opts.zo_raid_children > 1) {
	if (strcmp(oldvd->vdev_ops->vdev_op_type, "raidz") == 0)
	ASSERT3P(oldvd->vdev_ops, ==, &vdev_raidz_ops);
	else
	ASSERT3P(oldvd->vdev_ops, ==, &vdev_draid_ops);
	oldvd = oldvd->vdev_child[leaf % raidz_children];
	}

	/*
	* If we're already doing an attach or replace, oldvd may be a
	* mirror vdev -- in which case, pick a random child.
	*/
	while (oldvd->vdev_children != 0) {
	oldvd_has_siblings = B_TRUE;
	ASSERT3U(oldvd->vdev_children, >=, 2);
	oldvd = oldvd->vdev_child[ztest_random(oldvd->vdev_children)];
	}

	oldguid = oldvd->vdev_guid;
	oldsize = vdev_get_min_asize(oldvd);
	oldvd_is_log = oldvd->vdev_top->vdev_islog;
	oldvd_is_special =
	oldvd->vdev_top->vdev_alloc_bias == VDEV_BIAS_SPECIAL \|\|
	oldvd->vdev_top->vdev_alloc_bias == VDEV_BIAS_DEDUP;
	(void) strlcpy(oldpath, oldvd->vdev_path, MAXPATHLEN);
	pvd = oldvd->vdev_parent;
	pguid = pvd->vdev_guid;

	/*
	* If oldvd has siblings, then half of the time, detach it. Prior
	* to the detach the pool is scrubbed in order to prevent creating
	* unrepairable blocks as a result of the data corruption injection.
	*/
	if (oldvd_has_siblings && ztest_random(2) == 0) {
	spa_config_exit(spa, SCL_ALL, FTAG);

	error = ztest_scrub_impl(spa);
	if (error)
	goto out;

	error = spa_vdev_detach(spa, oldguid, pguid, B_FALSE);
	if (error != 0 && error != ENODEV && error != EBUSY &&
	error != ENOTSUP && error != ZFS_ERR_CHECKPOINT_EXISTS &&
	error != ZFS_ERR_DISCARDING_CHECKPOINT)
	fatal(B_FALSE, "detach (%s) returned %d",
	oldpath, error);
	goto out;
	}

	/*
	* For the new vdev, choose with equal probability between the two
	* standard paths (ending in either 'a' or 'b') or a random hot spare.
	*/
	if (sav->sav_count != 0 && ztest_random(3) == 0) {
	newvd = sav->sav_vdevs[ztest_random(sav->sav_count)];
	newvd_is_spare = B_TRUE;

	if (newvd->vdev_ops == &vdev_draid_spare_ops)
	newvd_is_dspare = B_TRUE;

	(void) strlcpy(newpath, newvd->vdev_path, MAXPATHLEN);
	} else {
	(void) snprintf(newpath, MAXPATHLEN, ztest_dev_template,
	ztest_opts.zo_dir, ztest_opts.zo_pool,
	top * leaves + leaf);
	if (ztest_random(2) == 0)
	newpath[strlen(newpath) - 1] = 'b';
	newvd = vdev_lookup_by_path(rvd, newpath);
	}

	if (newvd) {
	/*
	* Reopen to ensure the vdev's asize field isn't stale.
	*/
	vdev_reopen(newvd);
	newsize = vdev_get_min_asize(newvd);
	} else {
	/*
	* Make newsize a little bigger or smaller than oldsize.
	* If it's smaller, the attach should fail.
	* If it's larger, and we're doing a replace,
	* we should get dynamic LUN growth when we're done.
	*/
	newsize = 10 * oldsize / (9 + ztest_random(3));
	}

	/*
	* If pvd is not a mirror or root, the attach should fail with ENOTSUP,
	* unless it's a replace; in that case any non-replacing parent is OK.
	*
	* If newvd is already part of the pool, it should fail with EBUSY.
	*
	* If newvd is too small, it should fail with EOVERFLOW.
	*
	* If newvd is a distributed spare and it's being attached to a
	* dRAID which is not its parent it should fail with EINVAL.
	*/
	if (pvd->vdev_ops != &vdev_mirror_ops &&
	pvd->vdev_ops != &vdev_root_ops && (!replacing \|\|
	pvd->vdev_ops == &vdev_replacing_ops \|\|
	pvd->vdev_ops == &vdev_spare_ops))
	expected_error = ENOTSUP;
	else if (newvd_is_spare &&
	(!replacing \|\| oldvd_is_log \|\| oldvd_is_special))
	expected_error = ENOTSUP;
	else if (newvd == oldvd)
	expected_error = replacing ? 0 : EBUSY;
	else if (vdev_lookup_by_path(rvd, newpath) != NULL)
	expected_error = EBUSY;
	else if (!newvd_is_dspare && newsize < oldsize)
	expected_error = EOVERFLOW;
	else if (ashift > oldvd->vdev_top->vdev_ashift)
	expected_error = EDOM;
	else if (newvd_is_dspare && pvd != vdev_draid_spare_get_parent(newvd))
	expected_error = EINVAL;
	else
	expected_error = 0;

	spa_config_exit(spa, SCL_ALL, FTAG);

	/*
	* Build the nvlist describing newpath.
	*/
	root = make_vdev_root(newpath, NULL, NULL, newvd == NULL ? newsize : 0,
	ashift, NULL, 0, 0, 1);

	/*
	* When supported select either a healing or sequential resilver.
	*/
	boolean_t rebuilding = B_FALSE;
	if (pvd->vdev_ops == &vdev_mirror_ops \|\|
	pvd->vdev_ops == &vdev_root_ops) {
	rebuilding = !!ztest_random(2);
	}

	error = spa_vdev_attach(spa, oldguid, root, replacing, rebuilding);

	fnvlist_free(root);

	/*
	* If our parent was the replacing vdev, but the replace completed,
	* then instead of failing with ENOTSUP we may either succeed,
	* fail with ENODEV, or fail with EOVERFLOW.
	*/
	if (expected_error == ENOTSUP &&
	(error == 0 \|\| error == ENODEV \|\| error == EOVERFLOW))
	expected_error = error;

	/*
	* If someone grew the LUN, the replacement may be too small.
	*/
	if (error == EOVERFLOW \|\| error == EBUSY)
	expected_error = error;

	if (error == ZFS_ERR_CHECKPOINT_EXISTS \|\|
	error == ZFS_ERR_DISCARDING_CHECKPOINT \|\|
	error == ZFS_ERR_RESILVER_IN_PROGRESS \|\|
	error == ZFS_ERR_REBUILD_IN_PROGRESS)
	expected_error = error;

	if (error != expected_error && expected_error != EBUSY) {
	fatal(B_FALSE, "attach (%s %"PRIu64", %s %"PRIu64", %d) "
	"returned %d, expected %d",
	oldpath, oldsize, newpath,
	newsize, replacing, error, expected_error);
	}
	out:
	mutex_exit(&ztest_vdev_lock);

	umem_free(oldpath, MAXPATHLEN);
	umem_free(newpath, MAXPATHLEN);
	}

	static void
	raidz_scratch_verify(void)
	{
	spa_t *spa;
	uint64_t write_size, logical_size, offset;
	raidz_reflow_scratch_state_t state;
	vdev_raidz_expand_t *vre;
	vdev_t *raidvd;

	ASSERT(raidz_expand_pause_point == RAIDZ_EXPAND_PAUSE_NONE);

	if (ztest_scratch_state->zs_raidz_scratch_verify_pause == 0)
	return;

	kernel_init(SPA_MODE_READ);

	mutex_enter(&spa_namespace_lock);
	spa = spa_lookup(ztest_opts.zo_pool);
	ASSERT(spa);
	spa->spa_import_flags \|= ZFS_IMPORT_SKIP_MMP;
	mutex_exit(&spa_namespace_lock);

	VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG));

	ASSERT3U(RRSS_GET_OFFSET(&spa->spa_uberblock), !=, UINT64_MAX);

	mutex_enter(&ztest_vdev_lock);

	spa_config_enter(spa, SCL_ALL, FTAG, RW_READER);

	vre = spa->spa_raidz_expand;
	if (vre == NULL)
	goto out;

	raidvd = vdev_lookup_top(spa, vre->vre_vdev_id);
	offset = RRSS_GET_OFFSET(&spa->spa_uberblock);
	state = RRSS_GET_STATE(&spa->spa_uberblock);
	- write_size = P2ALIGN(VDEV_BOOT_SIZE, 1 << raidvd->vdev_ashift);
	+ write_size = P2ALIGN_TYPED(VDEV_BOOT_SIZE, 1 << raidvd->vdev_ashift,
	+ uint64_t);
	logical_size = write_size * raidvd->vdev_children;

	switch (state) {
	/*
	* Initial state of reflow process. RAIDZ expansion was
	* requested by user, but scratch object was not created.
	*/
	case RRSS_SCRATCH_NOT_IN_USE:
	ASSERT3U(offset, ==, 0);
	break;

	/*
	* Scratch object was synced and stored in boot area.
	*/
	case RRSS_SCRATCH_VALID:

	/*
	* Scratch object was synced back to raidz start offset,
	* raidz is ready for sector by sector reflow process.
	*/
	case RRSS_SCRATCH_INVALID_SYNCED:

	/*
	* Scratch object was synced back to raidz start offset
	* on zpool importing, raidz is ready for sector by sector
	* reflow process.
	*/
	case RRSS_SCRATCH_INVALID_SYNCED_ON_IMPORT:
	ASSERT3U(offset, ==, logical_size);
	break;

	/*
	* Sector by sector reflow process started.
	*/
	case RRSS_SCRATCH_INVALID_SYNCED_REFLOW:
	ASSERT3U(offset, >=, logical_size);
	break;
	}

	out:
	spa_config_exit(spa, SCL_ALL, FTAG);

	mutex_exit(&ztest_vdev_lock);

	ztest_scratch_state->zs_raidz_scratch_verify_pause = 0;

	spa_close(spa, FTAG);
	kernel_fini();
	}

	static void
	ztest_scratch_thread(void *arg)
	{
	(void) arg;

	/* wait up to 10 seconds */
	for (int t = 100; t > 0; t -= 1) {
	if (raidz_expand_pause_point == RAIDZ_EXPAND_PAUSE_NONE)
	thread_exit();

	(void) poll(NULL, 0, 100);
	}

	/* killed when the scratch area progress reached a certain point */
	ztest_kill(ztest_shared);
	}

	/*
	* Verify that we can attach raidz device.
	*/
	void
	ztest_vdev_raidz_attach(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	ztest_shared_t *zs = ztest_shared;
	spa_t *spa = ztest_spa;
	uint64_t leaves, raidz_children, newsize, ashift = ztest_get_ashift();
	kthread_t *scratch_thread = NULL;
	vdev_t newvd, pvd;
	nvlist_t *root;
	char *newpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
	int error, expected_error = 0;

	mutex_enter(&ztest_vdev_lock);

	spa_config_enter(spa, SCL_ALL, FTAG, RW_READER);

	/* Only allow attach when raid-kind = 'eraidz' */
	if (!ztest_opts.zo_raid_do_expand) {
	spa_config_exit(spa, SCL_ALL, FTAG);
	goto out;
	}

	if (ztest_opts.zo_mmp_test) {
	spa_config_exit(spa, SCL_ALL, FTAG);
	goto out;
	}

	if (ztest_device_removal_active) {
	spa_config_exit(spa, SCL_ALL, FTAG);
	goto out;
	}

	pvd = vdev_lookup_top(spa, 0);

	ASSERT(pvd->vdev_ops == &vdev_raidz_ops);

	/*
	* Get size of a child of the raidz group,
	* make sure device is a bit bigger
	*/
	newvd = pvd->vdev_child[ztest_random(pvd->vdev_children)];
	newsize = 10 * vdev_get_min_asize(newvd) / (9 + ztest_random(2));

	/*
	* Get next attached leaf id
	*/
	raidz_children = ztest_get_raidz_children(spa);
	leaves = MAX(zs->zs_mirrors + zs->zs_splits, 1) * raidz_children;
	zs->zs_vdev_next_leaf = spa_num_top_vdevs(spa) * leaves;

	if (spa->spa_raidz_expand)
	expected_error = ZFS_ERR_RAIDZ_EXPAND_IN_PROGRESS;

	spa_config_exit(spa, SCL_ALL, FTAG);

	/*
	* Path to vdev to be attached
	*/
	(void) snprintf(newpath, MAXPATHLEN, ztest_dev_template,
	ztest_opts.zo_dir, ztest_opts.zo_pool, zs->zs_vdev_next_leaf);

	/*
	* Build the nvlist describing newpath.
	*/
	root = make_vdev_root(newpath, NULL, NULL, newsize, ashift, NULL,
	0, 0, 1);

	/*
	* 50% of the time, set raidz_expand_pause_point to cause
	* raidz_reflow_scratch_sync() to pause at a certain point and
	* then kill the test after 10 seconds so raidz_scratch_verify()
	* can confirm consistency when the pool is imported.
	*/
	if (ztest_random(2) == 0 && expected_error == 0) {
	raidz_expand_pause_point =
	ztest_random(RAIDZ_EXPAND_PAUSE_SCRATCH_POST_REFLOW_2) + 1;
	scratch_thread = thread_create(NULL, 0, ztest_scratch_thread,
	ztest_shared, 0, NULL, TS_RUN \| TS_JOINABLE, defclsyspri);
	}

	error = spa_vdev_attach(spa, pvd->vdev_guid, root, B_FALSE, B_FALSE);

	nvlist_free(root);

	if (error == EOVERFLOW \|\| error == ENXIO \|\|
	error == ZFS_ERR_CHECKPOINT_EXISTS \|\|
	error == ZFS_ERR_DISCARDING_CHECKPOINT)
	expected_error = error;

	if (error != 0 && error != expected_error) {
	fatal(0, "raidz attach (%s %"PRIu64") returned %d, expected %d",
	newpath, newsize, error, expected_error);
	}

	if (raidz_expand_pause_point) {
	if (error != 0) {
	/*
	* Do not verify scratch object in case of error
	* returned by vdev attaching.
	*/
	raidz_expand_pause_point = RAIDZ_EXPAND_PAUSE_NONE;
	}

	VERIFY0(thread_join(scratch_thread));
	}
	out:
	mutex_exit(&ztest_vdev_lock);

	umem_free(newpath, MAXPATHLEN);
	}

	void
	ztest_device_removal(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	spa_t *spa = ztest_spa;
	vdev_t *vd;
	uint64_t guid;
	int error;

	mutex_enter(&ztest_vdev_lock);

	if (ztest_device_removal_active) {
	mutex_exit(&ztest_vdev_lock);
	return;
	}

	/*
	* Remove a random top-level vdev and wait for removal to finish.
	*/
	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
	vd = vdev_lookup_top(spa, ztest_random_vdev_top(spa, B_FALSE));
	guid = vd->vdev_guid;
	spa_config_exit(spa, SCL_VDEV, FTAG);

	error = spa_vdev_remove(spa, guid, B_FALSE);
	if (error == 0) {
	ztest_device_removal_active = B_TRUE;
	mutex_exit(&ztest_vdev_lock);

	/*
	* spa->spa_vdev_removal is created in a sync task that
	* is initiated via dsl_sync_task_nowait(). Since the
	* task may not run before spa_vdev_remove() returns, we
	* must wait at least 1 txg to ensure that the removal
	* struct has been created.
	*/
	txg_wait_synced(spa_get_dsl(spa), 0);

	while (spa->spa_removing_phys.sr_state == DSS_SCANNING)
	txg_wait_synced(spa_get_dsl(spa), 0);
	} else {
	mutex_exit(&ztest_vdev_lock);
	return;
	}

	/*
	* The pool needs to be scrubbed after completing device removal.
	* Failure to do so may result in checksum errors due to the
	* strategy employed by ztest_fault_inject() when selecting which
	* offset are redundant and can be damaged.
	*/
	error = spa_scan(spa, POOL_SCAN_SCRUB);
	if (error == 0) {
	while (dsl_scan_scrubbing(spa_get_dsl(spa)))
	txg_wait_synced(spa_get_dsl(spa), 0);
	}

	mutex_enter(&ztest_vdev_lock);
	ztest_device_removal_active = B_FALSE;
	mutex_exit(&ztest_vdev_lock);
	}

	/*
	* Callback function which expands the physical size of the vdev.
	*/
	static vdev_t *
	grow_vdev(vdev_t vd, void arg)
	{
	spa_t *spa __maybe_unused = vd->vdev_spa;
	size_t *newsize = arg;
	size_t fsize;
	int fd;

	ASSERT3S(spa_config_held(spa, SCL_STATE, RW_READER), ==, SCL_STATE);
	ASSERT(vd->vdev_ops->vdev_op_leaf);

	if ((fd = open(vd->vdev_path, O_RDWR)) == -1)
	return (vd);

	fsize = lseek(fd, 0, SEEK_END);
	VERIFY0(ftruncate(fd, *newsize));

	if (ztest_opts.zo_verbose >= 6) {
	(void) printf("%s grew from %lu to %lu bytes\n",
	vd->vdev_path, (ulong_t)fsize, (ulong_t)*newsize);
	}
	(void) close(fd);
	return (NULL);
	}

	/*
	* Callback function which expands a given vdev by calling vdev_online().
	*/
	static vdev_t *
	online_vdev(vdev_t vd, void arg)
	{
	(void) arg;
	spa_t *spa = vd->vdev_spa;
	vdev_t *tvd = vd->vdev_top;
	uint64_t guid = vd->vdev_guid;
	uint64_t generation = spa->spa_config_generation + 1;
	vdev_state_t newstate = VDEV_STATE_UNKNOWN;
	int error;

	ASSERT3S(spa_config_held(spa, SCL_STATE, RW_READER), ==, SCL_STATE);
	ASSERT(vd->vdev_ops->vdev_op_leaf);

	/* Calling vdev_online will initialize the new metaslabs */
	spa_config_exit(spa, SCL_STATE, spa);
	error = vdev_online(spa, guid, ZFS_ONLINE_EXPAND, &newstate);
	spa_config_enter(spa, SCL_STATE, spa, RW_READER);

	/*
	* If vdev_online returned an error or the underlying vdev_open
	* failed then we abort the expand. The only way to know that
	* vdev_open fails is by checking the returned newstate.
	*/
	if (error \|\| newstate != VDEV_STATE_HEALTHY) {
	if (ztest_opts.zo_verbose >= 5) {
	(void) printf("Unable to expand vdev, state %u, "
	"error %d\n", newstate, error);
	}
	return (vd);
	}
	ASSERT3U(newstate, ==, VDEV_STATE_HEALTHY);

	/*
	* Since we dropped the lock we need to ensure that we're
	* still talking to the original vdev. It's possible this
	* vdev may have been detached/replaced while we were
	* trying to online it.
	*/
	if (generation != spa->spa_config_generation) {
	if (ztest_opts.zo_verbose >= 5) {
	(void) printf("vdev configuration has changed, "
	"guid %"PRIu64", state %"PRIu64", "
	"expected gen %"PRIu64", got gen %"PRIu64"\n",
	guid,
	tvd->vdev_state,
	generation,
	spa->spa_config_generation);
	}
	return (vd);
	}
	return (NULL);
	}

	/*
	* Traverse the vdev tree calling the supplied function.
	* We continue to walk the tree until we either have walked all
	* children or we receive a non-NULL return from the callback.
	* If a NULL callback is passed, then we just return back the first
	* leaf vdev we encounter.
	*/
	static vdev_t *
	vdev_walk_tree(vdev_t vd, vdev_t (func)(vdev_t , void ), void arg)
	{
	uint_t c;

	if (vd->vdev_ops->vdev_op_leaf) {
	if (func == NULL)
	return (vd);
	else
	return (func(vd, arg));
	}

	for (c = 0; c < vd->vdev_children; c++) {
	vdev_t *cvd = vd->vdev_child[c];
	if ((cvd = vdev_walk_tree(cvd, func, arg)) != NULL)
	return (cvd);
	}
	return (NULL);
	}

	/*
	* Verify that dynamic LUN growth works as expected.
	*/
	void
	ztest_vdev_LUN_growth(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	spa_t *spa = ztest_spa;
	vdev_t vd, tvd;
	metaslab_class_t *mc;
	metaslab_group_t *mg;
	size_t psize, newsize;
	uint64_t top;
	uint64_t old_class_space, new_class_space, old_ms_count, new_ms_count;

	mutex_enter(&ztest_checkpoint_lock);
	mutex_enter(&ztest_vdev_lock);
	spa_config_enter(spa, SCL_STATE, spa, RW_READER);

	/*
	* If there is a vdev removal in progress, it could complete while
	* we are running, in which case we would not be able to verify
	* that the metaslab_class space increased (because it decreases
	* when the device removal completes).
	*/
	if (ztest_device_removal_active) {
	spa_config_exit(spa, SCL_STATE, spa);
	mutex_exit(&ztest_vdev_lock);
	mutex_exit(&ztest_checkpoint_lock);
	return;
	}

	/*
	* If we are under raidz expansion, the test can failed because the
	* metaslabs count will not increase immediately after the vdev is
	* expanded. It will happen only after raidz expansion completion.
	*/
	if (spa->spa_raidz_expand) {
	spa_config_exit(spa, SCL_STATE, spa);
	mutex_exit(&ztest_vdev_lock);
	mutex_exit(&ztest_checkpoint_lock);
	return;
	}

	top = ztest_random_vdev_top(spa, B_TRUE);

	tvd = spa->spa_root_vdev->vdev_child[top];
	mg = tvd->vdev_mg;
	mc = mg->mg_class;
	old_ms_count = tvd->vdev_ms_count;
	old_class_space = metaslab_class_get_space(mc);

	/*
	* Determine the size of the first leaf vdev associated with
	* our top-level device.
	*/
	vd = vdev_walk_tree(tvd, NULL, NULL);
	ASSERT3P(vd, !=, NULL);
	ASSERT(vd->vdev_ops->vdev_op_leaf);

	psize = vd->vdev_psize;

	/*
	* We only try to expand the vdev if it's healthy, less than 4x its
	* original size, and it has a valid psize.
	*/
	if (tvd->vdev_state != VDEV_STATE_HEALTHY \|\|
	psize == 0 \|\| psize >= 4 * ztest_opts.zo_vdev_size) {
	spa_config_exit(spa, SCL_STATE, spa);
	mutex_exit(&ztest_vdev_lock);
	mutex_exit(&ztest_checkpoint_lock);
	return;
	}
	ASSERT3U(psize, >, 0);
	newsize = psize + MAX(psize / 8, SPA_MAXBLOCKSIZE);
	ASSERT3U(newsize, >, psize);

	if (ztest_opts.zo_verbose >= 6) {
	(void) printf("Expanding LUN %s from %lu to %lu\n",
	vd->vdev_path, (ulong_t)psize, (ulong_t)newsize);
	}

	/*
	* Growing the vdev is a two step process:
	* 1). expand the physical size (i.e. relabel)
	* 2). online the vdev to create the new metaslabs
	*/
	if (vdev_walk_tree(tvd, grow_vdev, &newsize) != NULL \|\|
	vdev_walk_tree(tvd, online_vdev, NULL) != NULL \|\|
	tvd->vdev_state != VDEV_STATE_HEALTHY) {
	if (ztest_opts.zo_verbose >= 5) {
	(void) printf("Could not expand LUN because "
	"the vdev configuration changed.\n");
	}
	spa_config_exit(spa, SCL_STATE, spa);
	mutex_exit(&ztest_vdev_lock);
	mutex_exit(&ztest_checkpoint_lock);
	return;
	}

	spa_config_exit(spa, SCL_STATE, spa);

	/*
	* Expanding the LUN will update the config asynchronously,
	* thus we must wait for the async thread to complete any
	* pending tasks before proceeding.
	*/
	for (;;) {
	boolean_t done;
	mutex_enter(&spa->spa_async_lock);
	done = (spa->spa_async_thread == NULL && !spa->spa_async_tasks);
	mutex_exit(&spa->spa_async_lock);
	if (done)
	break;
	txg_wait_synced(spa_get_dsl(spa), 0);
	(void) poll(NULL, 0, 100);
	}

	spa_config_enter(spa, SCL_STATE, spa, RW_READER);

	tvd = spa->spa_root_vdev->vdev_child[top];
	new_ms_count = tvd->vdev_ms_count;
	new_class_space = metaslab_class_get_space(mc);

	if (tvd->vdev_mg != mg \|\| mg->mg_class != mc) {
	if (ztest_opts.zo_verbose >= 5) {
	(void) printf("Could not verify LUN expansion due to "
	"intervening vdev offline or remove.\n");
	}
	spa_config_exit(spa, SCL_STATE, spa);
	mutex_exit(&ztest_vdev_lock);
	mutex_exit(&ztest_checkpoint_lock);
	return;
	}

	/*
	* Make sure we were able to grow the vdev.
	*/
	if (new_ms_count <= old_ms_count) {
	fatal(B_FALSE,
	"LUN expansion failed: ms_count %"PRIu64" < %"PRIu64"\n",
	old_ms_count, new_ms_count);
	}

	/*
	* Make sure we were able to grow the pool.
	*/
	if (new_class_space <= old_class_space) {
	fatal(B_FALSE,
	"LUN expansion failed: class_space %"PRIu64" < %"PRIu64"\n",
	old_class_space, new_class_space);
	}

	if (ztest_opts.zo_verbose >= 5) {
	char oldnumbuf[NN_NUMBUF_SZ], newnumbuf[NN_NUMBUF_SZ];

	nicenum(old_class_space, oldnumbuf, sizeof (oldnumbuf));
	nicenum(new_class_space, newnumbuf, sizeof (newnumbuf));
	(void) printf("%s grew from %s to %s\n",
	spa->spa_name, oldnumbuf, newnumbuf);
	}

	spa_config_exit(spa, SCL_STATE, spa);
	mutex_exit(&ztest_vdev_lock);
	mutex_exit(&ztest_checkpoint_lock);
	}

	/*
	* Verify that dmu_objset_{create,destroy,open,close} work as expected.
	*/
	static void
	ztest_objset_create_cb(objset_t os, void arg, cred_t cr, dmu_tx_t tx)
	{
	(void) arg, (void) cr;

	/*
	* Create the objects common to all ztest datasets.
	*/
	VERIFY0(zap_create_claim(os, ZTEST_DIROBJ,
	DMU_OT_ZAP_OTHER, DMU_OT_NONE, 0, tx));
	}

	static int
	ztest_dataset_create(char *dsname)
	{
	int err;
	uint64_t rand;
	dsl_crypto_params_t *dcp = NULL;

	/*
	* 50% of the time, we create encrypted datasets
	* using a random cipher suite and a hard-coded
	* wrapping key.
	*/
	rand = ztest_random(2);
	if (rand != 0) {
	nvlist_t *crypto_args = fnvlist_alloc();
	nvlist_t *props = fnvlist_alloc();

	/* slight bias towards the default cipher suite */
	rand = ztest_random(ZIO_CRYPT_FUNCTIONS);
	if (rand < ZIO_CRYPT_AES_128_CCM)
	rand = ZIO_CRYPT_ON;

	fnvlist_add_uint64(props,
	zfs_prop_to_name(ZFS_PROP_ENCRYPTION), rand);
	fnvlist_add_uint8_array(crypto_args, "wkeydata",
	(uint8_t *)ztest_wkeydata, WRAPPING_KEY_LEN);

	/*
	* These parameters aren't really used by the kernel. They
	* are simply stored so that userspace knows how to load
	* the wrapping key.
	*/
	fnvlist_add_uint64(props,
	zfs_prop_to_name(ZFS_PROP_KEYFORMAT), ZFS_KEYFORMAT_RAW);
	fnvlist_add_string(props,
	zfs_prop_to_name(ZFS_PROP_KEYLOCATION), "prompt");
	fnvlist_add_uint64(props,
	zfs_prop_to_name(ZFS_PROP_PBKDF2_SALT), 0ULL);
	fnvlist_add_uint64(props,
	zfs_prop_to_name(ZFS_PROP_PBKDF2_ITERS), 0ULL);

	VERIFY0(dsl_crypto_params_create_nvlist(DCP_CMD_NONE, props,
	crypto_args, &dcp));

	/*
	* Cycle through all available encryption implementations
	* to verify interoperability.
	*/
	VERIFY0(gcm_impl_set("cycle"));
	VERIFY0(aes_impl_set("cycle"));

	fnvlist_free(crypto_args);
	fnvlist_free(props);
	}

	err = dmu_objset_create(dsname, DMU_OST_OTHER, 0, dcp,
	ztest_objset_create_cb, NULL);
	dsl_crypto_params_free(dcp, !!err);

	rand = ztest_random(100);
	if (err \|\| rand < 80)
	return (err);

	if (ztest_opts.zo_verbose >= 5)
	(void) printf("Setting dataset %s to sync always\n", dsname);
	return (ztest_dsl_prop_set_uint64(dsname, ZFS_PROP_SYNC,
	ZFS_SYNC_ALWAYS, B_FALSE));
	}

	static int
	ztest_objset_destroy_cb(const char name, void arg)
	{
	(void) arg;
	objset_t *os;
	dmu_object_info_t doi;
	int error;

	/*
	* Verify that the dataset contains a directory object.
	*/
	VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_TRUE,
	B_TRUE, FTAG, &os));
	error = dmu_object_info(os, ZTEST_DIROBJ, &doi);
	if (error != ENOENT) {
	/* We could have crashed in the middle of destroying it */
	ASSERT0(error);
	ASSERT3U(doi.doi_type, ==, DMU_OT_ZAP_OTHER);
	ASSERT3S(doi.doi_physical_blocks_512, >=, 0);
	}
	dmu_objset_disown(os, B_TRUE, FTAG);

	/*
	* Destroy the dataset.
	*/
	if (strchr(name, '@') != NULL) {
	error = dsl_destroy_snapshot(name, B_TRUE);
	if (error != ECHRNG) {
	/*
	* The program was executed, but encountered a runtime
	* error, such as insufficient slop, or a hold on the
	* dataset.
	*/
	ASSERT0(error);
	}
	} else {
	error = dsl_destroy_head(name);
	if (error == ENOSPC) {
	/* There could be checkpoint or insufficient slop */
	ztest_record_enospc(FTAG);
	} else if (error != EBUSY) {
	/* There could be a hold on this dataset */
	ASSERT0(error);
	}
	}
	return (0);
	}

	static boolean_t
	ztest_snapshot_create(char *osname, uint64_t id)
	{
	char snapname[ZFS_MAX_DATASET_NAME_LEN];
	int error;

	(void) snprintf(snapname, sizeof (snapname), "%"PRIu64"", id);

	error = dmu_objset_snapshot_one(osname, snapname);
	if (error == ENOSPC) {
	ztest_record_enospc(FTAG);
	return (B_FALSE);
	}
	if (error != 0 && error != EEXIST && error != ECHRNG) {
	fatal(B_FALSE, "ztest_snapshot_create(%s@%s) = %d", osname,
	snapname, error);
	}
	return (B_TRUE);
	}

	static boolean_t
	ztest_snapshot_destroy(char *osname, uint64_t id)
	{
	char snapname[ZFS_MAX_DATASET_NAME_LEN];
	int error;

	(void) snprintf(snapname, sizeof (snapname), "%s@%"PRIu64"",
	osname, id);

	error = dsl_destroy_snapshot(snapname, B_FALSE);
	if (error != 0 && error != ENOENT && error != ECHRNG)
	fatal(B_FALSE, "ztest_snapshot_destroy(%s) = %d",
	snapname, error);
	return (B_TRUE);
	}

	void
	ztest_dmu_objset_create_destroy(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd;
	ztest_ds_t *zdtmp;
	int iters;
	int error;
	objset_t os, os2;
	char name[ZFS_MAX_DATASET_NAME_LEN];
	zilog_t *zilog;
	int i;

	zdtmp = umem_alloc(sizeof (ztest_ds_t), UMEM_NOFAIL);

	(void) pthread_rwlock_rdlock(&ztest_name_lock);

	(void) snprintf(name, sizeof (name), "%s/temp_%"PRIu64"",
	ztest_opts.zo_pool, id);

	/*
	* If this dataset exists from a previous run, process its replay log
	* half of the time. If we don't replay it, then dsl_destroy_head()
	* (invoked from ztest_objset_destroy_cb()) should just throw it away.
	*/
	if (ztest_random(2) == 0 &&
	ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE,
	B_TRUE, FTAG, &os) == 0) {
	ztest_zd_init(zdtmp, NULL, os);
	zil_replay(os, zdtmp, ztest_replay_vector);
	ztest_zd_fini(zdtmp);
	dmu_objset_disown(os, B_TRUE, FTAG);
	}

	/*
	* There may be an old instance of the dataset we're about to
	* create lying around from a previous run. If so, destroy it
	* and all of its snapshots.
	*/
	(void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL,
	DS_FIND_CHILDREN \| DS_FIND_SNAPSHOTS);

	/*
	* Verify that the destroyed dataset is no longer in the namespace.
	* It may still be present if the destroy above fails with ENOSPC.
	*/
	error = ztest_dmu_objset_own(name, DMU_OST_OTHER, B_TRUE, B_TRUE,
	FTAG, &os);
	if (error == 0) {
	dmu_objset_disown(os, B_TRUE, FTAG);
	ztest_record_enospc(FTAG);
	goto out;
	}
	VERIFY3U(ENOENT, ==, error);

	/*
	* Verify that we can create a new dataset.
	*/
	error = ztest_dataset_create(name);
	if (error) {
	if (error == ENOSPC) {
	ztest_record_enospc(FTAG);
	goto out;
	}
	fatal(B_FALSE, "dmu_objset_create(%s) = %d", name, error);
	}

	VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, B_TRUE,
	FTAG, &os));

	ztest_zd_init(zdtmp, NULL, os);

	/*
	* Open the intent log for it.
	*/
	zilog = zil_open(os, ztest_get_data, NULL);

	/*
	* Put some objects in there, do a little I/O to them,
	* and randomly take a couple of snapshots along the way.
	*/
	iters = ztest_random(5);
	for (i = 0; i < iters; i++) {
	ztest_dmu_object_alloc_free(zdtmp, id);
	if (ztest_random(iters) == 0)
	(void) ztest_snapshot_create(name, i);
	}

	/*
	* Verify that we cannot create an existing dataset.
	*/
	VERIFY3U(EEXIST, ==,
	dmu_objset_create(name, DMU_OST_OTHER, 0, NULL, NULL, NULL));

	/*
	* Verify that we can hold an objset that is also owned.
	*/
	VERIFY0(dmu_objset_hold(name, FTAG, &os2));
	dmu_objset_rele(os2, FTAG);

	/*
	* Verify that we cannot own an objset that is already owned.
	*/
	VERIFY3U(EBUSY, ==, ztest_dmu_objset_own(name, DMU_OST_OTHER,
	B_FALSE, B_TRUE, FTAG, &os2));

	zil_close(zilog);
	dmu_objset_disown(os, B_TRUE, FTAG);
	ztest_zd_fini(zdtmp);
	out:
	(void) pthread_rwlock_unlock(&ztest_name_lock);

	umem_free(zdtmp, sizeof (ztest_ds_t));
	}

	/*
	* Verify that dmu_snapshot_{create,destroy,open,close} work as expected.
	*/
	void
	ztest_dmu_snapshot_create_destroy(ztest_ds_t *zd, uint64_t id)
	{
	(void) pthread_rwlock_rdlock(&ztest_name_lock);
	(void) ztest_snapshot_destroy(zd->zd_name, id);
	(void) ztest_snapshot_create(zd->zd_name, id);
	(void) pthread_rwlock_unlock(&ztest_name_lock);
	}

	/*
	* Cleanup non-standard snapshots and clones.
	*/
	static void
	ztest_dsl_dataset_cleanup(char *osname, uint64_t id)
	{
	char *snap1name;
	char *clone1name;
	char *snap2name;
	char *clone2name;
	char *snap3name;
	int error;

	snap1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
	clone1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
	snap2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
	clone2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
	snap3name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);

	(void) snprintf(snap1name, ZFS_MAX_DATASET_NAME_LEN, "%s@s1_%"PRIu64"",
	osname, id);
	(void) snprintf(clone1name, ZFS_MAX_DATASET_NAME_LEN, "%s/c1_%"PRIu64"",
	osname, id);
	(void) snprintf(snap2name, ZFS_MAX_DATASET_NAME_LEN, "%s@s2_%"PRIu64"",
	clone1name, id);
	(void) snprintf(clone2name, ZFS_MAX_DATASET_NAME_LEN, "%s/c2_%"PRIu64"",
	osname, id);
	(void) snprintf(snap3name, ZFS_MAX_DATASET_NAME_LEN, "%s@s3_%"PRIu64"",
	clone1name, id);

	error = dsl_destroy_head(clone2name);
	if (error && error != ENOENT)
	fatal(B_FALSE, "dsl_destroy_head(%s) = %d", clone2name, error);
	error = dsl_destroy_snapshot(snap3name, B_FALSE);
	if (error && error != ENOENT)
	fatal(B_FALSE, "dsl_destroy_snapshot(%s) = %d",
	snap3name, error);
	error = dsl_destroy_snapshot(snap2name, B_FALSE);
	if (error && error != ENOENT)
	fatal(B_FALSE, "dsl_destroy_snapshot(%s) = %d",
	snap2name, error);
	error = dsl_destroy_head(clone1name);
	if (error && error != ENOENT)
	fatal(B_FALSE, "dsl_destroy_head(%s) = %d", clone1name, error);
	error = dsl_destroy_snapshot(snap1name, B_FALSE);
	if (error && error != ENOENT)
	fatal(B_FALSE, "dsl_destroy_snapshot(%s) = %d",
	snap1name, error);

	umem_free(snap1name, ZFS_MAX_DATASET_NAME_LEN);
	umem_free(clone1name, ZFS_MAX_DATASET_NAME_LEN);
	umem_free(snap2name, ZFS_MAX_DATASET_NAME_LEN);
	umem_free(clone2name, ZFS_MAX_DATASET_NAME_LEN);
	umem_free(snap3name, ZFS_MAX_DATASET_NAME_LEN);
	}

	/*
	* Verify dsl_dataset_promote handles EBUSY
	*/
	void
	ztest_dsl_dataset_promote_busy(ztest_ds_t *zd, uint64_t id)
	{
	objset_t *os;
	char *snap1name;
	char *clone1name;
	char *snap2name;
	char *clone2name;
	char *snap3name;
	char *osname = zd->zd_name;
	int error;

	snap1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
	clone1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
	snap2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
	clone2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);
	snap3name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL);

	(void) pthread_rwlock_rdlock(&ztest_name_lock);

	ztest_dsl_dataset_cleanup(osname, id);

	(void) snprintf(snap1name, ZFS_MAX_DATASET_NAME_LEN, "%s@s1_%"PRIu64"",
	osname, id);
	(void) snprintf(clone1name, ZFS_MAX_DATASET_NAME_LEN, "%s/c1_%"PRIu64"",
	osname, id);
	(void) snprintf(snap2name, ZFS_MAX_DATASET_NAME_LEN, "%s@s2_%"PRIu64"",
	clone1name, id);
	(void) snprintf(clone2name, ZFS_MAX_DATASET_NAME_LEN, "%s/c2_%"PRIu64"",
	osname, id);
	(void) snprintf(snap3name, ZFS_MAX_DATASET_NAME_LEN, "%s@s3_%"PRIu64"",
	clone1name, id);

	error = dmu_objset_snapshot_one(osname, strchr(snap1name, '@') + 1);
	if (error && error != EEXIST) {
	if (error == ENOSPC) {
	ztest_record_enospc(FTAG);
	goto out;
	}
	fatal(B_FALSE, "dmu_take_snapshot(%s) = %d", snap1name, error);
	}

	error = dmu_objset_clone(clone1name, snap1name);
	if (error) {
	if (error == ENOSPC) {
	ztest_record_enospc(FTAG);
	goto out;
	}
	fatal(B_FALSE, "dmu_objset_create(%s) = %d", clone1name, error);
	}

	error = dmu_objset_snapshot_one(clone1name, strchr(snap2name, '@') + 1);
	if (error && error != EEXIST) {
	if (error == ENOSPC) {
	ztest_record_enospc(FTAG);
	goto out;
	}
	fatal(B_FALSE, "dmu_open_snapshot(%s) = %d", snap2name, error);
	}

	error = dmu_objset_snapshot_one(clone1name, strchr(snap3name, '@') + 1);
	if (error && error != EEXIST) {
	if (error == ENOSPC) {
	ztest_record_enospc(FTAG);
	goto out;
	}
	fatal(B_FALSE, "dmu_open_snapshot(%s) = %d", snap3name, error);
	}

	error = dmu_objset_clone(clone2name, snap3name);
	if (error) {
	if (error == ENOSPC) {
	ztest_record_enospc(FTAG);
	goto out;
	}
	fatal(B_FALSE, "dmu_objset_create(%s) = %d", clone2name, error);
	}

	error = ztest_dmu_objset_own(snap2name, DMU_OST_ANY, B_TRUE, B_TRUE,
	FTAG, &os);
	if (error)
	fatal(B_FALSE, "dmu_objset_own(%s) = %d", snap2name, error);
	error = dsl_dataset_promote(clone2name, NULL);
	if (error == ENOSPC) {
	dmu_objset_disown(os, B_TRUE, FTAG);
	ztest_record_enospc(FTAG);
	goto out;
	}
	if (error != EBUSY)
	fatal(B_FALSE, "dsl_dataset_promote(%s), %d, not EBUSY",
	clone2name, error);
	dmu_objset_disown(os, B_TRUE, FTAG);

	out:
	ztest_dsl_dataset_cleanup(osname, id);

	(void) pthread_rwlock_unlock(&ztest_name_lock);

	umem_free(snap1name, ZFS_MAX_DATASET_NAME_LEN);
	umem_free(clone1name, ZFS_MAX_DATASET_NAME_LEN);
	umem_free(snap2name, ZFS_MAX_DATASET_NAME_LEN);
	umem_free(clone2name, ZFS_MAX_DATASET_NAME_LEN);
	umem_free(snap3name, ZFS_MAX_DATASET_NAME_LEN);
	}

	#undef OD_ARRAY_SIZE
	#define OD_ARRAY_SIZE 4

	/*
	* Verify that dmu_object_{alloc,free} work as expected.
	*/
	void
	ztest_dmu_object_alloc_free(ztest_ds_t *zd, uint64_t id)
	{
	ztest_od_t *od;
	int batchsize;
	int size;
	int b;

	size = sizeof (ztest_od_t) * OD_ARRAY_SIZE;
	od = umem_alloc(size, UMEM_NOFAIL);
	batchsize = OD_ARRAY_SIZE;

	for (b = 0; b < batchsize; b++)
	ztest_od_init(od + b, id, FTAG, b, DMU_OT_UINT64_OTHER,
	0, 0, 0);

	/*
	* Destroy the previous batch of objects, create a new batch,
	* and do some I/O on the new objects.
	*/
	if (ztest_object_init(zd, od, size, B_TRUE) != 0) {
	zd->zd_od = NULL;
	umem_free(od, size);
	return;
	}

	while (ztest_random(4 * batchsize) != 0)
	ztest_io(zd, od[ztest_random(batchsize)].od_object,
	ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT);

	umem_free(od, size);
	}

	/*
	* Rewind the global allocator to verify object allocation backfilling.
	*/
	void
	ztest_dmu_object_next_chunk(ztest_ds_t *zd, uint64_t id)
	{
	(void) id;
	objset_t *os = zd->zd_os;
	uint_t dnodes_per_chunk = 1 << dmu_object_alloc_chunk_shift;
	uint64_t object;

	/*
	* Rewind the global allocator randomly back to a lower object number
	* to force backfilling and reclamation of recently freed dnodes.
	*/
	mutex_enter(&os->os_obj_lock);
	object = ztest_random(os->os_obj_next_chunk);
	- os->os_obj_next_chunk = P2ALIGN(object, dnodes_per_chunk);
	+ os->os_obj_next_chunk = P2ALIGN_TYPED(object, dnodes_per_chunk,
	+ uint64_t);
	mutex_exit(&os->os_obj_lock);
	}

	#undef OD_ARRAY_SIZE
	#define OD_ARRAY_SIZE 2

	/*
	* Verify that dmu_{read,write} work as expected.
	*/
	void
	ztest_dmu_read_write(ztest_ds_t *zd, uint64_t id)
	{
	int size;
	ztest_od_t *od;

	objset_t *os = zd->zd_os;
	size = sizeof (ztest_od_t) * OD_ARRAY_SIZE;
	od = umem_alloc(size, UMEM_NOFAIL);
	dmu_tx_t *tx;
	int freeit, error;
	uint64_t i, n, s, txg;
	bufwad_t packbuf, bigbuf, pack, bigH, *bigT;
	uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize;
	uint64_t chunksize = (1000 + ztest_random(1000)) * sizeof (uint64_t);
	uint64_t regions = 997;
	uint64_t stride = 123456789ULL;
	uint64_t width = 40;
	int free_percent = 5;

	/*
	* This test uses two objects, packobj and bigobj, that are always
	* updated together (i.e. in the same tx) so that their contents are
	* in sync and can be compared. Their contents relate to each other
	* in a simple way: packobj is a dense array of 'bufwad' structures,
	* while bigobj is a sparse array of the same bufwads. Specifically,
	* for any index n, there are three bufwads that should be identical:
	*
	* packobj, at offset n * sizeof (bufwad_t)
	* bigobj, at the head of the nth chunk
	* bigobj, at the tail of the nth chunk
	*
	* The chunk size is arbitrary. It doesn't have to be a power of two,
	* and it doesn't have any relation to the object blocksize.
	* The only requirement is that it can hold at least two bufwads.
	*
	* Normally, we write the bufwad to each of these locations.
	* However, free_percent of the time we instead write zeroes to
	* packobj and perform a dmu_free_range() on bigobj. By comparing
	* bigobj to packobj, we can verify that the DMU is correctly
	* tracking which parts of an object are allocated and free,
	* and that the contents of the allocated blocks are correct.
	*/

	/*
	* Read the directory info. If it's the first time, set things up.
	*/
	ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, chunksize);
	ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, 0,
	chunksize);

	if (ztest_object_init(zd, od, size, B_FALSE) != 0) {
	umem_free(od, size);
	return;
	}

	bigobj = od[0].od_object;
	packobj = od[1].od_object;
	chunksize = od[0].od_gen;
	ASSERT3U(chunksize, ==, od[1].od_gen);

	/*
	* Prefetch a random chunk of the big object.
	* Our aim here is to get some async reads in flight
	* for blocks that we may free below; the DMU should
	* handle this race correctly.
	*/
	n = ztest_random(regions) * stride + ztest_random(width);
	s = 1 + ztest_random(2 * width - 1);
	dmu_prefetch(os, bigobj, 0, n * chunksize, s * chunksize,
	ZIO_PRIORITY_SYNC_READ);

	/*
	* Pick a random index and compute the offsets into packobj and bigobj.
	*/
	n = ztest_random(regions) * stride + ztest_random(width);
	s = 1 + ztest_random(width - 1);

	packoff = n * sizeof (bufwad_t);
	packsize = s * sizeof (bufwad_t);

	bigoff = n * chunksize;
	bigsize = s * chunksize;

	packbuf = umem_alloc(packsize, UMEM_NOFAIL);
	bigbuf = umem_alloc(bigsize, UMEM_NOFAIL);

	/*
	* free_percent of the time, free a range of bigobj rather than
	* overwriting it.
	*/
	freeit = (ztest_random(100) < free_percent);

	/*
	* Read the current contents of our objects.
	*/
	error = dmu_read(os, packobj, packoff, packsize, packbuf,
	DMU_READ_PREFETCH);
	ASSERT0(error);
	error = dmu_read(os, bigobj, bigoff, bigsize, bigbuf,
	DMU_READ_PREFETCH);
	ASSERT0(error);

	/*
	* Get a tx for the mods to both packobj and bigobj.
	*/
	tx = dmu_tx_create(os);

	dmu_tx_hold_write(tx, packobj, packoff, packsize);

	if (freeit)
	dmu_tx_hold_free(tx, bigobj, bigoff, bigsize);
	else
	dmu_tx_hold_write(tx, bigobj, bigoff, bigsize);

	/* This accounts for setting the checksum/compression. */
	dmu_tx_hold_bonus(tx, bigobj);

	txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
	if (txg == 0) {
	umem_free(packbuf, packsize);
	umem_free(bigbuf, bigsize);
	umem_free(od, size);
	return;
	}

	enum zio_checksum cksum;
	do {
	cksum = (enum zio_checksum)
	ztest_random_dsl_prop(ZFS_PROP_CHECKSUM);
	} while (cksum >= ZIO_CHECKSUM_LEGACY_FUNCTIONS);
	dmu_object_set_checksum(os, bigobj, cksum, tx);

	enum zio_compress comp;
	do {
	comp = (enum zio_compress)
	ztest_random_dsl_prop(ZFS_PROP_COMPRESSION);
	} while (comp >= ZIO_COMPRESS_LEGACY_FUNCTIONS);
	dmu_object_set_compress(os, bigobj, comp, tx);

	/*
	* For each index from n to n + s, verify that the existing bufwad
	* in packobj matches the bufwads at the head and tail of the
	* corresponding chunk in bigobj. Then update all three bufwads
	* with the new values we want to write out.
	*/
	for (i = 0; i < s; i++) {
	/* LINTED */
	pack = (bufwad_t )((char )packbuf + i * sizeof (bufwad_t));
	/* LINTED */
	bigH = (bufwad_t )((char )bigbuf + i * chunksize);
	/* LINTED */
	bigT = (bufwad_t )((char )bigH + chunksize) - 1;

	ASSERT3U((uintptr_t)bigH - (uintptr_t)bigbuf, <, bigsize);
	ASSERT3U((uintptr_t)bigT - (uintptr_t)bigbuf, <, bigsize);

	if (pack->bw_txg > txg)
	fatal(B_FALSE,
	"future leak: got %"PRIx64", open txg is %"PRIx64"",
	pack->bw_txg, txg);

	if (pack->bw_data != 0 && pack->bw_index != n + i)
	fatal(B_FALSE, "wrong index: "
	"got %"PRIx64", wanted %"PRIx64"+%"PRIx64"",
	pack->bw_index, n, i);

	if (memcmp(pack, bigH, sizeof (bufwad_t)) != 0)
	fatal(B_FALSE, "pack/bigH mismatch in %p/%p",
	pack, bigH);

	if (memcmp(pack, bigT, sizeof (bufwad_t)) != 0)
	fatal(B_FALSE, "pack/bigT mismatch in %p/%p",
	pack, bigT);

	if (freeit) {
	memset(pack, 0, sizeof (bufwad_t));
	} else {
	pack->bw_index = n + i;
	pack->bw_txg = txg;
	pack->bw_data = 1 + ztest_random(-2ULL);
	}
	bigH = pack;
	bigT = pack;
	}

	/*
	* We've verified all the old bufwads, and made new ones.
	* Now write them out.
	*/
	dmu_write(os, packobj, packoff, packsize, packbuf, tx);

	if (freeit) {
	if (ztest_opts.zo_verbose >= 7) {
	(void) printf("freeing offset %"PRIx64" size %"PRIx64""
	" txg %"PRIx64"\n",
	bigoff, bigsize, txg);
	}
	VERIFY0(dmu_free_range(os, bigobj, bigoff, bigsize, tx));
	} else {
	if (ztest_opts.zo_verbose >= 7) {
	(void) printf("writing offset %"PRIx64" size %"PRIx64""
	" txg %"PRIx64"\n",
	bigoff, bigsize, txg);
	}
	dmu_write(os, bigobj, bigoff, bigsize, bigbuf, tx);
	}

	dmu_tx_commit(tx);

	/*
	* Sanity check the stuff we just wrote.
	*/
	{
	void *packcheck = umem_alloc(packsize, UMEM_NOFAIL);
	void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL);

	VERIFY0(dmu_read(os, packobj, packoff,
	packsize, packcheck, DMU_READ_PREFETCH));
	VERIFY0(dmu_read(os, bigobj, bigoff,
	bigsize, bigcheck, DMU_READ_PREFETCH));

	ASSERT0(memcmp(packbuf, packcheck, packsize));
	ASSERT0(memcmp(bigbuf, bigcheck, bigsize));

	umem_free(packcheck, packsize);
	umem_free(bigcheck, bigsize);
	}

	umem_free(packbuf, packsize);
	umem_free(bigbuf, bigsize);
	umem_free(od, size);
	}

	static void
	compare_and_update_pbbufs(uint64_t s, bufwad_t packbuf, bufwad_t bigbuf,
	uint64_t bigsize, uint64_t n, uint64_t chunksize, uint64_t txg)
	{
	uint64_t i;
	bufwad_t *pack;
	bufwad_t *bigH;
	bufwad_t *bigT;

	/*
	* For each index from n to n + s, verify that the existing bufwad
	* in packobj matches the bufwads at the head and tail of the
	* corresponding chunk in bigobj. Then update all three bufwads
	* with the new values we want to write out.
	*/
	for (i = 0; i < s; i++) {
	/* LINTED */
	pack = (bufwad_t )((char )packbuf + i * sizeof (bufwad_t));
	/* LINTED */
	bigH = (bufwad_t )((char )bigbuf + i * chunksize);
	/* LINTED */
	bigT = (bufwad_t )((char )bigH + chunksize) - 1;

	ASSERT3U((uintptr_t)bigH - (uintptr_t)bigbuf, <, bigsize);
	ASSERT3U((uintptr_t)bigT - (uintptr_t)bigbuf, <, bigsize);

	if (pack->bw_txg > txg)
	fatal(B_FALSE,
	"future leak: got %"PRIx64", open txg is %"PRIx64"",
	pack->bw_txg, txg);

	if (pack->bw_data != 0 && pack->bw_index != n + i)
	fatal(B_FALSE, "wrong index: "
	"got %"PRIx64", wanted %"PRIx64"+%"PRIx64"",
	pack->bw_index, n, i);

	if (memcmp(pack, bigH, sizeof (bufwad_t)) != 0)
	fatal(B_FALSE, "pack/bigH mismatch in %p/%p",
	pack, bigH);

	if (memcmp(pack, bigT, sizeof (bufwad_t)) != 0)
	fatal(B_FALSE, "pack/bigT mismatch in %p/%p",
	pack, bigT);

	pack->bw_index = n + i;
	pack->bw_txg = txg;
	pack->bw_data = 1 + ztest_random(-2ULL);

	bigH = pack;
	bigT = pack;
	}
	}

	#undef OD_ARRAY_SIZE
	#define OD_ARRAY_SIZE 2

	void
	ztest_dmu_read_write_zcopy(ztest_ds_t *zd, uint64_t id)
	{
	objset_t *os = zd->zd_os;
	ztest_od_t *od;
	dmu_tx_t *tx;
	uint64_t i;
	int error;
	int size;
	uint64_t n, s, txg;
	bufwad_t packbuf, bigbuf;
	uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize;
	uint64_t blocksize = ztest_random_blocksize();
	uint64_t chunksize = blocksize;
	uint64_t regions = 997;
	uint64_t stride = 123456789ULL;
	uint64_t width = 9;
	dmu_buf_t *bonus_db;
	arc_buf_t **bigbuf_arcbufs;
	dmu_object_info_t doi;

	size = sizeof (ztest_od_t) * OD_ARRAY_SIZE;
	od = umem_alloc(size, UMEM_NOFAIL);

	/*
	* This test uses two objects, packobj and bigobj, that are always
	* updated together (i.e. in the same tx) so that their contents are
	* in sync and can be compared. Their contents relate to each other
	* in a simple way: packobj is a dense array of 'bufwad' structures,
	* while bigobj is a sparse array of the same bufwads. Specifically,
	* for any index n, there are three bufwads that should be identical:
	*
	* packobj, at offset n * sizeof (bufwad_t)
	* bigobj, at the head of the nth chunk
	* bigobj, at the tail of the nth chunk
	*
	* The chunk size is set equal to bigobj block size so that
	* dmu_assign_arcbuf_by_dbuf() can be tested for object updates.
	*/

	/*
	* Read the directory info. If it's the first time, set things up.
	*/
	ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0);
	ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, 0,
	chunksize);


	if (ztest_object_init(zd, od, size, B_FALSE) != 0) {
	umem_free(od, size);
	return;
	}

	bigobj = od[0].od_object;
	packobj = od[1].od_object;
	blocksize = od[0].od_blocksize;
	chunksize = blocksize;
	ASSERT3U(chunksize, ==, od[1].od_gen);

	VERIFY0(dmu_object_info(os, bigobj, &doi));
	VERIFY(ISP2(doi.doi_data_block_size));
	VERIFY3U(chunksize, ==, doi.doi_data_block_size);
	VERIFY3U(chunksize, >=, 2 * sizeof (bufwad_t));

	/*
	* Pick a random index and compute the offsets into packobj and bigobj.
	*/
	n = ztest_random(regions) * stride + ztest_random(width);
	s = 1 + ztest_random(width - 1);

	packoff = n * sizeof (bufwad_t);
	packsize = s * sizeof (bufwad_t);

	bigoff = n * chunksize;
	bigsize = s * chunksize;

	packbuf = umem_zalloc(packsize, UMEM_NOFAIL);
	bigbuf = umem_zalloc(bigsize, UMEM_NOFAIL);

	VERIFY0(dmu_bonus_hold(os, bigobj, FTAG, &bonus_db));

	bigbuf_arcbufs = umem_zalloc(2 * s * sizeof (arc_buf_t *), UMEM_NOFAIL);

	/*
	* Iteration 0 test zcopy for DB_UNCACHED dbufs.
	* Iteration 1 test zcopy to already referenced dbufs.
	* Iteration 2 test zcopy to dirty dbuf in the same txg.
	* Iteration 3 test zcopy to dbuf dirty in previous txg.
	* Iteration 4 test zcopy when dbuf is no longer dirty.
	* Iteration 5 test zcopy when it can't be done.
	* Iteration 6 one more zcopy write.
	*/
	for (i = 0; i < 7; i++) {
	uint64_t j;
	uint64_t off;

	/*
	* In iteration 5 (i == 5) use arcbufs
	* that don't match bigobj blksz to test
	* dmu_assign_arcbuf_by_dbuf() when it can't directly
	* assign an arcbuf to a dbuf.
	*/
	for (j = 0; j < s; j++) {
	if (i != 5 \|\| chunksize < (SPA_MINBLOCKSIZE * 2)) {
	bigbuf_arcbufs[j] =
	dmu_request_arcbuf(bonus_db, chunksize);
	} else {
	bigbuf_arcbufs[2 * j] =
	dmu_request_arcbuf(bonus_db, chunksize / 2);
	bigbuf_arcbufs[2 * j + 1] =
	dmu_request_arcbuf(bonus_db, chunksize / 2);
	}
	}

	/*
	* Get a tx for the mods to both packobj and bigobj.
	*/
	tx = dmu_tx_create(os);

	dmu_tx_hold_write(tx, packobj, packoff, packsize);
	dmu_tx_hold_write(tx, bigobj, bigoff, bigsize);

	txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
	if (txg == 0) {
	umem_free(packbuf, packsize);
	umem_free(bigbuf, bigsize);
	for (j = 0; j < s; j++) {
	if (i != 5 \|\|
	chunksize < (SPA_MINBLOCKSIZE * 2)) {
	dmu_return_arcbuf(bigbuf_arcbufs[j]);
	} else {
	dmu_return_arcbuf(
	bigbuf_arcbufs[2 * j]);
	dmu_return_arcbuf(
	bigbuf_arcbufs[2 * j + 1]);
	}
	}
	umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *));
	umem_free(od, size);
	dmu_buf_rele(bonus_db, FTAG);
	return;
	}

	/*
	* 50% of the time don't read objects in the 1st iteration to
	* test dmu_assign_arcbuf_by_dbuf() for the case when there are
	* no existing dbufs for the specified offsets.
	*/
	if (i != 0 \|\| ztest_random(2) != 0) {
	error = dmu_read(os, packobj, packoff,
	packsize, packbuf, DMU_READ_PREFETCH);
	ASSERT0(error);
	error = dmu_read(os, bigobj, bigoff, bigsize,
	bigbuf, DMU_READ_PREFETCH);
	ASSERT0(error);
	}
	compare_and_update_pbbufs(s, packbuf, bigbuf, bigsize,
	n, chunksize, txg);

	/*
	* We've verified all the old bufwads, and made new ones.
	* Now write them out.
	*/
	dmu_write(os, packobj, packoff, packsize, packbuf, tx);
	if (ztest_opts.zo_verbose >= 7) {
	(void) printf("writing offset %"PRIx64" size %"PRIx64""
	" txg %"PRIx64"\n",
	bigoff, bigsize, txg);
	}
	for (off = bigoff, j = 0; j < s; j++, off += chunksize) {
	dmu_buf_t *dbt;
	if (i != 5 \|\| chunksize < (SPA_MINBLOCKSIZE * 2)) {
	memcpy(bigbuf_arcbufs[j]->b_data,
	(caddr_t)bigbuf + (off - bigoff),
	chunksize);
	} else {
	memcpy(bigbuf_arcbufs[2 * j]->b_data,
	(caddr_t)bigbuf + (off - bigoff),
	chunksize / 2);
	memcpy(bigbuf_arcbufs[2 * j + 1]->b_data,
	(caddr_t)bigbuf + (off - bigoff) +
	chunksize / 2,
	chunksize / 2);
	}

	if (i == 1) {
	VERIFY(dmu_buf_hold(os, bigobj, off,
	FTAG, &dbt, DMU_READ_NO_PREFETCH) == 0);
	}
	if (i != 5 \|\| chunksize < (SPA_MINBLOCKSIZE * 2)) {
	VERIFY0(dmu_assign_arcbuf_by_dbuf(bonus_db,
	off, bigbuf_arcbufs[j], tx));
	} else {
	VERIFY0(dmu_assign_arcbuf_by_dbuf(bonus_db,
	off, bigbuf_arcbufs[2 * j], tx));
	VERIFY0(dmu_assign_arcbuf_by_dbuf(bonus_db,
	off + chunksize / 2,
	bigbuf_arcbufs[2 * j + 1], tx));
	}
	if (i == 1) {
	dmu_buf_rele(dbt, FTAG);
	}
	}
	dmu_tx_commit(tx);

	/*
	* Sanity check the stuff we just wrote.
	*/
	{
	void *packcheck = umem_alloc(packsize, UMEM_NOFAIL);
	void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL);

	VERIFY0(dmu_read(os, packobj, packoff,
	packsize, packcheck, DMU_READ_PREFETCH));
	VERIFY0(dmu_read(os, bigobj, bigoff,
	bigsize, bigcheck, DMU_READ_PREFETCH));

	ASSERT0(memcmp(packbuf, packcheck, packsize));
	ASSERT0(memcmp(bigbuf, bigcheck, bigsize));

	umem_free(packcheck, packsize);
	umem_free(bigcheck, bigsize);
	}
	if (i == 2) {
	txg_wait_open(dmu_objset_pool(os), 0, B_TRUE);
	} else if (i == 3) {
	txg_wait_synced(dmu_objset_pool(os), 0);
	}
	}

	dmu_buf_rele(bonus_db, FTAG);
	umem_free(packbuf, packsize);
	umem_free(bigbuf, bigsize);
	umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *));
	umem_free(od, size);
	}

	void
	ztest_dmu_write_parallel(ztest_ds_t *zd, uint64_t id)
	{
	(void) id;
	ztest_od_t *od;

	od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL);
	uint64_t offset = (1ULL << (ztest_random(20) + 43)) +
	(ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT);

	/*
	* Have multiple threads write to large offsets in an object
	* to verify that parallel writes to an object -- even to the
	* same blocks within the object -- doesn't cause any trouble.
	*/
	ztest_od_init(od, ID_PARALLEL, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0);

	if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0)
	return;

	while (ztest_random(10) != 0)
	ztest_io(zd, od->od_object, offset);

	umem_free(od, sizeof (ztest_od_t));
	}

	void
	ztest_dmu_prealloc(ztest_ds_t *zd, uint64_t id)
	{
	ztest_od_t *od;
	uint64_t offset = (1ULL << (ztest_random(4) + SPA_MAXBLOCKSHIFT)) +
	(ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT);
	uint64_t count = ztest_random(20) + 1;
	uint64_t blocksize = ztest_random_blocksize();
	void *data;

	od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL);

	ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0);

	if (ztest_object_init(zd, od, sizeof (ztest_od_t),
	!ztest_random(2)) != 0) {
	umem_free(od, sizeof (ztest_od_t));
	return;
	}

	if (ztest_truncate(zd, od->od_object, offset, count * blocksize) != 0) {
	umem_free(od, sizeof (ztest_od_t));
	return;
	}

	ztest_prealloc(zd, od->od_object, offset, count * blocksize);

	data = umem_zalloc(blocksize, UMEM_NOFAIL);

	while (ztest_random(count) != 0) {
	uint64_t randoff = offset + (ztest_random(count) * blocksize);
	if (ztest_write(zd, od->od_object, randoff, blocksize,
	data) != 0)
	break;
	while (ztest_random(4) != 0)
	ztest_io(zd, od->od_object, randoff);
	}

	umem_free(data, blocksize);
	umem_free(od, sizeof (ztest_od_t));
	}

	/*
	* Verify that zap_{create,destroy,add,remove,update} work as expected.
	*/
	#define ZTEST_ZAP_MIN_INTS 1
	#define ZTEST_ZAP_MAX_INTS 4
	#define ZTEST_ZAP_MAX_PROPS 1000

	void
	ztest_zap(ztest_ds_t *zd, uint64_t id)
	{
	objset_t *os = zd->zd_os;
	ztest_od_t *od;
	uint64_t object;
	uint64_t txg, last_txg;
	uint64_t value[ZTEST_ZAP_MAX_INTS];
	uint64_t zl_ints, zl_intsize, prop;
	int i, ints;
	dmu_tx_t *tx;
	char propname[100], txgname[100];
	int error;
	const char *const hc[2] = { "s.acl.h", ".s.open.h.hyLZlg" };

	od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL);
	ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0, 0);

	if (ztest_object_init(zd, od, sizeof (ztest_od_t),
	!ztest_random(2)) != 0)
	goto out;

	object = od->od_object;

	/*
	* Generate a known hash collision, and verify that
	* we can lookup and remove both entries.
	*/
	tx = dmu_tx_create(os);
	dmu_tx_hold_zap(tx, object, B_TRUE, NULL);
	txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
	if (txg == 0)
	goto out;
	for (i = 0; i < 2; i++) {
	value[i] = i;
	VERIFY0(zap_add(os, object, hc[i], sizeof (uint64_t),
	1, &value[i], tx));
	}
	for (i = 0; i < 2; i++) {
	VERIFY3U(EEXIST, ==, zap_add(os, object, hc[i],
	sizeof (uint64_t), 1, &value[i], tx));
	VERIFY0(
	zap_length(os, object, hc[i], &zl_intsize, &zl_ints));
	ASSERT3U(zl_intsize, ==, sizeof (uint64_t));
	ASSERT3U(zl_ints, ==, 1);
	}
	for (i = 0; i < 2; i++) {
	VERIFY0(zap_remove(os, object, hc[i], tx));
	}
	dmu_tx_commit(tx);

	/*
	* Generate a bunch of random entries.
	*/
	ints = MAX(ZTEST_ZAP_MIN_INTS, object % ZTEST_ZAP_MAX_INTS);

	prop = ztest_random(ZTEST_ZAP_MAX_PROPS);
	(void) sprintf(propname, "prop_%"PRIu64"", prop);
	(void) sprintf(txgname, "txg_%"PRIu64"", prop);
	memset(value, 0, sizeof (value));
	last_txg = 0;

	/*
	* If these zap entries already exist, validate their contents.
	*/
	error = zap_length(os, object, txgname, &zl_intsize, &zl_ints);
	if (error == 0) {
	ASSERT3U(zl_intsize, ==, sizeof (uint64_t));
	ASSERT3U(zl_ints, ==, 1);

	VERIFY0(zap_lookup(os, object, txgname, zl_intsize,
	zl_ints, &last_txg));

	VERIFY0(zap_length(os, object, propname, &zl_intsize,
	&zl_ints));

	ASSERT3U(zl_intsize, ==, sizeof (uint64_t));
	ASSERT3U(zl_ints, ==, ints);

	VERIFY0(zap_lookup(os, object, propname, zl_intsize,
	zl_ints, value));

	for (i = 0; i < ints; i++) {
	ASSERT3U(value[i], ==, last_txg + object + i);
	}
	} else {
	ASSERT3U(error, ==, ENOENT);
	}

	/*
	* Atomically update two entries in our zap object.
	* The first is named txg_%llu, and contains the txg
	* in which the property was last updated. The second
	* is named prop_%llu, and the nth element of its value
	* should be txg + object + n.
	*/
	tx = dmu_tx_create(os);
	dmu_tx_hold_zap(tx, object, B_TRUE, NULL);
	txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
	if (txg == 0)
	goto out;

	if (last_txg > txg)
	fatal(B_FALSE, "zap future leak: old %"PRIu64" new %"PRIu64"",
	last_txg, txg);

	for (i = 0; i < ints; i++)
	value[i] = txg + object + i;

	VERIFY0(zap_update(os, object, txgname, sizeof (uint64_t),
	1, &txg, tx));
	VERIFY0(zap_update(os, object, propname, sizeof (uint64_t),
	ints, value, tx));

	dmu_tx_commit(tx);

	/*
	* Remove a random pair of entries.
	*/
	prop = ztest_random(ZTEST_ZAP_MAX_PROPS);
	(void) sprintf(propname, "prop_%"PRIu64"", prop);
	(void) sprintf(txgname, "txg_%"PRIu64"", prop);

	error = zap_length(os, object, txgname, &zl_intsize, &zl_ints);

	if (error == ENOENT)
	goto out;

	ASSERT0(error);

	tx = dmu_tx_create(os);
	dmu_tx_hold_zap(tx, object, B_TRUE, NULL);
	txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
	if (txg == 0)
	goto out;
	VERIFY0(zap_remove(os, object, txgname, tx));
	VERIFY0(zap_remove(os, object, propname, tx));
	dmu_tx_commit(tx);
	out:
	umem_free(od, sizeof (ztest_od_t));
	}

	/*
	* Test case to test the upgrading of a microzap to fatzap.
	*/
	void
	ztest_fzap(ztest_ds_t *zd, uint64_t id)
	{
	objset_t *os = zd->zd_os;
	ztest_od_t *od;
	uint64_t object, txg, value;

	od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL);
	ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0, 0);

	if (ztest_object_init(zd, od, sizeof (ztest_od_t),
	!ztest_random(2)) != 0)
	goto out;
	object = od->od_object;

	/*
	* Add entries to this ZAP and make sure it spills over
	* and gets upgraded to a fatzap. Also, since we are adding
	* 2050 entries we should see ptrtbl growth and leaf-block split.
	*/
	for (value = 0; value < 2050; value++) {
	char name[ZFS_MAX_DATASET_NAME_LEN];
	dmu_tx_t *tx;
	int error;

	(void) snprintf(name, sizeof (name), "fzap-%"PRIu64"-%"PRIu64"",
	id, value);

	tx = dmu_tx_create(os);
	dmu_tx_hold_zap(tx, object, B_TRUE, name);
	txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
	if (txg == 0)
	goto out;
	error = zap_add(os, object, name, sizeof (uint64_t), 1,
	&value, tx);
	ASSERT(error == 0 \|\| error == EEXIST);
	dmu_tx_commit(tx);
	}
	out:
	umem_free(od, sizeof (ztest_od_t));
	}

	void
	ztest_zap_parallel(ztest_ds_t *zd, uint64_t id)
	{
	(void) id;
	objset_t *os = zd->zd_os;
	ztest_od_t *od;
	uint64_t txg, object, count, wsize, wc, zl_wsize, zl_wc;
	dmu_tx_t *tx;
	int i, namelen, error;
	int micro = ztest_random(2);
	char name[20], string_value[20];
	void *data;

	od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL);
	ztest_od_init(od, ID_PARALLEL, FTAG, micro, DMU_OT_ZAP_OTHER, 0, 0, 0);

	if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) {
	umem_free(od, sizeof (ztest_od_t));
	return;
	}

	object = od->od_object;

	/*
	* Generate a random name of the form 'xxx.....' where each
	* x is a random printable character and the dots are dots.
	* There are 94 such characters, and the name length goes from
	* 6 to 20, so there are 94^3 * 15 = 12,458,760 possible names.
	*/
	namelen = ztest_random(sizeof (name) - 5) + 5 + 1;

	for (i = 0; i < 3; i++)
	name[i] = '!' + ztest_random('~' - '!' + 1);
	for (; i < namelen - 1; i++)
	name[i] = '.';
	name[i] = '\0';

	if ((namelen & 1) \|\| micro) {
	wsize = sizeof (txg);
	wc = 1;
	data = &txg;
	} else {
	wsize = 1;
	wc = namelen;
	data = string_value;
	}

	count = -1ULL;
	VERIFY0(zap_count(os, object, &count));
	ASSERT3S(count, !=, -1ULL);

	/*
	* Select an operation: length, lookup, add, update, remove.
	*/
	i = ztest_random(5);

	if (i >= 2) {
	tx = dmu_tx_create(os);
	dmu_tx_hold_zap(tx, object, B_TRUE, NULL);
	txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
	if (txg == 0) {
	umem_free(od, sizeof (ztest_od_t));
	return;
	}
	memcpy(string_value, name, namelen);
	} else {
	tx = NULL;
	txg = 0;
	memset(string_value, 0, namelen);
	}

	switch (i) {

	case 0:
	error = zap_length(os, object, name, &zl_wsize, &zl_wc);
	if (error == 0) {
	ASSERT3U(wsize, ==, zl_wsize);
	ASSERT3U(wc, ==, zl_wc);
	} else {
	ASSERT3U(error, ==, ENOENT);
	}
	break;

	case 1:
	error = zap_lookup(os, object, name, wsize, wc, data);
	if (error == 0) {
	if (data == string_value &&
	memcmp(name, data, namelen) != 0)
	fatal(B_FALSE, "name '%s' != val '%s' len %d",
	name, (char *)data, namelen);
	} else {
	ASSERT3U(error, ==, ENOENT);
	}
	break;

	case 2:
	error = zap_add(os, object, name, wsize, wc, data, tx);
	ASSERT(error == 0 \|\| error == EEXIST);
	break;

	case 3:
	VERIFY0(zap_update(os, object, name, wsize, wc, data, tx));
	break;

	case 4:
	error = zap_remove(os, object, name, tx);
	ASSERT(error == 0 \|\| error == ENOENT);
	break;
	}

	if (tx != NULL)
	dmu_tx_commit(tx);

	umem_free(od, sizeof (ztest_od_t));
	}

	/*
	* Commit callback data.
	*/
	typedef struct ztest_cb_data {
	list_node_t zcd_node;
	uint64_t zcd_txg;
	int zcd_expected_err;
	boolean_t zcd_added;
	boolean_t zcd_called;
	spa_t *zcd_spa;
	} ztest_cb_data_t;

	/* This is the actual commit callback function */
	static void
	ztest_commit_callback(void *arg, int error)
	{
	ztest_cb_data_t *data = arg;
	uint64_t synced_txg;

	VERIFY3P(data, !=, NULL);
	VERIFY3S(data->zcd_expected_err, ==, error);
	VERIFY(!data->zcd_called);

	synced_txg = spa_last_synced_txg(data->zcd_spa);
	if (data->zcd_txg > synced_txg)
	fatal(B_FALSE,
	"commit callback of txg %"PRIu64" called prematurely, "
	"last synced txg = %"PRIu64"\n",
	data->zcd_txg, synced_txg);

	data->zcd_called = B_TRUE;

	if (error == ECANCELED) {
	ASSERT0(data->zcd_txg);
	ASSERT(!data->zcd_added);

	/*
	* The private callback data should be destroyed here, but
	* since we are going to check the zcd_called field after
	* dmu_tx_abort(), we will destroy it there.
	*/
	return;
	}

	ASSERT(data->zcd_added);
	ASSERT3U(data->zcd_txg, !=, 0);

	(void) mutex_enter(&zcl.zcl_callbacks_lock);

	/* See if this cb was called more quickly */
	if ((synced_txg - data->zcd_txg) < zc_min_txg_delay)
	zc_min_txg_delay = synced_txg - data->zcd_txg;

	/* Remove our callback from the list */
	list_remove(&zcl.zcl_callbacks, data);

	(void) mutex_exit(&zcl.zcl_callbacks_lock);

	umem_free(data, sizeof (ztest_cb_data_t));
	}

	/* Allocate and initialize callback data structure */
	static ztest_cb_data_t *
	ztest_create_cb_data(objset_t *os, uint64_t txg)
	{
	ztest_cb_data_t *cb_data;

	cb_data = umem_zalloc(sizeof (ztest_cb_data_t), UMEM_NOFAIL);

	cb_data->zcd_txg = txg;
	cb_data->zcd_spa = dmu_objset_spa(os);
	list_link_init(&cb_data->zcd_node);

	return (cb_data);
	}

	/*
	* Commit callback test.
	*/
	void
	ztest_dmu_commit_callbacks(ztest_ds_t *zd, uint64_t id)
	{
	objset_t *os = zd->zd_os;
	ztest_od_t *od;
	dmu_tx_t *tx;
	ztest_cb_data_t cb_data[3], tmp_cb;
	uint64_t old_txg, txg;
	int i, error = 0;

	od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL);
	ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0);

	if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) {
	umem_free(od, sizeof (ztest_od_t));
	return;
	}

	tx = dmu_tx_create(os);

	cb_data[0] = ztest_create_cb_data(os, 0);
	dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[0]);

	dmu_tx_hold_write(tx, od->od_object, 0, sizeof (uint64_t));

	/* Every once in a while, abort the transaction on purpose */
	if (ztest_random(100) == 0)
	error = -1;

	if (!error)
	error = dmu_tx_assign(tx, TXG_NOWAIT);

	txg = error ? 0 : dmu_tx_get_txg(tx);

	cb_data[0]->zcd_txg = txg;
	cb_data[1] = ztest_create_cb_data(os, txg);
	dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[1]);

	if (error) {
	/*
	* It's not a strict requirement to call the registered
	* callbacks from inside dmu_tx_abort(), but that's what
	* it's supposed to happen in the current implementation
	* so we will check for that.
	*/
	for (i = 0; i < 2; i++) {
	cb_data[i]->zcd_expected_err = ECANCELED;
	VERIFY(!cb_data[i]->zcd_called);
	}

	dmu_tx_abort(tx);

	for (i = 0; i < 2; i++) {
	VERIFY(cb_data[i]->zcd_called);
	umem_free(cb_data[i], sizeof (ztest_cb_data_t));
	}

	umem_free(od, sizeof (ztest_od_t));
	return;
	}

	cb_data[2] = ztest_create_cb_data(os, txg);
	dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[2]);

	/*
	* Read existing data to make sure there isn't a future leak.
	*/
	VERIFY0(dmu_read(os, od->od_object, 0, sizeof (uint64_t),
	&old_txg, DMU_READ_PREFETCH));

	if (old_txg > txg)
	fatal(B_FALSE,
	"future leak: got %"PRIu64", open txg is %"PRIu64"",
	old_txg, txg);

	dmu_write(os, od->od_object, 0, sizeof (uint64_t), &txg, tx);

	(void) mutex_enter(&zcl.zcl_callbacks_lock);

	/*
	* Since commit callbacks don't have any ordering requirement and since
	* it is theoretically possible for a commit callback to be called
	* after an arbitrary amount of time has elapsed since its txg has been
	* synced, it is difficult to reliably determine whether a commit
	* callback hasn't been called due to high load or due to a flawed
	* implementation.
	*
	* In practice, we will assume that if after a certain number of txgs a
	* commit callback hasn't been called, then most likely there's an
	* implementation bug..
	*/
	tmp_cb = list_head(&zcl.zcl_callbacks);
	if (tmp_cb != NULL &&
	tmp_cb->zcd_txg + ZTEST_COMMIT_CB_THRESH < txg) {
	fatal(B_FALSE,
	"Commit callback threshold exceeded, "
	"oldest txg: %"PRIu64", open txg: %"PRIu64"\n",
	tmp_cb->zcd_txg, txg);
	}

	/*
	* Let's find the place to insert our callbacks.
	*
	* Even though the list is ordered by txg, it is possible for the
	* insertion point to not be the end because our txg may already be
	* quiescing at this point and other callbacks in the open txg
	* (from other objsets) may have sneaked in.
	*/
	tmp_cb = list_tail(&zcl.zcl_callbacks);
	while (tmp_cb != NULL && tmp_cb->zcd_txg > txg)
	tmp_cb = list_prev(&zcl.zcl_callbacks, tmp_cb);

	/* Add the 3 callbacks to the list */
	for (i = 0; i < 3; i++) {
	if (tmp_cb == NULL)
	list_insert_head(&zcl.zcl_callbacks, cb_data[i]);
	else
	list_insert_after(&zcl.zcl_callbacks, tmp_cb,
	cb_data[i]);

	cb_data[i]->zcd_added = B_TRUE;
	VERIFY(!cb_data[i]->zcd_called);

	tmp_cb = cb_data[i];
	}

	zc_cb_counter += 3;

	(void) mutex_exit(&zcl.zcl_callbacks_lock);

	dmu_tx_commit(tx);

	umem_free(od, sizeof (ztest_od_t));
	}

	/*
	* Visit each object in the dataset. Verify that its properties
	* are consistent what was stored in the block tag when it was created,
	* and that its unused bonus buffer space has not been overwritten.
	*/
	void
	ztest_verify_dnode_bt(ztest_ds_t *zd, uint64_t id)
	{
	(void) id;
	objset_t *os = zd->zd_os;
	uint64_t obj;
	int err = 0;

	for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) {
	ztest_block_tag_t *bt = NULL;
	dmu_object_info_t doi;
	dmu_buf_t *db;

	ztest_object_lock(zd, obj, ZTRL_READER);
	if (dmu_bonus_hold(os, obj, FTAG, &db) != 0) {
	ztest_object_unlock(zd, obj);
	continue;
	}

	dmu_object_info_from_db(db, &doi);
	if (doi.doi_bonus_size >= sizeof (*bt))
	bt = ztest_bt_bonus(db);

	if (bt && bt->bt_magic == BT_MAGIC) {
	ztest_bt_verify(bt, os, obj, doi.doi_dnodesize,
	bt->bt_offset, bt->bt_gen, bt->bt_txg,
	bt->bt_crtxg);
	ztest_verify_unused_bonus(db, bt, obj, os, bt->bt_gen);
	}

	dmu_buf_rele(db, FTAG);
	ztest_object_unlock(zd, obj);
	}
	}

	void
	ztest_dsl_prop_get_set(ztest_ds_t *zd, uint64_t id)
	{
	(void) id;
	zfs_prop_t proplist[] = {
	ZFS_PROP_CHECKSUM,
	ZFS_PROP_COMPRESSION,
	ZFS_PROP_COPIES,
	ZFS_PROP_DEDUP
	};

	(void) pthread_rwlock_rdlock(&ztest_name_lock);

	for (int p = 0; p < sizeof (proplist) / sizeof (proplist[0]); p++) {
	int error = ztest_dsl_prop_set_uint64(zd->zd_name, proplist[p],
	ztest_random_dsl_prop(proplist[p]), (int)ztest_random(2));
	ASSERT(error == 0 \|\| error == ENOSPC);
	}

	int error = ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_RECORDSIZE,
	ztest_random_blocksize(), (int)ztest_random(2));
	ASSERT(error == 0 \|\| error == ENOSPC);

	(void) pthread_rwlock_unlock(&ztest_name_lock);
	}

	void
	ztest_spa_prop_get_set(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	nvlist_t *props = NULL;

	(void) pthread_rwlock_rdlock(&ztest_name_lock);

	(void) ztest_spa_prop_set_uint64(ZPOOL_PROP_AUTOTRIM, ztest_random(2));

	VERIFY0(spa_prop_get(ztest_spa, &props));

	if (ztest_opts.zo_verbose >= 6)
	dump_nvlist(props, 4);

	fnvlist_free(props);

	(void) pthread_rwlock_unlock(&ztest_name_lock);
	}

	static int
	user_release_one(const char snapname, const char holdname)
	{
	nvlist_t snaps, holds;
	int error;

	snaps = fnvlist_alloc();
	holds = fnvlist_alloc();
	fnvlist_add_boolean(holds, holdname);
	fnvlist_add_nvlist(snaps, snapname, holds);
	fnvlist_free(holds);
	error = dsl_dataset_user_release(snaps, NULL);
	fnvlist_free(snaps);
	return (error);
	}

	/*
	* Test snapshot hold/release and deferred destroy.
	*/
	void
	ztest_dmu_snapshot_hold(ztest_ds_t *zd, uint64_t id)
	{
	int error;
	objset_t *os = zd->zd_os;
	objset_t *origin;
	char snapname[100];
	char fullname[100];
	char clonename[100];
	char tag[100];
	char osname[ZFS_MAX_DATASET_NAME_LEN];
	nvlist_t *holds;

	(void) pthread_rwlock_rdlock(&ztest_name_lock);

	dmu_objset_name(os, osname);

	(void) snprintf(snapname, sizeof (snapname), "sh1_%"PRIu64"", id);
	(void) snprintf(fullname, sizeof (fullname), "%s@%s", osname, snapname);
	(void) snprintf(clonename, sizeof (clonename), "%s/ch1_%"PRIu64"",
	osname, id);
	(void) snprintf(tag, sizeof (tag), "tag_%"PRIu64"", id);

	/*
	* Clean up from any previous run.
	*/
	error = dsl_destroy_head(clonename);
	if (error != ENOENT)
	ASSERT0(error);
	error = user_release_one(fullname, tag);
	if (error != ESRCH && error != ENOENT)
	ASSERT0(error);
	error = dsl_destroy_snapshot(fullname, B_FALSE);
	if (error != ENOENT)
	ASSERT0(error);

	/*
	* Create snapshot, clone it, mark snap for deferred destroy,
	* destroy clone, verify snap was also destroyed.
	*/
	error = dmu_objset_snapshot_one(osname, snapname);
	if (error) {
	if (error == ENOSPC) {
	ztest_record_enospc("dmu_objset_snapshot");
	goto out;
	}
	fatal(B_FALSE, "dmu_objset_snapshot(%s) = %d", fullname, error);
	}

	error = dmu_objset_clone(clonename, fullname);
	if (error) {
	if (error == ENOSPC) {
	ztest_record_enospc("dmu_objset_clone");
	goto out;
	}
	fatal(B_FALSE, "dmu_objset_clone(%s) = %d", clonename, error);
	}

	error = dsl_destroy_snapshot(fullname, B_TRUE);
	if (error) {
	fatal(B_FALSE, "dsl_destroy_snapshot(%s, B_TRUE) = %d",
	fullname, error);
	}

	error = dsl_destroy_head(clonename);
	if (error)
	fatal(B_FALSE, "dsl_destroy_head(%s) = %d", clonename, error);

	error = dmu_objset_hold(fullname, FTAG, &origin);
	if (error != ENOENT)
	fatal(B_FALSE, "dmu_objset_hold(%s) = %d", fullname, error);

	/*
	* Create snapshot, add temporary hold, verify that we can't
	* destroy a held snapshot, mark for deferred destroy,
	* release hold, verify snapshot was destroyed.
	*/
	error = dmu_objset_snapshot_one(osname, snapname);
	if (error) {
	if (error == ENOSPC) {
	ztest_record_enospc("dmu_objset_snapshot");
	goto out;
	}
	fatal(B_FALSE, "dmu_objset_snapshot(%s) = %d", fullname, error);
	}

	holds = fnvlist_alloc();
	fnvlist_add_string(holds, fullname, tag);
	error = dsl_dataset_user_hold(holds, 0, NULL);
	fnvlist_free(holds);

	if (error == ENOSPC) {
	ztest_record_enospc("dsl_dataset_user_hold");
	goto out;
	} else if (error) {
	fatal(B_FALSE, "dsl_dataset_user_hold(%s, %s) = %u",
	fullname, tag, error);
	}

	error = dsl_destroy_snapshot(fullname, B_FALSE);
	if (error != EBUSY) {
	fatal(B_FALSE, "dsl_destroy_snapshot(%s, B_FALSE) = %d",
	fullname, error);
	}

	error = dsl_destroy_snapshot(fullname, B_TRUE);
	if (error) {
	fatal(B_FALSE, "dsl_destroy_snapshot(%s, B_TRUE) = %d",
	fullname, error);
	}

	error = user_release_one(fullname, tag);
	if (error)
	fatal(B_FALSE, "user_release_one(%s, %s) = %d",
	fullname, tag, error);

	VERIFY3U(dmu_objset_hold(fullname, FTAG, &origin), ==, ENOENT);

	out:
	(void) pthread_rwlock_unlock(&ztest_name_lock);
	}

	/*
	* Inject random faults into the on-disk data.
	*/
	void
	ztest_fault_inject(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	ztest_shared_t *zs = ztest_shared;
	spa_t *spa = ztest_spa;
	int fd;
	uint64_t offset;
	uint64_t leaves;
	uint64_t bad = 0x1990c0ffeedecadeull;
	uint64_t top, leaf;
	uint64_t raidz_children;
	char *path0;
	char *pathrand;
	size_t fsize;
	int bshift = SPA_MAXBLOCKSHIFT + 2;
	int iters = 1000;
	int maxfaults;
	int mirror_save;
	vdev_t *vd0 = NULL;
	uint64_t guid0 = 0;
	boolean_t islog = B_FALSE;
	boolean_t injected = B_FALSE;

	path0 = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
	pathrand = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);

	mutex_enter(&ztest_vdev_lock);

	/*
	* Device removal is in progress, fault injection must be disabled
	* until it completes and the pool is scrubbed. The fault injection
	* strategy for damaging blocks does not take in to account evacuated
	* blocks which may have already been damaged.
	*/
	if (ztest_device_removal_active)
	goto out;

	/*
	* The fault injection strategy for damaging blocks cannot be used
	* if raidz expansion is in progress. The leaves value
	* (attached raidz children) is variable and strategy for damaging
	* blocks will corrupt same data blocks on different child vdevs
	* because of the reflow process.
	*/
	if (spa->spa_raidz_expand != NULL)
	goto out;

	maxfaults = MAXFAULTS(zs);
	raidz_children = ztest_get_raidz_children(spa);
	leaves = MAX(zs->zs_mirrors, 1) * raidz_children;
	mirror_save = zs->zs_mirrors;

	ASSERT3U(leaves, >=, 1);

	/*
	* While ztest is running the number of leaves will not change. This
	* is critical for the fault injection logic as it determines where
	* errors can be safely injected such that they are always repairable.
	*
	* When restarting ztest a different number of leaves may be requested
	* which will shift the regions to be damaged. This is fine as long
	* as the pool has been scrubbed prior to using the new mapping.
	* Failure to do can result in non-repairable damage being injected.
	*/
	if (ztest_pool_scrubbed == B_FALSE)
	goto out;

	/*
	* Grab the name lock as reader. There are some operations
	* which don't like to have their vdevs changed while
	* they are in progress (i.e. spa_change_guid). Those
	* operations will have grabbed the name lock as writer.
	*/
	(void) pthread_rwlock_rdlock(&ztest_name_lock);

	/*
	* We need SCL_STATE here because we're going to look at vd0->vdev_tsd.
	*/
	spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);

	if (ztest_random(2) == 0) {
	/*
	* Inject errors on a normal data device or slog device.
	*/
	top = ztest_random_vdev_top(spa, B_TRUE);
	leaf = ztest_random(leaves) + zs->zs_splits;

	/*
	* Generate paths to the first leaf in this top-level vdev,
	* and to the random leaf we selected. We'll induce transient
	* write failures and random online/offline activity on leaf 0,
	* and we'll write random garbage to the randomly chosen leaf.
	*/
	(void) snprintf(path0, MAXPATHLEN, ztest_dev_template,
	ztest_opts.zo_dir, ztest_opts.zo_pool,
	top * leaves + zs->zs_splits);
	(void) snprintf(pathrand, MAXPATHLEN, ztest_dev_template,
	ztest_opts.zo_dir, ztest_opts.zo_pool,
	top * leaves + leaf);

	vd0 = vdev_lookup_by_path(spa->spa_root_vdev, path0);
	if (vd0 != NULL && vd0->vdev_top->vdev_islog)
	islog = B_TRUE;

	/*
	* If the top-level vdev needs to be resilvered
	* then we only allow faults on the device that is
	* resilvering.
	*/
	if (vd0 != NULL && maxfaults != 1 &&
	(!vdev_resilver_needed(vd0->vdev_top, NULL, NULL) \|\|
	vd0->vdev_resilver_txg != 0)) {
	/*
	* Make vd0 explicitly claim to be unreadable,
	* or unwritable, or reach behind its back
	* and close the underlying fd. We can do this if
	* maxfaults == 0 because we'll fail and reexecute,
	* and we can do it if maxfaults >= 2 because we'll
	* have enough redundancy. If maxfaults == 1, the
	* combination of this with injection of random data
	* corruption below exceeds the pool's fault tolerance.
	*/
	vdev_file_t *vf = vd0->vdev_tsd;

	zfs_dbgmsg("injecting fault to vdev %llu; maxfaults=%d",
	(long long)vd0->vdev_id, (int)maxfaults);

	if (vf != NULL && ztest_random(3) == 0) {
	(void) close(vf->vf_file->f_fd);
	vf->vf_file->f_fd = -1;
	} else if (ztest_random(2) == 0) {
	vd0->vdev_cant_read = B_TRUE;
	} else {
	vd0->vdev_cant_write = B_TRUE;
	}
	guid0 = vd0->vdev_guid;
	}
	} else {
	/*
	* Inject errors on an l2cache device.
	*/
	spa_aux_vdev_t *sav = &spa->spa_l2cache;

	if (sav->sav_count == 0) {
	spa_config_exit(spa, SCL_STATE, FTAG);
	(void) pthread_rwlock_unlock(&ztest_name_lock);
	goto out;
	}
	vd0 = sav->sav_vdevs[ztest_random(sav->sav_count)];
	guid0 = vd0->vdev_guid;
	(void) strlcpy(path0, vd0->vdev_path, MAXPATHLEN);
	(void) strlcpy(pathrand, vd0->vdev_path, MAXPATHLEN);

	leaf = 0;
	leaves = 1;
	maxfaults = INT_MAX; /* no limit on cache devices */
	}

	spa_config_exit(spa, SCL_STATE, FTAG);
	(void) pthread_rwlock_unlock(&ztest_name_lock);

	/*
	* If we can tolerate two or more faults, or we're dealing
	* with a slog, randomly online/offline vd0.
	*/
	if ((maxfaults >= 2 \|\| islog) && guid0 != 0) {
	if (ztest_random(10) < 6) {
	int flags = (ztest_random(2) == 0 ?
	ZFS_OFFLINE_TEMPORARY : 0);

	/*
	* We have to grab the zs_name_lock as writer to
	* prevent a race between offlining a slog and
	* destroying a dataset. Offlining the slog will
	* grab a reference on the dataset which may cause
	* dsl_destroy_head() to fail with EBUSY thus
	* leaving the dataset in an inconsistent state.
	*/
	if (islog)
	(void) pthread_rwlock_wrlock(&ztest_name_lock);

	VERIFY3U(vdev_offline(spa, guid0, flags), !=, EBUSY);

	if (islog)
	(void) pthread_rwlock_unlock(&ztest_name_lock);
	} else {
	/*
	* Ideally we would like to be able to randomly
	* call vdev_[on\|off]line without holding locks
	* to force unpredictable failures but the side
	* effects of vdev_[on\|off]line prevent us from
	* doing so.
	*/
	(void) vdev_online(spa, guid0, 0, NULL);
	}
	}

	if (maxfaults == 0)
	goto out;

	/*
	* We have at least single-fault tolerance, so inject data corruption.
	*/
	fd = open(pathrand, O_RDWR);

	if (fd == -1) /* we hit a gap in the device namespace */
	goto out;

	fsize = lseek(fd, 0, SEEK_END);

	while (--iters != 0) {
	/*
	* The offset must be chosen carefully to ensure that
	* we do not inject a given logical block with errors
	* on two different leaf devices, because ZFS can not
	* tolerate that (if maxfaults==1).
	*
	* To achieve this we divide each leaf device into
	* chunks of size (# leaves * SPA_MAXBLOCKSIZE * 4).
	* Each chunk is further divided into error-injection
	* ranges (can accept errors) and clear ranges (we do
	* not inject errors in those). Each error-injection
	* range can accept errors only for a single leaf vdev.
	* Error-injection ranges are separated by clear ranges.
	*
	* For example, with 3 leaves, each chunk looks like:
	* 0 to 32M: injection range for leaf 0
	* 32M to 64M: clear range - no injection allowed
	* 64M to 96M: injection range for leaf 1
	* 96M to 128M: clear range - no injection allowed
	* 128M to 160M: injection range for leaf 2
	* 160M to 192M: clear range - no injection allowed
	*
	* Each clear range must be large enough such that a
	* single block cannot straddle it. This way a block
	* can't be a target in two different injection ranges
	* (on different leaf vdevs).
	*/
	offset = ztest_random(fsize / (leaves << bshift)) *
	(leaves << bshift) + (leaf << bshift) +
	(ztest_random(1ULL << (bshift - 1)) & -8ULL);

	/*
	* Only allow damage to the labels at one end of the vdev.
	*
	* If all labels are damaged, the device will be totally
	* inaccessible, which will result in loss of data,
	* because we also damage (parts of) the other side of
	* the mirror/raidz.
	*
	* Additionally, we will always have both an even and an
	* odd label, so that we can handle crashes in the
	* middle of vdev_config_sync().
	*/
	if ((leaf & 1) == 0 && offset < VDEV_LABEL_START_SIZE)
	continue;

	/*
	* The two end labels are stored at the "end" of the disk, but
	* the end of the disk (vdev_psize) is aligned to
	* sizeof (vdev_label_t).
	*/
	- uint64_t psize = P2ALIGN(fsize, sizeof (vdev_label_t));
	+ uint64_t psize = P2ALIGN_TYPED(fsize, sizeof (vdev_label_t),
	+ uint64_t);
	if ((leaf & 1) == 1 &&
	offset + sizeof (bad) > psize - VDEV_LABEL_END_SIZE)
	continue;

	if (mirror_save != zs->zs_mirrors) {
	(void) close(fd);
	goto out;
	}

	if (pwrite(fd, &bad, sizeof (bad), offset) != sizeof (bad))
	fatal(B_TRUE,
	"can't inject bad word at 0x%"PRIx64" in %s",
	offset, pathrand);

	if (ztest_opts.zo_verbose >= 7)
	(void) printf("injected bad word into %s,"
	" offset 0x%"PRIx64"\n", pathrand, offset);

	injected = B_TRUE;
	}

	(void) close(fd);
	out:
	mutex_exit(&ztest_vdev_lock);

	if (injected && ztest_opts.zo_raid_do_expand) {
	int error = spa_scan(spa, POOL_SCAN_SCRUB);
	if (error == 0) {
	while (dsl_scan_scrubbing(spa_get_dsl(spa)))
	txg_wait_synced(spa_get_dsl(spa), 0);
	}
	}

	umem_free(path0, MAXPATHLEN);
	umem_free(pathrand, MAXPATHLEN);
	}

	/*
	* By design ztest will never inject uncorrectable damage in to the pool.
	* Issue a scrub, wait for it to complete, and verify there is never any
	* persistent damage.
	*
	* Only after a full scrub has been completed is it safe to start injecting
	* data corruption. See the comment in zfs_fault_inject().
	*/
	static int
	ztest_scrub_impl(spa_t *spa)
	{
	int error = spa_scan(spa, POOL_SCAN_SCRUB);
	if (error)
	return (error);

	while (dsl_scan_scrubbing(spa_get_dsl(spa)))
	txg_wait_synced(spa_get_dsl(spa), 0);

	if (spa_approx_errlog_size(spa) > 0)
	return (ECKSUM);

	ztest_pool_scrubbed = B_TRUE;

	return (0);
	}

	/*
	* Scrub the pool.
	*/
	void
	ztest_scrub(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	spa_t *spa = ztest_spa;
	int error;

	/*
	* Scrub in progress by device removal.
	*/
	if (ztest_device_removal_active)
	return;

	/*
	* Start a scrub, wait a moment, then force a restart.
	*/
	(void) spa_scan(spa, POOL_SCAN_SCRUB);
	(void) poll(NULL, 0, 100);

	error = ztest_scrub_impl(spa);
	if (error == EBUSY)
	error = 0;
	ASSERT0(error);
	}

	/*
	* Change the guid for the pool.
	*/
	void
	ztest_reguid(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	spa_t *spa = ztest_spa;
	uint64_t orig, load;
	int error;
	ztest_shared_t *zs = ztest_shared;

	if (ztest_opts.zo_mmp_test)
	return;

	orig = spa_guid(spa);
	load = spa_load_guid(spa);

	(void) pthread_rwlock_wrlock(&ztest_name_lock);
	error = spa_change_guid(spa);
	zs->zs_guid = spa_guid(spa);
	(void) pthread_rwlock_unlock(&ztest_name_lock);

	if (error != 0)
	return;

	if (ztest_opts.zo_verbose >= 4) {
	(void) printf("Changed guid old %"PRIu64" -> %"PRIu64"\n",
	orig, spa_guid(spa));
	}

	VERIFY3U(orig, !=, spa_guid(spa));
	VERIFY3U(load, ==, spa_load_guid(spa));
	}

	void
	ztest_blake3(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	hrtime_t end = gethrtime() + NANOSEC;
	zio_cksum_salt_t salt;
	void *salt_ptr = &salt.zcs_bytes;
	struct abd abd_data, abd_meta;
	void buf, templ;
	int i, *ptr;
	uint32_t size;
	BLAKE3_CTX ctx;
	const zfs_impl_t *blake3 = zfs_impl_get_ops("blake3");

	size = ztest_random_blocksize();
	buf = umem_alloc(size, UMEM_NOFAIL);
	abd_data = abd_alloc(size, B_FALSE);
	abd_meta = abd_alloc(size, B_TRUE);

	for (i = 0, ptr = buf; i < size / sizeof (*ptr); i++, ptr++)
	*ptr = ztest_random(UINT_MAX);
	memset(salt_ptr, 'A', 32);

	abd_copy_from_buf_off(abd_data, buf, 0, size);
	abd_copy_from_buf_off(abd_meta, buf, 0, size);

	while (gethrtime() <= end) {
	int run_count = 100;
	zio_cksum_t zc_ref1, zc_ref2;
	zio_cksum_t zc_res1, zc_res2;

	void *ref1 = &zc_ref1;
	void *ref2 = &zc_ref2;
	void *res1 = &zc_res1;
	void *res2 = &zc_res2;

	/* BLAKE3_KEY_LEN = 32 */
	VERIFY0(blake3->setname("generic"));
	templ = abd_checksum_blake3_tmpl_init(&salt);
	Blake3_InitKeyed(&ctx, salt_ptr);
	Blake3_Update(&ctx, buf, size);
	Blake3_Final(&ctx, ref1);
	zc_ref2 = zc_ref1;
	ZIO_CHECKSUM_BSWAP(&zc_ref2);
	abd_checksum_blake3_tmpl_free(templ);

	VERIFY0(blake3->setname("cycle"));
	while (run_count-- > 0) {

	/* Test current implementation */
	Blake3_InitKeyed(&ctx, salt_ptr);
	Blake3_Update(&ctx, buf, size);
	Blake3_Final(&ctx, res1);
	zc_res2 = zc_res1;
	ZIO_CHECKSUM_BSWAP(&zc_res2);

	VERIFY0(memcmp(ref1, res1, 32));
	VERIFY0(memcmp(ref2, res2, 32));

	/* Test ABD - data */
	templ = abd_checksum_blake3_tmpl_init(&salt);
	abd_checksum_blake3_native(abd_data, size,
	templ, &zc_res1);
	abd_checksum_blake3_byteswap(abd_data, size,
	templ, &zc_res2);

	VERIFY0(memcmp(ref1, res1, 32));
	VERIFY0(memcmp(ref2, res2, 32));

	/* Test ABD - metadata */
	abd_checksum_blake3_native(abd_meta, size,
	templ, &zc_res1);
	abd_checksum_blake3_byteswap(abd_meta, size,
	templ, &zc_res2);
	abd_checksum_blake3_tmpl_free(templ);

	VERIFY0(memcmp(ref1, res1, 32));
	VERIFY0(memcmp(ref2, res2, 32));

	}
	}

	abd_free(abd_data);
	abd_free(abd_meta);
	umem_free(buf, size);
	}

	void
	ztest_fletcher(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	hrtime_t end = gethrtime() + NANOSEC;

	while (gethrtime() <= end) {
	int run_count = 100;
	void *buf;
	struct abd abd_data, abd_meta;
	uint32_t size;
	int *ptr;
	int i;
	zio_cksum_t zc_ref;
	zio_cksum_t zc_ref_byteswap;

	size = ztest_random_blocksize();

	buf = umem_alloc(size, UMEM_NOFAIL);
	abd_data = abd_alloc(size, B_FALSE);
	abd_meta = abd_alloc(size, B_TRUE);

	for (i = 0, ptr = buf; i < size / sizeof (*ptr); i++, ptr++)
	*ptr = ztest_random(UINT_MAX);

	abd_copy_from_buf_off(abd_data, buf, 0, size);
	abd_copy_from_buf_off(abd_meta, buf, 0, size);

	VERIFY0(fletcher_4_impl_set("scalar"));
	fletcher_4_native(buf, size, NULL, &zc_ref);
	fletcher_4_byteswap(buf, size, NULL, &zc_ref_byteswap);

	VERIFY0(fletcher_4_impl_set("cycle"));
	while (run_count-- > 0) {
	zio_cksum_t zc;
	zio_cksum_t zc_byteswap;

	fletcher_4_byteswap(buf, size, NULL, &zc_byteswap);
	fletcher_4_native(buf, size, NULL, &zc);

	VERIFY0(memcmp(&zc, &zc_ref, sizeof (zc)));
	VERIFY0(memcmp(&zc_byteswap, &zc_ref_byteswap,
	sizeof (zc_byteswap)));

	/* Test ABD - data */
	abd_fletcher_4_byteswap(abd_data, size, NULL,
	&zc_byteswap);
	abd_fletcher_4_native(abd_data, size, NULL, &zc);

	VERIFY0(memcmp(&zc, &zc_ref, sizeof (zc)));
	VERIFY0(memcmp(&zc_byteswap, &zc_ref_byteswap,
	sizeof (zc_byteswap)));

	/* Test ABD - metadata */
	abd_fletcher_4_byteswap(abd_meta, size, NULL,
	&zc_byteswap);
	abd_fletcher_4_native(abd_meta, size, NULL, &zc);

	VERIFY0(memcmp(&zc, &zc_ref, sizeof (zc)));
	VERIFY0(memcmp(&zc_byteswap, &zc_ref_byteswap,
	sizeof (zc_byteswap)));

	}

	umem_free(buf, size);
	abd_free(abd_data);
	abd_free(abd_meta);
	}
	}

	void
	ztest_fletcher_incr(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	void *buf;
	size_t size;
	int *ptr;
	int i;
	zio_cksum_t zc_ref;
	zio_cksum_t zc_ref_bswap;

	hrtime_t end = gethrtime() + NANOSEC;

	while (gethrtime() <= end) {
	int run_count = 100;

	size = ztest_random_blocksize();
	buf = umem_alloc(size, UMEM_NOFAIL);

	for (i = 0, ptr = buf; i < size / sizeof (*ptr); i++, ptr++)
	*ptr = ztest_random(UINT_MAX);

	VERIFY0(fletcher_4_impl_set("scalar"));
	fletcher_4_native(buf, size, NULL, &zc_ref);
	fletcher_4_byteswap(buf, size, NULL, &zc_ref_bswap);

	VERIFY0(fletcher_4_impl_set("cycle"));

	while (run_count-- > 0) {
	zio_cksum_t zc;
	zio_cksum_t zc_bswap;
	size_t pos = 0;

	ZIO_SET_CHECKSUM(&zc, 0, 0, 0, 0);
	ZIO_SET_CHECKSUM(&zc_bswap, 0, 0, 0, 0);

	while (pos < size) {
	size_t inc = 64 * ztest_random(size / 67);
	/* sometimes add few bytes to test non-simd */
	if (ztest_random(100) < 10)
	- inc += P2ALIGN(ztest_random(64),
	- sizeof (uint32_t));
	+ inc += P2ALIGN_TYPED(ztest_random(64),
	+ sizeof (uint32_t), uint64_t);

	if (inc > (size - pos))
	inc = size - pos;

	fletcher_4_incremental_native(buf + pos, inc,
	&zc);
	fletcher_4_incremental_byteswap(buf + pos, inc,
	&zc_bswap);

	pos += inc;
	}

	VERIFY3U(pos, ==, size);

	VERIFY(ZIO_CHECKSUM_EQUAL(zc, zc_ref));
	VERIFY(ZIO_CHECKSUM_EQUAL(zc_bswap, zc_ref_bswap));

	/*
	* verify if incremental on the whole buffer is
	* equivalent to non-incremental version
	*/
	ZIO_SET_CHECKSUM(&zc, 0, 0, 0, 0);
	ZIO_SET_CHECKSUM(&zc_bswap, 0, 0, 0, 0);

	fletcher_4_incremental_native(buf, size, &zc);
	fletcher_4_incremental_byteswap(buf, size, &zc_bswap);

	VERIFY(ZIO_CHECKSUM_EQUAL(zc, zc_ref));
	VERIFY(ZIO_CHECKSUM_EQUAL(zc_bswap, zc_ref_bswap));
	}

	umem_free(buf, size);
	}
	}

	static int
	ztest_set_global_vars(void)
	{
	for (size_t i = 0; i < ztest_opts.zo_gvars_count; i++) {
	char *kv = ztest_opts.zo_gvars[i];
	VERIFY3U(strlen(kv), <=, ZO_GVARS_MAX_ARGLEN);
	VERIFY3U(strlen(kv), >, 0);
	int err = set_global_var(kv);
	if (ztest_opts.zo_verbose > 0) {
	(void) printf("setting global var %s ... %s\n", kv,
	err ? "failed" : "ok");
	}
	if (err != 0) {
	(void) fprintf(stderr,
	"failed to set global var '%s'\n", kv);
	return (err);
	}
	}
	return (0);
	}

	static char **
	ztest_global_vars_to_zdb_args(void)
	{
	char *args = calloc(2ztest_opts.zo_gvars_count + 1, sizeof (char *));
	char **cur = args;
	if (args == NULL)
	return (NULL);
	for (size_t i = 0; i < ztest_opts.zo_gvars_count; i++) {
	cur++ = (char )"-o";
	*cur++ = ztest_opts.zo_gvars[i];
	}
	ASSERT3P(cur, ==, &args[2*ztest_opts.zo_gvars_count]);
	*cur = NULL;
	return (args);
	}

	/* The end of strings is indicated by a NULL element */
	static char *
	join_strings(char *strings, const char sep)
	{
	size_t totallen = 0;
	for (char *sp = strings; sp != NULL; sp++) {
	totallen += strlen(*sp);
	totallen += strlen(sep);
	}
	if (totallen > 0) {
	ASSERT(totallen >= strlen(sep));
	totallen -= strlen(sep);
	}

	size_t buflen = totallen + 1;
	char o = umem_alloc(buflen, UMEM_NOFAIL); / trailing 0 byte */
	o[0] = '\0';
	for (char *sp = strings; sp != NULL; sp++) {
	size_t would;
	would = strlcat(o, *sp, buflen);
	VERIFY3U(would, <, buflen);
	if (*(sp+1) == NULL) {
	break;
	}
	would = strlcat(o, sep, buflen);
	VERIFY3U(would, <, buflen);
	}
	ASSERT3S(strlen(o), ==, totallen);
	return (o);
	}

	static int
	ztest_check_path(char *path)
	{
	struct stat s;
	/* return true on success */
	return (!stat(path, &s));
	}

	static void
	ztest_get_zdb_bin(char *bin, int len)
	{
	char *zdb_path;
	/*
	* Try to use $ZDB and in-tree zdb path. If not successful, just
	* let popen to search through PATH.
	*/
	if ((zdb_path = getenv("ZDB"))) {
	strlcpy(bin, zdb_path, len); /* In env */
	if (!ztest_check_path(bin)) {
	ztest_dump_core = 0;
	fatal(B_TRUE, "invalid ZDB '%s'", bin);
	}
	return;
	}

	VERIFY3P(realpath(getexecname(), bin), !=, NULL);
	if (strstr(bin, ".libs/ztest")) {
	strstr(bin, ".libs/ztest")[0] = '\0'; /* In-tree */
	strcat(bin, "zdb");
	if (ztest_check_path(bin))
	return;
	}
	strcpy(bin, "zdb");
	}

	static vdev_t *
	ztest_random_concrete_vdev_leaf(vdev_t *vd)
	{
	if (vd == NULL)
	return (NULL);

	if (vd->vdev_children == 0)
	return (vd);

	vdev_t *eligible[vd->vdev_children];
	int eligible_idx = 0, i;
	for (i = 0; i < vd->vdev_children; i++) {
	vdev_t *cvd = vd->vdev_child[i];
	if (cvd->vdev_top->vdev_removing)
	continue;
	if (cvd->vdev_children > 0 \|\|
	(vdev_is_concrete(cvd) && !cvd->vdev_detached)) {
	eligible[eligible_idx++] = cvd;
	}
	}
	VERIFY3S(eligible_idx, >, 0);

	uint64_t child_no = ztest_random(eligible_idx);
	return (ztest_random_concrete_vdev_leaf(eligible[child_no]));
	}

	void
	ztest_initialize(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	spa_t *spa = ztest_spa;
	int error = 0;

	mutex_enter(&ztest_vdev_lock);

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);

	/* Random leaf vdev */
	vdev_t *rand_vd = ztest_random_concrete_vdev_leaf(spa->spa_root_vdev);
	if (rand_vd == NULL) {
	spa_config_exit(spa, SCL_VDEV, FTAG);
	mutex_exit(&ztest_vdev_lock);
	return;
	}

	/*
	* The random vdev we've selected may change as soon as we
	* drop the spa_config_lock. We create local copies of things
	* we're interested in.
	*/
	uint64_t guid = rand_vd->vdev_guid;
	char *path = strdup(rand_vd->vdev_path);
	boolean_t active = rand_vd->vdev_initialize_thread != NULL;

	zfs_dbgmsg("vd %px, guid %llu", rand_vd, (u_longlong_t)guid);
	spa_config_exit(spa, SCL_VDEV, FTAG);

	uint64_t cmd = ztest_random(POOL_INITIALIZE_FUNCS);

	nvlist_t *vdev_guids = fnvlist_alloc();
	nvlist_t *vdev_errlist = fnvlist_alloc();
	fnvlist_add_uint64(vdev_guids, path, guid);
	error = spa_vdev_initialize(spa, vdev_guids, cmd, vdev_errlist);
	fnvlist_free(vdev_guids);
	fnvlist_free(vdev_errlist);

	switch (cmd) {
	case POOL_INITIALIZE_CANCEL:
	if (ztest_opts.zo_verbose >= 4) {
	(void) printf("Cancel initialize %s", path);
	if (!active)
	(void) printf(" failed (no initialize active)");
	(void) printf("\n");
	}
	break;
	case POOL_INITIALIZE_START:
	if (ztest_opts.zo_verbose >= 4) {
	(void) printf("Start initialize %s", path);
	if (active && error == 0)
	(void) printf(" failed (already active)");
	else if (error != 0)
	(void) printf(" failed (error %d)", error);
	(void) printf("\n");
	}
	break;
	case POOL_INITIALIZE_SUSPEND:
	if (ztest_opts.zo_verbose >= 4) {
	(void) printf("Suspend initialize %s", path);
	if (!active)
	(void) printf(" failed (no initialize active)");
	(void) printf("\n");
	}
	break;
	}
	free(path);
	mutex_exit(&ztest_vdev_lock);
	}

	void
	ztest_trim(ztest_ds_t *zd, uint64_t id)
	{
	(void) zd, (void) id;
	spa_t *spa = ztest_spa;
	int error = 0;

	mutex_enter(&ztest_vdev_lock);

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);

	/* Random leaf vdev */
	vdev_t *rand_vd = ztest_random_concrete_vdev_leaf(spa->spa_root_vdev);
	if (rand_vd == NULL) {
	spa_config_exit(spa, SCL_VDEV, FTAG);
	mutex_exit(&ztest_vdev_lock);
	return;
	}

	/*
	* The random vdev we've selected may change as soon as we
	* drop the spa_config_lock. We create local copies of things
	* we're interested in.
	*/
	uint64_t guid = rand_vd->vdev_guid;
	char *path = strdup(rand_vd->vdev_path);
	boolean_t active = rand_vd->vdev_trim_thread != NULL;

	zfs_dbgmsg("vd %p, guid %llu", rand_vd, (u_longlong_t)guid);
	spa_config_exit(spa, SCL_VDEV, FTAG);

	uint64_t cmd = ztest_random(POOL_TRIM_FUNCS);
	uint64_t rate = 1 << ztest_random(30);
	boolean_t partial = (ztest_random(5) > 0);
	boolean_t secure = (ztest_random(5) > 0);

	nvlist_t *vdev_guids = fnvlist_alloc();
	nvlist_t *vdev_errlist = fnvlist_alloc();
	fnvlist_add_uint64(vdev_guids, path, guid);
	error = spa_vdev_trim(spa, vdev_guids, cmd, rate, partial,
	secure, vdev_errlist);
	fnvlist_free(vdev_guids);
	fnvlist_free(vdev_errlist);

	switch (cmd) {
	case POOL_TRIM_CANCEL:
	if (ztest_opts.zo_verbose >= 4) {
	(void) printf("Cancel TRIM %s", path);
	if (!active)
	(void) printf(" failed (no TRIM active)");
	(void) printf("\n");
	}
	break;
	case POOL_TRIM_START:
	if (ztest_opts.zo_verbose >= 4) {
	(void) printf("Start TRIM %s", path);
	if (active && error == 0)
	(void) printf(" failed (already active)");
	else if (error != 0)
	(void) printf(" failed (error %d)", error);
	(void) printf("\n");
	}
	break;
	case POOL_TRIM_SUSPEND:
	if (ztest_opts.zo_verbose >= 4) {
	(void) printf("Suspend TRIM %s", path);
	if (!active)
	(void) printf(" failed (no TRIM active)");
	(void) printf("\n");
	}
	break;
	}
	free(path);
	mutex_exit(&ztest_vdev_lock);
	}

	/*
	* Verify pool integrity by running zdb.
	*/
	static void
	ztest_run_zdb(uint64_t guid)
	{
	int status;
	char *bin;
	char *zdb;
	char *zbuf;
	const int len = MAXPATHLEN + MAXNAMELEN + 20;
	FILE *fp;

	bin = umem_alloc(len, UMEM_NOFAIL);
	zdb = umem_alloc(len, UMEM_NOFAIL);
	zbuf = umem_alloc(1024, UMEM_NOFAIL);

	ztest_get_zdb_bin(bin, len);

	char **set_gvars_args = ztest_global_vars_to_zdb_args();
	if (set_gvars_args == NULL) {
	fatal(B_FALSE, "Failed to allocate memory in "
	"ztest_global_vars_to_zdb_args(). Cannot run zdb.\n");
	}
	char *set_gvars_args_joined = join_strings(set_gvars_args, " ");
	free(set_gvars_args);

	size_t would = snprintf(zdb, len,
	"%s -bcc%s%s -G -d -Y -e -y %s -p %s %"PRIu64,
	bin,
	ztest_opts.zo_verbose >= 3 ? "s" : "",
	ztest_opts.zo_verbose >= 4 ? "v" : "",
	set_gvars_args_joined,
	ztest_opts.zo_dir,
	guid);
	ASSERT3U(would, <, len);

	umem_free(set_gvars_args_joined, strlen(set_gvars_args_joined) + 1);

	if (ztest_opts.zo_verbose >= 5)
	(void) printf("Executing %s\n", zdb);

	fp = popen(zdb, "r");

	while (fgets(zbuf, 1024, fp) != NULL)
	if (ztest_opts.zo_verbose >= 3)
	(void) printf("%s", zbuf);

	status = pclose(fp);

	if (status == 0)
	goto out;

	ztest_dump_core = 0;
	if (WIFEXITED(status))
	fatal(B_FALSE, "'%s' exit code %d", zdb, WEXITSTATUS(status));
	else
	fatal(B_FALSE, "'%s' died with signal %d",
	zdb, WTERMSIG(status));
	out:
	umem_free(bin, len);
	umem_free(zdb, len);
	umem_free(zbuf, 1024);
	}

	static void
	ztest_walk_pool_directory(const char *header)
	{
	spa_t *spa = NULL;

	if (ztest_opts.zo_verbose >= 6)
	(void) puts(header);

	mutex_enter(&spa_namespace_lock);
	while ((spa = spa_next(spa)) != NULL)
	if (ztest_opts.zo_verbose >= 6)
	(void) printf("\t%s\n", spa_name(spa));
	mutex_exit(&spa_namespace_lock);
	}

	static void
	ztest_spa_import_export(char oldname, char newname)
	{
	nvlist_t config, newconfig;
	uint64_t pool_guid;
	spa_t *spa;
	int error;

	if (ztest_opts.zo_verbose >= 4) {
	(void) printf("import/export: old = %s, new = %s\n",
	oldname, newname);
	}

	/*
	* Clean up from previous runs.
	*/
	(void) spa_destroy(newname);

	/*
	* Get the pool's configuration and guid.
	*/
	VERIFY0(spa_open(oldname, &spa, FTAG));

	/*
	* Kick off a scrub to tickle scrub/export races.
	*/
	if (ztest_random(2) == 0)
	(void) spa_scan(spa, POOL_SCAN_SCRUB);

	pool_guid = spa_guid(spa);
	spa_close(spa, FTAG);

	ztest_walk_pool_directory("pools before export");

	/*
	* Export it.
	*/
	VERIFY0(spa_export(oldname, &config, B_FALSE, B_FALSE));

	ztest_walk_pool_directory("pools after export");

	/*
	* Try to import it.
	*/
	newconfig = spa_tryimport(config);
	ASSERT3P(newconfig, !=, NULL);
	fnvlist_free(newconfig);

	/*
	* Import it under the new name.
	*/
	error = spa_import(newname, config, NULL, 0);
	if (error != 0) {
	dump_nvlist(config, 0);
	fatal(B_FALSE, "couldn't import pool %s as %s: error %u",
	oldname, newname, error);
	}

	ztest_walk_pool_directory("pools after import");

	/*
	* Try to import it again -- should fail with EEXIST.
	*/
	VERIFY3U(EEXIST, ==, spa_import(newname, config, NULL, 0));

	/*
	* Try to import it under a different name -- should fail with EEXIST.
	*/
	VERIFY3U(EEXIST, ==, spa_import(oldname, config, NULL, 0));

	/*
	* Verify that the pool is no longer visible under the old name.
	*/
	VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG));

	/*
	* Verify that we can open and close the pool using the new name.
	*/
	VERIFY0(spa_open(newname, &spa, FTAG));
	ASSERT3U(pool_guid, ==, spa_guid(spa));
	spa_close(spa, FTAG);

	fnvlist_free(config);
	}

	static void
	ztest_resume(spa_t *spa)
	{
	if (spa_suspended(spa) && ztest_opts.zo_verbose >= 6)
	(void) printf("resuming from suspended state\n");
	spa_vdev_state_enter(spa, SCL_NONE);
	vdev_clear(spa, NULL);
	(void) spa_vdev_state_exit(spa, NULL, 0);
	(void) zio_resume(spa);
	}

	static __attribute__((noreturn)) void
	ztest_resume_thread(void *arg)
	{
	spa_t *spa = arg;

	while (!ztest_exiting) {
	if (spa_suspended(spa))
	ztest_resume(spa);
	(void) poll(NULL, 0, 100);

	/*
	* Periodically change the zfs_compressed_arc_enabled setting.
	*/
	if (ztest_random(10) == 0)
	zfs_compressed_arc_enabled = ztest_random(2);

	/*
	* Periodically change the zfs_abd_scatter_enabled setting.
	*/
	if (ztest_random(10) == 0)
	zfs_abd_scatter_enabled = ztest_random(2);
	}

	thread_exit();
	}

	static __attribute__((noreturn)) void
	ztest_deadman_thread(void *arg)
	{
	ztest_shared_t *zs = arg;
	spa_t *spa = ztest_spa;
	hrtime_t delay, overdue, last_run = gethrtime();

	delay = (zs->zs_thread_stop - zs->zs_thread_start) +
	MSEC2NSEC(zfs_deadman_synctime_ms);

	while (!ztest_exiting) {
	/*
	* Wait for the delay timer while checking occasionally
	* if we should stop.
	*/
	if (gethrtime() < last_run + delay) {
	(void) poll(NULL, 0, 1000);
	continue;
	}

	/*
	* If the pool is suspended then fail immediately. Otherwise,
	* check to see if the pool is making any progress. If
	* vdev_deadman() discovers that there hasn't been any recent
	* I/Os then it will end up aborting the tests.
	*/
	if (spa_suspended(spa) \|\| spa->spa_root_vdev == NULL) {
	fatal(B_FALSE,
	"aborting test after %llu seconds because "
	"pool has transitioned to a suspended state.",
	(u_longlong_t)zfs_deadman_synctime_ms / 1000);
	}
	vdev_deadman(spa->spa_root_vdev, FTAG);

	/*
	* If the process doesn't complete within a grace period of
	* zfs_deadman_synctime_ms over the expected finish time,
	* then it may be hung and is terminated.
	*/
	overdue = zs->zs_proc_stop + MSEC2NSEC(zfs_deadman_synctime_ms);
	if (gethrtime() > overdue) {
	fatal(B_FALSE,
	"aborting test after %llu seconds because "
	"the process is overdue for termination.",
	(gethrtime() - zs->zs_proc_start) / NANOSEC);
	}

	(void) printf("ztest has been running for %lld seconds\n",
	(gethrtime() - zs->zs_proc_start) / NANOSEC);

	last_run = gethrtime();
	delay = MSEC2NSEC(zfs_deadman_checktime_ms);
	}

	thread_exit();
	}

	static void
	ztest_execute(int test, ztest_info_t *zi, uint64_t id)
	{
	ztest_ds_t *zd = &ztest_ds[id % ztest_opts.zo_datasets];
	ztest_shared_callstate_t *zc = ZTEST_GET_SHARED_CALLSTATE(test);
	hrtime_t functime = gethrtime();
	int i;

	for (i = 0; i < zi->zi_iters; i++)
	zi->zi_func(zd, id);

	functime = gethrtime() - functime;

	atomic_add_64(&zc->zc_count, 1);
	atomic_add_64(&zc->zc_time, functime);

	if (ztest_opts.zo_verbose >= 4)
	(void) printf("%6.2f sec in %s\n",
	(double)functime / NANOSEC, zi->zi_funcname);
	}

	typedef struct ztest_raidz_expand_io {
	uint64_t rzx_id;
	uint64_t rzx_amount;
	uint64_t rzx_bufsize;
	const void *rzx_buffer;
	uint64_t rzx_alloc_max;
	spa_t *rzx_spa;
	} ztest_expand_io_t;

	#undef OD_ARRAY_SIZE
	#define OD_ARRAY_SIZE 10

	/*
	* Write a request amount of data to some dataset objects.
	* There will be ztest_opts.zo_threads count of these running in parallel.
	*/
	static __attribute__((noreturn)) void
	ztest_rzx_thread(void *arg)
	{
	ztest_expand_io_t info = (ztest_expand_io_t )arg;
	ztest_od_t *od;
	int batchsize;
	int od_size;
	ztest_ds_t *zd = &ztest_ds[info->rzx_id % ztest_opts.zo_datasets];
	spa_t *spa = info->rzx_spa;

	od_size = sizeof (ztest_od_t) * OD_ARRAY_SIZE;
	od = umem_alloc(od_size, UMEM_NOFAIL);
	batchsize = OD_ARRAY_SIZE;

	/* Create objects to write to */
	for (int b = 0; b < batchsize; b++) {
	ztest_od_init(od + b, info->rzx_id, FTAG, b,
	DMU_OT_UINT64_OTHER, 0, 0, 0);
	}
	if (ztest_object_init(zd, od, od_size, B_FALSE) != 0) {
	umem_free(od, od_size);
	thread_exit();
	}

	for (uint64_t offset = 0, written = 0; written < info->rzx_amount;
	offset += info->rzx_bufsize) {
	/* write to 10 objects */
	for (int i = 0; i < batchsize && written < info->rzx_amount;
	i++) {
	(void) pthread_rwlock_rdlock(&zd->zd_zilog_lock);
	ztest_write(zd, od[i].od_object, offset,
	info->rzx_bufsize, info->rzx_buffer);
	(void) pthread_rwlock_unlock(&zd->zd_zilog_lock);
	written += info->rzx_bufsize;
	}
	txg_wait_synced(spa_get_dsl(spa), 0);
	/* due to inflation, we'll typically bail here */
	if (metaslab_class_get_alloc(spa_normal_class(spa)) >
	info->rzx_alloc_max) {
	break;
	}
	}

	/* Remove a few objects to leave some holes in allocation space */
	mutex_enter(&zd->zd_dirobj_lock);
	(void) ztest_remove(zd, od, 2);
	mutex_exit(&zd->zd_dirobj_lock);

	umem_free(od, od_size);

	thread_exit();
	}

	static __attribute__((noreturn)) void
	ztest_thread(void *arg)
	{
	int rand;
	uint64_t id = (uintptr_t)arg;
	ztest_shared_t *zs = ztest_shared;
	uint64_t call_next;
	hrtime_t now;
	ztest_info_t *zi;
	ztest_shared_callstate_t *zc;

	while ((now = gethrtime()) < zs->zs_thread_stop) {
	/*
	* See if it's time to force a crash.
	*/
	if (now > zs->zs_thread_kill &&
	raidz_expand_pause_point == RAIDZ_EXPAND_PAUSE_NONE) {
	ztest_kill(zs);
	}

	/*
	* If we're getting ENOSPC with some regularity, stop.
	*/
	if (zs->zs_enospc_count > 10)
	break;

	/*
	* Pick a random function to execute.
	*/
	rand = ztest_random(ZTEST_FUNCS);
	zi = &ztest_info[rand];
	zc = ZTEST_GET_SHARED_CALLSTATE(rand);
	call_next = zc->zc_next;

	if (now >= call_next &&
	atomic_cas_64(&zc->zc_next, call_next, call_next +
	ztest_random(2 * zi->zi_interval[0] + 1)) == call_next) {
	ztest_execute(rand, zi, id);
	}
	}

	thread_exit();
	}

	static void
	ztest_dataset_name(char dsname, const char pool, int d)
	{
	(void) snprintf(dsname, ZFS_MAX_DATASET_NAME_LEN, "%s/ds_%d", pool, d);
	}

	static void
	ztest_dataset_destroy(int d)
	{
	char name[ZFS_MAX_DATASET_NAME_LEN];
	int t;

	ztest_dataset_name(name, ztest_opts.zo_pool, d);

	if (ztest_opts.zo_verbose >= 3)
	(void) printf("Destroying %s to free up space\n", name);

	/*
	* Cleanup any non-standard clones and snapshots. In general,
	* ztest thread t operates on dataset (t % zopt_datasets),
	* so there may be more than one thing to clean up.
	*/
	for (t = d; t < ztest_opts.zo_threads;
	t += ztest_opts.zo_datasets)
	ztest_dsl_dataset_cleanup(name, t);

	(void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL,
	DS_FIND_SNAPSHOTS \| DS_FIND_CHILDREN);
	}

	static void
	ztest_dataset_dirobj_verify(ztest_ds_t *zd)
	{
	uint64_t usedobjs, dirobjs, scratch;

	/*
	* ZTEST_DIROBJ is the object directory for the entire dataset.
	* Therefore, the number of objects in use should equal the
	* number of ZTEST_DIROBJ entries, +1 for ZTEST_DIROBJ itself.
	* If not, we have an object leak.
	*
	* Note that we can only check this in ztest_dataset_open(),
	* when the open-context and syncing-context values agree.
	* That's because zap_count() returns the open-context value,
	* while dmu_objset_space() returns the rootbp fill count.
	*/
	VERIFY0(zap_count(zd->zd_os, ZTEST_DIROBJ, &dirobjs));
	dmu_objset_space(zd->zd_os, &scratch, &scratch, &usedobjs, &scratch);
	ASSERT3U(dirobjs + 1, ==, usedobjs);
	}

	static int
	ztest_dataset_open(int d)
	{
	ztest_ds_t *zd = &ztest_ds[d];
	uint64_t committed_seq = ZTEST_GET_SHARED_DS(d)->zd_seq;
	objset_t *os;
	zilog_t *zilog;
	char name[ZFS_MAX_DATASET_NAME_LEN];
	int error;

	ztest_dataset_name(name, ztest_opts.zo_pool, d);

	(void) pthread_rwlock_rdlock(&ztest_name_lock);

	error = ztest_dataset_create(name);
	if (error == ENOSPC) {
	(void) pthread_rwlock_unlock(&ztest_name_lock);
	ztest_record_enospc(FTAG);
	return (error);
	}
	ASSERT(error == 0 \|\| error == EEXIST);

	VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE,
	B_TRUE, zd, &os));
	(void) pthread_rwlock_unlock(&ztest_name_lock);

	ztest_zd_init(zd, ZTEST_GET_SHARED_DS(d), os);

	zilog = zd->zd_zilog;

	if (zilog->zl_header->zh_claim_lr_seq != 0 &&
	zilog->zl_header->zh_claim_lr_seq < committed_seq)
	fatal(B_FALSE, "missing log records: "
	"claimed %"PRIu64" < committed %"PRIu64"",
	zilog->zl_header->zh_claim_lr_seq, committed_seq);

	ztest_dataset_dirobj_verify(zd);

	zil_replay(os, zd, ztest_replay_vector);

	ztest_dataset_dirobj_verify(zd);

	if (ztest_opts.zo_verbose >= 6)
	(void) printf("%s replay %"PRIu64" blocks, "
	"%"PRIu64" records, seq %"PRIu64"\n",
	zd->zd_name,
	zilog->zl_parse_blk_count,
	zilog->zl_parse_lr_count,
	zilog->zl_replaying_seq);

	zilog = zil_open(os, ztest_get_data, NULL);

	if (zilog->zl_replaying_seq != 0 &&
	zilog->zl_replaying_seq < committed_seq)
	fatal(B_FALSE, "missing log records: "
	"replayed %"PRIu64" < committed %"PRIu64"",
	zilog->zl_replaying_seq, committed_seq);

	return (0);
	}

	static void
	ztest_dataset_close(int d)
	{
	ztest_ds_t *zd = &ztest_ds[d];

	zil_close(zd->zd_zilog);
	dmu_objset_disown(zd->zd_os, B_TRUE, zd);

	ztest_zd_fini(zd);
	}

	static int
	ztest_replay_zil_cb(const char name, void arg)
	{
	(void) arg;
	objset_t *os;
	ztest_ds_t *zdtmp;

	VERIFY0(ztest_dmu_objset_own(name, DMU_OST_ANY, B_TRUE,
	B_TRUE, FTAG, &os));

	zdtmp = umem_alloc(sizeof (ztest_ds_t), UMEM_NOFAIL);

	ztest_zd_init(zdtmp, NULL, os);
	zil_replay(os, zdtmp, ztest_replay_vector);
	ztest_zd_fini(zdtmp);

	if (dmu_objset_zil(os)->zl_parse_lr_count != 0 &&
	ztest_opts.zo_verbose >= 6) {
	zilog_t *zilog = dmu_objset_zil(os);

	(void) printf("%s replay %"PRIu64" blocks, "
	"%"PRIu64" records, seq %"PRIu64"\n",
	name,
	zilog->zl_parse_blk_count,
	zilog->zl_parse_lr_count,
	zilog->zl_replaying_seq);
	}

	umem_free(zdtmp, sizeof (ztest_ds_t));

	dmu_objset_disown(os, B_TRUE, FTAG);
	return (0);
	}

	static void
	ztest_freeze(void)
	{
	ztest_ds_t *zd = &ztest_ds[0];
	spa_t *spa;
	int numloops = 0;

	/* freeze not supported during RAIDZ expansion */
	if (ztest_opts.zo_raid_do_expand)
	return;

	if (ztest_opts.zo_verbose >= 3)
	(void) printf("testing spa_freeze()...\n");

	raidz_scratch_verify();
	kernel_init(SPA_MODE_READ \| SPA_MODE_WRITE);
	VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG));
	VERIFY0(ztest_dataset_open(0));
	ztest_spa = spa;

	/*
	* Force the first log block to be transactionally allocated.
	* We have to do this before we freeze the pool -- otherwise
	* the log chain won't be anchored.
	*/
	while (BP_IS_HOLE(&zd->zd_zilog->zl_header->zh_log)) {
	ztest_dmu_object_alloc_free(zd, 0);
	zil_commit(zd->zd_zilog, 0);
	}

	txg_wait_synced(spa_get_dsl(spa), 0);

	/*
	* Freeze the pool. This stops spa_sync() from doing anything,
	* so that the only way to record changes from now on is the ZIL.
	*/
	spa_freeze(spa);

	/*
	* Because it is hard to predict how much space a write will actually
	* require beforehand, we leave ourselves some fudge space to write over
	* capacity.
	*/
	uint64_t capacity = metaslab_class_get_space(spa_normal_class(spa)) / 2;

	/*
	* Run tests that generate log records but don't alter the pool config
	* or depend on DSL sync tasks (snapshots, objset create/destroy, etc).
	* We do a txg_wait_synced() after each iteration to force the txg
	* to increase well beyond the last synced value in the uberblock.
	* The ZIL should be OK with that.
	*
	* Run a random number of times less than zo_maxloops and ensure we do
	* not run out of space on the pool.
	*/
	while (ztest_random(10) != 0 &&
	numloops++ < ztest_opts.zo_maxloops &&
	metaslab_class_get_alloc(spa_normal_class(spa)) < capacity) {
	ztest_od_t od;
	ztest_od_init(&od, 0, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0);
	VERIFY0(ztest_object_init(zd, &od, sizeof (od), B_FALSE));
	ztest_io(zd, od.od_object,
	ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT);
	txg_wait_synced(spa_get_dsl(spa), 0);
	}

	/*
	* Commit all of the changes we just generated.
	*/
	zil_commit(zd->zd_zilog, 0);
	txg_wait_synced(spa_get_dsl(spa), 0);

	/*
	* Close our dataset and close the pool.
	*/
	ztest_dataset_close(0);
	spa_close(spa, FTAG);
	kernel_fini();

	/*
	* Open and close the pool and dataset to induce log replay.
	*/
	raidz_scratch_verify();
	kernel_init(SPA_MODE_READ \| SPA_MODE_WRITE);
	VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG));
	ASSERT3U(spa_freeze_txg(spa), ==, UINT64_MAX);
	VERIFY0(ztest_dataset_open(0));
	ztest_spa = spa;
	txg_wait_synced(spa_get_dsl(spa), 0);
	ztest_dataset_close(0);
	ztest_reguid(NULL, 0);

	spa_close(spa, FTAG);
	kernel_fini();
	}

	static void
	ztest_import_impl(void)
	{
	importargs_t args = { 0 };
	nvlist_t *cfg = NULL;
	int nsearch = 1;
	char *searchdirs[nsearch];
	int flags = ZFS_IMPORT_MISSING_LOG;

	searchdirs[0] = ztest_opts.zo_dir;
	args.paths = nsearch;
	args.path = searchdirs;
	args.can_be_active = B_FALSE;

	libpc_handle_t lpch = {
	.lpc_lib_handle = NULL,
	.lpc_ops = &libzpool_config_ops,
	.lpc_printerr = B_TRUE
	};
	VERIFY0(zpool_find_config(&lpch, ztest_opts.zo_pool, &cfg, &args));
	VERIFY0(spa_import(ztest_opts.zo_pool, cfg, NULL, flags));
	fnvlist_free(cfg);
	}

	/*
	* Import a storage pool with the given name.
	*/
	static void
	ztest_import(ztest_shared_t *zs)
	{
	spa_t *spa;

	mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL);
	VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL));

	raidz_scratch_verify();
	kernel_init(SPA_MODE_READ \| SPA_MODE_WRITE);

	ztest_import_impl();

	VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG));
	zs->zs_metaslab_sz =
	1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift;
	zs->zs_guid = spa_guid(spa);
	spa_close(spa, FTAG);

	kernel_fini();

	if (!ztest_opts.zo_mmp_test) {
	ztest_run_zdb(zs->zs_guid);
	ztest_freeze();
	ztest_run_zdb(zs->zs_guid);
	}

	(void) pthread_rwlock_destroy(&ztest_name_lock);
	mutex_destroy(&ztest_vdev_lock);
	mutex_destroy(&ztest_checkpoint_lock);
	}

	/*
	* After the expansion was killed, check that the pool is healthy
	*/
	static void
	ztest_raidz_expand_check(spa_t *spa)
	{
	ASSERT3U(ztest_opts.zo_raidz_expand_test, ==, RAIDZ_EXPAND_KILLED);
	/*
	* Set pool check done flag, main program will run a zdb check
	* of the pool when we exit.
	*/
	ztest_shared_opts->zo_raidz_expand_test = RAIDZ_EXPAND_CHECKED;

	/* Wait for reflow to finish */
	if (ztest_opts.zo_verbose >= 1) {
	(void) printf("\nwaiting for reflow to finish ...\n");
	}
	pool_raidz_expand_stat_t rzx_stats;
	pool_raidz_expand_stat_t *pres = &rzx_stats;
	do {
	txg_wait_synced(spa_get_dsl(spa), 0);
	(void) poll(NULL, 0, 500); /* wait 1/2 second */

	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	(void) spa_raidz_expand_get_stats(spa, pres);
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	} while (pres->pres_state != DSS_FINISHED &&
	pres->pres_reflowed < pres->pres_to_reflow);

	if (ztest_opts.zo_verbose >= 1) {
	(void) printf("verifying an interrupted raidz "
	"expansion using a pool scrub ...\n");
	}
	/* Will fail here if there is non-recoverable corruption detected */
	VERIFY0(ztest_scrub_impl(spa));
	if (ztest_opts.zo_verbose >= 1) {
	(void) printf("raidz expansion scrub check complete\n");
	}
	}

	/*
	* Start a raidz expansion test. We run some I/O on the pool for a while
	* to get some data in the pool. Then we grow the raidz and
	* kill the test at the requested offset into the reflow, verifying that
	* doing such does not lead to pool corruption.
	*/
	static void
	ztest_raidz_expand_run(ztest_shared_t zs, spa_t spa)
	{
	nvlist_t *root;
	pool_raidz_expand_stat_t rzx_stats;
	pool_raidz_expand_stat_t *pres = &rzx_stats;
	kthread_t **run_threads;
	vdev_t cvd, rzvd = spa->spa_root_vdev->vdev_child[0];
	int total_disks = rzvd->vdev_children;
	int data_disks = total_disks - vdev_get_nparity(rzvd);
	uint64_t alloc_goal;
	uint64_t csize;
	int error, t;
	int threads = ztest_opts.zo_threads;
	ztest_expand_io_t *thread_args;

	ASSERT3U(ztest_opts.zo_raidz_expand_test, !=, RAIDZ_EXPAND_NONE);
	ASSERT3P(rzvd->vdev_ops, ==, &vdev_raidz_ops);
	ztest_opts.zo_raidz_expand_test = RAIDZ_EXPAND_STARTED;

	/* Setup a 1 MiB buffer of random data */
	uint64_t bufsize = 1024 * 1024;
	void *buffer = umem_alloc(bufsize, UMEM_NOFAIL);

	if (read(ztest_fd_rand, buffer, bufsize) != bufsize) {
	fatal(B_TRUE, "short read from /dev/urandom");
	}
	/*
	* Put some data in the pool and then attach a vdev to initiate
	* reflow.
	*/
	run_threads = umem_zalloc(threads * sizeof (kthread_t *), UMEM_NOFAIL);
	thread_args = umem_zalloc(threads * sizeof (ztest_expand_io_t),
	UMEM_NOFAIL);
	/* Aim for roughly 25% of allocatable space up to 1GB */
	alloc_goal = (vdev_get_min_asize(rzvd) * data_disks) / total_disks;
	alloc_goal = MIN(alloc_goal >> 2, 102410241024);
	if (ztest_opts.zo_verbose >= 1) {
	(void) printf("adding data to pool '%s', goal %llu bytes\n",
	ztest_opts.zo_pool, (u_longlong_t)alloc_goal);
	}

	/*
	* Kick off all the I/O generators that run in parallel.
	*/
	for (t = 0; t < threads; t++) {
	if (t < ztest_opts.zo_datasets && ztest_dataset_open(t) != 0) {
	umem_free(run_threads, threads * sizeof (kthread_t *));
	umem_free(buffer, bufsize);
	return;
	}
	thread_args[t].rzx_id = t;
	thread_args[t].rzx_amount = alloc_goal / threads;
	thread_args[t].rzx_bufsize = bufsize;
	thread_args[t].rzx_buffer = buffer;
	thread_args[t].rzx_alloc_max = alloc_goal;
	thread_args[t].rzx_spa = spa;
	run_threads[t] = thread_create(NULL, 0, ztest_rzx_thread,
	&thread_args[t], 0, NULL, TS_RUN \| TS_JOINABLE,
	defclsyspri);
	}

	/*
	* Wait for all of the writers to complete.
	*/
	for (t = 0; t < threads; t++)
	VERIFY0(thread_join(run_threads[t]));

	/*
	* Close all datasets. This must be done after all the threads
	* are joined so we can be sure none of the datasets are in-use
	* by any of the threads.
	*/
	for (t = 0; t < ztest_opts.zo_threads; t++) {
	if (t < ztest_opts.zo_datasets)
	ztest_dataset_close(t);
	}

	txg_wait_synced(spa_get_dsl(spa), 0);

	zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(spa));
	zs->zs_space = metaslab_class_get_space(spa_normal_class(spa));

	umem_free(buffer, bufsize);
	umem_free(run_threads, threads * sizeof (kthread_t *));
	umem_free(thread_args, threads * sizeof (ztest_expand_io_t));

	/* Set our reflow target to 25%, 50% or 75% of allocated size */
	uint_t multiple = ztest_random(3) + 1;
	uint64_t reflow_max = (rzvd->vdev_stat.vs_alloc * multiple) / 4;
	raidz_expand_max_reflow_bytes = reflow_max;

	if (ztest_opts.zo_verbose >= 1) {
	(void) printf("running raidz expansion test, killing when "
	"reflow reaches %llu bytes (%u/4 of allocated space)\n",
	(u_longlong_t)reflow_max, multiple);
	}

	/* XXX - do we want some I/O load during the reflow? */

	/*
	* Use a disk size that is larger than existing ones
	*/
	cvd = rzvd->vdev_child[0];
	csize = vdev_get_min_asize(cvd);
	csize += csize / 10;
	/*
	* Path to vdev to be attached
	*/
	char *newpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
	(void) snprintf(newpath, MAXPATHLEN, ztest_dev_template,
	ztest_opts.zo_dir, ztest_opts.zo_pool, rzvd->vdev_children);
	/*
	* Build the nvlist describing newpath.
	*/
	root = make_vdev_root(newpath, NULL, NULL, csize, ztest_get_ashift(),
	NULL, 0, 0, 1);
	/*
	* Expand the raidz vdev by attaching the new disk
	*/
	if (ztest_opts.zo_verbose >= 1) {
	(void) printf("expanding raidz: %d wide to %d wide with '%s'\n",
	(int)rzvd->vdev_children, (int)rzvd->vdev_children + 1,
	newpath);
	}
	error = spa_vdev_attach(spa, rzvd->vdev_guid, root, B_FALSE, B_FALSE);
	nvlist_free(root);
	if (error != 0) {
	fatal(0, "raidz expand: attach (%s %llu) returned %d",
	newpath, (long long)csize, error);
	}

	/*
	* Wait for reflow to begin
	*/
	while (spa->spa_raidz_expand == NULL) {
	txg_wait_synced(spa_get_dsl(spa), 0);
	(void) poll(NULL, 0, 100); /* wait 1/10 second */
	}
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	(void) spa_raidz_expand_get_stats(spa, pres);
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	while (pres->pres_state != DSS_SCANNING) {
	txg_wait_synced(spa_get_dsl(spa), 0);
	(void) poll(NULL, 0, 100); /* wait 1/10 second */
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	(void) spa_raidz_expand_get_stats(spa, pres);
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	}

	ASSERT3U(pres->pres_state, ==, DSS_SCANNING);
	ASSERT3U(pres->pres_to_reflow, !=, 0);
	/*
	* Set so when we are killed we go to raidz checking rather than
	* restarting test.
	*/
	ztest_shared_opts->zo_raidz_expand_test = RAIDZ_EXPAND_KILLED;
	if (ztest_opts.zo_verbose >= 1) {
	(void) printf("raidz expansion reflow started, waiting for "
	"%llu bytes to be copied\n", (u_longlong_t)reflow_max);
	}

	/*
	* Wait for reflow maximum to be reached and then kill the test
	*/
	while (pres->pres_reflowed < reflow_max) {
	txg_wait_synced(spa_get_dsl(spa), 0);
	(void) poll(NULL, 0, 100); /* wait 1/10 second */
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	(void) spa_raidz_expand_get_stats(spa, pres);
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	}

	/* Reset the reflow pause before killing */
	raidz_expand_max_reflow_bytes = 0;

	if (ztest_opts.zo_verbose >= 1) {
	(void) printf("killing raidz expansion test after reflow "
	"reached %llu bytes\n", (u_longlong_t)pres->pres_reflowed);
	}

	/*
	* Kill ourself to simulate a panic during a reflow. Our parent will
	* restart the test and the changed flag value will drive the test
	* through the scrub/check code to verify the pool is not corrupted.
	*/
	ztest_kill(zs);
	}

	static void
	ztest_generic_run(ztest_shared_t zs, spa_t spa)
	{
	kthread_t **run_threads;
	int t;

	run_threads = umem_zalloc(ztest_opts.zo_threads * sizeof (kthread_t *),
	UMEM_NOFAIL);

	/*
	* Kick off all the tests that run in parallel.
	*/
	for (t = 0; t < ztest_opts.zo_threads; t++) {
	if (t < ztest_opts.zo_datasets && ztest_dataset_open(t) != 0) {
	umem_free(run_threads, ztest_opts.zo_threads *
	sizeof (kthread_t *));
	return;
	}

	run_threads[t] = thread_create(NULL, 0, ztest_thread,
	(void *)(uintptr_t)t, 0, NULL, TS_RUN \| TS_JOINABLE,
	defclsyspri);
	}

	/*
	* Wait for all of the tests to complete.
	*/
	for (t = 0; t < ztest_opts.zo_threads; t++)
	VERIFY0(thread_join(run_threads[t]));

	/*
	* Close all datasets. This must be done after all the threads
	* are joined so we can be sure none of the datasets are in-use
	* by any of the threads.
	*/
	for (t = 0; t < ztest_opts.zo_threads; t++) {
	if (t < ztest_opts.zo_datasets)
	ztest_dataset_close(t);
	}

	txg_wait_synced(spa_get_dsl(spa), 0);

	zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(spa));
	zs->zs_space = metaslab_class_get_space(spa_normal_class(spa));

	umem_free(run_threads, ztest_opts.zo_threads * sizeof (kthread_t *));
	}

	/*
	* Setup our test context and kick off threads to run tests on all datasets
	* in parallel.
	*/
	static void
	ztest_run(ztest_shared_t *zs)
	{
	spa_t *spa;
	objset_t *os;
	kthread_t resume_thread, deadman_thread;
	uint64_t object;
	int error;
	int t, d;

	ztest_exiting = B_FALSE;

	/*
	* Initialize parent/child shared state.
	*/
	mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL);
	VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL));

	zs->zs_thread_start = gethrtime();
	zs->zs_thread_stop =
	zs->zs_thread_start + ztest_opts.zo_passtime * NANOSEC;
	zs->zs_thread_stop = MIN(zs->zs_thread_stop, zs->zs_proc_stop);
	zs->zs_thread_kill = zs->zs_thread_stop;
	if (ztest_random(100) < ztest_opts.zo_killrate) {
	zs->zs_thread_kill -=
	ztest_random(ztest_opts.zo_passtime * NANOSEC);
	}

	mutex_init(&zcl.zcl_callbacks_lock, NULL, MUTEX_DEFAULT, NULL);

	list_create(&zcl.zcl_callbacks, sizeof (ztest_cb_data_t),
	offsetof(ztest_cb_data_t, zcd_node));

	/*
	* Open our pool. It may need to be imported first depending on
	* what tests were running when the previous pass was terminated.
	*/
	raidz_scratch_verify();
	kernel_init(SPA_MODE_READ \| SPA_MODE_WRITE);
	error = spa_open(ztest_opts.zo_pool, &spa, FTAG);
	if (error) {
	VERIFY3S(error, ==, ENOENT);
	ztest_import_impl();
	VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG));
	zs->zs_metaslab_sz =
	1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift;
	}

	metaslab_preload_limit = ztest_random(20) + 1;
	ztest_spa = spa;

	/*
	* XXX - BUGBUG raidz expansion do not run this for generic for now
	*/
	if (ztest_opts.zo_raidz_expand_test != RAIDZ_EXPAND_NONE)
	VERIFY0(vdev_raidz_impl_set("cycle"));

	dmu_objset_stats_t dds;
	VERIFY0(ztest_dmu_objset_own(ztest_opts.zo_pool,
	DMU_OST_ANY, B_TRUE, B_TRUE, FTAG, &os));
	dsl_pool_config_enter(dmu_objset_pool(os), FTAG);
	dmu_objset_fast_stat(os, &dds);
	dsl_pool_config_exit(dmu_objset_pool(os), FTAG);
	dmu_objset_disown(os, B_TRUE, FTAG);

	/* Give the dedicated raidz expansion test more grace time */
	if (ztest_opts.zo_raidz_expand_test != RAIDZ_EXPAND_NONE)
	zfs_deadman_synctime_ms *= 2;

	/*
	* Create a thread to periodically resume suspended I/O.
	*/
	resume_thread = thread_create(NULL, 0, ztest_resume_thread,
	spa, 0, NULL, TS_RUN \| TS_JOINABLE, defclsyspri);

	/*
	* Create a deadman thread and set to panic if we hang.
	*/
	deadman_thread = thread_create(NULL, 0, ztest_deadman_thread,
	zs, 0, NULL, TS_RUN \| TS_JOINABLE, defclsyspri);

	spa->spa_deadman_failmode = ZIO_FAILURE_MODE_PANIC;

	/*
	* Verify that we can safely inquire about any object,
	* whether it's allocated or not. To make it interesting,
	* we probe a 5-wide window around each power of two.
	* This hits all edge cases, including zero and the max.
	*/
	for (t = 0; t < 64; t++) {
	for (d = -5; d <= 5; d++) {
	error = dmu_object_info(spa->spa_meta_objset,
	(1ULL << t) + d, NULL);
	ASSERT(error == 0 \|\| error == ENOENT \|\|
	error == EINVAL);
	}
	}

	/*
	* If we got any ENOSPC errors on the previous run, destroy something.
	*/
	if (zs->zs_enospc_count != 0) {
	/* Not expecting ENOSPC errors during raidz expansion tests */
	ASSERT3U(ztest_opts.zo_raidz_expand_test, ==,
	RAIDZ_EXPAND_NONE);

	int d = ztest_random(ztest_opts.zo_datasets);
	ztest_dataset_destroy(d);
	}
	zs->zs_enospc_count = 0;

	/*
	* If we were in the middle of ztest_device_removal() and were killed
	* we need to ensure the removal and scrub complete before running
	* any tests that check ztest_device_removal_active. The removal will
	* be restarted automatically when the spa is opened, but we need to
	* initiate the scrub manually if it is not already in progress. Note
	* that we always run the scrub whenever an indirect vdev exists
	* because we have no way of knowing for sure if ztest_device_removal()
	* fully completed its scrub before the pool was reimported.
	*
	* Does not apply for the RAIDZ expansion specific test runs
	*/
	if (ztest_opts.zo_raidz_expand_test == RAIDZ_EXPAND_NONE &&
	(spa->spa_removing_phys.sr_state == DSS_SCANNING \|\|
	spa->spa_removing_phys.sr_prev_indirect_vdev != -1)) {
	while (spa->spa_removing_phys.sr_state == DSS_SCANNING)
	txg_wait_synced(spa_get_dsl(spa), 0);

	error = ztest_scrub_impl(spa);
	if (error == EBUSY)
	error = 0;
	ASSERT0(error);
	}

	if (ztest_opts.zo_verbose >= 4)
	(void) printf("starting main threads...\n");

	/*
	* Replay all logs of all datasets in the pool. This is primarily for
	* temporary datasets which wouldn't otherwise get replayed, which
	* can trigger failures when attempting to offline a SLOG in
	* ztest_fault_inject().
	*/
	(void) dmu_objset_find(ztest_opts.zo_pool, ztest_replay_zil_cb,
	NULL, DS_FIND_CHILDREN);

	if (ztest_opts.zo_raidz_expand_test == RAIDZ_EXPAND_REQUESTED)
	ztest_raidz_expand_run(zs, spa);
	else if (ztest_opts.zo_raidz_expand_test == RAIDZ_EXPAND_KILLED)
	ztest_raidz_expand_check(spa);
	else
	ztest_generic_run(zs, spa);

	/* Kill the resume and deadman threads */
	ztest_exiting = B_TRUE;
	VERIFY0(thread_join(resume_thread));
	VERIFY0(thread_join(deadman_thread));
	ztest_resume(spa);

	/*
	* Right before closing the pool, kick off a bunch of async I/O;
	* spa_close() should wait for it to complete.
	*/
	for (object = 1; object < 50; object++) {
	dmu_prefetch(spa->spa_meta_objset, object, 0, 0, 1ULL << 20,
	ZIO_PRIORITY_SYNC_READ);
	}

	/* Verify that at least one commit cb was called in a timely fashion */
	if (zc_cb_counter >= ZTEST_COMMIT_CB_MIN_REG)
	VERIFY0(zc_min_txg_delay);

	spa_close(spa, FTAG);

	/*
	* Verify that we can loop over all pools.
	*/
	mutex_enter(&spa_namespace_lock);
	for (spa = spa_next(NULL); spa != NULL; spa = spa_next(spa))
	if (ztest_opts.zo_verbose > 3)
	(void) printf("spa_next: found %s\n", spa_name(spa));
	mutex_exit(&spa_namespace_lock);

	/*
	* Verify that we can export the pool and reimport it under a
	* different name.
	*/
	if ((ztest_random(2) == 0) && !ztest_opts.zo_mmp_test) {
	char name[ZFS_MAX_DATASET_NAME_LEN];
	(void) snprintf(name, sizeof (name), "%s_import",
	ztest_opts.zo_pool);
	ztest_spa_import_export(ztest_opts.zo_pool, name);
	ztest_spa_import_export(name, ztest_opts.zo_pool);
	}

	kernel_fini();

	list_destroy(&zcl.zcl_callbacks);
	mutex_destroy(&zcl.zcl_callbacks_lock);
	(void) pthread_rwlock_destroy(&ztest_name_lock);
	mutex_destroy(&ztest_vdev_lock);
	mutex_destroy(&ztest_checkpoint_lock);
	}

	static void
	print_time(hrtime_t t, char *timebuf)
	{
	hrtime_t s = t / NANOSEC;
	hrtime_t m = s / 60;
	hrtime_t h = m / 60;
	hrtime_t d = h / 24;

	s -= m * 60;
	m -= h * 60;
	h -= d * 24;

	timebuf[0] = '\0';

	if (d)
	(void) sprintf(timebuf,
	"%llud%02lluh%02llum%02llus", d, h, m, s);
	else if (h)
	(void) sprintf(timebuf, "%lluh%02llum%02llus", h, m, s);
	else if (m)
	(void) sprintf(timebuf, "%llum%02llus", m, s);
	else
	(void) sprintf(timebuf, "%llus", s);
	}

	static nvlist_t *
	make_random_props(void)
	{
	nvlist_t *props;

	props = fnvlist_alloc();

	if (ztest_random(2) == 0)
	return (props);

	fnvlist_add_uint64(props,
	zpool_prop_to_name(ZPOOL_PROP_AUTOREPLACE), 1);

	return (props);
	}

	/*
	* Create a storage pool with the given name and initial vdev size.
	* Then test spa_freeze() functionality.
	*/
	static void
	ztest_init(ztest_shared_t *zs)
	{
	spa_t *spa;
	nvlist_t nvroot, props;
	int i;

	mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL);
	VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL));

	raidz_scratch_verify();
	kernel_init(SPA_MODE_READ \| SPA_MODE_WRITE);

	/*
	* Create the storage pool.
	*/
	(void) spa_destroy(ztest_opts.zo_pool);
	ztest_shared->zs_vdev_next_leaf = 0;
	zs->zs_splits = 0;
	zs->zs_mirrors = ztest_opts.zo_mirrors;
	nvroot = make_vdev_root(NULL, NULL, NULL, ztest_opts.zo_vdev_size, 0,
	NULL, ztest_opts.zo_raid_children, zs->zs_mirrors, 1);
	props = make_random_props();

	/*
	* We don't expect the pool to suspend unless maxfaults == 0,
	* in which case ztest_fault_inject() temporarily takes away
	* the only valid replica.
	*/
	fnvlist_add_uint64(props,
	zpool_prop_to_name(ZPOOL_PROP_FAILUREMODE),
	MAXFAULTS(zs) ? ZIO_FAILURE_MODE_PANIC : ZIO_FAILURE_MODE_WAIT);

	for (i = 0; i < SPA_FEATURES; i++) {
	char *buf;

	if (!spa_feature_table[i].fi_zfs_mod_supported)
	continue;

	/*
	* 75% chance of using the log space map feature. We want ztest
	* to exercise both the code paths that use the log space map
	* feature and the ones that don't.
	*/
	if (i == SPA_FEATURE_LOG_SPACEMAP && ztest_random(4) == 0)
	continue;

	VERIFY3S(-1, !=, asprintf(&buf, "feature@%s",
	spa_feature_table[i].fi_uname));
	fnvlist_add_uint64(props, buf, 0);
	free(buf);
	}

	VERIFY0(spa_create(ztest_opts.zo_pool, nvroot, props, NULL, NULL));
	fnvlist_free(nvroot);
	fnvlist_free(props);

	VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG));
	zs->zs_metaslab_sz =
	1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift;
	zs->zs_guid = spa_guid(spa);
	spa_close(spa, FTAG);

	kernel_fini();

	if (!ztest_opts.zo_mmp_test) {
	ztest_run_zdb(zs->zs_guid);
	ztest_freeze();
	ztest_run_zdb(zs->zs_guid);
	}

	(void) pthread_rwlock_destroy(&ztest_name_lock);
	mutex_destroy(&ztest_vdev_lock);
	mutex_destroy(&ztest_checkpoint_lock);
	}

	static void
	setup_data_fd(void)
	{
	static char ztest_name_data[] = "/tmp/ztest.data.XXXXXX";

	ztest_fd_data = mkstemp(ztest_name_data);
	ASSERT3S(ztest_fd_data, >=, 0);
	(void) unlink(ztest_name_data);
	}

	static int
	shared_data_size(ztest_shared_hdr_t *hdr)
	{
	int size;

	size = hdr->zh_hdr_size;
	size += hdr->zh_opts_size;
	size += hdr->zh_size;
	size += hdr->zh_stats_size * hdr->zh_stats_count;
	size += hdr->zh_ds_size * hdr->zh_ds_count;
	size += hdr->zh_scratch_state_size;

	return (size);
	}

	static void
	setup_hdr(void)
	{
	int size;
	ztest_shared_hdr_t *hdr;

	hdr = (void )mmap(0, P2ROUNDUP(sizeof (hdr), getpagesize()),
	PROT_READ \| PROT_WRITE, MAP_SHARED, ztest_fd_data, 0);
	ASSERT3P(hdr, !=, MAP_FAILED);

	VERIFY0(ftruncate(ztest_fd_data, sizeof (ztest_shared_hdr_t)));

	hdr->zh_hdr_size = sizeof (ztest_shared_hdr_t);
	hdr->zh_opts_size = sizeof (ztest_shared_opts_t);
	hdr->zh_size = sizeof (ztest_shared_t);
	hdr->zh_stats_size = sizeof (ztest_shared_callstate_t);
	hdr->zh_stats_count = ZTEST_FUNCS;
	hdr->zh_ds_size = sizeof (ztest_shared_ds_t);
	hdr->zh_ds_count = ztest_opts.zo_datasets;
	hdr->zh_scratch_state_size = sizeof (ztest_shared_scratch_state_t);

	size = shared_data_size(hdr);
	VERIFY0(ftruncate(ztest_fd_data, size));

	(void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize()));
	}

	static void
	setup_data(void)
	{
	int size, offset;
	ztest_shared_hdr_t *hdr;
	uint8_t *buf;

	hdr = (void )mmap(0, P2ROUNDUP(sizeof (hdr), getpagesize()),
	PROT_READ, MAP_SHARED, ztest_fd_data, 0);
	ASSERT3P(hdr, !=, MAP_FAILED);

	size = shared_data_size(hdr);

	(void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize()));
	hdr = ztest_shared_hdr = (void *)mmap(0, P2ROUNDUP(size, getpagesize()),
	PROT_READ \| PROT_WRITE, MAP_SHARED, ztest_fd_data, 0);
	ASSERT3P(hdr, !=, MAP_FAILED);
	buf = (uint8_t *)hdr;

	offset = hdr->zh_hdr_size;
	ztest_shared_opts = (void *)&buf[offset];
	offset += hdr->zh_opts_size;
	ztest_shared = (void *)&buf[offset];
	offset += hdr->zh_size;
	ztest_shared_callstate = (void *)&buf[offset];
	offset += hdr->zh_stats_size * hdr->zh_stats_count;
	ztest_shared_ds = (void *)&buf[offset];
	offset += hdr->zh_ds_size * hdr->zh_ds_count;
	ztest_scratch_state = (void *)&buf[offset];
	}

	static boolean_t
	exec_child(char cmd, char libpath, boolean_t ignorekill, int *statusp)
	{
	pid_t pid;
	int status;
	char *cmdbuf = NULL;

	pid = fork();

	if (cmd == NULL) {
	cmdbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL);
	(void) strlcpy(cmdbuf, getexecname(), MAXPATHLEN);
	cmd = cmdbuf;
	}

	if (pid == -1)
	fatal(B_TRUE, "fork failed");

	if (pid == 0) { /* child */
	char fd_data_str[12];

	VERIFY3S(11, >=,
	snprintf(fd_data_str, 12, "%d", ztest_fd_data));
	VERIFY0(setenv("ZTEST_FD_DATA", fd_data_str, 1));

	if (libpath != NULL) {
	const char *curlp = getenv("LD_LIBRARY_PATH");
	if (curlp == NULL)
	VERIFY0(setenv("LD_LIBRARY_PATH", libpath, 1));
	else {
	char *newlp = NULL;
	VERIFY3S(-1, !=,
	asprintf(&newlp, "%s:%s", libpath, curlp));
	VERIFY0(setenv("LD_LIBRARY_PATH", newlp, 1));
	free(newlp);
	}
	}
	(void) execl(cmd, cmd, (char *)NULL);
	ztest_dump_core = B_FALSE;
	fatal(B_TRUE, "exec failed: %s", cmd);
	}

	if (cmdbuf != NULL) {
	umem_free(cmdbuf, MAXPATHLEN);
	cmd = NULL;
	}

	while (waitpid(pid, &status, 0) != pid)
	continue;
	if (statusp != NULL)
	*statusp = status;

	if (WIFEXITED(status)) {
	if (WEXITSTATUS(status) != 0) {
	(void) fprintf(stderr, "child exited with code %d\n",
	WEXITSTATUS(status));
	exit(2);
	}
	return (B_FALSE);
	} else if (WIFSIGNALED(status)) {
	if (!ignorekill \|\| WTERMSIG(status) != SIGKILL) {
	(void) fprintf(stderr, "child died with signal %d\n",
	WTERMSIG(status));
	exit(3);
	}
	return (B_TRUE);
	} else {
	(void) fprintf(stderr, "something strange happened to child\n");
	exit(4);
	}
	}

	static void
	ztest_run_init(void)
	{
	int i;

	ztest_shared_t *zs = ztest_shared;

	/*
	* Blow away any existing copy of zpool.cache
	*/
	(void) remove(spa_config_path);

	if (ztest_opts.zo_init == 0) {
	if (ztest_opts.zo_verbose >= 1)
	(void) printf("Importing pool %s\n",
	ztest_opts.zo_pool);
	ztest_import(zs);
	return;
	}

	/*
	* Create and initialize our storage pool.
	*/
	for (i = 1; i <= ztest_opts.zo_init; i++) {
	memset(zs, 0, sizeof (*zs));
	if (ztest_opts.zo_verbose >= 3 &&
	ztest_opts.zo_init != 1) {
	(void) printf("ztest_init(), pass %d\n", i);
	}
	ztest_init(zs);
	}
	}

	int
	main(int argc, char **argv)
	{
	int kills = 0;
	int iters = 0;
	int older = 0;
	int newer = 0;
	ztest_shared_t *zs;
	ztest_info_t *zi;
	ztest_shared_callstate_t *zc;
	char timebuf[100];
	char numbuf[NN_NUMBUF_SZ];
	char *cmd;
	boolean_t hasalt;
	int f, err;
	char *fd_data_str = getenv("ZTEST_FD_DATA");
	struct sigaction action;

	(void) setvbuf(stdout, NULL, _IOLBF, 0);

	dprintf_setup(&argc, argv);
	zfs_deadman_synctime_ms = 300000;
	zfs_deadman_checktime_ms = 30000;
	/*
	* As two-word space map entries may not come up often (especially
	* if pool and vdev sizes are small) we want to force at least some
	* of them so the feature get tested.
	*/
	zfs_force_some_double_word_sm_entries = B_TRUE;

	/*
	* Verify that even extensively damaged split blocks with many
	* segments can be reconstructed in a reasonable amount of time
	* when reconstruction is known to be possible.
	*
	* Note: the lower this value is, the more damage we inflict, and
	* the more time ztest spends in recovering that damage. We chose
	* to induce damage 1/100th of the time so recovery is tested but
	* not so frequently that ztest doesn't get to test other code paths.
	*/
	zfs_reconstruct_indirect_damage_fraction = 100;

	action.sa_handler = sig_handler;
	sigemptyset(&action.sa_mask);
	action.sa_flags = 0;

	if (sigaction(SIGSEGV, &action, NULL) < 0) {
	(void) fprintf(stderr, "ztest: cannot catch SIGSEGV: %s.\n",
	strerror(errno));
	exit(EXIT_FAILURE);
	}

	if (sigaction(SIGABRT, &action, NULL) < 0) {
	(void) fprintf(stderr, "ztest: cannot catch SIGABRT: %s.\n",
	strerror(errno));
	exit(EXIT_FAILURE);
	}

	/*
	* Force random_get_bytes() to use /dev/urandom in order to prevent
	* ztest from needlessly depleting the system entropy pool.
	*/
	random_path = "/dev/urandom";
	ztest_fd_rand = open(random_path, O_RDONLY \| O_CLOEXEC);
	ASSERT3S(ztest_fd_rand, >=, 0);

	if (!fd_data_str) {
	process_options(argc, argv);

	setup_data_fd();
	setup_hdr();
	setup_data();
	memcpy(ztest_shared_opts, &ztest_opts,
	sizeof (*ztest_shared_opts));
	} else {
	ztest_fd_data = atoi(fd_data_str);
	setup_data();
	memcpy(&ztest_opts, ztest_shared_opts, sizeof (ztest_opts));
	}
	ASSERT3U(ztest_opts.zo_datasets, ==, ztest_shared_hdr->zh_ds_count);

	err = ztest_set_global_vars();
	if (err != 0 && !fd_data_str) {
	/* error message done by ztest_set_global_vars */
	exit(EXIT_FAILURE);
	} else {
	/* children should not be spawned if setting gvars fails */
	VERIFY3S(err, ==, 0);
	}

	/* Override location of zpool.cache */
	VERIFY3S(asprintf((char **)&spa_config_path, "%s/zpool.cache",
	ztest_opts.zo_dir), !=, -1);

	ztest_ds = umem_alloc(ztest_opts.zo_datasets * sizeof (ztest_ds_t),
	UMEM_NOFAIL);
	zs = ztest_shared;

	if (fd_data_str) {
	metaslab_force_ganging = ztest_opts.zo_metaslab_force_ganging;
	metaslab_df_alloc_threshold =
	zs->zs_metaslab_df_alloc_threshold;

	if (zs->zs_do_init)
	ztest_run_init();
	else
	ztest_run(zs);
	exit(0);
	}

	hasalt = (strlen(ztest_opts.zo_alt_ztest) != 0);

	if (ztest_opts.zo_verbose >= 1) {
	(void) printf("%"PRIu64" vdevs, %d datasets, %d threads, "
	"%d %s disks, parity %d, %"PRIu64" seconds...\n\n",
	ztest_opts.zo_vdevs,
	ztest_opts.zo_datasets,
	ztest_opts.zo_threads,
	ztest_opts.zo_raid_children,
	ztest_opts.zo_raid_type,
	ztest_opts.zo_raid_parity,
	ztest_opts.zo_time);
	}

	cmd = umem_alloc(MAXNAMELEN, UMEM_NOFAIL);
	(void) strlcpy(cmd, getexecname(), MAXNAMELEN);

	zs->zs_do_init = B_TRUE;
	if (strlen(ztest_opts.zo_alt_ztest) != 0) {
	if (ztest_opts.zo_verbose >= 1) {
	(void) printf("Executing older ztest for "
	"initialization: %s\n", ztest_opts.zo_alt_ztest);
	}
	VERIFY(!exec_child(ztest_opts.zo_alt_ztest,
	ztest_opts.zo_alt_libpath, B_FALSE, NULL));
	} else {
	VERIFY(!exec_child(NULL, NULL, B_FALSE, NULL));
	}
	zs->zs_do_init = B_FALSE;

	zs->zs_proc_start = gethrtime();
	zs->zs_proc_stop = zs->zs_proc_start + ztest_opts.zo_time * NANOSEC;

	for (f = 0; f < ZTEST_FUNCS; f++) {
	zi = &ztest_info[f];
	zc = ZTEST_GET_SHARED_CALLSTATE(f);
	if (zs->zs_proc_start + zi->zi_interval[0] > zs->zs_proc_stop)
	zc->zc_next = UINT64_MAX;
	else
	zc->zc_next = zs->zs_proc_start +
	ztest_random(2 * zi->zi_interval[0] + 1);
	}

	/*
	* Run the tests in a loop. These tests include fault injection
	* to verify that self-healing data works, and forced crashes
	* to verify that we never lose on-disk consistency.
	*/
	while (gethrtime() < zs->zs_proc_stop) {
	int status;
	boolean_t killed;

	/*
	* Initialize the workload counters for each function.
	*/
	for (f = 0; f < ZTEST_FUNCS; f++) {
	zc = ZTEST_GET_SHARED_CALLSTATE(f);
	zc->zc_count = 0;
	zc->zc_time = 0;
	}

	/* Set the allocation switch size */
	zs->zs_metaslab_df_alloc_threshold =
	ztest_random(zs->zs_metaslab_sz / 4) + 1;

	if (!hasalt \|\| ztest_random(2) == 0) {
	if (hasalt && ztest_opts.zo_verbose >= 1) {
	(void) printf("Executing newer ztest: %s\n",
	cmd);
	}
	newer++;
	killed = exec_child(cmd, NULL, B_TRUE, &status);
	} else {
	if (hasalt && ztest_opts.zo_verbose >= 1) {
	(void) printf("Executing older ztest: %s\n",
	ztest_opts.zo_alt_ztest);
	}
	older++;
	killed = exec_child(ztest_opts.zo_alt_ztest,
	ztest_opts.zo_alt_libpath, B_TRUE, &status);
	}

	if (killed)
	kills++;
	iters++;

	if (ztest_opts.zo_verbose >= 1) {
	hrtime_t now = gethrtime();

	now = MIN(now, zs->zs_proc_stop);
	print_time(zs->zs_proc_stop - now, timebuf);
	nicenum(zs->zs_space, numbuf, sizeof (numbuf));

	(void) printf("Pass %3d, %8s, %3"PRIu64" ENOSPC, "
	"%4.1f%% of %5s used, %3.0f%% done, %8s to go\n",
	iters,
	WIFEXITED(status) ? "Complete" : "SIGKILL",
	zs->zs_enospc_count,
	100.0 * zs->zs_alloc / zs->zs_space,
	numbuf,
	100.0 * (now - zs->zs_proc_start) /
	(ztest_opts.zo_time * NANOSEC), timebuf);
	}

	if (ztest_opts.zo_verbose >= 2) {
	(void) printf("\nWorkload summary:\n\n");
	(void) printf("%7s %9s %s\n",
	"Calls", "Time", "Function");
	(void) printf("%7s %9s %s\n",
	"-----", "----", "--------");
	for (f = 0; f < ZTEST_FUNCS; f++) {
	zi = &ztest_info[f];
	zc = ZTEST_GET_SHARED_CALLSTATE(f);
	print_time(zc->zc_time, timebuf);
	(void) printf("%7"PRIu64" %9s %s\n",
	zc->zc_count, timebuf,
	zi->zi_funcname);
	}
	(void) printf("\n");
	}

	if (!ztest_opts.zo_mmp_test)
	ztest_run_zdb(zs->zs_guid);
	if (ztest_shared_opts->zo_raidz_expand_test ==
	RAIDZ_EXPAND_CHECKED)
	break; /* raidz expand test complete */
	}

	if (ztest_opts.zo_verbose >= 1) {
	if (hasalt) {
	(void) printf("%d runs of older ztest: %s\n", older,
	ztest_opts.zo_alt_ztest);
	(void) printf("%d runs of newer ztest: %s\n", newer,
	cmd);
	}
	(void) printf("%d killed, %d completed, %.0f%% kill rate\n",
	kills, iters - kills, (100.0 * kills) / MAX(1, iters));
	}

	umem_free(cmd, MAXNAMELEN);

	return (0);
	}
	diff --git a/sys/contrib/openzfs/config/kernel-types.m4 b/sys/contrib/openzfs/config/kernel-types.m4
	new file mode 100644
	index 000000000000..ed76af28337b
	--- /dev/null
	+++ b/sys/contrib/openzfs/config/kernel-types.m4
	@@ -0,0 +1,40 @@
	+dnl #
	+dnl # check if kernel provides definitions for given types
	+dnl #
	+
	+dnl _ZFS_AC_KERNEL_SRC_TYPE(type)
	+AC_DEFUN([_ZFS_AC_KERNEL_SRC_TYPE], [
	+ ZFS_LINUX_TEST_SRC([type_$1], [
	+ #include <linux/types.h>
	+ ],[
	+ const $1 __attribute__((unused)) x = ($1) 0;
	+ ])
	+])
	+
	+dnl _ZFS_AC_KERNEL_TYPE(type)
	+AC_DEFUN([_ZFS_AC_KERNEL_TYPE], [
	+ AC_MSG_CHECKING([whether kernel defines $1])
	+ ZFS_LINUX_TEST_RESULT([type_$1], [
	+ AC_MSG_RESULT([yes])
	+ AC_DEFINE([HAVE_KERNEL_]m4_quote(m4_translit([$1], [a-z], [A-Z])),
	+ 1, [kernel defines $1])
	+ ], [
	+ AC_MSG_RESULT([no])
	+ ])
	+])
	+
	+dnl ZFS_AC_KERNEL_TYPES([types...])
	+AC_DEFUN([ZFS_AC_KERNEL_TYPES], [
	+ AC_DEFUN([ZFS_AC_KERNEL_SRC_TYPES], [
	+ m4_foreach_w([type], [$1], [
	+ _ZFS_AC_KERNEL_SRC_TYPE(type)
	+ ])
	+ ])
	+ AC_DEFUN([ZFS_AC_KERNEL_TYPES], [
	+ m4_foreach_w([type], [$1], [
	+ _ZFS_AC_KERNEL_TYPE(type)
	+ ])
	+ ])
	+])
	+
	+ZFS_AC_KERNEL_TYPES([intptr_t])
	diff --git a/sys/contrib/openzfs/config/kernel.m4 b/sys/contrib/openzfs/config/kernel.m4
	index 548905ccd04d..b51477b6a951 100644
	--- a/sys/contrib/openzfs/config/kernel.m4
	+++ b/sys/contrib/openzfs/config/kernel.m4
	@@ -1,1046 +1,1048 @@
	dnl #
	dnl # Default ZFS kernel configuration
	dnl #
	AC_DEFUN([ZFS_AC_CONFIG_KERNEL], [
	AM_COND_IF([BUILD_LINUX], [
	dnl # Setup the kernel build environment.
	ZFS_AC_KERNEL
	ZFS_AC_QAT

	dnl # Sanity checks for module building and CONFIG_* defines
	ZFS_AC_KERNEL_CONFIG_DEFINED
	ZFS_AC_MODULE_SYMVERS

	dnl # Sequential ZFS_LINUX_TRY_COMPILE tests
	ZFS_AC_KERNEL_FPU_HEADER
	ZFS_AC_KERNEL_OBJTOOL_HEADER
	ZFS_AC_KERNEL_WAIT_QUEUE_ENTRY_T
	ZFS_AC_KERNEL_MISC_MINOR
	ZFS_AC_KERNEL_DECLARE_EVENT_CLASS

	dnl # Parallel ZFS_LINUX_TEST_SRC / ZFS_LINUX_TEST_RESULT tests
	ZFS_AC_KERNEL_TEST_SRC
	ZFS_AC_KERNEL_TEST_RESULT

	AS_IF([test "$LINUX_OBJ" != "$LINUX"], [
	KERNEL_MAKE="$KERNEL_MAKE O=$LINUX_OBJ"
	])

	AC_SUBST(KERNEL_MAKE)
	])
	])

	dnl #
	dnl # Generate and compile all of the kernel API test cases to determine
	dnl # which interfaces are available. By invoking the kernel build system
	dnl # only once the compilation can be done in parallel significantly
	dnl # speeding up the process.
	dnl #
	AC_DEFUN([ZFS_AC_KERNEL_TEST_SRC], [
	+ ZFS_AC_KERNEL_SRC_TYPES
	ZFS_AC_KERNEL_SRC_OBJTOOL
	ZFS_AC_KERNEL_SRC_GLOBAL_PAGE_STATE
	ZFS_AC_KERNEL_SRC_ACCESS_OK_TYPE
	ZFS_AC_KERNEL_SRC_PDE_DATA
	ZFS_AC_KERNEL_SRC_FALLOCATE
	ZFS_AC_KERNEL_SRC_FADVISE
	ZFS_AC_KERNEL_SRC_GENERIC_FADVISE
	ZFS_AC_KERNEL_SRC_2ARGS_ZLIB_DEFLATE_WORKSPACESIZE
	ZFS_AC_KERNEL_SRC_RWSEM
	ZFS_AC_KERNEL_SRC_SCHED
	ZFS_AC_KERNEL_SRC_USLEEP_RANGE
	ZFS_AC_KERNEL_SRC_KMEM_CACHE
	ZFS_AC_KERNEL_SRC_KVMALLOC
	ZFS_AC_KERNEL_SRC_VMALLOC_PAGE_KERNEL
	ZFS_AC_KERNEL_SRC_WAIT
	ZFS_AC_KERNEL_SRC_INODE_TIMES
	ZFS_AC_KERNEL_SRC_INODE_LOCK
	ZFS_AC_KERNEL_SRC_GROUP_INFO_GID
	ZFS_AC_KERNEL_SRC_RW
	ZFS_AC_KERNEL_SRC_TIMER_SETUP
	ZFS_AC_KERNEL_SRC_SUPER_USER_NS
	ZFS_AC_KERNEL_SRC_PROC_OPERATIONS
	ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS
	ZFS_AC_KERNEL_SRC_BIO
	ZFS_AC_KERNEL_SRC_BLKDEV
	ZFS_AC_KERNEL_SRC_BLK_QUEUE
	ZFS_AC_KERNEL_SRC_GENHD_FLAGS
	ZFS_AC_KERNEL_SRC_REVALIDATE_DISK
	ZFS_AC_KERNEL_SRC_GET_DISK_RO
	ZFS_AC_KERNEL_SRC_GENERIC_READLINK_GLOBAL
	ZFS_AC_KERNEL_SRC_DISCARD_GRANULARITY
	ZFS_AC_KERNEL_SRC_INODE_OWNER_OR_CAPABLE
	ZFS_AC_KERNEL_SRC_XATTR
	ZFS_AC_KERNEL_SRC_ACL
	ZFS_AC_KERNEL_SRC_INODE_SETATTR
	ZFS_AC_KERNEL_SRC_INODE_GETATTR
	ZFS_AC_KERNEL_SRC_INODE_SET_FLAGS
	ZFS_AC_KERNEL_SRC_INODE_SET_IVERSION
	ZFS_AC_KERNEL_SRC_SHOW_OPTIONS
	ZFS_AC_KERNEL_SRC_FILE_INODE
	ZFS_AC_KERNEL_SRC_FILE_DENTRY
	ZFS_AC_KERNEL_SRC_FSYNC
	ZFS_AC_KERNEL_SRC_AIO_FSYNC
	ZFS_AC_KERNEL_SRC_EVICT_INODE
	ZFS_AC_KERNEL_SRC_DIRTY_INODE
	ZFS_AC_KERNEL_SRC_SHRINKER
	ZFS_AC_KERNEL_SRC_MKDIR
	ZFS_AC_KERNEL_SRC_LOOKUP_FLAGS
	ZFS_AC_KERNEL_SRC_CREATE
	ZFS_AC_KERNEL_SRC_PERMISSION
	ZFS_AC_KERNEL_SRC_GET_LINK
	ZFS_AC_KERNEL_SRC_PUT_LINK
	ZFS_AC_KERNEL_SRC_TMPFILE
	ZFS_AC_KERNEL_SRC_AUTOMOUNT
	ZFS_AC_KERNEL_SRC_ENCODE_FH_WITH_INODE
	ZFS_AC_KERNEL_SRC_COMMIT_METADATA
	ZFS_AC_KERNEL_SRC_CLEAR_INODE
	ZFS_AC_KERNEL_SRC_SETATTR_PREPARE
	ZFS_AC_KERNEL_SRC_INSERT_INODE_LOCKED
	ZFS_AC_KERNEL_SRC_DENTRY
	ZFS_AC_KERNEL_SRC_DENTRY_ALIAS_D_U
	ZFS_AC_KERNEL_SRC_TRUNCATE_SETSIZE
	ZFS_AC_KERNEL_SRC_SECURITY_INODE
	ZFS_AC_KERNEL_SRC_FST_MOUNT
	ZFS_AC_KERNEL_SRC_BDI
	ZFS_AC_KERNEL_SRC_SET_NLINK
	ZFS_AC_KERNEL_SRC_SGET
	ZFS_AC_KERNEL_SRC_LSEEK_EXECUTE
	ZFS_AC_KERNEL_SRC_VFS_FILEMAP_DIRTY_FOLIO
	ZFS_AC_KERNEL_SRC_VFS_READ_FOLIO
	ZFS_AC_KERNEL_SRC_VFS_GETATTR
	ZFS_AC_KERNEL_SRC_VFS_FSYNC_2ARGS
	ZFS_AC_KERNEL_SRC_VFS_ITERATE
	ZFS_AC_KERNEL_SRC_VFS_DIRECT_IO
	ZFS_AC_KERNEL_SRC_VFS_READPAGES
	ZFS_AC_KERNEL_SRC_VFS_SET_PAGE_DIRTY_NOBUFFERS
	ZFS_AC_KERNEL_SRC_VFS_RW_ITERATE
	ZFS_AC_KERNEL_SRC_VFS_GENERIC_WRITE_CHECKS
	ZFS_AC_KERNEL_SRC_VFS_IOV_ITER
	ZFS_AC_KERNEL_SRC_VFS_COPY_FILE_RANGE
	ZFS_AC_KERNEL_SRC_VFS_GENERIC_COPY_FILE_RANGE
	ZFS_AC_KERNEL_SRC_VFS_SPLICE_COPY_FILE_RANGE
	ZFS_AC_KERNEL_SRC_VFS_REMAP_FILE_RANGE
	ZFS_AC_KERNEL_SRC_VFS_CLONE_FILE_RANGE
	ZFS_AC_KERNEL_SRC_VFS_DEDUPE_FILE_RANGE
	ZFS_AC_KERNEL_SRC_VFS_FILE_OPERATIONS_EXTEND
	ZFS_AC_KERNEL_SRC_KMAP_ATOMIC_ARGS
	ZFS_AC_KERNEL_SRC_FOLLOW_DOWN_ONE
	ZFS_AC_KERNEL_SRC_MAKE_REQUEST_FN
	ZFS_AC_KERNEL_SRC_GENERIC_IO_ACCT
	ZFS_AC_KERNEL_SRC_FPU
	ZFS_AC_KERNEL_SRC_FMODE_T
	ZFS_AC_KERNEL_SRC_KUIDGID_T
	ZFS_AC_KERNEL_SRC_KUID_HELPERS
	ZFS_AC_KERNEL_SRC_RENAME
	ZFS_AC_KERNEL_SRC_CURRENT_TIME
	ZFS_AC_KERNEL_SRC_USERNS_CAPABILITIES
	ZFS_AC_KERNEL_SRC_IN_COMPAT_SYSCALL
	ZFS_AC_KERNEL_SRC_KTIME
	ZFS_AC_KERNEL_SRC_TOTALRAM_PAGES_FUNC
	ZFS_AC_KERNEL_SRC_TOTALHIGH_PAGES
	ZFS_AC_KERNEL_SRC_KSTRTOUL
	ZFS_AC_KERNEL_SRC_PERCPU
	ZFS_AC_KERNEL_SRC_CPU_HOTPLUG
	ZFS_AC_KERNEL_SRC_GENERIC_FILLATTR
	ZFS_AC_KERNEL_SRC_MKNOD
	ZFS_AC_KERNEL_SRC_SYMLINK
	ZFS_AC_KERNEL_SRC_BIO_MAX_SEGS
	ZFS_AC_KERNEL_SRC_SIGNAL_STOP
	ZFS_AC_KERNEL_SRC_SIGINFO
	ZFS_AC_KERNEL_SRC_SYSFS
	ZFS_AC_KERNEL_SRC_SET_SPECIAL_STATE
	ZFS_AC_KERNEL_SRC_STANDALONE_LINUX_STDARG
	ZFS_AC_KERNEL_SRC_STRLCPY
	ZFS_AC_KERNEL_SRC_STRSCPY
	ZFS_AC_KERNEL_SRC_PAGEMAP_FOLIO_WAIT_BIT
	ZFS_AC_KERNEL_SRC_ADD_DISK
	ZFS_AC_KERNEL_SRC_KTHREAD
	ZFS_AC_KERNEL_SRC_ZERO_PAGE
	ZFS_AC_KERNEL_SRC___COPY_FROM_USER_INATOMIC
	ZFS_AC_KERNEL_SRC_USER_NS_COMMON_INUM
	ZFS_AC_KERNEL_SRC_IDMAP_MNT_API
	ZFS_AC_KERNEL_SRC_IDMAP_NO_USERNS
	ZFS_AC_KERNEL_SRC_IATTR_VFSID
	ZFS_AC_KERNEL_SRC_FILEMAP
	ZFS_AC_KERNEL_SRC_WRITEPAGE_T
	ZFS_AC_KERNEL_SRC_RECLAIMED
	ZFS_AC_KERNEL_SRC_REGISTER_SYSCTL_TABLE
	ZFS_AC_KERNEL_SRC_COPY_SPLICE_READ
	ZFS_AC_KERNEL_SRC_SYNC_BDEV
	ZFS_AC_KERNEL_SRC_MM_PAGE_SIZE
	case "$host_cpu" in
	powerpc*)
	ZFS_AC_KERNEL_SRC_CPU_HAS_FEATURE
	ZFS_AC_KERNEL_SRC_FLUSH_DCACHE_PAGE
	;;
	riscv*)
	ZFS_AC_KERNEL_SRC_FLUSH_DCACHE_PAGE
	;;
	esac

	AC_MSG_CHECKING([for available kernel interfaces])
	ZFS_LINUX_TEST_COMPILE_ALL([kabi])
	AC_MSG_RESULT([done])
	])

	dnl #
	dnl # Check results of kernel interface tests.
	dnl #
	AC_DEFUN([ZFS_AC_KERNEL_TEST_RESULT], [
	+ ZFS_AC_KERNEL_TYPES
	ZFS_AC_KERNEL_ACCESS_OK_TYPE
	ZFS_AC_KERNEL_GLOBAL_PAGE_STATE
	ZFS_AC_KERNEL_OBJTOOL
	ZFS_AC_KERNEL_PDE_DATA
	ZFS_AC_KERNEL_FALLOCATE
	ZFS_AC_KERNEL_FADVISE
	ZFS_AC_KERNEL_GENERIC_FADVISE
	ZFS_AC_KERNEL_2ARGS_ZLIB_DEFLATE_WORKSPACESIZE
	ZFS_AC_KERNEL_RWSEM
	ZFS_AC_KERNEL_SCHED
	ZFS_AC_KERNEL_USLEEP_RANGE
	ZFS_AC_KERNEL_KMEM_CACHE
	ZFS_AC_KERNEL_KVMALLOC
	ZFS_AC_KERNEL_VMALLOC_PAGE_KERNEL
	ZFS_AC_KERNEL_WAIT
	ZFS_AC_KERNEL_INODE_TIMES
	ZFS_AC_KERNEL_INODE_LOCK
	ZFS_AC_KERNEL_GROUP_INFO_GID
	ZFS_AC_KERNEL_RW
	ZFS_AC_KERNEL_TIMER_SETUP
	ZFS_AC_KERNEL_SUPER_USER_NS
	ZFS_AC_KERNEL_PROC_OPERATIONS
	ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS
	ZFS_AC_KERNEL_BIO
	ZFS_AC_KERNEL_BLKDEV
	ZFS_AC_KERNEL_BLK_QUEUE
	ZFS_AC_KERNEL_GENHD_FLAGS
	ZFS_AC_KERNEL_REVALIDATE_DISK
	ZFS_AC_KERNEL_GET_DISK_RO
	ZFS_AC_KERNEL_GENERIC_READLINK_GLOBAL
	ZFS_AC_KERNEL_DISCARD_GRANULARITY
	ZFS_AC_KERNEL_INODE_OWNER_OR_CAPABLE
	ZFS_AC_KERNEL_XATTR
	ZFS_AC_KERNEL_ACL
	ZFS_AC_KERNEL_INODE_SETATTR
	ZFS_AC_KERNEL_INODE_GETATTR
	ZFS_AC_KERNEL_INODE_SET_FLAGS
	ZFS_AC_KERNEL_INODE_SET_IVERSION
	ZFS_AC_KERNEL_SHOW_OPTIONS
	ZFS_AC_KERNEL_FILE_INODE
	ZFS_AC_KERNEL_FILE_DENTRY
	ZFS_AC_KERNEL_FSYNC
	ZFS_AC_KERNEL_AIO_FSYNC
	ZFS_AC_KERNEL_EVICT_INODE
	ZFS_AC_KERNEL_DIRTY_INODE
	ZFS_AC_KERNEL_SHRINKER
	ZFS_AC_KERNEL_MKDIR
	ZFS_AC_KERNEL_LOOKUP_FLAGS
	ZFS_AC_KERNEL_CREATE
	ZFS_AC_KERNEL_PERMISSION
	ZFS_AC_KERNEL_GET_LINK
	ZFS_AC_KERNEL_PUT_LINK
	ZFS_AC_KERNEL_TMPFILE
	ZFS_AC_KERNEL_AUTOMOUNT
	ZFS_AC_KERNEL_ENCODE_FH_WITH_INODE
	ZFS_AC_KERNEL_COMMIT_METADATA
	ZFS_AC_KERNEL_CLEAR_INODE
	ZFS_AC_KERNEL_SETATTR_PREPARE
	ZFS_AC_KERNEL_INSERT_INODE_LOCKED
	ZFS_AC_KERNEL_DENTRY
	ZFS_AC_KERNEL_DENTRY_ALIAS_D_U
	ZFS_AC_KERNEL_TRUNCATE_SETSIZE
	ZFS_AC_KERNEL_SECURITY_INODE
	ZFS_AC_KERNEL_FST_MOUNT
	ZFS_AC_KERNEL_BDI
	ZFS_AC_KERNEL_SET_NLINK
	ZFS_AC_KERNEL_SGET
	ZFS_AC_KERNEL_LSEEK_EXECUTE
	ZFS_AC_KERNEL_VFS_FILEMAP_DIRTY_FOLIO
	ZFS_AC_KERNEL_VFS_READ_FOLIO
	ZFS_AC_KERNEL_VFS_GETATTR
	ZFS_AC_KERNEL_VFS_FSYNC_2ARGS
	ZFS_AC_KERNEL_VFS_ITERATE
	ZFS_AC_KERNEL_VFS_DIRECT_IO
	ZFS_AC_KERNEL_VFS_READPAGES
	ZFS_AC_KERNEL_VFS_SET_PAGE_DIRTY_NOBUFFERS
	ZFS_AC_KERNEL_VFS_RW_ITERATE
	ZFS_AC_KERNEL_VFS_GENERIC_WRITE_CHECKS
	ZFS_AC_KERNEL_VFS_IOV_ITER
	ZFS_AC_KERNEL_VFS_COPY_FILE_RANGE
	ZFS_AC_KERNEL_VFS_GENERIC_COPY_FILE_RANGE
	ZFS_AC_KERNEL_VFS_SPLICE_COPY_FILE_RANGE
	ZFS_AC_KERNEL_VFS_REMAP_FILE_RANGE
	ZFS_AC_KERNEL_VFS_CLONE_FILE_RANGE
	ZFS_AC_KERNEL_VFS_DEDUPE_FILE_RANGE
	ZFS_AC_KERNEL_VFS_FILE_OPERATIONS_EXTEND
	ZFS_AC_KERNEL_KMAP_ATOMIC_ARGS
	ZFS_AC_KERNEL_FOLLOW_DOWN_ONE
	ZFS_AC_KERNEL_MAKE_REQUEST_FN
	ZFS_AC_KERNEL_GENERIC_IO_ACCT
	ZFS_AC_KERNEL_FPU
	ZFS_AC_KERNEL_FMODE_T
	ZFS_AC_KERNEL_KUIDGID_T
	ZFS_AC_KERNEL_KUID_HELPERS
	ZFS_AC_KERNEL_RENAME
	ZFS_AC_KERNEL_CURRENT_TIME
	ZFS_AC_KERNEL_USERNS_CAPABILITIES
	ZFS_AC_KERNEL_IN_COMPAT_SYSCALL
	ZFS_AC_KERNEL_KTIME
	ZFS_AC_KERNEL_TOTALRAM_PAGES_FUNC
	ZFS_AC_KERNEL_TOTALHIGH_PAGES
	ZFS_AC_KERNEL_KSTRTOUL
	ZFS_AC_KERNEL_PERCPU
	ZFS_AC_KERNEL_CPU_HOTPLUG
	ZFS_AC_KERNEL_GENERIC_FILLATTR
	ZFS_AC_KERNEL_MKNOD
	ZFS_AC_KERNEL_SYMLINK
	ZFS_AC_KERNEL_BIO_MAX_SEGS
	ZFS_AC_KERNEL_SIGNAL_STOP
	ZFS_AC_KERNEL_SIGINFO
	ZFS_AC_KERNEL_SYSFS
	ZFS_AC_KERNEL_SET_SPECIAL_STATE
	ZFS_AC_KERNEL_STANDALONE_LINUX_STDARG
	ZFS_AC_KERNEL_STRLCPY
	ZFS_AC_KERNEL_STRSCPY
	ZFS_AC_KERNEL_PAGEMAP_FOLIO_WAIT_BIT
	ZFS_AC_KERNEL_ADD_DISK
	ZFS_AC_KERNEL_KTHREAD
	ZFS_AC_KERNEL_ZERO_PAGE
	ZFS_AC_KERNEL___COPY_FROM_USER_INATOMIC
	ZFS_AC_KERNEL_USER_NS_COMMON_INUM
	ZFS_AC_KERNEL_IDMAP_MNT_API
	ZFS_AC_KERNEL_IDMAP_NO_USERNS
	ZFS_AC_KERNEL_IATTR_VFSID
	ZFS_AC_KERNEL_FILEMAP
	ZFS_AC_KERNEL_WRITEPAGE_T
	ZFS_AC_KERNEL_RECLAIMED
	ZFS_AC_KERNEL_REGISTER_SYSCTL_TABLE
	ZFS_AC_KERNEL_COPY_SPLICE_READ
	ZFS_AC_KERNEL_SYNC_BDEV
	ZFS_AC_KERNEL_MM_PAGE_SIZE
	case "$host_cpu" in
	powerpc*)
	ZFS_AC_KERNEL_CPU_HAS_FEATURE
	ZFS_AC_KERNEL_FLUSH_DCACHE_PAGE
	;;
	riscv*)
	ZFS_AC_KERNEL_FLUSH_DCACHE_PAGE
	;;
	esac
	])

	dnl #
	dnl # Detect name used for Module.symvers file in kernel
	dnl #
	AC_DEFUN([ZFS_AC_MODULE_SYMVERS], [
	modpost=$LINUX/scripts/Makefile.modpost
	AC_MSG_CHECKING([kernel file name for module symbols])
	AS_IF([test "x$enable_linux_builtin" != xyes -a -f "$modpost"], [
	AS_IF([grep -q Modules.symvers $modpost], [
	LINUX_SYMBOLS=Modules.symvers
	], [
	LINUX_SYMBOLS=Module.symvers
	])

	AS_IF([test ! -f "$LINUX_OBJ/$LINUX_SYMBOLS"], [
	AC_MSG_ERROR([
	*** Please make sure the kernel devel package for your distribution
	*** is installed. If you are building with a custom kernel, make sure
	*** the kernel is configured, built, and the '--with-linux=PATH'
	*** configure option refers to the location of the kernel source.
	])
	])
	], [
	LINUX_SYMBOLS=NONE
	])
	AC_MSG_RESULT($LINUX_SYMBOLS)
	AC_SUBST(LINUX_SYMBOLS)
	])

	dnl #
	dnl # Detect the kernel to be built against
	dnl #
	dnl # Most modern Linux distributions have separate locations for bare
	dnl # source (source) and prebuilt (build) files. Additionally, there are
	dnl # `source` and `build` symlinks in `/lib/modules/$(KERNEL_VERSION)`
	dnl # pointing to them. The directory search order is now:
	dnl #
	dnl # - `configure` command line values if both `--with-linux` and
	dnl # `--with-linux-obj` were defined
	dnl #
	dnl # - If only `--with-linux` was defined, `--with-linux-obj` is assumed
	dnl # to have the same value as `--with-linux`
	dnl #
	dnl # - If neither `--with-linux` nor `--with-linux-obj` were defined
	dnl # autodetection is used:
	dnl #
	dnl # - `/lib/modules/$(uname -r)/{source,build}` respectively, if exist.
	dnl #
	dnl # - If only `/lib/modules/$(uname -r)/build` exists, it is assumed
	dnl # to be both source and build directory.
	dnl #
	dnl # - The first directory in `/lib/modules` with the highest version
	dnl # number according to `sort -V` which contains both `source` and
	dnl # `build` symlinks/directories. If module directory contains only
	dnl # `build` component, it is assumed to be both source and build
	dnl # directory.
	dnl #
	dnl # - Last resort: the first directory matching `/usr/src/kernels/*`
	dnl # and `/usr/src/linux-*` with the highest version number according
	dnl # to `sort -V` is assumed to be both source and build directory.
	dnl #
	AC_DEFUN([ZFS_AC_KERNEL], [
	AC_ARG_WITH([linux],
	AS_HELP_STRING([--with-linux=PATH],
	[Path to kernel source]),
	[kernelsrc="$withval"])

	AC_ARG_WITH(linux-obj,
	AS_HELP_STRING([--with-linux-obj=PATH],
	[Path to kernel build objects]),
	[kernelbuild="$withval"])

	AC_MSG_CHECKING([kernel source and build directories])
	AS_IF([test -n "$kernelsrc" && test -z "$kernelbuild"], [
	kernelbuild="$kernelsrc"
	], [test -z "$kernelsrc"], [
	AS_IF([test -e "/lib/modules/$(uname -r)/source" && \
	test -e "/lib/modules/$(uname -r)/build"], [
	src="/lib/modules/$(uname -r)/source"
	build="/lib/modules/$(uname -r)/build"
	], [test -e "/lib/modules/$(uname -r)/build"], [
	build="/lib/modules/$(uname -r)/build"
	src="$build"
	], [
	src=

	for d in $(ls -1d /lib/modules/* 2>/dev/null \| sort -Vr); do
	if test -e "$d/source" && test -e "$d/build"; then
	src="$d/source"
	build="$d/build"
	break
	fi

	if test -e "$d/build"; then
	src="$d/build"
	build="$d/build"
	break
	fi
	done

	# the least reliable method
	if test -z "$src"; then
	src=$(ls -1d /usr/src/kernels/* /usr/src/linux-* \
	2>/dev/null \| grep -v obj \| sort -Vr \| head -1)
	build="$src"
	fi
	])

	AS_IF([test -n "$src" && test -e "$src"], [
	kernelsrc=$(readlink -e "$src")
	], [
	kernelsrc="[Not found]"
	])
	AS_IF([test -n "$build" && test -e "$build"], [
	kernelbuild=$(readlink -e "$build")
	], [
	kernelbuild="[Not found]"
	])
	], [
	AS_IF([test "$kernelsrc" = "NONE"], [
	kernsrcver=NONE
	])
	withlinux=yes
	])

	AC_MSG_RESULT([done])
	AC_MSG_CHECKING([kernel source directory])
	AC_MSG_RESULT([$kernelsrc])
	AC_MSG_CHECKING([kernel build directory])
	AC_MSG_RESULT([$kernelbuild])
	AS_IF([test ! -d "$kernelsrc" \|\| test ! -d "$kernelbuild"], [
	AC_MSG_ERROR([
	*** Please make sure the kernel devel package for your distribution
	*** is installed and then try again. If that fails, you can specify the
	*** location of the kernel source and build with the '--with-linux=PATH' and
	*** '--with-linux-obj=PATH' options respectively.])
	])

	AC_MSG_CHECKING([kernel source version])
	utsrelease1=$kernelbuild/include/linux/version.h
	utsrelease2=$kernelbuild/include/linux/utsrelease.h
	utsrelease3=$kernelbuild/include/generated/utsrelease.h
	AS_IF([test -r $utsrelease1 && grep -qF UTS_RELEASE $utsrelease1], [
	utsrelease=$utsrelease1
	], [test -r $utsrelease2 && grep -qF UTS_RELEASE $utsrelease2], [
	utsrelease=$utsrelease2
	], [test -r $utsrelease3 && grep -qF UTS_RELEASE $utsrelease3], [
	utsrelease=$utsrelease3
	])

	AS_IF([test -n "$utsrelease"], [
	kernsrcver=$($AWK '/UTS_RELEASE/ { gsub(/"/, "", $[3]); print $[3] }' $utsrelease)
	AS_IF([test -z "$kernsrcver"], [
	AC_MSG_RESULT([Not found])
	AC_MSG_ERROR([
	*** Cannot determine kernel version.
	])
	])
	], [
	AC_MSG_RESULT([Not found])
	if test "x$enable_linux_builtin" != xyes; then
	AC_MSG_ERROR([
	*** Cannot find UTS_RELEASE definition.
	])
	else
	AC_MSG_ERROR([
	*** Cannot find UTS_RELEASE definition.
	*** Please run 'make prepare' inside the kernel source tree.])
	fi
	])

	AC_MSG_RESULT([$kernsrcver])

	AS_VERSION_COMPARE([$kernsrcver], [$ZFS_META_KVER_MIN], [
	AC_MSG_ERROR([
	*** Cannot build against kernel version $kernsrcver.
	*** The minimum supported kernel version is $ZFS_META_KVER_MIN.
	])
	])

	LINUX=${kernelsrc}
	LINUX_OBJ=${kernelbuild}
	LINUX_VERSION=${kernsrcver}

	AC_SUBST(LINUX)
	AC_SUBST(LINUX_OBJ)
	AC_SUBST(LINUX_VERSION)
	])

	dnl #
	dnl # Detect the QAT module to be built against, QAT provides hardware
	dnl # acceleration for data compression:
	dnl #
	dnl # https://01.org/intel-quickassist-technology
	dnl #
	dnl # 1) Download and install QAT driver from the above link
	dnl # 2) Start QAT driver in your system:
	dnl # service qat_service start
	dnl # 3) Enable QAT in ZFS, e.g.:
	dnl # ./configure --with-qat=<qat-driver-path>/QAT1.6
	dnl # make
	dnl # 4) Set GZIP compression in ZFS dataset:
	dnl # zfs set compression = gzip <dataset>
	dnl #
	dnl # Then the data written to this ZFS pool is compressed by QAT accelerator
	dnl # automatically, and de-compressed by QAT when read from the pool.
	dnl #
	dnl # 1) Get QAT hardware statistics with:
	dnl # cat /proc/icp_dh895xcc_dev/qat
	dnl # 2) To disable QAT:
	dnl # insmod zfs.ko zfs_qat_disable=1
	dnl #
	AC_DEFUN([ZFS_AC_QAT], [
	AC_ARG_WITH([qat],
	AS_HELP_STRING([--with-qat=PATH],
	[Path to qat source]),
	AS_IF([test "$withval" = "yes"],
	AC_MSG_ERROR([--with-qat=PATH requires a PATH]),
	[qatsrc="$withval"]))

	AC_ARG_WITH([qat-obj],
	AS_HELP_STRING([--with-qat-obj=PATH],
	[Path to qat build objects]),
	[qatbuild="$withval"])

	AS_IF([test ! -z "${qatsrc}"], [
	AC_MSG_CHECKING([qat source directory])
	AC_MSG_RESULT([$qatsrc])
	QAT_SRC="${qatsrc}/quickassist"
	AS_IF([ test ! -e "$QAT_SRC/include/cpa.h"], [
	AC_MSG_ERROR([
	*** Please make sure the qat driver package is installed
	*** and specify the location of the qat source with the
	*** '--with-qat=PATH' option then try again. Failed to
	*** find cpa.h in:
	${QAT_SRC}/include])
	])
	])

	AS_IF([test ! -z "${qatsrc}"], [
	AC_MSG_CHECKING([qat build directory])
	AS_IF([test -z "$qatbuild"], [
	qatbuild="${qatsrc}/build"
	])

	AC_MSG_RESULT([$qatbuild])
	QAT_OBJ=${qatbuild}
	AS_IF([ ! test -e "$QAT_OBJ/icp_qa_al.ko" && ! test -e "$QAT_OBJ/qat_api.ko"], [
	AC_MSG_ERROR([
	*** Please make sure the qat driver is installed then try again.
	*** Failed to find icp_qa_al.ko or qat_api.ko in:
	$QAT_OBJ])
	])

	AC_SUBST(QAT_SRC)
	AC_SUBST(QAT_OBJ)

	AC_DEFINE(HAVE_QAT, 1,
	[qat is enabled and existed])
	])

	dnl #
	dnl # Detect the name used for the QAT Module.symvers file.
	dnl #
	AS_IF([test ! -z "${qatsrc}"], [
	AC_MSG_CHECKING([qat file for module symbols])
	QAT_SYMBOLS=$QAT_SRC/lookaside/access_layer/src/Module.symvers

	AS_IF([test -r $QAT_SYMBOLS], [
	AC_MSG_RESULT([$QAT_SYMBOLS])
	AC_SUBST(QAT_SYMBOLS)
	],[
	AC_MSG_ERROR([
	*** Please make sure the qat driver is installed then try again.
	*** Failed to find Module.symvers in:
	$QAT_SYMBOLS
	])
	])
	])
	])

	dnl #
	dnl # ZFS_LINUX_CONFTEST_H
	dnl #
	AC_DEFUN([ZFS_LINUX_CONFTEST_H], [
	test -d build/$2 \|\| mkdir -p build/$2
	cat - <<_ACEOF >build/$2/$2.h
	$1
	_ACEOF
	])

	dnl #
	dnl # ZFS_LINUX_CONFTEST_C
	dnl #
	AC_DEFUN([ZFS_LINUX_CONFTEST_C], [
	test -d build/$2 \|\| mkdir -p build/$2
	cat confdefs.h - <<_ACEOF >build/$2/$2.c
	$1
	_ACEOF
	])

	dnl #
	dnl # ZFS_LINUX_CONFTEST_MAKEFILE
	dnl #
	dnl # $1 - test case name
	dnl # $2 - add to top-level Makefile
	dnl # $3 - additional build flags
	dnl #
	AC_DEFUN([ZFS_LINUX_CONFTEST_MAKEFILE], [
	test -d build \|\| mkdir -p build
	test -d build/$1 \|\| mkdir -p build/$1

	file=build/$1/Makefile

	dnl # Example command line to manually build source.
	cat - <<_ACEOF >$file
	# Example command line to manually build source
	# make modules -C $LINUX_OBJ $ARCH_UM M=$PWD/build/$1

	ccflags-y := -Werror $FRAME_LARGER_THAN
	_ACEOF

	dnl # Additional custom CFLAGS as requested.
	m4_ifval($3, [echo "ccflags-y += $3" >>$file], [])

	dnl # Test case source
	echo "obj-m := $1.o" >>$file

	AS_IF([test "x$2" = "xyes"], [echo "obj-m += $1/" >>build/Makefile], [])
	])

	dnl #
	dnl # ZFS_LINUX_TEST_PROGRAM(C)([PROLOGUE], [BODY])
	dnl #
	m4_define([ZFS_LINUX_TEST_PROGRAM], [
	#include <linux/module.h>
	$1

	int
	main (void)
	{
	$2
	;
	return 0;
	}

	MODULE_DESCRIPTION("conftest");
	MODULE_AUTHOR(ZFS_META_AUTHOR);
	MODULE_VERSION(ZFS_META_VERSION "-" ZFS_META_RELEASE);
	MODULE_LICENSE($3);
	])

	dnl #
	dnl # ZFS_LINUX_TEST_REMOVE
	dnl #
	dnl # Removes the specified test source and results.
	dnl #
	AC_DEFUN([ZFS_LINUX_TEST_REMOVE], [
	test -d build/$1 && rm -Rf build/$1
	test -f build/Makefile && sed '/$1/d' build/Makefile
	])

	dnl #
	dnl # ZFS_LINUX_COMPILE
	dnl #
	dnl # $1 - build dir
	dnl # $2 - test command
	dnl # $3 - pass command
	dnl # $4 - fail command
	dnl # $5 - set KBUILD_MODPOST_NOFINAL='yes'
	dnl # $6 - set KBUILD_MODPOST_WARN='yes'
	dnl #
	dnl # Used internally by ZFS_LINUX_TEST_{COMPILE,MODPOST}
	dnl #
	AC_DEFUN([ZFS_LINUX_COMPILE], [
	AC_ARG_VAR([KERNEL_CC], [C compiler for
	building kernel modules])
	AC_ARG_VAR([KERNEL_LD], [Linker for
	building kernel modules])
	AC_ARG_VAR([KERNEL_LLVM], [Binary option to
	build kernel modules with LLVM/CLANG toolchain])
	AC_TRY_COMMAND([
	KBUILD_MODPOST_NOFINAL="$5" KBUILD_MODPOST_WARN="$6"
	make modules -k -j$TEST_JOBS ${KERNEL_CC:+CC=$KERNEL_CC}
	${KERNEL_LD:+LD=$KERNEL_LD} ${KERNEL_LLVM:+LLVM=$KERNEL_LLVM}
	CONFIG_MODULES=y CFLAGS_MODULE=-DCONFIG_MODULES
	-C $LINUX_OBJ $ARCH_UM M=$PWD/$1 >$1/build.log 2>&1])
	AS_IF([AC_TRY_COMMAND([$2])], [$3], [$4])
	])

	dnl #
	dnl # ZFS_LINUX_TEST_COMPILE
	dnl #
	dnl # Perform a full compile excluding the final modpost phase.
	dnl #
	AC_DEFUN([ZFS_LINUX_TEST_COMPILE], [
	ZFS_LINUX_COMPILE([$2], [test -f $2/build.log], [
	mv $2/Makefile $2/Makefile.compile.$1
	mv $2/build.log $2/build.log.$1
	],[
	AC_MSG_ERROR([
	*** Unable to compile test source to determine kernel interfaces.])
	], [yes], [])
	])

	dnl #
	dnl # ZFS_LINUX_TEST_MODPOST
	dnl #
	dnl # Perform a full compile including the modpost phase. This may
	dnl # be an incremental build if the objects have already been built.
	dnl #
	AC_DEFUN([ZFS_LINUX_TEST_MODPOST], [
	ZFS_LINUX_COMPILE([$2], [test -f $2/build.log], [
	mv $2/Makefile $2/Makefile.modpost.$1
	cat $2/build.log >>build/build.log.$1
	],[
	AC_MSG_ERROR([
	*** Unable to modpost test source to determine kernel interfaces.])
	], [], [yes])
	])

	dnl #
	dnl # Perform the compilation of the test cases in two phases.
	dnl #
	dnl # Phase 1) attempt to build the object files for all of the tests
	dnl # defined by the ZFS_LINUX_TEST_SRC macro. But do not
	dnl # perform the final modpost stage.
	dnl #
	dnl # Phase 2) disable all tests which failed the initial compilation,
	dnl # then invoke the final modpost step for the remaining tests.
	dnl #
	dnl # This allows us efficiently build the test cases in parallel while
	dnl # remaining resilient to build failures which are expected when
	dnl # detecting the available kernel interfaces.
	dnl #
	dnl # The maximum allowed parallelism can be controlled by setting the
	dnl # TEST_JOBS environment variable. Otherwise, it default to $(nproc).
	dnl #
	AC_DEFUN([ZFS_LINUX_TEST_COMPILE_ALL], [
	dnl # Phase 1 - Compilation only, final linking is skipped.
	ZFS_LINUX_TEST_COMPILE([$1], [build])

	dnl #
	dnl # Phase 2 - When building external modules disable test cases
	dnl # which failed to compile and invoke modpost to verify the
	dnl # final linking.
	dnl #
	dnl # Test names suffixed with '_license' call modpost independently
	dnl # to ensure that a single incompatibility does not result in the
	dnl # modpost phase exiting early. This check is not performed on
	dnl # every symbol since the majority are compatible and doing so
	dnl # would significantly slow down this phase.
	dnl #
	dnl # When configuring for builtin (--enable-linux-builtin)
	dnl # fake the linking step artificially create the expected .ko
	dnl # files for tests which did compile. This is required for
	dnl # kernels which do not have loadable module support or have
	dnl # not yet been built.
	dnl #
	AS_IF([test "x$enable_linux_builtin" = "xno"], [
	for dir in $(awk '/^obj-m/ { print [$]3 }' \
	build/Makefile.compile.$1); do
	name=${dir%/}
	AS_IF([test -f build/$name/$name.o], [
	AS_IF([test "${name##*_}" = "license"], [
	ZFS_LINUX_TEST_MODPOST([$1],
	[build/$name])
	echo "obj-n += $dir" >>build/Makefile
	], [
	echo "obj-m += $dir" >>build/Makefile
	])
	], [
	echo "obj-n += $dir" >>build/Makefile
	])
	done

	ZFS_LINUX_TEST_MODPOST([$1], [build])
	], [
	for dir in $(awk '/^obj-m/ { print [$]3 }' \
	build/Makefile.compile.$1); do
	name=${dir%/}
	AS_IF([test -f build/$name/$name.o], [
	touch build/$name/$name.ko
	])
	done
	])
	])

	dnl #
	dnl # ZFS_LINUX_TEST_SRC
	dnl #
	dnl # $1 - name
	dnl # $2 - global
	dnl # $3 - source
	dnl # $4 - extra cflags
	dnl # $5 - check license-compatibility
	dnl #
	dnl # Check if the test source is buildable at all and then if it is
	dnl # license compatible.
	dnl #
	dnl # N.B because all of the test cases are compiled in parallel they
	dnl # must never depend on the results of previous tests. Each test
	dnl # needs to be entirely independent.
	dnl #
	AC_DEFUN([ZFS_LINUX_TEST_SRC], [
	ZFS_LINUX_CONFTEST_C([ZFS_LINUX_TEST_PROGRAM([[$2]], [[$3]],
	[["Dual BSD/GPL"]])], [$1])
	ZFS_LINUX_CONFTEST_MAKEFILE([$1], [yes], [$4])

	AS_IF([ test -n "$5" ], [
	ZFS_LINUX_CONFTEST_C([ZFS_LINUX_TEST_PROGRAM(
	[[$2]], [[$3]], [[$5]])], [$1_license])
	ZFS_LINUX_CONFTEST_MAKEFILE([$1_license], [yes], [$4])
	])
	])

	dnl #
	dnl # ZFS_LINUX_TEST_RESULT
	dnl #
	dnl # $1 - name of a test source (ZFS_LINUX_TEST_SRC)
	dnl # $2 - run on success (valid .ko generated)
	dnl # $3 - run on failure (unable to compile)
	dnl #
	AC_DEFUN([ZFS_LINUX_TEST_RESULT], [
	AS_IF([test -d build/$1], [
	AS_IF([test -f build/$1/$1.ko], [$2], [$3])
	], [
	AC_MSG_ERROR([
	*** No matching source for the "$1" test, check that
	*** both the test source and result macros refer to the same name.
	])
	])
	])

	dnl #
	dnl # ZFS_LINUX_TEST_ERROR
	dnl #
	dnl # Generic error message which can be used when none of the expected
	dnl # kernel interfaces were detected.
	dnl #
	AC_DEFUN([ZFS_LINUX_TEST_ERROR], [
	AC_MSG_ERROR([
	*** None of the expected "$1" interfaces were detected.
	*** This may be because your kernel version is newer than what is
	*** supported, or you are using a patched custom kernel with
	*** incompatible modifications.
	***
	*** ZFS Version: $ZFS_META_ALIAS
	*** Compatible Kernels: $ZFS_META_KVER_MIN - $ZFS_META_KVER_MAX
	])
	])

	dnl #
	dnl # ZFS_LINUX_TEST_RESULT_SYMBOL
	dnl #
	dnl # Like ZFS_LINUX_TEST_RESULT except ZFS_CHECK_SYMBOL_EXPORT is called to
	dnl # verify symbol exports, unless --enable-linux-builtin was provided to
	dnl # configure.
	dnl #
	AC_DEFUN([ZFS_LINUX_TEST_RESULT_SYMBOL], [
	AS_IF([ ! test -f build/$1/$1.ko], [
	$5
	], [
	AS_IF([test "x$enable_linux_builtin" != "xyes"], [
	ZFS_CHECK_SYMBOL_EXPORT([$2], [$3], [$4], [$5])
	], [
	$4
	])
	])
	])

	dnl #
	dnl # ZFS_LINUX_COMPILE_IFELSE
	dnl #
	AC_DEFUN([ZFS_LINUX_COMPILE_IFELSE], [
	ZFS_LINUX_TEST_REMOVE([conftest])

	m4_ifvaln([$1], [ZFS_LINUX_CONFTEST_C([$1], [conftest])])
	m4_ifvaln([$5], [ZFS_LINUX_CONFTEST_H([$5], [conftest])],
	[ZFS_LINUX_CONFTEST_H([], [conftest])])

	ZFS_LINUX_CONFTEST_MAKEFILE([conftest], [no],
	[m4_ifvaln([$5], [-I$PWD/build/conftest], [])])
	ZFS_LINUX_COMPILE([build/conftest], [$2], [$3], [$4], [], [])
	])

	dnl #
	dnl # ZFS_LINUX_TRY_COMPILE
	dnl #
	dnl # $1 - global
	dnl # $2 - source
	dnl # $3 - run on success (valid .ko generated)
	dnl # $4 - run on failure (unable to compile)
	dnl #
	dnl # When configuring as builtin (--enable-linux-builtin) for kernels
	dnl # without loadable module support (CONFIG_MODULES=n) only the object
	dnl # file is created. See ZFS_LINUX_TEST_COMPILE_ALL for details.
	dnl #
	AC_DEFUN([ZFS_LINUX_TRY_COMPILE], [
	AS_IF([test "x$enable_linux_builtin" = "xyes"], [
	ZFS_LINUX_COMPILE_IFELSE(
	[ZFS_LINUX_TEST_PROGRAM([[$1]], [[$2]],
	[[ZFS_META_LICENSE]])],
	[test -f build/conftest/conftest.o], [$3], [$4])
	], [
	ZFS_LINUX_COMPILE_IFELSE(
	[ZFS_LINUX_TEST_PROGRAM([[$1]], [[$2]],
	[[ZFS_META_LICENSE]])],
	[test -f build/conftest/conftest.ko], [$3], [$4])
	])
	])

	dnl #
	dnl # ZFS_CHECK_SYMBOL_EXPORT
	dnl #
	dnl # Check if a symbol is exported on not by consulting the symbols
	dnl # file, or optionally the source code.
	dnl #
	AC_DEFUN([ZFS_CHECK_SYMBOL_EXPORT], [
	grep -q -E '[[[:space:]]]$1[[[:space:]]]' \
	$LINUX_OBJ/$LINUX_SYMBOLS 2>/dev/null
	rc=$?
	if test $rc -ne 0; then
	export=0
	for file in $2; do
	grep -q -E "EXPORT_SYMBOL.*($1)" \
	"$LINUX/$file" 2>/dev/null
	rc=$?
	if test $rc -eq 0; then
	export=1
	break;
	fi
	done
	if test $export -eq 0; then :
	$4
	else :
	$3
	fi
	else :
	$3
	fi
	])

	dnl #
	dnl # ZFS_LINUX_TRY_COMPILE_SYMBOL
	dnl #
	dnl # Like ZFS_LINUX_TRY_COMPILER except ZFS_CHECK_SYMBOL_EXPORT is called
	dnl # to verify symbol exports, unless --enable-linux-builtin was provided
	dnl # to configure.
	dnl #
	AC_DEFUN([ZFS_LINUX_TRY_COMPILE_SYMBOL], [
	ZFS_LINUX_TRY_COMPILE([$1], [$2], [rc=0], [rc=1])
	if test $rc -ne 0; then :
	$6
	else
	if test "x$enable_linux_builtin" != xyes; then
	ZFS_CHECK_SYMBOL_EXPORT([$3], [$4], [rc=0], [rc=1])
	fi
	if test $rc -ne 0; then :
	$6
	else :
	$5
	fi
	fi
	])

	dnl #
	dnl # ZFS_LINUX_TRY_COMPILE_HEADER
	dnl # like ZFS_LINUX_TRY_COMPILE, except the contents conftest.h are
	dnl # provided via the fifth parameter
	dnl #
	AC_DEFUN([ZFS_LINUX_TRY_COMPILE_HEADER], [
	AS_IF([test "x$enable_linux_builtin" = "xyes"], [
	ZFS_LINUX_COMPILE_IFELSE(
	[ZFS_LINUX_TEST_PROGRAM([[$1]], [[$2]],
	[[ZFS_META_LICENSE]])],
	[test -f build/conftest/conftest.o], [$3], [$4], [$5])
	], [
	ZFS_LINUX_COMPILE_IFELSE(
	[ZFS_LINUX_TEST_PROGRAM([[$1]], [[$2]],
	[[ZFS_META_LICENSE]])],
	[test -f build/conftest/conftest.ko], [$3], [$4], [$5])
	])
	])

	dnl #
	dnl # AS_VERSION_COMPARE_LE
	dnl # like AS_VERSION_COMPARE_LE, but runs $3 if (and only if) $1 <= $2
	dnl # AS_VERSION_COMPARE_LE (version-1, version-2, [action-if-less-or-equal], [action-if-greater])
	dnl #
	AC_DEFUN([AS_VERSION_COMPARE_LE], [
	AS_VERSION_COMPARE([$1], [$2], [$3], [$3], [$4])
	])

	dnl #
	dnl # ZFS_LINUX_REQUIRE_API
	dnl # like ZFS_LINUX_TEST_ERROR, except only fails if the kernel is
	dnl # at least some specified version.
	dnl #
	AC_DEFUN([ZFS_LINUX_REQUIRE_API], [
	AS_VERSION_COMPARE_LE([$2], [$kernsrcver], [
	AC_MSG_ERROR([
	*** None of the expected "$1" interfaces were detected. This
	*** interface is expected for kernels version "$2" and above.
	*** This may be because your kernel version is newer than what is
	*** supported, or you are using a patched custom kernel with
	*** incompatible modifications. Newer kernels may have incompatible
	*** APIs.
	***
	*** ZFS Version: $ZFS_META_ALIAS
	*** Compatible Kernels: $ZFS_META_KVER_MIN - $ZFS_META_KVER_MAX
	])
	], [
	AC_MSG_RESULT(no)
	])
	])
	diff --git a/sys/contrib/openzfs/config/zfs-build.m4 b/sys/contrib/openzfs/config/zfs-build.m4
	index bb5a85d815d1..368684e1c512 100644
	--- a/sys/contrib/openzfs/config/zfs-build.m4
	+++ b/sys/contrib/openzfs/config/zfs-build.m4
	@@ -1,649 +1,651 @@
	AC_DEFUN([ZFS_AC_LICENSE], [
	AC_MSG_CHECKING([zfs author])
	AC_MSG_RESULT([$ZFS_META_AUTHOR])

	AC_MSG_CHECKING([zfs license])
	AC_MSG_RESULT([$ZFS_META_LICENSE])
	])

	AC_DEFUN([ZFS_AC_DEBUG_ENABLE], [
	DEBUG_CFLAGS="-Werror"
	DEBUG_CPPFLAGS="-DDEBUG -UNDEBUG"
	DEBUG_LDFLAGS=""
	DEBUG_ZFS="_with_debug"
	WITH_DEBUG="true"
	AC_DEFINE(ZFS_DEBUG, 1, [zfs debugging enabled])

	KERNEL_DEBUG_CFLAGS="-Werror"
	KERNEL_DEBUG_CPPFLAGS="-DDEBUG -UNDEBUG"
	])

	AC_DEFUN([ZFS_AC_DEBUG_DISABLE], [
	DEBUG_CFLAGS=""
	DEBUG_CPPFLAGS="-UDEBUG -DNDEBUG"
	DEBUG_LDFLAGS=""
	DEBUG_ZFS="_without_debug"
	WITH_DEBUG=""

	KERNEL_DEBUG_CFLAGS=""
	KERNEL_DEBUG_CPPFLAGS="-UDEBUG -DNDEBUG"
	])

	dnl #
	dnl # When debugging is enabled:
	dnl # - Enable all ASSERTs (-DDEBUG)
	dnl # - Promote all compiler warnings to errors (-Werror)
	dnl #
	dnl # (If INVARIANTS is detected, we need to force DEBUG, or strange panics
	dnl # can ensue.)
	dnl #
	AC_DEFUN([ZFS_AC_DEBUG], [
	AC_MSG_CHECKING([whether assertion support will be enabled])
	AC_ARG_ENABLE([debug],
	[AS_HELP_STRING([--enable-debug],
	[Enable compiler and code assertions @<:@default=no@:>@])],
	[],
	[enable_debug=no])

	AS_CASE(["x$enable_debug"],
	["xyes"],
	[ZFS_AC_DEBUG_ENABLE],
	["xno"],
	[ZFS_AC_DEBUG_DISABLE],
	[AC_MSG_ERROR([Unknown option $enable_debug])])

	AS_CASE(["x$enable_invariants"],
	["xyes"],
	[],
	["xno"],
	[],
	[ZFS_AC_DEBUG_INVARIANTS_DETECT])

	AS_CASE(["x$enable_invariants"],
	["xyes"],
	[ZFS_AC_DEBUG_ENABLE],
	["xno"],
	[],
	[AC_MSG_ERROR([Unknown option $enable_invariants])])

	AC_SUBST(DEBUG_CFLAGS)
	AC_SUBST(DEBUG_CPPFLAGS)
	AC_SUBST(DEBUG_LDFLAGS)
	AC_SUBST(DEBUG_ZFS)
	AC_SUBST(WITH_DEBUG)

	AC_SUBST(KERNEL_DEBUG_CFLAGS)
	AC_SUBST(KERNEL_DEBUG_CPPFLAGS)

	AC_MSG_RESULT([$enable_debug])
	])

	AC_DEFUN([ZFS_AC_DEBUGINFO_ENABLE], [
	DEBUG_CFLAGS="$DEBUG_CFLAGS -g -fno-inline $NO_IPA_SRA"

	KERNEL_DEBUG_CFLAGS="$KERNEL_DEBUG_CFLAGS -fno-inline $KERNEL_NO_IPA_SRA"
	KERNEL_MAKE="$KERNEL_MAKE CONFIG_DEBUG_INFO=y"

	DEBUGINFO_ZFS="_with_debuginfo"
	])

	AC_DEFUN([ZFS_AC_DEBUGINFO_DISABLE], [
	DEBUGINFO_ZFS="_without_debuginfo"
	])

	AC_DEFUN([ZFS_AC_DEBUGINFO], [
	AC_MSG_CHECKING([whether debuginfo support will be forced])
	AC_ARG_ENABLE([debuginfo],
	[AS_HELP_STRING([--enable-debuginfo],
	[Force generation of debuginfo @<:@default=no@:>@])],
	[],
	[enable_debuginfo=no])

	AS_CASE(["x$enable_debuginfo"],
	["xyes"],
	[ZFS_AC_DEBUGINFO_ENABLE],
	["xno"],
	[ZFS_AC_DEBUGINFO_DISABLE],
	[AC_MSG_ERROR([Unknown option $enable_debuginfo])])

	AC_SUBST(DEBUG_CFLAGS)
	AC_SUBST(DEBUGINFO_ZFS)

	AC_SUBST(KERNEL_DEBUG_CFLAGS)
	AC_SUBST(KERNEL_MAKE)

	AC_MSG_RESULT([$enable_debuginfo])
	])

	dnl #
	dnl # Disabled by default, provides basic memory tracking. Track the total
	dnl # number of bytes allocated with kmem_alloc() and freed with kmem_free().
	dnl # Then at module unload time if any bytes were leaked it will be reported
	dnl # on the console.
	dnl #
	AC_DEFUN([ZFS_AC_DEBUG_KMEM], [
	AC_MSG_CHECKING([whether basic kmem accounting is enabled])
	AC_ARG_ENABLE([debug-kmem],
	[AS_HELP_STRING([--enable-debug-kmem],
	[Enable basic kmem accounting @<:@default=no@:>@])],
	[],
	[enable_debug_kmem=no])

	AS_IF([test "x$enable_debug_kmem" = xyes], [
	KERNEL_DEBUG_CPPFLAGS="${KERNEL_DEBUG_CPPFLAGS} -DDEBUG_KMEM"
	DEBUG_KMEM_ZFS="_with_debug_kmem"
	], [
	DEBUG_KMEM_ZFS="_without_debug_kmem"
	])

	AC_SUBST(KERNEL_DEBUG_CPPFLAGS)
	AC_SUBST(DEBUG_KMEM_ZFS)

	AC_MSG_RESULT([$enable_debug_kmem])
	])

	dnl #
	dnl # Disabled by default, provides detailed memory tracking. This feature
	dnl # also requires --enable-debug-kmem to be set. When enabled not only will
	dnl # total bytes be tracked but also the location of every kmem_alloc() and
	dnl # kmem_free(). When the module is unloaded a list of all leaked addresses
	dnl # and where they were allocated will be dumped to the console. Enabling
	dnl # this feature has a significant impact on performance but it makes finding
	dnl # memory leaks straight forward.
	dnl #
	AC_DEFUN([ZFS_AC_DEBUG_KMEM_TRACKING], [
	AC_MSG_CHECKING([whether detailed kmem tracking is enabled])
	AC_ARG_ENABLE([debug-kmem-tracking],
	[AS_HELP_STRING([--enable-debug-kmem-tracking],
	[Enable detailed kmem tracking @<:@default=no@:>@])],
	[],
	[enable_debug_kmem_tracking=no])

	AS_IF([test "x$enable_debug_kmem_tracking" = xyes], [
	KERNEL_DEBUG_CPPFLAGS="${KERNEL_DEBUG_CPPFLAGS} -DDEBUG_KMEM_TRACKING"
	DEBUG_KMEM_TRACKING_ZFS="_with_debug_kmem_tracking"
	], [
	DEBUG_KMEM_TRACKING_ZFS="_without_debug_kmem_tracking"
	])

	AC_SUBST(KERNEL_DEBUG_CPPFLAGS)
	AC_SUBST(DEBUG_KMEM_TRACKING_ZFS)

	AC_MSG_RESULT([$enable_debug_kmem_tracking])
	])

	AC_DEFUN([ZFS_AC_DEBUG_INVARIANTS_DETECT_FREEBSD], [
	AS_IF([sysctl -n kern.conftxt \| grep -Fqx $'options\tINVARIANTS'],
	[enable_invariants="yes"],
	[enable_invariants="no"])
	])

	AC_DEFUN([ZFS_AC_DEBUG_INVARIANTS_DETECT], [
	AM_COND_IF([BUILD_FREEBSD],
	[ZFS_AC_DEBUG_INVARIANTS_DETECT_FREEBSD],
	[enable_invariants="no"])
	])

	dnl #
	dnl # Detected for the running kernel by default, enables INVARIANTS features
	dnl # in the FreeBSD kernel module. This feature must be used when building
	dnl # for a FreeBSD kernel with "options INVARIANTS" in the KERNCONF and must
	dnl # not be used when the INVARIANTS option is absent.
	dnl #
	AC_DEFUN([ZFS_AC_DEBUG_INVARIANTS], [
	AC_MSG_CHECKING([whether FreeBSD kernel INVARIANTS checks are enabled])
	AC_ARG_ENABLE([invariants],
	[AS_HELP_STRING([--enable-invariants],
	[Enable FreeBSD kernel INVARIANTS checks [[default: detect]]])],
	[], [ZFS_AC_DEBUG_INVARIANTS_DETECT])

	AS_IF([test "x$enable_invariants" = xyes],
	[WITH_INVARIANTS="true"],
	[WITH_INVARIANTS=""])
	AC_SUBST(WITH_INVARIANTS)

	AC_MSG_RESULT([$enable_invariants])
	])

	AC_DEFUN([ZFS_AC_CONFIG_ALWAYS], [
	AX_COUNT_CPUS([])
	AC_SUBST(CPU_COUNT)

	ZFS_AC_CONFIG_ALWAYS_CC_NO_CLOBBERED
	ZFS_AC_CONFIG_ALWAYS_CC_INFINITE_RECURSION
	ZFS_AC_CONFIG_ALWAYS_KERNEL_CC_INFINITE_RECURSION
	ZFS_AC_CONFIG_ALWAYS_CC_IMPLICIT_FALLTHROUGH
	ZFS_AC_CONFIG_ALWAYS_CC_FRAME_LARGER_THAN
	ZFS_AC_CONFIG_ALWAYS_CC_NO_FORMAT_TRUNCATION
	ZFS_AC_CONFIG_ALWAYS_CC_NO_FORMAT_ZERO_LENGTH
	ZFS_AC_CONFIG_ALWAYS_CC_FORMAT_OVERFLOW
	ZFS_AC_CONFIG_ALWAYS_CC_NO_OMIT_FRAME_POINTER
	ZFS_AC_CONFIG_ALWAYS_CC_NO_IPA_SRA
	ZFS_AC_CONFIG_ALWAYS_KERNEL_CC_NO_IPA_SRA
	ZFS_AC_CONFIG_ALWAYS_CC_ASAN
	ZFS_AC_CONFIG_ALWAYS_CC_UBSAN
	ZFS_AC_CONFIG_ALWAYS_TOOLCHAIN_SIMD
	ZFS_AC_CONFIG_ALWAYS_SYSTEM
	ZFS_AC_CONFIG_ALWAYS_ARCH
	ZFS_AC_CONFIG_ALWAYS_PYTHON
	ZFS_AC_CONFIG_ALWAYS_PYZFS
	ZFS_AC_CONFIG_ALWAYS_SED
	ZFS_AC_CONFIG_ALWAYS_CPPCHECK
	ZFS_AC_CONFIG_ALWAYS_SHELLCHECK
	ZFS_AC_CONFIG_ALWAYS_PARALLEL
	])

	AC_DEFUN([ZFS_AC_CONFIG], [

	dnl # Remove the previous build test directory.
	rm -Rf build

	ZFS_CONFIG=all
	AC_ARG_WITH([config],
	AS_HELP_STRING([--with-config=CONFIG],
	[Config file 'kernel\|user\|all\|srpm']),
	[ZFS_CONFIG="$withval"])
	AC_ARG_ENABLE([linux-builtin],
	[AS_HELP_STRING([--enable-linux-builtin],
	[Configure for builtin in-tree kernel modules @<:@default=no@:>@])],
	[],
	[enable_linux_builtin=no])

	AC_MSG_CHECKING([zfs config])
	AC_MSG_RESULT([$ZFS_CONFIG]);
	AC_SUBST(ZFS_CONFIG)

	ZFS_AC_CONFIG_ALWAYS

	AM_COND_IF([BUILD_LINUX], [
	AC_ARG_VAR([TEST_JOBS], [simultaneous jobs during configure])
	if test "x$ac_cv_env_TEST_JOBS_set" != "xset"; then
	TEST_JOBS=$CPU_COUNT
	fi
	AC_SUBST(TEST_JOBS)
	])

	ZFS_INIT_SYSV=
	ZFS_INIT_SYSTEMD=
	ZFS_WANT_MODULES_LOAD_D=

	case "$ZFS_CONFIG" in
	kernel) ZFS_AC_CONFIG_KERNEL ;;
	user) ZFS_AC_CONFIG_USER ;;
	all) ZFS_AC_CONFIG_USER
	ZFS_AC_CONFIG_KERNEL ;;
	dist) ;;
	srpm) ;;
	*)
	AC_MSG_RESULT([Error!])
	AC_MSG_ERROR([Bad value "$ZFS_CONFIG" for --with-config,
	user kernel\|user\|all\|srpm]) ;;
	esac

	AM_CONDITIONAL([INIT_SYSV], [test "x$ZFS_INIT_SYSV" = "xyes"])
	AM_CONDITIONAL([INIT_SYSTEMD], [test "x$ZFS_INIT_SYSTEMD" = "xyes"])
	AM_CONDITIONAL([WANT_MODULES_LOAD_D], [test "x$ZFS_WANT_MODULES_LOAD_D" = "xyes"])

	AM_CONDITIONAL([CONFIG_USER],
	[test "$ZFS_CONFIG" = user -o "$ZFS_CONFIG" = all])
	AM_CONDITIONAL([CONFIG_KERNEL],
	[test "$ZFS_CONFIG" = kernel -o "$ZFS_CONFIG" = all] &&
	[test "x$enable_linux_builtin" != xyes ])
	AM_CONDITIONAL([CONFIG_QAT],
	[test "$ZFS_CONFIG" = kernel -o "$ZFS_CONFIG" = all] &&
	[test "x$qatsrc" != x ])
	AM_CONDITIONAL([WANT_DEVNAME2DEVID], [test "x$user_libudev" = xyes ])
	AM_CONDITIONAL([WANT_MMAP_LIBAIO], [test "x$user_libaio" = xyes ])
	AM_CONDITIONAL([PAM_ZFS_ENABLED], [test "x$enable_pam" = xyes])
	])

	dnl #
	dnl # Check for rpm+rpmbuild to build RPM packages. If these tools
	dnl # are missing it is non-fatal but you will not be able to build
	dnl # RPM packages and will be warned if you try too.
	dnl #
	dnl # By default the generic spec file will be used because it requires
	dnl # minimal dependencies. Distribution specific spec files can be
	dnl # placed under the 'rpm/<distribution>' directory and enabled using
	dnl # the --with-spec=<distribution> configure option.
	dnl #
	AC_DEFUN([ZFS_AC_RPM], [
	RPM=rpm
	RPMBUILD=rpmbuild

	AC_MSG_CHECKING([whether $RPM is available])
	AS_IF([tmp=$($RPM --version 2>/dev/null)], [
	RPM_VERSION=$(echo $tmp \| $AWK '/RPM/ { print $[3] }')
	HAVE_RPM=yes
	AC_MSG_RESULT([$HAVE_RPM ($RPM_VERSION)])
	],[
	HAVE_RPM=no
	AC_MSG_RESULT([$HAVE_RPM])
	])

	AC_MSG_CHECKING([whether $RPMBUILD is available])
	AS_IF([tmp=$($RPMBUILD --version 2>/dev/null)], [
	RPMBUILD_VERSION=$(echo $tmp \| $AWK '/RPM/ { print $[3] }')
	HAVE_RPMBUILD=yes
	AC_MSG_RESULT([$HAVE_RPMBUILD ($RPMBUILD_VERSION)])
	],[
	HAVE_RPMBUILD=no
	AC_MSG_RESULT([$HAVE_RPMBUILD])
	])

	RPM_DEFINE_COMMON='--define "$(DEBUG_ZFS) 1"'
	RPM_DEFINE_COMMON=${RPM_DEFINE_COMMON}' --define "$(DEBUGINFO_ZFS) 1"'
	RPM_DEFINE_COMMON=${RPM_DEFINE_COMMON}' --define "$(DEBUG_KMEM_ZFS) 1"'
	RPM_DEFINE_COMMON=${RPM_DEFINE_COMMON}' --define "$(DEBUG_KMEM_TRACKING_ZFS) 1"'
	RPM_DEFINE_COMMON=${RPM_DEFINE_COMMON}' --define "$(ASAN_ZFS) 1"'
	RPM_DEFINE_COMMON=${RPM_DEFINE_COMMON}' --define "$(UBSAN_ZFS) 1"'

	AS_IF([test "x$enable_debuginfo" = xyes], [
	RPM_DEFINE_COMMON=${RPM_DEFINE_COMMON}' --define "__strip /bin/true"'
	])

	RPM_DEFINE_UTIL=' --define "_initconfdir $(initconfdir)"'

	dnl # Make the next three RPM_DEFINE_UTIL additions conditional, since
	dnl # their values may not be set when running:
	dnl #
	dnl # ./configure --with-config=srpm
	dnl #
	AS_IF([test -n "$dracutdir" ], [
	RPM_DEFINE_UTIL=${RPM_DEFINE_UTIL}' --define "_dracutdir $(dracutdir)"'
	])
	AS_IF([test -n "$udevdir" ], [
	RPM_DEFINE_UTIL=${RPM_DEFINE_UTIL}' --define "_udevdir $(udevdir)"'
	])
	AS_IF([test -n "$udevruledir" ], [
	RPM_DEFINE_UTIL=${RPM_DEFINE_UTIL}' --define "_udevruledir $(udevruledir)"'
	])
	AS_IF([test -n "$bashcompletiondir" ], [
	RPM_DEFINE_UTIL=${RPM_DEFINE_UTIL}' --define "_bashcompletiondir $(bashcompletiondir)"'
	])
	RPM_DEFINE_UTIL=${RPM_DEFINE_UTIL}' $(DEFINE_SYSTEMD)'
	RPM_DEFINE_UTIL=${RPM_DEFINE_UTIL}' $(DEFINE_PYZFS)'
	RPM_DEFINE_UTIL=${RPM_DEFINE_UTIL}' $(DEFINE_PAM)'
	RPM_DEFINE_UTIL=${RPM_DEFINE_UTIL}' $(DEFINE_PYTHON_VERSION)'
	RPM_DEFINE_UTIL=${RPM_DEFINE_UTIL}' $(DEFINE_PYTHON_PKG_VERSION)'

	dnl # Override default lib directory on Debian/Ubuntu systems. The
	dnl # provided /usr/lib/rpm/platform/<arch>/macros files do not
	dnl # specify the correct path for multiarch systems as described
	dnl # by the packaging guidelines.
	dnl #
	dnl # https://wiki.ubuntu.com/MultiarchSpec
	dnl # https://wiki.debian.org/Multiarch/Implementation
	dnl #
	AS_IF([test "$DEFAULT_PACKAGE" = "deb"], [
	MULTIARCH_LIBDIR="lib/$(dpkg-architecture -qDEB_HOST_MULTIARCH)"
	RPM_DEFINE_UTIL=${RPM_DEFINE_UTIL}' --define "_lib $(MULTIARCH_LIBDIR)"'
	AC_SUBST(MULTIARCH_LIBDIR)
	])

	dnl # Make RPM_DEFINE_KMOD additions conditional on CONFIG_KERNEL,
	dnl # since the values will not be set otherwise. The spec files
	dnl # provide defaults for them.
	dnl #
	RPM_DEFINE_KMOD='--define "_wrong_version_format_terminate_build 0"'
	AM_COND_IF([CONFIG_KERNEL], [
	RPM_DEFINE_KMOD=${RPM_DEFINE_KMOD}' --define "kernels $(LINUX_VERSION)"'
	RPM_DEFINE_KMOD=${RPM_DEFINE_KMOD}' --define "ksrc $(LINUX)"'
	RPM_DEFINE_KMOD=${RPM_DEFINE_KMOD}' --define "kobj $(LINUX_OBJ)"'
	RPM_DEFINE_KMOD=${RPM_DEFINE_KMOD}' --define "kernel_cc KERNEL_CC=$(KERNEL_CC)"'
	RPM_DEFINE_KMOD=${RPM_DEFINE_KMOD}' --define "kernel_ld KERNEL_LD=$(KERNEL_LD)"'
	RPM_DEFINE_KMOD=${RPM_DEFINE_KMOD}' --define "kernel_llvm KERNEL_LLVM=$(KERNEL_LLVM)"'
	])

	RPM_DEFINE_DKMS=''

	SRPM_DEFINE_COMMON='--define "build_src_rpm 1"'
	SRPM_DEFINE_UTIL=
	SRPM_DEFINE_KMOD=
	SRPM_DEFINE_DKMS=

	RPM_SPEC_DIR="rpm/generic"
	AC_ARG_WITH([spec],
	AS_HELP_STRING([--with-spec=SPEC],
	[Spec files 'generic\|redhat']),
	[RPM_SPEC_DIR="rpm/$withval"])

	AC_MSG_CHECKING([whether spec files are available])
	AC_MSG_RESULT([yes ($RPM_SPEC_DIR/*.spec.in)])

	AC_SUBST(HAVE_RPM)
	AC_SUBST(RPM)
	AC_SUBST(RPM_VERSION)

	AC_SUBST(HAVE_RPMBUILD)
	AC_SUBST(RPMBUILD)
	AC_SUBST(RPMBUILD_VERSION)

	AC_SUBST(RPM_SPEC_DIR)
	AC_SUBST(RPM_DEFINE_UTIL)
	AC_SUBST(RPM_DEFINE_KMOD)
	AC_SUBST(RPM_DEFINE_DKMS)
	AC_SUBST(RPM_DEFINE_COMMON)
	AC_SUBST(SRPM_DEFINE_UTIL)
	AC_SUBST(SRPM_DEFINE_KMOD)
	AC_SUBST(SRPM_DEFINE_DKMS)
	AC_SUBST(SRPM_DEFINE_COMMON)
	])

	dnl #
	dnl # Check for dpkg+dpkg-buildpackage to build DEB packages. If these
	dnl # tools are missing it is non-fatal but you will not be able to build
	dnl # DEB packages and will be warned if you try too.
	dnl #
	AC_DEFUN([ZFS_AC_DPKG], [
	DPKG=dpkg
	DPKGBUILD=dpkg-buildpackage

	AC_MSG_CHECKING([whether $DPKG is available])
	AS_IF([tmp=$($DPKG --version 2>/dev/null)], [
	DPKG_VERSION=$(echo $tmp \| $AWK '/Debian/ { print $[7] }')
	HAVE_DPKG=yes
	AC_MSG_RESULT([$HAVE_DPKG ($DPKG_VERSION)])
	],[
	HAVE_DPKG=no
	AC_MSG_RESULT([$HAVE_DPKG])
	])

	AC_MSG_CHECKING([whether $DPKGBUILD is available])
	AS_IF([tmp=$($DPKGBUILD --version 2>/dev/null)], [
	DPKGBUILD_VERSION=$(echo $tmp \| \
	$AWK '/Debian/ { print $[4] }' \| cut -f-4 -d'.')
	HAVE_DPKGBUILD=yes
	AC_MSG_RESULT([$HAVE_DPKGBUILD ($DPKGBUILD_VERSION)])
	],[
	HAVE_DPKGBUILD=no
	AC_MSG_RESULT([$HAVE_DPKGBUILD])
	])

	AC_SUBST(HAVE_DPKG)
	AC_SUBST(DPKG)
	AC_SUBST(DPKG_VERSION)

	AC_SUBST(HAVE_DPKGBUILD)
	AC_SUBST(DPKGBUILD)
	AC_SUBST(DPKGBUILD_VERSION)
	AC_SUBST([CFGOPTS], ["$CFGOPTS"])
	])

	dnl #
	dnl # Until native packaging for various different packing systems
	dnl # can be added the least we can do is attempt to use alien to
	dnl # convert the RPM packages to the needed package type. This is
	dnl # a hack but so far it has worked reasonable well.
	dnl #
	AC_DEFUN([ZFS_AC_ALIEN], [
	ALIEN=alien

	AC_MSG_CHECKING([whether $ALIEN is available])
	AS_IF([tmp=$($ALIEN --version 2>/dev/null)], [
	ALIEN_VERSION=$(echo $tmp \| $AWK '{ print $[3] }')
	ALIEN_MAJOR=$(echo ${ALIEN_VERSION} \| $AWK -F'.' '{ print $[1] }')
	ALIEN_MINOR=$(echo ${ALIEN_VERSION} \| $AWK -F'.' '{ print $[2] }')
	ALIEN_POINT=$(echo ${ALIEN_VERSION} \| $AWK -F'.' '{ print $[3] }')
	HAVE_ALIEN=yes
	AC_MSG_RESULT([$HAVE_ALIEN ($ALIEN_VERSION)])
	],[
	HAVE_ALIEN=no
	AC_MSG_RESULT([$HAVE_ALIEN])
	])

	AC_SUBST(HAVE_ALIEN)
	AC_SUBST(ALIEN)
	AC_SUBST(ALIEN_VERSION)
	AC_SUBST(ALIEN_MAJOR)
	AC_SUBST(ALIEN_MINOR)
	AC_SUBST(ALIEN_POINT)
	])

	dnl #
	dnl # Using the VENDOR tag from config.guess set the default
	dnl # package type for 'make pkg': (rpm \| deb \| tgz)
	dnl #
	AC_DEFUN([ZFS_AC_DEFAULT_PACKAGE], [
	AC_MSG_CHECKING([os distribution])
	AC_ARG_WITH([vendor],
	[AS_HELP_STRING([--with-vendor],
	[Distribution vendor @<:@default=check@:>@])],
	[with_vendor=$withval],
	[with_vendor=check])
	AS_IF([test "x$with_vendor" = "xcheck"],[
	- if test -f /etc/toss-release ; then
	- VENDOR=toss ;
	+ if test -f /etc/alpine-release ; then
	+ VENDOR=alpine ;
	+ elif test -f /etc/arch-release ; then
	+ VENDOR=arch ;
	elif test -f /etc/fedora-release ; then
	VENDOR=fedora ;
	- elif test -f /etc/redhat-release ; then
	- VENDOR=redhat ;
	+ elif test -f /bin/freebsd-version ; then
	+ VENDOR=freebsd ;
	elif test -f /etc/gentoo-release ; then
	VENDOR=gentoo ;
	- elif test -f /etc/arch-release ; then
	- VENDOR=arch ;
	+ elif test -f /etc/lunar.release ; then
	+ VENDOR=lunar ;
	+ elif test -f /etc/openEuler-release ; then
	+ VENDOR=openeuler ;
	elif test -f /etc/SuSE-release ; then
	VENDOR=sles ;
	elif test -f /etc/slackware-version ; then
	VENDOR=slackware ;
	- elif test -f /etc/lunar.release ; then
	- VENDOR=lunar ;
	+ elif test -f /etc/toss-release ; then
	+ VENDOR=toss ;
	elif test -f /etc/lsb-release ; then
	VENDOR=ubuntu ;
	+ # put debian and redhat last as derivatives may have also their file
	elif test -f /etc/debian_version ; then
	VENDOR=debian ;
	- elif test -f /etc/alpine-release ; then
	- VENDOR=alpine ;
	- elif test -f /bin/freebsd-version ; then
	- VENDOR=freebsd ;
	- elif test -f /etc/openEuler-release ; then
	- VENDOR=openeuler ;
	+ elif test -f /etc/redhat-release ; then
	+ VENDOR=redhat ;
	else
	VENDOR= ;
	fi],
	[ test "x${with_vendor}" != x],[
	VENDOR="$with_vendor" ],
	[ VENDOR= ; ]
	)
	AC_MSG_RESULT([$VENDOR])
	AC_SUBST(VENDOR)

	AC_MSG_CHECKING([default package type])
	case "$VENDOR" in
	- toss) DEFAULT_PACKAGE=rpm ;;
	- redhat) DEFAULT_PACKAGE=rpm ;;
	- fedora) DEFAULT_PACKAGE=rpm ;;
	- gentoo) DEFAULT_PACKAGE=tgz ;;
	- alpine) DEFAULT_PACKAGE=tgz ;;
	- arch) DEFAULT_PACKAGE=tgz ;;
	- sles) DEFAULT_PACKAGE=rpm ;;
	- slackware) DEFAULT_PACKAGE=tgz ;;
	- lunar) DEFAULT_PACKAGE=tgz ;;
	- ubuntu) DEFAULT_PACKAGE=deb ;;
	- debian) DEFAULT_PACKAGE=deb ;;
	- freebsd) DEFAULT_PACKAGE=pkg ;;
	- openeuler) DEFAULT_PACKAGE=rpm ;;
	- *) DEFAULT_PACKAGE=rpm ;;
	+ alpine\|arch\|gentoo\|lunar\|slackware)
	+ DEFAULT_PACKAGE=tgz ;;
	+ debian\|ubuntu)
	+ DEFAULT_PACKAGE=deb ;;
	+ freebsd)
	+ DEFAULT_PACKAGE=pkg ;;
	+ *)
	+ # fedora\|openeuler\|redhat\|sles\|toss
	+ DEFAULT_PACKAGE=rpm ;;
	esac
	AC_MSG_RESULT([$DEFAULT_PACKAGE])
	AC_SUBST(DEFAULT_PACKAGE)

	AC_MSG_CHECKING([default init directory])
	case "$VENDOR" in
	freebsd) initdir=$sysconfdir/rc.d ;;
	*) initdir=$sysconfdir/init.d;;
	esac
	AC_MSG_RESULT([$initdir])
	AC_SUBST(initdir)

	AC_MSG_CHECKING([default shell])
	case "$VENDOR" in
	- gentoo\|alpine) DEFAULT_INIT_SHELL=/sbin/openrc-run
	+ alpine\|gentoo) DEFAULT_INIT_SHELL=/sbin/openrc-run
	IS_SYSV_RC=false ;;
	*) DEFAULT_INIT_SHELL=/bin/sh
	IS_SYSV_RC=true ;;
	esac

	AC_MSG_RESULT([$DEFAULT_INIT_SHELL])
	AC_SUBST(DEFAULT_INIT_SHELL)
	AC_SUBST(IS_SYSV_RC)

	AC_MSG_CHECKING([default nfs server init script])
	AS_IF([test "$VENDOR" = "debian"],
	[DEFAULT_INIT_NFS_SERVER="nfs-kernel-server"],
	[DEFAULT_INIT_NFS_SERVER="nfs"]
	)
	AC_MSG_RESULT([$DEFAULT_INIT_NFS_SERVER])
	AC_SUBST(DEFAULT_INIT_NFS_SERVER)

	AC_MSG_CHECKING([default init config directory])
	case "$VENDOR" in
	- alpine) initconfdir=/etc/conf.d ;;
	- gentoo) initconfdir=/etc/conf.d ;;
	- toss) initconfdir=/etc/sysconfig ;;
	- redhat) initconfdir=/etc/sysconfig ;;
	- fedora) initconfdir=/etc/sysconfig ;;
	- sles) initconfdir=/etc/sysconfig ;;
	- openeuler) initconfdir=/etc/sysconfig ;;
	- ubuntu) initconfdir=/etc/default ;;
	- debian) initconfdir=/etc/default ;;
	- freebsd) initconfdir=$sysconfdir/rc.conf.d;;
	- *) initconfdir=/etc/default ;;
	+ alpine\|gentoo)
	+ initconfdir=/etc/conf.d
	+ ;;
	+ fedora\|openeuler\|redhat\|sles\|toss)
	+ initconfdir=/etc/sysconfig
	+ ;;
	+ freebsd)
	+ initconfdir=$sysconfdir/rc.conf.d
	+ ;;
	+ *)
	+ # debian\|ubuntu
	+ initconfdir=/etc/default
	+ ;;
	esac
	AC_MSG_RESULT([$initconfdir])
	AC_SUBST(initconfdir)

	AC_MSG_CHECKING([whether initramfs-tools is available])
	if test -d /usr/share/initramfs-tools ; then
	RPM_DEFINE_INITRAMFS='--define "_initramfs 1"'
	AC_MSG_RESULT([yes])
	else
	RPM_DEFINE_INITRAMFS=''
	AC_MSG_RESULT([no])
	fi
	AC_SUBST(RPM_DEFINE_INITRAMFS)

	AC_MSG_CHECKING([default bash completion directory])
	case "$VENDOR" in
	- ubuntu) bashcompletiondir=/usr/share/bash-completion/completions ;;
	- debian) bashcompletiondir=/usr/share/bash-completion/completions ;;
	- freebsd) bashcompletiondir=$sysconfdir/bash_completion.d;;
	- gentoo) bashcompletiondir=/usr/share/bash-completion/completions ;;
	- *) bashcompletiondir=/etc/bash_completion.d ;;
	+ alpine\|debian\|gentoo\|ubuntu)
	+ bashcompletiondir=/usr/share/bash-completion/completions
	+ ;;
	+ freebsd)
	+ bashcompletiondir=$sysconfdir/bash_completion.d
	+ ;;
	+ *)
	+ bashcompletiondir=/etc/bash_completion.d
	+ ;;
	esac
	AC_MSG_RESULT([$bashcompletiondir])
	AC_SUBST(bashcompletiondir)

	])

	dnl #
	dnl # Default ZFS package configuration
	dnl #
	AC_DEFUN([ZFS_AC_PACKAGE], [
	ZFS_AC_DEFAULT_PACKAGE
	AS_IF([test x$VENDOR != xfreebsd], [
	ZFS_AC_RPM
	ZFS_AC_DPKG
	ZFS_AC_ALIEN
	])
	])
	diff --git a/sys/contrib/openzfs/include/libzfs.h b/sys/contrib/openzfs/include/libzfs.h
	index 2823b8845827..7836c2325f4e 100644
	--- a/sys/contrib/openzfs/include/libzfs.h
	+++ b/sys/contrib/openzfs/include/libzfs.h
	@@ -1,1054 +1,1055 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2024 by Delphix. All rights reserved.
	* Copyright Joyent, Inc.
	* Copyright (c) 2013 Steven Hartland. All rights reserved.
	* Copyright (c) 2016, Intel Corporation.
	* Copyright 2016 Nexenta Systems, Inc.
	* Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
	* Copyright (c) 2019 Datto Inc.
	* Copyright (c) 2021, Colm Buckley <colm@tuatha.org>
	*/

	#ifndef _LIBZFS_H
	#define _LIBZFS_H extern __attribute__((visibility("default")))

	#include <assert.h>
	#include <libshare.h>
	#include <libnvpair.h>
	#include <sys/mnttab.h>
	#include <sys/param.h>
	#include <sys/types.h>
	#include <sys/fs/zfs.h>
	#include <sys/avl.h>
	#include <libzfs_core.h>

	#ifdef __cplusplus
	extern "C" {
	#endif

	/*
	* Miscellaneous ZFS constants
	*/
	#define ZFS_MAXPROPLEN MAXPATHLEN
	#define ZPOOL_MAXPROPLEN MAXPATHLEN

	/*
	* libzfs errors
	*/
	typedef enum zfs_error {
	EZFS_SUCCESS = 0, /* no error -- success */
	EZFS_NOMEM = 2000, /* out of memory */
	EZFS_BADPROP, /* invalid property value */
	EZFS_PROPREADONLY, /* cannot set readonly property */
	EZFS_PROPTYPE, /* property does not apply to dataset type */
	EZFS_PROPNONINHERIT, /* property is not inheritable */
	EZFS_PROPSPACE, /* bad quota or reservation */
	EZFS_BADTYPE, /* dataset is not of appropriate type */
	EZFS_BUSY, /* pool or dataset is busy */
	EZFS_EXISTS, /* pool or dataset already exists */
	EZFS_NOENT, /* no such pool or dataset */
	EZFS_BADSTREAM, /* bad backup stream */
	EZFS_DSREADONLY, /* dataset is readonly */
	EZFS_VOLTOOBIG, /* volume is too large for 32-bit system */
	EZFS_INVALIDNAME, /* invalid dataset name */
	EZFS_BADRESTORE, /* unable to restore to destination */
	EZFS_BADBACKUP, /* backup failed */
	EZFS_BADTARGET, /* bad attach/detach/replace target */
	EZFS_NODEVICE, /* no such device in pool */
	EZFS_BADDEV, /* invalid device to add */
	EZFS_NOREPLICAS, /* no valid replicas */
	EZFS_RESILVERING, /* resilvering (healing reconstruction) */
	EZFS_BADVERSION, /* unsupported version */
	EZFS_POOLUNAVAIL, /* pool is currently unavailable */
	EZFS_DEVOVERFLOW, /* too many devices in one vdev */
	EZFS_BADPATH, /* must be an absolute path */
	EZFS_CROSSTARGET, /* rename or clone across pool or dataset */
	EZFS_ZONED, /* used improperly in local zone */
	EZFS_MOUNTFAILED, /* failed to mount dataset */
	EZFS_UMOUNTFAILED, /* failed to unmount dataset */
	EZFS_UNSHARENFSFAILED, /* failed to unshare over nfs */
	EZFS_SHARENFSFAILED, /* failed to share over nfs */
	EZFS_PERM, /* permission denied */
	EZFS_NOSPC, /* out of space */
	EZFS_FAULT, /* bad address */
	EZFS_IO, /* I/O error */
	EZFS_INTR, /* signal received */
	EZFS_ISSPARE, /* device is a hot spare */
	EZFS_INVALCONFIG, /* invalid vdev configuration */
	EZFS_RECURSIVE, /* recursive dependency */
	EZFS_NOHISTORY, /* no history object */
	EZFS_POOLPROPS, /* couldn't retrieve pool props */
	EZFS_POOL_NOTSUP, /* ops not supported for this type of pool */
	EZFS_POOL_INVALARG, /* invalid argument for this pool operation */
	EZFS_NAMETOOLONG, /* dataset name is too long */
	EZFS_OPENFAILED, /* open of device failed */
	EZFS_NOCAP, /* couldn't get capacity */
	EZFS_LABELFAILED, /* write of label failed */
	EZFS_BADWHO, /* invalid permission who */
	EZFS_BADPERM, /* invalid permission */
	EZFS_BADPERMSET, /* invalid permission set name */
	EZFS_NODELEGATION, /* delegated administration is disabled */
	EZFS_UNSHARESMBFAILED, /* failed to unshare over smb */
	EZFS_SHARESMBFAILED, /* failed to share over smb */
	EZFS_BADCACHE, /* bad cache file */
	EZFS_ISL2CACHE, /* device is for the level 2 ARC */
	EZFS_VDEVNOTSUP, /* unsupported vdev type */
	EZFS_NOTSUP, /* ops not supported on this dataset */
	EZFS_ACTIVE_SPARE, /* pool has active shared spare devices */
	EZFS_UNPLAYED_LOGS, /* log device has unplayed logs */
	EZFS_REFTAG_RELE, /* snapshot release: tag not found */
	EZFS_REFTAG_HOLD, /* snapshot hold: tag already exists */
	EZFS_TAGTOOLONG, /* snapshot hold/rele: tag too long */
	EZFS_PIPEFAILED, /* pipe create failed */
	EZFS_THREADCREATEFAILED, /* thread create failed */
	EZFS_POSTSPLIT_ONLINE, /* onlining a disk after splitting it */
	EZFS_SCRUBBING, /* currently scrubbing */
	EZFS_ERRORSCRUBBING, /* currently error scrubbing */
	EZFS_ERRORSCRUB_PAUSED, /* error scrub currently paused */
	EZFS_NO_SCRUB, /* no active scrub */
	EZFS_DIFF, /* general failure of zfs diff */
	EZFS_DIFFDATA, /* bad zfs diff data */
	EZFS_POOLREADONLY, /* pool is in read-only mode */
	EZFS_SCRUB_PAUSED, /* scrub currently paused */
	EZFS_SCRUB_PAUSED_TO_CANCEL, /* scrub currently paused */
	EZFS_ACTIVE_POOL, /* pool is imported on a different system */
	EZFS_CRYPTOFAILED, /* failed to setup encryption */
	EZFS_NO_PENDING, /* cannot cancel, no operation is pending */
	EZFS_CHECKPOINT_EXISTS, /* checkpoint exists */
	EZFS_DISCARDING_CHECKPOINT, /* currently discarding a checkpoint */
	EZFS_NO_CHECKPOINT, /* pool has no checkpoint */
	EZFS_DEVRM_IN_PROGRESS, /* a device is currently being removed */
	EZFS_VDEV_TOO_BIG, /* a device is too big to be used */
	EZFS_IOC_NOTSUPPORTED, /* operation not supported by zfs module */
	EZFS_TOOMANY, /* argument list too long */
	EZFS_INITIALIZING, /* currently initializing */
	EZFS_NO_INITIALIZE, /* no active initialize */
	EZFS_WRONG_PARENT, /* invalid parent dataset (e.g ZVOL) */
	EZFS_TRIMMING, /* currently trimming */
	EZFS_NO_TRIM, /* no active trim */
	EZFS_TRIM_NOTSUP, /* device does not support trim */
	EZFS_NO_RESILVER_DEFER, /* pool doesn't support resilver_defer */
	EZFS_EXPORT_IN_PROGRESS, /* currently exporting the pool */
	EZFS_REBUILDING, /* resilvering (sequential reconstrution) */
	EZFS_VDEV_NOTSUP, /* ops not supported for this type of vdev */
	EZFS_NOT_USER_NAMESPACE, /* a file is not a user namespace */
	EZFS_CKSUM, /* insufficient replicas */
	EZFS_RESUME_EXISTS, /* Resume on existing dataset without force */
	EZFS_SHAREFAILED, /* filesystem share failed */
	EZFS_RAIDZ_EXPAND_IN_PROGRESS, /* a raidz is currently expanding */
	EZFS_ASHIFT_MISMATCH, /* can't add vdevs with different ashifts */
	EZFS_UNKNOWN
	} zfs_error_t;

	/*
	* The following data structures are all part
	* of the zfs_allow_t data structure which is
	* used for printing 'allow' permissions.
	* It is a linked list of zfs_allow_t's which
	* then contain avl tree's for user/group/sets/...
	* and each one of the entries in those trees have
	* avl tree's for the permissions they belong to and
	* whether they are local,descendent or local+descendent
	* permissions. The AVL trees are used primarily for
	* sorting purposes, but also so that we can quickly find
	* a given user and or permission.
	*/
	typedef struct zfs_perm_node {
	avl_node_t z_node;
	char z_pname[MAXPATHLEN];
	} zfs_perm_node_t;

	typedef struct zfs_allow_node {
	avl_node_t z_node;
	char z_key[MAXPATHLEN]; /* name, such as joe */
	avl_tree_t z_localdescend; /* local+descendent perms */
	avl_tree_t z_local; /* local permissions */
	avl_tree_t z_descend; /* descendent permissions */
	} zfs_allow_node_t;

	typedef struct zfs_allow {
	struct zfs_allow *z_next;
	char z_setpoint[MAXPATHLEN];
	avl_tree_t z_sets;
	avl_tree_t z_crperms;
	avl_tree_t z_user;
	avl_tree_t z_group;
	avl_tree_t z_everyone;
	} zfs_allow_t;

	/*
	* Basic handle types
	*/
	typedef struct zfs_handle zfs_handle_t;
	typedef struct zpool_handle zpool_handle_t;
	typedef struct libzfs_handle libzfs_handle_t;

	_LIBZFS_H int zpool_wait(zpool_handle_t *, zpool_wait_activity_t);
	_LIBZFS_H int zpool_wait_status(zpool_handle_t *, zpool_wait_activity_t,
	boolean_t , boolean_t );

	/*
	* Library initialization
	*/
	_LIBZFS_H libzfs_handle_t *libzfs_init(void);
	_LIBZFS_H void libzfs_fini(libzfs_handle_t *);

	_LIBZFS_H libzfs_handle_t zpool_get_handle(zpool_handle_t );
	_LIBZFS_H libzfs_handle_t zfs_get_handle(zfs_handle_t );

	_LIBZFS_H void libzfs_print_on_error(libzfs_handle_t *, boolean_t);

	_LIBZFS_H void zfs_save_arguments(int argc, char *, char , int);
	_LIBZFS_H int zpool_log_history(libzfs_handle_t , const char );

	_LIBZFS_H int libzfs_errno(libzfs_handle_t *);
	_LIBZFS_H const char *libzfs_error_init(int);
	_LIBZFS_H const char libzfs_error_action(libzfs_handle_t );
	_LIBZFS_H const char libzfs_error_description(libzfs_handle_t );
	_LIBZFS_H int zfs_standard_error(libzfs_handle_t , int, const char );
	_LIBZFS_H void libzfs_mnttab_init(libzfs_handle_t *);
	_LIBZFS_H void libzfs_mnttab_fini(libzfs_handle_t *);
	_LIBZFS_H void libzfs_mnttab_cache(libzfs_handle_t *, boolean_t);
	_LIBZFS_H int libzfs_mnttab_find(libzfs_handle_t , const char ,
	struct mnttab *);
	_LIBZFS_H void libzfs_mnttab_add(libzfs_handle_t , const char ,
	const char , const char );
	_LIBZFS_H void libzfs_mnttab_remove(libzfs_handle_t , const char );

	/*
	* Basic handle functions
	*/
	_LIBZFS_H zpool_handle_t zpool_open(libzfs_handle_t , const char *);
	_LIBZFS_H zpool_handle_t zpool_open_canfail(libzfs_handle_t , const char *);
	_LIBZFS_H void zpool_close(zpool_handle_t *);
	_LIBZFS_H const char zpool_get_name(zpool_handle_t );
	_LIBZFS_H int zpool_get_state(zpool_handle_t *);
	_LIBZFS_H const char *zpool_state_to_name(vdev_state_t, vdev_aux_t);
	_LIBZFS_H const char *zpool_pool_state_to_name(pool_state_t);
	_LIBZFS_H void zpool_free_handles(libzfs_handle_t *);

	/*
	* Iterate over all active pools in the system.
	*/
	typedef int (zpool_iter_f)(zpool_handle_t , void *);
	_LIBZFS_H int zpool_iter(libzfs_handle_t , zpool_iter_f, void );
	_LIBZFS_H boolean_t zpool_skip_pool(const char *);

	/*
	* Functions to create and destroy pools
	*/
	_LIBZFS_H int zpool_create(libzfs_handle_t , const char , nvlist_t *,
	nvlist_t , nvlist_t );
	_LIBZFS_H int zpool_destroy(zpool_handle_t , const char );
	_LIBZFS_H int zpool_add(zpool_handle_t , nvlist_t , boolean_t check_ashift);

	typedef struct splitflags {
	/* do not split, but return the config that would be split off */
	unsigned int dryrun : 1;

	/* after splitting, import the pool */
	unsigned int import : 1;
	int name_flags;
	} splitflags_t;

	typedef struct trimflags {
	/* requested vdevs are for the entire pool */
	boolean_t fullpool;

	/* request a secure trim, requires support from device */
	boolean_t secure;

	/* after starting trim, block until trim completes */
	boolean_t wait;

	/* trim at the requested rate in bytes/second */
	uint64_t rate;
	} trimflags_t;

	/*
	* Functions to manipulate pool and vdev state
	*/
	_LIBZFS_H int zpool_scan(zpool_handle_t *, pool_scan_func_t, pool_scrub_cmd_t);
	_LIBZFS_H int zpool_initialize(zpool_handle_t *, pool_initialize_func_t,
	nvlist_t *);
	_LIBZFS_H int zpool_initialize_wait(zpool_handle_t *, pool_initialize_func_t,
	nvlist_t *);
	_LIBZFS_H int zpool_trim(zpool_handle_t , pool_trim_func_t, nvlist_t ,
	trimflags_t *);

	_LIBZFS_H int zpool_clear(zpool_handle_t , const char , nvlist_t *);
	_LIBZFS_H int zpool_reguid(zpool_handle_t *);
	_LIBZFS_H int zpool_reopen_one(zpool_handle_t , void );

	_LIBZFS_H int zpool_sync_one(zpool_handle_t , void );

	_LIBZFS_H int zpool_vdev_online(zpool_handle_t , const char , int,
	vdev_state_t *);
	_LIBZFS_H int zpool_vdev_offline(zpool_handle_t , const char , boolean_t);
	_LIBZFS_H int zpool_vdev_attach(zpool_handle_t , const char ,
	const char , nvlist_t , int, boolean_t);
	_LIBZFS_H int zpool_vdev_detach(zpool_handle_t , const char );
	_LIBZFS_H int zpool_vdev_remove(zpool_handle_t , const char );
	_LIBZFS_H int zpool_vdev_remove_cancel(zpool_handle_t *);
	_LIBZFS_H int zpool_vdev_indirect_size(zpool_handle_t , const char ,
	uint64_t *);
	_LIBZFS_H int zpool_vdev_split(zpool_handle_t , char , nvlist_t **,
	nvlist_t *, splitflags_t);
	_LIBZFS_H int zpool_vdev_remove_wanted(zpool_handle_t , const char );

	_LIBZFS_H int zpool_vdev_fault(zpool_handle_t *, uint64_t, vdev_aux_t);
	_LIBZFS_H int zpool_vdev_degrade(zpool_handle_t *, uint64_t, vdev_aux_t);
	_LIBZFS_H int zpool_vdev_set_removed_state(zpool_handle_t *, uint64_t,
	vdev_aux_t);

	_LIBZFS_H int zpool_vdev_clear(zpool_handle_t *, uint64_t);

	_LIBZFS_H nvlist_t zpool_find_vdev(zpool_handle_t , const char , boolean_t ,
	boolean_t , boolean_t );
	_LIBZFS_H nvlist_t zpool_find_vdev_by_physpath(zpool_handle_t , const char *,
	boolean_t , boolean_t , boolean_t *);
	_LIBZFS_H int zpool_label_disk(libzfs_handle_t , zpool_handle_t ,
	const char *);
	_LIBZFS_H int zpool_prepare_disk(zpool_handle_t zhp, nvlist_t vdev_nv,
	const char prepare_str, char lines[], int lines_cnt);
	_LIBZFS_H int zpool_prepare_and_label_disk(libzfs_handle_t *hdl,
	zpool_handle_t , const char , nvlist_t vdev_nv, const char prepare_str,
	char *lines[], int lines_cnt);
	_LIBZFS_H char ** zpool_vdev_script_alloc_env(const char *pool_name,
	const char vdev_path, const char vdev_upath,
	const char vdev_enc_sysfs_path, const char opt_key, const char *opt_val);
	_LIBZFS_H void zpool_vdev_script_free_env(char **env);
	_LIBZFS_H uint64_t zpool_vdev_path_to_guid(zpool_handle_t *zhp,
	const char *path);

	_LIBZFS_H const char zpool_get_state_str(zpool_handle_t );

	/*
	* Functions to manage pool properties
	*/
	_LIBZFS_H int zpool_set_prop(zpool_handle_t , const char , const char *);
	_LIBZFS_H int zpool_get_prop(zpool_handle_t , zpool_prop_t, char ,
	size_t proplen, zprop_source_t *, boolean_t literal);
	_LIBZFS_H int zpool_get_userprop(zpool_handle_t , const char , char *,
	size_t proplen, zprop_source_t *);
	_LIBZFS_H uint64_t zpool_get_prop_int(zpool_handle_t *, zpool_prop_t,
	zprop_source_t *);
	_LIBZFS_H int zpool_props_refresh(zpool_handle_t *);

	_LIBZFS_H const char *zpool_prop_to_name(zpool_prop_t);
	_LIBZFS_H const char *zpool_prop_values(zpool_prop_t);

	/*
	* Functions to manage vdev properties
	*/
	_LIBZFS_H int zpool_get_vdev_prop_value(nvlist_t , vdev_prop_t, char , char *,
	size_t, zprop_source_t *, boolean_t);
	_LIBZFS_H int zpool_get_vdev_prop(zpool_handle_t , const char , vdev_prop_t,
	char , char , size_t, zprop_source_t *, boolean_t);
	_LIBZFS_H int zpool_get_all_vdev_props(zpool_handle_t , const char ,
	nvlist_t **);
	_LIBZFS_H int zpool_set_vdev_prop(zpool_handle_t , const char , const char *,
	const char *);

	_LIBZFS_H const char *vdev_prop_to_name(vdev_prop_t);
	_LIBZFS_H const char *vdev_prop_values(vdev_prop_t);
	_LIBZFS_H boolean_t vdev_prop_user(const char *name);
	_LIBZFS_H const char *vdev_prop_column_name(vdev_prop_t);
	_LIBZFS_H boolean_t vdev_prop_align_right(vdev_prop_t);

	/*
	* Pool health statistics.
	*/
	typedef enum {
	/*
	* The following correspond to faults as defined in the (fault.fs.zfs.*)
	* event namespace. Each is associated with a corresponding message ID.
	* This must be kept in sync with the zfs_msgid_table in
	* lib/libzfs/libzfs_status.c.
	*/
	ZPOOL_STATUS_CORRUPT_CACHE, /* corrupt /kernel/drv/zpool.cache */
	ZPOOL_STATUS_MISSING_DEV_R, /* missing device with replicas */
	ZPOOL_STATUS_MISSING_DEV_NR, /* missing device with no replicas */
	ZPOOL_STATUS_CORRUPT_LABEL_R, /* bad device label with replicas */
	ZPOOL_STATUS_CORRUPT_LABEL_NR, /* bad device label with no replicas */
	ZPOOL_STATUS_BAD_GUID_SUM, /* sum of device guids didn't match */
	ZPOOL_STATUS_CORRUPT_POOL, /* pool metadata is corrupted */
	ZPOOL_STATUS_CORRUPT_DATA, /* data errors in user (meta)data */
	ZPOOL_STATUS_FAILING_DEV, /* device experiencing errors */
	ZPOOL_STATUS_VERSION_NEWER, /* newer on-disk version */
	ZPOOL_STATUS_HOSTID_MISMATCH, /* last accessed by another system */
	ZPOOL_STATUS_HOSTID_ACTIVE, /* currently active on another system */
	ZPOOL_STATUS_HOSTID_REQUIRED, /* multihost=on and hostid=0 */
	ZPOOL_STATUS_IO_FAILURE_WAIT, /* failed I/O, failmode 'wait' */
	ZPOOL_STATUS_IO_FAILURE_CONTINUE, /* failed I/O, failmode 'continue' */
	ZPOOL_STATUS_IO_FAILURE_MMP, /* failed MMP, failmode not 'panic' */
	ZPOOL_STATUS_BAD_LOG, /* cannot read log chain(s) */
	ZPOOL_STATUS_ERRATA, /* informational errata available */

	/*
	* If the pool has unsupported features but can still be opened in
	* read-only mode, its status is ZPOOL_STATUS_UNSUP_FEAT_WRITE. If the
	* pool has unsupported features but cannot be opened at all, its
	* status is ZPOOL_STATUS_UNSUP_FEAT_READ.
	*/
	ZPOOL_STATUS_UNSUP_FEAT_READ, /* unsupported features for read */
	ZPOOL_STATUS_UNSUP_FEAT_WRITE, /* unsupported features for write */

	/*
	* These faults have no corresponding message ID. At the time we are
	* checking the status, the original reason for the FMA fault (I/O or
	* checksum errors) has been lost.
	*/
	ZPOOL_STATUS_FAULTED_DEV_R, /* faulted device with replicas */
	ZPOOL_STATUS_FAULTED_DEV_NR, /* faulted device with no replicas */

	/*
	* The following are not faults per se, but still an error possibly
	* requiring administrative attention. There is no corresponding
	* message ID.
	*/
	ZPOOL_STATUS_VERSION_OLDER, /* older legacy on-disk version */
	ZPOOL_STATUS_FEAT_DISABLED, /* supported features are disabled */
	ZPOOL_STATUS_RESILVERING, /* device being resilvered */
	ZPOOL_STATUS_OFFLINE_DEV, /* device offline */
	ZPOOL_STATUS_REMOVED_DEV, /* removed device */
	ZPOOL_STATUS_REBUILDING, /* device being rebuilt */
	ZPOOL_STATUS_REBUILD_SCRUB, /* recommend scrubbing the pool */
	ZPOOL_STATUS_NON_NATIVE_ASHIFT, /* (e.g. 512e dev with ashift of 9) */
	ZPOOL_STATUS_COMPATIBILITY_ERR, /* bad 'compatibility' property */
	ZPOOL_STATUS_INCOMPATIBLE_FEAT, /* feature set outside compatibility */

	/*
	* Finally, the following indicates a healthy pool.
	*/
	ZPOOL_STATUS_OK
	} zpool_status_t;

	_LIBZFS_H zpool_status_t zpool_get_status(zpool_handle_t , const char *,
	zpool_errata_t *);
	_LIBZFS_H zpool_status_t zpool_import_status(nvlist_t , const char *,
	zpool_errata_t *);

	/*
	* Statistics and configuration functions.
	*/
	_LIBZFS_H nvlist_t zpool_get_config(zpool_handle_t , nvlist_t **);
	_LIBZFS_H nvlist_t zpool_get_features(zpool_handle_t );
	_LIBZFS_H int zpool_refresh_stats(zpool_handle_t , boolean_t );
	_LIBZFS_H int zpool_get_errlog(zpool_handle_t , nvlist_t *);

	/*
	* Import and export functions
	*/
	_LIBZFS_H int zpool_export(zpool_handle_t , boolean_t, const char );
	_LIBZFS_H int zpool_export_force(zpool_handle_t , const char );
	_LIBZFS_H int zpool_import(libzfs_handle_t , nvlist_t , const char *,
	char *altroot);
	_LIBZFS_H int zpool_import_props(libzfs_handle_t , nvlist_t , const char *,
	nvlist_t *, int);
	_LIBZFS_H void zpool_print_unsup_feat(nvlist_t *config);

	/*
	* Miscellaneous pool functions
	*/
	struct zfs_cmd;

	_LIBZFS_H const char *const zfs_history_event_names[];

	typedef enum {
	VDEV_NAME_PATH = 1 << 0,
	VDEV_NAME_GUID = 1 << 1,
	VDEV_NAME_FOLLOW_LINKS = 1 << 2,
	VDEV_NAME_TYPE_ID = 1 << 3,
	} vdev_name_t;

	_LIBZFS_H char zpool_vdev_name(libzfs_handle_t , zpool_handle_t , nvlist_t ,
	int name_flags);
	_LIBZFS_H int zpool_upgrade(zpool_handle_t *, uint64_t);
	_LIBZFS_H int zpool_get_history(zpool_handle_t , nvlist_t , uint64_t ,
	boolean_t *);
	_LIBZFS_H int zpool_events_next(libzfs_handle_t , nvlist_t , int , unsigned,
	int);
	_LIBZFS_H int zpool_events_clear(libzfs_handle_t , int );
	_LIBZFS_H int zpool_events_seek(libzfs_handle_t *, uint64_t, int);
	_LIBZFS_H void zpool_obj_to_path_ds(zpool_handle_t *, uint64_t, uint64_t,
	char *, size_t);
	_LIBZFS_H void zpool_obj_to_path(zpool_handle_t , uint64_t, uint64_t, char ,
	size_t);
	_LIBZFS_H int zfs_ioctl(libzfs_handle_t , int, struct zfs_cmd );
	_LIBZFS_H void zpool_explain_recover(libzfs_handle_t , const char , int,
	nvlist_t *);
	_LIBZFS_H int zpool_checkpoint(zpool_handle_t *);
	_LIBZFS_H int zpool_discard_checkpoint(zpool_handle_t *);
	_LIBZFS_H boolean_t zpool_is_draid_spare(const char *);

	/*
	* Basic handle manipulations. These functions do not create or destroy the
	* underlying datasets, only the references to them.
	*/
	_LIBZFS_H zfs_handle_t zfs_open(libzfs_handle_t , const char *, int);
	_LIBZFS_H zfs_handle_t zfs_handle_dup(zfs_handle_t );
	_LIBZFS_H void zfs_close(zfs_handle_t *);
	_LIBZFS_H zfs_type_t zfs_get_type(const zfs_handle_t *);
	_LIBZFS_H zfs_type_t zfs_get_underlying_type(const zfs_handle_t *);
	_LIBZFS_H const char zfs_get_name(const zfs_handle_t );
	_LIBZFS_H zpool_handle_t zfs_get_pool_handle(const zfs_handle_t );
	_LIBZFS_H const char zfs_get_pool_name(const zfs_handle_t );

	/*
	* Property management functions. Some functions are shared with the kernel,
	* and are found in sys/fs/zfs.h.
	*/

	/*
	* zfs dataset property management
	*/
	_LIBZFS_H const char *zfs_prop_default_string(zfs_prop_t);
	_LIBZFS_H uint64_t zfs_prop_default_numeric(zfs_prop_t);
	_LIBZFS_H const char *zfs_prop_column_name(zfs_prop_t);
	_LIBZFS_H boolean_t zfs_prop_align_right(zfs_prop_t);

	_LIBZFS_H nvlist_t zfs_valid_proplist(libzfs_handle_t , zfs_type_t,
	nvlist_t , uint64_t, zfs_handle_t , zpool_handle_t *, boolean_t,
	const char *);

	_LIBZFS_H const char *zfs_prop_to_name(zfs_prop_t);
	_LIBZFS_H int zfs_prop_set(zfs_handle_t , const char , const char *);
	_LIBZFS_H int zfs_prop_set_list(zfs_handle_t , nvlist_t );
	_LIBZFS_H int zfs_prop_set_list_flags(zfs_handle_t , nvlist_t , int);
	_LIBZFS_H int zfs_prop_get(zfs_handle_t , zfs_prop_t, char , size_t,
	zprop_source_t , char , size_t, boolean_t);
	_LIBZFS_H int zfs_prop_get_recvd(zfs_handle_t , const char , char *, size_t,
	boolean_t);
	_LIBZFS_H int zfs_prop_get_numeric(zfs_handle_t , zfs_prop_t, uint64_t ,
	zprop_source_t , char , size_t);
	_LIBZFS_H int zfs_prop_get_userquota_int(zfs_handle_t *zhp,
	const char propname, uint64_t propvalue);
	_LIBZFS_H int zfs_prop_get_userquota(zfs_handle_t zhp, const char propname,
	char *propbuf, int proplen, boolean_t literal);
	_LIBZFS_H int zfs_prop_get_written_int(zfs_handle_t zhp, const char propname,
	uint64_t *propvalue);
	_LIBZFS_H int zfs_prop_get_written(zfs_handle_t zhp, const char propname,
	char *propbuf, int proplen, boolean_t literal);
	_LIBZFS_H int zfs_prop_get_feature(zfs_handle_t zhp, const char propname,
	char *buf, size_t len);
	_LIBZFS_H uint64_t getprop_uint64(zfs_handle_t , zfs_prop_t, const char *);
	_LIBZFS_H uint64_t zfs_prop_get_int(zfs_handle_t *, zfs_prop_t);
	_LIBZFS_H int zfs_prop_inherit(zfs_handle_t , const char , boolean_t);
	_LIBZFS_H const char *zfs_prop_values(zfs_prop_t);
	_LIBZFS_H int zfs_prop_is_string(zfs_prop_t prop);
	_LIBZFS_H nvlist_t zfs_get_all_props(zfs_handle_t );
	_LIBZFS_H nvlist_t zfs_get_user_props(zfs_handle_t );
	_LIBZFS_H nvlist_t zfs_get_recvd_props(zfs_handle_t );
	_LIBZFS_H nvlist_t zfs_get_clones_nvl(zfs_handle_t );

	_LIBZFS_H int zfs_wait_status(zfs_handle_t *, zfs_wait_activity_t,
	boolean_t , boolean_t );

	/*
	* zfs encryption management
	*/
	_LIBZFS_H int zfs_crypto_get_encryption_root(zfs_handle_t , boolean_t ,
	char *);
	_LIBZFS_H int zfs_crypto_create(libzfs_handle_t , char , nvlist_t *,
	nvlist_t , boolean_t stdin_available, uint8_t , uint_t );
	_LIBZFS_H int zfs_crypto_clone_check(libzfs_handle_t , zfs_handle_t , char *,
	nvlist_t *);
	_LIBZFS_H int zfs_crypto_attempt_load_keys(libzfs_handle_t , const char );
	_LIBZFS_H int zfs_crypto_load_key(zfs_handle_t , boolean_t, const char );
	_LIBZFS_H int zfs_crypto_unload_key(zfs_handle_t *);
	_LIBZFS_H int zfs_crypto_rewrap(zfs_handle_t , nvlist_t , boolean_t);

	typedef struct zprop_list {
	int pl_prop;
	char *pl_user_prop;
	struct zprop_list *pl_next;
	boolean_t pl_all;
	size_t pl_width;
	size_t pl_recvd_width;
	boolean_t pl_fixed;
	} zprop_list_t;

	_LIBZFS_H int zfs_expand_proplist(zfs_handle_t , zprop_list_t *, boolean_t,
	boolean_t);
	_LIBZFS_H void zfs_prune_proplist(zfs_handle_t , uint8_t );
	_LIBZFS_H int vdev_expand_proplist(zpool_handle_t , const char ,
	zprop_list_t **);

	#define ZFS_MOUNTPOINT_NONE "none"
	#define ZFS_MOUNTPOINT_LEGACY "legacy"

	#define ZFS_FEATURE_DISABLED "disabled"
	#define ZFS_FEATURE_ENABLED "enabled"
	#define ZFS_FEATURE_ACTIVE "active"

	#define ZFS_UNSUPPORTED_INACTIVE "inactive"
	#define ZFS_UNSUPPORTED_READONLY "readonly"

	/*
	* zpool property management
	*/
	_LIBZFS_H int zpool_expand_proplist(zpool_handle_t , zprop_list_t *,
	zfs_type_t, boolean_t);
	_LIBZFS_H int zpool_prop_get_feature(zpool_handle_t , const char , char *,
	size_t);
	_LIBZFS_H const char *zpool_prop_default_string(zpool_prop_t);
	_LIBZFS_H uint64_t zpool_prop_default_numeric(zpool_prop_t);
	_LIBZFS_H const char *zpool_prop_column_name(zpool_prop_t);
	_LIBZFS_H boolean_t zpool_prop_align_right(zpool_prop_t);

	/*
	* Functions shared by zfs and zpool property management.
	*/
	_LIBZFS_H int zprop_iter(zprop_func func, void *cb, boolean_t show_all,
	boolean_t ordered, zfs_type_t type);
	_LIBZFS_H int zprop_get_list(libzfs_handle_t , char , zprop_list_t **,
	zfs_type_t);
	_LIBZFS_H void zprop_free_list(zprop_list_t *);

	#define ZFS_GET_NCOLS 5

	typedef enum {
	GET_COL_NONE,
	GET_COL_NAME,
	GET_COL_PROPERTY,
	GET_COL_VALUE,
	GET_COL_RECVD,
	GET_COL_SOURCE
	} zfs_get_column_t;

	/*
	* Functions for printing zfs or zpool properties
	*/
	typedef struct vdev_cbdata {
	int cb_name_flags;
	char **cb_names;
	unsigned int cb_names_count;
	} vdev_cbdata_t;

	typedef struct zprop_get_cbdata {
	int cb_sources;
	zfs_get_column_t cb_columns[ZFS_GET_NCOLS];
	int cb_colwidths[ZFS_GET_NCOLS + 1];
	boolean_t cb_scripted;
	boolean_t cb_literal;
	boolean_t cb_first;
	zprop_list_t *cb_proplist;
	zfs_type_t cb_type;
	vdev_cbdata_t cb_vdevs;
	} zprop_get_cbdata_t;

	#define ZFS_SET_NOMOUNT 1

	typedef struct zprop_set_cbdata {
	int cb_flags;
	nvlist_t *cb_proplist;
	} zprop_set_cbdata_t;

	_LIBZFS_H void zprop_print_one_property(const char , zprop_get_cbdata_t ,
	const char , const char , zprop_source_t, const char *,
	const char *);

	/*
	* Iterator functions.
	*/
	#define ZFS_ITER_RECURSE (1 << 0)
	#define ZFS_ITER_ARGS_CAN_BE_PATHS (1 << 1)
	#define ZFS_ITER_PROP_LISTSNAPS (1 << 2)
	#define ZFS_ITER_DEPTH_LIMIT (1 << 3)
	#define ZFS_ITER_RECVD_PROPS (1 << 4)
	#define ZFS_ITER_LITERAL_PROPS (1 << 5)
	#define ZFS_ITER_SIMPLE (1 << 6)

	typedef int (zfs_iter_f)(zfs_handle_t , void *);
	_LIBZFS_H int zfs_iter_root(libzfs_handle_t , zfs_iter_f, void );
	_LIBZFS_H int zfs_iter_children(zfs_handle_t , zfs_iter_f, void );
	_LIBZFS_H int zfs_iter_dependents(zfs_handle_t *, boolean_t, zfs_iter_f,
	void *);
	_LIBZFS_H int zfs_iter_filesystems(zfs_handle_t , zfs_iter_f, void );
	_LIBZFS_H int zfs_iter_snapshots(zfs_handle_t , boolean_t, zfs_iter_f, void ,
	uint64_t, uint64_t);
	_LIBZFS_H int zfs_iter_snapshots_sorted(zfs_handle_t , zfs_iter_f, void ,
	uint64_t, uint64_t);
	_LIBZFS_H int zfs_iter_snapspec(zfs_handle_t , const char , zfs_iter_f,
	void *);
	_LIBZFS_H int zfs_iter_bookmarks(zfs_handle_t , zfs_iter_f, void );

	_LIBZFS_H int zfs_iter_children_v2(zfs_handle_t , int, zfs_iter_f, void );
	_LIBZFS_H int zfs_iter_dependents_v2(zfs_handle_t *, int, boolean_t, zfs_iter_f,
	void *);
	_LIBZFS_H int zfs_iter_filesystems_v2(zfs_handle_t , int, zfs_iter_f, void );
	_LIBZFS_H int zfs_iter_snapshots_v2(zfs_handle_t , int, zfs_iter_f, void ,
	uint64_t, uint64_t);
	_LIBZFS_H int zfs_iter_snapshots_sorted_v2(zfs_handle_t *, int, zfs_iter_f,
	void *, uint64_t, uint64_t);
	_LIBZFS_H int zfs_iter_snapspec_v2(zfs_handle_t , int, const char ,
	zfs_iter_f, void *);
	_LIBZFS_H int zfs_iter_bookmarks_v2(zfs_handle_t , int, zfs_iter_f, void );
	_LIBZFS_H int zfs_iter_mounted(zfs_handle_t , zfs_iter_f, void );

	typedef struct get_all_cb {
	zfs_handle_t **cb_handles;
	size_t cb_alloc;
	size_t cb_used;
	} get_all_cb_t;

	_LIBZFS_H void zfs_foreach_mountpoint(libzfs_handle_t , zfs_handle_t *,
	- size_t, zfs_iter_f, void *, boolean_t);
	+ size_t, zfs_iter_f, void *, uint_t);
	_LIBZFS_H void libzfs_add_handle(get_all_cb_t , zfs_handle_t );

	/*
	* Functions to create and destroy datasets.
	*/
	_LIBZFS_H int zfs_create(libzfs_handle_t , const char , zfs_type_t,
	nvlist_t *);
	_LIBZFS_H int zfs_create_ancestors(libzfs_handle_t , const char );
	_LIBZFS_H int zfs_destroy(zfs_handle_t *, boolean_t);
	_LIBZFS_H int zfs_destroy_snaps(zfs_handle_t , char , boolean_t);
	_LIBZFS_H int zfs_destroy_snaps_nvl(libzfs_handle_t , nvlist_t , boolean_t);
	_LIBZFS_H int zfs_destroy_snaps_nvl_os(libzfs_handle_t , nvlist_t );
	_LIBZFS_H int zfs_clone(zfs_handle_t , const char , nvlist_t *);
	_LIBZFS_H int zfs_snapshot(libzfs_handle_t , const char , boolean_t,
	nvlist_t *);
	_LIBZFS_H int zfs_snapshot_nvl(libzfs_handle_t hdl, nvlist_t snaps,
	nvlist_t *props);
	_LIBZFS_H int zfs_rollback(zfs_handle_t , zfs_handle_t , boolean_t);

	typedef struct renameflags {
	/* recursive rename */
	unsigned int recursive : 1;

	/* don't unmount file systems */
	unsigned int nounmount : 1;

	/* force unmount file systems */
	unsigned int forceunmount : 1;
	} renameflags_t;

	_LIBZFS_H int zfs_rename(zfs_handle_t , const char , renameflags_t);

	typedef struct sendflags {
	/* Amount of extra information to print. */
	int verbosity;

	/* recursive send (ie, -R) */
	boolean_t replicate;

	/* for recursive send, skip sending missing snapshots */
	boolean_t skipmissing;

	/* for incrementals, do all intermediate snapshots */
	boolean_t doall;

	/* if dataset is a clone, do incremental from its origin */
	boolean_t fromorigin;

	/* field no longer used, maintained for backwards compatibility */
	boolean_t pad;

	/* send properties (ie, -p) */
	boolean_t props;

	/* do not send (no-op, ie. -n) */
	boolean_t dryrun;

	/* parsable verbose output (ie. -P) */
	boolean_t parsable;

	/* show progress (ie. -v) */
	boolean_t progress;

	/* show progress as process title (ie. -V) */
	boolean_t progressastitle;

	/* large blocks (>128K) are permitted */
	boolean_t largeblock;

	/* WRITE_EMBEDDED records of type DATA are permitted */
	boolean_t embed_data;

	/* compressed WRITE records are permitted */
	boolean_t compress;

	/* raw encrypted records are permitted */
	boolean_t raw;

	/* only send received properties (ie. -b) */
	boolean_t backup;

	/* include snapshot holds in send stream */
	boolean_t holds;

	/* stream represents a partially received dataset */
	boolean_t saved;
	} sendflags_t;

	typedef boolean_t (snapfilter_cb_t)(zfs_handle_t , void );

	_LIBZFS_H int zfs_send(zfs_handle_t , const char , const char *,
	sendflags_t , int, snapfilter_cb_t, void , nvlist_t **);
	_LIBZFS_H int zfs_send_one(zfs_handle_t , const char , int, sendflags_t *,
	const char *);
	_LIBZFS_H int zfs_send_progress(zfs_handle_t , int, uint64_t , uint64_t *);
	_LIBZFS_H int zfs_send_resume(libzfs_handle_t , sendflags_t , int outfd,
	const char *);
	_LIBZFS_H int zfs_send_saved(zfs_handle_t , sendflags_t , int, const char *);
	_LIBZFS_H nvlist_t zfs_send_resume_token_to_nvlist(libzfs_handle_t hdl,
	const char *token);

	_LIBZFS_H int zfs_promote(zfs_handle_t *);
	_LIBZFS_H int zfs_hold(zfs_handle_t , const char , const char *,
	boolean_t, int);
	_LIBZFS_H int zfs_hold_nvl(zfs_handle_t , int, nvlist_t );
	_LIBZFS_H int zfs_release(zfs_handle_t , const char , const char *,
	boolean_t);
	_LIBZFS_H int zfs_get_holds(zfs_handle_t , nvlist_t *);
	_LIBZFS_H uint64_t zvol_volsize_to_reservation(zpool_handle_t *, uint64_t,
	nvlist_t *);

	typedef int (zfs_userspace_cb_t)(void arg, const char *domain,
	uid_t rid, uint64_t space);

	_LIBZFS_H int zfs_userspace(zfs_handle_t *, zfs_userquota_prop_t,
	zfs_userspace_cb_t, void *);

	_LIBZFS_H int zfs_get_fsacl(zfs_handle_t , nvlist_t *);
	_LIBZFS_H int zfs_set_fsacl(zfs_handle_t , boolean_t, nvlist_t );

	typedef struct recvflags {
	/* print informational messages (ie, -v was specified) */
	boolean_t verbose;

	/* the destination is a prefix, not the exact fs (ie, -d) */
	boolean_t isprefix;

	/*
	* Only the tail of the sent snapshot path is appended to the
	* destination to determine the received snapshot name (ie, -e).
	*/
	boolean_t istail;

	/* do not actually do the recv, just check if it would work (ie, -n) */
	boolean_t dryrun;

	/* rollback/destroy filesystems as necessary (eg, -F) */
	boolean_t force;

	/* set "canmount=off" on all modified filesystems */
	boolean_t canmountoff;

	/*
	* Mark the file systems as "resumable" and do not destroy them if the
	* receive is interrupted
	*/
	boolean_t resumable;

	/* byteswap flag is used internally; callers need not specify */
	boolean_t byteswap;

	/* do not mount file systems as they are extracted (private) */
	boolean_t nomount;

	/* Was holds flag set in the compound header? */
	boolean_t holds;

	/* skip receive of snapshot holds */
	boolean_t skipholds;

	/* mount the filesystem unless nomount is specified */
	boolean_t domount;

	/* force unmount while recv snapshot (private) */
	boolean_t forceunmount;

	/* use this recv to check (and heal if needed) an existing snapshot */
	boolean_t heal;
	} recvflags_t;

	_LIBZFS_H int zfs_receive(libzfs_handle_t , const char , nvlist_t *,
	recvflags_t , int, avl_tree_t );

	typedef enum diff_flags {
	ZFS_DIFF_PARSEABLE = 1 << 0,
	ZFS_DIFF_TIMESTAMP = 1 << 1,
	ZFS_DIFF_CLASSIFY = 1 << 2,
	ZFS_DIFF_NO_MANGLE = 1 << 3
	} diff_flags_t;

	_LIBZFS_H int zfs_show_diffs(zfs_handle_t , int, const char , const char *,
	int);

	/*
	* Miscellaneous functions.
	*/
	_LIBZFS_H const char *zfs_type_to_name(zfs_type_t);
	_LIBZFS_H void zfs_refresh_properties(zfs_handle_t *);
	_LIBZFS_H int zfs_name_valid(const char *, zfs_type_t);
	_LIBZFS_H zfs_handle_t zfs_path_to_zhandle(libzfs_handle_t , const char *,
	zfs_type_t);
	_LIBZFS_H int zfs_parent_name(zfs_handle_t , char , size_t);
	_LIBZFS_H boolean_t zfs_dataset_exists(libzfs_handle_t , const char ,
	zfs_type_t);
	_LIBZFS_H int zfs_spa_version(zfs_handle_t , int );
	_LIBZFS_H boolean_t zfs_bookmark_exists(const char *path);

	/*
	* Mount support functions.
	*/
	_LIBZFS_H boolean_t is_mounted(libzfs_handle_t , const char special, char **);
	_LIBZFS_H boolean_t zfs_is_mounted(zfs_handle_t , char *);
	_LIBZFS_H int zfs_mount(zfs_handle_t , const char , int);
	_LIBZFS_H int zfs_mount_at(zfs_handle_t , const char , int, const char *);
	_LIBZFS_H int zfs_unmount(zfs_handle_t , const char , int);
	_LIBZFS_H int zfs_unmountall(zfs_handle_t *, int);
	_LIBZFS_H int zfs_mount_delegation_check(void);

	#if defined(__linux__) \|\| defined(__APPLE__)
	_LIBZFS_H int zfs_parse_mount_options(const char *mntopts,
	unsigned long mntflags, unsigned long zfsflags, int sloppy, char *badopt,
	char *mtabopt);
	_LIBZFS_H void zfs_adjust_mount_options(zfs_handle_t zhp, const char mntpoint,
	char mntopts, char mtabopt);
	#endif

	/*
	* Share support functions.
	*
	* enum sa_protocol * lists are terminated with SA_NO_PROTOCOL,
	* NULL means "all/any known to this libzfs".
	*/
	#define SA_NO_PROTOCOL -1

	_LIBZFS_H boolean_t zfs_is_shared(zfs_handle_t zhp, char *where,
	const enum sa_protocol *proto);
	_LIBZFS_H int zfs_share(zfs_handle_t zhp, const enum sa_protocol proto);
	_LIBZFS_H int zfs_unshare(zfs_handle_t zhp, const char mountpoint,
	const enum sa_protocol *proto);
	_LIBZFS_H int zfs_unshareall(zfs_handle_t *zhp,
	const enum sa_protocol *proto);
	_LIBZFS_H void zfs_commit_shares(const enum sa_protocol *proto);
	_LIBZFS_H void zfs_truncate_shares(const enum sa_protocol *proto);

	_LIBZFS_H int zfs_nicestrtonum(libzfs_handle_t , const char , uint64_t *);

	/*
	* Utility functions to run an external process.
	*/
	#define STDOUT_VERBOSE 0x01
	#define STDERR_VERBOSE 0x02
	#define NO_DEFAULT_PATH 0x04 /* Don't use $PATH to lookup the command */

	_LIBZFS_H int libzfs_run_process(const char , char *, int);
	_LIBZFS_H int libzfs_run_process_get_stdout(const char , char [], char *[],
	char *[], int );
	_LIBZFS_H int libzfs_run_process_get_stdout_nopath(const char , char [],
	char [], char [], int );

	_LIBZFS_H void libzfs_free_str_array(char **, int);

	_LIBZFS_H boolean_t libzfs_envvar_is_set(const char *);

	/*
	* Utility functions for zfs version
	*/
	_LIBZFS_H const char *zfs_version_userland(void);
	_LIBZFS_H char *zfs_version_kernel(void);
	_LIBZFS_H int zfs_version_print(void);

	/*
	* Given a device or file, determine if it is part of a pool.
	*/
	_LIBZFS_H int zpool_in_use(libzfs_handle_t , int, pool_state_t , char **,
	boolean_t *);

	/*
	* Label manipulation.
	*/
	_LIBZFS_H int zpool_clear_label(int);
	_LIBZFS_H int zpool_set_bootenv(zpool_handle_t , const nvlist_t );
	_LIBZFS_H int zpool_get_bootenv(zpool_handle_t , nvlist_t *);

	/*
	* Management interfaces for SMB ACL files
	*/

	_LIBZFS_H int zfs_smb_acl_add(libzfs_handle_t , char , char , char );
	_LIBZFS_H int zfs_smb_acl_remove(libzfs_handle_t , char , char , char );
	_LIBZFS_H int zfs_smb_acl_purge(libzfs_handle_t , char , char *);
	_LIBZFS_H int zfs_smb_acl_rename(libzfs_handle_t , char , char , char ,
	char *);

	/*
	* Enable and disable datasets within a pool by mounting/unmounting and
	* sharing/unsharing them.
	*/
	-_LIBZFS_H int zpool_enable_datasets(zpool_handle_t , const char , int);
	+_LIBZFS_H int zpool_enable_datasets(zpool_handle_t , const char , int,
	+ uint_t);
	_LIBZFS_H int zpool_disable_datasets(zpool_handle_t *, boolean_t);
	_LIBZFS_H void zpool_disable_datasets_os(zpool_handle_t *, boolean_t);
	_LIBZFS_H void zpool_disable_volume_os(const char *);

	/*
	* Parse a features file for -o compatibility
	*/
	typedef enum {
	ZPOOL_COMPATIBILITY_OK,
	ZPOOL_COMPATIBILITY_WARNTOKEN,
	ZPOOL_COMPATIBILITY_BADTOKEN,
	ZPOOL_COMPATIBILITY_BADFILE,
	ZPOOL_COMPATIBILITY_NOFILES
	} zpool_compat_status_t;

	_LIBZFS_H zpool_compat_status_t zpool_load_compat(const char *,
	boolean_t , char , size_t);

	#ifdef __FreeBSD__

	/*
	* Attach/detach the given filesystem to/from the given jail.
	*/
	_LIBZFS_H int zfs_jail(zfs_handle_t *zhp, int jailid, int attach);

	/*
	* Set loader options for next boot.
	*/
	_LIBZFS_H int zpool_nextboot(libzfs_handle_t *, uint64_t, uint64_t,
	const char *);

	#endif /* __FreeBSD__ */

	#ifdef __linux__

	/*
	* Add or delete the given filesystem to/from the given user namespace.
	*/
	_LIBZFS_H int zfs_userns(zfs_handle_t zhp, const char nspath, int attach);

	#endif

	#ifdef __cplusplus
	}
	#endif

	#endif /* _LIBZFS_H */
	diff --git a/sys/contrib/openzfs/include/os/freebsd/spl/sys/ccompile.h b/sys/contrib/openzfs/include/os/freebsd/spl/sys/ccompile.h
	index 26cf4db87aea..bebeb0db2452 100644
	--- a/sys/contrib/openzfs/include/os/freebsd/spl/sys/ccompile.h
	+++ b/sys/contrib/openzfs/include/os/freebsd/spl/sys/ccompile.h
	@@ -1,192 +1,193 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License, Version 1.0 only
	* (the "License"). You may not use this file except in compliance
	* with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright 2004 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/

	#ifndef _SYS_CCOMPILE_H
	#define _SYS_CCOMPILE_H

	/*
	* This file contains definitions designed to enable different compilers
	* to be used harmoniously on Solaris systems.
	*/

	#ifdef __cplusplus
	extern "C" {
	#endif

	#if defined(INVARIANTS) && !defined(ZFS_DEBUG)
	#define ZFS_DEBUG
	#undef NDEBUG
	#endif

	#define EXPORT_SYMBOL(x)
	#define module_param(a, b, c)
	#define module_param_call(a, b, c, d, e)
	#define module_param_named(a, b, c, d)
	#define MODULE_PARM_DESC(a, b)
	#define asm __asm
	#ifdef ZFS_DEBUG
	#undef NDEBUG
	#endif
	#if !defined(ZFS_DEBUG) && !defined(NDEBUG)
	#define NDEBUG
	#endif

	#ifndef EINTEGRITY
	#define EINTEGRITY 97 /* EINTEGRITY is new in 13 */
	#endif

	/*
	* These are bespoke errnos used in ZFS. We map them to their closest FreeBSD
	* equivalents. This gives us more useful error messages from strerror(3).
	*/
	#define ECKSUM EINTEGRITY
	#define EFRAGS ENOSPC

	/* Similar for ENOACTIVE */
	#define ENOTACTIVE ECANCELED

	#define EREMOTEIO EREMOTE
	#define ECHRNG ENXIO
	#define ETIME ETIMEDOUT

	#ifndef LOCORE
	#ifndef HAVE_RPC_TYPES
	#ifndef _KERNEL
	typedef int bool_t;
	typedef int enum_t;
	#endif
	#endif
	#endif

	#ifndef __cplusplus
	#define __init
	#define __exit
	#endif

	#if defined(_KERNEL) \|\| defined(_STANDALONE)
	#define param_set_charp(a, b) (0)
	#define ATTR_UID AT_UID
	#define ATTR_GID AT_GID
	#define ATTR_MODE AT_MODE
	#define ATTR_XVATTR AT_XVATTR
	#define ATTR_CTIME AT_CTIME
	#define ATTR_MTIME AT_MTIME
	#define ATTR_ATIME AT_ATIME
	#if defined(_STANDALONE)
	#define vmem_free kmem_free
	#define vmem_zalloc kmem_zalloc
	#define vmem_alloc kmem_zalloc
	#else
	#define vmem_free zfs_kmem_free
	#define vmem_zalloc(size, flags) zfs_kmem_alloc(size, flags \| M_ZERO)
	#define vmem_alloc zfs_kmem_alloc
	#endif
	#define MUTEX_NOLOCKDEP 0
	#define RW_NOLOCKDEP 0

	#else
	#define FALSE 0
	#define TRUE 1
	/*
	* XXX We really need to consolidate on standard
	* error codes in the common code
	*/
	#define ENOSTR ENOTCONN
	#define ENODATA EINVAL


	#define __BSD_VISIBLE 1
	#ifndef IN_BASE
	#define __POSIX_VISIBLE 201808
	#define __XSI_VISIBLE 1000
	#endif
	#define ARRAY_SIZE(a) (sizeof (a) / sizeof (a[0]))
	#define mmap64 mmap
	/* Note: this file can be used on linux/macOS when bootstrapping tools. */
	#if defined(__FreeBSD__)
	#define open64 open
	#define pwrite64 pwrite
	#define ftruncate64 ftruncate
	#define lseek64 lseek
	#define pread64 pread
	#define stat64 stat
	#define lstat64 lstat
	#define statfs64 statfs
	#define readdir64 readdir
	#define dirent64 dirent
	#endif
	-#define P2ALIGN(x, align) ((x) & -(align))
	+// Deprecated. Use P2ALIGN_TYPED instead.
	+// #define P2ALIGN(x, align) ((x) & -(align))
	#define P2CROSS(x, y, align) (((x) ^ (y)) > (align) - 1)
	#define P2ROUNDUP(x, align) ((((x) - 1) \| ((align) - 1)) + 1)
	#define P2PHASE(x, align) ((x) & ((align) - 1))
	#define P2NPHASE(x, align) (-(x) & ((align) - 1))
	#define ISP2(x) (((x) & ((x) - 1)) == 0)
	#define IS_P2ALIGNED(v, a) ((((uintptr_t)(v)) & ((uintptr_t)(a) - 1)) == 0)
	#define P2BOUNDARY(off, len, align) \
	(((off) ^ ((off) + (len) - 1)) > (align) - 1)

	/*
	* Typed version of the P2* macros. These macros should be used to ensure
	* that the result is correctly calculated based on the data type of (x),
	* which is passed in as the last argument, regardless of the data
	* type of the alignment. For example, if (x) is of type uint64_t,
	* and we want to round it up to a page boundary using "PAGESIZE" as
	* the alignment, we can do either
	*
	* P2ROUNDUP(x, (uint64_t)PAGESIZE)
	* or
	* P2ROUNDUP_TYPED(x, PAGESIZE, uint64_t)
	*/
	#define P2ALIGN_TYPED(x, align, type) \
	((type)(x) & -(type)(align))
	#define P2PHASE_TYPED(x, align, type) \
	((type)(x) & ((type)(align) - 1))
	#define P2NPHASE_TYPED(x, align, type) \
	(-(type)(x) & ((type)(align) - 1))
	#define P2ROUNDUP_TYPED(x, align, type) \
	((((type)(x) - 1) \| ((type)(align) - 1)) + 1)
	#define P2END_TYPED(x, align, type) \
	(-(~(type)(x) & -(type)(align)))
	#define P2PHASEUP_TYPED(x, align, phase, type) \
	((type)(phase) - (((type)(phase) - (type)(x)) & -(type)(align)))
	#define P2CROSS_TYPED(x, y, align, type) \
	(((type)(x) ^ (type)(y)) > (type)(align) - 1)
	#define P2SAMEHIGHBIT_TYPED(x, y, type) \
	(((type)(x) ^ (type)(y)) < ((type)(x) & (type)(y)))

	#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
	#define RLIM64_INFINITY RLIM_INFINITY
	#ifndef HAVE_ERESTART
	#define ERESTART EAGAIN
	#endif
	#define ABS(a) ((a) < 0 ? -(a) : (a))

	#endif
	#ifdef __cplusplus
	}
	#endif

	#endif /* _SYS_CCOMPILE_H */
	diff --git a/sys/contrib/openzfs/include/os/freebsd/spl/sys/misc.h b/sys/contrib/openzfs/include/os/freebsd/spl/sys/misc.h
	index 2e4efc60544a..894ccd8bf9b1 100644
	--- a/sys/contrib/openzfs/include/os/freebsd/spl/sys/misc.h
	+++ b/sys/contrib/openzfs/include/os/freebsd/spl/sys/misc.h
	@@ -1,58 +1,58 @@
	/*
	* Copyright (c) 2007 Pawel Jakub Dawidek <pjd@FreeBSD.org>
	* All rights reserved.
	*
	* Redistribution and use in source and binary forms, with or without
	* modification, are permitted provided that the following conditions
	* are met:
	* 1. Redistributions of source code must retain the above copyright
	* notice, this list of conditions and the following disclaimer.
	* 2. Redistributions in binary form must reproduce the above copyright
	* notice, this list of conditions and the following disclaimer in the
	* documentation and/or other materials provided with the distribution.
	*
	* THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
	* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
	* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
	* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
	* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
	* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
	* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
	* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
	* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
	* SUCH DAMAGE.
	*
	* $FreeBSD$
	*/

	#ifndef _OPENSOLARIS_SYS_MISC_H_
	#define _OPENSOLARIS_SYS_MISC_H_

	#include <sys/limits.h>
	#include <sys/filio.h>

	#define MAXUID UID_MAX

	#define _ACL_ACLENT_ENABLED 0x1
	#define _ACL_ACE_ENABLED 0x2

	#define _FIOFFS (INT_MIN)
	#define _FIOGDIO (INT_MIN+1)
	#define _FIOSDIO (INT_MIN+2)

	#define F_SEEK_DATA FIOSEEKDATA
	#define F_SEEK_HOLE FIOSEEKHOLE

	struct opensolaris_utsname {
	- char *sysname;
	- char *nodename;
	- char *release;
	- char version[32];
	- char *machine;
	+ const char *sysname;
	+ const char *nodename;
	+ const char *release;
	+ char version[32];
	+ const char *machine;
	};

	#define task_io_account_read(n)
	#define task_io_account_write(n)

	#endif /* _OPENSOLARIS_SYS_MISC_H_ */
	diff --git a/sys/contrib/openzfs/include/os/freebsd/spl/sys/sig.h b/sys/contrib/openzfs/include/os/freebsd/spl/sys/sig.h
	index a4d440d38326..17fc65cbe3e2 100644
	--- a/sys/contrib/openzfs/include/os/freebsd/spl/sys/sig.h
	+++ b/sys/contrib/openzfs/include/os/freebsd/spl/sys/sig.h
	@@ -1,70 +1,64 @@
	/*
	* Copyright (c) 2008 Pawel Jakub Dawidek <pjd@FreeBSD.org>
	* All rights reserved.
	*
	* Redistribution and use in source and binary forms, with or without
	* modification, are permitted provided that the following conditions
	* are met:
	* 1. Redistributions of source code must retain the above copyright
	* notice, this list of conditions and the following disclaimer.
	* 2. Redistributions in binary form must reproduce the above copyright
	* notice, this list of conditions and the following disclaimer in the
	* documentation and/or other materials provided with the distribution.
	*
	* THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
	* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
	* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
	* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
	* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
	* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
	* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
	* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
	* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
	* SUCH DAMAGE.
	*
	* $FreeBSD$
	*/

	#ifndef _OPENSOLARIS_SYS_SIG_H_
	#define _OPENSOLARIS_SYS_SIG_H_

	#ifndef _STANDALONE

	#include_next <sys/signal.h>
	#include <sys/param.h>
	#include <sys/lock.h>
	#include <sys/mutex.h>
	#include <sys/proc.h>
	#include <sys/signalvar.h>
	#include <sys/debug.h>

	-#define FORREAL 0
	-#define JUSTLOOKING 1
	-
	static __inline int
	-issig(int why)
	+issig(void)
	{
	struct thread *td = curthread;
	struct proc *p;
	int sig;

	- ASSERT(why == FORREAL \|\| why == JUSTLOOKING);
	if (SIGPENDING(td)) {
	- if (why == JUSTLOOKING)
	- return (1);
	p = td->td_proc;
	PROC_LOCK(p);
	mtx_lock(&p->p_sigacts->ps_mtx);
	sig = cursig(td);
	mtx_unlock(&p->p_sigacts->ps_mtx);
	PROC_UNLOCK(p);
	if (sig != 0)
	return (1);
	}
	return (0);
	}

	#endif /* !_STANDALONE */

	#endif /* _OPENSOLARIS_SYS_SIG_H_ */
	diff --git a/sys/contrib/openzfs/include/os/freebsd/spl/sys/sysmacros.h b/sys/contrib/openzfs/include/os/freebsd/spl/sys/sysmacros.h
	index 3e8841ae66bd..2c9f4438d769 100644
	--- a/sys/contrib/openzfs/include/os/freebsd/spl/sys/sysmacros.h
	+++ b/sys/contrib/openzfs/include/os/freebsd/spl/sys/sysmacros.h
	@@ -1,370 +1,371 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/* Copyright (c) 1984, 1986, 1987, 1988, 1989 AT&T */
	/* All Rights Reserved */


	/*
	* Copyright 2008 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/

	#ifndef _SYS_SYSMACROS_H
	#define _SYS_SYSMACROS_H

	#include <sys/param.h>
	#include <sys/systm.h>
	#include <sys/isa_defs.h>
	#include <sys/libkern.h>
	#include <sys/zone.h>
	#include <sys/condvar.h>

	#ifdef __cplusplus
	extern "C" {
	#endif

	/*
	* Some macros for units conversion
	*/
	/*
	* Disk blocks (sectors) and bytes.
	*/
	#define dtob(DD) ((DD) << DEV_BSHIFT)
	#define btod(BB) (((BB) + DEV_BSIZE - 1) >> DEV_BSHIFT)
	#define btodt(BB) ((BB) >> DEV_BSHIFT)
	#define lbtod(BB) (((offset_t)(BB) + DEV_BSIZE - 1) >> DEV_BSHIFT)

	/* common macros */
	#ifndef MIN
	#define MIN(a, b) ((a) < (b) ? (a) : (b))
	#endif
	#ifndef MAX
	#define MAX(a, b) ((a) < (b) ? (b) : (a))
	#endif
	#ifndef ABS
	#define ABS(a) ((a) < 0 ? -(a) : (a))
	#endif
	#ifndef SIGNOF
	#define SIGNOF(a) ((a) < 0 ? -1 : (a) > 0)
	#endif
	#ifndef ARRAY_SIZE
	#define ARRAY_SIZE(a) (sizeof (a) / sizeof (a[0]))
	#endif
	#ifndef DIV_ROUND_UP
	#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
	#endif

	#ifdef _STANDALONE
	#define boot_ncpus 1
	#else /* _STANDALONE */
	#define boot_ncpus mp_ncpus
	#endif /* _STANDALONE */
	#define kpreempt_disable() critical_enter()
	#define kpreempt_enable() critical_exit()
	#define CPU_SEQID curcpu
	#define CPU_SEQID_UNSTABLE curcpu
	#define is_system_labeled() 0
	/*
	* Convert a single byte to/from binary-coded decimal (BCD).
	*/
	extern unsigned char byte_to_bcd[256];
	extern unsigned char bcd_to_byte[256];

	#define BYTE_TO_BCD(x) byte_to_bcd[(x) & 0xff]
	#define BCD_TO_BYTE(x) bcd_to_byte[(x) & 0xff]

	/*
	* WARNING: The device number macros defined here should not be used by device
	* drivers or user software. Device drivers should use the device functions
	* defined in the DDI/DKI interface (see also ddi.h). Application software
	* should make use of the library routines available in makedev(3). A set of
	* new device macros are provided to operate on the expanded device number
	* format supported in SVR4. Macro versions of the DDI device functions are
	* provided for use by kernel proper routines only. Macro routines bmajor(),
	* major(), minor(), emajor(), eminor(), and makedev() will be removed or
	* their definitions changed at the next major release following SVR4.
	*/

	#define O_BITSMAJOR 7 /* # of SVR3 major device bits */
	#define O_BITSMINOR 8 /* # of SVR3 minor device bits */
	#define O_MAXMAJ 0x7f /* SVR3 max major value */
	#define O_MAXMIN 0xff /* SVR3 max minor value */


	#define L_BITSMAJOR32 14 /* # of SVR4 major device bits */
	#define L_BITSMINOR32 18 /* # of SVR4 minor device bits */
	#define L_MAXMAJ32 0x3fff /* SVR4 max major value */
	#define L_MAXMIN32 0x3ffff /* MAX minor for 3b2 software drivers. */
	/* For 3b2 hardware devices the minor is */
	/* restricted to 256 (0-255) */

	#ifdef _LP64
	#define L_BITSMAJOR 32 /* # of major device bits in 64-bit Solaris */
	#define L_BITSMINOR 32 /* # of minor device bits in 64-bit Solaris */
	#define L_MAXMAJ 0xfffffffful /* max major value */
	#define L_MAXMIN 0xfffffffful /* max minor value */
	#else
	#define L_BITSMAJOR L_BITSMAJOR32
	#define L_BITSMINOR L_BITSMINOR32
	#define L_MAXMAJ L_MAXMAJ32
	#define L_MAXMIN L_MAXMIN32
	#endif

	/*
	* These are versions of the kernel routines for compressing and
	* expanding long device numbers that don't return errors.
	*/
	#if (L_BITSMAJOR32 == L_BITSMAJOR) && (L_BITSMINOR32 == L_BITSMINOR)

	#define DEVCMPL(x) (x)
	#define DEVEXPL(x) (x)

	#else

	#define DEVCMPL(x) \
	(dev32_t)((((x) >> L_BITSMINOR) > L_MAXMAJ32 \|\| \
	((x) & L_MAXMIN) > L_MAXMIN32) ? NODEV32 : \
	((((x) >> L_BITSMINOR) << L_BITSMINOR32) \| ((x) & L_MAXMIN32)))

	#define DEVEXPL(x) \
	(((x) == NODEV32) ? NODEV : \
	makedevice(((x) >> L_BITSMINOR32) & L_MAXMAJ32, (x) & L_MAXMIN32))

	#endif /* L_BITSMAJOR32 ... */

	/* convert to old (SVR3.2) dev format */

	#define cmpdev(x) \
	(o_dev_t)((((x) >> L_BITSMINOR) > O_MAXMAJ \|\| \
	((x) & L_MAXMIN) > O_MAXMIN) ? NODEV : \
	((((x) >> L_BITSMINOR) << O_BITSMINOR) \| ((x) & O_MAXMIN)))

	/* convert to new (SVR4) dev format */

	#define expdev(x) \
	(dev_t)(((dev_t)(((x) >> O_BITSMINOR) & O_MAXMAJ) << L_BITSMINOR) \| \
	((x) & O_MAXMIN))

	/*
	* Macro for checking power of 2 address alignment.
	*/
	#define IS_P2ALIGNED(v, a) ((((uintptr_t)(v)) & ((uintptr_t)(a) - 1)) == 0)

	/*
	* Macros for counting and rounding.
	*/
	#define howmany(x, y) (((x)+((y)-1))/(y))
	#define roundup(x, y) ((((x)+((y)-1))/(y))*(y))

	/*
	* Macro to determine if value is a power of 2
	*/
	#define ISP2(x) (((x) & ((x) - 1)) == 0)

	/*
	* Macros for various sorts of alignment and rounding. The "align" must
	* be a power of 2. Often times it is a block, sector, or page.
	*/

	/*
	* return x rounded down to an align boundary
	* eg, P2ALIGN(1200, 1024) == 1024 (1*align)
	* eg, P2ALIGN(1024, 1024) == 1024 (1*align)
	* eg, P2ALIGN(0x1234, 0x100) == 0x1200 (0x12*align)
	* eg, P2ALIGN(0x5600, 0x100) == 0x5600 (0x56*align)
	*/
	-#define P2ALIGN(x, align) ((x) & -(align))
	+// Deprecated. Use P2ALIGN_TYPED instead.
	+// #define P2ALIGN(x, align) ((x) & -(align))

	/*
	* return x % (mod) align
	* eg, P2PHASE(0x1234, 0x100) == 0x34 (x-0x12*align)
	* eg, P2PHASE(0x5600, 0x100) == 0x00 (x-0x56*align)
	*/
	#define P2PHASE(x, align) ((x) & ((align) - 1))

	/*
	* return how much space is left in this block (but if it's perfectly
	* aligned, return 0).
	* eg, P2NPHASE(0x1234, 0x100) == 0xcc (0x13*align-x)
	* eg, P2NPHASE(0x5600, 0x100) == 0x00 (0x56*align-x)
	*/
	#define P2NPHASE(x, align) (-(x) & ((align) - 1))

	/*
	* return x rounded up to an align boundary
	* eg, P2ROUNDUP(0x1234, 0x100) == 0x1300 (0x13*align)
	* eg, P2ROUNDUP(0x5600, 0x100) == 0x5600 (0x56*align)
	*/
	#define P2ROUNDUP(x, align) (-(-(x) & -(align)))

	/*
	* return the ending address of the block that x is in
	* eg, P2END(0x1234, 0x100) == 0x12ff (0x13*align - 1)
	* eg, P2END(0x5600, 0x100) == 0x56ff (0x57*align - 1)
	*/
	#define P2END(x, align) (-(~(x) & -(align)))

	/*
	* return x rounded up to the next phase (offset) within align.
	* phase should be < align.
	* eg, P2PHASEUP(0x1234, 0x100, 0x10) == 0x1310 (0x13*align + phase)
	* eg, P2PHASEUP(0x5600, 0x100, 0x10) == 0x5610 (0x56*align + phase)
	*/
	#define P2PHASEUP(x, align, phase) ((phase) - (((phase) - (x)) & -(align)))

	/*
	* return TRUE if adding len to off would cause it to cross an align
	* boundary.
	* eg, P2BOUNDARY(0x1234, 0xe0, 0x100) == TRUE (0x1234 + 0xe0 == 0x1314)
	* eg, P2BOUNDARY(0x1234, 0x50, 0x100) == FALSE (0x1234 + 0x50 == 0x1284)
	*/
	#define P2BOUNDARY(off, len, align) \
	(((off) ^ ((off) + (len) - 1)) > (align) - 1)

	/*
	* Return TRUE if they have the same highest bit set.
	* eg, P2SAMEHIGHBIT(0x1234, 0x1001) == TRUE (the high bit is 0x1000)
	* eg, P2SAMEHIGHBIT(0x1234, 0x3010) == FALSE (high bit of 0x3010 is 0x2000)
	*/
	#define P2SAMEHIGHBIT(x, y) (((x) ^ (y)) < ((x) & (y)))

	/*
	* Typed version of the P2* macros. These macros should be used to ensure
	* that the result is correctly calculated based on the data type of (x),
	* which is passed in as the last argument, regardless of the data
	* type of the alignment. For example, if (x) is of type uint64_t,
	* and we want to round it up to a page boundary using "PAGESIZE" as
	* the alignment, we can do either
	* P2ROUNDUP(x, (uint64_t)PAGESIZE)
	* or
	* P2ROUNDUP_TYPED(x, PAGESIZE, uint64_t)
	*/
	#define P2ALIGN_TYPED(x, align, type) \
	((type)(x) & -(type)(align))
	#define P2PHASE_TYPED(x, align, type) \
	((type)(x) & ((type)(align) - 1))
	#define P2NPHASE_TYPED(x, align, type) \
	(-(type)(x) & ((type)(align) - 1))
	#define P2ROUNDUP_TYPED(x, align, type) \
	(-(-(type)(x) & -(type)(align)))
	#define P2END_TYPED(x, align, type) \
	(-(~(type)(x) & -(type)(align)))
	#define P2PHASEUP_TYPED(x, align, phase, type) \
	((type)(phase) - (((type)(phase) - (type)(x)) & -(type)(align)))
	#define P2CROSS_TYPED(x, y, align, type) \
	(((type)(x) ^ (type)(y)) > (type)(align) - 1)
	#define P2SAMEHIGHBIT_TYPED(x, y, type) \
	(((type)(x) ^ (type)(y)) < ((type)(x) & (type)(y)))

	/*
	* Macros to atomically increment/decrement a variable. mutex and var
	* must be pointers.
	*/
	#define INCR_COUNT(var, mutex) mutex_enter(mutex), (*(var))++, mutex_exit(mutex)
	#define DECR_COUNT(var, mutex) mutex_enter(mutex), (*(var))--, mutex_exit(mutex)

	#if !defined(_KMEMUSER) && !defined(offsetof)

	/* avoid any possibility of clashing with <stddef.h> version */

	#define offsetof(type, field) __offsetof(type, field)
	#endif

	/*
	* Find highest one bit set.
	* Returns bit number + 1 of highest bit that is set, otherwise returns 0.
	* High order bit is 31 (or 63 in _LP64 kernel).
	*/
	static __inline int
	highbit(ulong_t i)
	{
	#if defined(HAVE_INLINE_FLSL)
	return (flsl(i));
	#else
	int h = 1;

	if (i == 0)
	return (0);
	#ifdef _LP64
	if (i & 0xffffffff00000000ul) {
	h += 32; i >>= 32;
	}
	#endif
	if (i & 0xffff0000) {
	h += 16; i >>= 16;
	}
	if (i & 0xff00) {
	h += 8; i >>= 8;
	}
	if (i & 0xf0) {
	h += 4; i >>= 4;
	}
	if (i & 0xc) {
	h += 2; i >>= 2;
	}
	if (i & 0x2) {
	h += 1;
	}
	return (h);
	#endif
	}

	/*
	* Find highest one bit set.
	* Returns bit number + 1 of highest bit that is set, otherwise returns 0.
	*/
	static __inline int
	highbit64(uint64_t i)
	{
	#if defined(HAVE_INLINE_FLSLL)
	return (flsll(i));
	#else
	int h = 1;

	if (i == 0)
	return (0);
	if (i & 0xffffffff00000000ULL) {
	h += 32; i >>= 32;
	}
	if (i & 0xffff0000) {
	h += 16; i >>= 16;
	}
	if (i & 0xff00) {
	h += 8; i >>= 8;
	}
	if (i & 0xf0) {
	h += 4; i >>= 4;
	}
	if (i & 0xc) {
	h += 2; i >>= 2;
	}
	if (i & 0x2) {
	h += 1;
	}
	return (h);
	#endif
	}

	#ifdef __cplusplus
	}
	#endif

	#endif /* _SYS_SYSMACROS_H */
	diff --git a/sys/contrib/openzfs/include/os/linux/spl/sys/signal.h b/sys/contrib/openzfs/include/os/linux/spl/sys/signal.h
	index 6b538c8966f2..cb4b33261647 100644
	--- a/sys/contrib/openzfs/include/os/linux/spl/sys/signal.h
	+++ b/sys/contrib/openzfs/include/os/linux/spl/sys/signal.h
	@@ -1,38 +1,35 @@
	/*
	* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
	* Copyright (C) 2007 The Regents of the University of California.
	* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
	* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
	* UCRL-CODE-235197
	*
	* This file is part of the SPL, Solaris Porting Layer.
	*
	* The SPL is free software; you can redistribute it and/or modify it
	* under the terms of the GNU General Public License as published by the
	* Free Software Foundation; either version 2 of the License, or (at your
	* option) any later version.
	*
	* The SPL is distributed in the hope that it will be useful, but WITHOUT
	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
	* for more details.
	*
	* You should have received a copy of the GNU General Public License along
	* with the SPL. If not, see <http://www.gnu.org/licenses/>.
	*/

	#ifndef _SPL_SIGNAL_H
	#define _SPL_SIGNAL_H

	#include <linux/sched.h>

	#ifdef HAVE_SCHED_SIGNAL_HEADER
	#include <linux/sched/signal.h>
	#endif

	-#define FORREAL 0 /* Usual side-effects */
	-#define JUSTLOOKING 1 /* Don't stop the process */
	-
	-extern int issig(int why);
	+extern int issig(void);

	#endif /* SPL_SIGNAL_H */
	diff --git a/sys/contrib/openzfs/include/os/linux/spl/sys/sysmacros.h b/sys/contrib/openzfs/include/os/linux/spl/sys/sysmacros.h
	index 99e3a6fb41c6..0e8390736309 100644
	--- a/sys/contrib/openzfs/include/os/linux/spl/sys/sysmacros.h
	+++ b/sys/contrib/openzfs/include/os/linux/spl/sys/sysmacros.h
	@@ -1,215 +1,216 @@
	/*
	* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
	* Copyright (C) 2007 The Regents of the University of California.
	* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
	* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
	* UCRL-CODE-235197
	*
	* This file is part of the SPL, Solaris Porting Layer.
	*
	* The SPL is free software; you can redistribute it and/or modify it
	* under the terms of the GNU General Public License as published by the
	* Free Software Foundation; either version 2 of the License, or (at your
	* option) any later version.
	*
	* The SPL is distributed in the hope that it will be useful, but WITHOUT
	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
	* for more details.
	*
	* You should have received a copy of the GNU General Public License along
	* with the SPL. If not, see <http://www.gnu.org/licenses/>.
	*/

	#ifndef _SPL_SYSMACROS_H
	#define _SPL_SYSMACROS_H

	#include <linux/module.h>
	#include <linux/sched.h>
	#include <linux/sched/rt.h>
	#include <linux/cpumask.h>
	#include <sys/debug.h>
	#include <sys/zone.h>
	#include <sys/signal.h>
	#include <asm/page.h>


	#ifndef _KERNEL
	#define _KERNEL __KERNEL__
	#endif

	#define FALSE 0
	#define TRUE 1

	#define INT8_MAX (127)
	#define INT8_MIN (-128)
	#define UINT8_MAX (255)
	#define UINT8_MIN (0)

	#define INT16_MAX (32767)
	#define INT16_MIN (-32768)
	#define UINT16_MAX (65535)
	#define UINT16_MIN (0)

	#define INT32_MAX INT_MAX
	#define INT32_MIN INT_MIN
	#define UINT32_MAX UINT_MAX
	#define UINT32_MIN UINT_MIN

	#define INT64_MAX LLONG_MAX
	#define INT64_MIN LLONG_MIN
	#define UINT64_MAX ULLONG_MAX
	#define UINT64_MIN ULLONG_MIN

	#define NBBY 8

	#define MAXMSGLEN 256
	#define MAXNAMELEN 256
	#define MAXPATHLEN 4096
	#define MAXOFFSET_T LLONG_MAX
	#define MAXBSIZE 8192
	#define DEV_BSIZE 512
	#define DEV_BSHIFT 9 /* log2(DEV_BSIZE) */

	#define proc_pageout NULL
	#define curproc current
	#define max_ncpus num_possible_cpus()
	#define boot_ncpus num_online_cpus()
	#define CPU_SEQID smp_processor_id()
	#define CPU_SEQID_UNSTABLE raw_smp_processor_id()
	#define is_system_labeled() 0

	#ifndef RLIM64_INFINITY
	#define RLIM64_INFINITY (~0ULL)
	#endif

	/*
	* 0..MAX_PRIO-1: Process priority
	* 0..MAX_RT_PRIO-1: RT priority tasks
	* MAX_RT_PRIO..MAX_PRIO-1: SCHED_NORMAL tasks
	*
	* Treat shim tasks as SCHED_NORMAL tasks
	*/
	#define minclsyspri (MAX_PRIO-1)
	#define maxclsyspri (MAX_RT_PRIO)
	#define defclsyspri (DEFAULT_PRIO)

	#ifndef NICE_TO_PRIO
	#define NICE_TO_PRIO(nice) (MAX_RT_PRIO + (nice) + 20)
	#endif
	#ifndef PRIO_TO_NICE
	#define PRIO_TO_NICE(prio) ((prio) - MAX_RT_PRIO - 20)
	#endif

	/*
	* Missing macros
	*/
	#ifndef PAGESIZE
	#define PAGESIZE PAGE_SIZE
	#endif

	#ifndef PAGESHIFT
	#define PAGESHIFT PAGE_SHIFT
	#endif

	/* Missing globals */
	extern unsigned long spl_hostid;

	/* Missing misc functions */
	extern uint32_t zone_get_hostid(void *zone);
	extern void spl_setup(void);
	extern void spl_cleanup(void);

	/*
	* Only handles the first 4096 majors and first 256 minors. We don't have a
	* libc for the kernel module so we define this inline.
	*/
	static inline dev_t
	makedev(unsigned int major, unsigned int minor)
	{
	return ((major & 0xFFF) << 8) \| (minor & 0xFF);
	}

	#define highbit(x) __fls(x)
	#define lowbit(x) __ffs(x)

	#define highbit64(x) fls64(x)
	#define makedevice(maj, min) makedev(maj, min)

	/* common macros */
	#ifndef MIN
	#define MIN(a, b) ((a) < (b) ? (a) : (b))
	#endif
	#ifndef MAX
	#define MAX(a, b) ((a) < (b) ? (b) : (a))
	#endif
	#ifndef ABS
	#define ABS(a) ((a) < 0 ? -(a) : (a))
	#endif
	#ifndef DIV_ROUND_UP
	#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
	#endif
	#ifndef roundup
	#define roundup(x, y) ((((x) + ((y) - 1)) / (y)) * (y))
	#endif
	#ifndef howmany
	#define howmany(x, y) (((x) + ((y) - 1)) / (y))
	#endif

	/*
	* Compatibility macros/typedefs needed for Solaris -> Linux port
	*/
	-#define P2ALIGN(x, align) ((x) & -(align))
	+// Deprecated. Use P2ALIGN_TYPED instead.
	+// #define P2ALIGN(x, align) ((x) & -(align))
	#define P2CROSS(x, y, align) (((x) ^ (y)) > (align) - 1)
	#define P2ROUNDUP(x, align) ((((x) - 1) \| ((align) - 1)) + 1)
	#define P2PHASE(x, align) ((x) & ((align) - 1))
	#define P2NPHASE(x, align) (-(x) & ((align) - 1))
	#define ISP2(x) (((x) & ((x) - 1)) == 0)
	#define IS_P2ALIGNED(v, a) ((((uintptr_t)(v)) & ((uintptr_t)(a) - 1)) == 0)
	#define P2BOUNDARY(off, len, align) \
	(((off) ^ ((off) + (len) - 1)) > (align) - 1)

	/*
	* Typed version of the P2* macros. These macros should be used to ensure
	* that the result is correctly calculated based on the data type of (x),
	* which is passed in as the last argument, regardless of the data
	* type of the alignment. For example, if (x) is of type uint64_t,
	* and we want to round it up to a page boundary using "PAGESIZE" as
	* the alignment, we can do either
	*
	* P2ROUNDUP(x, (uint64_t)PAGESIZE)
	* or
	* P2ROUNDUP_TYPED(x, PAGESIZE, uint64_t)
	*/
	#define P2ALIGN_TYPED(x, align, type) \
	((type)(x) & -(type)(align))
	#define P2PHASE_TYPED(x, align, type) \
	((type)(x) & ((type)(align) - 1))
	#define P2NPHASE_TYPED(x, align, type) \
	(-(type)(x) & ((type)(align) - 1))
	#define P2ROUNDUP_TYPED(x, align, type) \
	((((type)(x) - 1) \| ((type)(align) - 1)) + 1)
	#define P2END_TYPED(x, align, type) \
	(-(~(type)(x) & -(type)(align)))
	#define P2PHASEUP_TYPED(x, align, phase, type) \
	((type)(phase) - (((type)(phase) - (type)(x)) & -(type)(align)))
	#define P2CROSS_TYPED(x, y, align, type) \
	(((type)(x) ^ (type)(y)) > (type)(align) - 1)
	#define P2SAMEHIGHBIT_TYPED(x, y, type) \
	(((type)(x) ^ (type)(y)) < ((type)(x) & (type)(y)))

	#define SET_ERROR(err) \
	(__set_error(__FILE__, __func__, __LINE__, err), err)

	#include <linux/sort.h>
	#define qsort(base, num, size, cmp) \
	sort(base, num, size, cmp, NULL)

	#if !defined(_KMEMUSER) && !defined(offsetof)

	/* avoid any possibility of clashing with <stddef.h> version */

	#define offsetof(s, m) ((size_t)(&(((s *)0)->m)))
	#endif

	#endif /* _SPL_SYSMACROS_H */
	diff --git a/sys/contrib/openzfs/include/os/linux/spl/sys/types.h b/sys/contrib/openzfs/include/os/linux/spl/sys/types.h
	index 20ba457f7efe..94ba7b6ad323 100644
	--- a/sys/contrib/openzfs/include/os/linux/spl/sys/types.h
	+++ b/sys/contrib/openzfs/include/os/linux/spl/sys/types.h
	@@ -1,84 +1,86 @@
	/*
	* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
	* Copyright (C) 2007 The Regents of the University of California.
	* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
	* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
	* UCRL-CODE-235197
	*
	* This file is part of the SPL, Solaris Porting Layer.
	*
	* The SPL is free software; you can redistribute it and/or modify it
	* under the terms of the GNU General Public License as published by the
	* Free Software Foundation; either version 2 of the License, or (at your
	* option) any later version.
	*
	* The SPL is distributed in the hope that it will be useful, but WITHOUT
	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
	* for more details.
	*
	* You should have received a copy of the GNU General Public License along
	* with the SPL. If not, see <http://www.gnu.org/licenses/>.
	*/

	#ifndef _SPL_TYPES_H
	#define _SPL_TYPES_H

	#include <linux/types.h>

	typedef enum {
	B_FALSE = 0,
	B_TRUE = 1
	} boolean_t;

	typedef unsigned char uchar_t;
	typedef unsigned short ushort_t;
	typedef unsigned int uint_t;
	typedef unsigned long ulong_t;
	typedef unsigned long long u_longlong_t;
	typedef long long longlong_t;

	+#ifndef HAVE_KERNEL_INTPTR_T
	typedef long intptr_t;
	+#endif
	typedef unsigned long long rlim64_t;

	typedef struct task_struct kthread_t;
	typedef struct task_struct proc_t;

	typedef int id_t;
	typedef short pri_t;
	typedef short index_t;
	typedef longlong_t offset_t;
	typedef u_longlong_t u_offset_t;
	typedef ulong_t pgcnt_t;

	typedef int major_t;
	typedef int minor_t;

	struct user_namespace;
	#ifdef HAVE_IOPS_CREATE_IDMAP
	#include <linux/refcount.h>
	#ifdef HAVE_IDMAP_NO_USERNS
	#include <linux/user_namespace.h>
	struct mnt_idmap {
	struct uid_gid_map uid_map;
	struct uid_gid_map gid_map;
	refcount_t count;
	};
	typedef struct mnt_idmap zidmap_t;
	#define idmap_owner(p) (NULL)
	#else
	struct mnt_idmap {
	struct user_namespace *owner;
	refcount_t count;
	};
	typedef struct mnt_idmap zidmap_t;
	#define idmap_owner(p) (((struct mnt_idmap *)p)->owner)
	#endif
	#else
	typedef struct user_namespace zidmap_t;
	#define idmap_owner(p) ((struct user_namespace *)p)
	#endif

	extern zidmap_t *zfs_init_idmap;

	#endif /* _SPL_TYPES_H */
	diff --git a/sys/contrib/openzfs/include/sys/spa_impl.h b/sys/contrib/openzfs/include/sys/spa_impl.h
	index a40914ec5fcb..5605a35b8641 100644
	--- a/sys/contrib/openzfs/include/sys/spa_impl.h
	+++ b/sys/contrib/openzfs/include/sys/spa_impl.h
	@@ -1,501 +1,500 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2024 by Delphix. All rights reserved.
	* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
	* Copyright 2013 Saso Kiselkov. All rights reserved.
	* Copyright (c) 2016 Actifio, Inc. All rights reserved.
	* Copyright (c) 2017, Intel Corporation.
	* Copyright (c) 2019 Datto Inc.
	*/

	#ifndef _SYS_SPA_IMPL_H
	#define _SYS_SPA_IMPL_H

	#include <sys/spa.h>
	#include <sys/spa_checkpoint.h>
	#include <sys/spa_log_spacemap.h>
	#include <sys/vdev.h>
	#include <sys/vdev_rebuild.h>
	#include <sys/vdev_removal.h>
	#include <sys/vdev_raidz.h>
	#include <sys/metaslab.h>
	#include <sys/dmu.h>
	#include <sys/dsl_pool.h>
	#include <sys/uberblock_impl.h>
	#include <sys/zfs_context.h>
	#include <sys/avl.h>
	#include <sys/zfs_refcount.h>
	#include <sys/bplist.h>
	#include <sys/bpobj.h>
	#include <sys/dsl_crypt.h>
	#include <sys/zfeature.h>
	#include <sys/zthr.h>
	#include <sys/dsl_deadlist.h>
	#include <zfeature_common.h>

	#ifdef __cplusplus
	extern "C" {
	#endif

	typedef struct spa_alloc {
	kmutex_t spaa_lock;
	avl_tree_t spaa_tree;
	} ____cacheline_aligned spa_alloc_t;

	typedef struct spa_allocs_use {
	kmutex_t sau_lock;
	uint_t sau_rotor;
	boolean_t sau_inuse[];
	} spa_allocs_use_t;

	typedef struct spa_error_entry {
	zbookmark_phys_t se_bookmark;
	char *se_name;
	avl_node_t se_avl;
	zbookmark_err_phys_t se_zep; /* not accounted in avl_find */
	} spa_error_entry_t;

	typedef struct spa_history_phys {
	uint64_t sh_pool_create_len; /* ending offset of zpool create */
	uint64_t sh_phys_max_off; /* physical EOF */
	uint64_t sh_bof; /* logical BOF */
	uint64_t sh_eof; /* logical EOF */
	uint64_t sh_records_lost; /* num of records overwritten */
	} spa_history_phys_t;

	/*
	* All members must be uint64_t, for byteswap purposes.
	*/
	typedef struct spa_removing_phys {
	uint64_t sr_state; /* dsl_scan_state_t */

	/*
	* The vdev ID that we most recently attempted to remove,
	* or -1 if no removal has been attempted.
	*/
	uint64_t sr_removing_vdev;

	/*
	* The vdev ID that we most recently successfully removed,
	* or -1 if no devices have been removed.
	*/
	uint64_t sr_prev_indirect_vdev;

	uint64_t sr_start_time;
	uint64_t sr_end_time;

	/*
	* Note that we can not use the space map's or indirect mapping's
	* accounting as a substitute for these values, because we need to
	* count frees of not-yet-copied data as though it did the copy.
	* Otherwise, we could get into a situation where copied > to_copy,
	* or we complete before copied == to_copy.
	*/
	uint64_t sr_to_copy; /* bytes that need to be copied */
	uint64_t sr_copied; /* bytes that have been copied or freed */
	} spa_removing_phys_t;

	/*
	* This struct is stored as an entry in the DMU_POOL_DIRECTORY_OBJECT
	* (with key DMU_POOL_CONDENSING_INDIRECT). It is present if a condense
	* of an indirect vdev's mapping object is in progress.
	*/
	typedef struct spa_condensing_indirect_phys {
	/*
	* The vdev ID of the indirect vdev whose indirect mapping is
	* being condensed.
	*/
	uint64_t scip_vdev;

	/*
	* The vdev's old obsolete spacemap. This spacemap's contents are
	* being integrated into the new mapping.
	*/
	uint64_t scip_prev_obsolete_sm_object;

	/*
	* The new mapping object that is being created.
	*/
	uint64_t scip_next_mapping_object;
	} spa_condensing_indirect_phys_t;

	struct spa_aux_vdev {
	uint64_t sav_object; /* MOS object for device list */
	nvlist_t sav_config; / cached device config */
	vdev_t *sav_vdevs; / devices */
	int sav_count; /* number devices */
	boolean_t sav_sync; /* sync the device list */
	nvlist_t *sav_pending; / pending device additions */
	uint_t sav_npending; /* # pending devices */
	};

	typedef struct spa_config_lock {
	kmutex_t scl_lock;
	kthread_t *scl_writer;
	int scl_write_wanted;
	int scl_count;
	kcondvar_t scl_cv;
	} ____cacheline_aligned spa_config_lock_t;

	typedef struct spa_config_dirent {
	list_node_t scd_link;
	char *scd_path;
	} spa_config_dirent_t;

	typedef enum zio_taskq_type {
	ZIO_TASKQ_ISSUE = 0,
	ZIO_TASKQ_ISSUE_HIGH,
	ZIO_TASKQ_INTERRUPT,
	ZIO_TASKQ_INTERRUPT_HIGH,
	ZIO_TASKQ_TYPES
	} zio_taskq_type_t;

	/*
	* State machine for the zpool-poolname process. The states transitions
	* are done as follows:
	*
	* From To Routine
	* PROC_NONE -> PROC_CREATED spa_activate()
	* PROC_CREATED -> PROC_ACTIVE spa_thread()
	* PROC_ACTIVE -> PROC_DEACTIVATE spa_deactivate()
	* PROC_DEACTIVATE -> PROC_GONE spa_thread()
	* PROC_GONE -> PROC_NONE spa_deactivate()
	*/
	typedef enum spa_proc_state {
	SPA_PROC_NONE, /* spa_proc = &p0, no process created */
	SPA_PROC_CREATED, /* spa_activate() has proc, is waiting */
	SPA_PROC_ACTIVE, /* taskqs created, spa_proc set */
	SPA_PROC_DEACTIVATE, /* spa_deactivate() requests process exit */
	SPA_PROC_GONE /* spa_thread() is exiting, spa_proc = &p0 */
	} spa_proc_state_t;

	typedef struct spa_taskqs {
	uint_t stqs_count;
	taskq_t **stqs_taskq;
	} spa_taskqs_t;

	/* one for each thread in the spa sync taskq */
	typedef struct spa_syncthread_info {
	kthread_t *sti_thread;
	uint_t sti_allocator;
	} spa_syncthread_info_t;

	typedef enum spa_all_vdev_zap_action {
	AVZ_ACTION_NONE = 0,
	AVZ_ACTION_DESTROY, /* Destroy all per-vdev ZAPs and the AVZ. */
	AVZ_ACTION_REBUILD, /* Populate the new AVZ, see spa_avz_rebuild */
	AVZ_ACTION_INITIALIZE
	} spa_avz_action_t;

	typedef enum spa_config_source {
	SPA_CONFIG_SRC_NONE = 0,
	SPA_CONFIG_SRC_SCAN, /* scan of path (default: /dev/dsk) */
	SPA_CONFIG_SRC_CACHEFILE, /* any cachefile */
	SPA_CONFIG_SRC_TRYIMPORT, /* returned from call to tryimport */
	SPA_CONFIG_SRC_SPLIT, /* new pool in a pool split */
	SPA_CONFIG_SRC_MOS /* MOS, but not always from right txg */
	} spa_config_source_t;

	struct spa {
	/*
	* Fields protected by spa_namespace_lock.
	*/
	char spa_name[ZFS_MAX_DATASET_NAME_LEN]; /* pool name */
	char spa_comment; / comment */
	avl_node_t spa_avl; /* node in spa_namespace_avl */
	nvlist_t spa_config; / last synced config */
	nvlist_t spa_config_syncing; / currently syncing config */
	nvlist_t spa_config_splitting; / config for splitting */
	nvlist_t spa_load_info; / info and errors from load */
	uint64_t spa_config_txg; /* txg of last config change */
	uint32_t spa_sync_pass; /* iterate-to-convergence */
	pool_state_t spa_state; /* pool state */
	int spa_inject_ref; /* injection references */
	uint8_t spa_sync_on; /* sync threads are running */
	spa_load_state_t spa_load_state; /* current load operation */
	boolean_t spa_indirect_vdevs_loaded; /* mappings loaded? */
	boolean_t spa_trust_config; /* do we trust vdev tree? */
	boolean_t spa_is_splitting; /* in the middle of a split? */
	spa_config_source_t spa_config_source; /* where config comes from? */
	uint64_t spa_import_flags; /* import specific flags */
	spa_taskqs_t spa_zio_taskq[ZIO_TYPES][ZIO_TASKQ_TYPES];
	dsl_pool_t *spa_dsl_pool;
	boolean_t spa_is_initializing; /* true while opening pool */
	boolean_t spa_is_exporting; /* true while exporting pool */
	+ kthread_t spa_export_thread; / valid during pool export */
	kthread_t spa_load_thread; / loading, no namespace lock */
	metaslab_class_t spa_normal_class; / normal data class */
	metaslab_class_t spa_log_class; / intent log data class */
	metaslab_class_t spa_embedded_log_class; / log on normal vdevs */
	metaslab_class_t spa_special_class; / special allocation class */
	metaslab_class_t spa_dedup_class; / dedup allocation class */
	uint64_t spa_first_txg; /* first txg after spa_open() */
	uint64_t spa_final_txg; /* txg of export/destroy */
	uint64_t spa_freeze_txg; /* freeze pool at this txg */
	uint64_t spa_load_max_txg; /* best initial ub_txg */
	uint64_t spa_claim_max_txg; /* highest claimed birth txg */
	inode_timespec_t spa_loaded_ts; /* 1st successful open time */
	objset_t spa_meta_objset; / copy of dp->dp_meta_objset */
	kmutex_t spa_evicting_os_lock; /* Evicting objset list lock */
	list_t spa_evicting_os_list; /* Objsets being evicted. */
	kcondvar_t spa_evicting_os_cv; /* Objset Eviction Completion */
	txg_list_t spa_vdev_txg_list; /* per-txg dirty vdev list */
	vdev_t spa_root_vdev; / top-level vdev container */
	uint64_t spa_min_ashift; /* of vdevs in normal class */
	uint64_t spa_max_ashift; /* of vdevs in normal class */
	uint64_t spa_min_alloc; /* of vdevs in normal class */
	uint64_t spa_gcd_alloc; /* of vdevs in normal class */
	uint64_t spa_config_guid; /* config pool guid */
	uint64_t spa_load_guid; /* spa_load initialized guid */
	uint64_t spa_last_synced_guid; /* last synced guid */
	list_t spa_config_dirty_list; /* vdevs with dirty config */
	list_t spa_state_dirty_list; /* vdevs with dirty state */
	/*
	* spa_allocs is an array, whose lengths is stored in spa_alloc_count.
	* There is one tree and one lock for each allocator, to help improve
	* allocation performance in write-heavy workloads.
	*/
	spa_alloc_t *spa_allocs;
	spa_allocs_use_t *spa_allocs_use;
	int spa_alloc_count;
	int spa_active_allocator; /* selectable allocator */

	/* per-allocator sync thread taskqs */
	taskq_t *spa_sync_tq;
	spa_syncthread_info_t *spa_syncthreads;

	spa_aux_vdev_t spa_spares; /* hot spares */
	spa_aux_vdev_t spa_l2cache; /* L2ARC cache devices */
	boolean_t spa_aux_sync_uber; /* need to sync aux uber */
	nvlist_t spa_label_features; / Features for reading MOS */
	uint64_t spa_config_object; /* MOS object for pool config */
	uint64_t spa_config_generation; /* config generation number */
	uint64_t spa_syncing_txg; /* txg currently syncing */
	bpobj_t spa_deferred_bpobj; /* deferred-free bplist */
	bplist_t spa_free_bplist[TXG_SIZE]; /* bplist of stuff to free */
	zio_cksum_salt_t spa_cksum_salt; /* secret salt for cksum */
	/* checksum context templates */
	kmutex_t spa_cksum_tmpls_lock;
	void *spa_cksum_tmpls[ZIO_CHECKSUM_FUNCTIONS];
	uberblock_t spa_ubsync; /* last synced uberblock */
	uberblock_t spa_uberblock; /* current uberblock */
	boolean_t spa_extreme_rewind; /* rewind past deferred frees */
	kmutex_t spa_scrub_lock; /* resilver/scrub lock */
	uint64_t spa_scrub_inflight; /* in-flight scrub bytes */

	/* in-flight verification bytes */
	uint64_t spa_load_verify_bytes;
	kcondvar_t spa_scrub_io_cv; /* scrub I/O completion */
	uint8_t spa_scrub_active; /* active or suspended? */
	uint8_t spa_scrub_type; /* type of scrub we're doing */
	uint8_t spa_scrub_finished; /* indicator to rotate logs */
	uint8_t spa_scrub_started; /* started since last boot */
	uint8_t spa_scrub_reopen; /* scrub doing vdev_reopen */
	uint64_t spa_scan_pass_start; /* start time per pass/reboot */
	uint64_t spa_scan_pass_scrub_pause; /* scrub pause time */
	uint64_t spa_scan_pass_scrub_spent_paused; /* total paused */
	uint64_t spa_scan_pass_exam; /* examined bytes per pass */
	uint64_t spa_scan_pass_issued; /* issued bytes per pass */

	/* error scrub pause time in milliseconds */
	uint64_t spa_scan_pass_errorscrub_pause;
	/* total error scrub paused time in milliseconds */
	uint64_t spa_scan_pass_errorscrub_spent_paused;
	/*
	* We are in the middle of a resilver, and another resilver
	* is needed once this one completes. This is set iff any
	* vdev_resilver_deferred is set.
	*/
	boolean_t spa_resilver_deferred;
	kmutex_t spa_async_lock; /* protect async state */
	kthread_t spa_async_thread; / thread doing async task */
	int spa_async_suspended; /* async tasks suspended */
	kcondvar_t spa_async_cv; /* wait for thread_exit() */
	uint16_t spa_async_tasks; /* async task mask */
	uint64_t spa_missing_tvds; /* unopenable tvds on load */
	uint64_t spa_missing_tvds_allowed; /* allow loading spa? */

	uint64_t spa_nonallocating_dspace;
	spa_removing_phys_t spa_removing_phys;
	spa_vdev_removal_t *spa_vdev_removal;

	spa_condensing_indirect_phys_t spa_condensing_indirect_phys;
	spa_condensing_indirect_t *spa_condensing_indirect;
	zthr_t spa_condense_zthr; / zthr doing condense. */

	vdev_raidz_expand_t *spa_raidz_expand;
	zthr_t *spa_raidz_expand_zthr;

	uint64_t spa_checkpoint_txg; /* the txg of the checkpoint */
	spa_checkpoint_info_t spa_checkpoint_info; /* checkpoint accounting */
	zthr_t *spa_checkpoint_discard_zthr;

	space_map_t spa_syncing_log_sm; / current log space map */
	avl_tree_t spa_sm_logs_by_txg;
	kmutex_t spa_flushed_ms_lock; /* for metaslabs_by_flushed */
	avl_tree_t spa_metaslabs_by_flushed;
	spa_unflushed_stats_t spa_unflushed_stats;
	list_t spa_log_summary;
	uint64_t spa_log_flushall_txg;

	zthr_t spa_livelist_delete_zthr; / deleting livelists */
	zthr_t spa_livelist_condense_zthr; / condensing livelists */
	uint64_t spa_livelists_to_delete; /* set of livelists to free */
	livelist_condense_entry_t spa_to_condense; /* next to condense */

	char spa_root; / alternate root directory */
	uint64_t spa_ena; /* spa-wide ereport ENA */
	int spa_last_open_failed; /* error if last open failed */
	uint64_t spa_last_ubsync_txg; /* "best" uberblock txg */
	uint64_t spa_last_ubsync_txg_ts; /* timestamp from that ub */
	uint64_t spa_load_txg; /* ub txg that loaded */
	uint64_t spa_load_txg_ts; /* timestamp from that ub */
	uint64_t spa_load_meta_errors; /* verify metadata err count */
	uint64_t spa_load_data_errors; /* verify data err count */
	uint64_t spa_verify_min_txg; /* start txg of verify scrub */
	kmutex_t spa_errlog_lock; /* error log lock */
	uint64_t spa_errlog_last; /* last error log object */
	uint64_t spa_errlog_scrub; /* scrub error log object */
	kmutex_t spa_errlist_lock; /* error list/ereport lock */
	avl_tree_t spa_errlist_last; /* last error list */
	avl_tree_t spa_errlist_scrub; /* scrub error list */
	avl_tree_t spa_errlist_healed; /* list of healed blocks */
	uint64_t spa_deflate; /* should we deflate? */
	uint64_t spa_history; /* history object */
	kmutex_t spa_history_lock; /* history lock */
	vdev_t spa_pending_vdev; / pending vdev additions */
	kmutex_t spa_props_lock; /* property lock */
	uint64_t spa_pool_props_object; /* object for properties */
	uint64_t spa_bootfs; /* default boot filesystem */
	uint64_t spa_failmode; /* failure mode for the pool */
	uint64_t spa_deadman_failmode; /* failure mode for deadman */
	uint64_t spa_delegation; /* delegation on/off */
	list_t spa_config_list; /* previous cache file(s) */
	/* per-CPU array of root of async I/O: */
	zio_t **spa_async_zio_root;
	zio_t spa_suspend_zio_root; / root of all suspended I/O */
	zio_t spa_txg_zio[TXG_SIZE]; / spa_sync() waits for this */
	kmutex_t spa_suspend_lock; /* protects suspend_zio_root */
	kcondvar_t spa_suspend_cv; /* notification of resume */
	zio_suspend_reason_t spa_suspended; /* pool is suspended */
	uint8_t spa_claiming; /* pool is doing zil_claim() */
	boolean_t spa_is_root; /* pool is root */
	int spa_minref; /* num refs when first opened */
	spa_mode_t spa_mode; /* SPA_MODE_{READ\|WRITE} */
	boolean_t spa_read_spacemaps; /* spacemaps available if ro */
	spa_log_state_t spa_log_state; /* log state */
	uint64_t spa_autoexpand; /* lun expansion on/off */
	ddt_t spa_ddt[ZIO_CHECKSUM_FUNCTIONS]; / in-core DDTs */
	uint64_t spa_ddt_stat_object; /* DDT statistics */
	uint64_t spa_dedup_dspace; /* Cache get_dedup_dspace() */
	uint64_t spa_dedup_checksum; /* default dedup checksum */
	uint64_t spa_dspace; /* dspace in normal class */
	struct brt spa_brt; / in-core BRT */
	kmutex_t spa_vdev_top_lock; /* dueling offline/remove */
	kmutex_t spa_proc_lock; /* protects spa_proc* */
	kcondvar_t spa_proc_cv; /* spa_proc_state transitions */
	spa_proc_state_t spa_proc_state; /* see definition */
	proc_t spa_proc; / "zpool-poolname" process */
	uintptr_t spa_did; /* if procp != p0, did of t1 */
	boolean_t spa_autoreplace; /* autoreplace set in open */
	int spa_vdev_locks; /* locks grabbed */
	uint64_t spa_creation_version; /* version at pool creation */
	uint64_t spa_prev_software_version; /* See ub_software_version */
	uint64_t spa_feat_for_write_obj; /* required to write to pool */
	uint64_t spa_feat_for_read_obj; /* required to read from pool */
	uint64_t spa_feat_desc_obj; /* Feature descriptions */
	uint64_t spa_feat_enabled_txg_obj; /* Feature enabled txg */
	kmutex_t spa_feat_stats_lock; /* protects spa_feat_stats */
	nvlist_t spa_feat_stats; / Cache of enabled features */
	/* cache feature refcounts */
	uint64_t spa_feat_refcount_cache[SPA_FEATURES];
	taskqid_t spa_deadman_tqid; /* Task id */
	uint64_t spa_deadman_calls; /* number of deadman calls */
	hrtime_t spa_sync_starttime; /* starting time of spa_sync */
	uint64_t spa_deadman_synctime; /* deadman sync expiration */
	uint64_t spa_deadman_ziotime; /* deadman zio expiration */
	uint64_t spa_all_vdev_zaps; /* ZAP of per-vd ZAP obj #s */
	spa_avz_action_t spa_avz_action; /* destroy/rebuild AVZ? */
	uint64_t spa_autotrim; /* automatic background trim? */
	uint64_t spa_errata; /* errata issues detected */
	spa_stats_t spa_stats; /* assorted spa statistics */
	spa_keystore_t spa_keystore; /* loaded crypto keys */

	/* arc_memory_throttle() parameters during low memory condition */
	uint64_t spa_lowmem_page_load; /* memory load during txg */
	uint64_t spa_lowmem_last_txg; /* txg window start */

	hrtime_t spa_ccw_fail_time; /* Conf cache write fail time */
	taskq_t spa_zvol_taskq; / Taskq for minor management */
	taskq_t spa_metaslab_taskq; / Taskq for metaslab preload */
	taskq_t spa_prefetch_taskq; / Taskq for prefetch threads */
	taskq_t spa_upgrade_taskq; / Taskq for upgrade jobs */
	uint64_t spa_multihost; /* multihost aware (mmp) */
	mmp_thread_t spa_mmp; /* multihost mmp thread */
	list_t spa_leaf_list; /* list of leaf vdevs */
	uint64_t spa_leaf_list_gen; /* track leaf_list changes */
	uint32_t spa_hostid; /* cached system hostid */

	/* synchronization for threads in spa_wait */
	kmutex_t spa_activities_lock;
	kcondvar_t spa_activities_cv;
	kcondvar_t spa_waiters_cv;
	int spa_waiters; /* number of waiting threads */
	boolean_t spa_waiters_cancel; /* waiters should return */

	char spa_compatibility; / compatibility file(s) */

	/*
	* spa_refcount & spa_config_lock must be the last elements
	* because zfs_refcount_t changes size based on compilation options.
	* In order for the MDB module to function correctly, the other
	* fields must remain in the same location.
	*/
	spa_config_lock_t spa_config_lock[SCL_LOCKS]; /* config changes */
	zfs_refcount_t spa_refcount; /* number of opens */
	};

	extern char *spa_config_path;
	extern const char *zfs_deadman_failmode;
	extern uint_t spa_slop_shift;
	-extern void spa_taskq_dispatch_ent(spa_t *spa, zio_type_t t, zio_taskq_type_t q,
	- task_func_t func, void arg, uint_t flags, taskq_ent_t ent, zio_t zio);
	-extern void spa_taskq_dispatch_sync(spa_t *, zio_type_t t, zio_taskq_type_t q,
	- task_func_t func, void arg, uint_t flags);
	+extern void spa_taskq_dispatch(spa_t *spa, zio_type_t t, zio_taskq_type_t q,
	+ task_func_t func, zio_t zio, boolean_t cutinline);
	extern void spa_load_spares(spa_t *spa);
	extern void spa_load_l2cache(spa_t *spa);
	extern sysevent_t spa_event_create(spa_t spa, vdev_t vd, nvlist_t hist_nvl,
	const char *name);
	extern void spa_event_post(sysevent_t *ev);
	extern int param_set_deadman_failmode_common(const char *val);
	extern void spa_set_deadman_synctime(hrtime_t ns);
	extern void spa_set_deadman_ziotime(hrtime_t ns);
	extern const char *spa_history_zone(void);
	extern const char *zfs_active_allocator;
	extern int param_set_active_allocator_common(const char *val);

	#ifdef __cplusplus
	}
	#endif

	#endif /* _SYS_SPA_IMPL_H */
	diff --git a/sys/contrib/openzfs/include/sys/zfs_context.h b/sys/contrib/openzfs/include/sys/zfs_context.h
	index 8f264b50e995..e4711ce4194a 100644
	--- a/sys/contrib/openzfs/include/sys/zfs_context.h
	+++ b/sys/contrib/openzfs/include/sys/zfs_context.h
	@@ -1,792 +1,791 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2012, 2018 by Delphix. All rights reserved.
	* Copyright (c) 2012, Joyent, Inc. All rights reserved.
	*/

	#ifndef _SYS_ZFS_CONTEXT_H
	#define _SYS_ZFS_CONTEXT_H

	#ifdef __cplusplus
	extern "C" {
	#endif

	/*
	* This code compiles in three different contexts. When __KERNEL__ is defined,
	* the code uses "unix-like" kernel interfaces. When _STANDALONE is defined, the
	* code is running in a reduced capacity environment of the boot loader which is
	* generally a subset of both POSIX and kernel interfaces (with a few unique
	* interfaces too). When neither are defined, it's in a userland POSIX or
	* similar environment.
	*/
	#if defined(__KERNEL__) \|\| defined(_STANDALONE)
	#include <sys/types.h>
	#include <sys/atomic.h>
	#include <sys/sysmacros.h>
	#include <sys/vmsystm.h>
	#include <sys/condvar.h>
	#include <sys/cmn_err.h>
	#include <sys/kmem.h>
	#include <sys/kmem_cache.h>
	#include <sys/vmem.h>
	#include <sys/misc.h>
	#include <sys/taskq.h>
	#include <sys/param.h>
	#include <sys/disp.h>
	#include <sys/debug.h>
	#include <sys/random.h>
	#include <sys/string.h>
	#include <sys/byteorder.h>
	#include <sys/list.h>
	#include <sys/time.h>
	#include <sys/zone.h>
	#include <sys/kstat.h>
	#include <sys/zfs_debug.h>
	#include <sys/sysevent.h>
	#include <sys/sysevent/eventdefs.h>
	#include <sys/zfs_delay.h>
	#include <sys/sunddi.h>
	#include <sys/ctype.h>
	#include <sys/disp.h>
	#include <sys/trace.h>
	#include <sys/procfs_list.h>
	#include <sys/mod.h>
	#include <sys/uio_impl.h>
	#include <sys/zfs_context_os.h>
	#else /* _KERNEL \|\| _STANDALONE */

	#define _SYS_MUTEX_H
	#define _SYS_RWLOCK_H
	#define _SYS_CONDVAR_H
	#define _SYS_VNODE_H
	#define _SYS_VFS_H
	#define _SYS_SUNDDI_H
	#define _SYS_CALLB_H

	#include <stdio.h>
	#include <stdlib.h>
	#include <stddef.h>
	#include <stdarg.h>
	#include <fcntl.h>
	#include <unistd.h>
	#include <errno.h>
	#include <string.h>
	#include <pthread.h>
	#include <setjmp.h>
	#include <assert.h>
	#include <umem.h>
	#include <limits.h>
	#include <atomic.h>
	#include <dirent.h>
	#include <time.h>
	#include <ctype.h>
	#include <signal.h>
	#include <sys/mman.h>
	#include <sys/types.h>
	#include <sys/cred.h>
	#include <sys/sysmacros.h>
	#include <sys/resource.h>
	#include <sys/byteorder.h>
	#include <sys/list.h>
	#include <sys/mod.h>
	#include <sys/uio.h>
	#include <sys/zfs_debug.h>
	#include <sys/kstat.h>
	#include <sys/u8_textprep.h>
	#include <sys/sysevent.h>
	#include <sys/sysevent/eventdefs.h>
	#include <sys/sunddi.h>
	#include <sys/debug.h>
	#include <sys/utsname.h>
	#include <sys/trace_zfs.h>

	#include <sys/zfs_context_os.h>

	/*
	* Stack
	*/

	#define noinline __attribute__((noinline))
	#define likely(x) __builtin_expect((x), 1)
	#define unlikely(x) __builtin_expect((x), 0)

	/*
	* Debugging
	*/

	/*
	* Note that we are not using the debugging levels.
	*/

	#define CE_CONT 0 /* continuation */
	#define CE_NOTE 1 /* notice */
	#define CE_WARN 2 /* warning */
	#define CE_PANIC 3 /* panic */
	#define CE_IGNORE 4 /* print nothing */

	/*
	* ZFS debugging
	*/

	extern void dprintf_setup(int argc, char *argv);

	extern void cmn_err(int, const char *, ...)
	__attribute__((format(printf, 2, 3)));
	extern void vcmn_err(int, const char *, va_list)
	__attribute__((format(printf, 2, 0)));
	extern void panic(const char *, ...)
	__attribute__((format(printf, 1, 2), noreturn));
	extern void vpanic(const char *, va_list)
	__attribute__((format(printf, 1, 0), noreturn));

	#define fm_panic panic

	/*
	* DTrace SDT probes have different signatures in userland than they do in
	* the kernel. If they're being used in kernel code, re-define them out of
	* existence for their counterparts in libzpool.
	*
	* Here's an example of how to use the set-error probes in userland:
	* zfs$target:::set-error /arg0 == EBUSY/ {stack();}
	*
	* Here's an example of how to use DTRACE_PROBE probes in userland:
	* If there is a probe declared as follows:
	* DTRACE_PROBE2(zfs__probe_name, uint64_t, blkid, dnode_t *, dn);
	* Then you can use it as follows:
	* zfs$target:::probe2 /copyinstr(arg0) == "zfs__probe_name"/
	* {printf("%u %p\n", arg1, arg2);}
	*/

	#ifdef DTRACE_PROBE
	#undef DTRACE_PROBE
	#endif /* DTRACE_PROBE */
	#define DTRACE_PROBE(a)

	#ifdef DTRACE_PROBE1
	#undef DTRACE_PROBE1
	#endif /* DTRACE_PROBE1 */
	#define DTRACE_PROBE1(a, b, c)

	#ifdef DTRACE_PROBE2
	#undef DTRACE_PROBE2
	#endif /* DTRACE_PROBE2 */
	#define DTRACE_PROBE2(a, b, c, d, e)

	#ifdef DTRACE_PROBE3
	#undef DTRACE_PROBE3
	#endif /* DTRACE_PROBE3 */
	#define DTRACE_PROBE3(a, b, c, d, e, f, g)

	#ifdef DTRACE_PROBE4
	#undef DTRACE_PROBE4
	#endif /* DTRACE_PROBE4 */
	#define DTRACE_PROBE4(a, b, c, d, e, f, g, h, i)

	/*
	* Tunables.
	*/
	typedef struct zfs_kernel_param {
	const char name; / unused stub */
	} zfs_kernel_param_t;

	#define ZFS_MODULE_PARAM(scope_prefix, name_prefix, name, type, perm, desc)
	#define ZFS_MODULE_PARAM_ARGS void
	#define ZFS_MODULE_PARAM_CALL(scope_prefix, name_prefix, name, setfunc, \
	getfunc, perm, desc)

	/*
	* Threads.
	*/
	typedef pthread_t kthread_t;

	#define TS_RUN 0x00000002
	#define TS_JOINABLE 0x00000004

	#define curthread ((void *)(uintptr_t)pthread_self())
	#define getcomm() "unknown"

	#define thread_create_named(name, stk, stksize, func, arg, len, \
	pp, state, pri) \
	zk_thread_create(name, func, arg, stksize, state)
	#define thread_create(stk, stksize, func, arg, len, pp, state, pri) \
	zk_thread_create(#func, func, arg, stksize, state)
	#define thread_exit() pthread_exit(NULL)
	#define thread_join(t) pthread_join((pthread_t)(t), NULL)

	#define newproc(f, a, cid, pri, ctp, pid) (ENOSYS)

	/* in libzpool, p0 exists only to have its address taken */
	typedef struct proc {
	uintptr_t this_is_never_used_dont_dereference_it;
	} proc_t;

	extern struct proc p0;
	#define curproc (&p0)

	#define PS_NONE -1

	extern kthread_t zk_thread_create(const char name, void (func)(void ),
	void *arg, size_t stksize, int state);

	-#define issig(why) (FALSE)
	-#define ISSIG(thr, why) (FALSE)
	+#define issig() (FALSE)

	#define KPREEMPT_SYNC (-1)

	#define kpreempt(x) sched_yield()
	#define kpreempt_disable() ((void)0)
	#define kpreempt_enable() ((void)0)

	/*
	* Mutexes
	*/
	typedef struct kmutex {
	pthread_mutex_t m_lock;
	pthread_t m_owner;
	} kmutex_t;

	#define MUTEX_DEFAULT 0
	#define MUTEX_NOLOCKDEP MUTEX_DEFAULT
	#define MUTEX_HELD(mp) pthread_equal((mp)->m_owner, pthread_self())
	#define MUTEX_NOT_HELD(mp) !MUTEX_HELD(mp)

	extern void mutex_init(kmutex_t mp, char name, int type, void *cookie);
	extern void mutex_destroy(kmutex_t *mp);
	extern void mutex_enter(kmutex_t *mp);
	extern int mutex_enter_check_return(kmutex_t *mp);
	extern void mutex_exit(kmutex_t *mp);
	extern int mutex_tryenter(kmutex_t *mp);

	#define NESTED_SINGLE 1
	#define mutex_enter_nested(mp, class) mutex_enter(mp)
	#define mutex_enter_interruptible(mp) mutex_enter_check_return(mp)
	/*
	* RW locks
	*/
	typedef struct krwlock {
	pthread_rwlock_t rw_lock;
	pthread_t rw_owner;
	uint_t rw_readers;
	} krwlock_t;

	typedef int krw_t;

	#define RW_READER 0
	#define RW_WRITER 1
	#define RW_DEFAULT RW_READER
	#define RW_NOLOCKDEP RW_READER

	#define RW_READ_HELD(rw) ((rw)->rw_readers > 0)
	#define RW_WRITE_HELD(rw) pthread_equal((rw)->rw_owner, pthread_self())
	#define RW_LOCK_HELD(rw) (RW_READ_HELD(rw) \|\| RW_WRITE_HELD(rw))

	extern void rw_init(krwlock_t rwlp, char name, int type, void *arg);
	extern void rw_destroy(krwlock_t *rwlp);
	extern void rw_enter(krwlock_t *rwlp, krw_t rw);
	extern int rw_tryenter(krwlock_t *rwlp, krw_t rw);
	extern int rw_tryupgrade(krwlock_t *rwlp);
	extern void rw_exit(krwlock_t *rwlp);
	#define rw_downgrade(rwlp) do { } while (0)

	/*
	* Credentials
	*/
	extern uid_t crgetuid(cred_t *cr);
	extern uid_t crgetruid(cred_t *cr);
	extern gid_t crgetgid(cred_t *cr);
	extern int crgetngroups(cred_t *cr);
	extern gid_t crgetgroups(cred_t cr);

	/*
	* Condition variables
	*/
	typedef pthread_cond_t kcondvar_t;

	#define CV_DEFAULT 0
	#define CALLOUT_FLAG_ABSOLUTE 0x2

	extern void cv_init(kcondvar_t cv, char name, int type, void *arg);
	extern void cv_destroy(kcondvar_t *cv);
	extern void cv_wait(kcondvar_t cv, kmutex_t mp);
	extern int cv_wait_sig(kcondvar_t cv, kmutex_t mp);
	extern int cv_timedwait(kcondvar_t cv, kmutex_t mp, clock_t abstime);
	extern int cv_timedwait_hires(kcondvar_t cvp, kmutex_t mp, hrtime_t tim,
	hrtime_t res, int flag);
	extern void cv_signal(kcondvar_t *cv);
	extern void cv_broadcast(kcondvar_t *cv);

	#define cv_timedwait_io(cv, mp, at) cv_timedwait(cv, mp, at)
	#define cv_timedwait_idle(cv, mp, at) cv_timedwait(cv, mp, at)
	#define cv_timedwait_sig(cv, mp, at) cv_timedwait(cv, mp, at)
	#define cv_wait_io(cv, mp) cv_wait(cv, mp)
	#define cv_wait_idle(cv, mp) cv_wait(cv, mp)
	#define cv_wait_io_sig(cv, mp) cv_wait_sig(cv, mp)
	#define cv_timedwait_sig_hires(cv, mp, t, r, f) \
	cv_timedwait_hires(cv, mp, t, r, f)
	#define cv_timedwait_idle_hires(cv, mp, t, r, f) \
	cv_timedwait_hires(cv, mp, t, r, f)

	/*
	* Thread-specific data
	*/
	#define tsd_get(k) pthread_getspecific(k)
	#define tsd_set(k, v) pthread_setspecific(k, v)
	#define tsd_create(kp, d) pthread_key_create((pthread_key_t *)kp, d)
	#define tsd_destroy(kp) /* nothing */
	#ifdef __FreeBSD__
	typedef off_t loff_t;
	#endif

	/*
	* kstat creation, installation and deletion
	*/
	extern kstat_t kstat_create(const char , int,
	const char , const char , uchar_t, ulong_t, uchar_t);
	extern void kstat_install(kstat_t *);
	extern void kstat_delete(kstat_t *);
	extern void kstat_set_raw_ops(kstat_t *ksp,
	int (headers)(char buf, size_t size),
	int (data)(char buf, size_t size, void *data),
	void (addr)(kstat_t *ksp, loff_t index));

	/*
	* procfs list manipulation
	*/

	typedef struct procfs_list {
	void *pl_private;
	kmutex_t pl_lock;
	list_t pl_list;
	uint64_t pl_next_id;
	size_t pl_node_offset;
	} procfs_list_t;

	#ifndef __cplusplus
	struct seq_file { };
	void seq_printf(struct seq_file m, const char fmt, ...);

	typedef struct procfs_list_node {
	list_node_t pln_link;
	uint64_t pln_id;
	} procfs_list_node_t;

	void procfs_list_install(const char *module,
	const char *submodule,
	const char *name,
	mode_t mode,
	procfs_list_t *procfs_list,
	int (show)(struct seq_file f, void *p),
	int (show_header)(struct seq_file f),
	int (clear)(procfs_list_t procfs_list),
	size_t procfs_list_node_off);
	void procfs_list_uninstall(procfs_list_t *procfs_list);
	void procfs_list_destroy(procfs_list_t *procfs_list);
	void procfs_list_add(procfs_list_t procfs_list, void p);
	#endif

	/*
	* Kernel memory
	*/
	#define KM_SLEEP UMEM_NOFAIL
	#define KM_PUSHPAGE KM_SLEEP
	#define KM_NOSLEEP UMEM_DEFAULT
	#define KM_NORMALPRI 0 /* not needed with UMEM_DEFAULT */
	#define KMC_NODEBUG UMC_NODEBUG
	#define KMC_KVMEM 0x0
	#define kmem_alloc(_s, _f) umem_alloc(_s, _f)
	#define kmem_zalloc(_s, _f) umem_zalloc(_s, _f)
	#define kmem_free(_b, _s) umem_free(_b, _s)
	#define vmem_alloc(_s, _f) kmem_alloc(_s, _f)
	#define vmem_zalloc(_s, _f) kmem_zalloc(_s, _f)
	#define vmem_free(_b, _s) kmem_free(_b, _s)
	#define kmem_cache_create(_a, _b, _c, _d, _e, _f, _g, _h, _i) \
	umem_cache_create(_a, _b, _c, _d, _e, _f, _g, _h, _i)
	#define kmem_cache_destroy(_c) umem_cache_destroy(_c)
	#define kmem_cache_alloc(_c, _f) umem_cache_alloc(_c, _f)
	#define kmem_cache_free(_c, _b) umem_cache_free(_c, _b)
	#define kmem_debugging() 0
	#define kmem_cache_reap_now(_c) umem_cache_reap_now(_c);
	#define kmem_cache_set_move(_c, _cb) /* nothing */
	#define POINTER_INVALIDATE(_pp) /* nothing */
	#define POINTER_IS_VALID(_p) 0

	typedef umem_cache_t kmem_cache_t;

	typedef enum kmem_cbrc {
	KMEM_CBRC_YES,
	KMEM_CBRC_NO,
	KMEM_CBRC_LATER,
	KMEM_CBRC_DONT_NEED,
	KMEM_CBRC_DONT_KNOW
	} kmem_cbrc_t;

	/*
	* Task queues
	*/

	#define TASKQ_NAMELEN 31

	typedef uintptr_t taskqid_t;
	typedef void (task_func_t)(void *);

	typedef struct taskq_ent {
	struct taskq_ent *tqent_next;
	struct taskq_ent *tqent_prev;
	task_func_t *tqent_func;
	void *tqent_arg;
	uintptr_t tqent_flags;
	} taskq_ent_t;

	typedef struct taskq {
	char tq_name[TASKQ_NAMELEN + 1];
	kmutex_t tq_lock;
	krwlock_t tq_threadlock;
	kcondvar_t tq_dispatch_cv;
	kcondvar_t tq_wait_cv;
	kthread_t **tq_threadlist;
	int tq_flags;
	int tq_active;
	int tq_nthreads;
	int tq_nalloc;
	int tq_minalloc;
	int tq_maxalloc;
	kcondvar_t tq_maxalloc_cv;
	int tq_maxalloc_wait;
	taskq_ent_t *tq_freelist;
	taskq_ent_t tq_task;
	} taskq_t;

	#define TQENT_FLAG_PREALLOC 0x1 /* taskq_dispatch_ent used */

	#define TASKQ_PREPOPULATE 0x0001
	#define TASKQ_CPR_SAFE 0x0002 /* Use CPR safe protocol */
	#define TASKQ_DYNAMIC 0x0004 /* Use dynamic thread scheduling */
	#define TASKQ_THREADS_CPU_PCT 0x0008 /* Scale # threads by # cpus */
	#define TASKQ_DC_BATCH 0x0010 /* Mark threads as batch */

	#define TQ_SLEEP KM_SLEEP /* Can block for memory */
	#define TQ_NOSLEEP KM_NOSLEEP /* cannot block for memory; may fail */
	#define TQ_NOQUEUE 0x02 /* Do not enqueue if can't dispatch */
	#define TQ_FRONT 0x08 /* Queue in front */

	#define TASKQID_INVALID ((taskqid_t)0)

	extern taskq_t *system_taskq;
	extern taskq_t *system_delay_taskq;

	extern taskq_t taskq_create(const char , int, pri_t, int, int, uint_t);
	extern taskq_t taskq_create_synced(const char , int, pri_t, int, int, uint_t,
	kthread_t ***);
	#define taskq_create_proc(a, b, c, d, e, p, f) \
	(taskq_create(a, b, c, d, e, f))
	#define taskq_create_sysdc(a, b, d, e, p, dc, f) \
	((void) sizeof (dc), taskq_create(a, b, maxclsyspri, d, e, f))
	extern taskqid_t taskq_dispatch(taskq_t , task_func_t, void , uint_t);
	extern taskqid_t taskq_dispatch_delay(taskq_t , task_func_t, void , uint_t,
	clock_t);
	extern void taskq_dispatch_ent(taskq_t , task_func_t, void , uint_t,
	taskq_ent_t *);
	extern int taskq_empty_ent(taskq_ent_t *);
	extern void taskq_init_ent(taskq_ent_t *);
	extern void taskq_destroy(taskq_t *);
	extern void taskq_wait(taskq_t *);
	extern void taskq_wait_id(taskq_t *, taskqid_t);
	extern void taskq_wait_outstanding(taskq_t *, taskqid_t);
	extern int taskq_member(taskq_t , kthread_t );
	extern taskq_t *taskq_of_curthread(void);
	extern int taskq_cancel_id(taskq_t *, taskqid_t);
	extern void system_taskq_init(void);
	extern void system_taskq_fini(void);

	#define XVA_MAPSIZE 3
	#define XVA_MAGIC 0x78766174

	extern char *vn_dumpdir;
	#define AV_SCANSTAMP_SZ 32 /* length of anti-virus scanstamp */

	typedef struct xoptattr {
	inode_timespec_t xoa_createtime; /* Create time of file */
	uint8_t xoa_archive;
	uint8_t xoa_system;
	uint8_t xoa_readonly;
	uint8_t xoa_hidden;
	uint8_t xoa_nounlink;
	uint8_t xoa_immutable;
	uint8_t xoa_appendonly;
	uint8_t xoa_nodump;
	uint8_t xoa_settable;
	uint8_t xoa_opaque;
	uint8_t xoa_av_quarantined;
	uint8_t xoa_av_modified;
	uint8_t xoa_av_scanstamp[AV_SCANSTAMP_SZ];
	uint8_t xoa_reparse;
	uint8_t xoa_offline;
	uint8_t xoa_sparse;
	} xoptattr_t;

	typedef struct vattr {
	uint_t va_mask; /* bit-mask of attributes */
	u_offset_t va_size; /* file size in bytes */
	} vattr_t;


	typedef struct xvattr {
	vattr_t xva_vattr; /* Embedded vattr structure */
	uint32_t xva_magic; /* Magic Number */
	uint32_t xva_mapsize; /* Size of attr bitmap (32-bit words) */
	uint32_t xva_rtnattrmapp; / Ptr to xva_rtnattrmap[] */
	uint32_t xva_reqattrmap[XVA_MAPSIZE]; /* Requested attrs */
	uint32_t xva_rtnattrmap[XVA_MAPSIZE]; /* Returned attrs */
	xoptattr_t xva_xoptattrs; /* Optional attributes */
	} xvattr_t;

	typedef struct vsecattr {
	uint_t vsa_mask; /* See below */
	int vsa_aclcnt; /* ACL entry count */
	void vsa_aclentp; / pointer to ACL entries */
	int vsa_dfaclcnt; /* default ACL entry count */
	void vsa_dfaclentp; / pointer to default ACL entries */
	size_t vsa_aclentsz; /* ACE size in bytes of vsa_aclentp */
	} vsecattr_t;

	#define AT_MODE 0x00002
	#define AT_UID 0x00004
	#define AT_GID 0x00008
	#define AT_FSID 0x00010
	#define AT_NODEID 0x00020
	#define AT_NLINK 0x00040
	#define AT_SIZE 0x00080
	#define AT_ATIME 0x00100
	#define AT_MTIME 0x00200
	#define AT_CTIME 0x00400
	#define AT_RDEV 0x00800
	#define AT_BLKSIZE 0x01000
	#define AT_NBLOCKS 0x02000
	#define AT_SEQ 0x08000
	#define AT_XVATTR 0x10000

	#define CRCREAT 0

	#define F_FREESP 11
	#define FIGNORECASE 0x80000 /* request case-insensitive lookups */

	/*
	* Random stuff
	*/
	#define ddi_get_lbolt() (gethrtime() >> 23)
	#define ddi_get_lbolt64() (gethrtime() >> 23)
	#define hz 119 /* frequency when using gethrtime() >> 23 for lbolt */

	#define ddi_time_before(a, b) (a < b)
	#define ddi_time_after(a, b) ddi_time_before(b, a)
	#define ddi_time_before_eq(a, b) (!ddi_time_after(a, b))
	#define ddi_time_after_eq(a, b) ddi_time_before_eq(b, a)

	#define ddi_time_before64(a, b) (a < b)
	#define ddi_time_after64(a, b) ddi_time_before64(b, a)
	#define ddi_time_before_eq64(a, b) (!ddi_time_after64(a, b))
	#define ddi_time_after_eq64(a, b) ddi_time_before_eq64(b, a)

	extern void delay(clock_t ticks);

	#define SEC_TO_TICK(sec) ((sec) * hz)
	#define MSEC_TO_TICK(msec) (howmany((hrtime_t)(msec) * hz, MILLISEC))
	#define USEC_TO_TICK(usec) (howmany((hrtime_t)(usec) * hz, MICROSEC))
	#define NSEC_TO_TICK(nsec) (howmany((hrtime_t)(nsec) * hz, NANOSEC))

	#define max_ncpus 64
	#define boot_ncpus (sysconf(_SC_NPROCESSORS_ONLN))

	/*
	* Process priorities as defined by setpriority(2) and getpriority(2).
	*/
	#define minclsyspri 19
	#define maxclsyspri -20
	#define defclsyspri 0

	#define CPU_SEQID ((uintptr_t)pthread_self() & (max_ncpus - 1))
	#define CPU_SEQID_UNSTABLE CPU_SEQID

	#define kcred NULL
	#define CRED() NULL

	#define ptob(x) ((x) * PAGESIZE)

	#define NN_DIVISOR_1000 (1U << 0)
	#define NN_NUMBUF_SZ (6)

	extern uint64_t physmem;
	extern const char *random_path;
	extern const char *urandom_path;

	extern int highbit64(uint64_t i);
	extern int lowbit64(uint64_t i);
	extern int random_get_bytes(uint8_t *ptr, size_t len);
	extern int random_get_pseudo_bytes(uint8_t *ptr, size_t len);

	static __inline__ uint32_t
	random_in_range(uint32_t range)
	{
	uint32_t r;

	ASSERT(range != 0);

	if (range == 1)
	return (0);

	(void) random_get_pseudo_bytes((uint8_t *)&r, sizeof (r));

	return (r % range);
	}

	extern void kernel_init(int mode);
	extern void kernel_fini(void);
	extern void random_init(void);
	extern void random_fini(void);

	struct spa;
	extern void show_pool_stats(struct spa *);
	extern int set_global_var(char const *arg);

	typedef struct callb_cpr {
	kmutex_t *cc_lockp;
	} callb_cpr_t;

	#define CALLB_CPR_INIT(cp, lockp, func, name) { \
	(cp)->cc_lockp = lockp; \
	}

	#define CALLB_CPR_SAFE_BEGIN(cp) { \
	ASSERT(MUTEX_HELD((cp)->cc_lockp)); \
	}

	#define CALLB_CPR_SAFE_END(cp, lockp) { \
	ASSERT(MUTEX_HELD((cp)->cc_lockp)); \
	}

	#define CALLB_CPR_EXIT(cp) { \
	ASSERT(MUTEX_HELD((cp)->cc_lockp)); \
	mutex_exit((cp)->cc_lockp); \
	}

	#define zone_dataset_visible(x, y) (1)
	#define INGLOBALZONE(z) (1)
	extern uint32_t zone_get_hostid(void *zonep);

	extern char kmem_vasprintf(const char fmt, va_list adx);
	extern char kmem_asprintf(const char fmt, ...);
	#define kmem_strfree(str) kmem_free((str), strlen(str) + 1)
	#define kmem_strdup(s) strdup(s)

	#ifndef __cplusplus
	extern int kmem_scnprintf(char *restrict str, size_t size,
	const char *restrict fmt, ...);
	#endif

	/*
	* Hostname information
	*/
	extern int ddi_strtoull(const char str, char *nptr, int base,
	u_longlong_t *result);

	typedef struct utsname utsname_t;
	extern utsname_t *utsname(void);

	/* ZFS Boot Related stuff. */

	struct _buf {
	intptr_t _fd;
	};

	struct bootstat {
	uint64_t st_size;
	};

	typedef struct ace_object {
	uid_t a_who;
	uint32_t a_access_mask;
	uint16_t a_flags;
	uint16_t a_type;
	uint8_t a_obj_type[16];
	uint8_t a_inherit_obj_type[16];
	} ace_object_t;


	#define ACE_ACCESS_ALLOWED_OBJECT_ACE_TYPE 0x05
	#define ACE_ACCESS_DENIED_OBJECT_ACE_TYPE 0x06
	#define ACE_SYSTEM_AUDIT_OBJECT_ACE_TYPE 0x07
	#define ACE_SYSTEM_ALARM_OBJECT_ACE_TYPE 0x08

	extern int zfs_secpolicy_snapshot_perms(const char name, cred_t cr);
	extern int zfs_secpolicy_rename_perms(const char from, const char to,
	cred_t *cr);
	extern int zfs_secpolicy_destroy_perms(const char name, cred_t cr);
	extern int secpolicy_zfs(const cred_t *cr);
	extern int secpolicy_zfs_proc(const cred_t cr, proc_t proc);
	extern zoneid_t getzoneid(void);

	/* SID stuff */
	typedef struct ksiddomain {
	uint_t kd_ref;
	uint_t kd_len;
	char *kd_name;
	} ksiddomain_t;

	ksiddomain_t ksid_lookupdomain(const char );
	void ksiddomain_rele(ksiddomain_t *);

	#define DDI_SLEEP KM_SLEEP
	#define ddi_log_sysevent(_a, _b, _c, _d, _e, _f, _g) \
	sysevent_post_event(_c, _d, _b, "libzpool", _e, _f)

	#define zfs_sleep_until(wakeup) \
	do { \
	hrtime_t delta = wakeup - gethrtime(); \
	struct timespec ts; \
	ts.tv_sec = delta / NANOSEC; \
	ts.tv_nsec = delta % NANOSEC; \
	(void) nanosleep(&ts, NULL); \
	} while (0)

	typedef int fstrans_cookie_t;

	extern fstrans_cookie_t spl_fstrans_mark(void);
	extern void spl_fstrans_unmark(fstrans_cookie_t);
	extern int __spl_pf_fstrans_check(void);
	extern int kmem_cache_reap_active(void);


	/*
	* Kernel modules
	*/
	#define __init
	#define __exit

	#endif /* _KERNEL \|\| _STANDALONE */

	#ifdef __cplusplus
	};
	#endif

	#endif /* _SYS_ZFS_CONTEXT_H */
	diff --git a/sys/contrib/openzfs/include/sys/zfs_debug.h b/sys/contrib/openzfs/include/sys/zfs_debug.h
	index 8d94557a5882..e509c8b7c638 100644
	--- a/sys/contrib/openzfs/include/sys/zfs_debug.h
	+++ b/sys/contrib/openzfs/include/sys/zfs_debug.h
	@@ -1,114 +1,114 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2019 by Delphix. All rights reserved.
	*/

	#ifndef _SYS_ZFS_DEBUG_H
	#define _SYS_ZFS_DEBUG_H

	#ifdef __cplusplus
	extern "C" {
	#endif

	#ifndef TRUE
	#define TRUE 1
	#endif

	#ifndef FALSE
	#define FALSE 0
	#endif

	extern int zfs_flags;
	extern int zfs_recover;
	extern int zfs_free_leak_on_eio;
	extern int zfs_dbgmsg_enable;

	#define ZFS_DEBUG_DPRINTF (1 << 0)
	#define ZFS_DEBUG_DBUF_VERIFY (1 << 1)
	#define ZFS_DEBUG_DNODE_VERIFY (1 << 2)
	#define ZFS_DEBUG_SNAPNAMES (1 << 3)
	#define ZFS_DEBUG_MODIFY (1 << 4)
	/* 1<<5 was previously used, try not to reuse */
	#define ZFS_DEBUG_ZIO_FREE (1 << 6)
	#define ZFS_DEBUG_HISTOGRAM_VERIFY (1 << 7)
	#define ZFS_DEBUG_METASLAB_VERIFY (1 << 8)
	#define ZFS_DEBUG_SET_ERROR (1 << 9)
	#define ZFS_DEBUG_INDIRECT_REMAP (1 << 10)
	#define ZFS_DEBUG_TRIM (1 << 11)
	#define ZFS_DEBUG_LOG_SPACEMAP (1 << 12)
	#define ZFS_DEBUG_METASLAB_ALLOC (1 << 13)
	#define ZFS_DEBUG_BRT (1 << 14)
	#define ZFS_DEBUG_RAIDZ_RECONSTRUCT (1 << 15)

	extern void __set_error(const char file, const char func, int line, int err);
	extern void __zfs_dbgmsg(char *buf);
	extern void __dprintf(boolean_t dprint, const char file, const char func,
	int line, const char *fmt, ...) __attribute__((format(printf, 5, 6)));

	/*
	* Some general principles for using zfs_dbgmsg():
	* 1. We don't want to pollute the log with typically-irrelevant messages,
	* so don't print too many messages in the "normal" code path - O(1)
	* per txg.
	* 2. We want to know for sure what happened, so make the message specific
	* (e.g. which thing am I operating on).
	* 3. Do print a message when something unusual or unexpected happens
	* (e.g. error cases).
	* 4. Print a message when making user-initiated on-disk changes.
	*
	* Note that besides principle 1, another reason that we don't want to
	* use zfs_dbgmsg in high-frequency routines is the potential impact
	* that it can have on performance.
	*/
	#define zfs_dbgmsg(...) \
	if (zfs_dbgmsg_enable) \
	__dprintf(B_FALSE, __FILE__, __func__, __LINE__, __VA_ARGS__)

	#ifdef ZFS_DEBUG
	/*
	* To enable this:
	*
	* $ echo 1 >/sys/module/zfs/parameters/zfs_flags
	*/
	#define dprintf(...) \
	if (zfs_flags & ZFS_DEBUG_DPRINTF) \
	__dprintf(B_TRUE, __FILE__, __func__, __LINE__, __VA_ARGS__)
	#else
	#define dprintf(...) ((void)0)
	#endif /* ZFS_DEBUG */

	extern void zfs_panic_recover(const char *fmt, ...);

	extern void zfs_dbgmsg_init(void);
	extern void zfs_dbgmsg_fini(void);

	#ifndef _KERNEL
	extern int dprintf_find_string(const char *string);
	-extern void zfs_dbgmsg_print(const char *tag);
	+extern void zfs_dbgmsg_print(int fd, const char *tag);
	#endif

	#ifdef __cplusplus
	}
	#endif

	#endif /* _SYS_ZFS_DEBUG_H */
	diff --git a/sys/contrib/openzfs/lib/libefi/rdwr_efi.c b/sys/contrib/openzfs/lib/libefi/rdwr_efi.c
	index 739219e0410f..63c91059ae5f 100644
	--- a/sys/contrib/openzfs/lib/libefi/rdwr_efi.c
	+++ b/sys/contrib/openzfs/lib/libefi/rdwr_efi.c
	@@ -1,1627 +1,1627 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright 2012 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2018 by Delphix. All rights reserved.
	*/

	#include <stdio.h>
	#include <stdlib.h>
	#include <errno.h>
	#include <string.h>
	#include <unistd.h>
	#include <uuid/uuid.h>
	#include <zlib.h>
	#include <libintl.h>
	#include <sys/types.h>
	#include <sys/dkio.h>
	#include <sys/mhd.h>
	#include <sys/param.h>
	#include <sys/dktp/fdisk.h>
	#include <sys/efi_partition.h>
	#include <sys/byteorder.h>
	#include <sys/vdev_disk.h>
	#include <linux/fs.h>
	#include <linux/blkpg.h>

	static struct uuid_to_ptag {
	struct uuid uuid;
	} conversion_array[] = {
	{ EFI_UNUSED },
	{ EFI_BOOT },
	{ EFI_ROOT },
	{ EFI_SWAP },
	{ EFI_USR },
	{ EFI_BACKUP },
	{ EFI_UNUSED }, /* STAND is never used */
	{ EFI_VAR },
	{ EFI_HOME },
	{ EFI_ALTSCTR },
	{ EFI_UNUSED }, /* CACHE (cachefs) is never used */
	{ EFI_RESERVED },
	{ EFI_SYSTEM },
	{ EFI_LEGACY_MBR },
	{ EFI_SYMC_PUB },
	{ EFI_SYMC_CDS },
	{ EFI_MSFT_RESV },
	{ EFI_DELL_BASIC },
	{ EFI_DELL_RAID },
	{ EFI_DELL_SWAP },
	{ EFI_DELL_LVM },
	{ EFI_DELL_RESV },
	{ EFI_AAPL_HFS },
	{ EFI_AAPL_UFS },
	{ EFI_FREEBSD_BOOT },
	{ EFI_FREEBSD_SWAP },
	{ EFI_FREEBSD_UFS },
	{ EFI_FREEBSD_VINUM },
	{ EFI_FREEBSD_ZFS },
	{ EFI_BIOS_BOOT },
	{ EFI_INTC_RS },
	{ EFI_SNE_BOOT },
	{ EFI_LENOVO_BOOT },
	{ EFI_MSFT_LDMM },
	{ EFI_MSFT_LDMD },
	{ EFI_MSFT_RE },
	{ EFI_IBM_GPFS },
	{ EFI_MSFT_STORAGESPACES },
	{ EFI_HPQ_DATA },
	{ EFI_HPQ_SVC },
	{ EFI_RHT_DATA },
	{ EFI_RHT_HOME },
	{ EFI_RHT_SRV },
	{ EFI_RHT_DMCRYPT },
	{ EFI_RHT_LUKS },
	{ EFI_FREEBSD_DISKLABEL },
	{ EFI_AAPL_RAID },
	{ EFI_AAPL_RAIDOFFLINE },
	{ EFI_AAPL_BOOT },
	{ EFI_AAPL_LABEL },
	{ EFI_AAPL_TVRECOVERY },
	{ EFI_AAPL_CORESTORAGE },
	{ EFI_NETBSD_SWAP },
	{ EFI_NETBSD_FFS },
	{ EFI_NETBSD_LFS },
	{ EFI_NETBSD_RAID },
	{ EFI_NETBSD_CAT },
	{ EFI_NETBSD_CRYPT },
	{ EFI_GOOG_KERN },
	{ EFI_GOOG_ROOT },
	{ EFI_GOOG_RESV },
	{ EFI_HAIKU_BFS },
	{ EFI_MIDNIGHTBSD_BOOT },
	{ EFI_MIDNIGHTBSD_DATA },
	{ EFI_MIDNIGHTBSD_SWAP },
	{ EFI_MIDNIGHTBSD_UFS },
	{ EFI_MIDNIGHTBSD_VINUM },
	{ EFI_MIDNIGHTBSD_ZFS },
	{ EFI_CEPH_JOURNAL },
	{ EFI_CEPH_DMCRYPTJOURNAL },
	{ EFI_CEPH_OSD },
	{ EFI_CEPH_DMCRYPTOSD },
	{ EFI_CEPH_CREATE },
	{ EFI_CEPH_DMCRYPTCREATE },
	{ EFI_OPENBSD_DISKLABEL },
	{ EFI_BBRY_QNX },
	{ EFI_BELL_PLAN9 },
	{ EFI_VMW_KCORE },
	{ EFI_VMW_VMFS },
	{ EFI_VMW_RESV },
	{ EFI_RHT_ROOTX86 },
	{ EFI_RHT_ROOTAMD64 },
	{ EFI_RHT_ROOTARM },
	{ EFI_RHT_ROOTARM64 },
	{ EFI_ACRONIS_SECUREZONE },
	{ EFI_ONIE_BOOT },
	{ EFI_ONIE_CONFIG },
	{ EFI_IBM_PPRPBOOT },
	{ EFI_FREEDESKTOP_BOOT }
	};

	int efi_debug = 0;

	static int efi_read(int, struct dk_gpt *);

	/*
	* Return a 32-bit CRC of the contents of the buffer. Pre-and-post
	* one's conditioning will be handled by crc32() internally.
	*/
	static uint32_t
	efi_crc32(const unsigned char *buf, unsigned int size)
	{
	uint32_t crc = crc32(0, Z_NULL, 0);

	crc = crc32(crc, buf, size);

	return (crc);
	}

	static int
	read_disk_info(int fd, diskaddr_t capacity, uint_t lbsize)
	{
	int sector_size;
	unsigned long long capacity_size;

	if (ioctl(fd, BLKSSZGET, &sector_size) < 0)
	return (-1);

	if (ioctl(fd, BLKGETSIZE64, &capacity_size) < 0)
	return (-1);

	*lbsize = (uint_t)sector_size;
	*capacity = (diskaddr_t)(capacity_size / sector_size);

	return (0);
	}

	/*
	* Return back the device name associated with the file descriptor. The
	* caller is responsible for freeing the memory associated with the
	* returned string.
	*/
	static char *
	efi_get_devname(int fd)
	{
	char path[32];

	/*
	* The libefi API only provides the open fd and not the file path.
	* To handle this realpath(3) is used to resolve the block device
	* name from /proc/self/fd/<fd>.
	*/
	(void) snprintf(path, sizeof (path), "/proc/self/fd/%d", fd);
	return (realpath(path, NULL));
	}

	static int
	efi_get_info(int fd, struct dk_cinfo *dki_info)
	{
	char *dev_path;
	int rval = 0;

	memset(dki_info, 0, sizeof (*dki_info));

	/*
	* The simplest way to get the partition number under linux is
	* to parse it out of the /dev/<disk><partition> block device name.
	* The kernel creates this using the partition number when it
	* populates /dev/ so it may be trusted. The tricky bit here is
	* that the naming convention is based on the block device type.
	* So we need to take this in to account when parsing out the
	* partition information. Aside from the partition number we collect
	* some additional device info.
	*/
	dev_path = efi_get_devname(fd);
	if (dev_path == NULL)
	goto error;

	if ((strncmp(dev_path, "/dev/sd", 7) == 0)) {
	strcpy(dki_info->dki_cname, "sd");
	dki_info->dki_ctype = DKC_SCSI_CCS;
	rval = sscanf(dev_path, "/dev/%[a-zA-Z]%hu",
	dki_info->dki_dname,
	&dki_info->dki_partition);
	} else if ((strncmp(dev_path, "/dev/hd", 7) == 0)) {
	strcpy(dki_info->dki_cname, "hd");
	dki_info->dki_ctype = DKC_DIRECT;
	rval = sscanf(dev_path, "/dev/%[a-zA-Z]%hu",
	dki_info->dki_dname,
	&dki_info->dki_partition);
	} else if ((strncmp(dev_path, "/dev/md", 7) == 0)) {
	strcpy(dki_info->dki_cname, "pseudo");
	dki_info->dki_ctype = DKC_MD;
	strcpy(dki_info->dki_dname, "md");
	rval = sscanf(dev_path, "/dev/md%[0-9]p%hu",
	dki_info->dki_dname + 2,
	&dki_info->dki_partition);
	} else if ((strncmp(dev_path, "/dev/vd", 7) == 0)) {
	strcpy(dki_info->dki_cname, "vd");
	dki_info->dki_ctype = DKC_MD;
	rval = sscanf(dev_path, "/dev/%[a-zA-Z]%hu",
	dki_info->dki_dname,
	&dki_info->dki_partition);
	} else if ((strncmp(dev_path, "/dev/xvd", 8) == 0)) {
	strcpy(dki_info->dki_cname, "xvd");
	dki_info->dki_ctype = DKC_MD;
	rval = sscanf(dev_path, "/dev/%[a-zA-Z]%hu",
	dki_info->dki_dname,
	&dki_info->dki_partition);
	} else if ((strncmp(dev_path, "/dev/zd", 7) == 0)) {
	strcpy(dki_info->dki_cname, "zd");
	dki_info->dki_ctype = DKC_MD;
	strcpy(dki_info->dki_dname, "zd");
	rval = sscanf(dev_path, "/dev/zd%[0-9]p%hu",
	dki_info->dki_dname + 2,
	&dki_info->dki_partition);
	} else if ((strncmp(dev_path, "/dev/dm-", 8) == 0)) {
	strcpy(dki_info->dki_cname, "pseudo");
	dki_info->dki_ctype = DKC_VBD;
	strcpy(dki_info->dki_dname, "dm-");
	rval = sscanf(dev_path, "/dev/dm-%[0-9]p%hu",
	dki_info->dki_dname + 3,
	&dki_info->dki_partition);
	} else if ((strncmp(dev_path, "/dev/ram", 8) == 0)) {
	strcpy(dki_info->dki_cname, "pseudo");
	dki_info->dki_ctype = DKC_PCMCIA_MEM;
	strcpy(dki_info->dki_dname, "ram");
	rval = sscanf(dev_path, "/dev/ram%[0-9]p%hu",
	dki_info->dki_dname + 3,
	&dki_info->dki_partition);
	} else if ((strncmp(dev_path, "/dev/loop", 9) == 0)) {
	strcpy(dki_info->dki_cname, "pseudo");
	dki_info->dki_ctype = DKC_VBD;
	strcpy(dki_info->dki_dname, "loop");
	rval = sscanf(dev_path, "/dev/loop%[0-9]p%hu",
	dki_info->dki_dname + 4,
	&dki_info->dki_partition);
	} else if ((strncmp(dev_path, "/dev/nvme", 9) == 0)) {
	strcpy(dki_info->dki_cname, "nvme");
	dki_info->dki_ctype = DKC_SCSI_CCS;
	strcpy(dki_info->dki_dname, "nvme");
	(void) sscanf(dev_path, "/dev/nvme%[0-9]",
	dki_info->dki_dname + 4);
	size_t controller_length = strlen(
	dki_info->dki_dname);
	strcpy(dki_info->dki_dname + controller_length,
	"n");
	rval = sscanf(dev_path,
	"/dev/nvme%*[0-9]n%[0-9]p%hu",
	dki_info->dki_dname + controller_length + 1,
	&dki_info->dki_partition);
	} else {
	strcpy(dki_info->dki_dname, "unknown");
	strcpy(dki_info->dki_cname, "unknown");
	dki_info->dki_ctype = DKC_UNKNOWN;
	}

	switch (rval) {
	case 0:
	errno = EINVAL;
	goto error;
	case 1:
	dki_info->dki_partition = 0;
	}

	free(dev_path);

	return (0);
	error:
	if (efi_debug)
	(void) fprintf(stderr, "DKIOCINFO errno 0x%x\n", errno);

	switch (errno) {
	case EIO:
	return (VT_EIO);
	case EINVAL:
	return (VT_EINVAL);
	default:
	return (VT_ERROR);
	}
	}

	/*
	* the number of blocks the EFI label takes up (round up to nearest
	* block)
	*/
	#define NBLOCKS(p, l) (1 + ((((p) * (int)sizeof (efi_gpe_t)) + \
	((l) - 1)) / (l)))
	/* number of partitions -- limited by what we can malloc */
	#define MAX_PARTS ((4294967295UL - sizeof (struct dk_gpt)) / \
	sizeof (struct dk_part))

	int
	efi_alloc_and_init(int fd, uint32_t nparts, struct dk_gpt **vtoc)
	{
	diskaddr_t capacity = 0;
	uint_t lbsize = 0;
	uint_t nblocks;
	size_t length;
	struct dk_gpt *vptr;
	struct uuid uuid;
	struct dk_cinfo dki_info;

	if (read_disk_info(fd, &capacity, &lbsize) != 0)
	return (-1);

	if (efi_get_info(fd, &dki_info) != 0)
	return (-1);

	if (dki_info.dki_partition != 0)
	return (-1);

	if ((dki_info.dki_ctype == DKC_PCMCIA_MEM) \|\|
	(dki_info.dki_ctype == DKC_VBD) \|\|
	(dki_info.dki_ctype == DKC_UNKNOWN))
	return (-1);

	nblocks = NBLOCKS(nparts, lbsize);
	if ((nblocks * lbsize) < EFI_MIN_ARRAY_SIZE + lbsize) {
	/* 16K plus one block for the GPT */
	nblocks = EFI_MIN_ARRAY_SIZE / lbsize + 1;
	}

	if (nparts > MAX_PARTS) {
	if (efi_debug) {
	(void) fprintf(stderr,
	"the maximum number of partitions supported is %lu\n",
	MAX_PARTS);
	}
	return (-1);
	}

	length = sizeof (struct dk_gpt) +
	sizeof (struct dk_part) * (nparts - 1);

	vptr = calloc(1, length);
	if (vptr == NULL)
	return (-1);

	*vtoc = vptr;

	vptr->efi_version = EFI_VERSION_CURRENT;
	vptr->efi_lbasize = lbsize;
	vptr->efi_nparts = nparts;
	/*
	* add one block here for the PMBR; on disks with a 512 byte
	* block size and 128 or fewer partitions, efi_first_u_lba
	* should work out to "34"
	*/
	vptr->efi_first_u_lba = nblocks + 1;
	vptr->efi_last_lba = capacity - 1;
	vptr->efi_altern_lba = capacity -1;
	vptr->efi_last_u_lba = vptr->efi_last_lba - nblocks;

	(void) uuid_generate((uchar_t *)&uuid);
	UUID_LE_CONVERT(vptr->efi_disk_uguid, uuid);
	return (0);
	}

	/*
	* Read EFI - return partition number upon success.
	*/
	int
	efi_alloc_and_read(int fd, struct dk_gpt **vtoc)
	{
	int rval;
	uint32_t nparts;
	int length;
	struct dk_gpt *vptr;

	/* figure out the number of entries that would fit into 16K */
	nparts = EFI_MIN_ARRAY_SIZE / sizeof (efi_gpe_t);
	length = (int) sizeof (struct dk_gpt) +
	(int) sizeof (struct dk_part) * (nparts - 1);
	vptr = calloc(1, length);

	if (vptr == NULL)
	return (VT_ERROR);

	vptr->efi_nparts = nparts;
	rval = efi_read(fd, vptr);

	if ((rval == VT_EINVAL) && vptr->efi_nparts > nparts) {
	void *tmp;
	length = (int) sizeof (struct dk_gpt) +
	(int) sizeof (struct dk_part) * (vptr->efi_nparts - 1);
	if ((tmp = realloc(vptr, length)) == NULL) {
	/* cppcheck-suppress doubleFree */
	free(vptr);
	*vtoc = NULL;
	return (VT_ERROR);
	} else {
	vptr = tmp;
	rval = efi_read(fd, vptr);
	}
	}

	if (rval < 0) {
	if (efi_debug) {
	(void) fprintf(stderr,
	"read of EFI table failed, rval=%d\n", rval);
	}
	free(vptr);
	*vtoc = NULL;
	} else {
	*vtoc = vptr;
	}

	return (rval);
	}

	static int
	efi_ioctl(int fd, int cmd, dk_efi_t *dk_ioc)
	{
	void *data = dk_ioc->dki_data;
	int error;
	diskaddr_t capacity;
	uint_t lbsize;

	/*
	* When the IO is not being performed in kernel as an ioctl we need
	* to know the sector size so we can seek to the proper byte offset.
	*/
	if (read_disk_info(fd, &capacity, &lbsize) == -1) {
	if (efi_debug)
	fprintf(stderr, "unable to read disk info: %d", errno);

	errno = EIO;
	return (-1);
	}

	switch (cmd) {
	case DKIOCGETEFI:
	if (lbsize == 0) {
	if (efi_debug)
	(void) fprintf(stderr, "DKIOCGETEFI assuming "
	"LBA %d bytes\n", DEV_BSIZE);

	lbsize = DEV_BSIZE;
	}

	error = lseek(fd, dk_ioc->dki_lba * lbsize, SEEK_SET);
	if (error == -1) {
	if (efi_debug)
	(void) fprintf(stderr, "DKIOCGETEFI lseek "
	"error: %d\n", errno);
	return (error);
	}

	error = read(fd, data, dk_ioc->dki_length);
	if (error == -1) {
	if (efi_debug)
	(void) fprintf(stderr, "DKIOCGETEFI read "
	"error: %d\n", errno);
	return (error);
	}

	if (error != dk_ioc->dki_length) {
	if (efi_debug)
	(void) fprintf(stderr, "DKIOCGETEFI short "
	"read of %d bytes\n", error);
	errno = EIO;
	return (-1);
	}
	error = 0;
	break;

	case DKIOCSETEFI:
	if (lbsize == 0) {
	if (efi_debug)
	(void) fprintf(stderr, "DKIOCSETEFI unknown "
	"LBA size\n");
	errno = EIO;
	return (-1);
	}

	error = lseek(fd, dk_ioc->dki_lba * lbsize, SEEK_SET);
	if (error == -1) {
	if (efi_debug)
	(void) fprintf(stderr, "DKIOCSETEFI lseek "
	"error: %d\n", errno);
	return (error);
	}

	error = write(fd, data, dk_ioc->dki_length);
	if (error == -1) {
	if (efi_debug)
	(void) fprintf(stderr, "DKIOCSETEFI write "
	"error: %d\n", errno);
	return (error);
	}

	if (error != dk_ioc->dki_length) {
	if (efi_debug)
	(void) fprintf(stderr, "DKIOCSETEFI short "
	"write of %d bytes\n", error);
	errno = EIO;
	return (-1);
	}

	/* Sync the new EFI table to disk */
	error = fsync(fd);
	if (error == -1)
	return (error);

	/* Ensure any local disk cache is also flushed */
	if (ioctl(fd, BLKFLSBUF, 0) == -1)
	return (error);

	error = 0;
	break;

	default:
	if (efi_debug)
	(void) fprintf(stderr, "unsupported ioctl()\n");

	errno = EIO;
	return (-1);
	}

	return (error);
	}

	int
	efi_rescan(int fd)
	{
	int retry = 10;

	/* Notify the kernel a devices partition table has been updated */
	while (ioctl(fd, BLKRRPART) != 0) {
	if ((--retry == 0) \|\| (errno != EBUSY)) {
	(void) fprintf(stderr, "the kernel failed to rescan "
	"the partition table: %d\n", errno);
	return (-1);
	}
	usleep(50000);
	}

	return (0);
	}

	static int
	check_label(int fd, dk_efi_t *dk_ioc)
	{
	efi_gpt_t *efi;
	uint_t crc;

	if (efi_ioctl(fd, DKIOCGETEFI, dk_ioc) == -1) {
	switch (errno) {
	case EIO:
	return (VT_EIO);
	default:
	return (VT_ERROR);
	}
	}
	efi = dk_ioc->dki_data;
	if (efi->efi_gpt_Signature != LE_64(EFI_SIGNATURE)) {
	if (efi_debug)
	(void) fprintf(stderr,
	"Bad EFI signature: 0x%llx != 0x%llx\n",
	(long long)efi->efi_gpt_Signature,
	(long long)LE_64(EFI_SIGNATURE));
	return (VT_EINVAL);
	}

	/*
	* check CRC of the header; the size of the header should
	* never be larger than one block
	*/
	crc = efi->efi_gpt_HeaderCRC32;
	efi->efi_gpt_HeaderCRC32 = 0;
	len_t headerSize = (len_t)LE_32(efi->efi_gpt_HeaderSize);

	if (headerSize < EFI_MIN_LABEL_SIZE \|\| headerSize > EFI_LABEL_SIZE) {
	if (efi_debug)
	(void) fprintf(stderr,
	"Invalid EFI HeaderSize %llu. Assuming %d.\n",
	headerSize, EFI_MIN_LABEL_SIZE);
	}

	if ((headerSize > dk_ioc->dki_length) \|\|
	crc != LE_32(efi_crc32((unsigned char *)efi, headerSize))) {
	if (efi_debug)
	(void) fprintf(stderr,
	"Bad EFI CRC: 0x%x != 0x%x\n",
	crc, LE_32(efi_crc32((unsigned char *)efi,
	headerSize)));
	return (VT_EINVAL);
	}

	return (0);
	}

	static int
	efi_read(int fd, struct dk_gpt *vtoc)
	{
	int i, j;
	int label_len;
	int rval = 0;
	int md_flag = 0;
	int vdc_flag = 0;
	diskaddr_t capacity = 0;
	uint_t lbsize = 0;
	struct dk_minfo disk_info;
	dk_efi_t dk_ioc;
	efi_gpt_t *efi;
	efi_gpe_t *efi_parts;
	struct dk_cinfo dki_info;
	uint32_t user_length;
	boolean_t legacy_label = B_FALSE;

	/*
	* get the partition number for this file descriptor.
	*/
	if ((rval = efi_get_info(fd, &dki_info)) != 0)
	return (rval);

	if ((strncmp(dki_info.dki_cname, "pseudo", 7) == 0) &&
	(strncmp(dki_info.dki_dname, "md", 3) == 0)) {
	md_flag++;
	} else if ((strncmp(dki_info.dki_cname, "vdc", 4) == 0) &&
	(strncmp(dki_info.dki_dname, "vdc", 4) == 0)) {
	/*
	* The controller and drive name "vdc" (virtual disk client)
	* indicates a LDoms virtual disk.
	*/
	vdc_flag++;
	}

	/* get the LBA size */
	if (read_disk_info(fd, &capacity, &lbsize) == -1) {
	if (efi_debug) {
	(void) fprintf(stderr,
	"unable to read disk info: %d",
	errno);
	}
	return (VT_EINVAL);
	}

	disk_info.dki_lbsize = lbsize;
	disk_info.dki_capacity = capacity;

	if (disk_info.dki_lbsize == 0) {
	if (efi_debug) {
	(void) fprintf(stderr,
	"efi_read: assuming LBA 512 bytes\n");
	}
	disk_info.dki_lbsize = DEV_BSIZE;
	}
	/*
	* Read the EFI GPT to figure out how many partitions we need
	* to deal with.
	*/
	dk_ioc.dki_lba = 1;
	if (NBLOCKS(vtoc->efi_nparts, disk_info.dki_lbsize) < 34) {
	label_len = EFI_MIN_ARRAY_SIZE + disk_info.dki_lbsize;
	} else {
	label_len = vtoc->efi_nparts * (int) sizeof (efi_gpe_t) +
	disk_info.dki_lbsize;
	if (label_len % disk_info.dki_lbsize) {
	/* pad to physical sector size */
	label_len += disk_info.dki_lbsize;
	label_len &= ~(disk_info.dki_lbsize - 1);
	}
	}

	if (posix_memalign((void **)&dk_ioc.dki_data,
	disk_info.dki_lbsize, label_len))
	return (VT_ERROR);

	memset(dk_ioc.dki_data, 0, label_len);
	dk_ioc.dki_length = disk_info.dki_lbsize;
	user_length = vtoc->efi_nparts;
	efi = dk_ioc.dki_data;
	if (md_flag) {
	dk_ioc.dki_length = label_len;
	if (efi_ioctl(fd, DKIOCGETEFI, &dk_ioc) == -1) {
	switch (errno) {
	case EIO:
	return (VT_EIO);
	default:
	return (VT_ERROR);
	}
	}
	} else if ((rval = check_label(fd, &dk_ioc)) == VT_EINVAL) {
	/*
	* No valid label here; try the alternate. Note that here
	* we just read GPT header and save it into dk_ioc.data,
	* Later, we will read GUID partition entry array if we
	* can get valid GPT header.
	*/

	/*
	* This is a workaround for legacy systems. In the past, the
	* last sector of SCSI disk was invisible on x86 platform. At
	* that time, backup label was saved on the next to the last
	* sector. It is possible for users to move a disk from previous
	* solaris system to present system. Here, we attempt to search
	* legacy backup EFI label first.
	*/
	dk_ioc.dki_lba = disk_info.dki_capacity - 2;
	dk_ioc.dki_length = disk_info.dki_lbsize;
	rval = check_label(fd, &dk_ioc);
	if (rval == VT_EINVAL) {
	/*
	* we didn't find legacy backup EFI label, try to
	* search backup EFI label in the last block.
	*/
	dk_ioc.dki_lba = disk_info.dki_capacity - 1;
	dk_ioc.dki_length = disk_info.dki_lbsize;
	rval = check_label(fd, &dk_ioc);
	if (rval == 0) {
	legacy_label = B_TRUE;
	if (efi_debug)
	(void) fprintf(stderr,
	"efi_read: primary label corrupt; "
	"using EFI backup label located on"
	" the last block\n");
	}
	} else {
	if ((efi_debug) && (rval == 0))
	(void) fprintf(stderr, "efi_read: primary label"
	" corrupt; using legacy EFI backup label "
	" located on the next to last block\n");
	}

	if (rval == 0) {
	dk_ioc.dki_lba = LE_64(efi->efi_gpt_PartitionEntryLBA);
	vtoc->efi_flags \|= EFI_GPT_PRIMARY_CORRUPT;
	vtoc->efi_nparts =
	LE_32(efi->efi_gpt_NumberOfPartitionEntries);
	/*
	* Partition tables are between backup GPT header
	* table and ParitionEntryLBA (the starting LBA of
	* the GUID partition entries array). Now that we
	* already got valid GPT header and saved it in
	* dk_ioc.dki_data, we try to get GUID partition
	* entry array here.
	*/
	/* LINTED */
	dk_ioc.dki_data = (efi_gpt_t )((char )dk_ioc.dki_data
	+ disk_info.dki_lbsize);
	if (legacy_label)
	dk_ioc.dki_length = disk_info.dki_capacity - 1 -
	dk_ioc.dki_lba;
	else
	dk_ioc.dki_length = disk_info.dki_capacity - 2 -
	dk_ioc.dki_lba;
	dk_ioc.dki_length *= disk_info.dki_lbsize;
	if (dk_ioc.dki_length >
	((len_t)label_len - sizeof (*dk_ioc.dki_data))) {
	rval = VT_EINVAL;
	} else {
	/*
	* read GUID partition entry array
	*/
	rval = efi_ioctl(fd, DKIOCGETEFI, &dk_ioc);
	}
	}

	} else if (rval == 0) {

	dk_ioc.dki_lba = LE_64(efi->efi_gpt_PartitionEntryLBA);
	/* LINTED */
	dk_ioc.dki_data = (efi_gpt_t )((char )dk_ioc.dki_data
	+ disk_info.dki_lbsize);
	dk_ioc.dki_length = label_len - disk_info.dki_lbsize;
	rval = efi_ioctl(fd, DKIOCGETEFI, &dk_ioc);

	} else if (vdc_flag && rval == VT_ERROR && errno == EINVAL) {
	/*
	* When the device is a LDoms virtual disk, the DKIOCGETEFI
	* ioctl can fail with EINVAL if the virtual disk backend
	* is a ZFS volume serviced by a domain running an old version
	* of Solaris. This is because the DKIOCGETEFI ioctl was
	* initially incorrectly implemented for a ZFS volume and it
	* expected the GPT and GPE to be retrieved with a single ioctl.
	* So we try to read the GPT and the GPE using that old style
	* ioctl.
	*/
	dk_ioc.dki_lba = 1;
	dk_ioc.dki_length = label_len;
	rval = check_label(fd, &dk_ioc);
	}

	if (rval < 0) {
	free(efi);
	return (rval);
	}

	/* LINTED -- always longlong aligned */
	efi_parts = (efi_gpe_t )(((char )efi) + disk_info.dki_lbsize);

	/*
	* Assemble this into a "dk_gpt" struct for easier
	* digestibility by applications.
	*/
	vtoc->efi_version = LE_32(efi->efi_gpt_Revision);
	vtoc->efi_nparts = LE_32(efi->efi_gpt_NumberOfPartitionEntries);
	vtoc->efi_part_size = LE_32(efi->efi_gpt_SizeOfPartitionEntry);
	vtoc->efi_lbasize = disk_info.dki_lbsize;
	vtoc->efi_last_lba = disk_info.dki_capacity - 1;
	vtoc->efi_first_u_lba = LE_64(efi->efi_gpt_FirstUsableLBA);
	vtoc->efi_last_u_lba = LE_64(efi->efi_gpt_LastUsableLBA);
	vtoc->efi_altern_lba = LE_64(efi->efi_gpt_AlternateLBA);
	UUID_LE_CONVERT(vtoc->efi_disk_uguid, efi->efi_gpt_DiskGUID);

	/*
	* If the array the user passed in is too small, set the length
	* to what it needs to be and return
	*/
	if (user_length < vtoc->efi_nparts) {
	return (VT_EINVAL);
	}

	for (i = 0; i < vtoc->efi_nparts; i++) {
	UUID_LE_CONVERT(vtoc->efi_parts[i].p_guid,
	efi_parts[i].efi_gpe_PartitionTypeGUID);

	for (j = 0;
	j < sizeof (conversion_array)
	/ sizeof (struct uuid_to_ptag); j++) {

	if (memcmp(&vtoc->efi_parts[i].p_guid,
	&conversion_array[j].uuid,
	sizeof (struct uuid)) == 0) {
	vtoc->efi_parts[i].p_tag = j;
	break;
	}
	}
	if (vtoc->efi_parts[i].p_tag == V_UNASSIGNED)
	continue;
	vtoc->efi_parts[i].p_flag =
	LE_16(efi_parts[i].efi_gpe_Attributes.PartitionAttrs);
	vtoc->efi_parts[i].p_start =
	LE_64(efi_parts[i].efi_gpe_StartingLBA);
	vtoc->efi_parts[i].p_size =
	LE_64(efi_parts[i].efi_gpe_EndingLBA) -
	vtoc->efi_parts[i].p_start + 1;
	for (j = 0; j < EFI_PART_NAME_LEN; j++) {
	vtoc->efi_parts[i].p_name[j] =
	(uchar_t)LE_16(
	efi_parts[i].efi_gpe_PartitionName[j]);
	}

	UUID_LE_CONVERT(vtoc->efi_parts[i].p_uguid,
	efi_parts[i].efi_gpe_UniquePartitionGUID);
	}
	free(efi);

	return (dki_info.dki_partition);
	}

	/* writes a "protective" MBR */
	static int
	write_pmbr(int fd, struct dk_gpt *vtoc)
	{
	dk_efi_t dk_ioc;
	struct mboot mb;
	uchar_t *cp;
	diskaddr_t size_in_lba;
	uchar_t *buf;
	int len;

	len = (vtoc->efi_lbasize == 0) ? sizeof (mb) : vtoc->efi_lbasize;
	if (posix_memalign((void **)&buf, len, len))
	return (VT_ERROR);

	/*
	* Preserve any boot code and disk signature if the first block is
	* already an MBR.
	*/
	memset(buf, 0, len);
	dk_ioc.dki_lba = 0;
	dk_ioc.dki_length = len;
	/* LINTED -- always longlong aligned */
	dk_ioc.dki_data = (efi_gpt_t *)buf;
	if (efi_ioctl(fd, DKIOCGETEFI, &dk_ioc) == -1) {
	memset(&mb, 0, sizeof (mb));
	mb.signature = LE_16(MBB_MAGIC);
	} else {
	(void) memcpy(&mb, buf, sizeof (mb));
	if (mb.signature != LE_16(MBB_MAGIC)) {
	memset(&mb, 0, sizeof (mb));
	mb.signature = LE_16(MBB_MAGIC);
	}
	}

	memset(&mb.parts, 0, sizeof (mb.parts));
	cp = (uchar_t *)&mb.parts[0];
	/* bootable or not */
	*cp++ = 0;
	/* beginning CHS; 0xffffff if not representable */
	*cp++ = 0xff;
	*cp++ = 0xff;
	*cp++ = 0xff;
	/* OS type */
	*cp++ = EFI_PMBR;
	/* ending CHS; 0xffffff if not representable */
	*cp++ = 0xff;
	*cp++ = 0xff;
	*cp++ = 0xff;
	/* starting LBA: 1 (little endian format) by EFI definition */
	*cp++ = 0x01;
	*cp++ = 0x00;
	*cp++ = 0x00;
	*cp++ = 0x00;
	/* ending LBA: last block on the disk (little endian format) */
	size_in_lba = vtoc->efi_last_lba;
	if (size_in_lba < 0xffffffff) {
	*cp++ = (size_in_lba & 0x000000ff);
	*cp++ = (size_in_lba & 0x0000ff00) >> 8;
	*cp++ = (size_in_lba & 0x00ff0000) >> 16;
	*cp++ = (size_in_lba & 0xff000000) >> 24;
	} else {
	*cp++ = 0xff;
	*cp++ = 0xff;
	*cp++ = 0xff;
	*cp++ = 0xff;
	}

	(void) memcpy(buf, &mb, sizeof (mb));
	/* LINTED -- always longlong aligned */
	dk_ioc.dki_data = (efi_gpt_t *)buf;
	dk_ioc.dki_lba = 0;
	dk_ioc.dki_length = len;
	if (efi_ioctl(fd, DKIOCSETEFI, &dk_ioc) == -1) {
	free(buf);
	switch (errno) {
	case EIO:
	return (VT_EIO);
	case EINVAL:
	return (VT_EINVAL);
	default:
	return (VT_ERROR);
	}
	}
	free(buf);
	return (0);
	}

	/* make sure the user specified something reasonable */
	static int
	check_input(struct dk_gpt *vtoc)
	{
	int resv_part = -1;
	int i, j;
	diskaddr_t istart, jstart, isize, jsize, endsect;

	/*
	* Sanity-check the input (make sure no partitions overlap)
	*/
	for (i = 0; i < vtoc->efi_nparts; i++) {
	/* It can't be unassigned and have an actual size */
	if ((vtoc->efi_parts[i].p_tag == V_UNASSIGNED) &&
	(vtoc->efi_parts[i].p_size != 0)) {
	if (efi_debug) {
	(void) fprintf(stderr, "partition %d is "
	"\"unassigned\" but has a size of %llu",
	i, vtoc->efi_parts[i].p_size);
	}
	return (VT_EINVAL);
	}
	if (vtoc->efi_parts[i].p_tag == V_UNASSIGNED) {
	if (uuid_is_null((uchar_t *)&vtoc->efi_parts[i].p_guid))
	continue;
	/* we have encountered an unknown uuid */
	vtoc->efi_parts[i].p_tag = 0xff;
	}
	if (vtoc->efi_parts[i].p_tag == V_RESERVED) {
	if (resv_part != -1) {
	if (efi_debug) {
	(void) fprintf(stderr, "found "
	"duplicate reserved partition "
	"at %d\n", i);
	}
	return (VT_EINVAL);
	}
	resv_part = i;
	}
	if ((vtoc->efi_parts[i].p_start < vtoc->efi_first_u_lba) \|\|
	(vtoc->efi_parts[i].p_start > vtoc->efi_last_u_lba)) {
	if (efi_debug) {
	(void) fprintf(stderr,
	"Partition %d starts at %llu. ",
	i,
	vtoc->efi_parts[i].p_start);
	(void) fprintf(stderr,
	"It must be between %llu and %llu.\n",
	vtoc->efi_first_u_lba,
	vtoc->efi_last_u_lba);
	}
	return (VT_EINVAL);
	}
	if ((vtoc->efi_parts[i].p_start +
	vtoc->efi_parts[i].p_size <
	vtoc->efi_first_u_lba) \|\|
	(vtoc->efi_parts[i].p_start +
	vtoc->efi_parts[i].p_size >
	vtoc->efi_last_u_lba + 1)) {
	if (efi_debug) {
	(void) fprintf(stderr,
	"Partition %d ends at %llu. ",
	i,
	vtoc->efi_parts[i].p_start +
	vtoc->efi_parts[i].p_size);
	(void) fprintf(stderr,
	"It must be between %llu and %llu.\n",
	vtoc->efi_first_u_lba,
	vtoc->efi_last_u_lba);
	}
	return (VT_EINVAL);
	}

	for (j = 0; j < vtoc->efi_nparts; j++) {
	isize = vtoc->efi_parts[i].p_size;
	jsize = vtoc->efi_parts[j].p_size;
	istart = vtoc->efi_parts[i].p_start;
	jstart = vtoc->efi_parts[j].p_start;
	if ((i != j) && (isize != 0) && (jsize != 0)) {
	endsect = jstart + jsize -1;
	if ((jstart <= istart) &&
	(istart <= endsect)) {
	if (efi_debug) {
	(void) fprintf(stderr,
	"Partition %d overlaps "
	"partition %d.", i, j);
	}
	return (VT_EINVAL);
	}
	}
	}
	}
	/* just a warning for now */
	if ((resv_part == -1) && efi_debug) {
	(void) fprintf(stderr,
	"no reserved partition found\n");
	}
	return (0);
	}

	static int
	call_blkpg_ioctl(int fd, int command, diskaddr_t start,
	diskaddr_t size, uint_t pno)
	{
	struct blkpg_ioctl_arg ioctl_arg;
	struct blkpg_partition linux_part;
	memset(&linux_part, 0, sizeof (linux_part));

	char *path = efi_get_devname(fd);
	if (path == NULL) {
	(void) fprintf(stderr, "failed to retrieve device name\n");
	return (VT_EINVAL);
	}

	linux_part.start = start;
	linux_part.length = size;
	linux_part.pno = pno;
	snprintf(linux_part.devname, BLKPG_DEVNAMELTH - 1, "%s%u", path, pno);
	linux_part.devname[BLKPG_DEVNAMELTH - 1] = '\0';
	free(path);

	ioctl_arg.op = command;
	ioctl_arg.flags = 0;
	ioctl_arg.datalen = sizeof (struct blkpg_partition);
	ioctl_arg.data = &linux_part;

	return (ioctl(fd, BLKPG, &ioctl_arg));
	}

	/*
	* add all the unallocated space to the current label
	*/
	int
	efi_use_whole_disk(int fd)
	{
	struct dk_gpt *efi_label = NULL;
	int rval;
	int i;
	uint_t resv_index = 0, data_index = 0;
	diskaddr_t resv_start = 0, data_start = 0;
	diskaddr_t data_size, limit, difference;
	boolean_t sync_needed = B_FALSE;
	uint_t nblocks;

	rval = efi_alloc_and_read(fd, &efi_label);
	if (rval < 0) {
	if (efi_label != NULL)
	efi_free(efi_label);
	return (rval);
	}

	/*
	* Find the last physically non-zero partition.
	* This should be the reserved partition.
	*/
	for (i = 0; i < efi_label->efi_nparts; i ++) {
	if (resv_start < efi_label->efi_parts[i].p_start) {
	resv_start = efi_label->efi_parts[i].p_start;
	resv_index = i;
	}
	}

	/*
	* Find the last physically non-zero partition before that.
	* This is the data partition.
	*/
	for (i = 0; i < resv_index; i ++) {
	if (data_start < efi_label->efi_parts[i].p_start) {
	data_start = efi_label->efi_parts[i].p_start;
	data_index = i;
	}
	}
	data_size = efi_label->efi_parts[data_index].p_size;

	/*
	* See the "efi_alloc_and_init" function for more information
	* about where this "nblocks" value comes from.
	*/
	nblocks = efi_label->efi_first_u_lba - 1;

	/*
	* Determine if the EFI label is out of sync. We check that:
	*
	* 1. the data partition ends at the limit we set, and
	* 2. the reserved partition starts at the limit we set.
	*
	* If either of these conditions is not met, then we need to
	* resync the EFI label.
	*
	* The limit is the last usable LBA, determined by the last LBA
	* and the first usable LBA fields on the EFI label of the disk
	* (see the lines directly above). Additionally, we factor in
	* EFI_MIN_RESV_SIZE (per its use in "zpool_label_disk") and
	* P2ALIGN it to ensure the partition boundaries are aligned
	* (for performance reasons). The alignment should match the
	* alignment used by the "zpool_label_disk" function.
	*/
	- limit = P2ALIGN(efi_label->efi_last_lba - nblocks - EFI_MIN_RESV_SIZE,
	- PARTITION_END_ALIGNMENT);
	+ limit = P2ALIGN_TYPED(efi_label->efi_last_lba - nblocks -
	+ EFI_MIN_RESV_SIZE, PARTITION_END_ALIGNMENT, diskaddr_t);
	if (data_start + data_size != limit \|\| resv_start != limit)
	sync_needed = B_TRUE;

	if (efi_debug && sync_needed)
	(void) fprintf(stderr, "efi_use_whole_disk: sync needed\n");

	/*
	* If alter_lba is 1, we are using the backup label.
	* Since we can locate the backup label by disk capacity,
	* there must be no unallocated space.
	*/
	if ((efi_label->efi_altern_lba == 1) \|\| (efi_label->efi_altern_lba
	>= efi_label->efi_last_lba && !sync_needed)) {
	if (efi_debug) {
	(void) fprintf(stderr,
	"efi_use_whole_disk: requested space not found\n");
	}
	efi_free(efi_label);
	return (VT_ENOSPC);
	}

	/*
	* Verify that we've found the reserved partition by checking
	* that it looks the way it did when we created it in zpool_label_disk.
	* If we've found the incorrect partition, then we know that this
	* device was reformatted and no longer is solely used by ZFS.
	*/
	if ((efi_label->efi_parts[resv_index].p_size != EFI_MIN_RESV_SIZE) \|\|
	(efi_label->efi_parts[resv_index].p_tag != V_RESERVED) \|\|
	(resv_index != 8)) {
	if (efi_debug) {
	(void) fprintf(stderr,
	"efi_use_whole_disk: wholedisk not available\n");
	}
	efi_free(efi_label);
	return (VT_ENOSPC);
	}

	if (data_start + data_size != resv_start) {
	if (efi_debug) {
	(void) fprintf(stderr,
	"efi_use_whole_disk: "
	"data_start (%lli) + "
	"data_size (%lli) != "
	"resv_start (%lli)\n",
	data_start, data_size, resv_start);
	}

	return (VT_EINVAL);
	}

	if (limit < resv_start) {
	if (efi_debug) {
	(void) fprintf(stderr,
	"efi_use_whole_disk: "
	"limit (%lli) < resv_start (%lli)\n",
	limit, resv_start);
	}

	return (VT_EINVAL);
	}

	difference = limit - resv_start;

	if (efi_debug)
	(void) fprintf(stderr,
	"efi_use_whole_disk: difference is %lli\n", difference);

	/*
	* Move the reserved partition. There is currently no data in
	* here except fabricated devids (which get generated via
	* efi_write()). So there is no need to copy data.
	*/
	efi_label->efi_parts[data_index].p_size += difference;
	efi_label->efi_parts[resv_index].p_start += difference;
	efi_label->efi_last_u_lba = efi_label->efi_last_lba - nblocks;

	/*
	* Rescanning the partition table in the kernel can result
	* in the device links to be removed (see comment in vdev_disk_open).
	* If BLKPG_RESIZE_PARTITION is available, then we can resize
	* the partition table online and avoid having to remove the device
	* links used by the pool. This provides a very deterministic
	* approach to resizing devices and does not require any
	* loops waiting for devices to reappear.
	*/
	#ifdef BLKPG_RESIZE_PARTITION
	/*
	* Delete the reserved partition since we're about to expand
	* the data partition and it would overlap with the reserved
	* partition.
	* NOTE: The starting index for the ioctl is 1 while for the
	* EFI partitions it's 0. For that reason we have to add one
	* whenever we make an ioctl call.
	*/
	rval = call_blkpg_ioctl(fd, BLKPG_DEL_PARTITION, 0, 0, resv_index + 1);
	if (rval != 0)
	goto out;

	/*
	* Expand the data partition
	*/
	rval = call_blkpg_ioctl(fd, BLKPG_RESIZE_PARTITION,
	efi_label->efi_parts[data_index].p_start * efi_label->efi_lbasize,
	efi_label->efi_parts[data_index].p_size * efi_label->efi_lbasize,
	data_index + 1);
	if (rval != 0) {
	(void) fprintf(stderr, "Unable to resize data "
	"partition: %d\n", rval);
	/*
	* Since we failed to resize, we need to reset the start
	* of the reserve partition and re-create it.
	*/
	efi_label->efi_parts[resv_index].p_start -= difference;
	}

	/*
	* Re-add the reserved partition. If we've expanded the data partition
	* then we'll move the reserve partition to the end of the data
	* partition. Otherwise, we'll recreate the partition in its original
	* location. Note that we do this as best-effort and ignore any
	* errors that may arise here. This will ensure that we finish writing
	* the EFI label.
	*/
	(void) call_blkpg_ioctl(fd, BLKPG_ADD_PARTITION,
	efi_label->efi_parts[resv_index].p_start * efi_label->efi_lbasize,
	efi_label->efi_parts[resv_index].p_size * efi_label->efi_lbasize,
	resv_index + 1);
	#endif

	/*
	* We're now ready to write the EFI label.
	*/
	if (rval == 0) {
	rval = efi_write(fd, efi_label);
	if (rval < 0 && efi_debug) {
	(void) fprintf(stderr, "efi_use_whole_disk:fail "
	"to write label, rval=%d\n", rval);
	}
	}

	out:
	efi_free(efi_label);
	return (rval);
	}

	/*
	* write EFI label and backup label
	*/
	int
	efi_write(int fd, struct dk_gpt *vtoc)
	{
	dk_efi_t dk_ioc;
	efi_gpt_t *efi;
	efi_gpe_t *efi_parts;
	int i, j;
	struct dk_cinfo dki_info;
	int rval;
	int md_flag = 0;
	int nblocks;
	diskaddr_t lba_backup_gpt_hdr;

	if ((rval = efi_get_info(fd, &dki_info)) != 0)
	return (rval);

	/* check if we are dealing with a metadevice */
	if ((strncmp(dki_info.dki_cname, "pseudo", 7) == 0) &&
	(strncmp(dki_info.dki_dname, "md", 3) == 0)) {
	md_flag = 1;
	}

	if (check_input(vtoc)) {
	/*
	* not valid; if it's a metadevice just pass it down
	* because SVM will do its own checking
	*/
	if (md_flag == 0) {
	return (VT_EINVAL);
	}
	}

	dk_ioc.dki_lba = 1;
	if (NBLOCKS(vtoc->efi_nparts, vtoc->efi_lbasize) < 34) {
	dk_ioc.dki_length = EFI_MIN_ARRAY_SIZE + vtoc->efi_lbasize;
	} else {
	dk_ioc.dki_length = (len_t)NBLOCKS(vtoc->efi_nparts,
	vtoc->efi_lbasize) *
	vtoc->efi_lbasize;
	}

	/*
	* the number of blocks occupied by GUID partition entry array
	*/
	nblocks = dk_ioc.dki_length / vtoc->efi_lbasize - 1;

	/*
	* Backup GPT header is located on the block after GUID
	* partition entry array. Here, we calculate the address
	* for backup GPT header.
	*/
	lba_backup_gpt_hdr = vtoc->efi_last_u_lba + 1 + nblocks;
	if (posix_memalign((void **)&dk_ioc.dki_data,
	vtoc->efi_lbasize, dk_ioc.dki_length))
	return (VT_ERROR);

	memset(dk_ioc.dki_data, 0, dk_ioc.dki_length);
	efi = dk_ioc.dki_data;

	/* stuff user's input into EFI struct */
	efi->efi_gpt_Signature = LE_64(EFI_SIGNATURE);
	efi->efi_gpt_Revision = LE_32(vtoc->efi_version); /* 0x02000100 */
	efi->efi_gpt_HeaderSize = LE_32(sizeof (struct efi_gpt) - LEN_EFI_PAD);
	efi->efi_gpt_Reserved1 = 0;
	efi->efi_gpt_MyLBA = LE_64(1ULL);
	efi->efi_gpt_AlternateLBA = LE_64(lba_backup_gpt_hdr);
	efi->efi_gpt_FirstUsableLBA = LE_64(vtoc->efi_first_u_lba);
	efi->efi_gpt_LastUsableLBA = LE_64(vtoc->efi_last_u_lba);
	efi->efi_gpt_PartitionEntryLBA = LE_64(2ULL);
	efi->efi_gpt_NumberOfPartitionEntries = LE_32(vtoc->efi_nparts);
	efi->efi_gpt_SizeOfPartitionEntry = LE_32(sizeof (struct efi_gpe));
	UUID_LE_CONVERT(efi->efi_gpt_DiskGUID, vtoc->efi_disk_uguid);

	/* LINTED -- always longlong aligned */
	efi_parts = (efi_gpe_t )((char )dk_ioc.dki_data + vtoc->efi_lbasize);

	for (i = 0; i < vtoc->efi_nparts; i++) {
	for (j = 0;
	j < sizeof (conversion_array) /
	sizeof (struct uuid_to_ptag); j++) {

	if (vtoc->efi_parts[i].p_tag == j) {
	UUID_LE_CONVERT(
	efi_parts[i].efi_gpe_PartitionTypeGUID,
	conversion_array[j].uuid);
	break;
	}
	}

	if (j == sizeof (conversion_array) /
	sizeof (struct uuid_to_ptag)) {
	/*
	* If we didn't have a matching uuid match, bail here.
	* Don't write a label with unknown uuid.
	*/
	if (efi_debug) {
	(void) fprintf(stderr,
	"Unknown uuid for p_tag %d\n",
	vtoc->efi_parts[i].p_tag);
	}
	return (VT_EINVAL);
	}

	/* Zero's should be written for empty partitions */
	if (vtoc->efi_parts[i].p_tag == V_UNASSIGNED)
	continue;

	efi_parts[i].efi_gpe_StartingLBA =
	LE_64(vtoc->efi_parts[i].p_start);
	efi_parts[i].efi_gpe_EndingLBA =
	LE_64(vtoc->efi_parts[i].p_start +
	vtoc->efi_parts[i].p_size - 1);
	efi_parts[i].efi_gpe_Attributes.PartitionAttrs =
	LE_16(vtoc->efi_parts[i].p_flag);
	for (j = 0; j < EFI_PART_NAME_LEN; j++) {
	efi_parts[i].efi_gpe_PartitionName[j] =
	LE_16((ushort_t)vtoc->efi_parts[i].p_name[j]);
	}
	if ((vtoc->efi_parts[i].p_tag != V_UNASSIGNED) &&
	uuid_is_null((uchar_t *)&vtoc->efi_parts[i].p_uguid)) {
	(void) uuid_generate((uchar_t *)
	&vtoc->efi_parts[i].p_uguid);
	}
	memcpy(&efi_parts[i].efi_gpe_UniquePartitionGUID,
	&vtoc->efi_parts[i].p_uguid,
	sizeof (uuid_t));
	}
	efi->efi_gpt_PartitionEntryArrayCRC32 =
	LE_32(efi_crc32((unsigned char *)efi_parts,
	vtoc->efi_nparts * (int)sizeof (struct efi_gpe)));
	efi->efi_gpt_HeaderCRC32 =
	LE_32(efi_crc32((unsigned char *)efi,
	LE_32(efi->efi_gpt_HeaderSize)));

	if (efi_ioctl(fd, DKIOCSETEFI, &dk_ioc) == -1) {
	free(dk_ioc.dki_data);
	switch (errno) {
	case EIO:
	return (VT_EIO);
	case EINVAL:
	return (VT_EINVAL);
	default:
	return (VT_ERROR);
	}
	}
	/* if it's a metadevice we're done */
	if (md_flag) {
	free(dk_ioc.dki_data);
	return (0);
	}

	/* write backup partition array */
	dk_ioc.dki_lba = vtoc->efi_last_u_lba + 1;
	dk_ioc.dki_length -= vtoc->efi_lbasize;
	/* LINTED */
	dk_ioc.dki_data = (efi_gpt_t )((char )dk_ioc.dki_data +
	vtoc->efi_lbasize);

	if (efi_ioctl(fd, DKIOCSETEFI, &dk_ioc) == -1) {
	/*
	* we wrote the primary label okay, so don't fail
	*/
	if (efi_debug) {
	(void) fprintf(stderr,
	"write of backup partitions to block %llu "
	"failed, errno %d\n",
	vtoc->efi_last_u_lba + 1,
	errno);
	}
	}
	/*
	* now swap MyLBA and AlternateLBA fields and write backup
	* partition table header
	*/
	dk_ioc.dki_lba = lba_backup_gpt_hdr;
	dk_ioc.dki_length = vtoc->efi_lbasize;
	/* LINTED */
	dk_ioc.dki_data = (efi_gpt_t )((char )dk_ioc.dki_data -
	vtoc->efi_lbasize);
	efi->efi_gpt_AlternateLBA = LE_64(1ULL);
	efi->efi_gpt_MyLBA = LE_64(lba_backup_gpt_hdr);
	efi->efi_gpt_PartitionEntryLBA = LE_64(vtoc->efi_last_u_lba + 1);
	efi->efi_gpt_HeaderCRC32 = 0;
	efi->efi_gpt_HeaderCRC32 =
	LE_32(efi_crc32((unsigned char *)dk_ioc.dki_data,
	LE_32(efi->efi_gpt_HeaderSize)));

	if (efi_ioctl(fd, DKIOCSETEFI, &dk_ioc) == -1) {
	if (efi_debug) {
	(void) fprintf(stderr,
	"write of backup header to block %llu failed, "
	"errno %d\n",
	lba_backup_gpt_hdr,
	errno);
	}
	}
	/* write the PMBR */
	(void) write_pmbr(fd, vtoc);
	free(dk_ioc.dki_data);

	return (0);
	}

	void
	efi_free(struct dk_gpt *ptr)
	{
	free(ptr);
	}

	void
	efi_err_check(struct dk_gpt *vtoc)
	{
	int resv_part = -1;
	int i, j;
	diskaddr_t istart, jstart, isize, jsize, endsect;
	int overlap = 0;

	/*
	* make sure no partitions overlap
	*/
	for (i = 0; i < vtoc->efi_nparts; i++) {
	/* It can't be unassigned and have an actual size */
	if ((vtoc->efi_parts[i].p_tag == V_UNASSIGNED) &&
	(vtoc->efi_parts[i].p_size != 0)) {
	(void) fprintf(stderr,
	"partition %d is \"unassigned\" but has a size "
	"of %llu\n", i, vtoc->efi_parts[i].p_size);
	}
	if (vtoc->efi_parts[i].p_tag == V_UNASSIGNED) {
	continue;
	}
	if (vtoc->efi_parts[i].p_tag == V_RESERVED) {
	if (resv_part != -1) {
	(void) fprintf(stderr,
	"found duplicate reserved partition at "
	"%d\n", i);
	}
	resv_part = i;
	if (vtoc->efi_parts[i].p_size != EFI_MIN_RESV_SIZE)
	(void) fprintf(stderr,
	"Warning: reserved partition size must "
	"be %d sectors\n", EFI_MIN_RESV_SIZE);
	}
	if ((vtoc->efi_parts[i].p_start < vtoc->efi_first_u_lba) \|\|
	(vtoc->efi_parts[i].p_start > vtoc->efi_last_u_lba)) {
	(void) fprintf(stderr,
	"Partition %d starts at %llu\n",
	i,
	vtoc->efi_parts[i].p_start);
	(void) fprintf(stderr,
	"It must be between %llu and %llu.\n",
	vtoc->efi_first_u_lba,
	vtoc->efi_last_u_lba);
	}
	if ((vtoc->efi_parts[i].p_start +
	vtoc->efi_parts[i].p_size <
	vtoc->efi_first_u_lba) \|\|
	(vtoc->efi_parts[i].p_start +
	vtoc->efi_parts[i].p_size >
	vtoc->efi_last_u_lba + 1)) {
	(void) fprintf(stderr,
	"Partition %d ends at %llu\n",
	i,
	vtoc->efi_parts[i].p_start +
	vtoc->efi_parts[i].p_size);
	(void) fprintf(stderr,
	"It must be between %llu and %llu.\n",
	vtoc->efi_first_u_lba,
	vtoc->efi_last_u_lba);
	}

	for (j = 0; j < vtoc->efi_nparts; j++) {
	isize = vtoc->efi_parts[i].p_size;
	jsize = vtoc->efi_parts[j].p_size;
	istart = vtoc->efi_parts[i].p_start;
	jstart = vtoc->efi_parts[j].p_start;
	if ((i != j) && (isize != 0) && (jsize != 0)) {
	endsect = jstart + jsize -1;
	if ((jstart <= istart) &&
	(istart <= endsect)) {
	if (!overlap) {
	(void) fprintf(stderr,
	"label error: EFI Labels do not "
	"support overlapping partitions\n");
	}
	(void) fprintf(stderr,
	"Partition %d overlaps partition "
	"%d.\n", i, j);
	overlap = 1;
	}
	}
	}
	}
	/* make sure there is a reserved partition */
	if (resv_part == -1) {
	(void) fprintf(stderr,
	"no reserved partition found\n");
	}
	}
	diff --git a/sys/contrib/openzfs/lib/libnvpair/libnvpair.abi b/sys/contrib/openzfs/lib/libnvpair/libnvpair.abi
	index ef92f3e9bda6..69009375e887 100644
	--- a/sys/contrib/openzfs/lib/libnvpair/libnvpair.abi
	+++ b/sys/contrib/openzfs/lib/libnvpair/libnvpair.abi
	@@ -1,3165 +1,3266 @@
	<abi-corpus version='2.0' architecture='elf-amd-x86_64' soname='libnvpair.so.3'>
	<elf-needed>
	<dependency name='libtirpc.so.3'/>
	<dependency name='libc.so.6'/>
	</elf-needed>
	<elf-function-symbols>
	<elf-symbol name='dump_nvlist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_boolean' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_boolean_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_boolean_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_byte' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_byte_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_int16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_int16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_int32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_int32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_int64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_int64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_int8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_int8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_nvlist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_nvlist_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_nvpair' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_string_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_uint16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_uint16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_uint32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_uint32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_uint64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_uint64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_uint8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_add_uint8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_alloc' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_dup' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_free' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_boolean' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_boolean_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_boolean_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_byte' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_byte_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_int16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_int16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_int32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_int32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_int64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_int64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_int8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_int8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_nvlist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_nvpair' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_uint16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_uint16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_uint32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_uint32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_uint64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_uint64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_uint8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_lookup_uint8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_merge' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_num_pairs' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_pack' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_pack_free' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_remove_nvpair' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_size' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvlist_unpack' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvpair_value_boolean_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvpair_value_byte' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvpair_value_int16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvpair_value_int32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvpair_value_int64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvpair_value_int8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvpair_value_nvlist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvpair_value_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvpair_value_uint16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvpair_value_uint32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvpair_value_uint64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fnvpair_value_uint8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libspl_assertf' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	+ <elf-symbol name='libspl_backtrace' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libspl_set_assert_ok' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nv_alloc_fini' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nv_alloc_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nv_alloc_reset' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_boolean' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_boolean_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_boolean_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_byte' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_byte_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_double' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_hrtime' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_int16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_int16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_int32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_int32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_int64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_int64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_int8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_int8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_nvlist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_nvlist_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_nvpair' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_string_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_uint16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_uint16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_uint32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_uint32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_uint64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_uint64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_uint8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_add_uint8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_alloc' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_dup' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_empty' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_exists' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_free' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_boolean' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_boolean_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_boolean_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_byte' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_byte_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_double' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_hrtime' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_int16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_int16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_int32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_int32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_int64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_int64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_int8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_int8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_nv_alloc' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_nvlist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_nvlist_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_nvpair' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_nvpair_embedded_index' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_pairs' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_string_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_uint16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_uint16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_uint32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_uint32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_uint64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_uint64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_uint8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_lookup_uint8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_merge' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_next_nvpair' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_nvflag' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_pack' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prev_nvpair' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_print' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_print_json' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prt' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctl_alloc' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctl_dofmt' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctl_doindent' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctl_free' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctl_getdest' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctl_setdest' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctl_setfmt' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctl_setindent' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_boolean' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_boolean_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_boolean_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_byte' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_byte_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_double' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_hrtime' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_int16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_int16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_int32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_int32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_int64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_int64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_int8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_int8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_nvlist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_nvlist_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_string_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_uint16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_uint16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_uint32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_uint32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_uint64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_uint64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_uint8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_prtctlop_uint8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_remove_all' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_remove_nvpair' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_size' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_unpack' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_xalloc' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_xdup' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_xpack' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvlist_xunpack' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_type_is_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_boolean_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_boolean_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_byte' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_byte_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_double' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_hrtime' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_int16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_int16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_int32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_int32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_int64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_int64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_int8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_int8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_match' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_match_regex' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_nvlist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_nvlist_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_string_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_uint16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_uint16_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_uint32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_uint32_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_uint64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_uint64_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_uint8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nvpair_value_uint8_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	</elf-function-symbols>
	<elf-variable-symbols>
	<elf-symbol name='nv_alloc_nosleep' size='8' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='nv_fixed_ops' size='8' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	</elf-variable-symbols>
	<abi-instr address-size='64' path='lib/libnvpair/libnvpair.c' language='LANG_C99'>
	<type-decl name='char' size-in-bits='8' id='a84c031d'/>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='8' id='89feb1ec'>
	<subrange length='1' type-id='7359adad' id='52f813b4'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='160' id='664ac0b7'>
	<subrange length='20' type-id='7359adad' id='fdca39cf'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='infinite' id='e84913bd'>
	<subrange length='infinite' type-id='7359adad' id='031f2035'/>
	</array-type-def>
	<class-decl name='_IO_codecvt' is-struct='yes' visibility='default' is-declaration-only='yes' id='a4036571'/>
	<class-decl name='_IO_marker' is-struct='yes' visibility='default' is-declaration-only='yes' id='010ae0b9'/>
	<class-decl name='_IO_wide_data' is-struct='yes' visibility='default' is-declaration-only='yes' id='79bd3751'/>
	<class-decl name='re_dfa_t' is-struct='yes' visibility='default' is-declaration-only='yes' id='b48d2441'/>
	<type-decl name='double' size-in-bits='64' id='a0eb0f08'/>
	<type-decl name='int' size-in-bits='32' id='95e97e5e'/>
	<type-decl name='long int' size-in-bits='64' id='bd54fe1a'/>
	<type-decl name='long long int' size-in-bits='64' id='1eb56b1e'/>
	<type-decl name='short int' size-in-bits='16' id='a2185560'/>
	<type-decl name='signed char' size-in-bits='8' id='28577a57'/>
	<type-decl name='unnamed-enum-underlying-type-32' is-anonymous='yes' size-in-bits='32' alignment-in-bits='32' id='9cac1fee'/>
	<type-decl name='unsigned char' size-in-bits='8' id='002ac4a6'/>
	<type-decl name='unsigned int' size-in-bits='32' id='f0981eeb'/>
	<type-decl name='unsigned long int' size-in-bits='64' id='7359adad'/>
	<type-decl name='unsigned short int' size-in-bits='16' id='8efea9e5'/>
	<type-decl name='variadic parameter type' id='2c1145c5'/>
	<type-decl name='void' id='48b5725f'/>
	<typedef-decl name='nvlist_prtctl_t' type-id='196db161' id='b0c1ff8d'/>
	<enum-decl name='nvlist_indent_mode' id='628aafab'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='NVLIST_INDENT_ABS' value='0'/>
	<enumerator name='NVLIST_INDENT_TABBED' value='1'/>
	</enum-decl>
	<enum-decl name='nvlist_prtctl_fmt' id='c8dcc53a'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='NVLIST_FMT_MEMBER_NAME' value='0'/>
	<enumerator name='NVLIST_FMT_MEMBER_POSTAMBLE' value='1'/>
	<enumerator name='NVLIST_FMT_BTWN_ARRAY' value='2'/>
	</enum-decl>
	<enum-decl name='data_type_t' naming-typedef-id='8d0687d2' id='aeeae136'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='DATA_TYPE_DONTCARE' value='-1'/>
	<enumerator name='DATA_TYPE_UNKNOWN' value='0'/>
	<enumerator name='DATA_TYPE_BOOLEAN' value='1'/>
	<enumerator name='DATA_TYPE_BYTE' value='2'/>
	<enumerator name='DATA_TYPE_INT16' value='3'/>
	<enumerator name='DATA_TYPE_UINT16' value='4'/>
	<enumerator name='DATA_TYPE_INT32' value='5'/>
	<enumerator name='DATA_TYPE_UINT32' value='6'/>
	<enumerator name='DATA_TYPE_INT64' value='7'/>
	<enumerator name='DATA_TYPE_UINT64' value='8'/>
	<enumerator name='DATA_TYPE_STRING' value='9'/>
	<enumerator name='DATA_TYPE_BYTE_ARRAY' value='10'/>
	<enumerator name='DATA_TYPE_INT16_ARRAY' value='11'/>
	<enumerator name='DATA_TYPE_UINT16_ARRAY' value='12'/>
	<enumerator name='DATA_TYPE_INT32_ARRAY' value='13'/>
	<enumerator name='DATA_TYPE_UINT32_ARRAY' value='14'/>
	<enumerator name='DATA_TYPE_INT64_ARRAY' value='15'/>
	<enumerator name='DATA_TYPE_UINT64_ARRAY' value='16'/>
	<enumerator name='DATA_TYPE_STRING_ARRAY' value='17'/>
	<enumerator name='DATA_TYPE_HRTIME' value='18'/>
	<enumerator name='DATA_TYPE_NVLIST' value='19'/>
	<enumerator name='DATA_TYPE_NVLIST_ARRAY' value='20'/>
	<enumerator name='DATA_TYPE_BOOLEAN_VALUE' value='21'/>
	<enumerator name='DATA_TYPE_INT8' value='22'/>
	<enumerator name='DATA_TYPE_UINT8' value='23'/>
	<enumerator name='DATA_TYPE_BOOLEAN_ARRAY' value='24'/>
	<enumerator name='DATA_TYPE_INT8_ARRAY' value='25'/>
	<enumerator name='DATA_TYPE_UINT8_ARRAY' value='26'/>
	<enumerator name='DATA_TYPE_DOUBLE' value='27'/>
	</enum-decl>
	<typedef-decl name='data_type_t' type-id='aeeae136' id='8d0687d2'/>
	<class-decl name='nvpair' size-in-bits='128' is-struct='yes' visibility='default' id='1c34e459'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='nvp_size' type-id='3ff5601b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='nvp_name_sz' type-id='23bd8cb5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='48'>
	<var-decl name='nvp_reserve' type-id='23bd8cb5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='nvp_value_elem' type-id='3ff5601b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='nvp_type' type-id='8d0687d2' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='nvp_name' type-id='e84913bd' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='nvpair_t' type-id='1c34e459' id='57928edf'/>
	<class-decl name='nvlist' size-in-bits='192' is-struct='yes' visibility='default' id='ac266fd9'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='nvl_version' type-id='3ff5601b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='nvl_nvflag' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='nvl_priv' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='nvl_flag' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='160'>
	<var-decl name='nvl_pad' type-id='3ff5601b' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='nvlist_t' type-id='ac266fd9' id='8e8d4be3'/>
	<enum-decl name='boolean_t' naming-typedef-id='c19b74c3' id='f58c8277'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='B_FALSE' value='0'/>
	<enumerator name='B_TRUE' value='1'/>
	</enum-decl>
	<typedef-decl name='boolean_t' type-id='f58c8277' id='c19b74c3'/>
	<typedef-decl name='uchar_t' type-id='002ac4a6' id='d8bf0010'/>
	<typedef-decl name='uint_t' type-id='f0981eeb' id='3502e3ff'/>
	<typedef-decl name='hrtime_t' type-id='1eb56b1e' id='cebdd548'/>
	<typedef-decl name='__re_long_size_t' type-id='7359adad' id='ba516949'/>
	<typedef-decl name='reg_syntax_t' type-id='7359adad' id='1b72c3b3'/>
	<class-decl name='re_pattern_buffer' size-in-bits='512' is-struct='yes' visibility='default' id='19fc9a8c'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='buffer' type-id='33976309' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='allocated' type-id='ba516949' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='used' type-id='ba516949' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='syntax' type-id='1b72c3b3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='fastmap' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='translate' type-id='cf536864' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='re_nsub' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='can_be_null' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='449'>
	<var-decl name='regs_allocated' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='451'>
	<var-decl name='fastmap_accurate' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='452'>
	<var-decl name='no_sub' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='453'>
	<var-decl name='not_bol' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='454'>
	<var-decl name='not_eol' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='455'>
	<var-decl name='newline_anchor' type-id='f0981eeb' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='regex_t' type-id='19fc9a8c' id='aca3bac8'/>
	<typedef-decl name='regoff_t' type-id='95e97e5e' id='54a2a2a8'/>
	<class-decl name='regmatch_t' size-in-bits='64' is-struct='yes' naming-typedef-id='1b941664' visibility='default' id='4f932615'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='rm_so' type-id='54a2a2a8' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='rm_eo' type-id='54a2a2a8' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='regmatch_t' type-id='4f932615' id='1b941664'/>
	<typedef-decl name='int8_t' type-id='2171a512' id='ee31ee44'/>
	<typedef-decl name='int16_t' type-id='03896e23' id='23bd8cb5'/>
	<typedef-decl name='int32_t' type-id='33f57a65' id='3ff5601b'/>
	<typedef-decl name='int64_t' type-id='0c9942d2' id='9da381c4'/>
	<typedef-decl name='uint8_t' type-id='c51d6389' id='b96825af'/>
	<typedef-decl name='uint16_t' type-id='253c2d2a' id='149c6638'/>
	<typedef-decl name='uint32_t' type-id='62f1140c' id='8f92235e'/>
	<typedef-decl name='uint64_t' type-id='8910171f' id='9c313c2d'/>
	<typedef-decl name='__int8_t' type-id='28577a57' id='2171a512'/>
	<typedef-decl name='__uint8_t' type-id='002ac4a6' id='c51d6389'/>
	<typedef-decl name='__int16_t' type-id='a2185560' id='03896e23'/>
	<typedef-decl name='__uint16_t' type-id='8efea9e5' id='253c2d2a'/>
	<typedef-decl name='__int32_t' type-id='95e97e5e' id='33f57a65'/>
	<typedef-decl name='__uint32_t' type-id='f0981eeb' id='62f1140c'/>
	<typedef-decl name='__int64_t' type-id='bd54fe1a' id='0c9942d2'/>
	<typedef-decl name='__uint64_t' type-id='7359adad' id='8910171f'/>
	<typedef-decl name='__off_t' type-id='bd54fe1a' id='79989e9c'/>
	<typedef-decl name='__off64_t' type-id='bd54fe1a' id='724e4de6'/>
	<typedef-decl name='FILE' type-id='ec1ed955' id='aa12d1ba'/>
	<typedef-decl name='_IO_lock_t' type-id='48b5725f' id='bb4788fa'/>
	<class-decl name='_IO_FILE' size-in-bits='1728' is-struct='yes' visibility='default' id='ec1ed955'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='_flags' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='_IO_read_ptr' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='_IO_read_end' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='_IO_read_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='_IO_write_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='_IO_write_ptr' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='_IO_write_end' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='_IO_buf_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='_IO_buf_end' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='576'>
	<var-decl name='_IO_save_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='640'>
	<var-decl name='_IO_backup_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='704'>
	<var-decl name='_IO_save_end' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='768'>
	<var-decl name='_markers' type-id='e4c6fa61' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='832'>
	<var-decl name='_chain' type-id='dca988a5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='896'>
	<var-decl name='_fileno' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='928'>
	<var-decl name='_flags2' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='960'>
	<var-decl name='_old_offset' type-id='79989e9c' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1024'>
	<var-decl name='_cur_column' type-id='8efea9e5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1040'>
	<var-decl name='_vtable_offset' type-id='28577a57' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1048'>
	<var-decl name='_shortbuf' type-id='89feb1ec' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1088'>
	<var-decl name='_lock' type-id='cecf4ea7' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1152'>
	<var-decl name='_offset' type-id='724e4de6' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1216'>
	<var-decl name='_codecvt' type-id='570f8c59' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1280'>
	<var-decl name='_wide_data' type-id='c65a1f29' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1344'>
	<var-decl name='_freeres_list' type-id='dca988a5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1408'>
	<var-decl name='_freeres_buf' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1472'>
	<var-decl name='__pad5' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1536'>
	<var-decl name='_mode' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1568'>
	<var-decl name='_unused2' type-id='664ac0b7' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='size_t' type-id='7359adad' id='b59d7dce'/>
	<class-decl name='nvlist_printops' size-in-bits='3456' is-struct='yes' visibility='default' id='ebc6735b'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='print_boolean' type-id='e7f43f72' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='print_boolean_value' type-id='e7f43f73' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='print_byte' type-id='e7f43f74' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='print_int8' type-id='e7f43f75' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='print_uint8' type-id='e7f43f76' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='640'>
	<var-decl name='print_int16' type-id='e7f43f77' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='768'>
	<var-decl name='print_uint16' type-id='e7f43f78' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='896'>
	<var-decl name='print_int32' type-id='e7f43f79' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1024'>
	<var-decl name='print_uint32' type-id='e7f43f7a' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1152'>
	<var-decl name='print_int64' type-id='e7f43f7b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1280'>
	<var-decl name='print_uint64' type-id='e7f43f7c' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1408'>
	<var-decl name='print_double' type-id='e7f43f7d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1536'>
	<var-decl name='print_string' type-id='e7f43f7e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1664'>
	<var-decl name='print_hrtime' type-id='e7f43f7f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1792'>
	<var-decl name='print_nvlist' type-id='e7f43f80' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1920'>
	<var-decl name='print_boolean_array' type-id='e7f43f81' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2048'>
	<var-decl name='print_byte_array' type-id='e7f43f82' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2176'>
	<var-decl name='print_int8_array' type-id='e7f43f83' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2304'>
	<var-decl name='print_uint8_array' type-id='e7f43f84' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2432'>
	<var-decl name='print_int16_array' type-id='e7f43f85' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2560'>
	<var-decl name='print_uint16_array' type-id='e7f43f86' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2688'>
	<var-decl name='print_int32_array' type-id='e7f43f87' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2816'>
	<var-decl name='print_uint32_array' type-id='e7f43f88' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2944'>
	<var-decl name='print_int64_array' type-id='e7f43f89' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='3072'>
	<var-decl name='print_uint64_array' type-id='e7f43f8a' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='3200'>
	<var-decl name='print_string_array' type-id='e7f43f8b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='3328'>
	<var-decl name='print_nvlist_array' type-id='e7f43f8c' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f72'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='6d994334' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__1' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f73'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='6a2f50c1' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__2' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f74'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='8a1fb33a' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__3' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f75'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='506696a8' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__4' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f76'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='39b623f9' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__5' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f77'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='ea6be4eb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__6' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f78'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='f10f1e84' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__7' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f79'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='1708018d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__8' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f7a'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='90174072' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__9' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f7b'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='d2af7f32' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__10' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f7c'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='0b22f759' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__11' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f7d'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='3be4d568' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__12' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f7e'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='bfc48fe4' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__13' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f7f'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='e1c54c3c' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__14' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f80'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='19ea27ae' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__15' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f81'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='7ef0e988' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__16' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f82'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='7391ed39' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__17' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f83'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='42257af5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__18' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f84'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='330cc0d0' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__19' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f85'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='506ab59a' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__20' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f86'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='ed6a3a3d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__21' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f87'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='750cc41c' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__22' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f88'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='292cdbcf' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__23' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f89'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='aaea91b5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__24' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f8a'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='7e85a9b6' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__25' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f8b'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='337f4a72' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__26' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f8c'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='op' type-id='2835af80' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='nvlist_prtctl' size-in-bits='576' is-struct='yes' visibility='default' id='d2e8bad9'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='nvprt_fp' type-id='822cd80b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='nvprt_indent_mode' type-id='628aafab' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='nvprt_indent' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='nvprt_indentinc' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='nvprt_nmfmt' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='nvprt_eomfmt' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='nvprt_btwnarrfmt' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='nvprt_btwnarrfmt_nl' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='nvprt_dfltops' type-id='7be54adb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='nvprt_custops' type-id='7be54adb' visibility='default'/>
	</data-member>
	</class-decl>
	<pointer-type-def type-id='aa12d1ba' size-in-bits='64' id='822cd80b'/>
	<qualified-type-def type-id='822cd80b' restrict='yes' id='e75a27e9'/>
	<pointer-type-def type-id='ec1ed955' size-in-bits='64' id='dca988a5'/>
	<pointer-type-def type-id='a4036571' size-in-bits='64' id='570f8c59'/>
	<pointer-type-def type-id='bb4788fa' size-in-bits='64' id='cecf4ea7'/>
	<pointer-type-def type-id='010ae0b9' size-in-bits='64' id='e4c6fa61'/>
	<pointer-type-def type-id='79bd3751' size-in-bits='64' id='c65a1f29'/>
	<pointer-type-def type-id='c19b74c3' size-in-bits='64' id='37e3bd22'/>
	<pointer-type-def type-id='37e3bd22' size-in-bits='64' id='03829398'/>
	<pointer-type-def type-id='a84c031d' size-in-bits='64' id='26a90f95'/>
	<qualified-type-def type-id='a84c031d' const='yes' id='9b45d938'/>
	<pointer-type-def type-id='9b45d938' size-in-bits='64' id='80f4b756'/>
	<qualified-type-def type-id='80f4b756' restrict='yes' id='9d26089a'/>
	<pointer-type-def type-id='80f4b756' size-in-bits='64' id='7d3cd834'/>
	<pointer-type-def type-id='7d3cd834' size-in-bits='64' id='44c8373a'/>
	<qualified-type-def type-id='57928edf' const='yes' id='642ee20f'/>
	<pointer-type-def type-id='642ee20f' size-in-bits='64' id='dace003f'/>
	<qualified-type-def type-id='aca3bac8' const='yes' id='2498fd78'/>
	<pointer-type-def type-id='2498fd78' size-in-bits='64' id='eed6c816'/>
	<qualified-type-def type-id='eed6c816' restrict='yes' id='a431a9da'/>
	<pointer-type-def type-id='a0eb0f08' size-in-bits='64' id='7408d286'/>
	<pointer-type-def type-id='cebdd548' size-in-bits='64' id='e379e62d'/>
	<pointer-type-def type-id='9f88f76e' size-in-bits='64' id='7ef0e988'/>
	<pointer-type-def type-id='889f3b42' size-in-bits='64' id='bfc48fe4'/>
	<pointer-type-def type-id='49055dd8' size-in-bits='64' id='337f4a72'/>
	<pointer-type-def type-id='70284cc6' size-in-bits='64' id='3be4d568'/>
	<pointer-type-def type-id='700c3bca' size-in-bits='64' id='6d994334'/>
	<pointer-type-def type-id='18ac1860' size-in-bits='64' id='506ab59a'/>
	<pointer-type-def type-id='328fee42' size-in-bits='64' id='750cc41c'/>
	<pointer-type-def type-id='7ba5cd31' size-in-bits='64' id='aaea91b5'/>
	<pointer-type-def type-id='a86d8029' size-in-bits='64' id='42257af5'/>
	<pointer-type-def type-id='0b4eb914' size-in-bits='64' id='19ea27ae'/>
	<pointer-type-def type-id='c6c8144e' size-in-bits='64' id='2835af80'/>
	<pointer-type-def type-id='20f7b475' size-in-bits='64' id='6a2f50c1'/>
	<pointer-type-def type-id='102ee17a' size-in-bits='64' id='e1c54c3c'/>
	<pointer-type-def type-id='49b69c77' size-in-bits='64' id='ea6be4eb'/>
	<pointer-type-def type-id='cb5d50f1' size-in-bits='64' id='1708018d'/>
	<pointer-type-def type-id='880d56b8' size-in-bits='64' id='d2af7f32'/>
	<pointer-type-def type-id='a739bfc6' size-in-bits='64' id='506696a8'/>
	<pointer-type-def type-id='234f35e8' size-in-bits='64' id='8a1fb33a'/>
	<pointer-type-def type-id='41f7168a' size-in-bits='64' id='f10f1e84'/>
	<pointer-type-def type-id='e8d6e508' size-in-bits='64' id='90174072'/>
	<pointer-type-def type-id='f3daafe5' size-in-bits='64' id='0b22f759'/>
	<pointer-type-def type-id='17ab04ad' size-in-bits='64' id='39b623f9'/>
	<pointer-type-def type-id='256cdd75' size-in-bits='64' id='7391ed39'/>
	<pointer-type-def type-id='cc10a041' size-in-bits='64' id='ed6a3a3d'/>
	<pointer-type-def type-id='9fd269d3' size-in-bits='64' id='292cdbcf'/>
	<pointer-type-def type-id='3bd73b0c' size-in-bits='64' id='7e85a9b6'/>
	<pointer-type-def type-id='0d445e26' size-in-bits='64' id='330cc0d0'/>
	<pointer-type-def type-id='e4b89f30' size-in-bits='64' id='ed8aa8ba'/>
	<pointer-type-def type-id='78e3bec8' size-in-bits='64' id='b38a1022'/>
	<pointer-type-def type-id='a5193bb2' size-in-bits='64' id='6263eda4'/>
	<pointer-type-def type-id='2783af3c' size-in-bits='64' id='e44553b6'/>
	<pointer-type-def type-id='33c6e3d8' size-in-bits='64' id='1263777a'/>
	<pointer-type-def type-id='dadb9eca' size-in-bits='64' id='cbda43ac'/>
	<pointer-type-def type-id='55b9e070' size-in-bits='64' id='b3fae562'/>
	<pointer-type-def type-id='8e63c78b' size-in-bits='64' id='8b41e457'/>
	<pointer-type-def type-id='c542ed33' size-in-bits='64' id='f9668a57'/>
	<pointer-type-def type-id='5dea179a' size-in-bits='64' id='001d8764'/>
	<pointer-type-def type-id='b6f659a0' size-in-bits='64' id='44f188f2'/>
	<pointer-type-def type-id='2765bd17' size-in-bits='64' id='976f721b'/>
	<pointer-type-def type-id='9e073b5c' size-in-bits='64' id='ee62ad8e'/>
	<pointer-type-def type-id='2c785071' size-in-bits='64' id='957d9f35'/>
	<pointer-type-def type-id='aad19bf7' size-in-bits='64' id='4db8acf3'/>
	<pointer-type-def type-id='0660e71a' size-in-bits='64' id='0ca7b13c'/>
	<pointer-type-def type-id='250287b8' size-in-bits='64' id='a91bad5a'/>
	<pointer-type-def type-id='e7344862' size-in-bits='64' id='519bf35c'/>
	<pointer-type-def type-id='32b6d968' size-in-bits='64' id='92988dea'/>
	<pointer-type-def type-id='5c975642' size-in-bits='64' id='7f8ee7e4'/>
	<pointer-type-def type-id='0155b993' size-in-bits='64' id='2c8c4457'/>
	<pointer-type-def type-id='6e8b02cb' size-in-bits='64' id='eb944897'/>
	<pointer-type-def type-id='d434b7d7' size-in-bits='64' id='108e6453'/>
	<pointer-type-def type-id='c645e10f' size-in-bits='64' id='5cbe16ab'/>
	<pointer-type-def type-id='de41f295' size-in-bits='64' id='d94cdfa1'/>
	<pointer-type-def type-id='b2fbf64a' size-in-bits='64' id='470a7fd4'/>
	<pointer-type-def type-id='cc22d314' size-in-bits='64' id='eddda806'/>
	<pointer-type-def type-id='23bd8cb5' size-in-bits='64' id='f76f73d0'/>
	<pointer-type-def type-id='f76f73d0' size-in-bits='64' id='7e73928e'/>
	<pointer-type-def type-id='3ff5601b' size-in-bits='64' id='4aafb922'/>
	<pointer-type-def type-id='4aafb922' size-in-bits='64' id='9aa04798'/>
	<pointer-type-def type-id='9da381c4' size-in-bits='64' id='cb785ebf'/>
	<pointer-type-def type-id='cb785ebf' size-in-bits='64' id='e37ce48f'/>
	<pointer-type-def type-id='ee31ee44' size-in-bits='64' id='256d5229'/>
	<pointer-type-def type-id='256d5229' size-in-bits='64' id='ee181ab9'/>
	<pointer-type-def type-id='ebc6735b' size-in-bits='64' id='7be54adb'/>
	<pointer-type-def type-id='d2e8bad9' size-in-bits='64' id='196db161'/>
	<pointer-type-def type-id='8e8d4be3' size-in-bits='64' id='5ce45b60'/>
	<pointer-type-def type-id='5ce45b60' size-in-bits='64' id='857bb57e'/>
	<pointer-type-def type-id='857bb57e' size-in-bits='64' id='75be733c'/>
	<pointer-type-def type-id='57928edf' size-in-bits='64' id='3fa542f0'/>
	<pointer-type-def type-id='b48d2441' size-in-bits='64' id='33976309'/>
	<pointer-type-def type-id='aca3bac8' size-in-bits='64' id='d33f11cb'/>
	<pointer-type-def type-id='1b941664' size-in-bits='64' id='7e2979d5'/>
	<qualified-type-def type-id='7e2979d5' restrict='yes' id='fc212857'/>
	<pointer-type-def type-id='d8bf0010' size-in-bits='64' id='45b65157'/>
	<pointer-type-def type-id='45b65157' size-in-bits='64' id='3b0247c7'/>
	<pointer-type-def type-id='149c6638' size-in-bits='64' id='8a121f49'/>
	<pointer-type-def type-id='8a121f49' size-in-bits='64' id='bd8768d9'/>
	<pointer-type-def type-id='8f92235e' size-in-bits='64' id='90421557'/>
	<pointer-type-def type-id='90421557' size-in-bits='64' id='9507d3c7'/>
	<pointer-type-def type-id='9c313c2d' size-in-bits='64' id='5d6479ae'/>
	<pointer-type-def type-id='5d6479ae' size-in-bits='64' id='892b4acc'/>
	<pointer-type-def type-id='b96825af' size-in-bits='64' id='ae3e8ca6'/>
	<pointer-type-def type-id='ae3e8ca6' size-in-bits='64' id='d8774064'/>
	<pointer-type-def type-id='3502e3ff' size-in-bits='64' id='4dd26a40'/>
	<pointer-type-def type-id='002ac4a6' size-in-bits='64' id='cf536864'/>
	<pointer-type-def type-id='48b5725f' size-in-bits='64' id='eaa32e2f'/>
	<class-decl name='_IO_codecvt' is-struct='yes' visibility='default' is-declaration-only='yes' id='a4036571'/>
	<class-decl name='_IO_marker' is-struct='yes' visibility='default' is-declaration-only='yes' id='010ae0b9'/>
	<class-decl name='_IO_wide_data' is-struct='yes' visibility='default' is-declaration-only='yes' id='79bd3751'/>
	<class-decl name='re_dfa_t' is-struct='yes' visibility='default' is-declaration-only='yes' id='b48d2441'/>
	<function-decl name='nvlist_next_nvpair' mangled-name='nvlist_next_nvpair' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_next_nvpair'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='dace003f'/>
	<return type-id='3fa542f0'/>
	</function-decl>
	<function-decl name='nvpair_name' mangled-name='nvpair_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_name'>
	<parameter type-id='dace003f'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='nvpair_type' mangled-name='nvpair_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_type'>
	<parameter type-id='dace003f'/>
	<return type-id='8d0687d2'/>
	</function-decl>
	<function-decl name='nvpair_type_is_array' mangled-name='nvpair_type_is_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_type_is_array'>
	<parameter type-id='dace003f'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_boolean_value' mangled-name='nvpair_value_boolean_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_boolean_value'>
	<parameter type-id='dace003f'/>
	<parameter type-id='37e3bd22'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_byte' mangled-name='nvpair_value_byte' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_byte'>
	<parameter type-id='dace003f'/>
	<parameter type-id='45b65157'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_int8' mangled-name='nvpair_value_int8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_int8'>
	<parameter type-id='dace003f'/>
	<parameter type-id='256d5229'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_uint8' mangled-name='nvpair_value_uint8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_uint8'>
	<parameter type-id='dace003f'/>
	<parameter type-id='ae3e8ca6'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_int16' mangled-name='nvpair_value_int16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_int16'>
	<parameter type-id='dace003f'/>
	<parameter type-id='f76f73d0'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_uint16' mangled-name='nvpair_value_uint16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_uint16'>
	<parameter type-id='dace003f'/>
	<parameter type-id='8a121f49'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_int32' mangled-name='nvpair_value_int32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_int32'>
	<parameter type-id='dace003f'/>
	<parameter type-id='4aafb922'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_uint32' mangled-name='nvpair_value_uint32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_uint32'>
	<parameter type-id='dace003f'/>
	<parameter type-id='90421557'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_int64' mangled-name='nvpair_value_int64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_int64'>
	<parameter type-id='dace003f'/>
	<parameter type-id='cb785ebf'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_uint64' mangled-name='nvpair_value_uint64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_uint64'>
	<parameter type-id='dace003f'/>
	<parameter type-id='5d6479ae'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_string' mangled-name='nvpair_value_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_string'>
	<parameter type-id='dace003f'/>
	<parameter type-id='7d3cd834'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_nvlist' mangled-name='nvpair_value_nvlist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_nvlist'>
	<parameter type-id='3fa542f0'/>
	<parameter type-id='857bb57e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_boolean_array' mangled-name='nvpair_value_boolean_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_boolean_array'>
	<parameter type-id='3fa542f0'/>
	<parameter type-id='03829398'/>
	<parameter type-id='4dd26a40'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_byte_array' mangled-name='nvpair_value_byte_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_byte_array'>
	<parameter type-id='3fa542f0'/>
	<parameter type-id='3b0247c7'/>
	<parameter type-id='4dd26a40'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_int8_array' mangled-name='nvpair_value_int8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_int8_array'>
	<parameter type-id='3fa542f0'/>
	<parameter type-id='ee181ab9'/>
	<parameter type-id='4dd26a40'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_uint8_array' mangled-name='nvpair_value_uint8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_uint8_array'>
	<parameter type-id='3fa542f0'/>
	<parameter type-id='d8774064'/>
	<parameter type-id='4dd26a40'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_int16_array' mangled-name='nvpair_value_int16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_int16_array'>
	<parameter type-id='3fa542f0'/>
	<parameter type-id='7e73928e'/>
	<parameter type-id='4dd26a40'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_uint16_array' mangled-name='nvpair_value_uint16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_uint16_array'>
	<parameter type-id='3fa542f0'/>
	<parameter type-id='bd8768d9'/>
	<parameter type-id='4dd26a40'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_int32_array' mangled-name='nvpair_value_int32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_int32_array'>
	<parameter type-id='3fa542f0'/>
	<parameter type-id='9aa04798'/>
	<parameter type-id='4dd26a40'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_uint32_array' mangled-name='nvpair_value_uint32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_uint32_array'>
	<parameter type-id='3fa542f0'/>
	<parameter type-id='9507d3c7'/>
	<parameter type-id='4dd26a40'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_int64_array' mangled-name='nvpair_value_int64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_int64_array'>
	<parameter type-id='3fa542f0'/>
	<parameter type-id='e37ce48f'/>
	<parameter type-id='4dd26a40'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_uint64_array' mangled-name='nvpair_value_uint64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_uint64_array'>
	<parameter type-id='3fa542f0'/>
	<parameter type-id='892b4acc'/>
	<parameter type-id='4dd26a40'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_string_array' mangled-name='nvpair_value_string_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_string_array'>
	<parameter type-id='3fa542f0'/>
	<parameter type-id='44c8373a'/>
	<parameter type-id='4dd26a40'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_nvlist_array' mangled-name='nvpair_value_nvlist_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_nvlist_array'>
	<parameter type-id='3fa542f0'/>
	<parameter type-id='75be733c'/>
	<parameter type-id='4dd26a40'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_hrtime' mangled-name='nvpair_value_hrtime' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_hrtime'>
	<parameter type-id='3fa542f0'/>
	<parameter type-id='e379e62d'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_double' mangled-name='nvpair_value_double' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_double'>
	<parameter type-id='dace003f'/>
	<parameter type-id='7408d286'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='dcgettext' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='regexec' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='a431a9da'/>
	<parameter type-id='9d26089a'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='fc212857'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='malloc' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b59d7dce'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='calloc' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='free' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='eaa32e2f'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='strcmp' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	+ <function-decl name='strchr' visibility='default' binding='global' size-in-bits='64'>
	+ <parameter type-id='80f4b756'/>
	+ <parameter type-id='95e97e5e'/>
	+ <return type-id='26a90f95'/>
	+ </function-decl>
	<function-decl name='strcspn' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='strspn' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='__fprintf_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='e75a27e9'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='9d26089a'/>
	<parameter is-variadic='yes'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='__printf_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<parameter is-variadic='yes'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_prtctl_setdest' mangled-name='nvlist_prtctl_setdest' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctl_setdest'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='822cd80b' name='fp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctl_getdest' mangled-name='nvlist_prtctl_getdest' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctl_getdest'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<return type-id='822cd80b'/>
	</function-decl>
	<function-decl name='nvlist_prtctl_setindent' mangled-name='nvlist_prtctl_setindent' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctl_setindent'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='628aafab' name='mode'/>
	<parameter type-id='95e97e5e' name='start'/>
	<parameter type-id='95e97e5e' name='inc'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctl_doindent' mangled-name='nvlist_prtctl_doindent' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctl_doindent'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='95e97e5e' name='onemore'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctl_setfmt' mangled-name='nvlist_prtctl_setfmt' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctl_setfmt'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='c8dcc53a' name='which'/>
	<parameter type-id='80f4b756' name='fmt'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctl_dofmt' mangled-name='nvlist_prtctl_dofmt' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctl_dofmt'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='c8dcc53a' name='which'/>
	<parameter is-variadic='yes'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_boolean' mangled-name='nvlist_prtctlop_boolean' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_boolean'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='1263777a' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_boolean_value' mangled-name='nvlist_prtctlop_boolean_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_boolean_value'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='976f721b' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_byte' mangled-name='nvlist_prtctlop_byte' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_byte'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='519bf35c' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_int8' mangled-name='nvlist_prtctlop_int8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_int8'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='a91bad5a' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_uint8' mangled-name='nvlist_prtctlop_uint8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_uint8'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='eb944897' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_int16' mangled-name='nvlist_prtctlop_int16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_int16'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='957d9f35' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_uint16' mangled-name='nvlist_prtctlop_uint16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_uint16'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='92988dea' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_int32' mangled-name='nvlist_prtctlop_int32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_int32'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='4db8acf3' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_uint32' mangled-name='nvlist_prtctlop_uint32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_uint32'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='7f8ee7e4' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_int64' mangled-name='nvlist_prtctlop_int64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_int64'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='0ca7b13c' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_uint64' mangled-name='nvlist_prtctlop_uint64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_uint64'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='2c8c4457' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_double' mangled-name='nvlist_prtctlop_double' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_double'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='e44553b6' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_string' mangled-name='nvlist_prtctlop_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_string'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='b38a1022' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_hrtime' mangled-name='nvlist_prtctlop_hrtime' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_hrtime'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='ee62ad8e' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_nvlist' mangled-name='nvlist_prtctlop_nvlist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_nvlist'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='001d8764' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_boolean_array' mangled-name='nvlist_prtctlop_boolean_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_boolean_array'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='ed8aa8ba' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_byte_array' mangled-name='nvlist_prtctlop_byte_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_byte_array'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='108e6453' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_int8_array' mangled-name='nvlist_prtctlop_int8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_int8_array'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='f9668a57' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_uint8_array' mangled-name='nvlist_prtctlop_uint8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_uint8_array'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='eddda806' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_int16_array' mangled-name='nvlist_prtctlop_int16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_int16_array'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='cbda43ac' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_uint16_array' mangled-name='nvlist_prtctlop_uint16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_uint16_array'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='5cbe16ab' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_int32_array' mangled-name='nvlist_prtctlop_int32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_int32_array'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='b3fae562' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_uint32_array' mangled-name='nvlist_prtctlop_uint32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_uint32_array'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='d94cdfa1' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_int64_array' mangled-name='nvlist_prtctlop_int64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_int64_array'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='8b41e457' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_uint64_array' mangled-name='nvlist_prtctlop_uint64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_uint64_array'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='470a7fd4' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_string_array' mangled-name='nvlist_prtctlop_string_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_string_array'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='6263eda4' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctlop_nvlist_array' mangled-name='nvlist_prtctlop_nvlist_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctlop_nvlist_array'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<parameter type-id='44f188f2' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prtctl_alloc' mangled-name='nvlist_prtctl_alloc' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctl_alloc'>
	<return type-id='b0c1ff8d'/>
	</function-decl>
	<function-decl name='nvlist_prtctl_free' mangled-name='nvlist_prtctl_free' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prtctl_free'>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_print' mangled-name='nvlist_print' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_print'>
	<parameter type-id='822cd80b' name='fp'/>
	<parameter type-id='5ce45b60' name='nvl'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_prt' mangled-name='nvlist_prt' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prt'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='b0c1ff8d' name='pctl'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='dump_nvlist' mangled-name='dump_nvlist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='dump_nvlist'>
	<parameter type-id='5ce45b60' name='list'/>
	<parameter type-id='95e97e5e' name='indent'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvpair_value_match_regex' mangled-name='nvpair_value_match_regex' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_match_regex'>
	<parameter type-id='3fa542f0' name='nvp'/>
	<parameter type-id='95e97e5e' name='ai'/>
	<parameter type-id='80f4b756' name='value'/>
	<parameter type-id='d33f11cb' name='value_regex'/>
	<parameter type-id='7d3cd834' name='ep'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_match' mangled-name='nvpair_value_match' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvpair_value_match'>
	<parameter type-id='3fa542f0' name='nvp'/>
	<parameter type-id='95e97e5e' name='ai'/>
	<parameter type-id='80f4b756' name='value'/>
	<parameter type-id='7d3cd834' name='ep'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-type size-in-bits='64' id='9f88f76e'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='37e3bd22'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='889f3b42'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='49055dd8'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='7d3cd834'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='70284cc6'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='a0eb0f08'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='700c3bca'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='18ac1860'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='f76f73d0'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='328fee42'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='4aafb922'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='7ba5cd31'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='cb785ebf'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='a86d8029'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='256d5229'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='0b4eb914'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='5ce45b60'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='c6c8144e'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='857bb57e'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='20f7b475'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='c19b74c3'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='102ee17a'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='cebdd548'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='49b69c77'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='23bd8cb5'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='cb5d50f1'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='3ff5601b'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='880d56b8'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='9da381c4'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='a739bfc6'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='ee31ee44'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='234f35e8'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='d8bf0010'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='41f7168a'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='149c6638'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='e8d6e508'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='8f92235e'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='f3daafe5'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='9c313c2d'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='17ab04ad'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='b96825af'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='256cdd75'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='45b65157'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='cc10a041'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='8a121f49'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='9fd269d3'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='90421557'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='3bd73b0c'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='5d6479ae'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='0d445e26'>
	<parameter type-id='196db161'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='ae3e8ca6'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='e4b89f30'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='37e3bd22'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='78e3bec8'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='a5193bb2'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='7d3cd834'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='2783af3c'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='a0eb0f08'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='33c6e3d8'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='dadb9eca'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='f76f73d0'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='55b9e070'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='4aafb922'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='8e63c78b'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='cb785ebf'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='c542ed33'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='256d5229'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='5dea179a'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='5ce45b60'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='b6f659a0'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='857bb57e'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='2765bd17'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='c19b74c3'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='9e073b5c'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='cebdd548'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='2c785071'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='23bd8cb5'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='aad19bf7'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='3ff5601b'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='0660e71a'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='9da381c4'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='250287b8'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='ee31ee44'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='e7344862'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='d8bf0010'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='32b6d968'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='149c6638'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='5c975642'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='8f92235e'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='0155b993'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='9c313c2d'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='6e8b02cb'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='b96825af'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='d434b7d7'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='45b65157'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='c645e10f'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='8a121f49'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='de41f295'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='90421557'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='b2fbf64a'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='5d6479ae'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='cc22d314'>
	<parameter type-id='b0c1ff8d'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='ae3e8ca6'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libnvpair/libnvpair_json.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='32' id='8e0573fd'>
	<subrange length='4' type-id='7359adad' id='16fe7105'/>
	</array-type-def>
	<class-decl name='__mbstate_t' size-in-bits='64' is-struct='yes' naming-typedef-id='55e5b2b5' visibility='default' id='87447a2e'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='__count' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='__value' type-id='ac5ab595' visibility='default'/>
	</data-member>
	</class-decl>
	<union-decl name='__anonymous_union__' size-in-bits='32' is-anonymous='yes' visibility='default' id='ac5ab595'>
	<data-member access='public'>
	<var-decl name='__wch' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='__wchb' type-id='8e0573fd' visibility='default'/>
	</data-member>
	</union-decl>
	<typedef-decl name='__mbstate_t' type-id='87447a2e' id='55e5b2b5'/>
	<typedef-decl name='mbstate_t' type-id='55e5b2b5' id='3d7d8cbf'/>
	<typedef-decl name='wchar_t' type-id='95e97e5e' id='928221d2'/>
	<pointer-type-def type-id='3d7d8cbf' size-in-bits='64' id='a68021ce'/>
	<qualified-type-def type-id='a68021ce' restrict='yes' id='03aaab72'/>
	<pointer-type-def type-id='928221d2' size-in-bits='64' id='323d93c1'/>
	<qualified-type-def type-id='323d93c1' restrict='yes' id='f1358bc3'/>
	<function-decl name='fnvpair_value_boolean_value' mangled-name='fnvpair_value_boolean_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_boolean_value'>
	<parameter type-id='dace003f'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='fnvpair_value_byte' mangled-name='fnvpair_value_byte' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_byte'>
	<parameter type-id='dace003f'/>
	<return type-id='d8bf0010'/>
	</function-decl>
	<function-decl name='fnvpair_value_int8' mangled-name='fnvpair_value_int8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_int8'>
	<parameter type-id='dace003f'/>
	<return type-id='ee31ee44'/>
	</function-decl>
	<function-decl name='fnvpair_value_int16' mangled-name='fnvpair_value_int16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_int16'>
	<parameter type-id='dace003f'/>
	<return type-id='23bd8cb5'/>
	</function-decl>
	<function-decl name='fnvpair_value_int32' mangled-name='fnvpair_value_int32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_int32'>
	<parameter type-id='dace003f'/>
	<return type-id='3ff5601b'/>
	</function-decl>
	<function-decl name='fnvpair_value_int64' mangled-name='fnvpair_value_int64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_int64'>
	<parameter type-id='dace003f'/>
	<return type-id='9da381c4'/>
	</function-decl>
	<function-decl name='fnvpair_value_uint8' mangled-name='fnvpair_value_uint8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_uint8'>
	<parameter type-id='dace003f'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='fnvpair_value_uint16' mangled-name='fnvpair_value_uint16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_uint16'>
	<parameter type-id='dace003f'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='fnvpair_value_uint32' mangled-name='fnvpair_value_uint32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_uint32'>
	<parameter type-id='dace003f'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='fnvpair_value_uint64' mangled-name='fnvpair_value_uint64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_uint64'>
	<parameter type-id='dace003f'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='fnvpair_value_string' mangled-name='fnvpair_value_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_string'>
	<parameter type-id='dace003f'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='fnvpair_value_nvlist' mangled-name='fnvpair_value_nvlist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvpair_value_nvlist'>
	<parameter type-id='3fa542f0'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='libspl_assertf' mangled-name='libspl_assertf' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libspl_assertf'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<parameter is-variadic='yes'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='__ctype_get_mb_cur_max' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='mbrtowc' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='f1358bc3'/>
	<parameter type-id='9d26089a'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='03aaab72'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='nvlist_print_json' mangled-name='nvlist_print_json' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_print_json'>
	<parameter type-id='822cd80b' name='fp'/>
	<parameter type-id='5ce45b60' name='nvl'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libnvpair/nvpair_alloc_system.c' language='LANG_C99'>
	<class-decl name='__va_list_tag' size-in-bits='192' is-struct='yes' visibility='default' id='d5027220'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='gp_offset' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='fp_offset' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='overflow_arg_area' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='reg_save_area' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='nv_alloc_ops_t' type-id='8f6cc4f4' id='03e8ffd6'/>
	<class-decl name='nv_alloc' size-in-bits='128' is-struct='yes' visibility='default' id='98213087'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='nva_ops' type-id='ee1d4944' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='nva_arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='nv_alloc_t' type-id='98213087' id='cca08635'/>
	<class-decl name='nv_alloc_ops' size-in-bits='320' is-struct='yes' visibility='default' id='8f6cc4f4'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='nv_ao_init' type-id='76da8447' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='nv_ao_fini' type-id='fe356f6f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='nv_ao_alloc' type-id='9ff7f508' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='nv_ao_free' type-id='520da3f4' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='nv_ao_reset' type-id='fe356f6f' visibility='default'/>
	</data-member>
	</class-decl>
	<pointer-type-def type-id='d5027220' size-in-bits='64' id='b7f2d5e6'/>
	<qualified-type-def type-id='03e8ffd6' const='yes' id='aca16c06'/>
	<pointer-type-def type-id='aca16c06' size-in-bits='64' id='ee1d4944'/>
	<pointer-type-def type-id='e9ff7293' size-in-bits='64' id='76da8447'/>
	<pointer-type-def type-id='cca08635' size-in-bits='64' id='11871392'/>
	<qualified-type-def type-id='11871392' const='yes' id='e1cbfe1d'/>
	<pointer-type-def type-id='51a21b4b' size-in-bits='64' id='fe356f6f'/>
	<pointer-type-def type-id='1169c032' size-in-bits='64' id='520da3f4'/>
	<pointer-type-def type-id='9fff962e' size-in-bits='64' id='9ff7f508'/>
	<var-decl name='nv_alloc_nosleep' type-id='e1cbfe1d' mangled-name='nv_alloc_nosleep' visibility='default' elf-symbol-id='nv_alloc_nosleep'/>
	<function-type size-in-bits='64' id='e9ff7293'>
	<parameter type-id='11871392'/>
	<parameter type-id='b7f2d5e6'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='51a21b4b'>
	<parameter type-id='11871392'/>
	<return type-id='48b5725f'/>
	</function-type>
	<function-type size-in-bits='64' id='1169c032'>
	<parameter type-id='11871392'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='48b5725f'/>
	</function-type>
	<function-type size-in-bits='64' id='9fff962e'>
	<parameter type-id='11871392'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='eaa32e2f'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/assert.c' language='LANG_C99'>
	+ <array-type-def dimensions='1' type-id='a84c031d' size-in-bits='320' id='36c46961'>
	+ <subrange length='40' type-id='7359adad' id='8f80b239'/>
	+ </array-type-def>
	+ <union-decl name='pthread_mutex_t' size-in-bits='320' naming-typedef-id='7a6844eb' visibility='default' id='70681f9b'>
	+ <data-member access='public'>
	+ <var-decl name='__data' type-id='4c734837' visibility='default'/>
	+ </data-member>
	+ <data-member access='public'>
	+ <var-decl name='__size' type-id='36c46961' visibility='default'/>
	+ </data-member>
	+ <data-member access='public'>
	+ <var-decl name='__align' type-id='bd54fe1a' visibility='default'/>
	+ </data-member>
	+ </union-decl>
	+ <typedef-decl name='pthread_mutex_t' type-id='70681f9b' id='7a6844eb'/>
	+ <class-decl name='__pthread_mutex_s' size-in-bits='320' is-struct='yes' visibility='default' id='4c734837'>
	+ <data-member access='public' layout-offset-in-bits='0'>
	+ <var-decl name='__lock' type-id='95e97e5e' visibility='default'/>
	+ </data-member>
	+ <data-member access='public' layout-offset-in-bits='32'>
	+ <var-decl name='__count' type-id='f0981eeb' visibility='default'/>
	+ </data-member>
	+ <data-member access='public' layout-offset-in-bits='64'>
	+ <var-decl name='__owner' type-id='95e97e5e' visibility='default'/>
	+ </data-member>
	+ <data-member access='public' layout-offset-in-bits='96'>
	+ <var-decl name='__nusers' type-id='f0981eeb' visibility='default'/>
	+ </data-member>
	+ <data-member access='public' layout-offset-in-bits='128'>
	+ <var-decl name='__kind' type-id='95e97e5e' visibility='default'/>
	+ </data-member>
	+ <data-member access='public' layout-offset-in-bits='160'>
	+ <var-decl name='__spins' type-id='a2185560' visibility='default'/>
	+ </data-member>
	+ <data-member access='public' layout-offset-in-bits='176'>
	+ <var-decl name='__elision' type-id='a2185560' visibility='default'/>
	+ </data-member>
	+ <data-member access='public' layout-offset-in-bits='192'>
	+ <var-decl name='__list' type-id='518fb49c' visibility='default'/>
	+ </data-member>
	+ </class-decl>
	+ <class-decl name='__pthread_internal_list' size-in-bits='128' is-struct='yes' visibility='default' id='0e01899c'>
	+ <data-member access='public' layout-offset-in-bits='0'>
	+ <var-decl name='__prev' type-id='4d98cd5a' visibility='default'/>
	+ </data-member>
	+ <data-member access='public' layout-offset-in-bits='64'>
	+ <var-decl name='__next' type-id='4d98cd5a' visibility='default'/>
	+ </data-member>
	+ </class-decl>
	+ <typedef-decl name='__pthread_list_t' type-id='0e01899c' id='518fb49c'/>
	+ <typedef-decl name='__pid_t' type-id='95e97e5e' id='3629bad8'/>
	+ <pointer-type-def type-id='0e01899c' size-in-bits='64' id='4d98cd5a'/>
	+ <pointer-type-def type-id='7a6844eb' size-in-bits='64' id='18c91f9e'/>
	+ <function-decl name='libspl_backtrace' mangled-name='libspl_backtrace' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libspl_backtrace'>
	+ <parameter type-id='95e97e5e'/>
	+ <return type-id='48b5725f'/>
	+ </function-decl>
	+ <function-decl name='pthread_mutex_lock' visibility='default' binding='global' size-in-bits='64'>
	+ <parameter type-id='18c91f9e'/>
	+ <return type-id='95e97e5e'/>
	+ </function-decl>
	+ <function-decl name='pthread_mutex_unlock' visibility='default' binding='global' size-in-bits='64'>
	+ <parameter type-id='18c91f9e'/>
	+ <return type-id='95e97e5e'/>
	+ </function-decl>
	<function-decl name='abort' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='48b5725f'/>
	</function-decl>
	+ <function-decl name='getpid' visibility='default' binding='global' size-in-bits='64'>
	+ <return type-id='3629bad8'/>
	+ </function-decl>
	<function-decl name='__vfprintf_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='e75a27e9'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='9d26089a'/>
	<parameter type-id='b7f2d5e6'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	+ <function-decl name='gettid' visibility='default' binding='global' size-in-bits='64'>
	+ <return type-id='3629bad8'/>
	+ </function-decl>
	+ <function-decl name='prctl' visibility='default' binding='global' size-in-bits='64'>
	+ <parameter type-id='95e97e5e'/>
	+ <parameter is-variadic='yes'/>
	+ <return type-id='95e97e5e'/>
	+ </function-decl>
	<function-decl name='libspl_set_assert_ok' mangled-name='libspl_set_assert_ok' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libspl_set_assert_ok'>
	<parameter type-id='c19b74c3' name='val'/>
	<return type-id='48b5725f'/>
	</function-decl>
	</abi-instr>
	+ <abi-instr address-size='64' path='lib/libspl/backtrace.c' language='LANG_C99'>
	+ <typedef-decl name='__ssize_t' type-id='bd54fe1a' id='41060289'/>
	+ <typedef-decl name='ssize_t' type-id='41060289' id='79a0948f'/>
	+ <qualified-type-def type-id='eaa32e2f' const='yes' id='83be723c'/>
	+ <pointer-type-def type-id='83be723c' size-in-bits='64' id='7acd98a2'/>
	+ <pointer-type-def type-id='eaa32e2f' size-in-bits='64' id='63e171df'/>
	+ <function-decl name='backtrace' visibility='default' binding='global' size-in-bits='64'>
	+ <parameter type-id='63e171df'/>
	+ <parameter type-id='95e97e5e'/>
	+ <return type-id='95e97e5e'/>
	+ </function-decl>
	+ <function-decl name='backtrace_symbols_fd' visibility='default' binding='global' size-in-bits='64'>
	+ <parameter type-id='7acd98a2'/>
	+ <parameter type-id='95e97e5e'/>
	+ <parameter type-id='95e97e5e'/>
	+ <return type-id='48b5725f'/>
	+ </function-decl>
	+ <function-decl name='write' visibility='default' binding='global' size-in-bits='64'>
	+ <parameter type-id='95e97e5e'/>
	+ <parameter type-id='eaa32e2f'/>
	+ <parameter type-id='b59d7dce'/>
	+ <return type-id='79a0948f'/>
	+ </function-decl>
	+ </abi-instr>
	<abi-instr address-size='64' path='module/nvpair/fnvpair.c' language='LANG_C99'>
	<function-decl name='fnvlist_alloc' mangled-name='fnvlist_alloc' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_alloc'>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='fnvlist_free' mangled-name='fnvlist_free' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_free'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_size' mangled-name='fnvlist_size' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_size'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='fnvlist_pack' mangled-name='fnvlist_pack' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_pack'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='78c01427' name='sizep'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='fnvlist_pack_free' mangled-name='fnvlist_pack_free' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_pack_free'>
	<parameter type-id='26a90f95' name='pack'/>
	<parameter type-id='b59d7dce' name='size'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_unpack' mangled-name='fnvlist_unpack' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_unpack'>
	<parameter type-id='26a90f95' name='buf'/>
	<parameter type-id='b59d7dce' name='buflen'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='fnvlist_dup' mangled-name='fnvlist_dup' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_dup'>
	<parameter type-id='22cce67b' name='nvl'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='fnvlist_merge' mangled-name='fnvlist_merge' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_merge'>
	<parameter type-id='5ce45b60' name='dst'/>
	<parameter type-id='5ce45b60' name='src'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_num_pairs' mangled-name='fnvlist_num_pairs' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_num_pairs'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='fnvlist_add_boolean' mangled-name='fnvlist_add_boolean' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_boolean'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_boolean_value' mangled-name='fnvlist_add_boolean_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_boolean_value'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='c19b74c3' name='val'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_byte' mangled-name='fnvlist_add_byte' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_byte'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='d8bf0010' name='val'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_int8' mangled-name='fnvlist_add_int8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_int8'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='ee31ee44' name='val'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_uint8' mangled-name='fnvlist_add_uint8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_uint8'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='b96825af' name='val'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_int16' mangled-name='fnvlist_add_int16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_int16'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='23bd8cb5' name='val'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_uint16' mangled-name='fnvlist_add_uint16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_uint16'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='149c6638' name='val'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_int32' mangled-name='fnvlist_add_int32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_int32'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='3ff5601b' name='val'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_uint32' mangled-name='fnvlist_add_uint32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_uint32'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='8f92235e' name='val'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_int64' mangled-name='fnvlist_add_int64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_int64'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='9da381c4' name='val'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_uint64' mangled-name='fnvlist_add_uint64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_uint64'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='9c313c2d' name='val'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_string' mangled-name='fnvlist_add_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_string'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='80f4b756' name='val'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_nvlist' mangled-name='fnvlist_add_nvlist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_nvlist'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='5ce45b60' name='val'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_nvpair' mangled-name='fnvlist_add_nvpair' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_nvpair'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='3fa542f0' name='pair'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_boolean_array' mangled-name='fnvlist_add_boolean_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_boolean_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='c5f6c15b' name='val'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_byte_array' mangled-name='fnvlist_add_byte_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_byte_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='d1db479e' name='val'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_int8_array' mangled-name='fnvlist_add_int8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_int8_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='a06445da' name='val'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_uint8_array' mangled-name='fnvlist_add_uint8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_uint8_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='9f7200cf' name='val'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_int16_array' mangled-name='fnvlist_add_int16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_int16_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='a3eb883d' name='val'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_uint16_array' mangled-name='fnvlist_add_uint16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_uint16_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='1b7d11c6' name='val'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_int32_array' mangled-name='fnvlist_add_int32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_int32_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='1f526493' name='val'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_uint32_array' mangled-name='fnvlist_add_uint32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_uint32_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='a6798dcc' name='val'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_int64_array' mangled-name='fnvlist_add_int64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_int64_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='505bed1a' name='val'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_uint64_array' mangled-name='fnvlist_add_uint64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_uint64_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='713a56f5' name='val'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_string_array' mangled-name='fnvlist_add_string_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_string_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='13956559' name='val'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_nvlist_array' mangled-name='fnvlist_add_nvlist_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_add_nvlist_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='3bbfee2e' name='val'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_remove' mangled-name='fnvlist_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_remove'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_remove_nvpair' mangled-name='fnvlist_remove_nvpair' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_remove_nvpair'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='3fa542f0' name='pair'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_nvpair' mangled-name='fnvlist_lookup_nvpair' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_nvpair'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='3fa542f0'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_boolean' mangled-name='fnvlist_lookup_boolean' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_boolean'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_boolean_value' mangled-name='fnvlist_lookup_boolean_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_boolean_value'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_byte' mangled-name='fnvlist_lookup_byte' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_byte'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='d8bf0010'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_int8' mangled-name='fnvlist_lookup_int8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_int8'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='ee31ee44'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_int16' mangled-name='fnvlist_lookup_int16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_int16'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='23bd8cb5'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_int32' mangled-name='fnvlist_lookup_int32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_int32'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='3ff5601b'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_int64' mangled-name='fnvlist_lookup_int64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_int64'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='9da381c4'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_uint8' mangled-name='fnvlist_lookup_uint8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_uint8'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_uint16' mangled-name='fnvlist_lookup_uint16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_uint16'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_uint32' mangled-name='fnvlist_lookup_uint32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_uint32'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_uint64' mangled-name='fnvlist_lookup_uint64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_uint64'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_string' mangled-name='fnvlist_lookup_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_string'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_nvlist' mangled-name='fnvlist_lookup_nvlist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_nvlist'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_boolean_array' mangled-name='fnvlist_lookup_boolean_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_boolean_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='4dd26a40' name='n'/>
	<return type-id='37e3bd22'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_byte_array' mangled-name='fnvlist_lookup_byte_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_byte_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='4dd26a40' name='n'/>
	<return type-id='45b65157'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_int8_array' mangled-name='fnvlist_lookup_int8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_int8_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='4dd26a40' name='n'/>
	<return type-id='256d5229'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_uint8_array' mangled-name='fnvlist_lookup_uint8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_uint8_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='4dd26a40' name='n'/>
	<return type-id='ae3e8ca6'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_int16_array' mangled-name='fnvlist_lookup_int16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_int16_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='4dd26a40' name='n'/>
	<return type-id='f76f73d0'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_uint16_array' mangled-name='fnvlist_lookup_uint16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_uint16_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='4dd26a40' name='n'/>
	<return type-id='8a121f49'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_int32_array' mangled-name='fnvlist_lookup_int32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_int32_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='4dd26a40' name='n'/>
	<return type-id='4aafb922'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_uint32_array' mangled-name='fnvlist_lookup_uint32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_uint32_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='4dd26a40' name='n'/>
	<return type-id='90421557'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_int64_array' mangled-name='fnvlist_lookup_int64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_int64_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='4dd26a40' name='n'/>
	<return type-id='cb785ebf'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_uint64_array' mangled-name='fnvlist_lookup_uint64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fnvlist_lookup_uint64_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='4dd26a40' name='n'/>
	<return type-id='5d6479ae'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='module/nvpair/nvpair.c' language='LANG_C99'>
	<typedef-decl name='bool_t' type-id='3ff5601b' id='310a70df'/>
	<enum-decl name='xdr_op' id='6badf1b8'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='XDR_ENCODE' value='0'/>
	<enumerator name='XDR_DECODE' value='1'/>
	<enumerator name='XDR_FREE' value='2'/>
	</enum-decl>
	<class-decl name='__rpc_xdr' size-in-bits='384' is-struct='yes' visibility='default' id='755707df'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='x_op' type-id='6badf1b8' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='x_ops' type-id='3457e9c5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='x_public' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='x_private' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='x_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='x_handy' type-id='48f7c3f5' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='xdr_ops' size-in-bits='576' is-struct='yes' visibility='default' id='ea1474f2'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='x_getlong' type-id='a4e6dd3f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='x_putlong' type-id='d447b08c' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='x_getbytes' type-id='b0b6ccaa' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='x_putbytes' type-id='581c24b7' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='x_getpostn' type-id='1c77bdb0' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='x_setpostn' type-id='13fdb43e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='x_inline' type-id='788e1f25' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='x_destroy' type-id='d781e3b2' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='x_control' type-id='733fc725' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='XDR' type-id='755707df' id='bc407f0e'/>
	<typedef-decl name='xdrproc_t' type-id='94d188f0' id='c28db3e9'/>
	<typedef-decl name='__u_short' type-id='8efea9e5' id='46c660f8'/>
	<typedef-decl name='__u_int' type-id='f0981eeb' id='8ae6822f'/>
	<typedef-decl name='__quad_t' type-id='bd54fe1a' id='2632227a'/>
	<typedef-decl name='__u_quad_t' type-id='7359adad' id='5f3d50a6'/>
	<typedef-decl name='u_short' type-id='46c660f8' id='32580e96'/>
	<typedef-decl name='u_int' type-id='8ae6822f' id='48f7c3f5'/>
	<typedef-decl name='quad_t' type-id='2632227a' id='f5ef0660'/>
	<typedef-decl name='u_quad_t' type-id='5f3d50a6' id='bd226ac0'/>
	<pointer-type-def type-id='bc407f0e' size-in-bits='64' id='17fd1621'/>
	<pointer-type-def type-id='755707df' size-in-bits='64' id='812c6697'/>
	<pointer-type-def type-id='26a90f95' size-in-bits='64' id='9b23c9ad'/>
	<qualified-type-def type-id='9b23c9ad' restrict='yes' id='8c85230f'/>
	<pointer-type-def type-id='9b23c9ad' size-in-bits='64' id='c0563f85'/>
	<qualified-type-def type-id='c19b74c3' const='yes' id='12373e33'/>
	<pointer-type-def type-id='12373e33' size-in-bits='64' id='c5f6c15b'/>
	<qualified-type-def type-id='80f4b756' const='yes' id='b99c00c9'/>
	<pointer-type-def type-id='b99c00c9' size-in-bits='64' id='13956559'/>
	<qualified-type-def type-id='23bd8cb5' const='yes' id='75f7b0c5'/>
	<pointer-type-def type-id='75f7b0c5' size-in-bits='64' id='a3eb883d'/>
	<qualified-type-def type-id='3ff5601b' const='yes' id='922df12b'/>
	<pointer-type-def type-id='922df12b' size-in-bits='64' id='1f526493'/>
	<qualified-type-def type-id='9da381c4' const='yes' id='f07b7694'/>
	<pointer-type-def type-id='f07b7694' size-in-bits='64' id='505bed1a'/>
	<qualified-type-def type-id='ee31ee44' const='yes' id='721c32d4'/>
	<pointer-type-def type-id='721c32d4' size-in-bits='64' id='a06445da'/>
	<qualified-type-def type-id='bd54fe1a' const='yes' id='9cb2385f'/>
	<pointer-type-def type-id='9cb2385f' size-in-bits='64' id='218ee02f'/>
	<qualified-type-def type-id='8e8d4be3' const='yes' id='693c3853'/>
	<pointer-type-def type-id='693c3853' size-in-bits='64' id='22cce67b'/>
	<qualified-type-def type-id='22cce67b' const='yes' id='d2816df0'/>
	<pointer-type-def type-id='d2816df0' size-in-bits='64' id='3bbfee2e'/>
	<qualified-type-def type-id='d8bf0010' const='yes' id='a9125480'/>
	<pointer-type-def type-id='a9125480' size-in-bits='64' id='d1db479e'/>
	<qualified-type-def type-id='149c6638' const='yes' id='b01a5ac8'/>
	<pointer-type-def type-id='b01a5ac8' size-in-bits='64' id='1b7d11c6'/>
	<qualified-type-def type-id='8f92235e' const='yes' id='b9930aae'/>
	<pointer-type-def type-id='b9930aae' size-in-bits='64' id='a6798dcc'/>
	<qualified-type-def type-id='9c313c2d' const='yes' id='c3b7ba7d'/>
	<pointer-type-def type-id='c3b7ba7d' size-in-bits='64' id='713a56f5'/>
	<qualified-type-def type-id='b96825af' const='yes' id='2b61797f'/>
	<pointer-type-def type-id='2b61797f' size-in-bits='64' id='9f7200cf'/>
	<qualified-type-def type-id='ea1474f2' const='yes' id='485ede6d'/>
	<pointer-type-def type-id='485ede6d' size-in-bits='64' id='3457e9c5'/>
	<pointer-type-def type-id='95e97e5e' size-in-bits='64' id='7292109c'/>
	<pointer-type-def type-id='441e0c31' size-in-bits='64' id='788e1f25'/>
	<pointer-type-def type-id='bd54fe1a' size-in-bits='64' id='3ccc2590'/>
	<pointer-type-def type-id='3fa542f0' size-in-bits='64' id='0b283d2e'/>
	<pointer-type-def type-id='f5ef0660' size-in-bits='64' id='de23782d'/>
	<pointer-type-def type-id='a2185560' size-in-bits='64' id='764e437e'/>
	<pointer-type-def type-id='b59d7dce' size-in-bits='64' id='78c01427'/>
	<pointer-type-def type-id='e80b47fe' size-in-bits='64' id='94d188f0'/>
	<pointer-type-def type-id='1c7a4858' size-in-bits='64' id='b0b6ccaa'/>
	<pointer-type-def type-id='f6358b93' size-in-bits='64' id='581c24b7'/>
	<pointer-type-def type-id='45354e42' size-in-bits='64' id='d447b08c'/>
	<pointer-type-def type-id='0760d6d1' size-in-bits='64' id='733fc725'/>
	<pointer-type-def type-id='be0f9e0b' size-in-bits='64' id='a4e6dd3f'/>
	<pointer-type-def type-id='16eb5704' size-in-bits='64' id='13fdb43e'/>
	<pointer-type-def type-id='46870456' size-in-bits='64' id='1c77bdb0'/>
	<pointer-type-def type-id='48f7c3f5' size-in-bits='64' id='b4c2e924'/>
	<pointer-type-def type-id='bd226ac0' size-in-bits='64' id='fce59795'/>
	<pointer-type-def type-id='32580e96' size-in-bits='64' id='ede2c2f5'/>
	<pointer-type-def type-id='c87cb1d0' size-in-bits='64' id='d781e3b2'/>
	<function-decl name='strtol' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9d26089a'/>
	<parameter type-id='8c85230f'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='bd54fe1a'/>
	</function-decl>
	<function-decl name='strncmp' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	- <function-decl name='strchr' visibility='default' binding='global' size-in-bits='64'>
	- <parameter type-id='80f4b756'/>
	- <parameter type-id='95e97e5e'/>
	- <return type-id='26a90f95'/>
	- </function-decl>
	<function-decl name='strlen' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='xdr_int' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='17fd1621'/>
	<parameter type-id='7292109c'/>
	<return type-id='310a70df'/>
	</function-decl>
	<function-decl name='xdr_u_int' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='17fd1621'/>
	<parameter type-id='b4c2e924'/>
	<return type-id='310a70df'/>
	</function-decl>
	<function-decl name='xdr_short' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='17fd1621'/>
	<parameter type-id='764e437e'/>
	<return type-id='310a70df'/>
	</function-decl>
	<function-decl name='xdr_u_short' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='17fd1621'/>
	<parameter type-id='ede2c2f5'/>
	<return type-id='310a70df'/>
	</function-decl>
	<function-decl name='xdr_array' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='17fd1621'/>
	<parameter type-id='9b23c9ad'/>
	<parameter type-id='b4c2e924'/>
	<parameter type-id='48f7c3f5'/>
	<parameter type-id='48f7c3f5'/>
	<parameter type-id='c28db3e9'/>
	<return type-id='310a70df'/>
	</function-decl>
	<function-decl name='xdr_opaque' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='17fd1621'/>
	<parameter type-id='26a90f95'/>
	<parameter type-id='48f7c3f5'/>
	<return type-id='310a70df'/>
	</function-decl>
	<function-decl name='xdr_string' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='17fd1621'/>
	<parameter type-id='9b23c9ad'/>
	<parameter type-id='48f7c3f5'/>
	<return type-id='310a70df'/>
	</function-decl>
	<function-decl name='xdr_char' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='17fd1621'/>
	<parameter type-id='26a90f95'/>
	<return type-id='310a70df'/>
	</function-decl>
	<function-decl name='xdr_double' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='17fd1621'/>
	<parameter type-id='7408d286'/>
	<return type-id='310a70df'/>
	</function-decl>
	<function-decl name='xdr_longlong_t' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='17fd1621'/>
	<parameter type-id='de23782d'/>
	<return type-id='310a70df'/>
	</function-decl>
	<function-decl name='xdr_u_longlong_t' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='17fd1621'/>
	<parameter type-id='fce59795'/>
	<return type-id='310a70df'/>
	</function-decl>
	<function-decl name='xdrmem_create' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='17fd1621'/>
	<parameter type-id='26a90f95'/>
	<parameter type-id='48f7c3f5'/>
	<parameter type-id='6badf1b8'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nv_alloc_init' mangled-name='nv_alloc_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nv_alloc_init'>
	<parameter type-id='11871392' name='nva'/>
	<parameter type-id='ee1d4944' name='nvo'/>
	<parameter is-variadic='yes'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nv_alloc_reset' mangled-name='nv_alloc_reset' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nv_alloc_reset'>
	<parameter type-id='11871392' name='nva'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nv_alloc_fini' mangled-name='nv_alloc_fini' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nv_alloc_fini'>
	<parameter type-id='11871392' name='nva'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_lookup_nv_alloc' mangled-name='nvlist_lookup_nv_alloc' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_nv_alloc'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<return type-id='11871392'/>
	</function-decl>
	<function-decl name='nvlist_nvflag' mangled-name='nvlist_nvflag' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_nvflag'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<return type-id='3502e3ff'/>
	</function-decl>
	<function-decl name='nvlist_alloc' mangled-name='nvlist_alloc' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_alloc'>
	<parameter type-id='857bb57e' name='nvlp'/>
	<parameter type-id='3502e3ff' name='nvflag'/>
	<parameter type-id='95e97e5e' name='kmflag'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_xalloc' mangled-name='nvlist_xalloc' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_xalloc'>
	<parameter type-id='857bb57e' name='nvlp'/>
	<parameter type-id='3502e3ff' name='nvflag'/>
	<parameter type-id='11871392' name='nva'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_free' mangled-name='nvlist_free' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_free'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_dup' mangled-name='nvlist_dup' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_dup'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='857bb57e' name='nvlp'/>
	<parameter type-id='95e97e5e' name='kmflag'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_xdup' mangled-name='nvlist_xdup' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_xdup'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='857bb57e' name='nvlp'/>
	<parameter type-id='11871392' name='nva'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_remove_all' mangled-name='nvlist_remove_all' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_remove_all'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_remove' mangled-name='nvlist_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_remove'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='8d0687d2' name='type'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_remove_nvpair' mangled-name='nvlist_remove_nvpair' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_remove_nvpair'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='3fa542f0' name='nvp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_boolean' mangled-name='nvlist_add_boolean' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_boolean'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_boolean_value' mangled-name='nvlist_add_boolean_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_boolean_value'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='c19b74c3' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_byte' mangled-name='nvlist_add_byte' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_byte'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='d8bf0010' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_int8' mangled-name='nvlist_add_int8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_int8'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='ee31ee44' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_uint8' mangled-name='nvlist_add_uint8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_uint8'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='b96825af' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_int16' mangled-name='nvlist_add_int16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_int16'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='23bd8cb5' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_uint16' mangled-name='nvlist_add_uint16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_uint16'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='149c6638' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_int32' mangled-name='nvlist_add_int32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_int32'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='3ff5601b' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_uint32' mangled-name='nvlist_add_uint32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_uint32'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='8f92235e' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_int64' mangled-name='nvlist_add_int64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_int64'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='9da381c4' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_uint64' mangled-name='nvlist_add_uint64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_uint64'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='9c313c2d' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_double' mangled-name='nvlist_add_double' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_double'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='a0eb0f08' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_string' mangled-name='nvlist_add_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_string'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='80f4b756' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_boolean_array' mangled-name='nvlist_add_boolean_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_boolean_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='c5f6c15b' name='a'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_byte_array' mangled-name='nvlist_add_byte_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_byte_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='d1db479e' name='a'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_int8_array' mangled-name='nvlist_add_int8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_int8_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='a06445da' name='a'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_uint8_array' mangled-name='nvlist_add_uint8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_uint8_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='9f7200cf' name='a'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_int16_array' mangled-name='nvlist_add_int16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_int16_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='a3eb883d' name='a'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_uint16_array' mangled-name='nvlist_add_uint16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_uint16_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='1b7d11c6' name='a'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_int32_array' mangled-name='nvlist_add_int32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_int32_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='1f526493' name='a'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_uint32_array' mangled-name='nvlist_add_uint32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_uint32_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='a6798dcc' name='a'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_int64_array' mangled-name='nvlist_add_int64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_int64_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='505bed1a' name='a'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_uint64_array' mangled-name='nvlist_add_uint64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_uint64_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='713a56f5' name='a'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_string_array' mangled-name='nvlist_add_string_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_string_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='13956559' name='a'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_hrtime' mangled-name='nvlist_add_hrtime' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_hrtime'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='cebdd548' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_nvlist' mangled-name='nvlist_add_nvlist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_nvlist'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='22cce67b' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_nvlist_array' mangled-name='nvlist_add_nvlist_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_nvlist_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='3bbfee2e' name='a'/>
	<parameter type-id='3502e3ff' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_prev_nvpair' mangled-name='nvlist_prev_nvpair' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_prev_nvpair'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='dace003f' name='nvp'/>
	<return type-id='3fa542f0'/>
	</function-decl>
	<function-decl name='nvlist_empty' mangled-name='nvlist_empty' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_empty'>
	<parameter type-id='22cce67b' name='nvl'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='nvlist_lookup_boolean' mangled-name='nvlist_lookup_boolean' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_boolean'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_boolean_value' mangled-name='nvlist_lookup_boolean_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_boolean_value'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='37e3bd22' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_byte' mangled-name='nvlist_lookup_byte' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_byte'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='45b65157' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_int8' mangled-name='nvlist_lookup_int8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_int8'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='256d5229' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_uint8' mangled-name='nvlist_lookup_uint8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_uint8'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='ae3e8ca6' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_int16' mangled-name='nvlist_lookup_int16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_int16'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='f76f73d0' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_uint16' mangled-name='nvlist_lookup_uint16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_uint16'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='8a121f49' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_int32' mangled-name='nvlist_lookup_int32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_int32'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='4aafb922' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_uint32' mangled-name='nvlist_lookup_uint32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_uint32'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='90421557' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_int64' mangled-name='nvlist_lookup_int64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_int64'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='cb785ebf' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_uint64' mangled-name='nvlist_lookup_uint64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_uint64'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='5d6479ae' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_double' mangled-name='nvlist_lookup_double' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_double'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='7408d286' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_string' mangled-name='nvlist_lookup_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_string'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='7d3cd834' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_nvlist' mangled-name='nvlist_lookup_nvlist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_nvlist'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='857bb57e' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_boolean_array' mangled-name='nvlist_lookup_boolean_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_boolean_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='03829398' name='a'/>
	<parameter type-id='4dd26a40' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_byte_array' mangled-name='nvlist_lookup_byte_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_byte_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='3b0247c7' name='a'/>
	<parameter type-id='4dd26a40' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_int8_array' mangled-name='nvlist_lookup_int8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_int8_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='ee181ab9' name='a'/>
	<parameter type-id='4dd26a40' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_uint8_array' mangled-name='nvlist_lookup_uint8_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_uint8_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='d8774064' name='a'/>
	<parameter type-id='4dd26a40' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_int16_array' mangled-name='nvlist_lookup_int16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_int16_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='7e73928e' name='a'/>
	<parameter type-id='4dd26a40' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_uint16_array' mangled-name='nvlist_lookup_uint16_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_uint16_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='bd8768d9' name='a'/>
	<parameter type-id='4dd26a40' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_int32_array' mangled-name='nvlist_lookup_int32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_int32_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='9aa04798' name='a'/>
	<parameter type-id='4dd26a40' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_uint32_array' mangled-name='nvlist_lookup_uint32_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_uint32_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='9507d3c7' name='a'/>
	<parameter type-id='4dd26a40' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_int64_array' mangled-name='nvlist_lookup_int64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_int64_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='e37ce48f' name='a'/>
	<parameter type-id='4dd26a40' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_uint64_array' mangled-name='nvlist_lookup_uint64_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_uint64_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='892b4acc' name='a'/>
	<parameter type-id='4dd26a40' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_string_array' mangled-name='nvlist_lookup_string_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_string_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='c0563f85' name='a'/>
	<parameter type-id='4dd26a40' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_nvlist_array' mangled-name='nvlist_lookup_nvlist_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_nvlist_array'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='75be733c' name='a'/>
	<parameter type-id='4dd26a40' name='n'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_hrtime' mangled-name='nvlist_lookup_hrtime' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_hrtime'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='e379e62d' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_pairs' mangled-name='nvlist_lookup_pairs' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_pairs'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='95e97e5e' name='flag'/>
	<parameter is-variadic='yes'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_nvpair' mangled-name='nvlist_lookup_nvpair' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_nvpair'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='0b283d2e' name='ret'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_nvpair_embedded_index' mangled-name='nvlist_lookup_nvpair_embedded_index' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_lookup_nvpair_embedded_index'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='0b283d2e' name='ret'/>
	<parameter type-id='7292109c' name='ip'/>
	<parameter type-id='7d3cd834' name='ep'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_exists' mangled-name='nvlist_exists' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_exists'>
	<parameter type-id='22cce67b' name='nvl'/>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='nvlist_add_nvpair' mangled-name='nvlist_add_nvpair' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_add_nvpair'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='3fa542f0' name='nvp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_merge' mangled-name='nvlist_merge' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_merge'>
	<parameter type-id='5ce45b60' name='dst'/>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='95e97e5e' name='flag'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_size' mangled-name='nvlist_size' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_size'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='78c01427' name='size'/>
	<parameter type-id='95e97e5e' name='encoding'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_pack' mangled-name='nvlist_pack' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_pack'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='9b23c9ad' name='bufp'/>
	<parameter type-id='78c01427' name='buflen'/>
	<parameter type-id='95e97e5e' name='encoding'/>
	<parameter type-id='95e97e5e' name='kmflag'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_xpack' mangled-name='nvlist_xpack' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_xpack'>
	<parameter type-id='5ce45b60' name='nvl'/>
	<parameter type-id='9b23c9ad' name='bufp'/>
	<parameter type-id='78c01427' name='buflen'/>
	<parameter type-id='95e97e5e' name='encoding'/>
	<parameter type-id='11871392' name='nva'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_unpack' mangled-name='nvlist_unpack' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_unpack'>
	<parameter type-id='26a90f95' name='buf'/>
	<parameter type-id='b59d7dce' name='buflen'/>
	<parameter type-id='857bb57e' name='nvlp'/>
	<parameter type-id='95e97e5e' name='kmflag'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_xunpack' mangled-name='nvlist_xunpack' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='nvlist_xunpack'>
	<parameter type-id='26a90f95' name='buf'/>
	<parameter type-id='b59d7dce' name='buflen'/>
	<parameter type-id='857bb57e' name='nvlp'/>
	<parameter type-id='11871392' name='nva'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-type size-in-bits='64' id='441e0c31'>
	<parameter type-id='812c6697'/>
	<parameter type-id='48f7c3f5'/>
	<return type-id='4aafb922'/>
	</function-type>
	<function-type size-in-bits='64' id='e80b47fe'>
	<parameter type-id='17fd1621'/>
	<parameter is-variadic='yes'/>
	<return type-id='310a70df'/>
	</function-type>
	<function-type size-in-bits='64' id='1c7a4858'>
	<parameter type-id='812c6697'/>
	<parameter type-id='26a90f95'/>
	<parameter type-id='48f7c3f5'/>
	<return type-id='310a70df'/>
	</function-type>
	<function-type size-in-bits='64' id='f6358b93'>
	<parameter type-id='812c6697'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='48f7c3f5'/>
	<return type-id='310a70df'/>
	</function-type>
	<function-type size-in-bits='64' id='45354e42'>
	<parameter type-id='812c6697'/>
	<parameter type-id='218ee02f'/>
	<return type-id='310a70df'/>
	</function-type>
	<function-type size-in-bits='64' id='0760d6d1'>
	<parameter type-id='812c6697'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='310a70df'/>
	</function-type>
	<function-type size-in-bits='64' id='be0f9e0b'>
	<parameter type-id='812c6697'/>
	<parameter type-id='3ccc2590'/>
	<return type-id='310a70df'/>
	</function-type>
	<function-type size-in-bits='64' id='16eb5704'>
	<parameter type-id='812c6697'/>
	<parameter type-id='48f7c3f5'/>
	<return type-id='310a70df'/>
	</function-type>
	<function-type size-in-bits='64' id='46870456'>
	<parameter type-id='812c6697'/>
	<return type-id='48f7c3f5'/>
	</function-type>
	<function-type size-in-bits='64' id='c87cb1d0'>
	<parameter type-id='812c6697'/>
	<return type-id='48b5725f'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='module/nvpair/nvpair_alloc_fixed.c' language='LANG_C99'>
	<qualified-type-def type-id='ee1d4944' const='yes' id='4b95388f'/>
	<var-decl name='nv_fixed_ops' type-id='4b95388f' mangled-name='nv_fixed_ops' visibility='default' elf-symbol-id='nv_fixed_ops'/>
	</abi-instr>
	</abi-corpus>
	diff --git a/sys/contrib/openzfs/lib/libspl/Makefile.am b/sys/contrib/openzfs/lib/libspl/Makefile.am
	index eb2377305aca..f8943572bf29 100644
	--- a/sys/contrib/openzfs/lib/libspl/Makefile.am
	+++ b/sys/contrib/openzfs/lib/libspl/Makefile.am
	@@ -1,47 +1,52 @@
	include $(srcdir)/%D%/include/Makefile.am

	libspl_assert_la_CFLAGS = $(AM_CFLAGS) $(LIBRARY_CFLAGS) $(LIBUNWIND_CFLAGS)
	libspl_la_CFLAGS = $(libspl_assert_la_CFLAGS)

	noinst_LTLIBRARIES += libspl_assert.la libspl.la
	CPPCHECKTARGETS += libspl_assert.la libspl.la

	libspl_assert_la_SOURCES = \
	- %D%/assert.c
	+ %D%/assert.c \
	+ %D%/backtrace.c

	libspl_la_SOURCES = \
	%D%/libspl_impl.h \
	%D%/atomic.c \
	%D%/getexecname.c \
	%D%/list.c \
	%D%/mkdirp.c \
	%D%/page.c \
	%D%/strlcat.c \
	%D%/strlcpy.c \
	%D%/timestamp.c \
	%D%/include/sys/list.h \
	%D%/include/sys/list_impl.h

	if BUILD_LINUX
	libspl_la_SOURCES += \
	%D%/os/linux/getexecname.c \
	%D%/os/linux/gethostid.c \
	%D%/os/linux/getmntany.c \
	%D%/os/linux/zone.c
	endif

	if BUILD_FREEBSD
	libspl_la_SOURCES += \
	%D%/os/freebsd/getexecname.c \
	%D%/os/freebsd/gethostid.c \
	%D%/os/freebsd/getmntany.c \
	%D%/os/freebsd/mnttab.c \
	%D%/os/freebsd/zone.c
	endif

	libspl_la_LIBADD = \
	libspl_assert.la

	libspl_la_LIBADD += $(LIBATOMIC_LIBS) $(LIBCLOCK_GETTIME)

	libspl_assert_la_LIBADD = $(BACKTRACE_LIBS) $(LIBUNWIND_LIBS)
	+
	+if BUILD_FREEBSD
	+libspl_assert_la_LIBADD += -lpthread
	+endif
	diff --git a/sys/contrib/openzfs/lib/libspl/assert.c b/sys/contrib/openzfs/lib/libspl/assert.c
	index 5b12c14acd6e..315ddd6b9a9d 100644
	--- a/sys/contrib/openzfs/lib/libspl/assert.c
	+++ b/sys/contrib/openzfs/lib/libspl/assert.c
	@@ -1,163 +1,128 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright 2008 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/
	/*
	* Copyright (c) 2024, Rob Norris <robn@despairlabs.com>
	*/

	#include <assert.h>
	#include <pthread.h>
	+#include <sys/backtrace.h>

	#if defined(__linux__)
	#include <errno.h>
	#include <sys/prctl.h>
	#ifdef HAVE_GETTID
	#define libspl_gettid() gettid()
	#else
	#include <sys/syscall.h>
	#define libspl_gettid() ((pid_t)syscall(__NR_gettid))
	#endif
	#define libspl_getprogname() (program_invocation_short_name)
	#define libspl_getthreadname(buf, len) \
	prctl(PR_GET_NAME, (unsigned long)(buf), 0, 0, 0)
	#elif defined(__FreeBSD__) \|\| defined(__APPLE__)
	#if !defined(__APPLE__)
	#include <pthread_np.h>
	#define libspl_gettid() pthread_getthreadid_np()
	#endif
	#define libspl_getprogname() getprogname()
	#define libspl_getthreadname(buf, len) \
	pthread_getname_np(pthread_self(), buf, len);
	#endif

	-#if defined(HAVE_LIBUNWIND)
	-#define UNW_LOCAL_ONLY
	-#include <libunwind.h>
	-
	-static inline void
	-libspl_dump_backtrace(void)
	+#if defined(__APPLE__)
	+static inline uint64_t
	+libspl_gettid(void)
	{
	- unw_context_t uc;
	- unw_cursor_t cp;
	- unw_word_t ip, off;
	- char funcname[128];
	-#ifdef HAVE_LIBUNWIND_ELF
	- char objname[128];
	- unw_word_t objoff;
	-#endif
	+ uint64_t tid;

	- fprintf(stderr, "Call trace:\n");
	- unw_getcontext(&uc);
	- unw_init_local(&cp, &uc);
	- while (unw_step(&cp) > 0) {
	- unw_get_reg(&cp, UNW_REG_IP, &ip);
	- unw_get_proc_name(&cp, funcname, sizeof (funcname), &off);
	-#ifdef HAVE_LIBUNWIND_ELF
	- unw_get_elf_filename(&cp, objname, sizeof (objname), &objoff);
	- fprintf(stderr, " [0x%08lx] %s+0x%2lx (in %s +0x%2lx)\n",
	- ip, funcname, off, objname, objoff);
	-#else
	- fprintf(stderr, " [0x%08lx] %s+0x%2lx\n", ip, funcname, off);
	-#endif
	- }
	-}
	-#elif defined(HAVE_BACKTRACE)
	-#include <execinfo.h>
	+ if (pthread_threadid_np(NULL, &tid) != 0)
	+ tid = 0;

	-static inline void
	-libspl_dump_backtrace(void)
	-{
	- void *btptrs[100];
	- size_t nptrs = backtrace(btptrs, 100);
	- char **bt = backtrace_symbols(btptrs, nptrs);
	- fprintf(stderr, "Call trace:\n");
	- for (size_t i = 0; i < nptrs; i++)
	- fprintf(stderr, " %s\n", bt[i]);
	- free(bt);
	+ return (tid);
	}
	-#else
	-#define libspl_dump_backtrace()
	#endif

	#if defined(__APPLE__)
	static inline uint64_t
	libspl_gettid(void)
	{
	uint64_t tid;

	if (pthread_threadid_np(NULL, &tid) != 0)
	tid = 0;

	return (tid);
	}
	#endif

	static boolean_t libspl_assert_ok = B_FALSE;

	void
	libspl_set_assert_ok(boolean_t val)
	{
	libspl_assert_ok = val;
	}

	static pthread_mutex_t assert_lock = PTHREAD_MUTEX_INITIALIZER;

	/* printf version of libspl_assert */
	void
	libspl_assertf(const char file, const char func, int line,
	const char *format, ...)
	{
	pthread_mutex_lock(&assert_lock);

	va_list args;
	char tname[64];

	libspl_getthreadname(tname, sizeof (tname));

	fprintf(stderr, "ASSERT at %s:%d:%s()\n", file, line, func);

	va_start(args, format);
	vfprintf(stderr, format, args);
	va_end(args);

	fprintf(stderr, "\n"
	" PID: %-8u COMM: %s\n"
	#if defined(__APPLE__)
	" TID: %-8" PRIu64 " NAME: %s\n",
	#else
	" TID: %-8u NAME: %s\n",
	#endif
	getpid(), libspl_getprogname(),
	libspl_gettid(), tname);

	- libspl_dump_backtrace();
	+ libspl_backtrace(STDERR_FILENO);

	#if !__has_feature(attribute_analyzer_noreturn) && !defined(__COVERITY__)
	if (libspl_assert_ok) {
	pthread_mutex_unlock(&assert_lock);
	return;
	}
	#endif
	abort();
	}
	diff --git a/sys/contrib/openzfs/lib/libspl/backtrace.c b/sys/contrib/openzfs/lib/libspl/backtrace.c
	new file mode 100644
	index 000000000000..d26d742106e2
	--- /dev/null
	+++ b/sys/contrib/openzfs/lib/libspl/backtrace.c
	@@ -0,0 +1,119 @@
	+/*
	+ * CDDL HEADER START
	+ *
	+ * The contents of this file are subject to the terms of the
	+ * Common Development and Distribution License (the "License").
	+ * You may not use this file except in compliance with the License.
	+ *
	+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	+ * or https://opensource.org/licenses/CDDL-1.0.
	+ * See the License for the specific language governing permissions
	+ * and limitations under the License.
	+ *
	+ * When distributing Covered Code, include this CDDL HEADER in each
	+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	+ * If applicable, add the following below this CDDL HEADER, with the
	+ * fields enclosed by brackets "[]" replaced with your own identifying
	+ * information: Portions Copyright [yyyy] [name of copyright owner]
	+ *
	+ * CDDL HEADER END
	+ */
	+/*
	+ * Copyright (c) 2024, Rob Norris <robn@despairlabs.com>
	+ * Copyright (c) 2024, Klara Inc.
	+ */
	+
	+#include <sys/backtrace.h>
	+#include <sys/types.h>
	+#include <unistd.h>
	+
	+/*
	+ * libspl_backtrace() must be safe to call from inside a signal hander. This
	+ * mostly means it must not allocate, and so we can't use things like printf.
	+ */
	+
	+#if defined(HAVE_LIBUNWIND)
	+#define UNW_LOCAL_ONLY
	+#include <libunwind.h>
	+
	+static size_t
	+libspl_u64_to_hex_str(uint64_t v, size_t digits, char *buf, size_t buflen)
	+{
	+ static const char hexdigits[] = {
	+ '0', '1', '2', '3', '4', '5', '6', '7',
	+ '8', '9', 'a', 'b', 'c', 'd', 'e', 'f'
	+ };
	+
	+ size_t pos = 0;
	+ boolean_t want = (digits == 0);
	+ for (int i = 15; i >= 0; i--) {
	+ const uint64_t d = v >> (i * 4) & 0xf;
	+ if (!want && (d != 0 \|\| digits > i))
	+ want = B_TRUE;
	+ if (want) {
	+ buf[pos++] = hexdigits[d];
	+ if (pos == buflen)
	+ break;
	+ }
	+ }
	+ return (pos);
	+}
	+
	+void
	+libspl_backtrace(int fd)
	+{
	+ ssize_t ret __attribute__((unused));
	+ unw_context_t uc;
	+ unw_cursor_t cp;
	+ unw_word_t loc;
	+ char buf[128];
	+ size_t n;
	+
	+ ret = write(fd, "Call trace:\n", 12);
	+ unw_getcontext(&uc);
	+ unw_init_local(&cp, &uc);
	+ while (unw_step(&cp) > 0) {
	+ unw_get_reg(&cp, UNW_REG_IP, &loc);
	+ ret = write(fd, " [0x", 5);
	+ n = libspl_u64_to_hex_str(loc, 10, buf, sizeof (buf));
	+ ret = write(fd, buf, n);
	+ ret = write(fd, "] ", 2);
	+ unw_get_proc_name(&cp, buf, sizeof (buf), &loc);
	+ for (n = 0; n < sizeof (buf) && buf[n] != '\0'; n++) {}
	+ ret = write(fd, buf, n);
	+ ret = write(fd, "+0x", 3);
	+ n = libspl_u64_to_hex_str(loc, 2, buf, sizeof (buf));
	+ ret = write(fd, buf, n);
	+#ifdef HAVE_LIBUNWIND_ELF
	+ ret = write(fd, " (in ", 5);
	+ unw_get_elf_filename(&cp, buf, sizeof (buf), &loc);
	+ for (n = 0; n < sizeof (buf) && buf[n] != '\0'; n++) {}
	+ ret = write(fd, buf, n);
	+ ret = write(fd, " +0x", 4);
	+ n = libspl_u64_to_hex_str(loc, 2, buf, sizeof (buf));
	+ ret = write(fd, buf, n);
	+ ret = write(fd, ")", 1);
	+#endif
	+ ret = write(fd, "\n", 1);
	+ }
	+}
	+#elif defined(HAVE_BACKTRACE)
	+#include <execinfo.h>
	+
	+void
	+libspl_backtrace(int fd)
	+{
	+ ssize_t ret __attribute__((unused));
	+ void *btptrs[64];
	+ size_t nptrs = backtrace(btptrs, 64);
	+ ret = write(fd, "Call trace:\n", 12);
	+ backtrace_symbols_fd(btptrs, nptrs, fd);
	+}
	+#else
	+#include <sys/debug.h>
	+
	+void
	+libspl_backtrace(int fd __maybe_unused)
	+{
	+}
	+#endif
	diff --git a/sys/contrib/openzfs/lib/libspl/include/Makefile.am b/sys/contrib/openzfs/lib/libspl/include/Makefile.am
	index 2c1d21edf19d..4ad3b854cbee 100644
	--- a/sys/contrib/openzfs/lib/libspl/include/Makefile.am
	+++ b/sys/contrib/openzfs/lib/libspl/include/Makefile.am
	@@ -1,103 +1,104 @@
	libspldir = $(includedir)/libspl
	libspl_HEADERS = \
	%D%/assert.h \
	%D%/atomic.h \
	%D%/libgen.h \
	%D%/libshare.h \
	%D%/statcommon.h \
	%D%/stdlib.h \
	%D%/string.h \
	%D%/umem.h \
	%D%/unistd.h \
	%D%/zone.h

	if BUILD_FREEBSD
	libspl_HEADERS += \
	%D%/os/freebsd/fcntl.h
	endif


	libspl_rpcdir = $(libspldir)/rpc
	libspl_rpc_HEADERS = \
	%D%/rpc/xdr.h


	libspl_sysdir = $(libspldir)/sys
	libspl_sys_HEADERS = \
	%D%/sys/acl.h \
	%D%/sys/acl_impl.h \
	%D%/sys/asm_linkage.h \
	+ %D%/sys/backtrace.h \
	%D%/sys/callb.h \
	%D%/sys/cmn_err.h \
	%D%/sys/cred.h \
	%D%/sys/debug.h \
	%D%/sys/dkio.h \
	%D%/sys/dklabel.h \
	%D%/sys/feature_tests.h \
	%D%/sys/inttypes.h \
	%D%/sys/isa_defs.h \
	%D%/sys/kmem.h \
	%D%/sys/kstat.h \
	%D%/sys/list.h \
	%D%/sys/list_impl.h \
	%D%/sys/mhd.h \
	%D%/sys/mkdev.h \
	%D%/sys/policy.h \
	%D%/sys/poll.h \
	%D%/sys/priv.h \
	%D%/sys/processor.h \
	%D%/sys/simd.h \
	%D%/sys/stack.h \
	%D%/sys/stdtypes.h \
	%D%/sys/string.h \
	%D%/sys/sunddi.h \
	%D%/sys/systeminfo.h \
	%D%/sys/time.h \
	%D%/sys/trace_spl.h \
	%D%/sys/trace_zfs.h \
	%D%/sys/types.h \
	%D%/sys/types32.h \
	%D%/sys/uio.h \
	%D%/sys/vnode.h \
	%D%/sys/wmsum.h \
	%D%/sys/zone.h

	libspl_ia32dir = $(libspldir)/sys/ia32

	if BUILD_LINUX
	libspl_sys_HEADERS += \
	%D%/os/linux/sys/byteorder.h \
	%D%/os/linux/sys/errno.h \
	%D%/os/linux/sys/mnttab.h \
	%D%/os/linux/sys/mount.h \
	%D%/os/linux/sys/param.h \
	%D%/os/linux/sys/stat.h \
	%D%/os/linux/sys/sysmacros.h \
	%D%/os/linux/sys/zfs_context_os.h

	libspl_ia32_HEADERS = \
	%D%/os/linux/sys/ia32/asm_linkage.h
	endif

	if BUILD_FREEBSD
	libspl_sys_HEADERS += \
	%D%/os/freebsd/sys/byteorder.h \
	%D%/os/freebsd/sys/fcntl.h \
	%D%/os/freebsd/sys/file.h \
	%D%/os/freebsd/sys/mnttab.h \
	%D%/os/freebsd/sys/mount.h \
	%D%/os/freebsd/sys/param.h \
	%D%/os/freebsd/sys/stat.h \
	%D%/os/freebsd/sys/sysmacros.h \
	%D%/os/freebsd/sys/vfs.h \
	%D%/os/freebsd/sys/zfs_context_os.h

	libspl_ia32_HEADERS = \
	%D%/os/freebsd/sys/ia32/asm_linkage.h
	endif


	libspl_sys_dktpdir = $(libspl_sysdir)/dktp
	libspl_sys_dktp_HEADERS = \
	%D%/sys/dktp/fdisk.h

	diff --git a/sys/contrib/openzfs/lib/libspl/include/os/linux/sys/sysmacros.h b/sys/contrib/openzfs/lib/libspl/include/os/linux/sys/sysmacros.h
	index 5765ee25c6cb..26e1c87a35af 100644
	--- a/sys/contrib/openzfs/lib/libspl/include/os/linux/sys/sysmacros.h
	+++ b/sys/contrib/openzfs/lib/libspl/include/os/linux/sys/sysmacros.h
	@@ -1,101 +1,102 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License, Version 1.0 only
	* (the "License"). You may not use this file except in compliance
	* with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright 2007 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/

	#ifndef _LIBSPL_SYS_SYSMACROS_H
	#define _LIBSPL_SYS_SYSMACROS_H

	#include_next <sys/sysmacros.h>

	/* common macros */
	#ifndef MIN
	#define MIN(a, b) ((a) < (b) ? (a) : (b))
	#endif
	#ifndef MAX
	#define MAX(a, b) ((a) < (b) ? (b) : (a))
	#endif
	#ifndef ABS
	#define ABS(a) ((a) < 0 ? -(a) : (a))
	#endif
	#ifndef ARRAY_SIZE
	#define ARRAY_SIZE(a) (sizeof (a) / sizeof (a[0]))
	#endif
	#ifndef DIV_ROUND_UP
	#define DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d))
	#endif

	#define makedevice(maj, min) makedev(maj, min)
	#define _sysconf(a) sysconf(a)

	/*
	* Compatibility macros/typedefs needed for Solaris -> Linux port
	*/
	-#define P2ALIGN(x, align) ((x) & -(align))
	+// Deprecated. Use P2ALIGN_TYPED instead.
	+// #define P2ALIGN(x, align) ((x) & -(align))
	#define P2CROSS(x, y, align) (((x) ^ (y)) > (align) - 1)
	#define P2ROUNDUP(x, align) ((((x) - 1) \| ((align) - 1)) + 1)
	#define P2BOUNDARY(off, len, align) \
	(((off) ^ ((off) + (len) - 1)) > (align) - 1)
	#define P2PHASE(x, align) ((x) & ((align) - 1))
	#define P2NPHASE(x, align) (-(x) & ((align) - 1))
	#define P2NPHASE_TYPED(x, align, type) \
	(-(type)(x) & ((type)(align) - 1))
	#define ISP2(x) (((x) & ((x) - 1)) == 0)
	#define IS_P2ALIGNED(v, a) ((((uintptr_t)(v)) & ((uintptr_t)(a) - 1)) == 0)

	/*
	* Typed version of the P2* macros. These macros should be used to ensure
	* that the result is correctly calculated based on the data type of (x),
	* which is passed in as the last argument, regardless of the data
	* type of the alignment. For example, if (x) is of type uint64_t,
	* and we want to round it up to a page boundary using "PAGESIZE" as
	* the alignment, we can do either
	* P2ROUNDUP(x, (uint64_t)PAGESIZE)
	* or
	* P2ROUNDUP_TYPED(x, PAGESIZE, uint64_t)
	*/
	#define P2ALIGN_TYPED(x, align, type) \
	((type)(x) & -(type)(align))
	#define P2PHASE_TYPED(x, align, type) \
	((type)(x) & ((type)(align) - 1))
	#define P2NPHASE_TYPED(x, align, type) \
	(-(type)(x) & ((type)(align) - 1))
	#define P2ROUNDUP_TYPED(x, align, type) \
	((((type)(x) - 1) \| ((type)(align) - 1)) + 1)
	#define P2END_TYPED(x, align, type) \
	(-(~(type)(x) & -(type)(align)))
	#define P2PHASEUP_TYPED(x, align, phase, type) \
	((type)(phase) - (((type)(phase) - (type)(x)) & -(type)(align)))
	#define P2CROSS_TYPED(x, y, align, type) \
	(((type)(x) ^ (type)(y)) > (type)(align) - 1)
	#define P2SAMEHIGHBIT_TYPED(x, y, type) \
	(((type)(x) ^ (type)(y)) < ((type)(x) & (type)(y)))


	/* avoid any possibility of clashing with <stddef.h> version */
	#if defined(_KERNEL) && !defined(_KMEMUSER) && !defined(offsetof)
	#define offsetof(s, m) ((size_t)(&(((s *)0)->m)))
	#endif

	#endif /* _LIBSPL_SYS_SYSMACROS_H */
	diff --git a/sys/contrib/openzfs/lib/libspl/include/sys/backtrace.h b/sys/contrib/openzfs/lib/libspl/include/sys/backtrace.h
	new file mode 100644
	index 000000000000..f9869ffc9e1a
	--- /dev/null
	+++ b/sys/contrib/openzfs/lib/libspl/include/sys/backtrace.h
	@@ -0,0 +1,32 @@
	+/*
	+ * CDDL HEADER START
	+ *
	+ * The contents of this file are subject to the terms of the
	+ * Common Development and Distribution License, Version 1.0 only
	+ * (the "License"). You may not use this file except in compliance
	+ * with the License.
	+ *
	+ * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	+ * or https://opensource.org/licenses/CDDL-1.0.
	+ * See the License for the specific language governing permissions
	+ * and limitations under the License.
	+ *
	+ * When distributing Covered Code, include this CDDL HEADER in each
	+ * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	+ * If applicable, add the following below this CDDL HEADER, with the
	+ * fields enclosed by brackets "[]" replaced with your own identifying
	+ * information: Portions Copyright [yyyy] [name of copyright owner]
	+ *
	+ * CDDL HEADER END
	+ */
	+/*
	+ * Copyright (c) 2024, Rob Norris <robn@despairlabs.com>
	+ * Copyright (c) 2024, Klara Inc.
	+ */
	+
	+#ifndef _LIBSPL_SYS_BACKTRACE_H
	+#define _LIBSPL_SYS_BACKTRACE_H
	+
	+void libspl_backtrace(int fd);
	+
	+#endif
	diff --git a/sys/contrib/openzfs/lib/libtpool/thread_pool.c b/sys/contrib/openzfs/lib/libtpool/thread_pool.c
	index 7802f8c1750f..9bf9cdf5dc84 100644
	--- a/sys/contrib/openzfs/lib/libtpool/thread_pool.c
	+++ b/sys/contrib/openzfs/lib/libtpool/thread_pool.c
	@@ -1,599 +1,611 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright 2008 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	*/

	#include <stdlib.h>
	#include <signal.h>
	#include <errno.h>
	#include <assert.h>
	#include <limits.h>
	#include "thread_pool_impl.h"

	static pthread_mutex_t thread_pool_lock = PTHREAD_MUTEX_INITIALIZER;
	static tpool_t *thread_pools = NULL;

	static void
	delete_pool(tpool_t *tpool)
	{
	tpool_job_t *job;

	ASSERT(tpool->tp_current == 0 && tpool->tp_active == NULL);

	/*
	* Unlink the pool from the global list of all pools.
	*/
	(void) pthread_mutex_lock(&thread_pool_lock);
	if (thread_pools == tpool)
	thread_pools = tpool->tp_forw;
	if (thread_pools == tpool)
	thread_pools = NULL;
	else {
	tpool->tp_back->tp_forw = tpool->tp_forw;
	tpool->tp_forw->tp_back = tpool->tp_back;
	}
	pthread_mutex_unlock(&thread_pool_lock);

	/*
	* There should be no pending jobs, but just in case...
	*/
	for (job = tpool->tp_head; job != NULL; job = tpool->tp_head) {
	tpool->tp_head = job->tpj_next;
	free(job);
	}
	(void) pthread_attr_destroy(&tpool->tp_attr);
	free(tpool);
	}

	/*
	* Worker thread is terminating.
	*/
	static void
	worker_cleanup(void *arg)
	{
	tpool_t tpool = (tpool_t )arg;

	if (--tpool->tp_current == 0 &&
	(tpool->tp_flags & (TP_DESTROY \| TP_ABANDON))) {
	if (tpool->tp_flags & TP_ABANDON) {
	pthread_mutex_unlock(&tpool->tp_mutex);
	delete_pool(tpool);
	return;
	}
	if (tpool->tp_flags & TP_DESTROY)
	(void) pthread_cond_broadcast(&tpool->tp_busycv);
	}
	pthread_mutex_unlock(&tpool->tp_mutex);
	}

	static void
	notify_waiters(tpool_t *tpool)
	{
	if (tpool->tp_head == NULL && tpool->tp_active == NULL) {
	tpool->tp_flags &= ~TP_WAIT;
	(void) pthread_cond_broadcast(&tpool->tp_waitcv);
	}
	}

	/*
	* Called by a worker thread on return from a tpool_dispatch()d job.
	*/
	static void
	job_cleanup(void *arg)
	{
	tpool_t tpool = (tpool_t )arg;

	pthread_t my_tid = pthread_self();
	tpool_active_t *activep;
	tpool_active_t **activepp;

	pthread_mutex_lock(&tpool->tp_mutex);
	for (activepp = &tpool->tp_active; ; activepp = &activep->tpa_next) {
	activep = *activepp;
	if (activep->tpa_tid == my_tid) {
	*activepp = activep->tpa_next;
	break;
	}
	}
	if (tpool->tp_flags & TP_WAIT)
	notify_waiters(tpool);
	}

	static void *
	tpool_worker(void *arg)
	{
	tpool_t tpool = (tpool_t )arg;
	int elapsed;
	tpool_job_t *job;
	void (func)(void );
	tpool_active_t active;

	pthread_mutex_lock(&tpool->tp_mutex);
	pthread_cleanup_push(worker_cleanup, tpool);

	/*
	* This is the worker's main loop.
	* It will only be left if a timeout or an error has occurred.
	*/
	active.tpa_tid = pthread_self();
	for (;;) {
	elapsed = 0;
	tpool->tp_idle++;
	if (tpool->tp_flags & TP_WAIT)
	notify_waiters(tpool);
	while ((tpool->tp_head == NULL \|\|
	(tpool->tp_flags & TP_SUSPEND)) &&
	!(tpool->tp_flags & (TP_DESTROY \| TP_ABANDON))) {
	if (tpool->tp_current <= tpool->tp_minimum \|\|
	tpool->tp_linger == 0) {
	(void) pthread_cond_wait(&tpool->tp_workcv,
	&tpool->tp_mutex);
	} else {
	struct timespec ts;

	clock_gettime(CLOCK_REALTIME, &ts);
	ts.tv_sec += tpool->tp_linger;

	if (pthread_cond_timedwait(&tpool->tp_workcv,
	&tpool->tp_mutex, &ts) != 0) {
	elapsed = 1;
	break;
	}
	}
	}
	tpool->tp_idle--;
	if (tpool->tp_flags & TP_DESTROY)
	break;
	if (tpool->tp_flags & TP_ABANDON) {
	/* can't abandon a suspended pool */
	if (tpool->tp_flags & TP_SUSPEND) {
	tpool->tp_flags &= ~TP_SUSPEND;
	(void) pthread_cond_broadcast(
	&tpool->tp_workcv);
	}
	if (tpool->tp_head == NULL)
	break;
	}
	if ((job = tpool->tp_head) != NULL &&
	!(tpool->tp_flags & TP_SUSPEND)) {
	elapsed = 0;
	func = job->tpj_func;
	arg = job->tpj_arg;
	tpool->tp_head = job->tpj_next;
	if (job == tpool->tp_tail)
	tpool->tp_tail = NULL;
	tpool->tp_njobs--;
	active.tpa_next = tpool->tp_active;
	tpool->tp_active = &active;
	pthread_mutex_unlock(&tpool->tp_mutex);
	pthread_cleanup_push(job_cleanup, tpool);
	free(job);

	sigset_t maskset;
	(void) pthread_sigmask(SIG_SETMASK, NULL, &maskset);

	/*
	* Call the specified function.
	*/
	func(arg);
	/*
	* We don't know what this thread has been doing,
	* so we reset its signal mask and cancellation
	* state back to the values prior to calling func().
	*/
	(void) pthread_sigmask(SIG_SETMASK, &maskset, NULL);
	(void) pthread_setcanceltype(PTHREAD_CANCEL_DEFERRED,
	NULL);
	(void) pthread_setcancelstate(PTHREAD_CANCEL_ENABLE,
	NULL);
	pthread_cleanup_pop(1);
	}
	if (elapsed && tpool->tp_current > tpool->tp_minimum) {
	/*
	* We timed out and there is no work to be done
	* and the number of workers exceeds the minimum.
	* Exit now to reduce the size of the pool.
	*/
	break;
	}
	}
	pthread_cleanup_pop(1);
	return (arg);
	}

	/*
	* Create a worker thread, with default signals blocked.
	*/
	static int
	create_worker(tpool_t *tpool)
	{
	pthread_t thread;
	sigset_t oset;
	int error;

	(void) pthread_sigmask(SIG_SETMASK, NULL, &oset);
	error = pthread_create(&thread, &tpool->tp_attr, tpool_worker, tpool);
	(void) pthread_sigmask(SIG_SETMASK, &oset, NULL);
	return (error);
	}


	/*
	* pthread_attr_clone: make a copy of a pthread_attr_t. When old_attr
	* is NULL initialize the cloned attr using default values.
	*/
	static int
	pthread_attr_clone(pthread_attr_t attr, const pthread_attr_t old_attr)
	{
	int error;

	error = pthread_attr_init(attr);
	if (error \|\| (old_attr == NULL))
	return (error);

	#ifdef __GLIBC__
	cpu_set_t cpuset;
	size_t cpusetsize = sizeof (cpuset);
	error = pthread_attr_getaffinity_np(old_attr, cpusetsize, &cpuset);
	if (error == 0)
	error = pthread_attr_setaffinity_np(attr, cpusetsize, &cpuset);
	if (error)
	goto error;
	#endif /* __GLIBC__ */

	int detachstate;
	error = pthread_attr_getdetachstate(old_attr, &detachstate);
	if (error == 0)
	error = pthread_attr_setdetachstate(attr, detachstate);
	if (error)
	goto error;

	size_t guardsize;
	error = pthread_attr_getguardsize(old_attr, &guardsize);
	if (error == 0)
	error = pthread_attr_setguardsize(attr, guardsize);
	if (error)
	goto error;

	int inheritsched;
	error = pthread_attr_getinheritsched(old_attr, &inheritsched);
	if (error == 0)
	error = pthread_attr_setinheritsched(attr, inheritsched);
	if (error)
	goto error;

	struct sched_param param;
	error = pthread_attr_getschedparam(old_attr, &param);
	if (error == 0)
	error = pthread_attr_setschedparam(attr, &param);
	if (error)
	goto error;

	int policy;
	error = pthread_attr_getschedpolicy(old_attr, &policy);
	if (error == 0)
	error = pthread_attr_setschedpolicy(attr, policy);
	if (error)
	goto error;

	int scope;
	error = pthread_attr_getscope(old_attr, &scope);
	if (error == 0)
	error = pthread_attr_setscope(attr, scope);
	if (error)
	goto error;

	void *stackaddr;
	size_t stacksize;
	error = pthread_attr_getstack(old_attr, &stackaddr, &stacksize);
	if (error == 0)
	error = pthread_attr_setstack(attr, stackaddr, stacksize);
	if (error)
	goto error;

	return (0);
	error:
	pthread_attr_destroy(attr);
	return (error);
	}

	tpool_t *
	tpool_create(uint_t min_threads, uint_t max_threads, uint_t linger,
	pthread_attr_t *attr)
	{
	tpool_t *tpool;
	void *stackaddr;
	size_t stacksize;
	size_t minstack;
	int error;

	if (min_threads > max_threads \|\| max_threads < 1) {
	errno = EINVAL;
	return (NULL);
	}
	if (attr != NULL) {
	if (pthread_attr_getstack(attr, &stackaddr, &stacksize) != 0) {
	errno = EINVAL;
	return (NULL);
	}
	/*
	* Allow only one thread in the pool with a specified stack.
	* Require threads to have at least the minimum stack size.
	*/
	minstack = PTHREAD_STACK_MIN;
	if (stackaddr != NULL) {
	if (stacksize < minstack \|\| max_threads != 1) {
	errno = EINVAL;
	return (NULL);
	}
	} else if (stacksize != 0 && stacksize < minstack) {
	errno = EINVAL;
	return (NULL);
	}
	}

	tpool = calloc(1, sizeof (*tpool));
	if (tpool == NULL) {
	errno = ENOMEM;
	return (NULL);
	}
	(void) pthread_mutex_init(&tpool->tp_mutex, NULL);
	(void) pthread_cond_init(&tpool->tp_busycv, NULL);
	(void) pthread_cond_init(&tpool->tp_workcv, NULL);
	(void) pthread_cond_init(&tpool->tp_waitcv, NULL);
	tpool->tp_minimum = min_threads;
	tpool->tp_maximum = max_threads;
	tpool->tp_linger = linger;

	/*
	* We cannot just copy the attribute pointer.
	* We need to initialize a new pthread_attr_t structure
	* with the values from the user-supplied pthread_attr_t.
	* If the attribute pointer is NULL, we need to initialize
	* the new pthread_attr_t structure with default values.
	*/
	error = pthread_attr_clone(&tpool->tp_attr, attr);
	if (error) {
	free(tpool);
	errno = error;
	return (NULL);
	}

	/* make all pool threads be detached daemon threads */
	(void) pthread_attr_setdetachstate(&tpool->tp_attr,
	PTHREAD_CREATE_DETACHED);

	/* insert into the global list of all thread pools */
	pthread_mutex_lock(&thread_pool_lock);
	if (thread_pools == NULL) {
	tpool->tp_forw = tpool;
	tpool->tp_back = tpool;
	thread_pools = tpool;
	} else {
	thread_pools->tp_back->tp_forw = tpool;
	tpool->tp_forw = thread_pools;
	tpool->tp_back = thread_pools->tp_back;
	thread_pools->tp_back = tpool;
	}
	pthread_mutex_unlock(&thread_pool_lock);

	return (tpool);
	}

	/*
	* Dispatch a work request to the thread pool.
	* If there are idle workers, awaken one.
	* Else, if the maximum number of workers has
	* not been reached, spawn a new worker thread.
	* Else just return with the job added to the queue.
	*/
	int
	tpool_dispatch(tpool_t tpool, void (func)(void ), void arg)
	{
	tpool_job_t *job;

	ASSERT(!(tpool->tp_flags & (TP_DESTROY \| TP_ABANDON)));

	if ((job = calloc(1, sizeof (*job))) == NULL)
	return (-1);
	job->tpj_next = NULL;
	job->tpj_func = func;
	job->tpj_arg = arg;

	pthread_mutex_lock(&tpool->tp_mutex);

	+ if (!(tpool->tp_flags & TP_SUSPEND)) {
	+ if (tpool->tp_idle > 0)
	+ (void) pthread_cond_signal(&tpool->tp_workcv);
	+ else if (tpool->tp_current >= tpool->tp_maximum) {
	+ /* At worker limit. Leave task on queue */
	+ } else {
	+ if (create_worker(tpool) == 0) {
	+ /* Started a new worker thread */
	+ tpool->tp_current++;
	+ } else if (tpool->tp_current > 0) {
	+ /* Leave task on queue */
	+ } else {
	+ /* Cannot start a single worker! */
	+ pthread_mutex_unlock(&tpool->tp_mutex);
	+ free(job);
	+ return (-1);
	+ }
	+ }
	+ }
	+
	if (tpool->tp_head == NULL)
	tpool->tp_head = job;
	else
	tpool->tp_tail->tpj_next = job;
	tpool->tp_tail = job;
	tpool->tp_njobs++;

	- if (!(tpool->tp_flags & TP_SUSPEND)) {
	- if (tpool->tp_idle > 0)
	- (void) pthread_cond_signal(&tpool->tp_workcv);
	- else if (tpool->tp_current < tpool->tp_maximum &&
	- create_worker(tpool) == 0)
	- tpool->tp_current++;
	- }
	-
	pthread_mutex_unlock(&tpool->tp_mutex);
	return (0);
	}

	static void
	tpool_cleanup(void *arg)
	{
	tpool_t tpool = (tpool_t )arg;

	pthread_mutex_unlock(&tpool->tp_mutex);
	}

	/*
	* Assumes: by the time tpool_destroy() is called no one will use this
	* thread pool in any way and no one will try to dispatch entries to it.
	* Calling tpool_destroy() from a job in the pool will cause deadlock.
	*/
	void
	tpool_destroy(tpool_t *tpool)
	{
	tpool_active_t *activep;

	ASSERT(!tpool_member(tpool));
	ASSERT(!(tpool->tp_flags & (TP_DESTROY \| TP_ABANDON)));

	pthread_mutex_lock(&tpool->tp_mutex);
	pthread_cleanup_push(tpool_cleanup, tpool);

	/* mark the pool as being destroyed; wakeup idle workers */
	tpool->tp_flags \|= TP_DESTROY;
	tpool->tp_flags &= ~TP_SUSPEND;
	(void) pthread_cond_broadcast(&tpool->tp_workcv);

	/* cancel all active workers */
	for (activep = tpool->tp_active; activep; activep = activep->tpa_next)
	(void) pthread_cancel(activep->tpa_tid);

	/* wait for all active workers to finish */
	while (tpool->tp_active != NULL) {
	tpool->tp_flags \|= TP_WAIT;
	(void) pthread_cond_wait(&tpool->tp_waitcv, &tpool->tp_mutex);
	}

	/* the last worker to terminate will wake us up */
	while (tpool->tp_current != 0)
	(void) pthread_cond_wait(&tpool->tp_busycv, &tpool->tp_mutex);

	pthread_cleanup_pop(1); /* pthread_mutex_unlock(&tpool->tp_mutex); */
	delete_pool(tpool);
	}

	/*
	* Like tpool_destroy(), but don't cancel workers or wait for them to finish.
	* The last worker to terminate will delete the pool.
	*/
	void
	tpool_abandon(tpool_t *tpool)
	{
	ASSERT(!(tpool->tp_flags & (TP_DESTROY \| TP_ABANDON)));

	pthread_mutex_lock(&tpool->tp_mutex);
	if (tpool->tp_current == 0) {
	/* no workers, just delete the pool */
	pthread_mutex_unlock(&tpool->tp_mutex);
	delete_pool(tpool);
	} else {
	/* wake up all workers, last one will delete the pool */
	tpool->tp_flags \|= TP_ABANDON;
	tpool->tp_flags &= ~TP_SUSPEND;
	(void) pthread_cond_broadcast(&tpool->tp_workcv);
	pthread_mutex_unlock(&tpool->tp_mutex);
	}
	}

	/*
	* Wait for all jobs to complete.
	* Calling tpool_wait() from a job in the pool will cause deadlock.
	*/
	void
	tpool_wait(tpool_t *tpool)
	{
	ASSERT(!tpool_member(tpool));
	ASSERT(!(tpool->tp_flags & (TP_DESTROY \| TP_ABANDON)));

	pthread_mutex_lock(&tpool->tp_mutex);
	pthread_cleanup_push(tpool_cleanup, tpool);
	while (tpool->tp_head != NULL \|\| tpool->tp_active != NULL) {
	tpool->tp_flags \|= TP_WAIT;
	(void) pthread_cond_wait(&tpool->tp_waitcv, &tpool->tp_mutex);
	ASSERT(!(tpool->tp_flags & (TP_DESTROY \| TP_ABANDON)));
	}
	pthread_cleanup_pop(1); /* pthread_mutex_unlock(&tpool->tp_mutex); */
	}

	void
	tpool_suspend(tpool_t *tpool)
	{
	ASSERT(!(tpool->tp_flags & (TP_DESTROY \| TP_ABANDON)));

	pthread_mutex_lock(&tpool->tp_mutex);
	tpool->tp_flags \|= TP_SUSPEND;
	pthread_mutex_unlock(&tpool->tp_mutex);
	}

	int
	tpool_suspended(tpool_t *tpool)
	{
	int suspended;

	ASSERT(!(tpool->tp_flags & (TP_DESTROY \| TP_ABANDON)));

	pthread_mutex_lock(&tpool->tp_mutex);
	suspended = (tpool->tp_flags & TP_SUSPEND) != 0;
	pthread_mutex_unlock(&tpool->tp_mutex);

	return (suspended);
	}

	void
	tpool_resume(tpool_t *tpool)
	{
	int excess;

	ASSERT(!(tpool->tp_flags & (TP_DESTROY \| TP_ABANDON)));

	pthread_mutex_lock(&tpool->tp_mutex);
	if (!(tpool->tp_flags & TP_SUSPEND)) {
	pthread_mutex_unlock(&tpool->tp_mutex);
	return;
	}
	tpool->tp_flags &= ~TP_SUSPEND;
	(void) pthread_cond_broadcast(&tpool->tp_workcv);
	excess = tpool->tp_njobs - tpool->tp_idle;
	while (excess-- > 0 && tpool->tp_current < tpool->tp_maximum) {
	if (create_worker(tpool) != 0)
	break; /* pthread_create() failed */
	tpool->tp_current++;
	}
	pthread_mutex_unlock(&tpool->tp_mutex);
	}

	int
	tpool_member(tpool_t *tpool)
	{
	pthread_t my_tid = pthread_self();
	tpool_active_t *activep;

	ASSERT(!(tpool->tp_flags & (TP_DESTROY \| TP_ABANDON)));

	pthread_mutex_lock(&tpool->tp_mutex);
	for (activep = tpool->tp_active; activep; activep = activep->tpa_next) {
	if (activep->tpa_tid == my_tid) {
	pthread_mutex_unlock(&tpool->tp_mutex);
	return (1);
	}
	}
	pthread_mutex_unlock(&tpool->tp_mutex);
	return (0);
	}
	diff --git a/sys/contrib/openzfs/lib/libuutil/libuutil.abi b/sys/contrib/openzfs/lib/libuutil/libuutil.abi
	index e942d24c6531..2ed2fb2e41e6 100644
	--- a/sys/contrib/openzfs/lib/libuutil/libuutil.abi
	+++ b/sys/contrib/openzfs/lib/libuutil/libuutil.abi
	@@ -1,2113 +1,2186 @@
	<abi-corpus version='2.0' architecture='elf-amd-x86_64' soname='libuutil.so.3'>
	<elf-needed>
	<dependency name='libc.so.6'/>
	<dependency name='ld-linux-x86-64.so.2'/>
	</elf-needed>
	<elf-function-symbols>
	<elf-symbol name='_sol_getmntent' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_char' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_char_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_int' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_int_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_long' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_long_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_ptr' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_ptr_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_short' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_short_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_uchar_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_uint_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_ulong_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_ushort_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_ptr' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_clear_long_excl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_uchar_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_uint_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_ulong_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_ushort_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_uchar_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_uint_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_ulong_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_ushort_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_uchar_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_uint_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_ulong_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_ushort_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_set_long_excl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_char' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_char_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_int' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_int_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_long' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_long_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_ptr' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_ptr_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_short' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_short_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_ptr' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_add' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_destroy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_destroy_nodes' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_find' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_first' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_insert' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_insert_here' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_is_empty' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_last' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_nearest' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_numnodes' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_swap' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_update' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_update_gt' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_update_lt' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_walk' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='get_system_hostid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='getexecname' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='getextmntent' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='getmntany' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='getzoneid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libspl_assertf' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	+ <elf-symbol name='libspl_backtrace' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libspl_set_assert_ok' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_destroy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_head' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_insert_after' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_insert_before' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_insert_head' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_insert_tail' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_is_empty' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_link_active' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_link_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_link_replace' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_move_tail' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_next' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_prev' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_remove_head' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_remove_tail' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_tail' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='membar_consumer' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='membar_enter' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='membar_exit' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='membar_producer' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='membar_sync' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='mkdirp' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='print_timestamp' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='spl_pagesize' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='strlcat' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='strlcpy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_avl_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_avl_destroy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_avl_find' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_avl_first' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_avl_insert' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_avl_last' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_avl_lockup' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_avl_nearest_next' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_avl_nearest_prev' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_avl_next' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_avl_node_fini' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_avl_node_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_avl_numnodes' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_avl_pool_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_avl_pool_destroy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_avl_prev' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_avl_release' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_avl_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_avl_teardown' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_avl_walk' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_avl_walk_end' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_avl_walk_next' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_avl_walk_start' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_check_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_error' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_free' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_destroy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_find' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_first' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_insert' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_insert_after' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_insert_before' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_last' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_lockup' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_nearest_next' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_nearest_prev' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_next' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_node_fini' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_node_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_numnodes' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_pool_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_pool_destroy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_prev' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_release' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_teardown' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_walk' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_walk_end' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_walk_next' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_list_walk_start' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_memdup' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_msprintf' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_panic' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_set_error' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_strbw' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_strcaseeq' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_strdup' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_streq' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_strerror' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_strndup' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='uu_zalloc' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	</elf-function-symbols>
	<abi-instr address-size='64' path='lib/libspl/assert.c' language='LANG_C99'>
	+ <typedef-decl name='__pid_t' type-id='95e97e5e' id='3629bad8'/>
	+ <function-decl name='libspl_backtrace' mangled-name='libspl_backtrace' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libspl_backtrace'>
	+ <parameter type-id='95e97e5e'/>
	+ <return type-id='48b5725f'/>
	+ </function-decl>
	+ <function-decl name='getpid' visibility='default' binding='global' size-in-bits='64'>
	+ <return type-id='3629bad8'/>
	+ </function-decl>
	+ <function-decl name='gettid' visibility='default' binding='global' size-in-bits='64'>
	+ <return type-id='3629bad8'/>
	+ </function-decl>
	+ <function-decl name='prctl' visibility='default' binding='global' size-in-bits='64'>
	+ <parameter type-id='95e97e5e'/>
	+ <parameter is-variadic='yes'/>
	+ <return type-id='95e97e5e'/>
	+ </function-decl>
	<function-decl name='libspl_set_assert_ok' mangled-name='libspl_set_assert_ok' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libspl_set_assert_ok'>
	<parameter type-id='c19b74c3' name='val'/>
	<return type-id='48b5725f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/atomic.c' language='LANG_C99'>
	<typedef-decl name='int16_t' type-id='03896e23' id='23bd8cb5'/>
	<typedef-decl name='int32_t' type-id='33f57a65' id='3ff5601b'/>
	<typedef-decl name='uint16_t' type-id='253c2d2a' id='149c6638'/>
	<typedef-decl name='__int16_t' type-id='a2185560' id='03896e23'/>
	<typedef-decl name='__uint16_t' type-id='8efea9e5' id='253c2d2a'/>
	<typedef-decl name='__int32_t' type-id='95e97e5e' id='33f57a65'/>
	<typedef-decl name='__ssize_t' type-id='bd54fe1a' id='41060289'/>
	<typedef-decl name='ssize_t' type-id='41060289' id='79a0948f'/>
	<qualified-type-def type-id='149c6638' volatile='yes' id='5120c5f7'/>
	<pointer-type-def type-id='5120c5f7' size-in-bits='64' id='93977ae7'/>
	<qualified-type-def type-id='8f92235e' volatile='yes' id='430e0681'/>
	<pointer-type-def type-id='430e0681' size-in-bits='64' id='3a147f31'/>
	<qualified-type-def type-id='b96825af' volatile='yes' id='84ff7d66'/>
	<pointer-type-def type-id='84ff7d66' size-in-bits='64' id='aa323ea4'/>
	<qualified-type-def type-id='ee1f298e' volatile='yes' id='6f7e09cb'/>
	<pointer-type-def type-id='6f7e09cb' size-in-bits='64' id='64698d33'/>
	<qualified-type-def type-id='48b5725f' volatile='yes' id='b0b3cbf9'/>
	<pointer-type-def type-id='b0b3cbf9' size-in-bits='64' id='fe09dd29'/>
	<function-decl name='atomic_inc_8' mangled-name='atomic_inc_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_inc_16' mangled-name='atomic_inc_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_16'>
	<parameter type-id='93977ae7' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_inc_32' mangled-name='atomic_inc_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_32'>
	<parameter type-id='3a147f31' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_inc_ulong' mangled-name='atomic_inc_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_dec_8' mangled-name='atomic_dec_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_dec_16' mangled-name='atomic_dec_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_16'>
	<parameter type-id='93977ae7' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_dec_32' mangled-name='atomic_dec_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_32'>
	<parameter type-id='3a147f31' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_dec_ulong' mangled-name='atomic_dec_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_add_ptr' mangled-name='atomic_add_ptr' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_ptr'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='79a0948f' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_add_8' mangled-name='atomic_add_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='ee31ee44' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_add_16' mangled-name='atomic_add_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='23bd8cb5' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_add_32' mangled-name='atomic_add_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_32'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='3ff5601b' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_sub_ptr' mangled-name='atomic_sub_ptr' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_ptr'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='79a0948f' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_sub_8' mangled-name='atomic_sub_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='ee31ee44' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_sub_16' mangled-name='atomic_sub_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='23bd8cb5' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_sub_32' mangled-name='atomic_sub_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_32'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='3ff5601b' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_or_8' mangled-name='atomic_or_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_or_16' mangled-name='atomic_or_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_or_32' mangled-name='atomic_or_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_32'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='8f92235e' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_or_ulong' mangled-name='atomic_or_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_and_8' mangled-name='atomic_and_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_and_16' mangled-name='atomic_and_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_and_32' mangled-name='atomic_and_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_32'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='8f92235e' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_and_ulong' mangled-name='atomic_and_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_inc_8_nv' mangled-name='atomic_inc_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_inc_16_nv' mangled-name='atomic_inc_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_inc_32_nv' mangled-name='atomic_inc_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_inc_ulong_nv' mangled-name='atomic_inc_ulong_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_ulong_nv'>
	<parameter type-id='64698d33' name='target'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_dec_8_nv' mangled-name='atomic_dec_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_dec_16_nv' mangled-name='atomic_dec_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_dec_32_nv' mangled-name='atomic_dec_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_dec_ulong_nv' mangled-name='atomic_dec_ulong_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_ulong_nv'>
	<parameter type-id='64698d33' name='target'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_add_ptr_nv' mangled-name='atomic_add_ptr_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_ptr_nv'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='79a0948f' name='bits'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='atomic_add_8_nv' mangled-name='atomic_add_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='ee31ee44' name='bits'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_add_16_nv' mangled-name='atomic_add_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='23bd8cb5' name='bits'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_add_32_nv' mangled-name='atomic_add_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='3ff5601b' name='bits'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_add_long_nv' mangled-name='atomic_add_long_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_long_nv'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='bd54fe1a' name='bits'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_sub_ptr_nv' mangled-name='atomic_sub_ptr_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_ptr_nv'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='79a0948f' name='bits'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='atomic_sub_8_nv' mangled-name='atomic_sub_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='ee31ee44' name='bits'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_sub_16_nv' mangled-name='atomic_sub_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='23bd8cb5' name='bits'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_sub_32_nv' mangled-name='atomic_sub_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='3ff5601b' name='bits'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_sub_long_nv' mangled-name='atomic_sub_long_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_long_nv'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='bd54fe1a' name='bits'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_or_8_nv' mangled-name='atomic_or_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='bits'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_or_16_nv' mangled-name='atomic_or_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='bits'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_or_32_nv' mangled-name='atomic_or_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='8f92235e' name='bits'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_or_ulong_nv' mangled-name='atomic_or_ulong_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_ulong_nv'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='bits'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_and_8_nv' mangled-name='atomic_and_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='bits'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_and_16_nv' mangled-name='atomic_and_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='bits'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_and_32_nv' mangled-name='atomic_and_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='8f92235e' name='bits'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_and_ulong_nv' mangled-name='atomic_and_ulong_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_ulong_nv'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='bits'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_cas_ptr' mangled-name='atomic_cas_ptr' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_ptr'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='eaa32e2f' name='exp'/>
	<parameter type-id='eaa32e2f' name='des'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='atomic_cas_8' mangled-name='atomic_cas_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='exp'/>
	<parameter type-id='b96825af' name='des'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_cas_16' mangled-name='atomic_cas_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='exp'/>
	<parameter type-id='149c6638' name='des'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_cas_32' mangled-name='atomic_cas_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_32'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='8f92235e' name='exp'/>
	<parameter type-id='8f92235e' name='des'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_cas_ulong' mangled-name='atomic_cas_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='exp'/>
	<parameter type-id='ee1f298e' name='des'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_swap_8' mangled-name='atomic_swap_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='bits'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_swap_16' mangled-name='atomic_swap_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='bits'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_swap_32' mangled-name='atomic_swap_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_32'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='8f92235e' name='bits'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_swap_ulong' mangled-name='atomic_swap_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='bits'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_swap_ptr' mangled-name='atomic_swap_ptr' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_ptr'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='eaa32e2f' name='bits'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='atomic_set_long_excl' mangled-name='atomic_set_long_excl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_set_long_excl'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='3502e3ff' name='value'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='atomic_clear_long_excl' mangled-name='atomic_clear_long_excl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_clear_long_excl'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='3502e3ff' name='value'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='membar_enter' mangled-name='membar_enter' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='membar_enter'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='membar_producer' mangled-name='membar_producer' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='membar_producer'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='membar_consumer' mangled-name='membar_consumer' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='membar_consumer'>
	<return type-id='48b5725f'/>
	</function-decl>
	<type-decl name='unsigned short int' size-in-bits='16' id='8efea9e5'/>
	</abi-instr>
	+ <abi-instr address-size='64' path='lib/libspl/backtrace.c' language='LANG_C99'>
	+ <qualified-type-def type-id='eaa32e2f' const='yes' id='83be723c'/>
	+ <pointer-type-def type-id='83be723c' size-in-bits='64' id='7acd98a2'/>
	+ <function-decl name='backtrace' visibility='default' binding='global' size-in-bits='64'>
	+ <parameter type-id='63e171df'/>
	+ <parameter type-id='95e97e5e'/>
	+ <return type-id='95e97e5e'/>
	+ </function-decl>
	+ <function-decl name='backtrace_symbols_fd' visibility='default' binding='global' size-in-bits='64'>
	+ <parameter type-id='7acd98a2'/>
	+ <parameter type-id='95e97e5e'/>
	+ <parameter type-id='95e97e5e'/>
	+ <return type-id='48b5725f'/>
	+ </function-decl>
	+ <function-decl name='write' visibility='default' binding='global' size-in-bits='64'>
	+ <parameter type-id='95e97e5e'/>
	+ <parameter type-id='eaa32e2f'/>
	+ <parameter type-id='b59d7dce'/>
	+ <return type-id='79a0948f'/>
	+ </function-decl>
	+ </abi-instr>
	<abi-instr address-size='64' path='lib/libspl/getexecname.c' language='LANG_C99'>
	<function-decl name='getexecname' mangled-name='getexecname' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='getexecname'>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='getexecname_impl' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='26a90f95'/>
	<return type-id='79a0948f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/list.c' language='LANG_C99'>
	<typedef-decl name='list_node_t' type-id='b0b5e45e' id='b21843b2'/>
	<typedef-decl name='list_t' type-id='e824dae9' id='0899125f'/>
	<class-decl name='list_node' size-in-bits='128' is-struct='yes' visibility='default' id='b0b5e45e'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='next' type-id='b03eadb4' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='prev' type-id='b03eadb4' visibility='default'/>
	</data-member>
	</class-decl>
	- <class-decl name='list' size-in-bits='256' is-struct='yes' visibility='default' id='e824dae9'>
	+ <class-decl name='list' size-in-bits='192' is-struct='yes' visibility='default' id='e824dae9'>
	<data-member access='public' layout-offset-in-bits='0'>
	- <var-decl name='list_size' type-id='b59d7dce' visibility='default'/>
	- </data-member>
	- <data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='list_offset' type-id='b59d7dce' visibility='default'/>
	</data-member>
	- <data-member access='public' layout-offset-in-bits='128'>
	+ <data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='list_head' type-id='b0b5e45e' visibility='default'/>
	</data-member>
	</class-decl>
	<pointer-type-def type-id='b0b5e45e' size-in-bits='64' id='b03eadb4'/>
	<pointer-type-def type-id='b21843b2' size-in-bits='64' id='ccc38265'/>
	<pointer-type-def type-id='0899125f' size-in-bits='64' id='352ec160'/>
	<function-decl name='list_create' mangled-name='list_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_create'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='b59d7dce' name='size'/>
	<parameter type-id='b59d7dce' name='offset'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_destroy' mangled-name='list_destroy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_destroy'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_insert_after' mangled-name='list_insert_after' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_insert_after'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<parameter type-id='eaa32e2f' name='nobject'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_insert_before' mangled-name='list_insert_before' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_insert_before'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<parameter type-id='eaa32e2f' name='nobject'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_insert_head' mangled-name='list_insert_head' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_insert_head'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_insert_tail' mangled-name='list_insert_tail' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_insert_tail'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_remove' mangled-name='list_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_remove'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_remove_head' mangled-name='list_remove_head' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_remove_head'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_remove_tail' mangled-name='list_remove_tail' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_remove_tail'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_head' mangled-name='list_head' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_head'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_tail' mangled-name='list_tail' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_tail'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_next' mangled-name='list_next' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_next'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_prev' mangled-name='list_prev' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_prev'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_move_tail' mangled-name='list_move_tail' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_move_tail'>
	<parameter type-id='352ec160' name='dst'/>
	<parameter type-id='352ec160' name='src'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_link_replace' mangled-name='list_link_replace' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_link_replace'>
	<parameter type-id='ccc38265' name='lold'/>
	<parameter type-id='ccc38265' name='lnew'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_link_init' mangled-name='list_link_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_link_init'>
	<parameter type-id='ccc38265' name='ln'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_link_active' mangled-name='list_link_active' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_link_active'>
	<parameter type-id='ccc38265' name='ln'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='list_is_empty' mangled-name='list_is_empty' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_is_empty'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/mkdirp.c' language='LANG_C99'>
	<typedef-decl name='__mode_t' type-id='f0981eeb' id='e1c52942'/>
	<typedef-decl name='mode_t' type-id='e1c52942' id='d50d396c'/>
	<typedef-decl name='wchar_t' type-id='95e97e5e' id='928221d2'/>
	<qualified-type-def type-id='26a90f95' restrict='yes' id='266fe297'/>
	<qualified-type-def type-id='928221d2' const='yes' id='effb3702'/>
	<pointer-type-def type-id='effb3702' size-in-bits='64' id='f077d3f8'/>
	<qualified-type-def type-id='f077d3f8' restrict='yes' id='598aab80'/>
	<pointer-type-def type-id='928221d2' size-in-bits='64' id='323d93c1'/>
	<qualified-type-def type-id='323d93c1' restrict='yes' id='f1358bc3'/>
	<function-decl name='calloc' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='strdup' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='strrchr' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='access' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='__mbstowcs_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='f1358bc3'/>
	<parameter type-id='9d26089a'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='__wcstombs_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='266fe297'/>
	<parameter type-id='598aab80'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='mkdir' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='e1c52942'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='mkdirp' mangled-name='mkdirp' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='mkdirp'>
	<parameter type-id='80f4b756' name='d'/>
	<parameter type-id='d50d396c' name='mode'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/os/linux/getexecname.c' language='LANG_C99'>
	<function-decl name='__readlink_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9d26089a'/>
	<parameter type-id='266fe297'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='79a0948f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/os/linux/gethostid.c' language='LANG_C99'>
	<type-decl name='long long unsigned int' size-in-bits='64' id='3a47d82b'/>
	<pointer-type-def type-id='26a90f95' size-in-bits='64' id='9b23c9ad'/>
	<qualified-type-def type-id='9b23c9ad' restrict='yes' id='8c85230f'/>
	<function-decl name='fclose' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='822cd80b'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='strtoull' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9d26089a'/>
	<parameter type-id='8c85230f'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='3a47d82b'/>
	</function-decl>
	<function-decl name='getenv' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='close' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='__open_too_many_args' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='__open_missing_mode' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='__read_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='79a0948f'/>
	</function-decl>
	<function-decl name='get_system_hostid' mangled-name='get_system_hostid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='get_system_hostid'>
	<return type-id='7359adad'/>
	</function-decl>
	<type-decl name='unsigned long int' size-in-bits='64' id='7359adad'/>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/os/linux/getmntany.c' language='LANG_C99'>
	+ <array-type-def dimensions='1' type-id='38b51b3c' size-in-bits='832' id='02b72c00'>
	+ <subrange length='13' type-id='7359adad' id='487fded1'/>
	+ </array-type-def>
	<array-type-def dimensions='1' type-id='03085adc' size-in-bits='192' id='083f8d58'>
	<subrange length='3' type-id='7359adad' id='56f209d2'/>
	</array-type-def>
	+ <class-decl name='__locale_data' is-struct='yes' visibility='default' is-declaration-only='yes' id='23de8b96'/>
	+ <array-type-def dimensions='1' type-id='80f4b756' size-in-bits='832' id='39e6f84a'>
	+ <subrange length='13' type-id='7359adad' id='487fded1'/>
	+ </array-type-def>
	<class-decl name='mnttab' size-in-bits='256' is-struct='yes' visibility='default' id='1b055409'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='mnt_special' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='mnt_mountp' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='mnt_fstype' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='mnt_mntopts' type-id='26a90f95' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='extmnttab' size-in-bits='320' is-struct='yes' visibility='default' id='0c544dc0'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='mnt_special' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='mnt_mountp' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='mnt_fstype' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='mnt_mntopts' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='mnt_major' type-id='3502e3ff' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='mnt_minor' type-id='3502e3ff' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='mntent' size-in-bits='320' is-struct='yes' visibility='default' id='56fe4a37'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='mnt_fsname' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='mnt_dir' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='mnt_type' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='mnt_opts' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='mnt_freq' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='mnt_passno' type-id='95e97e5e' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='stat64' size-in-bits='1152' is-struct='yes' visibility='default' id='0bbec9cd'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='st_dev' type-id='35ed8932' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='st_ino' type-id='71288a47' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='st_nlink' type-id='80f0b9df' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='st_mode' type-id='e1c52942' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='224'>
	<var-decl name='st_uid' type-id='cc5fcceb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='st_gid' type-id='d94ec6d9' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='__pad0' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='st_rdev' type-id='35ed8932' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='st_size' type-id='79989e9c' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='st_blksize' type-id='d3f10a7f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='st_blocks' type-id='4e711bf1' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='576'>
	<var-decl name='st_atim' type-id='a9c79a1f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='704'>
	<var-decl name='st_mtim' type-id='a9c79a1f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='832'>
	<var-decl name='st_ctim' type-id='a9c79a1f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='960'>
	<var-decl name='__glibc_reserved' type-id='083f8d58' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='__dev_t' type-id='7359adad' id='35ed8932'/>
	<typedef-decl name='__uid_t' type-id='f0981eeb' id='cc5fcceb'/>
	<typedef-decl name='__gid_t' type-id='f0981eeb' id='d94ec6d9'/>
	<typedef-decl name='__ino64_t' type-id='7359adad' id='71288a47'/>
	<typedef-decl name='__nlink_t' type-id='7359adad' id='80f0b9df'/>
	<typedef-decl name='__blksize_t' type-id='bd54fe1a' id='d3f10a7f'/>
	<typedef-decl name='__blkcnt64_t' type-id='bd54fe1a' id='4e711bf1'/>
	<typedef-decl name='__syscall_slong_t' type-id='bd54fe1a' id='03085adc'/>
	+ <class-decl name='__locale_struct' size-in-bits='1856' is-struct='yes' visibility='default' id='90cc1ce3'>
	+ <data-member access='public' layout-offset-in-bits='0'>
	+ <var-decl name='__locales' type-id='02b72c00' visibility='default'/>
	+ </data-member>
	+ <data-member access='public' layout-offset-in-bits='832'>
	+ <var-decl name='__ctype_b' type-id='31347b7a' visibility='default'/>
	+ </data-member>
	+ <data-member access='public' layout-offset-in-bits='896'>
	+ <var-decl name='__ctype_tolower' type-id='6d60f45d' visibility='default'/>
	+ </data-member>
	+ <data-member access='public' layout-offset-in-bits='960'>
	+ <var-decl name='__ctype_toupper' type-id='6d60f45d' visibility='default'/>
	+ </data-member>
	+ <data-member access='public' layout-offset-in-bits='1024'>
	+ <var-decl name='__names' type-id='39e6f84a' visibility='default'/>
	+ </data-member>
	+ </class-decl>
	+ <typedef-decl name='__locale_t' type-id='f01e1813' id='b7ac9b5f'/>
	+ <typedef-decl name='locale_t' type-id='b7ac9b5f' id='973a4f8d'/>
	<class-decl name='timespec' size-in-bits='128' is-struct='yes' visibility='default' id='a9c79a1f'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='tv_sec' type-id='65eda9c0' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='tv_nsec' type-id='03085adc' visibility='default'/>
	</data-member>
	</class-decl>
	+ <pointer-type-def type-id='23de8b96' size-in-bits='64' id='38b51b3c'/>
	+ <pointer-type-def type-id='90cc1ce3' size-in-bits='64' id='f01e1813'/>
	+ <qualified-type-def type-id='95e97e5e' const='yes' id='2448a865'/>
	+ <pointer-type-def type-id='2448a865' size-in-bits='64' id='6d60f45d'/>
	+ <qualified-type-def type-id='8efea9e5' const='yes' id='3beb2af4'/>
	+ <pointer-type-def type-id='3beb2af4' size-in-bits='64' id='31347b7a'/>
	<pointer-type-def type-id='0c544dc0' size-in-bits='64' id='394fc496'/>
	<pointer-type-def type-id='56fe4a37' size-in-bits='64' id='b6b61d2f'/>
	<qualified-type-def type-id='b6b61d2f' restrict='yes' id='3cad23cd'/>
	<pointer-type-def type-id='1b055409' size-in-bits='64' id='9d424d31'/>
	<pointer-type-def type-id='0bbec9cd' size-in-bits='64' id='62f7a03d'/>
	<qualified-type-def type-id='62f7a03d' restrict='yes' id='f1cadedf'/>
	+ <class-decl name='__locale_data' is-struct='yes' visibility='default' is-declaration-only='yes' id='23de8b96'/>
	+ <function-decl name='uselocale' visibility='default' binding='global' size-in-bits='64'>
	+ <parameter type-id='973a4f8d'/>
	+ <return type-id='973a4f8d'/>
	+ </function-decl>
	<function-decl name='getmntent_r' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='e75a27e9'/>
	<parameter type-id='3cad23cd'/>
	<parameter type-id='266fe297'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='b6b61d2f'/>
	</function-decl>
	<function-decl name='feof' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='822cd80b'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	- <function-decl name='strerror' visibility='default' binding='global' size-in-bits='64'>
	+ <function-decl name='strerror_l' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	+ <parameter type-id='973a4f8d'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='__fprintf_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='e75a27e9'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='9d26089a'/>
	<parameter is-variadic='yes'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='stat64' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9d26089a'/>
	<parameter type-id='f1cadedf'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='getmntany' mangled-name='getmntany' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='getmntany'>
	<parameter type-id='822cd80b' name='fp'/>
	<parameter type-id='9d424d31' name='mgetp'/>
	<parameter type-id='9d424d31' name='mrefp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='_sol_getmntent' mangled-name='_sol_getmntent' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='_sol_getmntent'>
	<parameter type-id='822cd80b' name='fp'/>
	<parameter type-id='9d424d31' name='mgetp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='getextmntent' mangled-name='getextmntent' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='getextmntent'>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='394fc496' name='entry'/>
	<parameter type-id='62f7a03d' name='statbuf'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/os/linux/zone.c' language='LANG_C99'>
	<typedef-decl name='zoneid_t' type-id='3502e3ff' id='4da03624'/>
	<function-decl name='strtoul' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9d26089a'/>
	<parameter type-id='8c85230f'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='7359adad'/>
	</function-decl>
	<function-decl name='getzoneid' mangled-name='getzoneid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='getzoneid'>
	<return type-id='4da03624'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/page.c' language='LANG_C99'>
	<function-decl name='sysconf' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<return type-id='bd54fe1a'/>
	</function-decl>
	<function-decl name='spl_pagesize' mangled-name='spl_pagesize' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='spl_pagesize'>
	<return type-id='b59d7dce'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/strlcat.c' language='LANG_C99'>
	<function-decl name='strlcat' mangled-name='strlcat' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='strlcat'>
	<parameter type-id='26a90f95' name='dst'/>
	<parameter type-id='80f4b756' name='src'/>
	<parameter type-id='b59d7dce' name='dstsize'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/timestamp.c' language='LANG_C99'>
	<typedef-decl name='nl_item' type-id='95e97e5e' id='03b79a94'/>
	<typedef-decl name='__time_t' type-id='bd54fe1a' id='65eda9c0'/>
	<class-decl name='tm' size-in-bits='448' is-struct='yes' visibility='default' id='dddf6ca2'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='tm_sec' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='tm_min' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='tm_hour' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='tm_mday' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='tm_mon' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='160'>
	<var-decl name='tm_year' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='tm_wday' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='224'>
	<var-decl name='tm_yday' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='tm_isdst' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='tm_gmtoff' type-id='bd54fe1a' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='tm_zone' type-id='80f4b756' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='time_t' type-id='65eda9c0' id='c9d12d66'/>
	<qualified-type-def type-id='c9d12d66' const='yes' id='588b3216'/>
	<pointer-type-def type-id='588b3216' size-in-bits='64' id='9f201474'/>
	<qualified-type-def type-id='9f201474' restrict='yes' id='d6e2847c'/>
	<qualified-type-def type-id='dddf6ca2' const='yes' id='e824a34f'/>
	<pointer-type-def type-id='e824a34f' size-in-bits='64' id='d6ad37ff'/>
	<qualified-type-def type-id='d6ad37ff' restrict='yes' id='f8c6051d'/>
	<pointer-type-def type-id='c9d12d66' size-in-bits='64' id='b2eb2c3f'/>
	<pointer-type-def type-id='dddf6ca2' size-in-bits='64' id='d915a820'/>
	<qualified-type-def type-id='d915a820' restrict='yes' id='f099ad08'/>
	<function-decl name='nl_langinfo' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='03b79a94'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='time' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b2eb2c3f'/>
	<return type-id='c9d12d66'/>
	</function-decl>
	<function-decl name='strftime' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='266fe297'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='9d26089a'/>
	<parameter type-id='f8c6051d'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='localtime_r' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='d6e2847c'/>
	<parameter type-id='f099ad08'/>
	<return type-id='d915a820'/>
	</function-decl>
	<function-decl name='__printf_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<parameter is-variadic='yes'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='print_timestamp' mangled-name='print_timestamp' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='print_timestamp'>
	<parameter type-id='3502e3ff' name='timestamp_fmt'/>
	<return type-id='48b5725f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libuutil/uu_alloc.c' language='LANG_C99'>
	<type-decl name='char' size-in-bits='8' id='a84c031d'/>
	<type-decl name='unsigned int' size-in-bits='32' id='f0981eeb'/>
	<type-decl name='unsigned long int' size-in-bits='64' id='7359adad'/>
	<type-decl name='variadic parameter type' id='2c1145c5'/>
	<type-decl name='void' id='48b5725f'/>
	<typedef-decl name='uint_t' type-id='f0981eeb' id='3502e3ff'/>
	<typedef-decl name='size_t' type-id='7359adad' id='b59d7dce'/>
	<pointer-type-def type-id='a84c031d' size-in-bits='64' id='26a90f95'/>
	<qualified-type-def type-id='a84c031d' const='yes' id='9b45d938'/>
	<pointer-type-def type-id='9b45d938' size-in-bits='64' id='80f4b756'/>
	<pointer-type-def type-id='48b5725f' size-in-bits='64' id='eaa32e2f'/>
	<function-decl name='uu_set_error' mangled-name='uu_set_error' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_set_error'>
	<parameter type-id='3502e3ff'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='malloc' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b59d7dce'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='free' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='eaa32e2f'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='strlen' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='strnlen' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='uu_zalloc' mangled-name='uu_zalloc' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_zalloc'>
	<parameter type-id='b59d7dce' name='n'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_free' mangled-name='uu_free' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_free'>
	<parameter type-id='eaa32e2f' name='p'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='uu_strdup' mangled-name='uu_strdup' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_strdup'>
	<parameter type-id='80f4b756' name='str'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='uu_strndup' mangled-name='uu_strndup' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_strndup'>
	<parameter type-id='80f4b756' name='s'/>
	<parameter type-id='b59d7dce' name='n'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='uu_memdup' mangled-name='uu_memdup' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_memdup'>
	<parameter type-id='eaa32e2f' name='buf'/>
	<parameter type-id='b59d7dce' name='sz'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_msprintf' mangled-name='uu_msprintf' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_msprintf'>
	<parameter type-id='80f4b756' name='format'/>
	<parameter is-variadic='yes'/>
	<return type-id='26a90f95'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libuutil/uu_avl.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='bf311473' size-in-bits='128' id='f0f65199'>
	<subrange length='2' type-id='7359adad' id='52efc4ef'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='320' id='36c46961'>
	<subrange length='40' type-id='7359adad' id='8f80b239'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='32' id='8e0573fd'>
	<subrange length='4' type-id='7359adad' id='16fe7105'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='512' id='59daf3ef'>
	<subrange length='64' type-id='7359adad' id='b10be967'/>
	</array-type-def>
	<type-decl name='int' size-in-bits='32' id='95e97e5e'/>
	<type-decl name='long int' size-in-bits='64' id='bd54fe1a'/>
	<type-decl name='short int' size-in-bits='16' id='a2185560'/>
	<type-decl name='signed char' size-in-bits='8' id='28577a57'/>
	<array-type-def dimensions='1' type-id='e475ab95' size-in-bits='192' id='0ce65a8b'>
	<subrange length='3' type-id='7359adad' id='56f209d2'/>
	</array-type-def>
	<type-decl name='unsigned char' size-in-bits='8' id='002ac4a6'/>
	<typedef-decl name='uu_compare_fn_t' type-id='add6e811' id='40f93560'/>
	<typedef-decl name='uu_walk_fn_t' type-id='96ee24a5' id='9d1aa0dc'/>
	<typedef-decl name='uu_avl_pool_t' type-id='12a530a8' id='7f84e390'/>
	<typedef-decl name='uu_avl_t' type-id='4af029d1' id='bb7f0973'/>
	<class-decl name='uu_avl_node' size-in-bits='192' is-struct='yes' visibility='default' id='f65f4326'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='uan_opaque' type-id='0ce65a8b' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='uu_avl_node_t' type-id='f65f4326' id='73a65116'/>
	<typedef-decl name='uu_avl_walk_t' type-id='e70a39e3' id='edd8457b'/>
	<typedef-decl name='uu_avl_index_t' type-id='e475ab95' id='5d7f5fc8'/>
	<class-decl name='uu_avl_walk' size-in-bits='320' is-struct='yes' visibility='default' id='e70a39e3'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='uaw_next' type-id='5842d146' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='uaw_prev' type-id='5842d146' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='uaw_avl' type-id='a5c21a38' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='uaw_next_result' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='uaw_dir' type-id='ee31ee44' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='264'>
	<var-decl name='uaw_robust' type-id='b96825af' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='uu_avl' size-in-bits='960' is-struct='yes' visibility='default' id='4af029d1'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='ua_next' type-id='a5c21a38' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='ua_prev' type-id='a5c21a38' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='ua_pool' type-id='de82c773' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='ua_parent' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='ua_debug' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='264'>
	<var-decl name='ua_index' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='ua_tree' type-id='b351119f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='640'>
	<var-decl name='ua_null_walk' type-id='edd8457b' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='uu_avl_pool' size-in-bits='2176' is-struct='yes' visibility='default' id='12a530a8'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='uap_next' type-id='de82c773' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='uap_prev' type-id='de82c773' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='uap_name' type-id='59daf3ef' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='640'>
	<var-decl name='uap_nodeoffset' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='704'>
	<var-decl name='uap_objsize' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='768'>
	<var-decl name='uap_cmp' type-id='d502b39f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='832'>
	<var-decl name='uap_debug' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='840'>
	<var-decl name='uap_last_index' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='896'>
	<var-decl name='uap_lock' type-id='7a6844eb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1216'>
	<var-decl name='uap_null_avl' type-id='bb7f0973' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='avl_tree_t' type-id='b351119f' id='f20fbd51'/>
	<typedef-decl name='avl_index_t' type-id='e475ab95' id='fba6cb51'/>
	<class-decl name='avl_node' size-in-bits='192' is-struct='yes' visibility='default' id='428b67b3'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='avl_child' type-id='f0f65199' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='avl_pcb' type-id='e475ab95' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='avl_tree' size-in-bits='320' is-struct='yes' visibility='default' id='b351119f'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='avl_root' type-id='bf311473' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='avl_compar' type-id='585e1de9' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='avl_offset' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='avl_numnodes' type-id='ee1f298e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='avl_pad' type-id='b59d7dce' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='ulong_t' type-id='7359adad' id='ee1f298e'/>
	<typedef-decl name='uintptr_t' type-id='7359adad' id='e475ab95'/>
	<union-decl name='pthread_mutexattr_t' size-in-bits='32' naming-typedef-id='8afd6070' visibility='default' id='7300eb00'>
	<data-member access='public'>
	<var-decl name='__size' type-id='8e0573fd' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='__align' type-id='95e97e5e' visibility='default'/>
	</data-member>
	</union-decl>
	<typedef-decl name='pthread_mutexattr_t' type-id='7300eb00' id='8afd6070'/>
	<union-decl name='pthread_mutex_t' size-in-bits='320' naming-typedef-id='7a6844eb' visibility='default' id='70681f9b'>
	<data-member access='public'>
	<var-decl name='__data' type-id='4c734837' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='__size' type-id='36c46961' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='__align' type-id='bd54fe1a' visibility='default'/>
	</data-member>
	</union-decl>
	<typedef-decl name='pthread_mutex_t' type-id='70681f9b' id='7a6844eb'/>
	<typedef-decl name='int8_t' type-id='2171a512' id='ee31ee44'/>
	<typedef-decl name='uint8_t' type-id='c51d6389' id='b96825af'/>
	<typedef-decl name='uint32_t' type-id='62f1140c' id='8f92235e'/>
	<class-decl name='__pthread_mutex_s' size-in-bits='320' is-struct='yes' visibility='default' id='4c734837'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='__lock' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='__count' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='__owner' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='__nusers' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='__kind' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='160'>
	<var-decl name='__spins' type-id='a2185560' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='176'>
	<var-decl name='__elision' type-id='a2185560' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='__list' type-id='518fb49c' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__pthread_internal_list' size-in-bits='128' is-struct='yes' visibility='default' id='0e01899c'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='__prev' type-id='4d98cd5a' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='__next' type-id='4d98cd5a' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='__pthread_list_t' type-id='0e01899c' id='518fb49c'/>
	<typedef-decl name='__int8_t' type-id='28577a57' id='2171a512'/>
	<typedef-decl name='__uint8_t' type-id='002ac4a6' id='c51d6389'/>
	<typedef-decl name='__uint32_t' type-id='f0981eeb' id='62f1140c'/>
	<pointer-type-def type-id='0e01899c' size-in-bits='64' id='4d98cd5a'/>
	<pointer-type-def type-id='fba6cb51' size-in-bits='64' id='32adbf30'/>
	<pointer-type-def type-id='428b67b3' size-in-bits='64' id='bf311473'/>
	<pointer-type-def type-id='b351119f' size-in-bits='64' id='716943c7'/>
	<pointer-type-def type-id='f20fbd51' size-in-bits='64' id='a3681dea'/>
	<qualified-type-def type-id='8afd6070' const='yes' id='1d853360'/>
	<pointer-type-def type-id='1d853360' size-in-bits='64' id='c2afbd7e'/>
	<pointer-type-def type-id='96ee24a5' size-in-bits='64' id='585e1de9'/>
	<pointer-type-def type-id='7a6844eb' size-in-bits='64' id='18c91f9e'/>
	<pointer-type-def type-id='5d7f5fc8' size-in-bits='64' id='813a2225'/>
	<pointer-type-def type-id='73a65116' size-in-bits='64' id='2dc35b9d'/>
	<pointer-type-def type-id='7f84e390' size-in-bits='64' id='de82c773'/>
	<pointer-type-def type-id='bb7f0973' size-in-bits='64' id='a5c21a38'/>
	<pointer-type-def type-id='edd8457b' size-in-bits='64' id='5842d146'/>
	<pointer-type-def type-id='40f93560' size-in-bits='64' id='d502b39f'/>
	<pointer-type-def type-id='9d1aa0dc' size-in-bits='64' id='30a42b6d'/>
	<pointer-type-def type-id='eaa32e2f' size-in-bits='64' id='63e171df'/>
	<function-decl name='uu_check_name' mangled-name='uu_check_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_check_name'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='uu_panic' mangled-name='uu_panic' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_panic'>
	<parameter type-id='80f4b756'/>
	<parameter is-variadic='yes'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='avl_create' mangled-name='avl_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_create'>
	<parameter type-id='a3681dea'/>
	<parameter type-id='585e1de9'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='avl_find' mangled-name='avl_find' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_find'>
	<parameter type-id='a3681dea'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='32adbf30'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='avl_insert' mangled-name='avl_insert' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_insert'>
	<parameter type-id='a3681dea'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='fba6cb51'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='avl_first' mangled-name='avl_first' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_first'>
	<parameter type-id='a3681dea'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='avl_last' mangled-name='avl_last' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_last'>
	<parameter type-id='a3681dea'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='avl_nearest' mangled-name='avl_nearest' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_nearest'>
	<parameter type-id='a3681dea'/>
	<parameter type-id='fba6cb51'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='avl_remove' mangled-name='avl_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_remove'>
	<parameter type-id='a3681dea'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='avl_numnodes' mangled-name='avl_numnodes' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_numnodes'>
	<parameter type-id='a3681dea'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='avl_destroy_nodes' mangled-name='avl_destroy_nodes' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_destroy_nodes'>
	<parameter type-id='a3681dea'/>
	<parameter type-id='63e171df'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='avl_destroy' mangled-name='avl_destroy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_destroy'>
	<parameter type-id='a3681dea'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='avl_walk' mangled-name='avl_walk' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_walk'>
	<parameter type-id='716943c7'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='strlcpy' mangled-name='strlcpy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='strlcpy'>
	<parameter type-id='26a90f95'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='pthread_mutex_init' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='18c91f9e'/>
	<parameter type-id='c2afbd7e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_mutex_destroy' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='18c91f9e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_mutex_lock' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='18c91f9e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_mutex_unlock' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='18c91f9e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='uu_avl_pool_create' mangled-name='uu_avl_pool_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_avl_pool_create'>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='b59d7dce' name='objsize'/>
	<parameter type-id='b59d7dce' name='nodeoffset'/>
	<parameter type-id='d502b39f' name='compare_func'/>
	<parameter type-id='8f92235e' name='flags'/>
	<return type-id='de82c773'/>
	</function-decl>
	<function-decl name='uu_avl_pool_destroy' mangled-name='uu_avl_pool_destroy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_avl_pool_destroy'>
	<parameter type-id='de82c773' name='pp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='uu_avl_node_init' mangled-name='uu_avl_node_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_avl_node_init'>
	<parameter type-id='eaa32e2f' name='base'/>
	<parameter type-id='2dc35b9d' name='np'/>
	<parameter type-id='de82c773' name='pp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='uu_avl_node_fini' mangled-name='uu_avl_node_fini' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_avl_node_fini'>
	<parameter type-id='eaa32e2f' name='base'/>
	<parameter type-id='2dc35b9d' name='np'/>
	<parameter type-id='de82c773' name='pp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='uu_avl_create' mangled-name='uu_avl_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_avl_create'>
	<parameter type-id='de82c773' name='pp'/>
	<parameter type-id='eaa32e2f' name='parent'/>
	<parameter type-id='8f92235e' name='flags'/>
	<return type-id='a5c21a38'/>
	</function-decl>
	<function-decl name='uu_avl_destroy' mangled-name='uu_avl_destroy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_avl_destroy'>
	<parameter type-id='a5c21a38' name='ap'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='uu_avl_numnodes' mangled-name='uu_avl_numnodes' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_avl_numnodes'>
	<parameter type-id='a5c21a38' name='ap'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='uu_avl_first' mangled-name='uu_avl_first' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_avl_first'>
	<parameter type-id='a5c21a38' name='ap'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_avl_last' mangled-name='uu_avl_last' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_avl_last'>
	<parameter type-id='a5c21a38' name='ap'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_avl_next' mangled-name='uu_avl_next' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_avl_next'>
	<parameter type-id='a5c21a38' name='ap'/>
	<parameter type-id='eaa32e2f' name='node'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_avl_prev' mangled-name='uu_avl_prev' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_avl_prev'>
	<parameter type-id='a5c21a38' name='ap'/>
	<parameter type-id='eaa32e2f' name='node'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_avl_walk_start' mangled-name='uu_avl_walk_start' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_avl_walk_start'>
	<parameter type-id='a5c21a38' name='ap'/>
	<parameter type-id='8f92235e' name='flags'/>
	<return type-id='5842d146'/>
	</function-decl>
	<function-decl name='uu_avl_walk_next' mangled-name='uu_avl_walk_next' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_avl_walk_next'>
	<parameter type-id='5842d146' name='wp'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_avl_walk_end' mangled-name='uu_avl_walk_end' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_avl_walk_end'>
	<parameter type-id='5842d146' name='wp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='uu_avl_walk' mangled-name='uu_avl_walk' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_avl_walk'>
	<parameter type-id='a5c21a38' name='ap'/>
	<parameter type-id='30a42b6d' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<parameter type-id='8f92235e' name='flags'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='uu_avl_remove' mangled-name='uu_avl_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_avl_remove'>
	<parameter type-id='a5c21a38' name='ap'/>
	<parameter type-id='eaa32e2f' name='elem'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='uu_avl_teardown' mangled-name='uu_avl_teardown' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_avl_teardown'>
	<parameter type-id='a5c21a38' name='ap'/>
	<parameter type-id='63e171df' name='cookie'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_avl_find' mangled-name='uu_avl_find' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_avl_find'>
	<parameter type-id='a5c21a38' name='ap'/>
	<parameter type-id='eaa32e2f' name='elem'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<parameter type-id='813a2225' name='out'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_avl_insert' mangled-name='uu_avl_insert' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_avl_insert'>
	<parameter type-id='a5c21a38' name='ap'/>
	<parameter type-id='eaa32e2f' name='elem'/>
	<parameter type-id='5d7f5fc8' name='idx'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='uu_avl_nearest_next' mangled-name='uu_avl_nearest_next' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_avl_nearest_next'>
	<parameter type-id='a5c21a38' name='ap'/>
	<parameter type-id='5d7f5fc8' name='idx'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_avl_nearest_prev' mangled-name='uu_avl_nearest_prev' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_avl_nearest_prev'>
	<parameter type-id='a5c21a38' name='ap'/>
	<parameter type-id='5d7f5fc8' name='idx'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_avl_lockup' mangled-name='uu_avl_lockup' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_avl_lockup'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='uu_avl_release' mangled-name='uu_avl_release' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_avl_release'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-type size-in-bits='64' id='96ee24a5'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='add6e811'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libuutil/uu_ident.c' language='LANG_C99'>
	<function-decl name='strchr' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='26a90f95'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libuutil/uu_list.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='e475ab95' size-in-bits='128' id='d0e9cdae'>
	<subrange length='2' type-id='7359adad' id='52efc4ef'/>
	</array-type-def>
	<typedef-decl name='uu_list_pool_t' type-id='55168cab' id='38a2549d'/>
	<typedef-decl name='uu_list_t' type-id='1d04bdf0' id='82e88484'/>
	<class-decl name='uu_list_node' size-in-bits='128' is-struct='yes' visibility='default' id='f8f3cec5'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='uln_opaque' type-id='d0e9cdae' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='uu_list_node_t' type-id='f8f3cec5' id='c4dc472f'/>
	<typedef-decl name='uu_list_walk_t' type-id='b80e3208' id='9fed32d2'/>
	<typedef-decl name='uu_list_index_t' type-id='e475ab95' id='f0dd35ff'/>
	<class-decl name='uu_list_node_impl' size-in-bits='128' is-struct='yes' visibility='default' id='700a795c'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='uln_next' type-id='5af1298a' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='uln_prev' type-id='5af1298a' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='uu_list_node_impl_t' type-id='700a795c' id='8e5864b0'/>
	<class-decl name='uu_list_walk' size-in-bits='320' is-struct='yes' visibility='default' id='b80e3208'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='ulw_next' type-id='4d848103' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='ulw_prev' type-id='4d848103' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='ulw_list' type-id='0c0b229b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='ulw_dir' type-id='ee31ee44' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='200'>
	<var-decl name='ulw_robust' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='ulw_next_result' type-id='a085247f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='uu_list' size-in-bits='896' is-struct='yes' visibility='default' id='1d04bdf0'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='ul_next' type-id='0c0b229b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='ul_prev' type-id='0c0b229b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='ul_pool' type-id='0941e04e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='ul_parent' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='ul_offset' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='ul_numnodes' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='ul_debug' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='392'>
	<var-decl name='ul_sorted' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='400'>
	<var-decl name='ul_index' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='ul_null_node' type-id='8e5864b0' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='576'>
	<var-decl name='ul_null_walk' type-id='9fed32d2' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='uu_list_pool' size-in-bits='2112' is-struct='yes' visibility='default' id='55168cab'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='ulp_next' type-id='0941e04e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='ulp_prev' type-id='0941e04e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='ulp_name' type-id='59daf3ef' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='640'>
	<var-decl name='ulp_nodeoffset' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='704'>
	<var-decl name='ulp_objsize' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='768'>
	<var-decl name='ulp_cmp' type-id='d502b39f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='832'>
	<var-decl name='ulp_debug' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='840'>
	<var-decl name='ulp_last_index' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='896'>
	<var-decl name='ulp_lock' type-id='7a6844eb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1216'>
	<var-decl name='ulp_null_list' type-id='82e88484' visibility='default'/>
	</data-member>
	</class-decl>
	<pointer-type-def type-id='f0dd35ff' size-in-bits='64' id='ecbc0046'/>
	<pointer-type-def type-id='700a795c' size-in-bits='64' id='5af1298a'/>
	<pointer-type-def type-id='8e5864b0' size-in-bits='64' id='a085247f'/>
	<pointer-type-def type-id='c4dc472f' size-in-bits='64' id='dbe143f4'/>
	<pointer-type-def type-id='38a2549d' size-in-bits='64' id='0941e04e'/>
	<pointer-type-def type-id='82e88484' size-in-bits='64' id='0c0b229b'/>
	<pointer-type-def type-id='9fed32d2' size-in-bits='64' id='4d848103'/>
	<function-decl name='uu_list_pool_create' mangled-name='uu_list_pool_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_pool_create'>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='b59d7dce' name='objsize'/>
	<parameter type-id='b59d7dce' name='nodeoffset'/>
	<parameter type-id='d502b39f' name='compare_func'/>
	<parameter type-id='8f92235e' name='flags'/>
	<return type-id='0941e04e'/>
	</function-decl>
	<function-decl name='uu_list_pool_destroy' mangled-name='uu_list_pool_destroy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_pool_destroy'>
	<parameter type-id='0941e04e' name='pp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='uu_list_node_init' mangled-name='uu_list_node_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_node_init'>
	<parameter type-id='eaa32e2f' name='base'/>
	<parameter type-id='dbe143f4' name='np_arg'/>
	<parameter type-id='0941e04e' name='pp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='uu_list_node_fini' mangled-name='uu_list_node_fini' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_node_fini'>
	<parameter type-id='eaa32e2f' name='base'/>
	<parameter type-id='dbe143f4' name='np_arg'/>
	<parameter type-id='0941e04e' name='pp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='uu_list_create' mangled-name='uu_list_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_create'>
	<parameter type-id='0941e04e' name='pp'/>
	<parameter type-id='eaa32e2f' name='parent'/>
	<parameter type-id='8f92235e' name='flags'/>
	<return type-id='0c0b229b'/>
	</function-decl>
	<function-decl name='uu_list_destroy' mangled-name='uu_list_destroy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_destroy'>
	<parameter type-id='0c0b229b' name='lp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='uu_list_insert' mangled-name='uu_list_insert' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_insert'>
	<parameter type-id='0c0b229b' name='lp'/>
	<parameter type-id='eaa32e2f' name='elem'/>
	<parameter type-id='f0dd35ff' name='idx'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='uu_list_find' mangled-name='uu_list_find' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_find'>
	<parameter type-id='0c0b229b' name='lp'/>
	<parameter type-id='eaa32e2f' name='elem'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<parameter type-id='ecbc0046' name='out'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_list_nearest_next' mangled-name='uu_list_nearest_next' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_nearest_next'>
	<parameter type-id='0c0b229b' name='lp'/>
	<parameter type-id='f0dd35ff' name='idx'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_list_nearest_prev' mangled-name='uu_list_nearest_prev' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_nearest_prev'>
	<parameter type-id='0c0b229b' name='lp'/>
	<parameter type-id='f0dd35ff' name='idx'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_list_walk_start' mangled-name='uu_list_walk_start' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_walk_start'>
	<parameter type-id='0c0b229b' name='lp'/>
	<parameter type-id='8f92235e' name='flags'/>
	<return type-id='4d848103'/>
	</function-decl>
	<function-decl name='uu_list_walk_next' mangled-name='uu_list_walk_next' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_walk_next'>
	<parameter type-id='4d848103' name='wp'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_list_walk_end' mangled-name='uu_list_walk_end' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_walk_end'>
	<parameter type-id='4d848103' name='wp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='uu_list_walk' mangled-name='uu_list_walk' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_walk'>
	<parameter type-id='0c0b229b' name='lp'/>
	<parameter type-id='30a42b6d' name='func'/>
	<parameter type-id='eaa32e2f' name='private'/>
	<parameter type-id='8f92235e' name='flags'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='uu_list_remove' mangled-name='uu_list_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_remove'>
	<parameter type-id='0c0b229b' name='lp'/>
	<parameter type-id='eaa32e2f' name='elem'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='uu_list_teardown' mangled-name='uu_list_teardown' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_teardown'>
	<parameter type-id='0c0b229b' name='lp'/>
	<parameter type-id='63e171df' name='cookie'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_list_insert_before' mangled-name='uu_list_insert_before' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_insert_before'>
	<parameter type-id='0c0b229b' name='lp'/>
	<parameter type-id='eaa32e2f' name='target'/>
	<parameter type-id='eaa32e2f' name='elem'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='uu_list_insert_after' mangled-name='uu_list_insert_after' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_insert_after'>
	<parameter type-id='0c0b229b' name='lp'/>
	<parameter type-id='eaa32e2f' name='target'/>
	<parameter type-id='eaa32e2f' name='elem'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='uu_list_numnodes' mangled-name='uu_list_numnodes' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_numnodes'>
	<parameter type-id='0c0b229b' name='lp'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='uu_list_first' mangled-name='uu_list_first' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_first'>
	<parameter type-id='0c0b229b' name='lp'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_list_last' mangled-name='uu_list_last' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_last'>
	<parameter type-id='0c0b229b' name='lp'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_list_next' mangled-name='uu_list_next' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_next'>
	<parameter type-id='0c0b229b' name='lp'/>
	<parameter type-id='eaa32e2f' name='elem'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_list_prev' mangled-name='uu_list_prev' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_prev'>
	<parameter type-id='0c0b229b' name='lp'/>
	<parameter type-id='eaa32e2f' name='elem'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_list_lockup' mangled-name='uu_list_lockup' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_lockup'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='uu_list_release' mangled-name='uu_list_release' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_list_release'>
	<return type-id='48b5725f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libuutil/uu_misc.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='8' id='89feb1ec'>
	<subrange length='1' type-id='7359adad' id='52f813b4'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='160' id='664ac0b7'>
	<subrange length='20' type-id='7359adad' id='fdca39cf'/>
	</array-type-def>
	<class-decl name='_IO_codecvt' is-struct='yes' visibility='default' is-declaration-only='yes' id='a4036571'/>
	<class-decl name='_IO_marker' is-struct='yes' visibility='default' is-declaration-only='yes' id='010ae0b9'/>
	<class-decl name='_IO_wide_data' is-struct='yes' visibility='default' is-declaration-only='yes' id='79bd3751'/>
	<class-decl name='__va_list_tag' size-in-bits='192' is-struct='yes' visibility='default' id='d5027220'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='gp_offset' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='fp_offset' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='overflow_arg_area' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='reg_save_area' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<type-decl name='unsigned short int' size-in-bits='16' id='8efea9e5'/>
	<typedef-decl name='pthread_t' type-id='7359adad' id='4051f5e7'/>
	<typedef-decl name='pthread_key_t' type-id='f0981eeb' id='2de5383b'/>
	<typedef-decl name='__off_t' type-id='bd54fe1a' id='79989e9c'/>
	<typedef-decl name='__off64_t' type-id='bd54fe1a' id='724e4de6'/>
	<typedef-decl name='FILE' type-id='ec1ed955' id='aa12d1ba'/>
	<typedef-decl name='_IO_lock_t' type-id='48b5725f' id='bb4788fa'/>
	<class-decl name='_IO_FILE' size-in-bits='1728' is-struct='yes' visibility='default' id='ec1ed955'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='_flags' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='_IO_read_ptr' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='_IO_read_end' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='_IO_read_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='_IO_write_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='_IO_write_ptr' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='_IO_write_end' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='_IO_buf_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='_IO_buf_end' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='576'>
	<var-decl name='_IO_save_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='640'>
	<var-decl name='_IO_backup_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='704'>
	<var-decl name='_IO_save_end' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='768'>
	<var-decl name='_markers' type-id='e4c6fa61' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='832'>
	<var-decl name='_chain' type-id='dca988a5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='896'>
	<var-decl name='_fileno' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='928'>
	<var-decl name='_flags2' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='960'>
	<var-decl name='_old_offset' type-id='79989e9c' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1024'>
	<var-decl name='_cur_column' type-id='8efea9e5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1040'>
	<var-decl name='_vtable_offset' type-id='28577a57' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1048'>
	<var-decl name='_shortbuf' type-id='89feb1ec' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1088'>
	<var-decl name='_lock' type-id='cecf4ea7' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1152'>
	<var-decl name='_offset' type-id='724e4de6' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1216'>
	<var-decl name='_codecvt' type-id='570f8c59' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1280'>
	<var-decl name='_wide_data' type-id='c65a1f29' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1344'>
	<var-decl name='_freeres_list' type-id='dca988a5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1408'>
	<var-decl name='_freeres_buf' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1472'>
	<var-decl name='__pad5' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1536'>
	<var-decl name='_mode' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1568'>
	<var-decl name='_unused2' type-id='664ac0b7' visibility='default'/>
	</data-member>
	</class-decl>
	<pointer-type-def type-id='aa12d1ba' size-in-bits='64' id='822cd80b'/>
	<qualified-type-def type-id='822cd80b' restrict='yes' id='e75a27e9'/>
	<pointer-type-def type-id='ec1ed955' size-in-bits='64' id='dca988a5'/>
	<pointer-type-def type-id='a4036571' size-in-bits='64' id='570f8c59'/>
	<pointer-type-def type-id='bb4788fa' size-in-bits='64' id='cecf4ea7'/>
	<pointer-type-def type-id='010ae0b9' size-in-bits='64' id='e4c6fa61'/>
	<pointer-type-def type-id='79bd3751' size-in-bits='64' id='c65a1f29'/>
	<pointer-type-def type-id='d5027220' size-in-bits='64' id='b7f2d5e6'/>
	<qualified-type-def type-id='80f4b756' restrict='yes' id='9d26089a'/>
	<pointer-type-def type-id='95e97e5e' size-in-bits='64' id='7292109c'/>
	<pointer-type-def type-id='2de5383b' size-in-bits='64' id='ce04b822'/>
	<pointer-type-def type-id='ee076206' size-in-bits='64' id='953b12f8'/>
	<pointer-type-def type-id='c5c76c9c' size-in-bits='64' id='b7f9d8e6'/>
	<class-decl name='_IO_codecvt' is-struct='yes' visibility='default' is-declaration-only='yes' id='a4036571'/>
	<class-decl name='_IO_marker' is-struct='yes' visibility='default' is-declaration-only='yes' id='010ae0b9'/>
	<class-decl name='_IO_wide_data' is-struct='yes' visibility='default' is-declaration-only='yes' id='79bd3751'/>
	<function-decl name='__errno_location' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='7292109c'/>
	</function-decl>
	<function-decl name='dcgettext' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='pthread_self' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='4051f5e7'/>
	</function-decl>
	<function-decl name='pthread_key_create' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='ce04b822'/>
	<parameter type-id='b7f9d8e6'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_getspecific' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='2de5383b'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='pthread_setspecific' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='2de5383b'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_atfork' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='953b12f8'/>
	<parameter type-id='953b12f8'/>
	<parameter type-id='953b12f8'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='abort' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='pause' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='__vfprintf_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='e75a27e9'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='9d26089a'/>
	<parameter type-id='b7f2d5e6'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='uu_error' mangled-name='uu_error' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_error'>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='uu_strerror' mangled-name='uu_strerror' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_strerror'>
	<parameter type-id='8f92235e' name='code'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-type size-in-bits='64' id='ee076206'>
	<return type-id='48b5725f'/>
	</function-type>
	<function-type size-in-bits='64' id='c5c76c9c'>
	<parameter type-id='eaa32e2f'/>
	<return type-id='48b5725f'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libuutil/uu_string.c' language='LANG_C99'>
	<type-decl name='unnamed-enum-underlying-type-32' is-anonymous='yes' size-in-bits='32' alignment-in-bits='32' id='9cac1fee'/>
	<enum-decl name='boolean_t' naming-typedef-id='c19b74c3' id='f58c8277'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='B_FALSE' value='0'/>
	<enumerator name='B_TRUE' value='1'/>
	</enum-decl>
	<typedef-decl name='boolean_t' type-id='f58c8277' id='c19b74c3'/>
	<function-decl name='strcmp' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='strncmp' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='strcasecmp' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='uu_streq' mangled-name='uu_streq' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_streq'>
	<parameter type-id='80f4b756' name='a'/>
	<parameter type-id='80f4b756' name='b'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='uu_strcaseeq' mangled-name='uu_strcaseeq' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_strcaseeq'>
	<parameter type-id='80f4b756' name='a'/>
	<parameter type-id='80f4b756' name='b'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='uu_strbw' mangled-name='uu_strbw' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='uu_strbw'>
	<parameter type-id='80f4b756' name='a'/>
	<parameter type-id='80f4b756' name='b'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='module/avl/avl.c' language='LANG_C99'>
	<function-decl name='libspl_assertf' mangled-name='libspl_assertf' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libspl_assertf'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<parameter is-variadic='yes'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='avl_insert_here' mangled-name='avl_insert_here' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_insert_here'>
	<parameter type-id='a3681dea' name='tree'/>
	<parameter type-id='eaa32e2f' name='new_data'/>
	<parameter type-id='eaa32e2f' name='here'/>
	<parameter type-id='95e97e5e' name='direction'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='avl_add' mangled-name='avl_add' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_add'>
	<parameter type-id='a3681dea' name='tree'/>
	<parameter type-id='eaa32e2f' name='new_node'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='avl_update_lt' mangled-name='avl_update_lt' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_update_lt'>
	<parameter type-id='a3681dea' name='t'/>
	<parameter type-id='eaa32e2f' name='obj'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='avl_update_gt' mangled-name='avl_update_gt' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_update_gt'>
	<parameter type-id='a3681dea' name='t'/>
	<parameter type-id='eaa32e2f' name='obj'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='avl_update' mangled-name='avl_update' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_update'>
	<parameter type-id='a3681dea' name='t'/>
	<parameter type-id='eaa32e2f' name='obj'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='avl_swap' mangled-name='avl_swap' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_swap'>
	<parameter type-id='a3681dea' name='tree1'/>
	<parameter type-id='a3681dea' name='tree2'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='avl_is_empty' mangled-name='avl_is_empty' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_is_empty'>
	<parameter type-id='a3681dea' name='tree'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	</abi-instr>
	</abi-corpus>
	diff --git a/sys/contrib/openzfs/lib/libzfs/libzfs.abi b/sys/contrib/openzfs/lib/libzfs/libzfs.abi
	index 2bbaae6345ab..80f4b7439a55 100644
	--- a/sys/contrib/openzfs/lib/libzfs/libzfs.abi
	+++ b/sys/contrib/openzfs/lib/libzfs/libzfs.abi
	@@ -1,9697 +1,9735 @@
	<abi-corpus version='2.0' architecture='elf-amd-x86_64' soname='libzfs.so.4'>
	<elf-needed>
	<dependency name='libzfs_core.so.3'/>
	<dependency name='libnvpair.so.3'/>
	<dependency name='libuuid.so.1'/>
	<dependency name='libblkid.so.1'/>
	<dependency name='libudev.so.1'/>
	<dependency name='libuutil.so.3'/>
	<dependency name='libm.so.6'/>
	<dependency name='libcrypto.so.3'/>
	<dependency name='libz.so.1'/>
	<dependency name='libc.so.6'/>
	<dependency name='ld-linux-x86-64.so.2'/>
	</elf-needed>
	<elf-function-symbols>
	<elf-symbol name='_sol_getmntent' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_char' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_char_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_int' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_int_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_long' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_long_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_ptr' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_ptr_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_short' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_short_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_uchar_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_uint_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_ulong_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_ushort_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_ptr' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_clear_long_excl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_uchar_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_uint_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_ulong_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_ushort_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_uchar_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_uint_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_ulong_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_ushort_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_uchar_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_uint_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_ulong_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_ushort_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_set_long_excl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_char' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_char_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_int' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_int_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_long' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_long_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_ptr' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_ptr_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_short' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_short_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_ptr' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_add' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_destroy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_destroy_nodes' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_find' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_first' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_insert' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_insert_here' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_is_empty' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_last' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_nearest' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_numnodes' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_swap' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_update' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_update_gt' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_update_lt' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='avl_walk' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='bookmark_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='cityhash4' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='color_end' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='color_start' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='dataset_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='dataset_nestcheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='efi_alloc_and_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='efi_alloc_and_read' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='efi_err_check' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='efi_free' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='efi_rescan' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='efi_use_whole_disk' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='efi_write' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='entity_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_2_byteswap' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_2_incremental_byteswap' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_2_incremental_native' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_2_native' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_byteswap' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_fini' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_impl_set' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_incremental_byteswap' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_incremental_native' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_native' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_native_varsize' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fsleep' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='get_dataset_depth' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='get_system_hostid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='getexecname' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='getextmntent' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='getmntany' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='getprop_uint64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='getzoneid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='is_mounted' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='is_mpath_whole_disk' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libpc_error_description' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libspl_assertf' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	+ <elf-symbol name='libspl_backtrace' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libspl_set_assert_ok' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_add_handle' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_envvar_is_set' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_errno' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_error_action' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_error_description' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_error_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_fini' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_free_str_array' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_mnttab_add' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_mnttab_cache' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_mnttab_find' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_mnttab_fini' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_mnttab_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_mnttab_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_print_on_error' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_run_process' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_run_process_get_stdout' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_run_process_get_stdout_nopath' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_destroy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_head' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_insert_after' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_insert_before' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_insert_head' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_insert_tail' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_is_empty' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_link_active' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_link_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_link_replace' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_move_tail' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_next' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_prev' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_remove_head' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_remove_tail' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_tail' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='membar_consumer' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='membar_enter' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='membar_exit' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='membar_producer' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='membar_sync' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='mkdirp' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='mountpoint_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='permset_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='pool_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='print_timestamp' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='printf_color' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='sa_commit_shares' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='sa_disable_share' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='sa_enable_share' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='sa_errorstr' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='sa_is_shared' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='sa_truncate_shares' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='sa_validate_shareopts' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='snapshot_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='spl_pagesize' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='strlcat' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='strlcpy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='tpool_abandon' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='tpool_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='tpool_destroy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='tpool_dispatch' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='tpool_member' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='tpool_resume' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='tpool_suspend' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='tpool_suspended' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='tpool_wait' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='update_vdev_config_dev_strs' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='update_vdev_config_dev_sysfs_path' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='use_color' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_expand_proplist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_name_to_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_align_right' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_column_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_default_numeric' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_default_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_get_table' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_get_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_index_to_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_random_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_readonly' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_string_to_index' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_to_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_user' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='vdev_prop_values' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfeature_depends_on' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfeature_is_supported' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfeature_is_valid_guid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfeature_lookup_guid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfeature_lookup_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_adjust_mount_options' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_allocatable_devs' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_append_partition' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_basename' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_bookmark_exists' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_clone' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_close' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_commit_shares' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_component_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_create_ancestors' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_crypto_attempt_load_keys' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_crypto_clone_check' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_crypto_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_crypto_get_encryption_root' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_crypto_load_key' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_crypto_rewrap' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_crypto_unload_key' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_dataset_exists' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_dataset_name_hidden' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_deleg_canonicalize_perm' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_deleg_verify_nvlist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_deleg_whokey' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_destroy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_destroy_snaps' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_destroy_snaps_nvl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_destroy_snaps_nvl_os' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_dev_flush' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_dev_is_dm' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_dev_is_whole_disk' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_device_get_devid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_device_get_physical' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_dirnamelen' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_expand_proplist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_foreach_mountpoint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_all_props' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_clones_nvl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_enclosure_sysfs_path' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_fsacl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_handle' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_holds' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_pool_handle' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_pool_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_recvd_props' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_underlying_path' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_underlying_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_get_user_props' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_handle_dup' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_hold' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_hold_nvl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_ioctl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_is_mounted' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_is_shared' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_isnumber' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_bookmarks' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_bookmarks_v2' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_children' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_children_v2' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_dependents' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_dependents_v2' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_filesystems' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_filesystems_v2' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_mounted' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_root' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_snapshots' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_snapshots_sorted' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_snapshots_sorted_v2' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_snapshots_v2' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_snapspec' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_iter_snapspec_v2' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_mod_supported' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_mount' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_mount_at' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_mount_delegation_check' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_name_to_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_name_valid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_nicebytes' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_nicenum' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_nicenum_format' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_niceraw' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_nicestrtonum' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_nicetime' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_open' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_parent_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_parse_mount_options' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_path_to_zhandle' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_promote' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_align_right' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_column_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_default_numeric' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_default_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_delegatable' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_encryption_key_param' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_get' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_get_int' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_get_numeric' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_get_recvd' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_get_table' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_get_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_get_userquota' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_get_userquota_int' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_get_written' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_get_written_int' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_index_to_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_inherit' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_inheritable' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_is_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_random_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_readonly' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_set' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_set_list' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_set_list_flags' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_setonce' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_string_to_index' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_to_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_user' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_userquota' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_valid_for_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_valid_keylocation' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_values' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_visible' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prop_written' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_prune_proplist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_receive' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_refresh_properties' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_release' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_rename' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_resolve_shortname' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_rollback' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_save_arguments' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_send' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_send_one' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_send_progress' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_send_resume' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_send_resume_token_to_nvlist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_send_saved' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_set_fsacl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_setproctitle' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_setproctitle_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_share' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_show_diffs' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_smb_acl_add' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_smb_acl_purge' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_smb_acl_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_smb_acl_rename' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_snapshot' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_snapshot_nvl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_spa_version' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_spa_version_map' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_special_devs' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_standard_error' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_strcmp_pathname' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_strip_partition' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_strip_path' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_truncate_shares' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_type_to_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_unmount' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_unmountall' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_unshare' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_unshareall' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_userns' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_userspace' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_valid_proplist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_version_kernel' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_version_print' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_version_userland' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_wait_status' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_zpl_version_map' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_add' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_checkpoint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_clear' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_clear_label' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_close' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_default_search_paths' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_destroy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_disable_datasets' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_disable_datasets_os' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_disable_volume_os' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_discard_checkpoint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_disk_wait' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_dump_ddt' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_enable_datasets' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_events_clear' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_events_next' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_events_seek' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_expand_proplist' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_explain_recover' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_export' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_export_force' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_feature_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_find_config' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_find_vdev' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_find_vdev_by_physpath' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_free_handles' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_all_vdev_props' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_bootenv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_config' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_errlog' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_features' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_handle' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_history' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_load_policy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_prop_int' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_state' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_state_str' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_status' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_userprop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_vdev_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_get_vdev_prop_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_getenv_int' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_history_unpack' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_import' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_import_props' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_import_status' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_in_use' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_initialize' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_initialize_wait' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_is_draid_spare' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_iter' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_label_disk' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_label_disk_wait' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_load_compat' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_log_history' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_name_to_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_obj_to_path' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_obj_to_path_ds' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_open' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_open_canfail' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_pool_state_to_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prepare_and_label_disk' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prepare_disk' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_print_unsup_feat' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_align_right' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_column_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_default_numeric' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_default_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_feature' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_get_feature' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_get_table' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_get_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_index_to_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_random_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_readonly' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_setonce' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_string_to_index' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_to_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_unsupported' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_values' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_prop_vdev' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_props_refresh' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_read_label' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_refresh_stats' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_reguid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_reopen_one' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_scan' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_search_import' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_set_bootenv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_set_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_set_vdev_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_skip_pool' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_state_to_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_sync_one' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_trim' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_upgrade' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_attach' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_clear' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_degrade' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_detach' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_fault' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_indirect_size' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_name' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_offline' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_online' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_path_to_guid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_remove_cancel' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_remove_wanted' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_script_alloc_env' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_script_free_env' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_set_removed_state' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_vdev_split' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_wait' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zpool_wait_status' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_free_list' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_get_list' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_index_to_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_iter' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_iter_common' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_name_to_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_print_one_property' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_random_value' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_register_hidden' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_register_impl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_register_index' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_register_number' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_register_string' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_string_to_index' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_valid_char' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_valid_for_type' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_values' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zprop_width' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zvol_volsize_to_reservation' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	</elf-function-symbols>
	<elf-variable-symbols>
	<elf-symbol name='efi_debug' size='4' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_abd_ops' size='24' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_avx2_ops' size='128' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_avx512bw_ops' size='128' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_avx512f_ops' size='128' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_sse2_ops' size='128' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_ssse3_ops' size='128' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_superscalar4_ops' size='128' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='fletcher_4_superscalar_ops' size='128' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_config_ops' size='16' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='sa_protocol_names' size='16' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='spa_feature_table' size='2296' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfeature_checks_disable' size='4' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_deleg_perm_tab' size='512' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_history_event_names' size='328' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_max_dataset_nesting' size='4' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='zfs_userquota_prop_prefixes' size='96' type='object-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	</elf-variable-symbols>
	<abi-instr address-size='64' path='lib/libefi/rdwr_efi.c' language='LANG_C99'>
	<typedef-decl name='uInt' type-id='f0981eeb' id='09110a74'/>
	<var-decl name='efi_debug' type-id='95e97e5e' mangled-name='efi_debug' visibility='default' elf-symbol-id='efi_debug'/>
	<function-decl name='uuid_generate' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='cf536864'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='uuid_is_null' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='354f7eb9'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='crc32' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5bbcce85'/>
	<parameter type-id='e8cb3e0e'/>
	<parameter type-id='09110a74'/>
	<return type-id='5bbcce85'/>
	</function-decl>
	<function-decl name='efi_err_check' mangled-name='efi_err_check' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='efi_err_check'>
	<parameter type-id='0d8119a8' name='vtoc'/>
	<return type-id='48b5725f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libshare/libshare.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='b99c00c9' size-in-bits='128' id='2d6895a3'>
	<subrange length='2' type-id='7359adad' id='52efc4ef'/>
	</array-type-def>
	<var-decl name='sa_protocol_names' type-id='2d6895a3' mangled-name='sa_protocol_names' visibility='default' elf-symbol-id='sa_protocol_names'/>
	<type-decl name='unsigned long int' size-in-bits='64' id='7359adad'/>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libshare/nfs.c' language='LANG_C99'>
	<function-decl name='rename' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='memchr' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='flock' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='fchmod' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='e1c52942'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='mkdir' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='e1c52942'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libshare/os/linux/nfs.c' language='LANG_C99'>
	<class-decl name='sa_share_impl' size-in-bits='192' is-struct='yes' visibility='default' id='72b09bf8'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='sa_zfsname' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='sa_mountpoint' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='sa_shareopts' type-id='80f4b756' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='sa_share_impl_t' type-id='946a2c6b' id='a48b47d0'/>
	<class-decl name='sa_fstype_t' size-in-bits='384' is-struct='yes' naming-typedef-id='639af739' visibility='default' id='944afa86'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='enable_share' type-id='2f78a9c1' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='disable_share' type-id='2f78a9c1' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='is_shared' type-id='81020bc2' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='validate_shareopts' type-id='f194a8fb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='commit_shares' type-id='797ee7da' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='truncate_shares' type-id='5d51038b' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='sa_fstype_t' type-id='944afa86' id='639af739'/>
	<qualified-type-def type-id='639af739' const='yes' id='d19dbca9'/>
	<qualified-type-def type-id='72b09bf8' const='yes' id='484950e3'/>
	<pointer-type-def type-id='484950e3' size-in-bits='64' id='946a2c6b'/>
	<pointer-type-def type-id='276427e1' size-in-bits='64' id='1db260e5'/>
	<qualified-type-def type-id='1db260e5' const='yes' id='797ee7da'/>
	<pointer-type-def type-id='5113b296' size-in-bits='64' id='70487b28'/>
	<qualified-type-def type-id='70487b28' const='yes' id='f194a8fb'/>
	<pointer-type-def type-id='c13578bc' size-in-bits='64' id='fa1f29ce'/>
	<qualified-type-def type-id='fa1f29ce' const='yes' id='2f78a9c1'/>
	<pointer-type-def type-id='723e6cf2' size-in-bits='64' id='1d99e49c'/>
	<pointer-type-def type-id='86373eb1' size-in-bits='64' id='f337456d'/>
	<qualified-type-def type-id='f337456d' const='yes' id='81020bc2'/>
	<qualified-type-def type-id='953b12f8' const='yes' id='5d51038b'/>
	<var-decl name='libshare_nfs_type' type-id='d19dbca9' visibility='default'/>
	<function-decl name='nfs_escape_mountpoint' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='9b23c9ad'/>
	<parameter type-id='37e3bd22'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nfs_is_shared_impl' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='a48b47d0'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='nfs_toggle_share' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='a48b47d0'/>
	<parameter type-id='1d99e49c'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nfs_reset_shares' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-type size-in-bits='64' id='276427e1'>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='5113b296'>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='c13578bc'>
	<parameter type-id='a48b47d0'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='723e6cf2'>
	<parameter type-id='a48b47d0'/>
	<parameter type-id='822cd80b'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='86373eb1'>
	<parameter type-id='a48b47d0'/>
	<return type-id='c19b74c3'/>
	</function-type>
	<function-type size-in-bits='64' id='ee076206'>
	<return type-id='48b5725f'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libshare/os/linux/smb.c' language='LANG_C99'>
	<var-decl name='libshare_smb_type' type-id='d19dbca9' visibility='default'/>
	<function-decl name='__fgets_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='266fe297'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='e75a27e9'/>
	<return type-id='26a90f95'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/assert.c' language='LANG_C99'>
	+ <function-decl name='libspl_backtrace' mangled-name='libspl_backtrace' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libspl_backtrace'>
	+ <parameter type-id='95e97e5e'/>
	+ <return type-id='48b5725f'/>
	+ </function-decl>
	+ <function-decl name='gettid' visibility='default' binding='global' size-in-bits='64'>
	+ <return type-id='3629bad8'/>
	+ </function-decl>
	+ <function-decl name='prctl' visibility='default' binding='global' size-in-bits='64'>
	+ <parameter type-id='95e97e5e'/>
	+ <parameter is-variadic='yes'/>
	+ <return type-id='95e97e5e'/>
	+ </function-decl>
	<function-decl name='libspl_set_assert_ok' mangled-name='libspl_set_assert_ok' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libspl_set_assert_ok'>
	<parameter type-id='c19b74c3' name='val'/>
	<return type-id='48b5725f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/atomic.c' language='LANG_C99'>
	<typedef-decl name='int8_t' type-id='2171a512' id='ee31ee44'/>
	<typedef-decl name='__int8_t' type-id='28577a57' id='2171a512'/>
	<qualified-type-def type-id='149c6638' volatile='yes' id='5120c5f7'/>
	<pointer-type-def type-id='5120c5f7' size-in-bits='64' id='93977ae7'/>
	<qualified-type-def type-id='b96825af' volatile='yes' id='84ff7d66'/>
	<pointer-type-def type-id='84ff7d66' size-in-bits='64' id='aa323ea4'/>
	<qualified-type-def type-id='ee1f298e' volatile='yes' id='6f7e09cb'/>
	<pointer-type-def type-id='6f7e09cb' size-in-bits='64' id='64698d33'/>
	<function-decl name='atomic_inc_8' mangled-name='atomic_inc_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_inc_16' mangled-name='atomic_inc_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_16'>
	<parameter type-id='93977ae7' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_inc_32' mangled-name='atomic_inc_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_32'>
	<parameter type-id='3a147f31' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_inc_ulong' mangled-name='atomic_inc_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_dec_8' mangled-name='atomic_dec_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_dec_16' mangled-name='atomic_dec_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_16'>
	<parameter type-id='93977ae7' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_dec_32' mangled-name='atomic_dec_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_32'>
	<parameter type-id='3a147f31' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_dec_ulong' mangled-name='atomic_dec_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_add_ptr' mangled-name='atomic_add_ptr' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_ptr'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='79a0948f' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_add_8' mangled-name='atomic_add_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='ee31ee44' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_add_16' mangled-name='atomic_add_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='23bd8cb5' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_add_32' mangled-name='atomic_add_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_32'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='3ff5601b' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_sub_ptr' mangled-name='atomic_sub_ptr' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_ptr'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='79a0948f' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_sub_8' mangled-name='atomic_sub_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='ee31ee44' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_sub_16' mangled-name='atomic_sub_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='23bd8cb5' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_sub_32' mangled-name='atomic_sub_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_32'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='3ff5601b' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_or_8' mangled-name='atomic_or_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_or_16' mangled-name='atomic_or_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_or_32' mangled-name='atomic_or_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_32'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='8f92235e' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_or_ulong' mangled-name='atomic_or_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_and_8' mangled-name='atomic_and_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_and_16' mangled-name='atomic_and_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_and_32' mangled-name='atomic_and_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_32'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='8f92235e' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_and_ulong' mangled-name='atomic_and_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_inc_8_nv' mangled-name='atomic_inc_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_inc_16_nv' mangled-name='atomic_inc_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_inc_32_nv' mangled-name='atomic_inc_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_inc_ulong_nv' mangled-name='atomic_inc_ulong_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_ulong_nv'>
	<parameter type-id='64698d33' name='target'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_dec_8_nv' mangled-name='atomic_dec_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_dec_16_nv' mangled-name='atomic_dec_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_dec_32_nv' mangled-name='atomic_dec_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_dec_ulong_nv' mangled-name='atomic_dec_ulong_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_ulong_nv'>
	<parameter type-id='64698d33' name='target'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_add_ptr_nv' mangled-name='atomic_add_ptr_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_ptr_nv'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='79a0948f' name='bits'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='atomic_add_8_nv' mangled-name='atomic_add_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='ee31ee44' name='bits'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_add_16_nv' mangled-name='atomic_add_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='23bd8cb5' name='bits'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_add_32_nv' mangled-name='atomic_add_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='3ff5601b' name='bits'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_add_long_nv' mangled-name='atomic_add_long_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_long_nv'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='bd54fe1a' name='bits'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_sub_ptr_nv' mangled-name='atomic_sub_ptr_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_ptr_nv'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='79a0948f' name='bits'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='atomic_sub_8_nv' mangled-name='atomic_sub_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='ee31ee44' name='bits'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_sub_16_nv' mangled-name='atomic_sub_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='23bd8cb5' name='bits'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_sub_32_nv' mangled-name='atomic_sub_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='3ff5601b' name='bits'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_sub_long_nv' mangled-name='atomic_sub_long_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_long_nv'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='bd54fe1a' name='bits'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_or_8_nv' mangled-name='atomic_or_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='bits'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_or_16_nv' mangled-name='atomic_or_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='bits'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_or_32_nv' mangled-name='atomic_or_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='8f92235e' name='bits'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_or_ulong_nv' mangled-name='atomic_or_ulong_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_ulong_nv'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='bits'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_and_8_nv' mangled-name='atomic_and_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='bits'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_and_16_nv' mangled-name='atomic_and_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='bits'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_and_32_nv' mangled-name='atomic_and_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='8f92235e' name='bits'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_and_ulong_nv' mangled-name='atomic_and_ulong_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_ulong_nv'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='bits'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_cas_ptr' mangled-name='atomic_cas_ptr' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_ptr'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='eaa32e2f' name='exp'/>
	<parameter type-id='eaa32e2f' name='des'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='atomic_cas_8' mangled-name='atomic_cas_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='exp'/>
	<parameter type-id='b96825af' name='des'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_cas_16' mangled-name='atomic_cas_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='exp'/>
	<parameter type-id='149c6638' name='des'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_cas_32' mangled-name='atomic_cas_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_32'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='8f92235e' name='exp'/>
	<parameter type-id='8f92235e' name='des'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_cas_ulong' mangled-name='atomic_cas_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='exp'/>
	<parameter type-id='ee1f298e' name='des'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_swap_8' mangled-name='atomic_swap_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='bits'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_swap_16' mangled-name='atomic_swap_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='bits'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_swap_ulong' mangled-name='atomic_swap_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='bits'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_swap_ptr' mangled-name='atomic_swap_ptr' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_ptr'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='eaa32e2f' name='bits'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='atomic_set_long_excl' mangled-name='atomic_set_long_excl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_set_long_excl'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='3502e3ff' name='value'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='atomic_clear_long_excl' mangled-name='atomic_clear_long_excl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_clear_long_excl'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='3502e3ff' name='value'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='membar_enter' mangled-name='membar_enter' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='membar_enter'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='membar_consumer' mangled-name='membar_consumer' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='membar_consumer'>
	<return type-id='48b5725f'/>
	</function-decl>
	</abi-instr>
	+ <abi-instr address-size='64' path='lib/libspl/backtrace.c' language='LANG_C99'>
	+ <function-decl name='backtrace' visibility='default' binding='global' size-in-bits='64'>
	+ <parameter type-id='63e171df'/>
	+ <parameter type-id='95e97e5e'/>
	+ <return type-id='95e97e5e'/>
	+ </function-decl>
	+ <function-decl name='backtrace_symbols_fd' visibility='default' binding='global' size-in-bits='64'>
	+ <parameter type-id='7acd98a2'/>
	+ <parameter type-id='95e97e5e'/>
	+ <parameter type-id='95e97e5e'/>
	+ <return type-id='48b5725f'/>
	+ </function-decl>
	+ </abi-instr>
	<abi-instr address-size='64' path='lib/libspl/getexecname.c' language='LANG_C99'>
	<function-decl name='getexecname' mangled-name='getexecname' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='getexecname'>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='getexecname_impl' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='26a90f95'/>
	<return type-id='79a0948f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/list.c' language='LANG_C99'>
	<typedef-decl name='list_node_t' type-id='b0b5e45e' id='b21843b2'/>
	<typedef-decl name='list_t' type-id='e824dae9' id='0899125f'/>
	<class-decl name='list_node' size-in-bits='128' is-struct='yes' visibility='default' id='b0b5e45e'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='next' type-id='b03eadb4' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='prev' type-id='b03eadb4' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='list' size-in-bits='192' is-struct='yes' visibility='default' id='e824dae9'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='list_offset' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='list_head' type-id='b0b5e45e' visibility='default'/>
	</data-member>
	</class-decl>
	<pointer-type-def type-id='b0b5e45e' size-in-bits='64' id='b03eadb4'/>
	<pointer-type-def type-id='b21843b2' size-in-bits='64' id='ccc38265'/>
	<pointer-type-def type-id='0899125f' size-in-bits='64' id='352ec160'/>
	<function-decl name='list_create' mangled-name='list_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_create'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='b59d7dce' name='size'/>
	<parameter type-id='b59d7dce' name='offset'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_destroy' mangled-name='list_destroy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_destroy'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_insert_after' mangled-name='list_insert_after' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_insert_after'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<parameter type-id='eaa32e2f' name='nobject'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_insert_before' mangled-name='list_insert_before' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_insert_before'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<parameter type-id='eaa32e2f' name='nobject'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_insert_head' mangled-name='list_insert_head' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_insert_head'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_insert_tail' mangled-name='list_insert_tail' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_insert_tail'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_remove' mangled-name='list_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_remove'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_remove_head' mangled-name='list_remove_head' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_remove_head'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_remove_tail' mangled-name='list_remove_tail' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_remove_tail'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_head' mangled-name='list_head' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_head'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_tail' mangled-name='list_tail' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_tail'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_next' mangled-name='list_next' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_next'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_prev' mangled-name='list_prev' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_prev'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_move_tail' mangled-name='list_move_tail' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_move_tail'>
	<parameter type-id='352ec160' name='dst'/>
	<parameter type-id='352ec160' name='src'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_link_replace' mangled-name='list_link_replace' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_link_replace'>
	<parameter type-id='ccc38265' name='lold'/>
	<parameter type-id='ccc38265' name='lnew'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_link_init' mangled-name='list_link_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_link_init'>
	<parameter type-id='ccc38265' name='ln'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_link_active' mangled-name='list_link_active' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_link_active'>
	<parameter type-id='ccc38265' name='ln'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='list_is_empty' mangled-name='list_is_empty' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_is_empty'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/mkdirp.c' language='LANG_C99'>
	<typedef-decl name='wchar_t' type-id='95e97e5e' id='928221d2'/>
	<qualified-type-def type-id='928221d2' const='yes' id='effb3702'/>
	<pointer-type-def type-id='effb3702' size-in-bits='64' id='f077d3f8'/>
	<qualified-type-def type-id='f077d3f8' restrict='yes' id='598aab80'/>
	<pointer-type-def type-id='928221d2' size-in-bits='64' id='323d93c1'/>
	<qualified-type-def type-id='323d93c1' restrict='yes' id='f1358bc3'/>
	<function-decl name='__mbstowcs_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='f1358bc3'/>
	<parameter type-id='9d26089a'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='__wcstombs_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='266fe297'/>
	<parameter type-id='598aab80'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/os/linux/getmntany.c' language='LANG_C99'>
	<pointer-type-def type-id='56fe4a37' size-in-bits='64' id='b6b61d2f'/>
	<qualified-type-def type-id='b6b61d2f' restrict='yes' id='3cad23cd'/>
	<function-decl name='getmntent_r' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='e75a27e9'/>
	<parameter type-id='3cad23cd'/>
	<parameter type-id='266fe297'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='b6b61d2f'/>
	</function-decl>
	<function-decl name='feof' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='822cd80b'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/timestamp.c' language='LANG_C99'>
	<typedef-decl name='nl_item' type-id='95e97e5e' id='03b79a94'/>
	<function-decl name='nl_langinfo' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='03b79a94'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='print_timestamp' mangled-name='print_timestamp' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='print_timestamp'>
	<parameter type-id='3502e3ff' name='timestamp_fmt'/>
	<return type-id='48b5725f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libtpool/thread_pool.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='49ef3ffd' size-in-bits='1024' id='a14403f5'>
	<subrange length='16' type-id='7359adad' id='848d0938'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='384' id='36d7f119'>
	<subrange length='48' type-id='7359adad' id='8f6d2a81'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='f0981eeb' size-in-bits='64' id='0d532ec1'>
	<subrange length='2' type-id='7359adad' id='52efc4ef'/>
	</array-type-def>
	<union-decl name='__atomic_wide_counter' size-in-bits='64' naming-typedef-id='f3b40860' visibility='default' id='613ce450'>
	<data-member access='public'>
	<var-decl name='__value64' type-id='3a47d82b' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='__value32' type-id='e7f43f72' visibility='default'/>
	</data-member>
	</union-decl>
	<class-decl name='__anonymous_struct__' size-in-bits='64' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f72'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='__low' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='__high' type-id='f0981eeb' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='__atomic_wide_counter' type-id='613ce450' id='f3b40860'/>
	<typedef-decl name='__cpu_mask' type-id='7359adad' id='49ef3ffd'/>
	<class-decl name='cpu_set_t' size-in-bits='1024' is-struct='yes' naming-typedef-id='8037c762' visibility='default' id='1f20d231'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='__bits' type-id='a14403f5' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='cpu_set_t' type-id='1f20d231' id='8037c762'/>
	<union-decl name='pthread_condattr_t' size-in-bits='32' naming-typedef-id='836265dd' visibility='default' id='33dd3aad'>
	<data-member access='public'>
	<var-decl name='__size' type-id='8e0573fd' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='__align' type-id='95e97e5e' visibility='default'/>
	</data-member>
	</union-decl>
	<typedef-decl name='pthread_condattr_t' type-id='33dd3aad' id='836265dd'/>
	<union-decl name='pthread_cond_t' size-in-bits='384' naming-typedef-id='62fab762' visibility='default' id='cbb12c12'>
	<data-member access='public'>
	<var-decl name='__data' type-id='c987b47c' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='__size' type-id='36d7f119' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='__align' type-id='1eb56b1e' visibility='default'/>
	</data-member>
	</union-decl>
	<typedef-decl name='pthread_cond_t' type-id='cbb12c12' id='62fab762'/>
	<class-decl name='__pthread_cond_s' size-in-bits='384' is-struct='yes' visibility='default' id='c987b47c'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='__wseq' type-id='f3b40860' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='__g1_start' type-id='f3b40860' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='__g_refs' type-id='0d532ec1' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='__g_size' type-id='0d532ec1' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='__g1_orig_size' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='__wrefs' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='__g_signals' type-id='0d532ec1' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='sched_param' size-in-bits='32' is-struct='yes' visibility='default' id='0897719a'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='sched_priority' type-id='95e97e5e' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='tpool_job_t' type-id='3b8579e5' id='66a0afc9'/>
	<class-decl name='tpool_job' size-in-bits='192' is-struct='yes' visibility='default' id='3b8579e5'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='tpj_next' type-id='f32b30e4' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='tpj_func' type-id='b7f9d8e6' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='tpj_arg' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='tpool_active_t' type-id='c8d086f4' id='6fcda10e'/>
	<class-decl name='tpool_active' size-in-bits='128' is-struct='yes' visibility='default' id='c8d086f4'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='tpa_next' type-id='ad33e5e7' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='tpa_tid' type-id='4051f5e7' visibility='default'/>
	</data-member>
	</class-decl>
	<qualified-type-def type-id='8037c762' const='yes' id='f50ea9b2'/>
	<pointer-type-def type-id='f50ea9b2' size-in-bits='64' id='5e14fa48'/>
	<qualified-type-def type-id='836265dd' const='yes' id='7d24c58d'/>
	<pointer-type-def type-id='7d24c58d' size-in-bits='64' id='a7e325e5'/>
	<qualified-type-def type-id='a7e325e5' restrict='yes' id='4c428e67'/>
	<qualified-type-def type-id='0897719a' const='yes' id='c4a7b189'/>
	<pointer-type-def type-id='c4a7b189' size-in-bits='64' id='36fca399'/>
	<qualified-type-def type-id='36fca399' restrict='yes' id='37e4897b'/>
	<qualified-type-def type-id='e05e8614' restrict='yes' id='0be2e71c'/>
	<pointer-type-def type-id='8037c762' size-in-bits='64' id='d74a6869'/>
	<qualified-type-def type-id='7292109c' restrict='yes' id='6942f6a4'/>
	<qualified-type-def type-id='7347a39e' restrict='yes' id='578ba182'/>
	<pointer-type-def type-id='62fab762' size-in-bits='64' id='db285b03'/>
	<qualified-type-def type-id='db285b03' restrict='yes' id='2a468b41'/>
	<qualified-type-def type-id='18c91f9e' restrict='yes' id='6e745582'/>
	<pointer-type-def type-id='0897719a' size-in-bits='64' id='23cbcb08'/>
	<qualified-type-def type-id='23cbcb08' restrict='yes' id='b09b2050'/>
	<pointer-type-def type-id='6fcda10e' size-in-bits='64' id='ad33e5e7'/>
	<pointer-type-def type-id='66a0afc9' size-in-bits='64' id='f32b30e4'/>
	<qualified-type-def type-id='63e171df' restrict='yes' id='9e7a3a7d'/>
	<function-decl name='pthread_self' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='4051f5e7'/>
	</function-decl>
	<function-decl name='pthread_attr_init' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='7347a39e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_attr_destroy' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='7347a39e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_attr_getdetachstate' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='540db505'/>
	<parameter type-id='7292109c'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_attr_setdetachstate' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='7347a39e'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_attr_getguardsize' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='540db505'/>
	<parameter type-id='78c01427'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_attr_setguardsize' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='7347a39e'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_attr_getschedparam' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='e1815e87'/>
	<parameter type-id='b09b2050'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_attr_setschedparam' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='578ba182'/>
	<parameter type-id='37e4897b'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_attr_getschedpolicy' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='e1815e87'/>
	<parameter type-id='6942f6a4'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_attr_setschedpolicy' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='7347a39e'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_attr_getinheritsched' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='e1815e87'/>
	<parameter type-id='6942f6a4'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_attr_setinheritsched' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='7347a39e'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_attr_getscope' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='e1815e87'/>
	<parameter type-id='6942f6a4'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_attr_setscope' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='7347a39e'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_attr_getstack' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='e1815e87'/>
	<parameter type-id='9e7a3a7d'/>
	<parameter type-id='d19b2c25'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_attr_setstack' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='7347a39e'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_attr_setaffinity_np' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='7347a39e'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='5e14fa48'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_attr_getaffinity_np' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='540db505'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='d74a6869'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_setcancelstate' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='7292109c'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_setcanceltype' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='7292109c'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='__pthread_unregister_cancel' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='ba7c727c'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='pthread_cond_init' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='2a468b41'/>
	<parameter type-id='4c428e67'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_cond_signal' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='db285b03'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_cond_broadcast' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='db285b03'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_cond_wait' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='2a468b41'/>
	<parameter type-id='6e745582'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_cond_timedwait' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='2a468b41'/>
	<parameter type-id='6e745582'/>
	<parameter type-id='0be2e71c'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='__sysconf' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<return type-id='bd54fe1a'/>
	</function-decl>
	<function-decl name='tpool_abandon' mangled-name='tpool_abandon' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_abandon'>
	<parameter type-id='9cf59a50' name='tpool'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='tpool_suspend' mangled-name='tpool_suspend' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_suspend'>
	<parameter type-id='9cf59a50' name='tpool'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='tpool_suspended' mangled-name='tpool_suspended' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_suspended'>
	<parameter type-id='9cf59a50' name='tpool'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='tpool_resume' mangled-name='tpool_resume' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_resume'>
	<parameter type-id='9cf59a50' name='tpool'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='tpool_member' mangled-name='tpool_member' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_member'>
	<parameter type-id='9cf59a50' name='tpool'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<pointer-type-def type-id='b1bbf10d' size-in-bits='64' id='9cf59a50'/>
	<typedef-decl name='tpool_t' type-id='88d1b7f9' id='b1bbf10d'/>
	<class-decl name='tpool' size-in-bits='2496' is-struct='yes' visibility='default' id='88d1b7f9'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='tp_forw' type-id='9cf59a50' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='tp_back' type-id='9cf59a50' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='tp_mutex' type-id='7a6844eb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='tp_busycv' type-id='62fab762' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='832'>
	<var-decl name='tp_workcv' type-id='62fab762' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1216'>
	<var-decl name='tp_waitcv' type-id='62fab762' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1600'>
	<var-decl name='tp_active' type-id='ad33e5e7' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1664'>
	<var-decl name='tp_head' type-id='f32b30e4' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1728'>
	<var-decl name='tp_tail' type-id='f32b30e4' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1792'>
	<var-decl name='tp_attr' type-id='7d8569fd' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2240'>
	<var-decl name='tp_flags' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2272'>
	<var-decl name='tp_linger' type-id='3502e3ff' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2304'>
	<var-decl name='tp_njobs' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2336'>
	<var-decl name='tp_minimum' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2368'>
	<var-decl name='tp_maximum' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2400'>
	<var-decl name='tp_current' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2432'>
	<var-decl name='tp_idle' type-id='95e97e5e' visibility='default'/>
	</data-member>
	</class-decl>
	<function-type size-in-bits='64' id='c5c76c9c'>
	<parameter type-id='eaa32e2f'/>
	<return type-id='48b5725f'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_changelist.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='bf311473' size-in-bits='128' id='f0f65199'>
	<subrange length='2' type-id='7359adad' id='52efc4ef'/>
	</array-type-def>
	<type-decl name='char' size-in-bits='8' id='a84c031d'/>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='8192' id='b54ce520'>
	<subrange length='1024' type-id='7359adad' id='c60446f8'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='2048' id='d1617432'>
	<subrange length='256' type-id='7359adad' id='36e5b9fa'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='320' id='36c46961'>
	<subrange length='40' type-id='7359adad' id='8f80b239'/>
	</array-type-def>
	<class-decl name='re_dfa_t' is-struct='yes' visibility='default' is-declaration-only='yes' id='b48d2441'/>
	<class-decl name='uu_avl' is-struct='yes' visibility='default' is-declaration-only='yes' id='4af029d1'/>
	<class-decl name='uu_avl_pool' is-struct='yes' visibility='default' is-declaration-only='yes' id='12a530a8'/>
	<class-decl name='uu_avl_walk' is-struct='yes' visibility='default' is-declaration-only='yes' id='e70a39e3'/>
	<type-decl name='int' size-in-bits='32' id='95e97e5e'/>
	<type-decl name='long int' size-in-bits='64' id='bd54fe1a'/>
	<type-decl name='long long int' size-in-bits='64' id='1eb56b1e'/>
	<type-decl name='short int' size-in-bits='16' id='a2185560'/>
	<array-type-def dimensions='1' type-id='e475ab95' size-in-bits='192' id='0ce65a8b'>
	<subrange length='3' type-id='7359adad' id='56f209d2'/>
	</array-type-def>
	<type-decl name='unnamed-enum-underlying-type-32' is-anonymous='yes' size-in-bits='32' alignment-in-bits='32' id='9cac1fee'/>
	<type-decl name='unsigned char' size-in-bits='8' id='002ac4a6'/>
	<type-decl name='unsigned int' size-in-bits='32' id='f0981eeb'/>
	<type-decl name='unsigned long int' size-in-bits='64' id='7359adad'/>
	<type-decl name='void' id='48b5725f'/>
	<typedef-decl name='uu_compare_fn_t' type-id='add6e811' id='40f93560'/>
	<typedef-decl name='uu_avl_pool_t' type-id='12a530a8' id='7f84e390'/>
	<typedef-decl name='uu_avl_t' type-id='4af029d1' id='bb7f0973'/>
	<class-decl name='uu_avl_node' size-in-bits='192' is-struct='yes' visibility='default' id='f65f4326'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='uan_opaque' type-id='0ce65a8b' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='uu_avl_node_t' type-id='f65f4326' id='73a65116'/>
	<typedef-decl name='uu_avl_walk_t' type-id='e70a39e3' id='edd8457b'/>
	<typedef-decl name='uu_avl_index_t' type-id='e475ab95' id='5d7f5fc8'/>
	<typedef-decl name='zfs_handle_t' type-id='f6ee4445' id='775509eb'/>
	<typedef-decl name='zpool_handle_t' type-id='67002a8a' id='b1efc708'/>
	<typedef-decl name='libzfs_handle_t' type-id='c8a9d9d8' id='95942d0c'/>
	<typedef-decl name='zfs_iter_f' type-id='5571cde4' id='d8e49ab9'/>
	<typedef-decl name='avl_tree_t' type-id='b351119f' id='f20fbd51'/>
	<class-decl name='avl_node' size-in-bits='192' is-struct='yes' visibility='default' id='428b67b3'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='avl_child' type-id='f0f65199' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='avl_pcb' type-id='e475ab95' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='avl_tree' size-in-bits='320' is-struct='yes' visibility='default' id='b351119f'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='avl_root' type-id='bf311473' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='avl_compar' type-id='585e1de9' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='avl_offset' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='avl_numnodes' type-id='ee1f298e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='avl_pad' type-id='b59d7dce' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='dmu_objset_stats' size-in-bits='2304' is-struct='yes' visibility='default' id='098f0221'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='dds_num_clones' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='dds_creation_txg' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='dds_guid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='dds_type' type-id='230f1e16' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='224'>
	<var-decl name='dds_is_snapshot' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='232'>
	<var-decl name='dds_inconsistent' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='240'>
	<var-decl name='dds_redacted' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='248'>
	<var-decl name='dds_origin' type-id='d1617432' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='dmu_objset_stats_t' type-id='098f0221' id='b2c14f17'/>
	<enum-decl name='zfs_type_t' naming-typedef-id='2e45de5d' id='5d8f7321'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZFS_TYPE_INVALID' value='0'/>
	<enumerator name='ZFS_TYPE_FILESYSTEM' value='1'/>
	<enumerator name='ZFS_TYPE_SNAPSHOT' value='2'/>
	<enumerator name='ZFS_TYPE_VOLUME' value='4'/>
	<enumerator name='ZFS_TYPE_POOL' value='8'/>
	<enumerator name='ZFS_TYPE_BOOKMARK' value='16'/>
	<enumerator name='ZFS_TYPE_VDEV' value='32'/>
	</enum-decl>
	<typedef-decl name='zfs_type_t' type-id='5d8f7321' id='2e45de5d'/>
	<enum-decl name='dmu_objset_type' id='6b1b19f9'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='DMU_OST_NONE' value='0'/>
	<enumerator name='DMU_OST_META' value='1'/>
	<enumerator name='DMU_OST_ZFS' value='2'/>
	<enumerator name='DMU_OST_ZVOL' value='3'/>
	<enumerator name='DMU_OST_OTHER' value='4'/>
	<enumerator name='DMU_OST_ANY' value='5'/>
	<enumerator name='DMU_OST_NUMTYPES' value='6'/>
	</enum-decl>
	<typedef-decl name='dmu_objset_type_t' type-id='6b1b19f9' id='230f1e16'/>
	<enum-decl name='zfs_prop_t' naming-typedef-id='58603c44' id='4b000d60'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZPROP_CONT' value='-2'/>
	<enumerator name='ZPROP_INVAL' value='-1'/>
	<enumerator name='ZPROP_USERPROP' value='-1'/>
	<enumerator name='ZFS_PROP_TYPE' value='0'/>
	<enumerator name='ZFS_PROP_CREATION' value='1'/>
	<enumerator name='ZFS_PROP_USED' value='2'/>
	<enumerator name='ZFS_PROP_AVAILABLE' value='3'/>
	<enumerator name='ZFS_PROP_REFERENCED' value='4'/>
	<enumerator name='ZFS_PROP_COMPRESSRATIO' value='5'/>
	<enumerator name='ZFS_PROP_MOUNTED' value='6'/>
	<enumerator name='ZFS_PROP_ORIGIN' value='7'/>
	<enumerator name='ZFS_PROP_QUOTA' value='8'/>
	<enumerator name='ZFS_PROP_RESERVATION' value='9'/>
	<enumerator name='ZFS_PROP_VOLSIZE' value='10'/>
	<enumerator name='ZFS_PROP_VOLBLOCKSIZE' value='11'/>
	<enumerator name='ZFS_PROP_RECORDSIZE' value='12'/>
	<enumerator name='ZFS_PROP_MOUNTPOINT' value='13'/>
	<enumerator name='ZFS_PROP_SHARENFS' value='14'/>
	<enumerator name='ZFS_PROP_CHECKSUM' value='15'/>
	<enumerator name='ZFS_PROP_COMPRESSION' value='16'/>
	<enumerator name='ZFS_PROP_ATIME' value='17'/>
	<enumerator name='ZFS_PROP_DEVICES' value='18'/>
	<enumerator name='ZFS_PROP_EXEC' value='19'/>
	<enumerator name='ZFS_PROP_SETUID' value='20'/>
	<enumerator name='ZFS_PROP_READONLY' value='21'/>
	<enumerator name='ZFS_PROP_ZONED' value='22'/>
	<enumerator name='ZFS_PROP_SNAPDIR' value='23'/>
	<enumerator name='ZFS_PROP_ACLMODE' value='24'/>
	<enumerator name='ZFS_PROP_ACLINHERIT' value='25'/>
	<enumerator name='ZFS_PROP_CREATETXG' value='26'/>
	<enumerator name='ZFS_PROP_NAME' value='27'/>
	<enumerator name='ZFS_PROP_CANMOUNT' value='28'/>
	<enumerator name='ZFS_PROP_ISCSIOPTIONS' value='29'/>
	<enumerator name='ZFS_PROP_XATTR' value='30'/>
	<enumerator name='ZFS_PROP_NUMCLONES' value='31'/>
	<enumerator name='ZFS_PROP_COPIES' value='32'/>
	<enumerator name='ZFS_PROP_VERSION' value='33'/>
	<enumerator name='ZFS_PROP_UTF8ONLY' value='34'/>
	<enumerator name='ZFS_PROP_NORMALIZE' value='35'/>
	<enumerator name='ZFS_PROP_CASE' value='36'/>
	<enumerator name='ZFS_PROP_VSCAN' value='37'/>
	<enumerator name='ZFS_PROP_NBMAND' value='38'/>
	<enumerator name='ZFS_PROP_SHARESMB' value='39'/>
	<enumerator name='ZFS_PROP_REFQUOTA' value='40'/>
	<enumerator name='ZFS_PROP_REFRESERVATION' value='41'/>
	<enumerator name='ZFS_PROP_GUID' value='42'/>
	<enumerator name='ZFS_PROP_PRIMARYCACHE' value='43'/>
	<enumerator name='ZFS_PROP_SECONDARYCACHE' value='44'/>
	<enumerator name='ZFS_PROP_USEDSNAP' value='45'/>
	<enumerator name='ZFS_PROP_USEDDS' value='46'/>
	<enumerator name='ZFS_PROP_USEDCHILD' value='47'/>
	<enumerator name='ZFS_PROP_USEDREFRESERV' value='48'/>
	<enumerator name='ZFS_PROP_USERACCOUNTING' value='49'/>
	<enumerator name='ZFS_PROP_STMF_SHAREINFO' value='50'/>
	<enumerator name='ZFS_PROP_DEFER_DESTROY' value='51'/>
	<enumerator name='ZFS_PROP_USERREFS' value='52'/>
	<enumerator name='ZFS_PROP_LOGBIAS' value='53'/>
	<enumerator name='ZFS_PROP_UNIQUE' value='54'/>
	<enumerator name='ZFS_PROP_OBJSETID' value='55'/>
	<enumerator name='ZFS_PROP_DEDUP' value='56'/>
	<enumerator name='ZFS_PROP_MLSLABEL' value='57'/>
	<enumerator name='ZFS_PROP_SYNC' value='58'/>
	<enumerator name='ZFS_PROP_DNODESIZE' value='59'/>
	<enumerator name='ZFS_PROP_REFRATIO' value='60'/>
	<enumerator name='ZFS_PROP_WRITTEN' value='61'/>
	<enumerator name='ZFS_PROP_CLONES' value='62'/>
	<enumerator name='ZFS_PROP_LOGICALUSED' value='63'/>
	<enumerator name='ZFS_PROP_LOGICALREFERENCED' value='64'/>
	<enumerator name='ZFS_PROP_INCONSISTENT' value='65'/>
	<enumerator name='ZFS_PROP_VOLMODE' value='66'/>
	<enumerator name='ZFS_PROP_FILESYSTEM_LIMIT' value='67'/>
	<enumerator name='ZFS_PROP_SNAPSHOT_LIMIT' value='68'/>
	<enumerator name='ZFS_PROP_FILESYSTEM_COUNT' value='69'/>
	<enumerator name='ZFS_PROP_SNAPSHOT_COUNT' value='70'/>
	<enumerator name='ZFS_PROP_SNAPDEV' value='71'/>
	<enumerator name='ZFS_PROP_ACLTYPE' value='72'/>
	<enumerator name='ZFS_PROP_SELINUX_CONTEXT' value='73'/>
	<enumerator name='ZFS_PROP_SELINUX_FSCONTEXT' value='74'/>
	<enumerator name='ZFS_PROP_SELINUX_DEFCONTEXT' value='75'/>
	<enumerator name='ZFS_PROP_SELINUX_ROOTCONTEXT' value='76'/>
	<enumerator name='ZFS_PROP_RELATIME' value='77'/>
	<enumerator name='ZFS_PROP_REDUNDANT_METADATA' value='78'/>
	<enumerator name='ZFS_PROP_OVERLAY' value='79'/>
	<enumerator name='ZFS_PROP_PREV_SNAP' value='80'/>
	<enumerator name='ZFS_PROP_RECEIVE_RESUME_TOKEN' value='81'/>
	<enumerator name='ZFS_PROP_ENCRYPTION' value='82'/>
	<enumerator name='ZFS_PROP_KEYLOCATION' value='83'/>
	<enumerator name='ZFS_PROP_KEYFORMAT' value='84'/>
	<enumerator name='ZFS_PROP_PBKDF2_SALT' value='85'/>
	<enumerator name='ZFS_PROP_PBKDF2_ITERS' value='86'/>
	<enumerator name='ZFS_PROP_ENCRYPTION_ROOT' value='87'/>
	<enumerator name='ZFS_PROP_KEY_GUID' value='88'/>
	<enumerator name='ZFS_PROP_KEYSTATUS' value='89'/>
	<enumerator name='ZFS_PROP_REMAPTXG' value='90'/>
	<enumerator name='ZFS_PROP_SPECIAL_SMALL_BLOCKS' value='91'/>
	<enumerator name='ZFS_PROP_IVSET_GUID' value='92'/>
	<enumerator name='ZFS_PROP_REDACTED' value='93'/>
	<enumerator name='ZFS_PROP_REDACT_SNAPS' value='94'/>
	<enumerator name='ZFS_PROP_SNAPSHOTS_CHANGED' value='95'/>
	<enumerator name='ZFS_PROP_PREFETCH' value='96'/>
	<enumerator name='ZFS_PROP_VOLTHREADING' value='97'/>
	<enumerator name='ZFS_NUM_PROPS' value='98'/>
	</enum-decl>
	<typedef-decl name='zfs_prop_t' type-id='4b000d60' id='58603c44'/>
	<enum-decl name='zprop_source_t' naming-typedef-id='a2256d42' id='5903f80e'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZPROP_SRC_NONE' value='1'/>
	<enumerator name='ZPROP_SRC_DEFAULT' value='2'/>
	<enumerator name='ZPROP_SRC_TEMPORARY' value='4'/>
	<enumerator name='ZPROP_SRC_LOCAL' value='8'/>
	<enumerator name='ZPROP_SRC_INHERITED' value='16'/>
	<enumerator name='ZPROP_SRC_RECEIVED' value='32'/>
	</enum-decl>
	<typedef-decl name='zprop_source_t' type-id='5903f80e' id='a2256d42'/>
	<class-decl name='nvlist' size-in-bits='192' is-struct='yes' visibility='default' id='ac266fd9'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='nvl_version' type-id='3ff5601b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='nvl_nvflag' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='nvl_priv' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='nvl_flag' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='160'>
	<var-decl name='nvl_pad' type-id='3ff5601b' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='nvlist_t' type-id='ac266fd9' id='8e8d4be3'/>
	<enum-decl name='sa_protocol' id='9155d4b5'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='SA_PROTOCOL_NFS' value='0'/>
	<enumerator name='SA_PROTOCOL_SMB' value='1'/>
	<enumerator name='SA_PROTOCOL_COUNT' value='2'/>
	</enum-decl>
	<enum-decl name='boolean_t' naming-typedef-id='c19b74c3' id='f58c8277'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='B_FALSE' value='0'/>
	<enumerator name='B_TRUE' value='1'/>
	</enum-decl>
	<typedef-decl name='boolean_t' type-id='f58c8277' id='c19b74c3'/>
	<typedef-decl name='uint_t' type-id='f0981eeb' id='3502e3ff'/>
	<typedef-decl name='ulong_t' type-id='7359adad' id='ee1f298e'/>
	<typedef-decl name='longlong_t' type-id='1eb56b1e' id='9b3ff54f'/>
	<typedef-decl name='diskaddr_t' type-id='9b3ff54f' id='804dc465'/>
	<typedef-decl name='zoneid_t' type-id='3502e3ff' id='4da03624'/>
	<typedef-decl name='__re_long_size_t' type-id='7359adad' id='ba516949'/>
	<typedef-decl name='reg_syntax_t' type-id='7359adad' id='1b72c3b3'/>
	<class-decl name='re_pattern_buffer' size-in-bits='512' is-struct='yes' visibility='default' id='19fc9a8c'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='buffer' type-id='33976309' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='allocated' type-id='ba516949' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='used' type-id='ba516949' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='syntax' type-id='1b72c3b3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='fastmap' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='translate' type-id='cf536864' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='re_nsub' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='can_be_null' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='449'>
	<var-decl name='regs_allocated' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='451'>
	<var-decl name='fastmap_accurate' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='452'>
	<var-decl name='no_sub' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='453'>
	<var-decl name='not_bol' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='454'>
	<var-decl name='not_eol' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='455'>
	<var-decl name='newline_anchor' type-id='f0981eeb' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='regex_t' type-id='19fc9a8c' id='aca3bac8'/>
	<typedef-decl name='uintptr_t' type-id='7359adad' id='e475ab95'/>
	<union-decl name='pthread_mutex_t' size-in-bits='320' naming-typedef-id='7a6844eb' visibility='default' id='70681f9b'>
	<data-member access='public'>
	<var-decl name='__data' type-id='4c734837' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='__size' type-id='36c46961' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='__align' type-id='bd54fe1a' visibility='default'/>
	</data-member>
	</union-decl>
	<typedef-decl name='pthread_mutex_t' type-id='70681f9b' id='7a6844eb'/>
	<typedef-decl name='int32_t' type-id='33f57a65' id='3ff5601b'/>
	<typedef-decl name='uint8_t' type-id='c51d6389' id='b96825af'/>
	<typedef-decl name='uint32_t' type-id='62f1140c' id='8f92235e'/>
	<typedef-decl name='uint64_t' type-id='8910171f' id='9c313c2d'/>
	<class-decl name='__pthread_mutex_s' size-in-bits='320' is-struct='yes' visibility='default' id='4c734837'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='__lock' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='__count' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='__owner' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='__nusers' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='__kind' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='160'>
	<var-decl name='__spins' type-id='a2185560' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='176'>
	<var-decl name='__elision' type-id='a2185560' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='__list' type-id='518fb49c' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__pthread_internal_list' size-in-bits='128' is-struct='yes' visibility='default' id='0e01899c'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='__prev' type-id='4d98cd5a' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='__next' type-id='4d98cd5a' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='__pthread_list_t' type-id='0e01899c' id='518fb49c'/>
	<typedef-decl name='__uint8_t' type-id='002ac4a6' id='c51d6389'/>
	<typedef-decl name='__int32_t' type-id='95e97e5e' id='33f57a65'/>
	<typedef-decl name='__uint32_t' type-id='f0981eeb' id='62f1140c'/>
	<typedef-decl name='__uint64_t' type-id='7359adad' id='8910171f'/>
	<typedef-decl name='size_t' type-id='7359adad' id='b59d7dce'/>
	<class-decl name='libzfs_handle' size-in-bits='18240' is-struct='yes' visibility='default' id='c8a9d9d8'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='libzfs_error' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='libzfs_fd' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='libzfs_pool_handles' type-id='4c81de99' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='libzfs_ns_avlpool' type-id='de82c773' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='libzfs_ns_avl' type-id='a5c21a38' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='libzfs_ns_gen' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='libzfs_desc_active' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='352'>
	<var-decl name='libzfs_action' type-id='b54ce520' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='8544'>
	<var-decl name='libzfs_desc' type-id='b54ce520' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='16736'>
	<var-decl name='libzfs_printerr' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='16768'>
	<var-decl name='libzfs_mnttab_enable' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='16832'>
	<var-decl name='libzfs_mnttab_cache_lock' type-id='7a6844eb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='17152'>
	<var-decl name='libzfs_mnttab_cache' type-id='f20fbd51' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='17472'>
	<var-decl name='libzfs_pool_iter' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='17504'>
	<var-decl name='libzfs_prop_debug' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='17536'>
	<var-decl name='libzfs_urire' type-id='aca3bac8' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='18048'>
	<var-decl name='libzfs_max_nvlist' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='18112'>
	<var-decl name='libfetch' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='18176'>
	<var-decl name='libfetch_load_error' type-id='26a90f95' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='zfs_handle' size-in-bits='4928' is-struct='yes' visibility='default' id='f6ee4445'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='zfs_hdl' type-id='b0382bb3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='zpool_hdl' type-id='4c81de99' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='zfs_name' type-id='d1617432' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2176'>
	<var-decl name='zfs_type' type-id='2e45de5d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2208'>
	<var-decl name='zfs_head_type' type-id='2e45de5d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2240'>
	<var-decl name='zfs_dmustats' type-id='b2c14f17' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='4544'>
	<var-decl name='zfs_props' type-id='5ce45b60' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='4608'>
	<var-decl name='zfs_user_props' type-id='5ce45b60' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='4672'>
	<var-decl name='zfs_recvd_props' type-id='5ce45b60' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='4736'>
	<var-decl name='zfs_mntcheck' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='4800'>
	<var-decl name='zfs_mntopts' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='4864'>
	<var-decl name='zfs_props_table' type-id='ae3e8ca6' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='zpool_handle' size-in-bits='2560' is-struct='yes' visibility='default' id='67002a8a'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='zpool_hdl' type-id='b0382bb3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='zpool_next' type-id='4c81de99' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='zpool_name' type-id='d1617432' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2176'>
	<var-decl name='zpool_state' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2240'>
	<var-decl name='zpool_config_size' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2304'>
	<var-decl name='zpool_config' type-id='5ce45b60' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2368'>
	<var-decl name='zpool_old_config' type-id='5ce45b60' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2432'>
	<var-decl name='zpool_props' type-id='5ce45b60' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2496'>
	<var-decl name='zpool_start_block' type-id='804dc465' visibility='default'/>
	</data-member>
	</class-decl>
	<pointer-type-def type-id='0e01899c' size-in-bits='64' id='4d98cd5a'/>
	<pointer-type-def type-id='428b67b3' size-in-bits='64' id='bf311473'/>
	<pointer-type-def type-id='a84c031d' size-in-bits='64' id='26a90f95'/>
	<pointer-type-def type-id='26a90f95' size-in-bits='64' id='9b23c9ad'/>
	<qualified-type-def type-id='a84c031d' const='yes' id='9b45d938'/>
	<pointer-type-def type-id='9b45d938' size-in-bits='64' id='80f4b756'/>
	<qualified-type-def type-id='9155d4b5' const='yes' id='9f2c1699'/>
	<pointer-type-def type-id='9f2c1699' size-in-bits='64' id='4567bbc9'/>
	<qualified-type-def type-id='775509eb' const='yes' id='5eadf2db'/>
	<pointer-type-def type-id='5eadf2db' size-in-bits='64' id='fcd57163'/>
	<pointer-type-def type-id='96ee24a5' size-in-bits='64' id='585e1de9'/>
	<pointer-type-def type-id='cb9628fa' size-in-bits='64' id='5571cde4'/>
	<pointer-type-def type-id='95942d0c' size-in-bits='64' id='b0382bb3'/>
	<pointer-type-def type-id='8e8d4be3' size-in-bits='64' id='5ce45b60'/>
	<pointer-type-def type-id='b48d2441' size-in-bits='64' id='33976309'/>
	<pointer-type-def type-id='b96825af' size-in-bits='64' id='ae3e8ca6'/>
	<pointer-type-def type-id='002ac4a6' size-in-bits='64' id='cf536864'/>
	<pointer-type-def type-id='5d7f5fc8' size-in-bits='64' id='813a2225'/>
	<pointer-type-def type-id='73a65116' size-in-bits='64' id='2dc35b9d'/>
	<pointer-type-def type-id='7f84e390' size-in-bits='64' id='de82c773'/>
	<pointer-type-def type-id='bb7f0973' size-in-bits='64' id='a5c21a38'/>
	<pointer-type-def type-id='edd8457b' size-in-bits='64' id='5842d146'/>
	<pointer-type-def type-id='40f93560' size-in-bits='64' id='d502b39f'/>
	<pointer-type-def type-id='48b5725f' size-in-bits='64' id='eaa32e2f'/>
	<pointer-type-def type-id='775509eb' size-in-bits='64' id='9200a744'/>
	<pointer-type-def type-id='b1efc708' size-in-bits='64' id='4c81de99'/>
	<pointer-type-def type-id='a2256d42' size-in-bits='64' id='debc6aa3'/>
	<class-decl name='re_dfa_t' is-struct='yes' visibility='default' is-declaration-only='yes' id='b48d2441'/>
	<class-decl name='uu_avl' is-struct='yes' visibility='default' is-declaration-only='yes' id='4af029d1'/>
	<class-decl name='uu_avl_pool' is-struct='yes' visibility='default' is-declaration-only='yes' id='12a530a8'/>
	<class-decl name='uu_avl_walk' is-struct='yes' visibility='default' is-declaration-only='yes' id='e70a39e3'/>
	<function-decl name='uu_avl_pool_create' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='d502b39f'/>
	<parameter type-id='8f92235e'/>
	<return type-id='de82c773'/>
	</function-decl>
	<function-decl name='uu_avl_pool_destroy' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='de82c773'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='uu_avl_node_init' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='2dc35b9d'/>
	<parameter type-id='de82c773'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='uu_avl_create' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='de82c773'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='8f92235e'/>
	<return type-id='a5c21a38'/>
	</function-decl>
	<function-decl name='uu_avl_destroy' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='a5c21a38'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='uu_avl_last' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='a5c21a38'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_avl_walk_start' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='a5c21a38'/>
	<parameter type-id='8f92235e'/>
	<return type-id='5842d146'/>
	</function-decl>
	<function-decl name='uu_avl_walk_next' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5842d146'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_avl_walk_end' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5842d146'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='uu_avl_find' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='a5c21a38'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='813a2225'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_avl_insert' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='a5c21a38'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5d7f5fc8'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='uu_avl_remove' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='a5c21a38'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_get_handle' mangled-name='zfs_get_handle' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_handle'>
	<parameter type-id='9200a744'/>
	<return type-id='b0382bb3'/>
	</function-decl>
	<function-decl name='zfs_open' mangled-name='zfs_open' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_open'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='9200a744'/>
	</function-decl>
	<function-decl name='zfs_close' mangled-name='zfs_close' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_close'>
	<parameter type-id='9200a744'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_get_name' mangled-name='zfs_get_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_name'>
	<parameter type-id='fcd57163'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_prop_get' mangled-name='zfs_prop_get' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get'>
	<parameter type-id='9200a744'/>
	<parameter type-id='58603c44'/>
	<parameter type-id='26a90f95'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='debc6aa3'/>
	<parameter type-id='26a90f95'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='c19b74c3'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_get_int' mangled-name='zfs_prop_get_int' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_int'>
	<parameter type-id='9200a744'/>
	<parameter type-id='58603c44'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='zfs_iter_children_v2' mangled-name='zfs_iter_children_v2' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_children_v2'>
	<parameter type-id='9200a744'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='d8e49ab9'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_iter_dependents_v2' mangled-name='zfs_iter_dependents_v2' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_dependents_v2'>
	<parameter type-id='9200a744'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='c19b74c3'/>
	<parameter type-id='d8e49ab9'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_iter_mounted' mangled-name='zfs_iter_mounted' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_mounted'>
	<parameter type-id='9200a744'/>
	<parameter type-id='d8e49ab9'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_refresh_properties' mangled-name='zfs_refresh_properties' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_refresh_properties'>
	<parameter type-id='9200a744'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_is_mounted' mangled-name='zfs_is_mounted' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_is_mounted'>
	<parameter type-id='9200a744'/>
	<parameter type-id='9b23c9ad'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_mount' mangled-name='zfs_mount' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_mount'>
	<parameter type-id='9200a744'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_unmount' mangled-name='zfs_unmount' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_unmount'>
	<parameter type-id='9200a744'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_is_shared' mangled-name='zfs_is_shared' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_is_shared'>
	<parameter type-id='9200a744'/>
	<parameter type-id='9b23c9ad'/>
	<parameter type-id='4567bbc9'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_share' mangled-name='zfs_share' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_share'>
	<parameter type-id='9200a744'/>
	<parameter type-id='4567bbc9'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_unshare' mangled-name='zfs_unshare' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_unshare'>
	<parameter type-id='9200a744'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='4567bbc9'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_commit_shares' mangled-name='zfs_commit_shares' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_commit_shares'>
	<parameter type-id='4567bbc9'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='getzoneid' mangled-name='getzoneid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='getzoneid'>
	<return type-id='4da03624'/>
	</function-decl>
	<function-decl name='sa_commit_shares' mangled-name='sa_commit_shares' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='sa_commit_shares'>
	<parameter type-id='9155d4b5'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='strlcat' mangled-name='strlcat' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='strlcat'>
	<parameter type-id='26a90f95'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='strlcpy' mangled-name='strlcpy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='strlcpy'>
	<parameter type-id='26a90f95'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='free' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='eaa32e2f'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='strcmp' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='strncmp' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='strlen' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='zfs_error' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_alloc' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='remove_mountpoint' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9200a744'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-type size-in-bits='64' id='96ee24a5'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='add6e811'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='cb9628fa'>
	<parameter type-id='9200a744'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_config.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='32768' id='d16c6df4'>
	<subrange length='4096' type-id='7359adad' id='bc1b5ddc'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='65536' id='163f6aa5'>
	<subrange length='8192' type-id='7359adad' id='c88f397d'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='infinite' id='e84913bd'>
	<subrange length='infinite' type-id='7359adad' id='031f2035'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='9c313c2d' size-in-bits='128' id='c1c22e6c'>
	<subrange length='2' type-id='7359adad' id='52efc4ef'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='b96825af' size-in-bits='24' id='d3490169'>
	<subrange length='3' type-id='7359adad' id='56f209d2'/>
	</array-type-def>
	<type-decl name='variadic parameter type' id='2c1145c5'/>
	<typedef-decl name='zpool_iter_f' type-id='3aebb66f' id='fa476e62'/>
	<enum-decl name='data_type_t' naming-typedef-id='8d0687d2' id='aeeae136'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='DATA_TYPE_DONTCARE' value='-1'/>
	<enumerator name='DATA_TYPE_UNKNOWN' value='0'/>
	<enumerator name='DATA_TYPE_BOOLEAN' value='1'/>
	<enumerator name='DATA_TYPE_BYTE' value='2'/>
	<enumerator name='DATA_TYPE_INT16' value='3'/>
	<enumerator name='DATA_TYPE_UINT16' value='4'/>
	<enumerator name='DATA_TYPE_INT32' value='5'/>
	<enumerator name='DATA_TYPE_UINT32' value='6'/>
	<enumerator name='DATA_TYPE_INT64' value='7'/>
	<enumerator name='DATA_TYPE_UINT64' value='8'/>
	<enumerator name='DATA_TYPE_STRING' value='9'/>
	<enumerator name='DATA_TYPE_BYTE_ARRAY' value='10'/>
	<enumerator name='DATA_TYPE_INT16_ARRAY' value='11'/>
	<enumerator name='DATA_TYPE_UINT16_ARRAY' value='12'/>
	<enumerator name='DATA_TYPE_INT32_ARRAY' value='13'/>
	<enumerator name='DATA_TYPE_UINT32_ARRAY' value='14'/>
	<enumerator name='DATA_TYPE_INT64_ARRAY' value='15'/>
	<enumerator name='DATA_TYPE_UINT64_ARRAY' value='16'/>
	<enumerator name='DATA_TYPE_STRING_ARRAY' value='17'/>
	<enumerator name='DATA_TYPE_HRTIME' value='18'/>
	<enumerator name='DATA_TYPE_NVLIST' value='19'/>
	<enumerator name='DATA_TYPE_NVLIST_ARRAY' value='20'/>
	<enumerator name='DATA_TYPE_BOOLEAN_VALUE' value='21'/>
	<enumerator name='DATA_TYPE_INT8' value='22'/>
	<enumerator name='DATA_TYPE_UINT8' value='23'/>
	<enumerator name='DATA_TYPE_BOOLEAN_ARRAY' value='24'/>
	<enumerator name='DATA_TYPE_INT8_ARRAY' value='25'/>
	<enumerator name='DATA_TYPE_UINT8_ARRAY' value='26'/>
	<enumerator name='DATA_TYPE_DOUBLE' value='27'/>
	</enum-decl>
	<typedef-decl name='data_type_t' type-id='aeeae136' id='8d0687d2'/>
	<class-decl name='nvpair' size-in-bits='128' is-struct='yes' visibility='default' id='1c34e459'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='nvp_size' type-id='3ff5601b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='nvp_name_sz' type-id='23bd8cb5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='48'>
	<var-decl name='nvp_reserve' type-id='23bd8cb5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='nvp_value_elem' type-id='3ff5601b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='nvp_type' type-id='8d0687d2' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='nvp_name' type-id='e84913bd' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='nvpair_t' type-id='1c34e459' id='57928edf'/>
	<class-decl name='drr_begin' size-in-bits='2432' is-struct='yes' visibility='default' id='09fcdc01'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_magic' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='drr_versioninfo' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='drr_creation_time' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='drr_type' type-id='230f1e16' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='224'>
	<var-decl name='drr_flags' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='drr_fromguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='drr_toname' type-id='d1617432' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='zinject_record' size-in-bits='2816' is-struct='yes' visibility='default' id='3216f820'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='zi_objset' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='zi_object' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='zi_start' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='zi_end' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='zi_guid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='zi_level' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='352'>
	<var-decl name='zi_error' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='zi_type' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='zi_freq' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='480'>
	<var-decl name='zi_failfast' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='zi_func' type-id='d1617432' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2560'>
	<var-decl name='zi_iotype' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2592'>
	<var-decl name='zi_duration' type-id='3ff5601b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2624'>
	<var-decl name='zi_timer' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2688'>
	<var-decl name='zi_nlanes' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2752'>
	<var-decl name='zi_cmd' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2784'>
	<var-decl name='zi_dvas' type-id='8f92235e' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zinject_record_t' type-id='3216f820' id='a4301ca6'/>
	<class-decl name='zfs_share' size-in-bits='256' is-struct='yes' visibility='default' id='feb6f2da'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='z_exportdata' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='z_sharedata' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='z_sharetype' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='z_sharemax' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zfs_share_t' type-id='feb6f2da' id='ee5cec36'/>
	<class-decl name='zfs_cmd' size-in-bits='109952' is-struct='yes' visibility='default' id='3522cd69'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='zc_name' type-id='d16c6df4' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32768'>
	<var-decl name='zc_nvlist_src' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32832'>
	<var-decl name='zc_nvlist_src_size' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32896'>
	<var-decl name='zc_nvlist_dst' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32960'>
	<var-decl name='zc_nvlist_dst_size' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='33024'>
	<var-decl name='zc_nvlist_dst_filled' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='33056'>
	<var-decl name='zc_pad2' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='33088'>
	<var-decl name='zc_history' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='33152'>
	<var-decl name='zc_value' type-id='163f6aa5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='98688'>
	<var-decl name='zc_string' type-id='d1617432' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='100736'>
	<var-decl name='zc_guid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='100800'>
	<var-decl name='zc_nvlist_conf' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='100864'>
	<var-decl name='zc_nvlist_conf_size' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='100928'>
	<var-decl name='zc_cookie' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='100992'>
	<var-decl name='zc_objset_type' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='101056'>
	<var-decl name='zc_perm_action' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='101120'>
	<var-decl name='zc_history_len' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='101184'>
	<var-decl name='zc_history_offset' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='101248'>
	<var-decl name='zc_obj' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='101312'>
	<var-decl name='zc_iflags' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='101376'>
	<var-decl name='zc_share' type-id='ee5cec36' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='101632'>
	<var-decl name='zc_objset_stats' type-id='b2c14f17' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='103936'>
	<var-decl name='zc_begin_record' type-id='09fcdc01' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='106368'>
	<var-decl name='zc_inject_record' type-id='a4301ca6' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109184'>
	<var-decl name='zc_defer_destroy' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109216'>
	<var-decl name='zc_flags' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109248'>
	<var-decl name='zc_action_handle' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109312'>
	<var-decl name='zc_cleanup_fd' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109344'>
	<var-decl name='zc_simple' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109352'>
	<var-decl name='zc_pad' type-id='d3490169' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109376'>
	<var-decl name='zc_sendobj' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109440'>
	<var-decl name='zc_fromobj' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109504'>
	<var-decl name='zc_createtxg' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109568'>
	<var-decl name='zc_stat' type-id='0371a9c7' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109888'>
	<var-decl name='zc_zoneid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zfs_cmd_t' type-id='3522cd69' id='a5559cdd'/>
	<class-decl name='zfs_stat' size-in-bits='320' is-struct='yes' visibility='default' id='6417f0b9'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='zs_gen' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='zs_mode' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='zs_links' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='zs_ctime' type-id='c1c22e6c' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zfs_stat_t' type-id='6417f0b9' id='0371a9c7'/>
	<typedef-decl name='int16_t' type-id='03896e23' id='23bd8cb5'/>
	<typedef-decl name='__int16_t' type-id='a2185560' id='03896e23'/>
	<pointer-type-def type-id='c19b74c3' size-in-bits='64' id='37e3bd22'/>
	<qualified-type-def type-id='8e8d4be3' const='yes' id='693c3853'/>
	<pointer-type-def type-id='693c3853' size-in-bits='64' id='22cce67b'/>
	<qualified-type-def type-id='57928edf' const='yes' id='642ee20f'/>
	<pointer-type-def type-id='642ee20f' size-in-bits='64' id='dace003f'/>
	<pointer-type-def type-id='2bce87e3' size-in-bits='64' id='3aebb66f'/>
	<pointer-type-def type-id='95e97e5e' size-in-bits='64' id='7292109c'/>
	<pointer-type-def type-id='5ce45b60' size-in-bits='64' id='857bb57e'/>
	<pointer-type-def type-id='57928edf' size-in-bits='64' id='3fa542f0'/>
	<pointer-type-def type-id='eaa32e2f' size-in-bits='64' id='63e171df'/>
	<pointer-type-def type-id='3522cd69' size-in-bits='64' id='b65f7fd1'/>
	<pointer-type-def type-id='a5559cdd' size-in-bits='64' id='e4ec4540'/>
	<pointer-type-def type-id='4c81de99' size-in-bits='64' id='237193c9'/>
	<function-decl name='uu_avl_first' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='a5c21a38'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_avl_next' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='a5c21a38'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='uu_avl_teardown' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='a5c21a38'/>
	<parameter type-id='63e171df'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='zfs_standard_error' mangled-name='zfs_standard_error' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_standard_error'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_ioctl' mangled-name='zfs_ioctl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_ioctl'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='b65f7fd1'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_free' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_dup' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='22cce67b'/>
	<parameter type-id='857bb57e'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_nvlist' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='857bb57e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_exists' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='22cce67b'/>
	<parameter type-id='80f4b756'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='nvlist_next_nvpair' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='dace003f'/>
	<return type-id='3fa542f0'/>
	</function-decl>
	<function-decl name='nvpair_name' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='dace003f'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='fnvpair_value_nvlist' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='3fa542f0'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='libspl_assertf' mangled-name='libspl_assertf' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libspl_assertf'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<parameter is-variadic='yes'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='__errno_location' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='7292109c'/>
	</function-decl>
	<function-decl name='dcgettext' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='getenv' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='strchr' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='zpool_get_config' mangled-name='zpool_get_config' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_config'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='857bb57e' name='oldconfig'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='zpool_get_features' mangled-name='zpool_get_features' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_features'>
	<parameter type-id='4c81de99' name='zhp'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='zpool_refresh_stats' mangled-name='zpool_refresh_stats' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_refresh_stats'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='37e3bd22' name='missing'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_skip_pool' mangled-name='zpool_skip_pool' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_skip_pool'>
	<parameter type-id='80f4b756' name='poolname'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zpool_iter' mangled-name='zpool_iter' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_iter'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='fa476e62' name='func'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_iter_root' mangled-name='zfs_iter_root' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_root'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='d8e49ab9' name='func'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_strdup' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='80f4b756'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='no_memory' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zcmd_alloc_dst_nvlist' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='e4ec4540'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zcmd_expand_dst_nvlist' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='e4ec4540'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zcmd_read_dst_nvlist' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='e4ec4540'/>
	<parameter type-id='857bb57e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zcmd_free_nvlists' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='e4ec4540'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='make_dataset_handle' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='80f4b756'/>
	<return type-id='9200a744'/>
	</function-decl>
	<function-decl name='zpool_open_silent' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='237193c9'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-type size-in-bits='64' id='2bce87e3'>
	<parameter type-id='4c81de99'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_crypto.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='38b51b3c' size-in-bits='832' id='02b72c00'>
	<subrange length='13' type-id='7359adad' id='487fded1'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='fb7c6451' size-in-bits='256' id='64177143'>
	<subrange length='32' type-id='7359adad' id='ae5bde82'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='8' id='89feb1ec'>
	<subrange length='1' type-id='7359adad' id='52f813b4'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='160' id='664ac0b7'>
	<subrange length='20' type-id='7359adad' id='fdca39cf'/>
	</array-type-def>
	<class-decl name='_IO_codecvt' is-struct='yes' visibility='default' is-declaration-only='yes' id='a4036571'/>
	<class-decl name='_IO_marker' is-struct='yes' visibility='default' is-declaration-only='yes' id='010ae0b9'/>
	<class-decl name='_IO_wide_data' is-struct='yes' visibility='default' is-declaration-only='yes' id='79bd3751'/>
	<class-decl name='__locale_data' is-struct='yes' visibility='default' is-declaration-only='yes' id='23de8b96'/>
	<array-type-def dimensions='1' type-id='80f4b756' size-in-bits='832' id='39e6f84a'>
	<subrange length='13' type-id='7359adad' id='487fded1'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='95e97e5e' size-in-bits='896' id='47394ee0'>
	<subrange length='28' type-id='7359adad' id='3db583d7'/>
	</array-type-def>
	<type-decl name='signed char' size-in-bits='8' id='28577a57'/>
	<array-type-def dimensions='1' type-id='7359adad' size-in-bits='1024' id='d2baa450'>
	<subrange length='16' type-id='7359adad' id='848d0938'/>
	</array-type-def>
	<type-decl name='unsigned short int' size-in-bits='16' id='8efea9e5'/>
	<enum-decl name='zpool_prop_t' naming-typedef-id='5d0c23fb' id='af1ba157'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZPOOL_PROP_INVAL' value='-1'/>
	<enumerator name='ZPOOL_PROP_NAME' value='0'/>
	<enumerator name='ZPOOL_PROP_SIZE' value='1'/>
	<enumerator name='ZPOOL_PROP_CAPACITY' value='2'/>
	<enumerator name='ZPOOL_PROP_ALTROOT' value='3'/>
	<enumerator name='ZPOOL_PROP_HEALTH' value='4'/>
	<enumerator name='ZPOOL_PROP_GUID' value='5'/>
	<enumerator name='ZPOOL_PROP_VERSION' value='6'/>
	<enumerator name='ZPOOL_PROP_BOOTFS' value='7'/>
	<enumerator name='ZPOOL_PROP_DELEGATION' value='8'/>
	<enumerator name='ZPOOL_PROP_AUTOREPLACE' value='9'/>
	<enumerator name='ZPOOL_PROP_CACHEFILE' value='10'/>
	<enumerator name='ZPOOL_PROP_FAILUREMODE' value='11'/>
	<enumerator name='ZPOOL_PROP_LISTSNAPS' value='12'/>
	<enumerator name='ZPOOL_PROP_AUTOEXPAND' value='13'/>
	<enumerator name='ZPOOL_PROP_DEDUPDITTO' value='14'/>
	<enumerator name='ZPOOL_PROP_DEDUPRATIO' value='15'/>
	<enumerator name='ZPOOL_PROP_FREE' value='16'/>
	<enumerator name='ZPOOL_PROP_ALLOCATED' value='17'/>
	<enumerator name='ZPOOL_PROP_READONLY' value='18'/>
	<enumerator name='ZPOOL_PROP_ASHIFT' value='19'/>
	<enumerator name='ZPOOL_PROP_COMMENT' value='20'/>
	<enumerator name='ZPOOL_PROP_EXPANDSZ' value='21'/>
	<enumerator name='ZPOOL_PROP_FREEING' value='22'/>
	<enumerator name='ZPOOL_PROP_FRAGMENTATION' value='23'/>
	<enumerator name='ZPOOL_PROP_LEAKED' value='24'/>
	<enumerator name='ZPOOL_PROP_MAXBLOCKSIZE' value='25'/>
	<enumerator name='ZPOOL_PROP_TNAME' value='26'/>
	<enumerator name='ZPOOL_PROP_MAXDNODESIZE' value='27'/>
	<enumerator name='ZPOOL_PROP_MULTIHOST' value='28'/>
	<enumerator name='ZPOOL_PROP_CHECKPOINT' value='29'/>
	<enumerator name='ZPOOL_PROP_LOAD_GUID' value='30'/>
	<enumerator name='ZPOOL_PROP_AUTOTRIM' value='31'/>
	<enumerator name='ZPOOL_PROP_COMPATIBILITY' value='32'/>
	<enumerator name='ZPOOL_PROP_BCLONEUSED' value='33'/>
	<enumerator name='ZPOOL_PROP_BCLONESAVED' value='34'/>
	<enumerator name='ZPOOL_PROP_BCLONERATIO' value='35'/>
	<enumerator name='ZPOOL_NUM_PROPS' value='36'/>
	</enum-decl>
	<typedef-decl name='zpool_prop_t' type-id='af1ba157' id='5d0c23fb'/>
	<typedef-decl name='regoff_t' type-id='95e97e5e' id='54a2a2a8'/>
	<class-decl name='regmatch_t' size-in-bits='64' is-struct='yes' naming-typedef-id='1b941664' visibility='default' id='4f932615'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='rm_so' type-id='54a2a2a8' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='rm_eo' type-id='54a2a2a8' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='regmatch_t' type-id='4f932615' id='1b941664'/>
	<typedef-decl name='__sighandler_t' type-id='03347643' id='8cdd9566'/>
	<typedef-decl name='ssize_t' type-id='41060289' id='79a0948f'/>
	<class-decl name='sigaction' size-in-bits='1216' is-struct='yes' visibility='default' id='fe391c48'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='__sigaction_handler' type-id='ac5ab595' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='sa_mask' type-id='b9c97942' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1088'>
	<var-decl name='sa_flags' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1152'>
	<var-decl name='sa_restorer' type-id='953b12f8' visibility='default'/>
	</data-member>
	</class-decl>
	<union-decl name='__anonymous_union__' size-in-bits='64' is-anonymous='yes' visibility='default' id='ac5ab595'>
	<data-member access='public'>
	<var-decl name='sa_handler' type-id='8cdd9566' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='sa_sigaction' type-id='6e756877' visibility='default'/>
	</data-member>
	</union-decl>
	<class-decl name='termios' size-in-bits='480' is-struct='yes' visibility='default' id='ad55d2bc'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='c_iflag' type-id='241ce6f8' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='c_oflag' type-id='241ce6f8' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='c_cflag' type-id='241ce6f8' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='c_lflag' type-id='241ce6f8' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='c_line' type-id='fb7c6451' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='136'>
	<var-decl name='c_cc' type-id='64177143' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='416'>
	<var-decl name='c_ispeed' type-id='6a8e8a14' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='c_ospeed' type-id='6a8e8a14' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='cc_t' type-id='002ac4a6' id='fb7c6451'/>
	<typedef-decl name='speed_t' type-id='f0981eeb' id='6a8e8a14'/>
	<typedef-decl name='tcflag_t' type-id='f0981eeb' id='241ce6f8'/>
	<typedef-decl name='__uid_t' type-id='f0981eeb' id='cc5fcceb'/>
	<typedef-decl name='__off_t' type-id='bd54fe1a' id='79989e9c'/>
	<typedef-decl name='__off64_t' type-id='bd54fe1a' id='724e4de6'/>
	<typedef-decl name='__pid_t' type-id='95e97e5e' id='3629bad8'/>
	<typedef-decl name='__clock_t' type-id='bd54fe1a' id='4d66c6d7'/>
	<typedef-decl name='__ssize_t' type-id='bd54fe1a' id='41060289'/>
	<typedef-decl name='FILE' type-id='ec1ed955' id='aa12d1ba'/>
	<class-decl name='__locale_struct' size-in-bits='1856' is-struct='yes' visibility='default' id='90cc1ce3'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='__locales' type-id='02b72c00' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='832'>
	<var-decl name='__ctype_b' type-id='31347b7a' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='896'>
	<var-decl name='__ctype_tolower' type-id='6d60f45d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='960'>
	<var-decl name='__ctype_toupper' type-id='6d60f45d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1024'>
	<var-decl name='__names' type-id='39e6f84a' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='__locale_t' type-id='f01e1813' id='b7ac9b5f'/>
	<class-decl name='__sigset_t' size-in-bits='1024' is-struct='yes' naming-typedef-id='b9c97942' visibility='default' id='2616147f'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='__val' type-id='d2baa450' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='__sigset_t' type-id='2616147f' id='b9c97942'/>
	<union-decl name='sigval' size-in-bits='64' visibility='default' id='a094b870'>
	<data-member access='public'>
	<var-decl name='sival_int' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='sival_ptr' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</union-decl>
	<typedef-decl name='__sigval_t' type-id='a094b870' id='eabacd01'/>
	<typedef-decl name='locale_t' type-id='b7ac9b5f' id='973a4f8d'/>
	<class-decl name='siginfo_t' size-in-bits='1024' is-struct='yes' naming-typedef-id='cb681f62' visibility='default' id='d8149419'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='si_signo' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='si_errno' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='si_code' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='__pad0' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='_sifields' type-id='ac5ab596' visibility='default'/>
	</data-member>
	</class-decl>
	<union-decl name='__anonymous_union__1' size-in-bits='896' is-anonymous='yes' visibility='default' id='ac5ab596'>
	<data-member access='public'>
	<var-decl name='_pad' type-id='47394ee0' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='_kill' type-id='e7f43f73' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='_timer' type-id='e7f43f74' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='_rt' type-id='e7f43f75' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='_sigchld' type-id='e7f43f76' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='_sigfault' type-id='e7f43f77' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='_sigpoll' type-id='e7f43f78' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='_sigsys' type-id='e7f43f79' visibility='default'/>
	</data-member>
	</union-decl>
	<class-decl name='__anonymous_struct__1' size-in-bits='64' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f73'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='si_pid' type-id='3629bad8' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='si_uid' type-id='cc5fcceb' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__2' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f74'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='si_tid' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='si_overrun' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='si_sigval' type-id='eabacd01' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__3' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f75'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='si_pid' type-id='3629bad8' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='si_uid' type-id='cc5fcceb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='si_sigval' type-id='eabacd01' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__4' size-in-bits='256' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f76'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='si_pid' type-id='3629bad8' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='si_uid' type-id='cc5fcceb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='si_status' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='si_utime' type-id='4d66c6d7' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='si_stime' type-id='4d66c6d7' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__5' size-in-bits='256' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f77'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='si_addr' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='si_addr_lsb' type-id='a2185560' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='_bounds' type-id='ac5ab597' visibility='default'/>
	</data-member>
	</class-decl>
	<union-decl name='__anonymous_union__2' size-in-bits='128' is-anonymous='yes' visibility='default' id='ac5ab597'>
	<data-member access='public'>
	<var-decl name='_addr_bnd' type-id='e7f43f7a' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='_pkey' type-id='62f1140c' visibility='default'/>
	</data-member>
	</union-decl>
	<class-decl name='__anonymous_struct__6' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f7a'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='_lower' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='_upper' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__7' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f78'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='si_band' type-id='bd54fe1a' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='si_fd' type-id='95e97e5e' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__anonymous_struct__8' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f79'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='_call_addr' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='_syscall' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='_arch' type-id='f0981eeb' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='siginfo_t' type-id='d8149419' id='cb681f62'/>
	<typedef-decl name='sigset_t' type-id='b9c97942' id='daf33c64'/>
	<typedef-decl name='_IO_lock_t' type-id='48b5725f' id='bb4788fa'/>
	<class-decl name='_IO_FILE' size-in-bits='1728' is-struct='yes' visibility='default' id='ec1ed955'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='_flags' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='_IO_read_ptr' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='_IO_read_end' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='_IO_read_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='_IO_write_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='_IO_write_ptr' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='_IO_write_end' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='_IO_buf_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='_IO_buf_end' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='576'>
	<var-decl name='_IO_save_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='640'>
	<var-decl name='_IO_backup_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='704'>
	<var-decl name='_IO_save_end' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='768'>
	<var-decl name='_markers' type-id='e4c6fa61' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='832'>
	<var-decl name='_chain' type-id='dca988a5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='896'>
	<var-decl name='_fileno' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='928'>
	<var-decl name='_flags2' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='960'>
	<var-decl name='_old_offset' type-id='79989e9c' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1024'>
	<var-decl name='_cur_column' type-id='8efea9e5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1040'>
	<var-decl name='_vtable_offset' type-id='28577a57' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1048'>
	<var-decl name='_shortbuf' type-id='89feb1ec' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1088'>
	<var-decl name='_lock' type-id='cecf4ea7' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1152'>
	<var-decl name='_offset' type-id='724e4de6' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1216'>
	<var-decl name='_codecvt' type-id='570f8c59' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1280'>
	<var-decl name='_wide_data' type-id='c65a1f29' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1344'>
	<var-decl name='_freeres_list' type-id='dca988a5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1408'>
	<var-decl name='_freeres_buf' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1472'>
	<var-decl name='__pad5' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1536'>
	<var-decl name='_mode' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1568'>
	<var-decl name='_unused2' type-id='664ac0b7' visibility='default'/>
	</data-member>
	</class-decl>
	<pointer-type-def type-id='aa12d1ba' size-in-bits='64' id='822cd80b'/>
	<qualified-type-def type-id='822cd80b' restrict='yes' id='e75a27e9'/>
	<pointer-type-def type-id='ec1ed955' size-in-bits='64' id='dca988a5'/>
	<pointer-type-def type-id='a4036571' size-in-bits='64' id='570f8c59'/>
	<pointer-type-def type-id='bb4788fa' size-in-bits='64' id='cecf4ea7'/>
	<pointer-type-def type-id='010ae0b9' size-in-bits='64' id='e4c6fa61'/>
	<pointer-type-def type-id='79bd3751' size-in-bits='64' id='c65a1f29'/>
	<pointer-type-def type-id='23de8b96' size-in-bits='64' id='38b51b3c'/>
	<pointer-type-def type-id='90cc1ce3' size-in-bits='64' id='f01e1813'/>
	<qualified-type-def type-id='9b23c9ad' restrict='yes' id='8c85230f'/>
	<qualified-type-def type-id='80f4b756' restrict='yes' id='9d26089a'/>
	<pointer-type-def type-id='80f4b756' size-in-bits='64' id='7d3cd834'/>
	<qualified-type-def type-id='95e97e5e' const='yes' id='2448a865'/>
	<pointer-type-def type-id='2448a865' size-in-bits='64' id='6d60f45d'/>
	<qualified-type-def type-id='aca3bac8' const='yes' id='2498fd78'/>
	<pointer-type-def type-id='2498fd78' size-in-bits='64' id='eed6c816'/>
	<qualified-type-def type-id='eed6c816' restrict='yes' id='a431a9da'/>
	<qualified-type-def type-id='fe391c48' const='yes' id='14a93b33'/>
	<pointer-type-def type-id='14a93b33' size-in-bits='64' id='9f68085b'/>
	<qualified-type-def type-id='9f68085b' restrict='yes' id='e2a5e6f9'/>
	<qualified-type-def type-id='ad55d2bc' const='yes' id='a46bf13f'/>
	<pointer-type-def type-id='a46bf13f' size-in-bits='64' id='eaec840f'/>
	<qualified-type-def type-id='002ac4a6' const='yes' id='ea86de29'/>
	<pointer-type-def type-id='ea86de29' size-in-bits='64' id='354f7eb9'/>
	<qualified-type-def type-id='8efea9e5' const='yes' id='3beb2af4'/>
	<pointer-type-def type-id='3beb2af4' size-in-bits='64' id='31347b7a'/>
	<pointer-type-def type-id='31347b7a' size-in-bits='64' id='c59e1ef0'/>
	<pointer-type-def type-id='1b941664' size-in-bits='64' id='7e2979d5'/>
	<qualified-type-def type-id='7e2979d5' restrict='yes' id='fc212857'/>
	<pointer-type-def type-id='fe391c48' size-in-bits='64' id='568dd84e'/>
	<qualified-type-def type-id='568dd84e' restrict='yes' id='3d8ee6f2'/>
	<pointer-type-def type-id='cb681f62' size-in-bits='64' id='185869c1'/>
	<pointer-type-def type-id='daf33c64' size-in-bits='64' id='9e80f729'/>
	<pointer-type-def type-id='b59d7dce' size-in-bits='64' id='78c01427'/>
	<qualified-type-def type-id='78c01427' restrict='yes' id='d19b2c25'/>
	<pointer-type-def type-id='ad55d2bc' size-in-bits='64' id='665a4eda'/>
	<pointer-type-def type-id='9c313c2d' size-in-bits='64' id='5d6479ae'/>
	<pointer-type-def type-id='ae3e8ca6' size-in-bits='64' id='d8774064'/>
	<pointer-type-def type-id='3502e3ff' size-in-bits='64' id='4dd26a40'/>
	<pointer-type-def type-id='ee076206' size-in-bits='64' id='953b12f8'/>
	<pointer-type-def type-id='f712e2b7' size-in-bits='64' id='03347643'/>
	<pointer-type-def type-id='ef70d893' size-in-bits='64' id='6e756877'/>
	<qualified-type-def type-id='eaa32e2f' restrict='yes' id='1b7446cd'/>
	<class-decl name='_IO_codecvt' is-struct='yes' visibility='default' is-declaration-only='yes' id='a4036571'/>
	<class-decl name='_IO_marker' is-struct='yes' visibility='default' is-declaration-only='yes' id='010ae0b9'/>
	<class-decl name='_IO_wide_data' is-struct='yes' visibility='default' is-declaration-only='yes' id='79bd3751'/>
	<class-decl name='__locale_data' is-struct='yes' visibility='default' is-declaration-only='yes' id='23de8b96'/>
	<function-decl name='zpool_get_prop_int' mangled-name='zpool_get_prop_int' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_prop_int'>
	<parameter type-id='4c81de99'/>
	<parameter type-id='5d0c23fb'/>
	<parameter type-id='debc6aa3'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='zfs_handle_dup' mangled-name='zfs_handle_dup' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_handle_dup'>
	<parameter type-id='9200a744'/>
	<return type-id='9200a744'/>
	</function-decl>
	<function-decl name='zfs_valid_proplist' mangled-name='zfs_valid_proplist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_valid_proplist'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='2e45de5d'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='9200a744'/>
	<parameter type-id='4c81de99'/>
	<parameter type-id='c19b74c3'/>
	<parameter type-id='80f4b756'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='zfs_prop_to_name' mangled-name='zfs_prop_to_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_to_name'>
	<parameter type-id='58603c44'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_iter_filesystems_v2' mangled-name='zfs_iter_filesystems_v2' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_filesystems_v2'>
	<parameter type-id='9200a744'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='d8e49ab9'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_parent_name' mangled-name='zfs_parent_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_parent_name'>
	<parameter type-id='9200a744'/>
	<parameter type-id='26a90f95'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_load_key' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='c19b74c3'/>
	<parameter type-id='ae3e8ca6'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_unload_key' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_change_key' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='ae3e8ca6'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_name_to_prop' mangled-name='zfs_name_to_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_name_to_prop'>
	<parameter type-id='80f4b756'/>
	<return type-id='58603c44'/>
	</function-decl>
	<function-decl name='nvlist_add_uint64' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='9c313c2d'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_string' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_uint64' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='22cce67b'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='5d6479ae'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_string' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='22cce67b'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='7d3cd834'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='fnvlist_alloc' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='__ctype_b_loc' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='c59e1ef0'/>
	</function-decl>
	<function-decl name='dlopen' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='dlsym' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='1b7446cd'/>
	<parameter type-id='9d26089a'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='dlerror' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='uselocale' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='973a4f8d'/>
	<return type-id='973a4f8d'/>
	</function-decl>
	<function-decl name='PKCS5_PBKDF2_HMAC_SHA1' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='354f7eb9'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='cf536864'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='regexec' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='a431a9da'/>
	<parameter type-id='9d26089a'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='fc212857'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='kill' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='3629bad8'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='sigemptyset' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9e80f729'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='sigaction' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='e2a5e6f9'/>
	<parameter type-id='3d8ee6f2'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='fclose' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='822cd80b'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='fflush' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='822cd80b'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='fdopen' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<return type-id='822cd80b'/>
	</function-decl>
	<function-decl name='fputc' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='822cd80b'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='__getdelim' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='8c85230f'/>
	<parameter type-id='d19b2c25'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='e75a27e9'/>
	<return type-id='41060289'/>
	</function-decl>
	<function-decl name='rewind' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='822cd80b'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='ferror' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='822cd80b'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='fileno' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='822cd80b'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='malloc' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b59d7dce'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='calloc' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='strdup' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='strerror_l' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='973a4f8d'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='tcgetattr' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='665a4eda'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='tcsetattr' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='eaec840f'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='close' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='getpid' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='3629bad8'/>
	</function-decl>
	<function-decl name='isatty' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='unlink' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='__open_too_many_args' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='__open_missing_mode' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='__printf_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<parameter is-variadic='yes'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='__asprintf_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='8c85230f'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='9d26089a'/>
	<parameter is-variadic='yes'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='__fread_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='1b7446cd'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='e75a27e9'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='__read_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='79a0948f'/>
	</function-decl>
	<function-decl name='zfs_crypto_get_encryption_root' mangled-name='zfs_crypto_get_encryption_root' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_crypto_get_encryption_root'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='37e3bd22' name='is_encroot'/>
	<parameter type-id='26a90f95' name='buf'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_crypto_create' mangled-name='zfs_crypto_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_crypto_create'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='26a90f95' name='parent_name'/>
	<parameter type-id='5ce45b60' name='props'/>
	<parameter type-id='5ce45b60' name='pool_props'/>
	<parameter type-id='c19b74c3' name='stdin_available'/>
	<parameter type-id='d8774064' name='wkeydata_out'/>
	<parameter type-id='4dd26a40' name='wkeylen_out'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_crypto_clone_check' mangled-name='zfs_crypto_clone_check' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_crypto_clone_check'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='9200a744' name='origin_zhp'/>
	<parameter type-id='26a90f95' name='parent_name'/>
	<parameter type-id='5ce45b60' name='props'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_crypto_attempt_load_keys' mangled-name='zfs_crypto_attempt_load_keys' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_crypto_attempt_load_keys'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='fsname'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_crypto_load_key' mangled-name='zfs_crypto_load_key' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_crypto_load_key'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='c19b74c3' name='noop'/>
	<parameter type-id='80f4b756' name='alt_keylocation'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_crypto_unload_key' mangled-name='zfs_crypto_unload_key' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_crypto_unload_key'>
	<parameter type-id='9200a744' name='zhp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_crypto_rewrap' mangled-name='zfs_crypto_rewrap' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_crypto_rewrap'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='5ce45b60' name='raw_props'/>
	<parameter type-id='c19b74c3' name='inheritkey'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_error_aux' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='80f4b756'/>
	<parameter is-variadic='yes'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-type size-in-bits='64' id='f712e2b7'>
	<parameter type-id='95e97e5e'/>
	<return type-id='48b5725f'/>
	</function-type>
	<function-type size-in-bits='64' id='ef70d893'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='185869c1'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='48b5725f'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_dataset.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='32' id='8e0573fd'>
	<subrange length='4' type-id='7359adad' id='16fe7105'/>
	</array-type-def>
	<class-decl name='prop_changelist' is-struct='yes' visibility='default' is-declaration-only='yes' id='d86edc51'/>
	<class-decl name='zprop_list' size-in-bits='448' is-struct='yes' visibility='default' id='bd9b4291'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='pl_prop' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='pl_user_prop' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='pl_next' type-id='9f1a1109' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='pl_all' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='pl_width' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='pl_recvd_width' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='pl_fixed' type-id='c19b74c3' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zprop_list_t' type-id='bd9b4291' id='bdb8ac4f'/>
	<class-decl name='renameflags' size-in-bits='32' is-struct='yes' visibility='default' id='7aee5792'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='recursive' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1'>
	<var-decl name='nounmount' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2'>
	<var-decl name='forceunmount' type-id='f0981eeb' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='renameflags_t' type-id='7aee5792' id='067170c2'/>
	<typedef-decl name='zfs_userspace_cb_t' type-id='ca64ff60' id='16c5f410'/>
	<enum-decl name='lzc_dataset_type' id='bc9887f1'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='LZC_DATSET_TYPE_ZFS' value='2'/>
	<enumerator name='LZC_DATSET_TYPE_ZVOL' value='3'/>
	</enum-decl>
	<typedef-decl name='avl_index_t' type-id='e475ab95' id='fba6cb51'/>
	<enum-decl name='zfs_userquota_prop_t' naming-typedef-id='279fde6a' id='5258d2f6'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZFS_PROP_USERUSED' value='0'/>
	<enumerator name='ZFS_PROP_USERQUOTA' value='1'/>
	<enumerator name='ZFS_PROP_GROUPUSED' value='2'/>
	<enumerator name='ZFS_PROP_GROUPQUOTA' value='3'/>
	<enumerator name='ZFS_PROP_USEROBJUSED' value='4'/>
	<enumerator name='ZFS_PROP_USEROBJQUOTA' value='5'/>
	<enumerator name='ZFS_PROP_GROUPOBJUSED' value='6'/>
	<enumerator name='ZFS_PROP_GROUPOBJQUOTA' value='7'/>
	<enumerator name='ZFS_PROP_PROJECTUSED' value='8'/>
	<enumerator name='ZFS_PROP_PROJECTQUOTA' value='9'/>
	<enumerator name='ZFS_PROP_PROJECTOBJUSED' value='10'/>
	<enumerator name='ZFS_PROP_PROJECTOBJQUOTA' value='11'/>
	<enumerator name='ZFS_NUM_USERQUOTA_PROPS' value='12'/>
	</enum-decl>
	<typedef-decl name='zfs_userquota_prop_t' type-id='5258d2f6' id='279fde6a'/>
	<enum-decl name='zfs_wait_activity_t' naming-typedef-id='3024501a' id='527d5dc6'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZFS_WAIT_DELETEQ' value='0'/>
	<enumerator name='ZFS_WAIT_NUM_ACTIVITIES' value='1'/>
	</enum-decl>
	<typedef-decl name='zfs_wait_activity_t' type-id='527d5dc6' id='3024501a'/>
	<enum-decl name='namecheck_err_t' naming-typedef-id='8e0af06e' id='f43bbcda'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='NAME_ERR_LEADING_SLASH' value='0'/>
	<enumerator name='NAME_ERR_EMPTY_COMPONENT' value='1'/>
	<enumerator name='NAME_ERR_TRAILING_SLASH' value='2'/>
	<enumerator name='NAME_ERR_INVALCHAR' value='3'/>
	<enumerator name='NAME_ERR_MULTIPLE_DELIMITERS' value='4'/>
	<enumerator name='NAME_ERR_NOLETTER' value='5'/>
	<enumerator name='NAME_ERR_RESERVED' value='6'/>
	<enumerator name='NAME_ERR_DISKLIKE' value='7'/>
	<enumerator name='NAME_ERR_TOOLONG' value='8'/>
	<enumerator name='NAME_ERR_SELF_REF' value='9'/>
	<enumerator name='NAME_ERR_PARENT_REF' value='10'/>
	<enumerator name='NAME_ERR_NO_AT' value='11'/>
	<enumerator name='NAME_ERR_NO_POUND' value='12'/>
	</enum-decl>
	<typedef-decl name='namecheck_err_t' type-id='f43bbcda' id='8e0af06e'/>
	<enum-decl name='zprop_type_t' naming-typedef-id='31429eff' id='87676253'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='PROP_TYPE_NUMBER' value='0'/>
	<enumerator name='PROP_TYPE_STRING' value='1'/>
	<enumerator name='PROP_TYPE_INDEX' value='2'/>
	</enum-decl>
	<typedef-decl name='zprop_type_t' type-id='87676253' id='31429eff'/>
	<class-decl name='mnttab' size-in-bits='256' is-struct='yes' visibility='default' id='1b055409'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='mnt_special' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='mnt_mountp' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='mnt_fstype' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='mnt_mntopts' type-id='26a90f95' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='group' size-in-bits='256' is-struct='yes' visibility='default' id='01a1b934'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='gr_name' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='gr_passwd' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='gr_gid' type-id='d94ec6d9' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='gr_mem' type-id='9b23c9ad' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='mntent' size-in-bits='320' is-struct='yes' visibility='default' id='56fe4a37'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='mnt_fsname' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='mnt_dir' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='mnt_type' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='mnt_opts' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='mnt_freq' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='mnt_passno' type-id='95e97e5e' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='passwd' size-in-bits='384' is-struct='yes' visibility='default' id='a63d15a3'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='pw_name' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='pw_passwd' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='pw_uid' type-id='cc5fcceb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='160'>
	<var-decl name='pw_gid' type-id='d94ec6d9' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='pw_gecos' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='pw_dir' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='pw_shell' type-id='26a90f95' visibility='default'/>
	</data-member>
	</class-decl>
	<union-decl name='pthread_mutexattr_t' size-in-bits='32' naming-typedef-id='8afd6070' visibility='default' id='7300eb00'>
	<data-member access='public'>
	<var-decl name='__size' type-id='8e0573fd' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='__align' type-id='95e97e5e' visibility='default'/>
	</data-member>
	</union-decl>
	<typedef-decl name='pthread_mutexattr_t' type-id='7300eb00' id='8afd6070'/>
	<typedef-decl name='int64_t' type-id='0c9942d2' id='9da381c4'/>
	<typedef-decl name='__int64_t' type-id='bd54fe1a' id='0c9942d2'/>
	<typedef-decl name='__gid_t' type-id='f0981eeb' id='d94ec6d9'/>
	<typedef-decl name='__time_t' type-id='bd54fe1a' id='65eda9c0'/>
	<class-decl name='tm' size-in-bits='448' is-struct='yes' visibility='default' id='dddf6ca2'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='tm_sec' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='tm_min' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='tm_hour' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='tm_mday' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='tm_mon' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='160'>
	<var-decl name='tm_year' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='tm_wday' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='224'>
	<var-decl name='tm_yday' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='tm_isdst' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='tm_gmtoff' type-id='bd54fe1a' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='tm_zone' type-id='80f4b756' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='time_t' type-id='65eda9c0' id='c9d12d66'/>
	<typedef-decl name='uid_t' type-id='cc5fcceb' id='354978ed'/>
	<typedef-decl name='prop_changelist_t' type-id='d86edc51' id='eae6431d'/>
	<pointer-type-def type-id='fba6cb51' size-in-bits='64' id='32adbf30'/>
	<pointer-type-def type-id='f20fbd51' size-in-bits='64' id='a3681dea'/>
	<qualified-type-def type-id='26a90f95' restrict='yes' id='266fe297'/>
	<qualified-type-def type-id='56fe4a37' const='yes' id='a75125ce'/>
	<pointer-type-def type-id='a75125ce' size-in-bits='64' id='48bea5ec'/>
	<qualified-type-def type-id='8afd6070' const='yes' id='1d853360'/>
	<pointer-type-def type-id='1d853360' size-in-bits='64' id='c2afbd7e'/>
	<qualified-type-def type-id='c9d12d66' const='yes' id='588b3216'/>
	<pointer-type-def type-id='588b3216' size-in-bits='64' id='9f201474'/>
	<qualified-type-def type-id='9f201474' restrict='yes' id='d6e2847c'/>
	<qualified-type-def type-id='dddf6ca2' const='yes' id='e824a34f'/>
	<pointer-type-def type-id='e824a34f' size-in-bits='64' id='d6ad37ff'/>
	<qualified-type-def type-id='d6ad37ff' restrict='yes' id='f8c6051d'/>
	<qualified-type-def type-id='9c313c2d' const='yes' id='c3b7ba7d'/>
	<pointer-type-def type-id='c3b7ba7d' size-in-bits='64' id='713a56f5'/>
	<pointer-type-def type-id='01a1b934' size-in-bits='64' id='566b3f52'/>
	<qualified-type-def type-id='566b3f52' restrict='yes' id='c878edd6'/>
	<pointer-type-def type-id='566b3f52' size-in-bits='64' id='82d4e9e8'/>
	<qualified-type-def type-id='82d4e9e8' restrict='yes' id='aa19c230'/>
	<pointer-type-def type-id='7e291ce6' size-in-bits='64' id='ca64ff60'/>
	<pointer-type-def type-id='9da381c4' size-in-bits='64' id='cb785ebf'/>
	<pointer-type-def type-id='1b055409' size-in-bits='64' id='9d424d31'/>
	<pointer-type-def type-id='8e0af06e' size-in-bits='64' id='053457bd'/>
	<pointer-type-def type-id='857bb57e' size-in-bits='64' id='75be733c'/>
	<pointer-type-def type-id='a63d15a3' size-in-bits='64' id='a195f4a3'/>
	<qualified-type-def type-id='a195f4a3' restrict='yes' id='33518961'/>
	<pointer-type-def type-id='a195f4a3' size-in-bits='64' id='e80ff3ab'/>
	<qualified-type-def type-id='e80ff3ab' restrict='yes' id='8f2c7109'/>
	<pointer-type-def type-id='eae6431d' size-in-bits='64' id='0d41d328'/>
	<pointer-type-def type-id='7a6844eb' size-in-bits='64' id='18c91f9e'/>
	<pointer-type-def type-id='dddf6ca2' size-in-bits='64' id='d915a820'/>
	<qualified-type-def type-id='d915a820' restrict='yes' id='f099ad08'/>
	<pointer-type-def type-id='5d6479ae' size-in-bits='64' id='892b4acc'/>
	<pointer-type-def type-id='bd9b4291' size-in-bits='64' id='9f1a1109'/>
	<pointer-type-def type-id='bdb8ac4f' size-in-bits='64' id='3a9b2288'/>
	<pointer-type-def type-id='3a9b2288' size-in-bits='64' id='e4378506'/>
	<class-decl name='prop_changelist' is-struct='yes' visibility='default' is-declaration-only='yes' id='d86edc51'/>
	<function-decl name='zpool_open' mangled-name='zpool_open' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_open'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='80f4b756'/>
	<return type-id='4c81de99'/>
	</function-decl>
	<function-decl name='zpool_open_canfail' mangled-name='zpool_open_canfail' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_open_canfail'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='80f4b756'/>
	<return type-id='4c81de99'/>
	</function-decl>
	<function-decl name='zpool_close' mangled-name='zpool_close' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_close'>
	<parameter type-id='4c81de99'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zpool_get_name' mangled-name='zpool_get_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_name'>
	<parameter type-id='4c81de99'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zpool_get_prop' mangled-name='zpool_get_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_prop'>
	<parameter type-id='4c81de99'/>
	<parameter type-id='5d0c23fb'/>
	<parameter type-id='26a90f95'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='debc6aa3'/>
	<parameter type-id='c19b74c3'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_default_string' mangled-name='zfs_prop_default_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_default_string'>
	<parameter type-id='58603c44'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_prop_default_numeric' mangled-name='zfs_prop_default_numeric' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_default_numeric'>
	<parameter type-id='58603c44'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='zpool_prop_get_feature' mangled-name='zpool_prop_get_feature' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_get_feature'>
	<parameter type-id='4c81de99'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='26a90f95'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_iter_snapshots_v2' mangled-name='zfs_iter_snapshots_v2' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_snapshots_v2'>
	<parameter type-id='9200a744'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='d8e49ab9'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='9c313c2d'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_iter_bookmarks_v2' mangled-name='zfs_iter_bookmarks_v2' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_bookmarks_v2'>
	<parameter type-id='9200a744'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='d8e49ab9'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_destroy_snaps_nvl_os' mangled-name='zfs_destroy_snaps_nvl_os' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_destroy_snaps_nvl_os'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='5ce45b60'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_nicestrtonum' mangled-name='zfs_nicestrtonum' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_nicestrtonum'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='5d6479ae'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_snapshot' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='857bb57e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_create' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='bc9887f1'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='ae3e8ca6'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_clone' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='5ce45b60'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_promote' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='26a90f95'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_destroy_snaps' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='c19b74c3'/>
	<parameter type-id='857bb57e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_get_bookmarks' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='857bb57e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_destroy_bookmarks' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='857bb57e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_hold' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='857bb57e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_release' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='857bb57e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_get_holds' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='857bb57e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_exists' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='lzc_rollback_to' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_destroy' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_channel_program_nosync' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='857bb57e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_wait_fs' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='3024501a'/>
	<parameter type-id='37e3bd22'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_nicebytes' mangled-name='zfs_nicebytes' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_nicebytes'>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='26a90f95'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_nicenum' mangled-name='zfs_nicenum' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_nicenum'>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='26a90f95'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='avl_create' mangled-name='avl_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_create'>
	<parameter type-id='a3681dea'/>
	<parameter type-id='585e1de9'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='avl_find' mangled-name='avl_find' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_find'>
	<parameter type-id='a3681dea'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='32adbf30'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='avl_add' mangled-name='avl_add' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_add'>
	<parameter type-id='a3681dea'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='avl_remove' mangled-name='avl_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_remove'>
	<parameter type-id='a3681dea'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='avl_numnodes' mangled-name='avl_numnodes' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_numnodes'>
	<parameter type-id='a3681dea'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='avl_destroy_nodes' mangled-name='avl_destroy_nodes' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_destroy_nodes'>
	<parameter type-id='a3681dea'/>
	<parameter type-id='63e171df'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='avl_destroy' mangled-name='avl_destroy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_destroy'>
	<parameter type-id='a3681dea'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_prop_readonly' mangled-name='zfs_prop_readonly' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_readonly'>
	<parameter type-id='58603c44'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_prop_inheritable' mangled-name='zfs_prop_inheritable' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_inheritable'>
	<parameter type-id='58603c44'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_prop_setonce' mangled-name='zfs_prop_setonce' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_setonce'>
	<parameter type-id='58603c44'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_prop_encryption_key_param' mangled-name='zfs_prop_encryption_key_param' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_encryption_key_param'>
	<parameter type-id='58603c44'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_prop_valid_keylocation' mangled-name='zfs_prop_valid_keylocation' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_valid_keylocation'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='c19b74c3'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_prop_user' mangled-name='zfs_prop_user' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_user'>
	<parameter type-id='80f4b756'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_prop_userquota' mangled-name='zfs_prop_userquota' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_userquota'>
	<parameter type-id='80f4b756'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_prop_written' mangled-name='zfs_prop_written' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_written'>
	<parameter type-id='80f4b756'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_prop_index_to_string' mangled-name='zfs_prop_index_to_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_index_to_string'>
	<parameter type-id='58603c44'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='7d3cd834'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_valid_for_type' mangled-name='zfs_prop_valid_for_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_valid_for_type'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='2e45de5d'/>
	<parameter type-id='c19b74c3'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='nvlist_alloc' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='857bb57e'/>
	<parameter type-id='3502e3ff'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_size' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='78c01427'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_pack' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='9b23c9ad'/>
	<parameter type-id='78c01427'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_unpack' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='26a90f95'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='857bb57e'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_nvlist' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='22cce67b'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_uint64_array' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='713a56f5'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_remove' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='8d0687d2'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_remove_all' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_int64' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='22cce67b'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='cb785ebf'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_uint64_array' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='892b4acc'/>
	<parameter type-id='4dd26a40'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_nvlist_array' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='75be733c'/>
	<parameter type-id='4dd26a40'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_empty' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='22cce67b'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='nvpair_type' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='dace003f'/>
	<return type-id='8d0687d2'/>
	</function-decl>
	<function-decl name='nvpair_value_uint64' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='dace003f'/>
	<parameter type-id='5d6479ae'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_string' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='dace003f'/>
	<parameter type-id='7d3cd834'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='fnvlist_free' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_boolean' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_uint64' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='9c313c2d'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_string' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_nvlist' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='5ce45b60'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_uint64' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='22cce67b'/>
	<parameter type-id='80f4b756'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_string' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='22cce67b'/>
	<parameter type-id='80f4b756'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_nvlist' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='fnvpair_value_int32' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='dace003f'/>
	<return type-id='3ff5601b'/>
	</function-decl>
	<function-decl name='fnvpair_value_uint64' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='dace003f'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='entity_namecheck' mangled-name='entity_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='entity_namecheck'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='053457bd'/>
	<parameter type-id='26a90f95'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='dataset_nestcheck' mangled-name='dataset_nestcheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='dataset_nestcheck'>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='mountpoint_namecheck' mangled-name='mountpoint_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='mountpoint_namecheck'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='053457bd'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_get_type' mangled-name='zfs_prop_get_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_type'>
	<parameter type-id='58603c44'/>
	<return type-id='31429eff'/>
	</function-decl>
	<function-decl name='sa_validate_shareopts' mangled-name='sa_validate_shareopts' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='sa_validate_shareopts'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='9155d4b5'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='getmntany' mangled-name='getmntany' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='getmntany'>
	<parameter type-id='822cd80b'/>
	<parameter type-id='9d424d31'/>
	<parameter type-id='9d424d31'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='_sol_getmntent' mangled-name='_sol_getmntent' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='_sol_getmntent'>
	<parameter type-id='822cd80b'/>
	<parameter type-id='9d424d31'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='getgrnam_r' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9d26089a'/>
	<parameter type-id='c878edd6'/>
	<parameter type-id='266fe297'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='aa19c230'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='hasmntopt' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='48bea5ec'/>
	<parameter type-id='80f4b756'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='pthread_mutex_init' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='18c91f9e'/>
	<parameter type-id='c2afbd7e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_mutex_destroy' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='18c91f9e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_mutex_lock' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='18c91f9e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_mutex_unlock' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='18c91f9e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='getpwnam_r' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9d26089a'/>
	<parameter type-id='33518961'/>
	<parameter type-id='266fe297'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='8f2c7109'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='strtol' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9d26089a'/>
	<parameter type-id='8c85230f'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='bd54fe1a'/>
	</function-decl>
	<function-decl name='strtoul' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9d26089a'/>
	<parameter type-id='8c85230f'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='7359adad'/>
	</function-decl>
	<function-decl name='abort' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='strrchr' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='strcspn' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='strstr' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='strsep' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='8c85230f'/>
	<parameter type-id='9d26089a'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='strftime' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='266fe297'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='9d26089a'/>
	<parameter type-id='f8c6051d'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='localtime_r' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='d6e2847c'/>
	<parameter type-id='f099ad08'/>
	<return type-id='d915a820'/>
	</function-decl>
	<function-decl name='__fprintf_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='e75a27e9'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='9d26089a'/>
	<parameter is-variadic='yes'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='ioctl' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='7359adad'/>
	<parameter is-variadic='yes'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_type_to_name' mangled-name='zfs_type_to_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_type_to_name'>
	<parameter type-id='2e45de5d' name='type'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_name_valid' mangled-name='zfs_name_valid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_name_valid'>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='2e45de5d' name='type'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_free_handles' mangled-name='zpool_free_handles' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_free_handles'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_bookmark_exists' mangled-name='zfs_bookmark_exists' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_bookmark_exists'>
	<parameter type-id='80f4b756' name='path'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='libzfs_mnttab_init' mangled-name='libzfs_mnttab_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_mnttab_init'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='libzfs_mnttab_fini' mangled-name='libzfs_mnttab_fini' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_mnttab_fini'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='libzfs_mnttab_cache' mangled-name='libzfs_mnttab_cache' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_mnttab_cache'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='c19b74c3' name='enable'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='libzfs_mnttab_find' mangled-name='libzfs_mnttab_find' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_mnttab_find'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='fsname'/>
	<parameter type-id='9d424d31' name='entry'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='libzfs_mnttab_add' mangled-name='libzfs_mnttab_add' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_mnttab_add'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='special'/>
	<parameter type-id='80f4b756' name='mountp'/>
	<parameter type-id='80f4b756' name='mntopts'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='libzfs_mnttab_remove' mangled-name='libzfs_mnttab_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_mnttab_remove'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='fsname'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_spa_version' mangled-name='zfs_spa_version' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_spa_version'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='7292109c' name='spa_version'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_set' mangled-name='zfs_prop_set' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_set'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='propname'/>
	<parameter type-id='80f4b756' name='propval'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_set_list' mangled-name='zfs_prop_set_list' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_set_list'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='5ce45b60' name='props'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_set_list_flags' mangled-name='zfs_prop_set_list_flags' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_set_list_flags'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='5ce45b60' name='props'/>
	<parameter type-id='95e97e5e' name='flags'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_inherit' mangled-name='zfs_prop_inherit' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_inherit'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='propname'/>
	<parameter type-id='c19b74c3' name='received'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='getprop_uint64' mangled-name='getprop_uint64' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='getprop_uint64'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='58603c44' name='prop'/>
	<parameter type-id='7d3cd834' name='source'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='zfs_prop_get_recvd' mangled-name='zfs_prop_get_recvd' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_recvd'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='propname'/>
	<parameter type-id='26a90f95' name='propbuf'/>
	<parameter type-id='b59d7dce' name='proplen'/>
	<parameter type-id='c19b74c3' name='literal'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_get_clones_nvl' mangled-name='zfs_get_clones_nvl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_clones_nvl'>
	<parameter type-id='9200a744' name='zhp'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='zfs_prop_get_numeric' mangled-name='zfs_prop_get_numeric' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_numeric'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='58603c44' name='prop'/>
	<parameter type-id='5d6479ae' name='value'/>
	<parameter type-id='debc6aa3' name='src'/>
	<parameter type-id='26a90f95' name='statbuf'/>
	<parameter type-id='b59d7dce' name='statlen'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_get_userquota_int' mangled-name='zfs_prop_get_userquota_int' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_userquota_int'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='propname'/>
	<parameter type-id='5d6479ae' name='propvalue'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_get_userquota' mangled-name='zfs_prop_get_userquota' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_userquota'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='propname'/>
	<parameter type-id='26a90f95' name='propbuf'/>
	<parameter type-id='95e97e5e' name='proplen'/>
	<parameter type-id='c19b74c3' name='literal'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_get_written_int' mangled-name='zfs_prop_get_written_int' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_written_int'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='propname'/>
	<parameter type-id='5d6479ae' name='propvalue'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_get_written' mangled-name='zfs_prop_get_written' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_written'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='propname'/>
	<parameter type-id='26a90f95' name='propbuf'/>
	<parameter type-id='95e97e5e' name='proplen'/>
	<parameter type-id='c19b74c3' name='literal'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_get_pool_name' mangled-name='zfs_get_pool_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_pool_name'>
	<parameter type-id='fcd57163' name='zhp'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_get_type' mangled-name='zfs_get_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_type'>
	<parameter type-id='fcd57163' name='zhp'/>
	<return type-id='2e45de5d'/>
	</function-decl>
	<function-decl name='zfs_get_underlying_type' mangled-name='zfs_get_underlying_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_underlying_type'>
	<parameter type-id='fcd57163' name='zhp'/>
	<return type-id='2e45de5d'/>
	</function-decl>
	<function-decl name='zfs_dataset_exists' mangled-name='zfs_dataset_exists' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_dataset_exists'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='2e45de5d' name='types'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_create_ancestors' mangled-name='zfs_create_ancestors' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_create_ancestors'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='path'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_create' mangled-name='zfs_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_create'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='2e45de5d' name='type'/>
	<parameter type-id='5ce45b60' name='props'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_destroy' mangled-name='zfs_destroy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_destroy'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='c19b74c3' name='defer'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_destroy_snaps' mangled-name='zfs_destroy_snaps' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_destroy_snaps'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='26a90f95' name='snapname'/>
	<parameter type-id='c19b74c3' name='defer'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_destroy_snaps_nvl' mangled-name='zfs_destroy_snaps_nvl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_destroy_snaps_nvl'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='5ce45b60' name='snaps'/>
	<parameter type-id='c19b74c3' name='defer'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_clone' mangled-name='zfs_clone' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_clone'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='target'/>
	<parameter type-id='5ce45b60' name='props'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_promote' mangled-name='zfs_promote' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_promote'>
	<parameter type-id='9200a744' name='zhp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_snapshot_nvl' mangled-name='zfs_snapshot_nvl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_snapshot_nvl'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='5ce45b60' name='snaps'/>
	<parameter type-id='5ce45b60' name='props'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_snapshot' mangled-name='zfs_snapshot' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_snapshot'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='c19b74c3' name='recursive'/>
	<parameter type-id='5ce45b60' name='props'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_rollback' mangled-name='zfs_rollback' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_rollback'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='9200a744' name='snap'/>
	<parameter type-id='c19b74c3' name='force'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_rename' mangled-name='zfs_rename' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_rename'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='target'/>
	<parameter type-id='067170c2' name='flags'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_get_all_props' mangled-name='zfs_get_all_props' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_all_props'>
	<parameter type-id='9200a744' name='zhp'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='zfs_get_recvd_props' mangled-name='zfs_get_recvd_props' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_recvd_props'>
	<parameter type-id='9200a744' name='zhp'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='zfs_get_user_props' mangled-name='zfs_get_user_props' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_user_props'>
	<parameter type-id='9200a744' name='zhp'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='zfs_expand_proplist' mangled-name='zfs_expand_proplist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_expand_proplist'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='e4378506' name='plp'/>
	<parameter type-id='c19b74c3' name='received'/>
	<parameter type-id='c19b74c3' name='literal'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prune_proplist' mangled-name='zfs_prune_proplist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prune_proplist'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='ae3e8ca6' name='props'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_smb_acl_add' mangled-name='zfs_smb_acl_add' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_smb_acl_add'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='26a90f95' name='dataset'/>
	<parameter type-id='26a90f95' name='path'/>
	<parameter type-id='26a90f95' name='resource'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_smb_acl_remove' mangled-name='zfs_smb_acl_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_smb_acl_remove'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='26a90f95' name='dataset'/>
	<parameter type-id='26a90f95' name='path'/>
	<parameter type-id='26a90f95' name='resource'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_smb_acl_purge' mangled-name='zfs_smb_acl_purge' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_smb_acl_purge'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='26a90f95' name='dataset'/>
	<parameter type-id='26a90f95' name='path'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_smb_acl_rename' mangled-name='zfs_smb_acl_rename' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_smb_acl_rename'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='26a90f95' name='dataset'/>
	<parameter type-id='26a90f95' name='path'/>
	<parameter type-id='26a90f95' name='oldname'/>
	<parameter type-id='26a90f95' name='newname'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_userspace' mangled-name='zfs_userspace' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_userspace'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='279fde6a' name='type'/>
	<parameter type-id='16c5f410' name='func'/>
	<parameter type-id='eaa32e2f' name='arg'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_hold' mangled-name='zfs_hold' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_hold'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='snapname'/>
	<parameter type-id='80f4b756' name='tag'/>
	<parameter type-id='c19b74c3' name='recursive'/>
	<parameter type-id='95e97e5e' name='cleanup_fd'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_hold_nvl' mangled-name='zfs_hold_nvl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_hold_nvl'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='95e97e5e' name='cleanup_fd'/>
	<parameter type-id='5ce45b60' name='holds'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_release' mangled-name='zfs_release' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_release'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='snapname'/>
	<parameter type-id='80f4b756' name='tag'/>
	<parameter type-id='c19b74c3' name='recursive'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_get_fsacl' mangled-name='zfs_get_fsacl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_fsacl'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='857bb57e' name='nvl'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_set_fsacl' mangled-name='zfs_set_fsacl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_set_fsacl'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='c19b74c3' name='un'/>
	<parameter type-id='5ce45b60' name='nvl'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_get_holds' mangled-name='zfs_get_holds' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_holds'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='857bb57e' name='nvl'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zvol_volsize_to_reservation' mangled-name='zvol_volsize_to_reservation' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zvol_volsize_to_reservation'>
	<parameter type-id='4c81de99' name='zph'/>
	<parameter type-id='9c313c2d' name='volsize'/>
	<parameter type-id='5ce45b60' name='props'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='zfs_wait_status' mangled-name='zfs_wait_status' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_wait_status'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='3024501a' name='activity'/>
	<parameter type-id='37e3bd22' name='missing'/>
	<parameter type-id='37e3bd22' name='waited'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_error_fmt' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<parameter is-variadic='yes'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_standard_error_fmt' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<parameter is-variadic='yes'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_setprop_error' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='58603c44'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='26a90f95'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zprop_parse_value' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='3fa542f0'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='2e45de5d'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='7d3cd834'/>
	<parameter type-id='5d6479ae'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zprop_expand_list' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='e4378506'/>
	<parameter type-id='2e45de5d'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zcmd_write_src_nvlist' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='e4ec4540'/>
	<parameter type-id='5ce45b60'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='changelist_prefix' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='0d41d328'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='changelist_postfix' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='0d41d328'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='changelist_rename' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='0d41d328'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='changelist_remove' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='0d41d328'/>
	<parameter type-id='80f4b756'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='changelist_free' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='0d41d328'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='changelist_gather' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9200a744'/>
	<parameter type-id='58603c44'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='0d41d328'/>
	</function-decl>
	<function-decl name='changelist_haszonedchild' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='0d41d328'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_name_valid' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='c19b74c3'/>
	<parameter type-id='80f4b756'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-type size-in-bits='64' id='7e291ce6'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='354978ed'/>
	<parameter type-id='9c313c2d'/>
	<return type-id='95e97e5e'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_diff.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='448' id='6093ff7c'>
	<subrange length='56' type-id='7359adad' id='f8137894'/>
	</array-type-def>
	<typedef-decl name='pthread_t' type-id='7359adad' id='4051f5e7'/>
	<union-decl name='pthread_attr_t' size-in-bits='448' visibility='default' id='b63afacd'>
	<data-member access='public'>
	<var-decl name='__size' type-id='6093ff7c' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='__align' type-id='bd54fe1a' visibility='default'/>
	</data-member>
	</union-decl>
	<typedef-decl name='pthread_attr_t' type-id='b63afacd' id='7d8569fd'/>
	<class-decl name='differ_info' size-in-bits='9088' is-struct='yes' visibility='default' id='d41965ee'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='zhp' type-id='9200a744' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='fromsnap' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='frommnt' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='tosnap' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='tomnt' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='ds' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='dsmnt' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='tmpsnap' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='errbuf' type-id='b54ce520' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='8704'>
	<var-decl name='isclone' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='8736'>
	<var-decl name='scripted' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='8768'>
	<var-decl name='classify' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='8800'>
	<var-decl name='timestamped' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='8832'>
	<var-decl name='no_mangle' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='8896'>
	<var-decl name='shares' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='8960'>
	<var-decl name='zerr' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='8992'>
	<var-decl name='cleanupfd' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='9024'>
	<var-decl name='outputfd' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='9056'>
	<var-decl name='datafd' type-id='95e97e5e' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='differ_info_t' type-id='d41965ee' id='e8525f0e'/>
	<qualified-type-def type-id='7d8569fd' const='yes' id='e06dee2d'/>
	<pointer-type-def type-id='e06dee2d' size-in-bits='64' id='540db505'/>
	<qualified-type-def type-id='540db505' restrict='yes' id='e1815e87'/>
	<pointer-type-def type-id='e8525f0e' size-in-bits='64' id='ee78f675'/>
	<pointer-type-def type-id='4051f5e7' size-in-bits='64' id='e01b5462'/>
	<qualified-type-def type-id='e01b5462' restrict='yes' id='cc338b26'/>
	<pointer-type-def type-id='cd5d79f4' size-in-bits='64' id='5ad9edb6'/>
	<function-decl name='is_mounted' mangled-name='is_mounted' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='is_mounted'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='9b23c9ad'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='color_start' mangled-name='color_start' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='color_start'>
	<parameter type-id='80f4b756'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='color_end' mangled-name='color_end' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='color_end'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='pthread_create' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='cc338b26'/>
	<parameter type-id='e1815e87'/>
	<parameter type-id='5ad9edb6'/>
	<parameter type-id='1b7446cd'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_join' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='4051f5e7'/>
	<parameter type-id='63e171df'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_cancel' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='4051f5e7'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='fputs' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9d26089a'/>
	<parameter type-id='e75a27e9'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pipe2' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='7292109c'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_show_diffs' mangled-name='zfs_show_diffs' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_show_diffs'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='95e97e5e' name='outfd'/>
	<parameter type-id='80f4b756' name='fromsnap'/>
	<parameter type-id='80f4b756' name='tosnap'/>
	<parameter type-id='95e97e5e' name='flags'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_asprintf' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='80f4b756'/>
	<parameter is-variadic='yes'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='zfs_validate_name' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='c19b74c3'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='find_shares_object' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='ee78f675'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-type size-in-bits='64' id='cd5d79f4'>
	<parameter type-id='eaa32e2f'/>
	<return type-id='eaa32e2f'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_import.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='03085adc' size-in-bits='192' id='083f8d58'>
	<subrange length='3' type-id='7359adad' id='56f209d2'/>
	</array-type-def>
	<typedef-decl name='refresh_config_func_t' type-id='29f040d2' id='b7c58eaa'/>
	<typedef-decl name='pool_active_func_t' type-id='baa42fef' id='de5d1d8f'/>
	<class-decl name='pool_config_ops' size-in-bits='128' is-struct='yes' visibility='default' id='8b092c69'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='pco_refresh_config' type-id='e7c00489' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='pco_pool_active' type-id='9eadf5e0' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='pool_config_ops_t' type-id='1a21babe' id='b1e62775'/>
	<enum-decl name='pool_state' id='4871ac24'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='POOL_STATE_ACTIVE' value='0'/>
	<enumerator name='POOL_STATE_EXPORTED' value='1'/>
	<enumerator name='POOL_STATE_DESTROYED' value='2'/>
	<enumerator name='POOL_STATE_SPARE' value='3'/>
	<enumerator name='POOL_STATE_L2CACHE' value='4'/>
	<enumerator name='POOL_STATE_UNINITIALIZED' value='5'/>
	<enumerator name='POOL_STATE_UNAVAIL' value='6'/>
	<enumerator name='POOL_STATE_POTENTIALLY_ACTIVE' value='7'/>
	</enum-decl>
	<typedef-decl name='pool_state_t' type-id='4871ac24' id='084a08a3'/>
	<class-decl name='stat64' size-in-bits='1152' is-struct='yes' visibility='default' id='0bbec9cd'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='st_dev' type-id='35ed8932' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='st_ino' type-id='71288a47' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='st_nlink' type-id='80f0b9df' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='st_mode' type-id='e1c52942' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='224'>
	<var-decl name='st_uid' type-id='cc5fcceb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='st_gid' type-id='d94ec6d9' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='__pad0' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='st_rdev' type-id='35ed8932' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='st_size' type-id='79989e9c' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='st_blksize' type-id='d3f10a7f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='st_blocks' type-id='4e711bf1' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='576'>
	<var-decl name='st_atim' type-id='a9c79a1f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='704'>
	<var-decl name='st_mtim' type-id='a9c79a1f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='832'>
	<var-decl name='st_ctim' type-id='a9c79a1f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='960'>
	<var-decl name='__glibc_reserved' type-id='083f8d58' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='__dev_t' type-id='7359adad' id='35ed8932'/>
	<typedef-decl name='__ino64_t' type-id='7359adad' id='71288a47'/>
	<typedef-decl name='__mode_t' type-id='f0981eeb' id='e1c52942'/>
	<typedef-decl name='__nlink_t' type-id='7359adad' id='80f0b9df'/>
	<typedef-decl name='__blksize_t' type-id='bd54fe1a' id='d3f10a7f'/>
	<typedef-decl name='__blkcnt64_t' type-id='bd54fe1a' id='4e711bf1'/>
	<typedef-decl name='__syscall_slong_t' type-id='bd54fe1a' id='03085adc'/>
	<class-decl name='timespec' size-in-bits='128' is-struct='yes' visibility='default' id='a9c79a1f'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='tv_sec' type-id='65eda9c0' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='tv_nsec' type-id='03085adc' visibility='default'/>
	</data-member>
	</class-decl>
	<qualified-type-def type-id='8b092c69' const='yes' id='1a21babe'/>
	<pointer-type-def type-id='de5d1d8f' size-in-bits='64' id='9eadf5e0'/>
	<pointer-type-def type-id='084a08a3' size-in-bits='64' id='b9ea57b8'/>
	<pointer-type-def type-id='b7c58eaa' size-in-bits='64' id='e7c00489'/>
	<pointer-type-def type-id='0bbec9cd' size-in-bits='64' id='62f7a03d'/>
	<var-decl name='libzfs_config_ops' type-id='b1e62775' mangled-name='libzfs_config_ops' visibility='default' elf-symbol-id='libzfs_config_ops'/>
	<function-decl name='zpool_read_label' mangled-name='zpool_read_label' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_read_label'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='857bb57e'/>
	<parameter type-id='7292109c'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pwrite64' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='724e4de6'/>
	<return type-id='79a0948f'/>
	</function-decl>
	<function-decl name='__pread64_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='724e4de6'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='79a0948f'/>
	</function-decl>
	<function-decl name='fstat64' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='62f7a03d'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zcmd_write_conf_nvlist' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='e4ec4540'/>
	<parameter type-id='5ce45b60'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zpool_clear_label' mangled-name='zpool_clear_label' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_clear_label'>
	<parameter type-id='95e97e5e' name='fd'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_in_use' mangled-name='zpool_in_use' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_in_use'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='95e97e5e' name='fd'/>
	<parameter type-id='b9ea57b8' name='state'/>
	<parameter type-id='9b23c9ad' name='namestr'/>
	<parameter type-id='37e3bd22' name='inuse'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-type size-in-bits='64' id='baa42fef'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='37e3bd22'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='29f040d2'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<return type-id='5ce45b60'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_iter.c' language='LANG_C99'>
	<pointer-type-def type-id='b351119f' size-in-bits='64' id='716943c7'/>
	<function-decl name='avl_first' mangled-name='avl_first' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_first'>
	<parameter type-id='a3681dea'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='avl_walk' mangled-name='avl_walk' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_walk'>
	<parameter type-id='716943c7'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='make_dataset_handle_zc' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='e4ec4540'/>
	<return type-id='9200a744'/>
	</function-decl>
	<function-decl name='make_dataset_simple_handle_zc' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9200a744'/>
	<parameter type-id='e4ec4540'/>
	<return type-id='9200a744'/>
	</function-decl>
	<function-decl name='make_bookmark_handle' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9200a744'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='5ce45b60'/>
	<return type-id='9200a744'/>
	</function-decl>
	<function-decl name='zfs_iter_filesystems' mangled-name='zfs_iter_filesystems' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_filesystems'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='d8e49ab9' name='func'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_iter_snapshots' mangled-name='zfs_iter_snapshots' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_snapshots'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='c19b74c3' name='simple'/>
	<parameter type-id='d8e49ab9' name='func'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<parameter type-id='9c313c2d' name='min_txg'/>
	<parameter type-id='9c313c2d' name='max_txg'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_iter_bookmarks' mangled-name='zfs_iter_bookmarks' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_bookmarks'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='d8e49ab9' name='func'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_iter_snapshots_sorted' mangled-name='zfs_iter_snapshots_sorted' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_snapshots_sorted'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='d8e49ab9' name='callback'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<parameter type-id='9c313c2d' name='min_txg'/>
	<parameter type-id='9c313c2d' name='max_txg'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_iter_snapshots_sorted_v2' mangled-name='zfs_iter_snapshots_sorted_v2' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_snapshots_sorted_v2'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='95e97e5e' name='flags'/>
	<parameter type-id='d8e49ab9' name='callback'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<parameter type-id='9c313c2d' name='min_txg'/>
	<parameter type-id='9c313c2d' name='max_txg'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_iter_snapspec' mangled-name='zfs_iter_snapspec' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_snapspec'>
	<parameter type-id='9200a744' name='fs_zhp'/>
	<parameter type-id='80f4b756' name='spec_orig'/>
	<parameter type-id='d8e49ab9' name='func'/>
	<parameter type-id='eaa32e2f' name='arg'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_iter_snapspec_v2' mangled-name='zfs_iter_snapspec_v2' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_snapspec_v2'>
	<parameter type-id='9200a744' name='fs_zhp'/>
	<parameter type-id='95e97e5e' name='flags'/>
	<parameter type-id='80f4b756' name='spec_orig'/>
	<parameter type-id='d8e49ab9' name='func'/>
	<parameter type-id='eaa32e2f' name='arg'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_iter_children' mangled-name='zfs_iter_children' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_children'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='d8e49ab9' name='func'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_iter_dependents' mangled-name='zfs_iter_dependents' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_iter_dependents'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='c19b74c3' name='allowrecursion'/>
	<parameter type-id='d8e49ab9' name='func'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_mount.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='6028cbfe' size-in-bits='256' id='b39b9aa7'>
	<subrange length='4' type-id='7359adad' id='16fe7105'/>
	</array-type-def>
	<class-decl name='__dirstream' is-struct='yes' visibility='default' is-declaration-only='yes' id='20cd73f2'/>
	<class-decl name='tpool' size-in-bits='2496' is-struct='yes' visibility='default' id='88d1b7f9'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='tp_forw' type-id='9cf59a50' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='tp_back' type-id='9cf59a50' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='tp_mutex' type-id='7a6844eb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='tp_busycv' type-id='62fab762' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='832'>
	<var-decl name='tp_workcv' type-id='62fab762' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1216'>
	<var-decl name='tp_waitcv' type-id='62fab762' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1600'>
	<var-decl name='tp_active' type-id='ad33e5e7' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1664'>
	<var-decl name='tp_head' type-id='f32b30e4' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1728'>
	<var-decl name='tp_tail' type-id='f32b30e4' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1792'>
	<var-decl name='tp_attr' type-id='7d8569fd' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2240'>
	<var-decl name='tp_flags' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2272'>
	<var-decl name='tp_linger' type-id='3502e3ff' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2304'>
	<var-decl name='tp_njobs' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2336'>
	<var-decl name='tp_minimum' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2368'>
	<var-decl name='tp_maximum' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2400'>
	<var-decl name='tp_current' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2432'>
	<var-decl name='tp_idle' type-id='95e97e5e' visibility='default'/>
	</data-member>
	</class-decl>
	<array-type-def dimensions='1' type-id='95e97e5e' size-in-bits='64' id='e4266c7e'>
	<subrange length='2' type-id='7359adad' id='52efc4ef'/>
	</array-type-def>
	<class-decl name='get_all_cb' size-in-bits='192' is-struct='yes' visibility='default' id='803dac95'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='cb_handles' type-id='4507922a' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='cb_alloc' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='cb_used' type-id='b59d7dce' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='get_all_cb_t' type-id='803dac95' id='9b293607'/>
	<typedef-decl name='tpool_t' type-id='88d1b7f9' id='b1bbf10d'/>
	<typedef-decl name='DIR' type-id='20cd73f2' id='54a5d683'/>
	<typedef-decl name='mode_t' type-id='e1c52942' id='d50d396c'/>
	<typedef-decl name='__compar_fn_t' type-id='585e1de9' id='aba7edd8'/>
	<class-decl name='dirent64' size-in-bits='2240' is-struct='yes' visibility='default' id='5725d813'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='d_ino' type-id='71288a47' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='d_off' type-id='724e4de6' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='d_reclen' type-id='8efea9e5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='144'>
	<var-decl name='d_type' type-id='002ac4a6' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='152'>
	<var-decl name='d_name' type-id='d1617432' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='statfs64' size-in-bits='960' is-struct='yes' visibility='default' id='a2a6be1a'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='f_type' type-id='6028cbfe' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='f_bsize' type-id='6028cbfe' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='f_blocks' type-id='95fe1a02' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='f_bfree' type-id='95fe1a02' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='f_bavail' type-id='95fe1a02' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='f_files' type-id='0c3a4dde' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='f_ffree' type-id='0c3a4dde' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='f_fsid' type-id='0f35d263' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='f_namelen' type-id='6028cbfe' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='576'>
	<var-decl name='f_frsize' type-id='6028cbfe' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='640'>
	<var-decl name='f_flags' type-id='6028cbfe' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='704'>
	<var-decl name='f_spare' type-id='b39b9aa7' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='stat' size-in-bits='1152' is-struct='yes' visibility='default' id='aafc373f'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='st_dev' type-id='35ed8932' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='st_ino' type-id='e43e523d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='st_nlink' type-id='80f0b9df' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='st_mode' type-id='e1c52942' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='224'>
	<var-decl name='st_uid' type-id='cc5fcceb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='st_gid' type-id='d94ec6d9' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='__pad0' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='st_rdev' type-id='35ed8932' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='st_size' type-id='79989e9c' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='st_blksize' type-id='d3f10a7f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='st_blocks' type-id='dbc43803' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='576'>
	<var-decl name='st_atim' type-id='a9c79a1f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='704'>
	<var-decl name='st_mtim' type-id='a9c79a1f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='832'>
	<var-decl name='st_ctim' type-id='a9c79a1f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='960'>
	<var-decl name='__glibc_reserved' type-id='083f8d58' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='__ino_t' type-id='7359adad' id='e43e523d'/>
	<class-decl name='__fsid_t' size-in-bits='64' is-struct='yes' naming-typedef-id='0f35d263' visibility='default' id='ea35c84a'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='__val' type-id='e4266c7e' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='__fsid_t' type-id='ea35c84a' id='0f35d263'/>
	<typedef-decl name='__blkcnt_t' type-id='bd54fe1a' id='dbc43803'/>
	<typedef-decl name='__fsblkcnt64_t' type-id='7359adad' id='95fe1a02'/>
	<typedef-decl name='__fsfilcnt64_t' type-id='7359adad' id='0c3a4dde'/>
	<typedef-decl name='__fsword_t' type-id='bd54fe1a' id='6028cbfe'/>
	<pointer-type-def type-id='54a5d683' size-in-bits='64' id='f09217ba'/>
	<pointer-type-def type-id='5725d813' size-in-bits='64' id='07b96073'/>
	<pointer-type-def type-id='9b293607' size-in-bits='64' id='77bf1784'/>
	<pointer-type-def type-id='7d8569fd' size-in-bits='64' id='7347a39e'/>
	<pointer-type-def type-id='aafc373f' size-in-bits='64' id='4330df87'/>
	<qualified-type-def type-id='4330df87' restrict='yes' id='73665405'/>
	<pointer-type-def type-id='a2a6be1a' size-in-bits='64' id='7fd094c8'/>
	<pointer-type-def type-id='b1bbf10d' size-in-bits='64' id='9cf59a50'/>
	<pointer-type-def type-id='c5c76c9c' size-in-bits='64' id='b7f9d8e6'/>
	<pointer-type-def type-id='9200a744' size-in-bits='64' id='4507922a'/>
	<class-decl name='__dirstream' is-struct='yes' visibility='default' is-declaration-only='yes' id='20cd73f2'/>
	<function-decl name='zpool_disable_datasets_os' mangled-name='zpool_disable_datasets_os' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_disable_datasets_os'>
	<parameter type-id='4c81de99'/>
	<parameter type-id='c19b74c3'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zpool_disable_volume_os' mangled-name='zpool_disable_volume_os' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_disable_volume_os'>
	<parameter type-id='80f4b756'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='tpool_create' mangled-name='tpool_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_create'>
	<parameter type-id='3502e3ff'/>
	<parameter type-id='3502e3ff'/>
	<parameter type-id='3502e3ff'/>
	<parameter type-id='7347a39e'/>
	<return type-id='9cf59a50'/>
	</function-decl>
	<function-decl name='tpool_dispatch' mangled-name='tpool_dispatch' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_dispatch'>
	<parameter type-id='9cf59a50'/>
	<parameter type-id='b7f9d8e6'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='tpool_destroy' mangled-name='tpool_destroy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_destroy'>
	<parameter type-id='9cf59a50'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='tpool_wait' mangled-name='tpool_wait' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='tpool_wait'>
	<parameter type-id='9cf59a50'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='mkdirp' mangled-name='mkdirp' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='mkdirp'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='d50d396c'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='sa_errorstr' mangled-name='sa_errorstr' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='sa_errorstr'>
	<parameter type-id='95e97e5e'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='sa_enable_share' mangled-name='sa_enable_share' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='sa_enable_share'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='9155d4b5'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='sa_disable_share' mangled-name='sa_disable_share' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='sa_disable_share'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='9155d4b5'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='sa_is_shared' mangled-name='sa_is_shared' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='sa_is_shared'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='9155d4b5'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='sa_truncate_shares' mangled-name='sa_truncate_shares' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='sa_truncate_shares'>
	<parameter type-id='9155d4b5'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fdopendir' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<return type-id='f09217ba'/>
	</function-decl>
	<function-decl name='closedir' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='f09217ba'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='readdir64' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='f09217ba'/>
	<return type-id='07b96073'/>
	</function-decl>
	<function-decl name='qsort' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='aba7edd8'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='rmdir' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='__openat_too_many_args' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='__openat_missing_mode' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='statfs64' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='7fd094c8'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_realloc' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='changelist_unshare' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='0d41d328'/>
	<parameter type-id='4567bbc9'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='do_mount' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9200a744'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='do_unmount' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9200a744'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_mount_at' mangled-name='zfs_mount_at' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_mount_at'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='options'/>
	<parameter type-id='95e97e5e' name='flags'/>
	<parameter type-id='80f4b756' name='mountpoint'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_unmountall' mangled-name='zfs_unmountall' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_unmountall'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='95e97e5e' name='flags'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_truncate_shares' mangled-name='zfs_truncate_shares' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_truncate_shares'>
	<parameter type-id='4567bbc9' name='proto'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_unshareall' mangled-name='zfs_unshareall' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_unshareall'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='4567bbc9' name='proto'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='libzfs_add_handle' mangled-name='libzfs_add_handle' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_add_handle'>
	<parameter type-id='77bf1784' name='cbp'/>
	<parameter type-id='9200a744' name='zhp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_foreach_mountpoint' mangled-name='zfs_foreach_mountpoint' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_foreach_mountpoint'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='4507922a' name='handles'/>
	<parameter type-id='b59d7dce' name='num_handles'/>
	<parameter type-id='d8e49ab9' name='func'/>
	<parameter type-id='eaa32e2f' name='data'/>
	- <parameter type-id='c19b74c3' name='parallel'/>
	+ <parameter type-id='3502e3ff' name='nthr'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zpool_enable_datasets' mangled-name='zpool_enable_datasets' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_enable_datasets'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='mntopts'/>
	<parameter type-id='95e97e5e' name='flags'/>
	+ <parameter type-id='3502e3ff' name='nthr'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_disable_datasets' mangled-name='zpool_disable_datasets' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_disable_datasets'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='c19b74c3' name='force'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_pool.c' language='LANG_C99'>
	<type-decl name='long long unsigned int' size-in-bits='64' id='3a47d82b'/>
	<class-decl name='splitflags' size-in-bits='64' is-struct='yes' visibility='default' id='dc01bf52'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='dryrun' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1'>
	<var-decl name='import' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='name_flags' type-id='95e97e5e' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='splitflags_t' type-id='dc01bf52' id='325c1e34'/>
	<class-decl name='trimflags' size-in-bits='192' is-struct='yes' visibility='default' id='8ef58008'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='fullpool' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='secure' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='wait' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='rate' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='trimflags_t' type-id='8ef58008' id='a093cbb8'/>
	<enum-decl name='zpool_status_t' naming-typedef-id='d3dd6294' id='5e770b40'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZPOOL_STATUS_CORRUPT_CACHE' value='0'/>
	<enumerator name='ZPOOL_STATUS_MISSING_DEV_R' value='1'/>
	<enumerator name='ZPOOL_STATUS_MISSING_DEV_NR' value='2'/>
	<enumerator name='ZPOOL_STATUS_CORRUPT_LABEL_R' value='3'/>
	<enumerator name='ZPOOL_STATUS_CORRUPT_LABEL_NR' value='4'/>
	<enumerator name='ZPOOL_STATUS_BAD_GUID_SUM' value='5'/>
	<enumerator name='ZPOOL_STATUS_CORRUPT_POOL' value='6'/>
	<enumerator name='ZPOOL_STATUS_CORRUPT_DATA' value='7'/>
	<enumerator name='ZPOOL_STATUS_FAILING_DEV' value='8'/>
	<enumerator name='ZPOOL_STATUS_VERSION_NEWER' value='9'/>
	<enumerator name='ZPOOL_STATUS_HOSTID_MISMATCH' value='10'/>
	<enumerator name='ZPOOL_STATUS_HOSTID_ACTIVE' value='11'/>
	<enumerator name='ZPOOL_STATUS_HOSTID_REQUIRED' value='12'/>
	<enumerator name='ZPOOL_STATUS_IO_FAILURE_WAIT' value='13'/>
	<enumerator name='ZPOOL_STATUS_IO_FAILURE_CONTINUE' value='14'/>
	<enumerator name='ZPOOL_STATUS_IO_FAILURE_MMP' value='15'/>
	<enumerator name='ZPOOL_STATUS_BAD_LOG' value='16'/>
	<enumerator name='ZPOOL_STATUS_ERRATA' value='17'/>
	<enumerator name='ZPOOL_STATUS_UNSUP_FEAT_READ' value='18'/>
	<enumerator name='ZPOOL_STATUS_UNSUP_FEAT_WRITE' value='19'/>
	<enumerator name='ZPOOL_STATUS_FAULTED_DEV_R' value='20'/>
	<enumerator name='ZPOOL_STATUS_FAULTED_DEV_NR' value='21'/>
	<enumerator name='ZPOOL_STATUS_VERSION_OLDER' value='22'/>
	<enumerator name='ZPOOL_STATUS_FEAT_DISABLED' value='23'/>
	<enumerator name='ZPOOL_STATUS_RESILVERING' value='24'/>
	<enumerator name='ZPOOL_STATUS_OFFLINE_DEV' value='25'/>
	<enumerator name='ZPOOL_STATUS_REMOVED_DEV' value='26'/>
	<enumerator name='ZPOOL_STATUS_REBUILDING' value='27'/>
	<enumerator name='ZPOOL_STATUS_REBUILD_SCRUB' value='28'/>
	<enumerator name='ZPOOL_STATUS_NON_NATIVE_ASHIFT' value='29'/>
	<enumerator name='ZPOOL_STATUS_COMPATIBILITY_ERR' value='30'/>
	<enumerator name='ZPOOL_STATUS_INCOMPATIBLE_FEAT' value='31'/>
	<enumerator name='ZPOOL_STATUS_OK' value='32'/>
	</enum-decl>
	<typedef-decl name='zpool_status_t' type-id='5e770b40' id='d3dd6294'/>
	<enum-decl name='zpool_compat_status_t' naming-typedef-id='901b78d1' id='20676925'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZPOOL_COMPATIBILITY_OK' value='0'/>
	<enumerator name='ZPOOL_COMPATIBILITY_WARNTOKEN' value='1'/>
	<enumerator name='ZPOOL_COMPATIBILITY_BADTOKEN' value='2'/>
	<enumerator name='ZPOOL_COMPATIBILITY_BADFILE' value='3'/>
	<enumerator name='ZPOOL_COMPATIBILITY_NOFILES' value='4'/>
	</enum-decl>
	<typedef-decl name='zpool_compat_status_t' type-id='20676925' id='901b78d1'/>
	<enum-decl name='vdev_prop_t' naming-typedef-id='5aa5c90c' id='1573bec8'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='VDEV_PROP_INVAL' value='-1'/>
	<enumerator name='VDEV_PROP_USERPROP' value='-1'/>
	<enumerator name='VDEV_PROP_NAME' value='0'/>
	<enumerator name='VDEV_PROP_CAPACITY' value='1'/>
	<enumerator name='VDEV_PROP_STATE' value='2'/>
	<enumerator name='VDEV_PROP_GUID' value='3'/>
	<enumerator name='VDEV_PROP_ASIZE' value='4'/>
	<enumerator name='VDEV_PROP_PSIZE' value='5'/>
	<enumerator name='VDEV_PROP_ASHIFT' value='6'/>
	<enumerator name='VDEV_PROP_SIZE' value='7'/>
	<enumerator name='VDEV_PROP_FREE' value='8'/>
	<enumerator name='VDEV_PROP_ALLOCATED' value='9'/>
	<enumerator name='VDEV_PROP_COMMENT' value='10'/>
	<enumerator name='VDEV_PROP_EXPANDSZ' value='11'/>
	<enumerator name='VDEV_PROP_FRAGMENTATION' value='12'/>
	<enumerator name='VDEV_PROP_BOOTSIZE' value='13'/>
	<enumerator name='VDEV_PROP_PARITY' value='14'/>
	<enumerator name='VDEV_PROP_PATH' value='15'/>
	<enumerator name='VDEV_PROP_DEVID' value='16'/>
	<enumerator name='VDEV_PROP_PHYS_PATH' value='17'/>
	<enumerator name='VDEV_PROP_ENC_PATH' value='18'/>
	<enumerator name='VDEV_PROP_FRU' value='19'/>
	<enumerator name='VDEV_PROP_PARENT' value='20'/>
	<enumerator name='VDEV_PROP_CHILDREN' value='21'/>
	<enumerator name='VDEV_PROP_NUMCHILDREN' value='22'/>
	<enumerator name='VDEV_PROP_READ_ERRORS' value='23'/>
	<enumerator name='VDEV_PROP_WRITE_ERRORS' value='24'/>
	<enumerator name='VDEV_PROP_CHECKSUM_ERRORS' value='25'/>
	<enumerator name='VDEV_PROP_INITIALIZE_ERRORS' value='26'/>
	<enumerator name='VDEV_PROP_OPS_NULL' value='27'/>
	<enumerator name='VDEV_PROP_OPS_READ' value='28'/>
	<enumerator name='VDEV_PROP_OPS_WRITE' value='29'/>
	<enumerator name='VDEV_PROP_OPS_FREE' value='30'/>
	<enumerator name='VDEV_PROP_OPS_CLAIM' value='31'/>
	<enumerator name='VDEV_PROP_OPS_TRIM' value='32'/>
	<enumerator name='VDEV_PROP_BYTES_NULL' value='33'/>
	<enumerator name='VDEV_PROP_BYTES_READ' value='34'/>
	<enumerator name='VDEV_PROP_BYTES_WRITE' value='35'/>
	<enumerator name='VDEV_PROP_BYTES_FREE' value='36'/>
	<enumerator name='VDEV_PROP_BYTES_CLAIM' value='37'/>
	<enumerator name='VDEV_PROP_BYTES_TRIM' value='38'/>
	<enumerator name='VDEV_PROP_REMOVING' value='39'/>
	<enumerator name='VDEV_PROP_ALLOCATING' value='40'/>
	<enumerator name='VDEV_PROP_FAILFAST' value='41'/>
	<enumerator name='VDEV_PROP_CHECKSUM_N' value='42'/>
	<enumerator name='VDEV_PROP_CHECKSUM_T' value='43'/>
	<enumerator name='VDEV_PROP_IO_N' value='44'/>
	<enumerator name='VDEV_PROP_IO_T' value='45'/>
	<enumerator name='VDEV_PROP_RAIDZ_EXPANDING' value='46'/>
	<enumerator name='VDEV_PROP_SLOW_IO_N' value='47'/>
	<enumerator name='VDEV_PROP_SLOW_IO_T' value='48'/>
	<enumerator name='VDEV_NUM_PROPS' value='49'/>
	</enum-decl>
	<typedef-decl name='vdev_prop_t' type-id='1573bec8' id='5aa5c90c'/>
	<class-decl name='zpool_load_policy' size-in-bits='256' is-struct='yes' visibility='default' id='2f65b36f'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='zlp_rewind' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='zlp_maxmeta' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='zlp_maxdata' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='zlp_txg' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zpool_load_policy_t' type-id='2f65b36f' id='d11b7617'/>
	<enum-decl name='vdev_state' id='21566197'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='VDEV_STATE_UNKNOWN' value='0'/>
	<enumerator name='VDEV_STATE_CLOSED' value='1'/>
	<enumerator name='VDEV_STATE_OFFLINE' value='2'/>
	<enumerator name='VDEV_STATE_REMOVED' value='3'/>
	<enumerator name='VDEV_STATE_CANT_OPEN' value='4'/>
	<enumerator name='VDEV_STATE_FAULTED' value='5'/>
	<enumerator name='VDEV_STATE_DEGRADED' value='6'/>
	<enumerator name='VDEV_STATE_HEALTHY' value='7'/>
	</enum-decl>
	<typedef-decl name='vdev_state_t' type-id='21566197' id='35acf840'/>
	<enum-decl name='vdev_aux' id='7f5bcca4'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='VDEV_AUX_NONE' value='0'/>
	<enumerator name='VDEV_AUX_OPEN_FAILED' value='1'/>
	<enumerator name='VDEV_AUX_CORRUPT_DATA' value='2'/>
	<enumerator name='VDEV_AUX_NO_REPLICAS' value='3'/>
	<enumerator name='VDEV_AUX_BAD_GUID_SUM' value='4'/>
	<enumerator name='VDEV_AUX_TOO_SMALL' value='5'/>
	<enumerator name='VDEV_AUX_BAD_LABEL' value='6'/>
	<enumerator name='VDEV_AUX_VERSION_NEWER' value='7'/>
	<enumerator name='VDEV_AUX_VERSION_OLDER' value='8'/>
	<enumerator name='VDEV_AUX_UNSUP_FEAT' value='9'/>
	<enumerator name='VDEV_AUX_SPARED' value='10'/>
	<enumerator name='VDEV_AUX_ERR_EXCEEDED' value='11'/>
	<enumerator name='VDEV_AUX_IO_FAILURE' value='12'/>
	<enumerator name='VDEV_AUX_BAD_LOG' value='13'/>
	<enumerator name='VDEV_AUX_EXTERNAL' value='14'/>
	<enumerator name='VDEV_AUX_SPLIT_POOL' value='15'/>
	<enumerator name='VDEV_AUX_BAD_ASHIFT' value='16'/>
	<enumerator name='VDEV_AUX_EXTERNAL_PERSIST' value='17'/>
	<enumerator name='VDEV_AUX_ACTIVE' value='18'/>
	<enumerator name='VDEV_AUX_CHILDREN_OFFLINE' value='19'/>
	<enumerator name='VDEV_AUX_ASHIFT_TOO_BIG' value='20'/>
	</enum-decl>
	<typedef-decl name='vdev_aux_t' type-id='7f5bcca4' id='9d774e0b'/>
	<enum-decl name='pool_scan_func' id='1b092565'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='POOL_SCAN_NONE' value='0'/>
	<enumerator name='POOL_SCAN_SCRUB' value='1'/>
	<enumerator name='POOL_SCAN_RESILVER' value='2'/>
	<enumerator name='POOL_SCAN_ERRORSCRUB' value='3'/>
	<enumerator name='POOL_SCAN_FUNCS' value='4'/>
	</enum-decl>
	<typedef-decl name='pool_scan_func_t' type-id='1b092565' id='7313fbe2'/>
	<enum-decl name='pool_scrub_cmd' id='a1474cbd'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='POOL_SCRUB_NORMAL' value='0'/>
	<enumerator name='POOL_SCRUB_PAUSE' value='1'/>
	<enumerator name='POOL_SCRUB_FLAGS_END' value='2'/>
	</enum-decl>
	<typedef-decl name='pool_scrub_cmd_t' type-id='a1474cbd' id='b51cf3c2'/>
	<enum-decl name='zpool_errata' id='d9abbf54'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZPOOL_ERRATA_NONE' value='0'/>
	<enumerator name='ZPOOL_ERRATA_ZOL_2094_SCRUB' value='1'/>
	<enumerator name='ZPOOL_ERRATA_ZOL_2094_ASYNC_DESTROY' value='2'/>
	<enumerator name='ZPOOL_ERRATA_ZOL_6845_ENCRYPTION' value='3'/>
	<enumerator name='ZPOOL_ERRATA_ZOL_8308_ENCRYPTION' value='4'/>
	</enum-decl>
	<typedef-decl name='zpool_errata_t' type-id='d9abbf54' id='688c495b'/>
	<enum-decl name='pool_initialize_func' id='5c246ad4'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='POOL_INITIALIZE_START' value='0'/>
	<enumerator name='POOL_INITIALIZE_CANCEL' value='1'/>
	<enumerator name='POOL_INITIALIZE_SUSPEND' value='2'/>
	<enumerator name='POOL_INITIALIZE_UNINIT' value='3'/>
	<enumerator name='POOL_INITIALIZE_FUNCS' value='4'/>
	</enum-decl>
	<typedef-decl name='pool_initialize_func_t' type-id='5c246ad4' id='7063e1ab'/>
	<enum-decl name='pool_trim_func' id='54ed608a'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='POOL_TRIM_START' value='0'/>
	<enumerator name='POOL_TRIM_CANCEL' value='1'/>
	<enumerator name='POOL_TRIM_SUSPEND' value='2'/>
	<enumerator name='POOL_TRIM_FUNCS' value='3'/>
	</enum-decl>
	<typedef-decl name='pool_trim_func_t' type-id='54ed608a' id='b1146b8d'/>
	<enum-decl name='zfs_ioc' id='12033f13'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZFS_IOC_FIRST' value='23040'/>
	<enumerator name='ZFS_IOC' value='23040'/>
	<enumerator name='ZFS_IOC_POOL_CREATE' value='23040'/>
	<enumerator name='ZFS_IOC_POOL_DESTROY' value='23041'/>
	<enumerator name='ZFS_IOC_POOL_IMPORT' value='23042'/>
	<enumerator name='ZFS_IOC_POOL_EXPORT' value='23043'/>
	<enumerator name='ZFS_IOC_POOL_CONFIGS' value='23044'/>
	<enumerator name='ZFS_IOC_POOL_STATS' value='23045'/>
	<enumerator name='ZFS_IOC_POOL_TRYIMPORT' value='23046'/>
	<enumerator name='ZFS_IOC_POOL_SCAN' value='23047'/>
	<enumerator name='ZFS_IOC_POOL_FREEZE' value='23048'/>
	<enumerator name='ZFS_IOC_POOL_UPGRADE' value='23049'/>
	<enumerator name='ZFS_IOC_POOL_GET_HISTORY' value='23050'/>
	<enumerator name='ZFS_IOC_VDEV_ADD' value='23051'/>
	<enumerator name='ZFS_IOC_VDEV_REMOVE' value='23052'/>
	<enumerator name='ZFS_IOC_VDEV_SET_STATE' value='23053'/>
	<enumerator name='ZFS_IOC_VDEV_ATTACH' value='23054'/>
	<enumerator name='ZFS_IOC_VDEV_DETACH' value='23055'/>
	<enumerator name='ZFS_IOC_VDEV_SETPATH' value='23056'/>
	<enumerator name='ZFS_IOC_VDEV_SETFRU' value='23057'/>
	<enumerator name='ZFS_IOC_OBJSET_STATS' value='23058'/>
	<enumerator name='ZFS_IOC_OBJSET_ZPLPROPS' value='23059'/>
	<enumerator name='ZFS_IOC_DATASET_LIST_NEXT' value='23060'/>
	<enumerator name='ZFS_IOC_SNAPSHOT_LIST_NEXT' value='23061'/>
	<enumerator name='ZFS_IOC_SET_PROP' value='23062'/>
	<enumerator name='ZFS_IOC_CREATE' value='23063'/>
	<enumerator name='ZFS_IOC_DESTROY' value='23064'/>
	<enumerator name='ZFS_IOC_ROLLBACK' value='23065'/>
	<enumerator name='ZFS_IOC_RENAME' value='23066'/>
	<enumerator name='ZFS_IOC_RECV' value='23067'/>
	<enumerator name='ZFS_IOC_SEND' value='23068'/>
	<enumerator name='ZFS_IOC_INJECT_FAULT' value='23069'/>
	<enumerator name='ZFS_IOC_CLEAR_FAULT' value='23070'/>
	<enumerator name='ZFS_IOC_INJECT_LIST_NEXT' value='23071'/>
	<enumerator name='ZFS_IOC_ERROR_LOG' value='23072'/>
	<enumerator name='ZFS_IOC_CLEAR' value='23073'/>
	<enumerator name='ZFS_IOC_PROMOTE' value='23074'/>
	<enumerator name='ZFS_IOC_SNAPSHOT' value='23075'/>
	<enumerator name='ZFS_IOC_DSOBJ_TO_DSNAME' value='23076'/>
	<enumerator name='ZFS_IOC_OBJ_TO_PATH' value='23077'/>
	<enumerator name='ZFS_IOC_POOL_SET_PROPS' value='23078'/>
	<enumerator name='ZFS_IOC_POOL_GET_PROPS' value='23079'/>
	<enumerator name='ZFS_IOC_SET_FSACL' value='23080'/>
	<enumerator name='ZFS_IOC_GET_FSACL' value='23081'/>
	<enumerator name='ZFS_IOC_SHARE' value='23082'/>
	<enumerator name='ZFS_IOC_INHERIT_PROP' value='23083'/>
	<enumerator name='ZFS_IOC_SMB_ACL' value='23084'/>
	<enumerator name='ZFS_IOC_USERSPACE_ONE' value='23085'/>
	<enumerator name='ZFS_IOC_USERSPACE_MANY' value='23086'/>
	<enumerator name='ZFS_IOC_USERSPACE_UPGRADE' value='23087'/>
	<enumerator name='ZFS_IOC_HOLD' value='23088'/>
	<enumerator name='ZFS_IOC_RELEASE' value='23089'/>
	<enumerator name='ZFS_IOC_GET_HOLDS' value='23090'/>
	<enumerator name='ZFS_IOC_OBJSET_RECVD_PROPS' value='23091'/>
	<enumerator name='ZFS_IOC_VDEV_SPLIT' value='23092'/>
	<enumerator name='ZFS_IOC_NEXT_OBJ' value='23093'/>
	<enumerator name='ZFS_IOC_DIFF' value='23094'/>
	<enumerator name='ZFS_IOC_TMP_SNAPSHOT' value='23095'/>
	<enumerator name='ZFS_IOC_OBJ_TO_STATS' value='23096'/>
	<enumerator name='ZFS_IOC_SPACE_WRITTEN' value='23097'/>
	<enumerator name='ZFS_IOC_SPACE_SNAPS' value='23098'/>
	<enumerator name='ZFS_IOC_DESTROY_SNAPS' value='23099'/>
	<enumerator name='ZFS_IOC_POOL_REGUID' value='23100'/>
	<enumerator name='ZFS_IOC_POOL_REOPEN' value='23101'/>
	<enumerator name='ZFS_IOC_SEND_PROGRESS' value='23102'/>
	<enumerator name='ZFS_IOC_LOG_HISTORY' value='23103'/>
	<enumerator name='ZFS_IOC_SEND_NEW' value='23104'/>
	<enumerator name='ZFS_IOC_SEND_SPACE' value='23105'/>
	<enumerator name='ZFS_IOC_CLONE' value='23106'/>
	<enumerator name='ZFS_IOC_BOOKMARK' value='23107'/>
	<enumerator name='ZFS_IOC_GET_BOOKMARKS' value='23108'/>
	<enumerator name='ZFS_IOC_DESTROY_BOOKMARKS' value='23109'/>
	<enumerator name='ZFS_IOC_RECV_NEW' value='23110'/>
	<enumerator name='ZFS_IOC_POOL_SYNC' value='23111'/>
	<enumerator name='ZFS_IOC_CHANNEL_PROGRAM' value='23112'/>
	<enumerator name='ZFS_IOC_LOAD_KEY' value='23113'/>
	<enumerator name='ZFS_IOC_UNLOAD_KEY' value='23114'/>
	<enumerator name='ZFS_IOC_CHANGE_KEY' value='23115'/>
	<enumerator name='ZFS_IOC_REMAP' value='23116'/>
	<enumerator name='ZFS_IOC_POOL_CHECKPOINT' value='23117'/>
	<enumerator name='ZFS_IOC_POOL_DISCARD_CHECKPOINT' value='23118'/>
	<enumerator name='ZFS_IOC_POOL_INITIALIZE' value='23119'/>
	<enumerator name='ZFS_IOC_POOL_TRIM' value='23120'/>
	<enumerator name='ZFS_IOC_REDACT' value='23121'/>
	<enumerator name='ZFS_IOC_GET_BOOKMARK_PROPS' value='23122'/>
	<enumerator name='ZFS_IOC_WAIT' value='23123'/>
	<enumerator name='ZFS_IOC_WAIT_FS' value='23124'/>
	<enumerator name='ZFS_IOC_VDEV_GET_PROPS' value='23125'/>
	<enumerator name='ZFS_IOC_VDEV_SET_PROPS' value='23126'/>
	<enumerator name='ZFS_IOC_POOL_SCRUB' value='23127'/>
	<enumerator name='ZFS_IOC_PLATFORM' value='23168'/>
	<enumerator name='ZFS_IOC_EVENTS_NEXT' value='23169'/>
	<enumerator name='ZFS_IOC_EVENTS_CLEAR' value='23170'/>
	<enumerator name='ZFS_IOC_EVENTS_SEEK' value='23171'/>
	<enumerator name='ZFS_IOC_NEXTBOOT' value='23172'/>
	<enumerator name='ZFS_IOC_JAIL' value='23173'/>
	<enumerator name='ZFS_IOC_USERNS_ATTACH' value='23173'/>
	<enumerator name='ZFS_IOC_UNJAIL' value='23174'/>
	<enumerator name='ZFS_IOC_USERNS_DETACH' value='23174'/>
	<enumerator name='ZFS_IOC_SET_BOOTENV' value='23175'/>
	<enumerator name='ZFS_IOC_GET_BOOTENV' value='23176'/>
	<enumerator name='ZFS_IOC_LAST' value='23177'/>
	</enum-decl>
	<typedef-decl name='zfs_ioc_t' type-id='12033f13' id='5b35941c'/>
	<enum-decl name='zpool_wait_activity_t' naming-typedef-id='73446457' id='849338e3'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZPOOL_WAIT_CKPT_DISCARD' value='0'/>
	<enumerator name='ZPOOL_WAIT_FREE' value='1'/>
	<enumerator name='ZPOOL_WAIT_INITIALIZE' value='2'/>
	<enumerator name='ZPOOL_WAIT_REPLACE' value='3'/>
	<enumerator name='ZPOOL_WAIT_REMOVE' value='4'/>
	<enumerator name='ZPOOL_WAIT_RESILVER' value='5'/>
	<enumerator name='ZPOOL_WAIT_SCRUB' value='6'/>
	<enumerator name='ZPOOL_WAIT_TRIM' value='7'/>
	<enumerator name='ZPOOL_WAIT_RAIDZ_EXPAND' value='8'/>
	<enumerator name='ZPOOL_WAIT_NUM_ACTIVITIES' value='9'/>
	</enum-decl>
	<typedef-decl name='zpool_wait_activity_t' type-id='849338e3' id='73446457'/>
	<enum-decl name='spa_feature' id='33ecb627'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='SPA_FEATURE_NONE' value='-1'/>
	<enumerator name='SPA_FEATURE_ASYNC_DESTROY' value='0'/>
	<enumerator name='SPA_FEATURE_EMPTY_BPOBJ' value='1'/>
	<enumerator name='SPA_FEATURE_LZ4_COMPRESS' value='2'/>
	<enumerator name='SPA_FEATURE_MULTI_VDEV_CRASH_DUMP' value='3'/>
	<enumerator name='SPA_FEATURE_SPACEMAP_HISTOGRAM' value='4'/>
	<enumerator name='SPA_FEATURE_ENABLED_TXG' value='5'/>
	<enumerator name='SPA_FEATURE_HOLE_BIRTH' value='6'/>
	<enumerator name='SPA_FEATURE_EXTENSIBLE_DATASET' value='7'/>
	<enumerator name='SPA_FEATURE_EMBEDDED_DATA' value='8'/>
	<enumerator name='SPA_FEATURE_BOOKMARKS' value='9'/>
	<enumerator name='SPA_FEATURE_FS_SS_LIMIT' value='10'/>
	<enumerator name='SPA_FEATURE_LARGE_BLOCKS' value='11'/>
	<enumerator name='SPA_FEATURE_LARGE_DNODE' value='12'/>
	<enumerator name='SPA_FEATURE_SHA512' value='13'/>
	<enumerator name='SPA_FEATURE_SKEIN' value='14'/>
	<enumerator name='SPA_FEATURE_EDONR' value='15'/>
	<enumerator name='SPA_FEATURE_USEROBJ_ACCOUNTING' value='16'/>
	<enumerator name='SPA_FEATURE_ENCRYPTION' value='17'/>
	<enumerator name='SPA_FEATURE_PROJECT_QUOTA' value='18'/>
	<enumerator name='SPA_FEATURE_DEVICE_REMOVAL' value='19'/>
	<enumerator name='SPA_FEATURE_OBSOLETE_COUNTS' value='20'/>
	<enumerator name='SPA_FEATURE_POOL_CHECKPOINT' value='21'/>
	<enumerator name='SPA_FEATURE_SPACEMAP_V2' value='22'/>
	<enumerator name='SPA_FEATURE_ALLOCATION_CLASSES' value='23'/>
	<enumerator name='SPA_FEATURE_RESILVER_DEFER' value='24'/>
	<enumerator name='SPA_FEATURE_BOOKMARK_V2' value='25'/>
	<enumerator name='SPA_FEATURE_REDACTION_BOOKMARKS' value='26'/>
	<enumerator name='SPA_FEATURE_REDACTED_DATASETS' value='27'/>
	<enumerator name='SPA_FEATURE_BOOKMARK_WRITTEN' value='28'/>
	<enumerator name='SPA_FEATURE_LOG_SPACEMAP' value='29'/>
	<enumerator name='SPA_FEATURE_LIVELIST' value='30'/>
	<enumerator name='SPA_FEATURE_DEVICE_REBUILD' value='31'/>
	<enumerator name='SPA_FEATURE_ZSTD_COMPRESS' value='32'/>
	<enumerator name='SPA_FEATURE_DRAID' value='33'/>
	<enumerator name='SPA_FEATURE_ZILSAXATTR' value='34'/>
	<enumerator name='SPA_FEATURE_HEAD_ERRLOG' value='35'/>
	<enumerator name='SPA_FEATURE_BLAKE3' value='36'/>
	<enumerator name='SPA_FEATURE_BLOCK_CLONING' value='37'/>
	<enumerator name='SPA_FEATURE_AVZ_V2' value='38'/>
	<enumerator name='SPA_FEATURE_REDACTION_LIST_SPILL' value='39'/>
	<enumerator name='SPA_FEATURE_RAIDZ_EXPANSION' value='40'/>
	<enumerator name='SPA_FEATURES' value='41'/>
	</enum-decl>
	<typedef-decl name='spa_feature_t' type-id='33ecb627' id='d6618c78'/>
	<qualified-type-def type-id='22cce67b' const='yes' id='d2816df0'/>
	<pointer-type-def type-id='d2816df0' size-in-bits='64' id='3bbfee2e'/>
	<qualified-type-def type-id='b96825af' const='yes' id='2b61797f'/>
	<pointer-type-def type-id='2b61797f' size-in-bits='64' id='9f7200cf'/>
	<pointer-type-def type-id='d6618c78' size-in-bits='64' id='a8425263'/>
	<qualified-type-def type-id='62f7a03d' restrict='yes' id='f1cadedf'/>
	<pointer-type-def type-id='a093cbb8' size-in-bits='64' id='b13f38c3'/>
	<pointer-type-def type-id='35acf840' size-in-bits='64' id='17f3480d'/>
	<pointer-type-def type-id='688c495b' size-in-bits='64' id='cec6f2e4'/>
	<pointer-type-def type-id='d11b7617' size-in-bits='64' id='23432aaa'/>
	<function-decl name='zpool_get_handle' mangled-name='zpool_get_handle' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_handle'>
	<parameter type-id='4c81de99'/>
	<return type-id='b0382bb3'/>
	</function-decl>
	<function-decl name='zpool_prop_to_name' mangled-name='zpool_prop_to_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_to_name'>
	<parameter type-id='5d0c23fb'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='vdev_prop_to_name' mangled-name='vdev_prop_to_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_to_name'>
	<parameter type-id='5aa5c90c'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='vdev_prop_user' mangled-name='vdev_prop_user' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_user'>
	<parameter type-id='80f4b756'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zpool_get_status' mangled-name='zpool_get_status' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_status'>
	<parameter type-id='4c81de99'/>
	<parameter type-id='7d3cd834'/>
	<parameter type-id='cec6f2e4'/>
	<return type-id='d3dd6294'/>
	</function-decl>
	<function-decl name='zpool_prop_default_string' mangled-name='zpool_prop_default_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_default_string'>
	<parameter type-id='5d0c23fb'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zpool_prop_default_numeric' mangled-name='zpool_prop_default_numeric' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_default_numeric'>
	<parameter type-id='5d0c23fb'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='libzfs_envvar_is_set' mangled-name='libzfs_envvar_is_set' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_envvar_is_set'>
	<parameter type-id='80f4b756'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='lzc_initialize' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='7063e1ab'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='857bb57e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_trim' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='b1146b8d'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='c19b74c3'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='857bb57e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_sync' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='857bb57e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_reopen' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='c19b74c3'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_pool_checkpoint' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_pool_checkpoint_discard' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_wait' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='73446457'/>
	<parameter type-id='37e3bd22'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_wait_tag' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='73446457'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='37e3bd22'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_set_bootenv' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='22cce67b'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_get_bootenv' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='857bb57e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_get_vdev_prop' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='857bb57e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_set_vdev_prop' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='857bb57e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_scrub' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5b35941c'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='857bb57e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_resolve_shortname' mangled-name='zfs_resolve_shortname' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_resolve_shortname'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='26a90f95'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_strip_partition' mangled-name='zfs_strip_partition' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_strip_partition'>
	<parameter type-id='80f4b756'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='zfs_strip_path' mangled-name='zfs_strip_path' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_strip_path'>
	<parameter type-id='80f4b756'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_strcmp_pathname' mangled-name='zfs_strcmp_pathname' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_strcmp_pathname'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_history_unpack' mangled-name='zpool_history_unpack' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_history_unpack'>
	<parameter type-id='26a90f95'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='5d6479ae'/>
	<parameter type-id='75be733c'/>
	<parameter type-id='4dd26a40'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_basename' mangled-name='zfs_basename' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_basename'>
	<parameter type-id='80f4b756'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zpool_name_to_prop' mangled-name='zpool_name_to_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_name_to_prop'>
	<parameter type-id='80f4b756'/>
	<return type-id='5d0c23fb'/>
	</function-decl>
	<function-decl name='zpool_prop_readonly' mangled-name='zpool_prop_readonly' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_readonly'>
	<parameter type-id='5d0c23fb'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zpool_prop_setonce' mangled-name='zpool_prop_setonce' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_setonce'>
	<parameter type-id='5d0c23fb'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zpool_prop_feature' mangled-name='zpool_prop_feature' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_feature'>
	<parameter type-id='80f4b756'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zpool_prop_index_to_string' mangled-name='zpool_prop_index_to_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_index_to_string'>
	<parameter type-id='5d0c23fb'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='7d3cd834'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='vdev_name_to_prop' mangled-name='vdev_name_to_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_name_to_prop'>
	<parameter type-id='80f4b756'/>
	<return type-id='5aa5c90c'/>
	</function-decl>
	<function-decl name='vdev_prop_default_string' mangled-name='vdev_prop_default_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_default_string'>
	<parameter type-id='5aa5c90c'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='vdev_prop_default_numeric' mangled-name='vdev_prop_default_numeric' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_default_numeric'>
	<parameter type-id='5aa5c90c'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='vdev_prop_readonly' mangled-name='vdev_prop_readonly' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_readonly'>
	<parameter type-id='5aa5c90c'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='vdev_prop_index_to_string' mangled-name='vdev_prop_index_to_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_index_to_string'>
	<parameter type-id='5aa5c90c'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='7d3cd834'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_prop_vdev' mangled-name='zpool_prop_vdev' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_vdev'>
	<parameter type-id='80f4b756'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='nvlist_add_nvpair' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='3fa542f0'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_uint8_array' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='9f7200cf'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_add_nvlist_array' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='3bbfee2e'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_nvlist' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='3fa542f0'/>
	<parameter type-id='857bb57e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='fnvlist_add_boolean_value' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='c19b74c3'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_int64' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='9da381c4'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_nvlist_array' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='3bbfee2e'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_uint64_array' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='4dd26a40'/>
	<return type-id='5d6479ae'/>
	</function-decl>
	<function-decl name='fnvpair_value_int64' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='dace003f'/>
	<return type-id='9da381c4'/>
	</function-decl>
	<function-decl name='fnvpair_value_string' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='dace003f'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfeature_is_supported' mangled-name='zfeature_is_supported' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfeature_is_supported'>
	<parameter type-id='80f4b756'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfeature_lookup_guid' mangled-name='zfeature_lookup_guid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfeature_lookup_guid'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='a8425263'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfeature_lookup_name' mangled-name='zfeature_lookup_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfeature_lookup_name'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='a8425263'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_get_load_policy' mangled-name='zpool_get_load_policy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_load_policy'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='23432aaa'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='pool_namecheck' mangled-name='pool_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='pool_namecheck'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='053457bd'/>
	<parameter type-id='26a90f95'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_prop_get_type' mangled-name='zpool_prop_get_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_get_type'>
	<parameter type-id='5d0c23fb'/>
	<return type-id='31429eff'/>
	</function-decl>
	<function-decl name='vdev_prop_get_type' mangled-name='vdev_prop_get_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_get_type'>
	<parameter type-id='5aa5c90c'/>
	<return type-id='31429eff'/>
	</function-decl>
	<function-decl name='get_system_hostid' mangled-name='get_system_hostid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='get_system_hostid'>
	<return type-id='7359adad'/>
	</function-decl>
	<function-decl name='strtoull' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9d26089a'/>
	<parameter type-id='8c85230f'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='3a47d82b'/>
	</function-decl>
	<function-decl name='memcmp' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='strtok_r' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='266fe297'/>
	<parameter type-id='9d26089a'/>
	<parameter type-id='8c85230f'/>
	<return type-id='26a90f95'/>
	</function-decl>
	+ <function-decl name='ctime_r' visibility='default' binding='global' size-in-bits='64'>
	+ <parameter type-id='d6e2847c'/>
	+ <parameter type-id='266fe297'/>
	+ <return type-id='26a90f95'/>
	+ </function-decl>
	<function-decl name='__realpath_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9d26089a'/>
	<parameter type-id='266fe297'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='munmap' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='stat64' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9d26089a'/>
	<parameter type-id='f1cadedf'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_standard_error' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_standard_error_fmt' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<parameter is-variadic='yes'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_relabel_disk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_props_refresh' mangled-name='zpool_props_refresh' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_props_refresh'>
	<parameter type-id='4c81de99' name='zhp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_state_to_name' mangled-name='zpool_state_to_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_state_to_name'>
	<parameter type-id='35acf840' name='state'/>
	<parameter type-id='9d774e0b' name='aux'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zpool_pool_state_to_name' mangled-name='zpool_pool_state_to_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_pool_state_to_name'>
	<parameter type-id='084a08a3' name='state'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zpool_get_state_str' mangled-name='zpool_get_state_str' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_state_str'>
	<parameter type-id='4c81de99' name='zhp'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zpool_get_userprop' mangled-name='zpool_get_userprop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_userprop'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='propname'/>
	<parameter type-id='26a90f95' name='buf'/>
	<parameter type-id='b59d7dce' name='len'/>
	<parameter type-id='debc6aa3' name='srctype'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_set_prop' mangled-name='zpool_set_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_set_prop'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='propname'/>
	<parameter type-id='80f4b756' name='propval'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_expand_proplist' mangled-name='zpool_expand_proplist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_expand_proplist'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='e4378506' name='plp'/>
	<parameter type-id='2e45de5d' name='type'/>
	<parameter type-id='c19b74c3' name='literal'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='vdev_expand_proplist' mangled-name='vdev_expand_proplist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_expand_proplist'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='vdevname'/>
	<parameter type-id='e4378506' name='plp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_get_state' mangled-name='zpool_get_state' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_state'>
	<parameter type-id='4c81de99' name='zhp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_is_draid_spare' mangled-name='zpool_is_draid_spare' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_is_draid_spare'>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zpool_create' mangled-name='zpool_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_create'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='pool'/>
	<parameter type-id='5ce45b60' name='nvroot'/>
	<parameter type-id='5ce45b60' name='props'/>
	<parameter type-id='5ce45b60' name='fsprops'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_destroy' mangled-name='zpool_destroy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_destroy'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='log_str'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_checkpoint' mangled-name='zpool_checkpoint' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_checkpoint'>
	<parameter type-id='4c81de99' name='zhp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_discard_checkpoint' mangled-name='zpool_discard_checkpoint' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_discard_checkpoint'>
	<parameter type-id='4c81de99' name='zhp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_add' mangled-name='zpool_add' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_add'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='5ce45b60' name='nvroot'/>
	- <parameter type-id='c19b74c3' name='ashift_check'/>
	+ <parameter type-id='c19b74c3' name='check_ashift'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_export' mangled-name='zpool_export' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_export'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='c19b74c3' name='force'/>
	<parameter type-id='80f4b756' name='log_str'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_export_force' mangled-name='zpool_export_force' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_export_force'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='log_str'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_explain_recover' mangled-name='zpool_explain_recover' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_explain_recover'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='95e97e5e' name='reason'/>
	<parameter type-id='5ce45b60' name='config'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zpool_import' mangled-name='zpool_import' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_import'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='5ce45b60' name='config'/>
	<parameter type-id='80f4b756' name='newname'/>
	<parameter type-id='26a90f95' name='altroot'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_print_unsup_feat' mangled-name='zpool_print_unsup_feat' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_print_unsup_feat'>
	<parameter type-id='5ce45b60' name='config'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zpool_import_props' mangled-name='zpool_import_props' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_import_props'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='5ce45b60' name='config'/>
	<parameter type-id='80f4b756' name='newname'/>
	<parameter type-id='5ce45b60' name='props'/>
	<parameter type-id='95e97e5e' name='flags'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_initialize' mangled-name='zpool_initialize' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_initialize'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='7063e1ab' name='cmd_type'/>
	<parameter type-id='5ce45b60' name='vds'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_initialize_wait' mangled-name='zpool_initialize_wait' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_initialize_wait'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='7063e1ab' name='cmd_type'/>
	<parameter type-id='5ce45b60' name='vds'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_trim' mangled-name='zpool_trim' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_trim'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='b1146b8d' name='cmd_type'/>
	<parameter type-id='5ce45b60' name='vds'/>
	<parameter type-id='b13f38c3' name='trim_flags'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_scan' mangled-name='zpool_scan' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_scan'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='7313fbe2' name='func'/>
	<parameter type-id='b51cf3c2' name='cmd'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_find_vdev_by_physpath' mangled-name='zpool_find_vdev_by_physpath' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_find_vdev_by_physpath'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='ppath'/>
	<parameter type-id='37e3bd22' name='avail_spare'/>
	<parameter type-id='37e3bd22' name='l2cache'/>
	<parameter type-id='37e3bd22' name='log'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='zpool_find_vdev' mangled-name='zpool_find_vdev' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_find_vdev'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='37e3bd22' name='avail_spare'/>
	<parameter type-id='37e3bd22' name='l2cache'/>
	<parameter type-id='37e3bd22' name='log'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='zpool_vdev_path_to_guid' mangled-name='zpool_vdev_path_to_guid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_path_to_guid'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='path'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='zpool_vdev_online' mangled-name='zpool_vdev_online' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_online'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='95e97e5e' name='flags'/>
	<parameter type-id='17f3480d' name='newstate'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_offline' mangled-name='zpool_vdev_offline' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_offline'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='c19b74c3' name='istmp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_remove_wanted' mangled-name='zpool_vdev_remove_wanted' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_remove_wanted'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='path'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_fault' mangled-name='zpool_vdev_fault' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_fault'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='9c313c2d' name='guid'/>
	<parameter type-id='9d774e0b' name='aux'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_degrade' mangled-name='zpool_vdev_degrade' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_degrade'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='9c313c2d' name='guid'/>
	<parameter type-id='9d774e0b' name='aux'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_set_removed_state' mangled-name='zpool_vdev_set_removed_state' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_set_removed_state'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='9c313c2d' name='guid'/>
	<parameter type-id='9d774e0b' name='aux'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_attach' mangled-name='zpool_vdev_attach' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_attach'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='old_disk'/>
	<parameter type-id='80f4b756' name='new_disk'/>
	<parameter type-id='5ce45b60' name='nvroot'/>
	<parameter type-id='95e97e5e' name='replacing'/>
	<parameter type-id='c19b74c3' name='rebuild'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_detach' mangled-name='zpool_vdev_detach' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_detach'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='path'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_split' mangled-name='zpool_vdev_split' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_split'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='26a90f95' name='newname'/>
	<parameter type-id='857bb57e' name='newroot'/>
	<parameter type-id='5ce45b60' name='props'/>
	<parameter type-id='325c1e34' name='flags'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_remove' mangled-name='zpool_vdev_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_remove'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='path'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_remove_cancel' mangled-name='zpool_vdev_remove_cancel' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_remove_cancel'>
	<parameter type-id='4c81de99' name='zhp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_indirect_size' mangled-name='zpool_vdev_indirect_size' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_indirect_size'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='5d6479ae' name='sizep'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_clear' mangled-name='zpool_clear' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_clear'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='5ce45b60' name='rewindnvl'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_clear' mangled-name='zpool_vdev_clear' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_clear'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='9c313c2d' name='guid'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_reguid' mangled-name='zpool_reguid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_reguid'>
	<parameter type-id='4c81de99' name='zhp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_reopen_one' mangled-name='zpool_reopen_one' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_reopen_one'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_sync_one' mangled-name='zpool_sync_one' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_sync_one'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_name' mangled-name='zpool_vdev_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_name'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='5ce45b60' name='nv'/>
	<parameter type-id='95e97e5e' name='name_flags'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='zpool_get_errlog' mangled-name='zpool_get_errlog' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_errlog'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='857bb57e' name='nverrlistp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_upgrade' mangled-name='zpool_upgrade' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_upgrade'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='9c313c2d' name='new_version'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_save_arguments' mangled-name='zfs_save_arguments' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_save_arguments'>
	<parameter type-id='95e97e5e' name='argc'/>
	<parameter type-id='9b23c9ad' name='argv'/>
	<parameter type-id='26a90f95' name='string'/>
	<parameter type-id='95e97e5e' name='len'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zpool_log_history' mangled-name='zpool_log_history' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_log_history'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='message'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_get_history' mangled-name='zpool_get_history' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_history'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='857bb57e' name='nvhisp'/>
	<parameter type-id='5d6479ae' name='off'/>
	<parameter type-id='37e3bd22' name='eof'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_events_next' mangled-name='zpool_events_next' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_events_next'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='857bb57e' name='nvp'/>
	<parameter type-id='7292109c' name='dropped'/>
	<parameter type-id='f0981eeb' name='flags'/>
	<parameter type-id='95e97e5e' name='zevent_fd'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_events_clear' mangled-name='zpool_events_clear' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_events_clear'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='7292109c' name='count'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_events_seek' mangled-name='zpool_events_seek' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_events_seek'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='9c313c2d' name='eid'/>
	<parameter type-id='95e97e5e' name='zevent_fd'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_obj_to_path' mangled-name='zpool_obj_to_path' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_obj_to_path'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='9c313c2d' name='dsobj'/>
	<parameter type-id='9c313c2d' name='obj'/>
	<parameter type-id='26a90f95' name='pathname'/>
	<parameter type-id='b59d7dce' name='len'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zpool_obj_to_path_ds' mangled-name='zpool_obj_to_path_ds' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_obj_to_path_ds'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='9c313c2d' name='dsobj'/>
	<parameter type-id='9c313c2d' name='obj'/>
	<parameter type-id='26a90f95' name='pathname'/>
	<parameter type-id='b59d7dce' name='len'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zpool_wait' mangled-name='zpool_wait' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_wait'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='73446457' name='activity'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_wait_status' mangled-name='zpool_wait_status' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_wait_status'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='73446457' name='activity'/>
	<parameter type-id='37e3bd22' name='missing'/>
	<parameter type-id='37e3bd22' name='waited'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_set_bootenv' mangled-name='zpool_set_bootenv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_set_bootenv'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='22cce67b' name='envmap'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_get_bootenv' mangled-name='zpool_get_bootenv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_bootenv'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='857bb57e' name='nvlp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_load_compat' mangled-name='zpool_load_compat' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_load_compat'>
	<parameter type-id='80f4b756' name='compat'/>
	<parameter type-id='37e3bd22' name='features'/>
	<parameter type-id='26a90f95' name='report'/>
	<parameter type-id='b59d7dce' name='rlen'/>
	<return type-id='901b78d1'/>
	</function-decl>
	<function-decl name='zpool_get_vdev_prop_value' mangled-name='zpool_get_vdev_prop_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_vdev_prop_value'>
	<parameter type-id='5ce45b60' name='nvprop'/>
	<parameter type-id='5aa5c90c' name='prop'/>
	<parameter type-id='26a90f95' name='prop_name'/>
	<parameter type-id='26a90f95' name='buf'/>
	<parameter type-id='b59d7dce' name='len'/>
	<parameter type-id='debc6aa3' name='srctype'/>
	<parameter type-id='c19b74c3' name='literal'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_get_vdev_prop' mangled-name='zpool_get_vdev_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_vdev_prop'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='vdevname'/>
	<parameter type-id='5aa5c90c' name='prop'/>
	<parameter type-id='26a90f95' name='prop_name'/>
	<parameter type-id='26a90f95' name='buf'/>
	<parameter type-id='b59d7dce' name='len'/>
	<parameter type-id='debc6aa3' name='srctype'/>
	<parameter type-id='c19b74c3' name='literal'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_get_all_vdev_props' mangled-name='zpool_get_all_vdev_props' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_get_all_vdev_props'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='vdevname'/>
	<parameter type-id='857bb57e' name='outnvl'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_set_vdev_prop' mangled-name='zpool_set_vdev_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_set_vdev_prop'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='vdevname'/>
	<parameter type-id='80f4b756' name='propname'/>
	<parameter type-id='80f4b756' name='propval'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_sendrecv.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='8901473c' size-in-bits='576' id='f5da478b'>
	<subrange length='1' type-id='7359adad' id='52f813b4'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='95e97e5e' size-in-bits='384' id='73b82f0f'>
	<subrange length='12' type-id='7359adad' id='84827bdc'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='bd54fe1a' size-in-bits='512' id='5d4efd44'>
	<subrange length='8' type-id='7359adad' id='56e0c0b1'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='9c313c2d' size-in-bits='2176' id='8c2bcad1'>
	<subrange length='34' type-id='7359adad' id='6a6a7e00'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='9c313c2d' size-in-bits='256' id='85c64d26'>
	<subrange length='4' type-id='7359adad' id='16fe7105'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='b96825af' size-in-bits='96' id='fa8ef949'>
	<subrange length='12' type-id='7359adad' id='84827bdc'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='b96825af' size-in-bits='128' id='fa9986a5'>
	<subrange length='16' type-id='7359adad' id='848d0938'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='b96825af' size-in-bits='40' id='0f4ddd0b'>
	<subrange length='5' type-id='7359adad' id='53010e10'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='b96825af' size-in-bits='48' id='0f562bd0'>
	<subrange length='6' type-id='7359adad' id='52fa524b'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='b96825af' size-in-bits='64' id='13339fda'>
	<subrange length='8' type-id='7359adad' id='56e0c0b1'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='eaa32e2f' size-in-bits='256' id='209ef23f'>
	<subrange length='4' type-id='7359adad' id='16fe7105'/>
	</array-type-def>
	<class-decl name='sendflags' size-in-bits='576' is-struct='yes' visibility='default' id='f6aa15be'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='verbosity' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='replicate' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='skipmissing' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='doall' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='fromorigin' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='160'>
	<var-decl name='pad' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='props' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='224'>
	<var-decl name='dryrun' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='parsable' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='progress' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='progressastitle' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='352'>
	<var-decl name='largeblock' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='embed_data' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='416'>
	<var-decl name='compress' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='raw' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='480'>
	<var-decl name='backup' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='holds' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='544'>
	<var-decl name='saved' type-id='c19b74c3' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='sendflags_t' type-id='f6aa15be' id='945467e6'/>
	<typedef-decl name='snapfilter_cb_t' type-id='d2a5e211' id='3d3ffb69'/>
	<class-decl name='recvflags' size-in-bits='448' is-struct='yes' visibility='default' id='34a384dc'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='verbose' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='isprefix' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='istail' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='dryrun' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='force' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='160'>
	<var-decl name='canmountoff' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='resumable' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='224'>
	<var-decl name='byteswap' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='nomount' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='holds' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='skipholds' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='352'>
	<var-decl name='domount' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='forceunmount' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='416'>
	<var-decl name='heal' type-id='c19b74c3' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='recvflags_t' type-id='34a384dc' id='9e59d1d4'/>
	<enum-decl name='lzc_send_flags' id='bfbd3c8e'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='LZC_SEND_FLAG_EMBED_DATA' value='1'/>
	<enumerator name='LZC_SEND_FLAG_LARGE_BLOCK' value='2'/>
	<enumerator name='LZC_SEND_FLAG_COMPRESS' value='4'/>
	<enumerator name='LZC_SEND_FLAG_RAW' value='8'/>
	<enumerator name='LZC_SEND_FLAG_SAVED' value='16'/>
	</enum-decl>
	<class-decl name='ddt_key_t' size-in-bits='320' is-struct='yes' naming-typedef-id='67f6d2cf' visibility='default' id='5fae1718'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='ddk_cksum' type-id='39730d0b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='ddk_prop' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='ddt_key_t' type-id='5fae1718' id='67f6d2cf'/>
	<enum-decl name='dmu_object_type' id='04b3b0b9'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='DMU_OT_NONE' value='0'/>
	<enumerator name='DMU_OT_OBJECT_DIRECTORY' value='1'/>
	<enumerator name='DMU_OT_OBJECT_ARRAY' value='2'/>
	<enumerator name='DMU_OT_PACKED_NVLIST' value='3'/>
	<enumerator name='DMU_OT_PACKED_NVLIST_SIZE' value='4'/>
	<enumerator name='DMU_OT_BPOBJ' value='5'/>
	<enumerator name='DMU_OT_BPOBJ_HDR' value='6'/>
	<enumerator name='DMU_OT_SPACE_MAP_HEADER' value='7'/>
	<enumerator name='DMU_OT_SPACE_MAP' value='8'/>
	<enumerator name='DMU_OT_INTENT_LOG' value='9'/>
	<enumerator name='DMU_OT_DNODE' value='10'/>
	<enumerator name='DMU_OT_OBJSET' value='11'/>
	<enumerator name='DMU_OT_DSL_DIR' value='12'/>
	<enumerator name='DMU_OT_DSL_DIR_CHILD_MAP' value='13'/>
	<enumerator name='DMU_OT_DSL_DS_SNAP_MAP' value='14'/>
	<enumerator name='DMU_OT_DSL_PROPS' value='15'/>
	<enumerator name='DMU_OT_DSL_DATASET' value='16'/>
	<enumerator name='DMU_OT_ZNODE' value='17'/>
	<enumerator name='DMU_OT_OLDACL' value='18'/>
	<enumerator name='DMU_OT_PLAIN_FILE_CONTENTS' value='19'/>
	<enumerator name='DMU_OT_DIRECTORY_CONTENTS' value='20'/>
	<enumerator name='DMU_OT_MASTER_NODE' value='21'/>
	<enumerator name='DMU_OT_UNLINKED_SET' value='22'/>
	<enumerator name='DMU_OT_ZVOL' value='23'/>
	<enumerator name='DMU_OT_ZVOL_PROP' value='24'/>
	<enumerator name='DMU_OT_PLAIN_OTHER' value='25'/>
	<enumerator name='DMU_OT_UINT64_OTHER' value='26'/>
	<enumerator name='DMU_OT_ZAP_OTHER' value='27'/>
	<enumerator name='DMU_OT_ERROR_LOG' value='28'/>
	<enumerator name='DMU_OT_SPA_HISTORY' value='29'/>
	<enumerator name='DMU_OT_SPA_HISTORY_OFFSETS' value='30'/>
	<enumerator name='DMU_OT_POOL_PROPS' value='31'/>
	<enumerator name='DMU_OT_DSL_PERMS' value='32'/>
	<enumerator name='DMU_OT_ACL' value='33'/>
	<enumerator name='DMU_OT_SYSACL' value='34'/>
	<enumerator name='DMU_OT_FUID' value='35'/>
	<enumerator name='DMU_OT_FUID_SIZE' value='36'/>
	<enumerator name='DMU_OT_NEXT_CLONES' value='37'/>
	<enumerator name='DMU_OT_SCAN_QUEUE' value='38'/>
	<enumerator name='DMU_OT_USERGROUP_USED' value='39'/>
	<enumerator name='DMU_OT_USERGROUP_QUOTA' value='40'/>
	<enumerator name='DMU_OT_USERREFS' value='41'/>
	<enumerator name='DMU_OT_DDT_ZAP' value='42'/>
	<enumerator name='DMU_OT_DDT_STATS' value='43'/>
	<enumerator name='DMU_OT_SA' value='44'/>
	<enumerator name='DMU_OT_SA_MASTER_NODE' value='45'/>
	<enumerator name='DMU_OT_SA_ATTR_REGISTRATION' value='46'/>
	<enumerator name='DMU_OT_SA_ATTR_LAYOUTS' value='47'/>
	<enumerator name='DMU_OT_SCAN_XLATE' value='48'/>
	<enumerator name='DMU_OT_DEDUP' value='49'/>
	<enumerator name='DMU_OT_DEADLIST' value='50'/>
	<enumerator name='DMU_OT_DEADLIST_HDR' value='51'/>
	<enumerator name='DMU_OT_DSL_CLONES' value='52'/>
	<enumerator name='DMU_OT_BPOBJ_SUBOBJ' value='53'/>
	<enumerator name='DMU_OT_NUMTYPES' value='54'/>
	<enumerator name='DMU_OTN_UINT8_DATA' value='128'/>
	<enumerator name='DMU_OTN_UINT8_METADATA' value='192'/>
	<enumerator name='DMU_OTN_UINT16_DATA' value='129'/>
	<enumerator name='DMU_OTN_UINT16_METADATA' value='193'/>
	<enumerator name='DMU_OTN_UINT32_DATA' value='130'/>
	<enumerator name='DMU_OTN_UINT32_METADATA' value='194'/>
	<enumerator name='DMU_OTN_UINT64_DATA' value='131'/>
	<enumerator name='DMU_OTN_UINT64_METADATA' value='195'/>
	<enumerator name='DMU_OTN_ZAP_DATA' value='132'/>
	<enumerator name='DMU_OTN_ZAP_METADATA' value='196'/>
	<enumerator name='DMU_OTN_UINT8_ENC_DATA' value='160'/>
	<enumerator name='DMU_OTN_UINT8_ENC_METADATA' value='224'/>
	<enumerator name='DMU_OTN_UINT16_ENC_DATA' value='161'/>
	<enumerator name='DMU_OTN_UINT16_ENC_METADATA' value='225'/>
	<enumerator name='DMU_OTN_UINT32_ENC_DATA' value='162'/>
	<enumerator name='DMU_OTN_UINT32_ENC_METADATA' value='226'/>
	<enumerator name='DMU_OTN_UINT64_ENC_DATA' value='163'/>
	<enumerator name='DMU_OTN_UINT64_ENC_METADATA' value='227'/>
	<enumerator name='DMU_OTN_ZAP_ENC_DATA' value='164'/>
	<enumerator name='DMU_OTN_ZAP_ENC_METADATA' value='228'/>
	</enum-decl>
	<typedef-decl name='dmu_object_type_t' type-id='04b3b0b9' id='5c9d8906'/>
	<class-decl name='zio_cksum' size-in-bits='256' is-struct='yes' visibility='default' id='1d53e28b'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='zc_word' type-id='85c64d26' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zio_cksum_t' type-id='1d53e28b' id='39730d0b'/>
	<class-decl name='dmu_replay_record' size-in-bits='2496' is-struct='yes' visibility='default' id='781a52d7'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_type' type-id='08f5ca17' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='drr_payloadlen' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='drr_u' type-id='ac5ab598' visibility='default'/>
	</data-member>
	</class-decl>
	<enum-decl name='__anonymous_enum__' is-anonymous='yes' id='08f5ca17'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='DRR_BEGIN' value='0'/>
	<enumerator name='DRR_OBJECT' value='1'/>
	<enumerator name='DRR_FREEOBJECTS' value='2'/>
	<enumerator name='DRR_WRITE' value='3'/>
	<enumerator name='DRR_FREE' value='4'/>
	<enumerator name='DRR_END' value='5'/>
	<enumerator name='DRR_WRITE_BYREF' value='6'/>
	<enumerator name='DRR_SPILL' value='7'/>
	<enumerator name='DRR_WRITE_EMBEDDED' value='8'/>
	<enumerator name='DRR_OBJECT_RANGE' value='9'/>
	<enumerator name='DRR_REDACT' value='10'/>
	<enumerator name='DRR_NUMTYPES' value='11'/>
	</enum-decl>
	<union-decl name='__anonymous_union__' size-in-bits='2432' is-anonymous='yes' visibility='default' id='ac5ab598'>
	<data-member access='public'>
	<var-decl name='drr_begin' type-id='09fcdc01' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='drr_end' type-id='6ee25631' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='drr_object' type-id='f9ad530b' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='drr_freeobjects' type-id='a27d958e' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='drr_write' type-id='4cc69e4b' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='drr_free' type-id='c836cfd2' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='drr_write_byref' type-id='e511cdce' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='drr_spill' type-id='1e69a80a' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='drr_write_embedded' type-id='98b1345e' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='drr_object_range' type-id='aba1f9e1' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='drr_redact' type-id='50389039' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='drr_checksum' type-id='a5fe3647' visibility='default'/>
	</data-member>
	</union-decl>
	<class-decl name='drr_end' size-in-bits='320' is-struct='yes' visibility='default' id='6ee25631'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_checksum' type-id='39730d0b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='drr_object' size-in-bits='448' is-struct='yes' visibility='default' id='f9ad530b'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_object' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='drr_type' type-id='5c9d8906' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='drr_bonustype' type-id='5c9d8906' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='drr_blksz' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='160'>
	<var-decl name='drr_bonuslen' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='drr_checksumtype' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='200'>
	<var-decl name='drr_compress' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='208'>
	<var-decl name='drr_dn_slots' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='216'>
	<var-decl name='drr_flags' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='224'>
	<var-decl name='drr_raw_bonuslen' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='drr_indblkshift' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='328'>
	<var-decl name='drr_nlevels' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='336'>
	<var-decl name='drr_nblkptr' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='344'>
	<var-decl name='drr_pad' type-id='0f4ddd0b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='drr_maxblkid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='drr_freeobjects' size-in-bits='192' is-struct='yes' visibility='default' id='a27d958e'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_firstobj' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='drr_numobjs' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='drr_write' size-in-bits='1088' is-struct='yes' visibility='default' id='4cc69e4b'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_object' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='drr_type' type-id='5c9d8906' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='drr_pad' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='drr_offset' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='drr_logical_size' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='drr_checksumtype' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='328'>
	<var-decl name='drr_flags' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='336'>
	<var-decl name='drr_compressiontype' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='344'>
	<var-decl name='drr_pad2' type-id='0f4ddd0b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='drr_key' type-id='67f6d2cf' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='704'>
	<var-decl name='drr_compressed_size' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='768'>
	<var-decl name='drr_salt' type-id='13339fda' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='832'>
	<var-decl name='drr_iv' type-id='fa8ef949' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='928'>
	<var-decl name='drr_mac' type-id='fa9986a5' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='drr_free' size-in-bits='256' is-struct='yes' visibility='default' id='c836cfd2'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_object' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='drr_offset' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='drr_length' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='drr_write_byref' size-in-bits='832' is-struct='yes' visibility='default' id='e511cdce'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_object' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='drr_offset' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='drr_length' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='drr_refguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='drr_refobject' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='drr_refoffset' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='drr_checksumtype' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='456'>
	<var-decl name='drr_flags' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='464'>
	<var-decl name='drr_pad2' type-id='0f562bd0' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='drr_key' type-id='67f6d2cf' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='drr_spill' size-in-bits='640' is-struct='yes' visibility='default' id='1e69a80a'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_object' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='drr_length' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='drr_flags' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='200'>
	<var-decl name='drr_compressiontype' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='208'>
	<var-decl name='drr_pad' type-id='0f562bd0' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='drr_compressed_size' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='drr_salt' type-id='13339fda' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='drr_iv' type-id='fa8ef949' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='480'>
	<var-decl name='drr_mac' type-id='fa9986a5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='608'>
	<var-decl name='drr_type' type-id='5c9d8906' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='drr_write_embedded' size-in-bits='384' is-struct='yes' visibility='default' id='98b1345e'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_object' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='drr_offset' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='drr_length' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='drr_compression' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='264'>
	<var-decl name='drr_etype' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='272'>
	<var-decl name='drr_pad' type-id='0f562bd0' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='drr_lsize' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='352'>
	<var-decl name='drr_psize' type-id='8f92235e' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='drr_object_range' size-in-bits='512' is-struct='yes' visibility='default' id='aba1f9e1'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_firstobj' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='drr_numslots' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='drr_salt' type-id='13339fda' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='drr_iv' type-id='fa8ef949' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='352'>
	<var-decl name='drr_mac' type-id='fa9986a5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='480'>
	<var-decl name='drr_flags' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='488'>
	<var-decl name='drr_pad' type-id='d3490169' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='drr_redact' size-in-bits='256' is-struct='yes' visibility='default' id='50389039'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_object' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='drr_offset' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='drr_length' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='drr_checksum' size-in-bits='2432' is-struct='yes' visibility='default' id='a5fe3647'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_pad' type-id='8c2bcad1' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2176'>
	<var-decl name='drr_checksum' type-id='39730d0b' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__cancel_jmp_buf_tag' size-in-bits='576' is-struct='yes' visibility='default' id='8901473c'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='__cancel_jmp_buf' type-id='379a1ab7' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='__mask_was_saved' type-id='95e97e5e' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__pthread_unwind_buf_t' size-in-bits='832' is-struct='yes' naming-typedef-id='4423cf7f' visibility='default' id='a0abc656'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='__cancel_jmp_buf' type-id='f5da478b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='576'>
	<var-decl name='__pad' type-id='209ef23f' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='__pthread_unwind_buf_t' type-id='a0abc656' id='4423cf7f'/>
	<typedef-decl name='__jmp_buf' type-id='5d4efd44' id='379a1ab7'/>
	<typedef-decl name='__clockid_t' type-id='95e97e5e' id='08f9a87a'/>
	<typedef-decl name='__timer_t' type-id='eaa32e2f' id='df209b60'/>
	<typedef-decl name='clockid_t' type-id='08f9a87a' id='a1c3b834'/>
	<class-decl name='sigevent' size-in-bits='512' is-struct='yes' visibility='default' id='519bc206'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='sigev_value' type-id='eabacd01' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='sigev_signo' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='sigev_notify' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='_sigev_un' type-id='ac5ab599' visibility='default'/>
	</data-member>
	</class-decl>
	<union-decl name='__anonymous_union__1' size-in-bits='384' is-anonymous='yes' visibility='default' id='ac5ab599'>
	<data-member access='public'>
	<var-decl name='_pad' type-id='73b82f0f' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='_tid' type-id='3629bad8' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='_sigev_thread' type-id='e7f43f7b' visibility='default'/>
	</data-member>
	</union-decl>
	<class-decl name='__anonymous_struct__' size-in-bits='128' is-struct='yes' is-anonymous='yes' visibility='default' id='e7f43f7b'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='_function' type-id='5f147c28' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='_attribute' type-id='7347a39e' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='itimerspec' size-in-bits='256' is-struct='yes' visibility='default' id='acbdbcc6'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='it_interval' type-id='a9c79a1f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='it_value' type-id='a9c79a1f' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='timer_t' type-id='df209b60' id='b07ae406'/>
	<typedef-decl name='Byte' type-id='002ac4a6' id='efb9ba06'/>
	<typedef-decl name='uLong' type-id='7359adad' id='5bbcce85'/>
	<typedef-decl name='Bytef' type-id='efb9ba06' id='c1606520'/>
	<typedef-decl name='uLongf' type-id='5bbcce85' id='4d39af59'/>
	<pointer-type-def type-id='c1606520' size-in-bits='64' id='4c667223'/>
	<pointer-type-def type-id='8901473c' size-in-bits='64' id='eb91b7ea'/>
	<pointer-type-def type-id='4423cf7f' size-in-bits='64' id='ba7c727c'/>
	<pointer-type-def type-id='b9c97942' size-in-bits='64' id='bbf06c47'/>
	<qualified-type-def type-id='bbf06c47' restrict='yes' id='65e6ec45'/>
	<qualified-type-def type-id='c1606520' const='yes' id='a6124a50'/>
	<pointer-type-def type-id='a6124a50' size-in-bits='64' id='e8cb3e0e'/>
	<qualified-type-def type-id='b9c97942' const='yes' id='191f6b72'/>
	<pointer-type-def type-id='191f6b72' size-in-bits='64' id='e475fb88'/>
	<qualified-type-def type-id='e475fb88' restrict='yes' id='5a8729d0'/>
	<qualified-type-def type-id='781a52d7' const='yes' id='413ab2b8'/>
	<pointer-type-def type-id='413ab2b8' size-in-bits='64' id='41671bd6'/>
	<qualified-type-def type-id='acbdbcc6' const='yes' id='4ba62af7'/>
	<pointer-type-def type-id='4ba62af7' size-in-bits='64' id='f39579e7'/>
	<qualified-type-def type-id='f39579e7' restrict='yes' id='9b23e165'/>
	<pointer-type-def type-id='c70fa2e8' size-in-bits='64' id='2e711a2a'/>
	<pointer-type-def type-id='3ff5601b' size-in-bits='64' id='4aafb922'/>
	<pointer-type-def type-id='acbdbcc6' size-in-bits='64' id='116842ac'/>
	<qualified-type-def type-id='116842ac' restrict='yes' id='3d3c4cf4'/>
	<pointer-type-def type-id='9e59d1d4' size-in-bits='64' id='4ea84b4f'/>
	<pointer-type-def type-id='945467e6' size-in-bits='64' id='8def7735'/>
	<pointer-type-def type-id='519bc206' size-in-bits='64' id='ef2f159c'/>
	<qualified-type-def type-id='ef2f159c' restrict='yes' id='de0eb5a4'/>
	<pointer-type-def type-id='3d3ffb69' size-in-bits='64' id='72a26210'/>
	<pointer-type-def type-id='c9d12d66' size-in-bits='64' id='b2eb2c3f'/>
	<pointer-type-def type-id='b07ae406' size-in-bits='64' id='36e89359'/>
	<qualified-type-def type-id='36e89359' restrict='yes' id='de98c2bb'/>
	<pointer-type-def type-id='a9c79a1f' size-in-bits='64' id='3d83ba87'/>
	<pointer-type-def type-id='4d39af59' size-in-bits='64' id='60db3356'/>
	<pointer-type-def type-id='f1abb096' size-in-bits='64' id='5f147c28'/>
	<pointer-type-def type-id='39730d0b' size-in-bits='64' id='c24fc2ee'/>
	<function-decl name='nvlist_print' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='822cd80b'/>
	<parameter type-id='5ce45b60'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_get_pool_handle' mangled-name='zfs_get_pool_handle' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_pool_handle'>
	<parameter type-id='fcd57163'/>
	<return type-id='4c81de99'/>
	</function-decl>
	<function-decl name='lzc_send_wrapper' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='2e711a2a'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_send_redacted' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='bfbd3c8e'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_send_resume_redacted' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='bfbd3c8e'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_receive_with_cmdprops' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='ae3e8ca6'/>
	<parameter type-id='3502e3ff'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='c19b74c3'/>
	<parameter type-id='c19b74c3'/>
	<parameter type-id='c19b74c3'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='41671bd6'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='5d6479ae'/>
	<parameter type-id='5d6479ae'/>
	<parameter type-id='5d6479ae'/>
	<parameter type-id='857bb57e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_receive_with_heal' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='ae3e8ca6'/>
	<parameter type-id='3502e3ff'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='c19b74c3'/>
	<parameter type-id='c19b74c3'/>
	<parameter type-id='c19b74c3'/>
	<parameter type-id='c19b74c3'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='41671bd6'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='5d6479ae'/>
	<parameter type-id='5d6479ae'/>
	<parameter type-id='5d6479ae'/>
	<parameter type-id='857bb57e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_send_space' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='bfbd3c8e'/>
	<parameter type-id='5d6479ae'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_send_space_resume_redacted' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='bfbd3c8e'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='5d6479ae'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_rename' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_setproctitle' mangled-name='zfs_setproctitle' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_setproctitle'>
	<parameter type-id='80f4b756'/>
	<parameter is-variadic='yes'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='avl_insert' mangled-name='avl_insert' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_insert'>
	<parameter type-id='a3681dea'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='fba6cb51'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_add_boolean' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_boolean' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='22cce67b'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvpair_value_int32' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='dace003f'/>
	<parameter type-id='4aafb922'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='fnvlist_size' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='fnvlist_merge' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='5ce45b60'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_remove' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_boolean_value' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='22cce67b'/>
	<parameter type-id='80f4b756'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='fletcher_4_native_varsize' mangled-name='fletcher_4_native_varsize' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_native_varsize'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='c24fc2ee'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fletcher_4_incremental_native' mangled-name='fletcher_4_incremental_native' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_incremental_native'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='fletcher_4_incremental_byteswap' mangled-name='fletcher_4_incremental_byteswap' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_incremental_byteswap'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_exit' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='eaa32e2f'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='__pthread_register_cancel' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='ba7c727c'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='__pthread_unwind_next' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='ba7c727c'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='sigaddset' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9e80f729'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='perror' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='strndup' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='time' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b2eb2c3f'/>
	<return type-id='c9d12d66'/>
	</function-decl>
	<function-decl name='clock_gettime' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='a1c3b834'/>
	<parameter type-id='3d83ba87'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='timer_create' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='a1c3b834'/>
	<parameter type-id='de0eb5a4'/>
	<parameter type-id='de98c2bb'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='timer_delete' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b07ae406'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='timer_settime' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b07ae406'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='9b23e165'/>
	<parameter type-id='3d3c4cf4'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='write' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='79a0948f'/>
	</function-decl>
	<function-decl name='pause' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_sigmask' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='5a8729d0'/>
	<parameter type-id='65e6ec45'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='uncompress' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='4c667223'/>
	<parameter type-id='60db3356'/>
	<parameter type-id='e8cb3e0e'/>
	<parameter type-id='5bbcce85'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='create_parents' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='26a90f95'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_send_progress' mangled-name='zfs_send_progress' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_send_progress'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='95e97e5e' name='fd'/>
	<parameter type-id='5d6479ae' name='bytes_written'/>
	<parameter type-id='5d6479ae' name='blocks_visited'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_send_resume_token_to_nvlist' mangled-name='zfs_send_resume_token_to_nvlist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_send_resume_token_to_nvlist'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='token'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='zfs_send_resume' mangled-name='zfs_send_resume' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_send_resume'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='8def7735' name='flags'/>
	<parameter type-id='95e97e5e' name='outfd'/>
	<parameter type-id='80f4b756' name='resume_token'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_send_saved' mangled-name='zfs_send_saved' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_send_saved'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='8def7735' name='flags'/>
	<parameter type-id='95e97e5e' name='outfd'/>
	<parameter type-id='80f4b756' name='resume_token'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_send' mangled-name='zfs_send' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_send'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='fromsnap'/>
	<parameter type-id='80f4b756' name='tosnap'/>
	<parameter type-id='8def7735' name='flags'/>
	<parameter type-id='95e97e5e' name='outfd'/>
	<parameter type-id='72a26210' name='filter_func'/>
	<parameter type-id='eaa32e2f' name='cb_arg'/>
	<parameter type-id='857bb57e' name='debugnvp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_send_one' mangled-name='zfs_send_one' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_send_one'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='from'/>
	<parameter type-id='95e97e5e' name='fd'/>
	<parameter type-id='8def7735' name='flags'/>
	<parameter type-id='80f4b756' name='redactbook'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_receive' mangled-name='zfs_receive' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_receive'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='tosnap'/>
	<parameter type-id='5ce45b60' name='props'/>
	<parameter type-id='4ea84b4f' name='flags'/>
	<parameter type-id='95e97e5e' name='infd'/>
	<parameter type-id='a3681dea' name='stream_avl'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-type size-in-bits='64' id='c70fa2e8'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='d2a5e211'>
	<parameter type-id='9200a744'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='c19b74c3'/>
	</function-type>
	<function-type size-in-bits='64' id='f1abb096'>
	<parameter type-id='eabacd01'/>
	<return type-id='48b5725f'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_status.c' language='LANG_C99'>
	<function-decl name='zpool_import_status' mangled-name='zpool_import_status' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_import_status'>
	<parameter type-id='5ce45b60' name='config'/>
	<parameter type-id='7d3cd834' name='msgid'/>
	<parameter type-id='cec6f2e4' name='errata'/>
	<return type-id='d3dd6294'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/libzfs_util.c' language='LANG_C99'>
	<class-decl name='__va_list_tag' size-in-bits='192' is-struct='yes' visibility='default' id='d5027220'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='gp_offset' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='fp_offset' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='overflow_arg_area' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='reg_save_area' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<type-decl name='double' size-in-bits='64' id='a0eb0f08'/>
	<array-type-def dimensions='1' type-id='95e97e5e' size-in-bits='192' id='e41bdf22'>
	<subrange length='6' type-id='7359adad' id='52fa524b'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='19cefcee' size-in-bits='160' alignment-in-bits='32' id='3fcf57d2'>
	<subrange length='5' type-id='7359adad' id='53010e10'/>
	</array-type-def>
	<enum-decl name='zfs_get_column_t' naming-typedef-id='19cefcee' id='223bdcaa'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='GET_COL_NONE' value='0'/>
	<enumerator name='GET_COL_NAME' value='1'/>
	<enumerator name='GET_COL_PROPERTY' value='2'/>
	<enumerator name='GET_COL_VALUE' value='3'/>
	<enumerator name='GET_COL_RECVD' value='4'/>
	<enumerator name='GET_COL_SOURCE' value='5'/>
	</enum-decl>
	<typedef-decl name='zfs_get_column_t' type-id='223bdcaa' id='19cefcee'/>
	<class-decl name='vdev_cbdata' size-in-bits='192' is-struct='yes' visibility='default' id='b8006be8'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='cb_name_flags' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='cb_names' type-id='9b23c9ad' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='cb_names_count' type-id='f0981eeb' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='vdev_cbdata_t' type-id='b8006be8' id='a9679c94'/>
	<class-decl name='zprop_get_cbdata' size-in-bits='832' is-struct='yes' visibility='default' id='f3d3c319'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='cb_sources' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='cb_columns' type-id='3fcf57d2' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='cb_colwidths' type-id='e41bdf22' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='cb_scripted' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='416'>
	<var-decl name='cb_literal' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='cb_first' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='cb_proplist' type-id='3a9b2288' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='576'>
	<var-decl name='cb_type' type-id='2e45de5d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='640'>
	<var-decl name='cb_vdevs' type-id='a9679c94' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zprop_get_cbdata_t' type-id='f3d3c319' id='f3d87113'/>
	<typedef-decl name='zprop_func' type-id='2e711a2a' id='1ec3747a'/>
	<enum-decl name='zprop_attr_t' naming-typedef-id='999701cc' id='77d05200'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='PROP_DEFAULT' value='0'/>
	<enumerator name='PROP_READONLY' value='1'/>
	<enumerator name='PROP_INHERIT' value='2'/>
	<enumerator name='PROP_ONETIME' value='3'/>
	<enumerator name='PROP_ONETIME_DEFAULT' value='4'/>
	</enum-decl>
	<typedef-decl name='zprop_attr_t' type-id='77d05200' id='999701cc'/>
	<class-decl name='zfs_index' size-in-bits='128' is-struct='yes' visibility='default' id='87957af9'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='pi_name' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='pi_value' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zprop_index_t' type-id='87957af9' id='64636ce3'/>
	<class-decl name='zprop_desc_t' size-in-bits='640' is-struct='yes' naming-typedef-id='ffa52b96' visibility='default' id='bbff5e4b'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='pd_name' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='pd_propnum' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='pd_proptype' type-id='31429eff' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='pd_strdefault' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='pd_numdefault' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='pd_attr' type-id='999701cc' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='pd_types' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='pd_values' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='pd_colname' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='pd_rightalign' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='449'>
	<var-decl name='pd_visible' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='450'>
	<var-decl name='pd_zfs_mod_supported' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='451'>
	<var-decl name='pd_always_flex' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='pd_table' type-id='c8bc397b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='576'>
	<var-decl name='pd_table_size' type-id='b59d7dce' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zprop_desc_t' type-id='bbff5e4b' id='ffa52b96'/>
	<class-decl name='extmnttab' size-in-bits='320' is-struct='yes' visibility='default' id='0c544dc0'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='mnt_special' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='mnt_mountp' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='mnt_fstype' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='mnt_mntopts' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='mnt_major' type-id='3502e3ff' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='mnt_minor' type-id='3502e3ff' visibility='default'/>
	</data-member>
	</class-decl>
	<pointer-type-def type-id='d5027220' size-in-bits='64' id='b7f2d5e6'/>
	<qualified-type-def type-id='26a90f95' const='yes' id='57de658a'/>
	<pointer-type-def type-id='57de658a' size-in-bits='64' id='f319fae0'/>
	<pointer-type-def type-id='9b23c9ad' size-in-bits='64' id='c0563f85'/>
	<qualified-type-def type-id='33f57a65' const='yes' id='21fd6035'/>
	<pointer-type-def type-id='21fd6035' size-in-bits='64' id='a0de50cd'/>
	<pointer-type-def type-id='a0de50cd' size-in-bits='64' id='24f95ba5'/>
	<qualified-type-def type-id='64636ce3' const='yes' id='072f7953'/>
	<pointer-type-def type-id='072f7953' size-in-bits='64' id='c8bc397b'/>
	<pointer-type-def type-id='0c544dc0' size-in-bits='64' id='394fc496'/>
	<pointer-type-def type-id='aca3bac8' size-in-bits='64' id='d33f11cb'/>
	<qualified-type-def type-id='d33f11cb' restrict='yes' id='5c53ba29'/>
	<pointer-type-def type-id='ffa52b96' size-in-bits='64' id='76c8174b'/>
	<pointer-type-def type-id='f3d87113' size-in-bits='64' id='0d2a0670'/>
	<function-decl name='zpool_label_disk' mangled-name='zpool_label_disk' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_label_disk'>
	<parameter type-id='b0382bb3'/>
	<parameter type-id='4c81de99'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_version_kernel' mangled-name='zfs_version_kernel' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_version_kernel'>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='libzfs_core_init' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='libzfs_core_fini' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_get_underlying_path' mangled-name='zfs_get_underlying_path' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_underlying_path'>
	<parameter type-id='80f4b756'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='zpool_prop_unsupported' mangled-name='zpool_prop_unsupported' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_unsupported'>
	<parameter type-id='80f4b756'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zpool_feature_init' mangled-name='zpool_feature_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_feature_init'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fletcher_4_init' mangled-name='fletcher_4_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_init'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fletcher_4_fini' mangled-name='fletcher_4_fini' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_fini'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_prop_init' mangled-name='zfs_prop_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_init'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_prop_get_table' mangled-name='zfs_prop_get_table' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_get_table'>
	<return type-id='76c8174b'/>
	</function-decl>
	<function-decl name='zpool_prop_init' mangled-name='zpool_prop_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_init'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zpool_prop_get_table' mangled-name='zpool_prop_get_table' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_get_table'>
	<return type-id='76c8174b'/>
	</function-decl>
	<function-decl name='vdev_prop_init' mangled-name='vdev_prop_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_init'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zprop_iter_common' mangled-name='zprop_iter_common' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_iter_common'>
	<parameter type-id='1ec3747a'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='c19b74c3'/>
	<parameter type-id='c19b74c3'/>
	<parameter type-id='2e45de5d'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zprop_name_to_prop' mangled-name='zprop_name_to_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_name_to_prop'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='2e45de5d'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zprop_string_to_index' mangled-name='zprop_string_to_index' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_string_to_index'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='5d6479ae'/>
	<parameter type-id='2e45de5d'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zprop_values' mangled-name='zprop_values' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_values'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='2e45de5d'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zprop_width' mangled-name='zprop_width' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_width'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='37e3bd22'/>
	<parameter type-id='2e45de5d'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='zprop_valid_for_type' mangled-name='zprop_valid_for_type' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_valid_for_type'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='2e45de5d'/>
	<parameter type-id='c19b74c3'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='getextmntent' mangled-name='getextmntent' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='getextmntent'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='394fc496'/>
	<parameter type-id='62f7a03d'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='__ctype_toupper_loc' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='24f95ba5'/>
	</function-decl>
	<function-decl name='dlclose' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='regcomp' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5c53ba29'/>
	<parameter type-id='9d26089a'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='regfree' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='d33f11cb'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='puts' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='strtod' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9d26089a'/>
	<parameter type-id='8c85230f'/>
	<return type-id='a0eb0f08'/>
	</function-decl>
	<function-decl name='realloc' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='exit' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='strnlen' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='strncasecmp' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='access' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='dup2' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='execve' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='f319fae0'/>
	<parameter type-id='f319fae0'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='execv' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='f319fae0'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='execvp' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='f319fae0'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='execvpe' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='f319fae0'/>
	<parameter type-id='f319fae0'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='_exit' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fork' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='3629bad8'/>
	</function-decl>
	<function-decl name='pow' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='a0eb0f08'/>
	<parameter type-id='a0eb0f08'/>
	<return type-id='a0eb0f08'/>
	</function-decl>
	<function-decl name='__vfprintf_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='e75a27e9'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='9d26089a'/>
	<parameter type-id='b7f2d5e6'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='__vasprintf_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='8c85230f'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='9d26089a'/>
	<parameter type-id='b7f2d5e6'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='waitpid' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='3629bad8'/>
	<parameter type-id='7292109c'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='3629bad8'/>
	</function-decl>
	<function-decl name='namespace_clear' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b0382bb3'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='libzfs_load_module' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='libzfs_errno' mangled-name='libzfs_errno' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_errno'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='libzfs_error_action' mangled-name='libzfs_error_action' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_error_action'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='libzfs_error_description' mangled-name='libzfs_error_description' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_error_description'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='libzfs_print_on_error' mangled-name='libzfs_print_on_error' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_print_on_error'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='c19b74c3' name='printerr'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='libzfs_run_process' mangled-name='libzfs_run_process' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_run_process'>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='9b23c9ad' name='argv'/>
	<parameter type-id='95e97e5e' name='flags'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='libzfs_run_process_get_stdout' mangled-name='libzfs_run_process_get_stdout' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_run_process_get_stdout'>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='9b23c9ad' name='argv'/>
	<parameter type-id='9b23c9ad' name='env'/>
	<parameter type-id='c0563f85' name='lines'/>
	<parameter type-id='7292109c' name='lines_cnt'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='libzfs_run_process_get_stdout_nopath' mangled-name='libzfs_run_process_get_stdout_nopath' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_run_process_get_stdout_nopath'>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='9b23c9ad' name='argv'/>
	<parameter type-id='9b23c9ad' name='env'/>
	<parameter type-id='c0563f85' name='lines'/>
	<parameter type-id='7292109c' name='lines_cnt'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='libzfs_free_str_array' mangled-name='libzfs_free_str_array' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_free_str_array'>
	<parameter type-id='9b23c9ad' name='strs'/>
	<parameter type-id='95e97e5e' name='count'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='libzfs_init' mangled-name='libzfs_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_init'>
	<return type-id='b0382bb3'/>
	</function-decl>
	<function-decl name='libzfs_fini' mangled-name='libzfs_fini' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_fini'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_path_to_zhandle' mangled-name='zfs_path_to_zhandle' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_path_to_zhandle'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='2e45de5d' name='argtype'/>
	<return type-id='9200a744'/>
	</function-decl>
	<function-decl name='zprop_print_one_property' mangled-name='zprop_print_one_property' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_print_one_property'>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='0d2a0670' name='cbp'/>
	<parameter type-id='80f4b756' name='propname'/>
	<parameter type-id='80f4b756' name='value'/>
	<parameter type-id='a2256d42' name='sourcetype'/>
	<parameter type-id='80f4b756' name='source'/>
	<parameter type-id='80f4b756' name='recvd_value'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zprop_get_list' mangled-name='zprop_get_list' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_get_list'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='26a90f95' name='props'/>
	<parameter type-id='e4378506' name='listp'/>
	<parameter type-id='2e45de5d' name='type'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zprop_free_list' mangled-name='zprop_free_list' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_free_list'>
	<parameter type-id='3a9b2288' name='pl'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zprop_iter' mangled-name='zprop_iter' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_iter'>
	<parameter type-id='1ec3747a' name='func'/>
	<parameter type-id='eaa32e2f' name='cb'/>
	<parameter type-id='c19b74c3' name='show_all'/>
	<parameter type-id='c19b74c3' name='ordered'/>
	<parameter type-id='2e45de5d' name='type'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_version_userland' mangled-name='zfs_version_userland' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_version_userland'>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_version_print' mangled-name='zfs_version_print' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_version_print'>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='use_color' mangled-name='use_color' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='use_color'>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='printf_color' mangled-name='printf_color' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='printf_color'>
	<parameter type-id='80f4b756' name='color'/>
	<parameter type-id='80f4b756' name='format'/>
	<parameter is-variadic='yes'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_vdev_script_alloc_env' mangled-name='zpool_vdev_script_alloc_env' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_script_alloc_env'>
	<parameter type-id='80f4b756' name='pool_name'/>
	<parameter type-id='80f4b756' name='vdev_path'/>
	<parameter type-id='80f4b756' name='vdev_upath'/>
	<parameter type-id='80f4b756' name='vdev_enc_sysfs_path'/>
	<parameter type-id='80f4b756' name='opt_key'/>
	<parameter type-id='80f4b756' name='opt_val'/>
	<return type-id='9b23c9ad'/>
	</function-decl>
	<function-decl name='zpool_vdev_script_free_env' mangled-name='zpool_vdev_script_free_env' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_vdev_script_free_env'>
	<parameter type-id='9b23c9ad' name='env'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zpool_prepare_disk' mangled-name='zpool_prepare_disk' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prepare_disk'>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='5ce45b60' name='vdev_nv'/>
	<parameter type-id='80f4b756' name='prepare_str'/>
	<parameter type-id='c0563f85' name='lines'/>
	<parameter type-id='7292109c' name='lines_cnt'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_prepare_and_label_disk' mangled-name='zpool_prepare_and_label_disk' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prepare_and_label_disk'>
	<parameter type-id='b0382bb3' name='hdl'/>
	<parameter type-id='4c81de99' name='zhp'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='5ce45b60' name='vdev_nv'/>
	<parameter type-id='80f4b756' name='prepare_str'/>
	<parameter type-id='c0563f85' name='lines'/>
	<parameter type-id='7292109c' name='lines_cnt'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/os/linux/libzfs_mount_os.c' language='LANG_C99'>
	<pointer-type-def type-id='7359adad' size-in-bits='64' id='1d2c2b85'/>
	<function-decl name='geteuid' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='cc5fcceb'/>
	</function-decl>
	<function-decl name='mount' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='7359adad'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='umount2' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_parse_mount_options' mangled-name='zfs_parse_mount_options' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_parse_mount_options'>
	<parameter type-id='80f4b756' name='mntopts'/>
	<parameter type-id='1d2c2b85' name='mntflags'/>
	<parameter type-id='1d2c2b85' name='zfsflags'/>
	<parameter type-id='95e97e5e' name='sloppy'/>
	<parameter type-id='26a90f95' name='badopt'/>
	<parameter type-id='26a90f95' name='mtabopt'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_adjust_mount_options' mangled-name='zfs_adjust_mount_options' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_adjust_mount_options'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='mntpoint'/>
	<parameter type-id='26a90f95' name='mntopts'/>
	<parameter type-id='26a90f95' name='mtabopt'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_mount_delegation_check' mangled-name='zfs_mount_delegation_check' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_mount_delegation_check'>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/os/linux/libzfs_pool_os.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='288' id='16e6f2c6'>
	<subrange length='36' type-id='7359adad' id='ae666bde'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a65ae39c' size-in-bits='960' id='fa198beb'>
	<subrange length='1' type-id='7359adad' id='52f813b4'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='3502e3ff' size-in-bits='384' id='dba89ba3'>
	<subrange length='12' type-id='7359adad' id='84827bdc'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='3502e3ff' size-in-bits='256' id='01d84ed4'>
	<subrange length='8' type-id='7359adad' id='56e0c0b1'/>
	</array-type-def>
	<class-decl name='dk_part' size-in-bits='960' is-struct='yes' visibility='default' id='a65ae39c'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='p_start' type-id='804dc465' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='p_size' type-id='804dc465' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='p_guid' type-id='214f32ea' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='p_tag' type-id='d908a348' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='272'>
	<var-decl name='p_flag' type-id='d908a348' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='p_name' type-id='16e6f2c6' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='576'>
	<var-decl name='p_uguid' type-id='214f32ea' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='704'>
	<var-decl name='p_resv' type-id='01d84ed4' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='dk_gpt' size-in-bits='1920' is-struct='yes' visibility='default' id='dd4a2e5a'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='efi_version' type-id='3502e3ff' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='efi_nparts' type-id='3502e3ff' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='efi_part_size' type-id='3502e3ff' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='efi_lbasize' type-id='3502e3ff' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='efi_last_lba' type-id='804dc465' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='efi_first_u_lba' type-id='804dc465' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='efi_last_u_lba' type-id='804dc465' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='efi_disk_uguid' type-id='214f32ea' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='efi_flags' type-id='3502e3ff' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='480'>
	<var-decl name='efi_reserved1' type-id='3502e3ff' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='efi_altern_lba' type-id='804dc465' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='576'>
	<var-decl name='efi_reserved' type-id='dba89ba3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='960'>
	<var-decl name='efi_parts' type-id='fa198beb' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='uuid' size-in-bits='128' is-struct='yes' visibility='default' id='214f32ea'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='time_low' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='time_mid' type-id='149c6638' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='48'>
	<var-decl name='time_hi_and_version' type-id='149c6638' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='clock_seq_hi_and_reserved' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='72'>
	<var-decl name='clock_seq_low' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='80'>
	<var-decl name='node_addr' type-id='0f562bd0' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='ushort_t' type-id='8efea9e5' id='d908a348'/>
	<typedef-decl name='uint16_t' type-id='253c2d2a' id='149c6638'/>
	<typedef-decl name='__uint16_t' type-id='8efea9e5' id='253c2d2a'/>
	<pointer-type-def type-id='dd4a2e5a' size-in-bits='64' id='0d8119a8'/>
	<pointer-type-def type-id='0d8119a8' size-in-bits='64' id='c43b27a6'/>
	<function-decl name='zpool_label_disk_wait' mangled-name='zpool_label_disk_wait' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_label_disk_wait'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_append_partition' mangled-name='zfs_append_partition' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_append_partition'>
	<parameter type-id='26a90f95'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='efi_alloc_and_init' mangled-name='efi_alloc_and_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='efi_alloc_and_init'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='8f92235e'/>
	<parameter type-id='c43b27a6'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='efi_alloc_and_read' mangled-name='efi_alloc_and_read' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='efi_alloc_and_read'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='c43b27a6'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='efi_write' mangled-name='efi_write' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='efi_write'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='0d8119a8'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='efi_rescan' mangled-name='efi_rescan' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='efi_rescan'>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='efi_free' mangled-name='efi_free' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='efi_free'>
	<parameter type-id='0d8119a8'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='efi_use_whole_disk' mangled-name='efi_use_whole_disk' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='efi_use_whole_disk'>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='rand' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='fsync' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs/os/linux/libzfs_util_os.c' language='LANG_C99'>
	<typedef-decl name='nfds_t' type-id='7359adad' id='555eef66'/>
	<class-decl name='pollfd' size-in-bits='64' is-struct='yes' visibility='default' id='b440e872'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='fd' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='events' type-id='a2185560' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='48'>
	<var-decl name='revents' type-id='a2185560' visibility='default'/>
	</data-member>
	</class-decl>
	<pointer-type-def type-id='b440e872' size-in-bits='64' id='3ac36db0'/>
	<function-decl name='__poll_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='3ac36db0'/>
	<parameter type-id='555eef66'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='7359adad'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='inotify_init1' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='inotify_add_watch' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='8f92235e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='timerfd_create' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='08f9a87a'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='timerfd_settime' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='f39579e7'/>
	<parameter type-id='116842ac'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='libzfs_error_init' mangled-name='libzfs_error_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_error_init'>
	<parameter type-id='95e97e5e' name='error'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_userns' mangled-name='zfs_userns' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_userns'>
	<parameter type-id='9200a744' name='zhp'/>
	<parameter type-id='80f4b756' name='nspath'/>
	<parameter type-id='95e97e5e' name='attach'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzutil/os/linux/zutil_device_path_os.c' language='LANG_C99'>
	<class-decl name='udev' is-struct='yes' visibility='default' is-declaration-only='yes' id='e4a7fb7f'/>
	<class-decl name='udev_device' is-struct='yes' visibility='default' is-declaration-only='yes' id='640b33ca'/>
	<pointer-type-def type-id='e4a7fb7f' size-in-bits='64' id='025eefe7'/>
	<pointer-type-def type-id='640b33ca' size-in-bits='64' id='b32bae08'/>
	<class-decl name='udev' is-struct='yes' visibility='default' is-declaration-only='yes' id='e4a7fb7f'/>
	<class-decl name='udev_device' is-struct='yes' visibility='default' is-declaration-only='yes' id='640b33ca'/>
	<function-decl name='udev_new' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='025eefe7'/>
	</function-decl>
	<function-decl name='udev_device_unref' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b32bae08'/>
	<return type-id='b32bae08'/>
	</function-decl>
	<function-decl name='udev_device_new_from_subsystem_sysname' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='025eefe7'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='b32bae08'/>
	</function-decl>
	<function-decl name='udev_device_get_property_value' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b32bae08'/>
	<parameter type-id='80f4b756'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='__readlink_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9d26089a'/>
	<parameter type-id='266fe297'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='79a0948f'/>
	</function-decl>
	<function-decl name='zfs_get_enclosure_sysfs_path' mangled-name='zfs_get_enclosure_sysfs_path' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_get_enclosure_sysfs_path'>
	<parameter type-id='80f4b756' name='dev_name'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='zfs_dev_is_dm' mangled-name='zfs_dev_is_dm' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_dev_is_dm'>
	<parameter type-id='80f4b756' name='dev_name'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_dev_is_whole_disk' mangled-name='zfs_dev_is_whole_disk' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_dev_is_whole_disk'>
	<parameter type-id='80f4b756' name='dev_name'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='is_mpath_whole_disk' mangled-name='is_mpath_whole_disk' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='is_mpath_whole_disk'>
	<parameter type-id='80f4b756' name='path'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzutil/os/linux/zutil_import_os.c' language='LANG_C99'>
	<class-decl name='blkid_struct_cache' is-struct='yes' visibility='default' is-declaration-only='yes' id='09286066'/>
	<class-decl name='blkid_struct_dev' is-struct='yes' visibility='default' is-declaration-only='yes' id='86223623'/>
	<class-decl name='blkid_struct_dev_iterate' is-struct='yes' visibility='default' is-declaration-only='yes' id='d88420d6'/>
	<class-decl name='udev_list_entry' is-struct='yes' visibility='default' is-declaration-only='yes' id='e7dbdca3'/>
	<typedef-decl name='pool_vdev_iter_f' type-id='6c16a6c8' id='dff793e0'/>
	<typedef-decl name='blkid_dev' type-id='8433f053' id='f47b023a'/>
	<typedef-decl name='blkid_cache' type-id='940e3afc' id='0882dfdf'/>
	<typedef-decl name='blkid_dev_iterate' type-id='b8fa2efc' id='f4760fa7'/>
	<typedef-decl name='__useconds_t' type-id='f0981eeb' id='4e80d4b1'/>
	<pointer-type-def type-id='0882dfdf' size-in-bits='64' id='2e3e7caa'/>
	<pointer-type-def type-id='f47b023a' size-in-bits='64' id='d87f9b75'/>
	<pointer-type-def type-id='09286066' size-in-bits='64' id='940e3afc'/>
	<pointer-type-def type-id='86223623' size-in-bits='64' id='8433f053'/>
	<pointer-type-def type-id='d88420d6' size-in-bits='64' id='b8fa2efc'/>
	<pointer-type-def type-id='2ec2411e' size-in-bits='64' id='6c16a6c8'/>
	<pointer-type-def type-id='e7dbdca3' size-in-bits='64' id='deabd0d3'/>
	<class-decl name='blkid_struct_cache' is-struct='yes' visibility='default' is-declaration-only='yes' id='09286066'/>
	<class-decl name='blkid_struct_dev' is-struct='yes' visibility='default' is-declaration-only='yes' id='86223623'/>
	<class-decl name='blkid_struct_dev_iterate' is-struct='yes' visibility='default' is-declaration-only='yes' id='d88420d6'/>
	<class-decl name='udev_list_entry' is-struct='yes' visibility='default' is-declaration-only='yes' id='e7dbdca3'/>
	<function-decl name='for_each_vdev_in_nvlist' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='dff793e0'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='label_paths' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5507783b'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='7d3cd834'/>
	<parameter type-id='7d3cd834'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zutil_alloc' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5507783b'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='zutil_strdup' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5507783b'/>
	<parameter type-id='80f4b756'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='slice_cache_compare' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='blkid_put_cache' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='0882dfdf'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='blkid_get_cache' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='2e3e7caa'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='blkid_dev_devname' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='f47b023a'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='blkid_dev_iterate_begin' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='0882dfdf'/>
	<return type-id='f4760fa7'/>
	</function-decl>
	<function-decl name='blkid_dev_set_search' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='f4760fa7'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='blkid_dev_next' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='f4760fa7'/>
	<parameter type-id='d87f9b75'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='blkid_dev_iterate_end' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='f4760fa7'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='blkid_probe_all_new' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='0882dfdf'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='udev_unref' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='025eefe7'/>
	<return type-id='025eefe7'/>
	</function-decl>
	<function-decl name='udev_list_entry_get_next' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='deabd0d3'/>
	<return type-id='deabd0d3'/>
	</function-decl>
	<function-decl name='udev_list_entry_get_name' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='deabd0d3'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='udev_device_get_parent_with_subsystem_devtype' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b32bae08'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='b32bae08'/>
	</function-decl>
	<function-decl name='udev_device_get_devlinks_list_entry' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b32bae08'/>
	<return type-id='deabd0d3'/>
	</function-decl>
	<function-decl name='sched_yield' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='usleep' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='4e80d4b1'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_dev_flush' mangled-name='zfs_dev_flush' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_dev_flush'>
	<parameter type-id='95e97e5e' name='fd'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_device_get_devid' mangled-name='zfs_device_get_devid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_device_get_devid'>
	<parameter type-id='b32bae08' name='dev'/>
	<parameter type-id='26a90f95' name='bufptr'/>
	<parameter type-id='b59d7dce' name='buflen'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_device_get_physical' mangled-name='zfs_device_get_physical' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_device_get_physical'>
	<parameter type-id='b32bae08' name='dev'/>
	<parameter type-id='26a90f95' name='bufptr'/>
	<parameter type-id='b59d7dce' name='buflen'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_disk_wait' mangled-name='zpool_disk_wait' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_disk_wait'>
	<parameter type-id='80f4b756' name='path'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='update_vdev_config_dev_sysfs_path' mangled-name='update_vdev_config_dev_sysfs_path' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='update_vdev_config_dev_sysfs_path'>
	<parameter type-id='5ce45b60' name='nv'/>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='80f4b756' name='key'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-type size-in-bits='64' id='2ec2411e'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzutil/os/linux/zutil_setproctitle.c' language='LANG_C99'>
	<function-decl name='warnx' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter is-variadic='yes'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='setenv' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='clearenv' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_setproctitle_init' mangled-name='zfs_setproctitle_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_setproctitle_init'>
	<parameter type-id='95e97e5e' name='argc'/>
	<parameter type-id='9b23c9ad' name='argv'/>
	<parameter type-id='9b23c9ad' name='envp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzutil/zutil_device_path.c' language='LANG_C99'>
	<pointer-type-def type-id='b99c00c9' size-in-bits='64' id='13956559'/>
	<function-decl name='zpool_default_search_paths' mangled-name='zpool_default_search_paths' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_default_search_paths'>
	<parameter type-id='78c01427'/>
	<return type-id='13956559'/>
	</function-decl>
	<function-decl name='strspn' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='zfs_dirnamelen' mangled-name='zfs_dirnamelen' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_dirnamelen'>
	<parameter type-id='80f4b756' name='path'/>
	<return type-id='79a0948f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzutil/zutil_import.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='256' id='16dc656a'>
	<subrange length='32' type-id='7359adad' id='ae5bde82'/>
	</array-type-def>
	- <class-decl name='importargs' size-in-bits='448' is-struct='yes' visibility='default' id='7ac83801'>
	+ <class-decl name='importargs' size-in-bits='512' is-struct='yes' visibility='default' id='7ac83801'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='path' type-id='9b23c9ad' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='paths' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='poolname' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='guid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='cachefile' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='can_be_active' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='352'>
	<var-decl name='scan' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='policy' type-id='5ce45b60' visibility='default'/>
	</data-member>
	+ <data-member access='public' layout-offset-in-bits='448'>
	+ <var-decl name='do_destroyed' type-id='c19b74c3' visibility='default'/>
	+ </data-member>
	+ <data-member access='public' layout-offset-in-bits='480'>
	+ <var-decl name='do_all' type-id='c19b74c3' visibility='default'/>
	+ </data-member>
	</class-decl>
	<typedef-decl name='importargs_t' type-id='7ac83801' id='7a842a6b'/>
	<class-decl name='libpc_handle' size-in-bits='8448' is-struct='yes' visibility='default' id='7c8737f0'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='lpc_error' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='lpc_printerr' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='lpc_open_access_error' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='lpc_desc_active' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='lpc_desc' type-id='b54ce520' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='8320'>
	<var-decl name='lpc_ops' type-id='f095e320' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='8384'>
	<var-decl name='lpc_lib_handle' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='libpc_handle_t' type-id='7c8737f0' id='8a70a786'/>
	<class-decl name='aiocb' size-in-bits='1344' is-struct='yes' visibility='default' id='e4957c49'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='aio_fildes' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='aio_lio_opcode' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='aio_reqprio' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='aio_buf' type-id='fe09dd29' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='aio_nbytes' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='aio_sigevent' type-id='519bc206' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='768'>
	<var-decl name='__next_prio' type-id='924bbc81' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='832'>
	<var-decl name='__abs_prio' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='864'>
	<var-decl name='__policy' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='896'>
	<var-decl name='__error_code' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='960'>
	<var-decl name='__return_value' type-id='41060289' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1024'>
	<var-decl name='aio_offset' type-id='724e4de6' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1088'>
	<var-decl name='__glibc_reserved' type-id='16dc656a' visibility='default'/>
	</data-member>
	</class-decl>
	<pointer-type-def type-id='e4957c49' size-in-bits='64' id='924bbc81'/>
	<qualified-type-def type-id='924bbc81' const='yes' id='5499dcde'/>
	<pointer-type-def type-id='5499dcde' size-in-bits='64' id='2236d41c'/>
	<qualified-type-def type-id='2236d41c' restrict='yes' id='31488924'/>
	<pointer-type-def type-id='a3681dea' size-in-bits='64' id='fce6d540'/>
	<qualified-type-def type-id='e4957c49' const='yes' id='fced9da2'/>
	<pointer-type-def type-id='fced9da2' size-in-bits='64' id='b20efd18'/>
	<pointer-type-def type-id='7a842a6b' size-in-bits='64' id='07ee4a58'/>
	<pointer-type-def type-id='8a70a786' size-in-bits='64' id='5507783b'/>
	<pointer-type-def type-id='b1e62775' size-in-bits='64' id='f095e320'/>
	<qualified-type-def type-id='48b5725f' volatile='yes' id='b0b3cbf9'/>
	<pointer-type-def type-id='b0b3cbf9' size-in-bits='64' id='fe09dd29'/>
	<function-decl name='update_vdev_config_dev_strs' mangled-name='update_vdev_config_dev_strs' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='update_vdev_config_dev_strs'>
	<parameter type-id='5ce45b60'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='update_vdevs_config_dev_sysfs_path' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_dup' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='22cce67b'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='spl_pagesize' mangled-name='spl_pagesize' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='spl_pagesize'>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='posix_memalign' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='63e171df'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='sysconf' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<return type-id='bd54fe1a'/>
	</function-decl>
	<function-decl name='libpc_error_description' mangled-name='libpc_error_description' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libpc_error_description'>
	<parameter type-id='5507783b' name='hdl'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zpool_search_import' mangled-name='zpool_search_import' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_search_import'>
	<parameter type-id='5507783b' name='hdl'/>
	<parameter type-id='07ee4a58' name='import'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='zpool_find_config' mangled-name='zpool_find_config' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_find_config'>
	<parameter type-id='5507783b' name='hdl'/>
	<parameter type-id='80f4b756' name='target'/>
	<parameter type-id='857bb57e' name='configp'/>
	<parameter type-id='07ee4a58' name='args'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_find_import_blkid' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5507783b'/>
	<parameter type-id='18c91f9e'/>
	<parameter type-id='fce6d540'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_open_func' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='eaa32e2f'/>
	<return type-id='48b5725f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzutil/zutil_nicenum.c' language='LANG_C99'>
	<type-decl name='long double' size-in-bits='128' id='e095c704'/>
	<enum-decl name='zfs_nicenum_format' id='29cf1969'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZFS_NICENUM_1024' value='0'/>
	<enumerator name='ZFS_NICENUM_BYTES' value='1'/>
	<enumerator name='ZFS_NICENUM_TIME' value='2'/>
	<enumerator name='ZFS_NICENUM_RAW' value='3'/>
	<enumerator name='ZFS_NICENUM_RAWTIME' value='4'/>
	</enum-decl>
	<function-decl name='powl' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='e095c704'/>
	<parameter type-id='e095c704'/>
	<return type-id='e095c704'/>
	</function-decl>
	<function-decl name='floor' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='a0eb0f08'/>
	<return type-id='a0eb0f08'/>
	</function-decl>
	<function-decl name='zfs_isnumber' mangled-name='zfs_isnumber' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_isnumber'>
	<parameter type-id='80f4b756' name='str'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_nicenum_format' mangled-name='zfs_nicenum_format' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_nicenum_format'>
	<parameter type-id='9c313c2d' name='num'/>
	<parameter type-id='26a90f95' name='buf'/>
	<parameter type-id='b59d7dce' name='buflen'/>
	<parameter type-id='29cf1969' name='format'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_nicetime' mangled-name='zfs_nicetime' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_nicetime'>
	<parameter type-id='9c313c2d' name='num'/>
	<parameter type-id='26a90f95' name='buf'/>
	<parameter type-id='b59d7dce' name='buflen'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfs_niceraw' mangled-name='zfs_niceraw' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_niceraw'>
	<parameter type-id='9c313c2d' name='num'/>
	<parameter type-id='26a90f95' name='buf'/>
	<parameter type-id='b59d7dce' name='buflen'/>
	<return type-id='48b5725f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzutil/zutil_pool.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='853fd5dc' size-in-bits='32768' id='b505fc2f'>
	<subrange length='64' type-id='7359adad' id='b10be967'/>
	</array-type-def>
	<type-decl name='float' size-in-bits='32' id='a6c45d85'/>
	<class-decl name='ddt_stat' size-in-bits='512' is-struct='yes' visibility='default' id='65242dfe'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='dds_blocks' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='dds_lsize' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='dds_psize' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='dds_dsize' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='dds_ref_blocks' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='dds_ref_lsize' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='dds_ref_psize' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='dds_ref_dsize' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='ddt_stat_t' type-id='65242dfe' id='853fd5dc'/>
	<class-decl name='ddt_histogram' size-in-bits='32768' is-struct='yes' visibility='default' id='bc2b3086'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='ddh_stat' type-id='b505fc2f' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='ddt_histogram_t' type-id='bc2b3086' id='2d7fe832'/>
	<qualified-type-def type-id='2d7fe832' const='yes' id='ec92d602'/>
	<pointer-type-def type-id='ec92d602' size-in-bits='64' id='932720f8'/>
	<qualified-type-def type-id='853fd5dc' const='yes' id='764c298c'/>
	<pointer-type-def type-id='764c298c' size-in-bits='64' id='dfe59052'/>
	<qualified-type-def type-id='a9c79a1f' const='yes' id='cd087e36'/>
	<pointer-type-def type-id='cd087e36' size-in-bits='64' id='e05e8614'/>
	<function-decl name='nanosleep' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='e05e8614'/>
	<parameter type-id='3d83ba87'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_dump_ddt' mangled-name='zpool_dump_ddt' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_dump_ddt'>
	<parameter type-id='dfe59052' name='dds_total'/>
	<parameter type-id='932720f8' name='ddh'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fsleep' mangled-name='fsleep' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fsleep'>
	<parameter type-id='a6c45d85' name='sec'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zpool_getenv_int' mangled-name='zpool_getenv_int' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_getenv_int'>
	<parameter type-id='80f4b756' name='env'/>
	<parameter type-id='95e97e5e' name='default_val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='module/avl/avl.c' language='LANG_C99'>
	<function-decl name='avl_last' mangled-name='avl_last' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_last'>
	<parameter type-id='a3681dea' name='tree'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='avl_nearest' mangled-name='avl_nearest' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_nearest'>
	<parameter type-id='a3681dea' name='tree'/>
	<parameter type-id='fba6cb51' name='where'/>
	<parameter type-id='95e97e5e' name='direction'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='avl_insert_here' mangled-name='avl_insert_here' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_insert_here'>
	<parameter type-id='a3681dea' name='tree'/>
	<parameter type-id='eaa32e2f' name='new_data'/>
	<parameter type-id='eaa32e2f' name='here'/>
	<parameter type-id='95e97e5e' name='direction'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='avl_update_lt' mangled-name='avl_update_lt' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_update_lt'>
	<parameter type-id='a3681dea' name='t'/>
	<parameter type-id='eaa32e2f' name='obj'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='avl_update_gt' mangled-name='avl_update_gt' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_update_gt'>
	<parameter type-id='a3681dea' name='t'/>
	<parameter type-id='eaa32e2f' name='obj'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='avl_update' mangled-name='avl_update' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_update'>
	<parameter type-id='a3681dea' name='t'/>
	<parameter type-id='eaa32e2f' name='obj'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='avl_swap' mangled-name='avl_swap' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_swap'>
	<parameter type-id='a3681dea' name='tree1'/>
	<parameter type-id='a3681dea' name='tree2'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='avl_is_empty' mangled-name='avl_is_empty' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='avl_is_empty'>
	<parameter type-id='a3681dea' name='tree'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/cityhash.c' language='LANG_C99'>
	<function-decl name='cityhash4' mangled-name='cityhash4' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='cityhash4'>
	<parameter type-id='9c313c2d' name='w1'/>
	<parameter type-id='9c313c2d' name='w2'/>
	<parameter type-id='9c313c2d' name='w3'/>
	<parameter type-id='9c313c2d' name='w4'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zfeature_common.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='83f29ca2' size-in-bits='18368' id='b93e4d14'>
	<subrange length='41' type-id='7359adad' id='cb834f44'/>
	</array-type-def>
	<enum-decl name='zfeature_flags' id='6db816a4'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZFEATURE_FLAG_READONLY_COMPAT' value='1'/>
	<enumerator name='ZFEATURE_FLAG_MOS' value='2'/>
	<enumerator name='ZFEATURE_FLAG_ACTIVATE_ON_ENABLE' value='4'/>
	<enumerator name='ZFEATURE_FLAG_PER_DATASET' value='8'/>
	</enum-decl>
	<typedef-decl name='zfeature_flags_t' type-id='6db816a4' id='fc329033'/>
	<enum-decl name='zfeature_type' id='c4fa2355'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZFEATURE_TYPE_BOOLEAN' value='0'/>
	<enumerator name='ZFEATURE_TYPE_UINT64_ARRAY' value='1'/>
	<enumerator name='ZFEATURE_NUM_TYPES' value='2'/>
	</enum-decl>
	<typedef-decl name='zfeature_type_t' type-id='c4fa2355' id='732d2bb2'/>
	<class-decl name='zfeature_info' size-in-bits='448' is-struct='yes' visibility='default' id='1178d146'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='fi_feature' type-id='d6618c78' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='fi_uname' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='fi_guid' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='fi_desc' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='fi_flags' type-id='fc329033' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='fi_zfs_mod_supported' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='fi_type' type-id='732d2bb2' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='fi_depends' type-id='1acff326' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zfeature_info_t' type-id='1178d146' id='83f29ca2'/>
	<typedef-decl name='__free_fn_t' type-id='b7f9d8e6' id='3ff5e51e'/>
	<class-decl name='dirent' size-in-bits='2240' is-struct='yes' visibility='default' id='611586a1'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='d_ino' type-id='71288a47' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='d_off' type-id='724e4de6' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='d_reclen' type-id='8efea9e5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='144'>
	<var-decl name='d_type' type-id='002ac4a6' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='152'>
	<var-decl name='d_name' type-id='d1617432' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='zfs_mod_supported_features' size-in-bits='128' is-struct='yes' visibility='default' id='3eee3342'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='tree' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='all_features' type-id='c19b74c3' visibility='default'/>
	</data-member>
	</class-decl>
	<qualified-type-def type-id='d6618c78' const='yes' id='81a65028'/>
	<pointer-type-def type-id='81a65028' size-in-bits='64' id='1acff326'/>
	<qualified-type-def type-id='3eee3342' const='yes' id='0c1d5bbb'/>
	<pointer-type-def type-id='0c1d5bbb' size-in-bits='64' id='a3372543'/>
	<pointer-type-def type-id='611586a1' size-in-bits='64' id='2e243169'/>
	<qualified-type-def type-id='eaa32e2f' const='yes' id='83be723c'/>
	<pointer-type-def type-id='83be723c' size-in-bits='64' id='7acd98a2'/>
	<var-decl name='spa_feature_table' type-id='b93e4d14' mangled-name='spa_feature_table' visibility='default' elf-symbol-id='spa_feature_table'/>
	<var-decl name='zfeature_checks_disable' type-id='c19b74c3' mangled-name='zfeature_checks_disable' visibility='default' elf-symbol-id='zfeature_checks_disable'/>
	<function-decl name='opendir' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<return type-id='f09217ba'/>
	</function-decl>
	<function-decl name='tsearch' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='63e171df'/>
	<parameter type-id='aba7edd8'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='tfind' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='7acd98a2'/>
	<parameter type-id='aba7edd8'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='tdestroy' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='3ff5e51e'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zfeature_is_valid_guid' mangled-name='zfeature_is_valid_guid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfeature_is_valid_guid'>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfeature_depends_on' mangled-name='zfeature_depends_on' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfeature_depends_on'>
	<parameter type-id='d6618c78' name='fid'/>
	<parameter type-id='d6618c78' name='check'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_mod_supported' mangled-name='zfs_mod_supported' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_mod_supported'>
	<parameter type-id='80f4b756' name='scope'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='a3372543' name='sfeatures'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zfs_comutil.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='b99c00c9' size-in-bits='2624' id='5ce15418'>
	<subrange length='41' type-id='7359adad' id='cb834f44'/>
	</array-type-def>
	<qualified-type-def type-id='80f4b756' const='yes' id='b99c00c9'/>
	<pointer-type-def type-id='8f92235e' size-in-bits='64' id='90421557'/>
	<function-decl name='nvpair_value_uint32' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='dace003f'/>
	<parameter type-id='90421557'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<var-decl name='zfs_history_event_names' type-id='5ce15418' mangled-name='zfs_history_event_names' visibility='default' elf-symbol-id='zfs_history_event_names'/>
	<function-decl name='strpbrk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='zfs_allocatable_devs' mangled-name='zfs_allocatable_devs' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_allocatable_devs'>
	<parameter type-id='5ce45b60' name='nv'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_special_devs' mangled-name='zfs_special_devs' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_special_devs'>
	<parameter type-id='5ce45b60' name='nv'/>
	<parameter type-id='80f4b756' name='type'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_zpl_version_map' mangled-name='zfs_zpl_version_map' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_zpl_version_map'>
	<parameter type-id='95e97e5e' name='spa_version'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_spa_version_map' mangled-name='zfs_spa_version_map' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_spa_version_map'>
	<parameter type-id='95e97e5e' name='zpl_version'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_dataset_name_hidden' mangled-name='zfs_dataset_name_hidden' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_dataset_name_hidden'>
	<parameter type-id='80f4b756' name='name'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zfs_deleg.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='fa1870fd' size-in-bits='4096' id='59e94aca'>
	<subrange length='32' type-id='7359adad' id='ae5bde82'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='fa1870fd' size-in-bits='infinite' id='7c00e69d'>
	<subrange length='infinite' id='031f2035'/>
	</array-type-def>
	<enum-decl name='zfs_deleg_who_type_t' naming-typedef-id='36d4bd5a' id='b5fa5816'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZFS_DELEG_WHO_UNKNOWN' value='0'/>
	<enumerator name='ZFS_DELEG_USER' value='117'/>
	<enumerator name='ZFS_DELEG_USER_SETS' value='85'/>
	<enumerator name='ZFS_DELEG_GROUP' value='103'/>
	<enumerator name='ZFS_DELEG_GROUP_SETS' value='71'/>
	<enumerator name='ZFS_DELEG_EVERYONE' value='101'/>
	<enumerator name='ZFS_DELEG_EVERYONE_SETS' value='69'/>
	<enumerator name='ZFS_DELEG_CREATE' value='99'/>
	<enumerator name='ZFS_DELEG_CREATE_SETS' value='67'/>
	<enumerator name='ZFS_DELEG_NAMED_SET' value='115'/>
	<enumerator name='ZFS_DELEG_NAMED_SET_SETS' value='83'/>
	</enum-decl>
	<typedef-decl name='zfs_deleg_who_type_t' type-id='b5fa5816' id='36d4bd5a'/>
	<enum-decl name='zfs_deleg_note_t' naming-typedef-id='4613c173' id='729d4547'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZFS_DELEG_NOTE_CREATE' value='0'/>
	<enumerator name='ZFS_DELEG_NOTE_DESTROY' value='1'/>
	<enumerator name='ZFS_DELEG_NOTE_SNAPSHOT' value='2'/>
	<enumerator name='ZFS_DELEG_NOTE_ROLLBACK' value='3'/>
	<enumerator name='ZFS_DELEG_NOTE_CLONE' value='4'/>
	<enumerator name='ZFS_DELEG_NOTE_PROMOTE' value='5'/>
	<enumerator name='ZFS_DELEG_NOTE_RENAME' value='6'/>
	<enumerator name='ZFS_DELEG_NOTE_SEND' value='7'/>
	<enumerator name='ZFS_DELEG_NOTE_RECEIVE' value='8'/>
	<enumerator name='ZFS_DELEG_NOTE_ALLOW' value='9'/>
	<enumerator name='ZFS_DELEG_NOTE_USERPROP' value='10'/>
	<enumerator name='ZFS_DELEG_NOTE_MOUNT' value='11'/>
	<enumerator name='ZFS_DELEG_NOTE_SHARE' value='12'/>
	<enumerator name='ZFS_DELEG_NOTE_USERQUOTA' value='13'/>
	<enumerator name='ZFS_DELEG_NOTE_GROUPQUOTA' value='14'/>
	<enumerator name='ZFS_DELEG_NOTE_USERUSED' value='15'/>
	<enumerator name='ZFS_DELEG_NOTE_GROUPUSED' value='16'/>
	<enumerator name='ZFS_DELEG_NOTE_USEROBJQUOTA' value='17'/>
	<enumerator name='ZFS_DELEG_NOTE_GROUPOBJQUOTA' value='18'/>
	<enumerator name='ZFS_DELEG_NOTE_USEROBJUSED' value='19'/>
	<enumerator name='ZFS_DELEG_NOTE_GROUPOBJUSED' value='20'/>
	<enumerator name='ZFS_DELEG_NOTE_HOLD' value='21'/>
	<enumerator name='ZFS_DELEG_NOTE_RELEASE' value='22'/>
	<enumerator name='ZFS_DELEG_NOTE_DIFF' value='23'/>
	<enumerator name='ZFS_DELEG_NOTE_BOOKMARK' value='24'/>
	<enumerator name='ZFS_DELEG_NOTE_LOAD_KEY' value='25'/>
	<enumerator name='ZFS_DELEG_NOTE_CHANGE_KEY' value='26'/>
	<enumerator name='ZFS_DELEG_NOTE_PROJECTUSED' value='27'/>
	<enumerator name='ZFS_DELEG_NOTE_PROJECTQUOTA' value='28'/>
	<enumerator name='ZFS_DELEG_NOTE_PROJECTOBJUSED' value='29'/>
	<enumerator name='ZFS_DELEG_NOTE_PROJECTOBJQUOTA' value='30'/>
	<enumerator name='ZFS_DELEG_NOTE_NONE' value='31'/>
	</enum-decl>
	<typedef-decl name='zfs_deleg_note_t' type-id='729d4547' id='4613c173'/>
	<class-decl name='zfs_deleg_perm_tab' size-in-bits='128' is-struct='yes' visibility='default' id='5aa05c1f'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='z_perm' type-id='80f4b756' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='z_note' type-id='4613c173' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zfs_deleg_perm_tab_t' type-id='5aa05c1f' id='f3f851ad'/>
	<qualified-type-def type-id='f3f851ad' const='yes' id='fa1870fd'/>
	<var-decl name='zfs_deleg_perm_tab' type-id='7c00e69d' mangled-name='zfs_deleg_perm_tab' visibility='default' elf-symbol-id='zfs_deleg_perm_tab'/>
	<function-decl name='permset_namecheck' mangled-name='permset_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='permset_namecheck'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='053457bd'/>
	<parameter type-id='26a90f95'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_delegatable' mangled-name='zfs_prop_delegatable' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_delegatable'>
	<parameter type-id='58603c44'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_deleg_canonicalize_perm' mangled-name='zfs_deleg_canonicalize_perm' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_deleg_canonicalize_perm'>
	<parameter type-id='80f4b756' name='perm'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_deleg_verify_nvlist' mangled-name='zfs_deleg_verify_nvlist' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_deleg_verify_nvlist'>
	<parameter type-id='5ce45b60' name='nvp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_deleg_whokey' mangled-name='zfs_deleg_whokey' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_deleg_whokey'>
	<parameter type-id='26a90f95' name='attr'/>
	<parameter type-id='36d4bd5a' name='type'/>
	<parameter type-id='a84c031d' name='inheritchr'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='48b5725f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zfs_fletcher.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='9c313c2d' size-in-bits='512' id='c5d13f42'>
	<subrange length='8' type-id='7359adad' id='56e0c0b1'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='90dbb6d6' size-in-bits='2048' id='16582e69'>
	<subrange length='4' type-id='7359adad' id='16fe7105'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='8240361c' size-in-bits='1024' id='481f90b1'>
	<subrange length='4' type-id='7359adad' id='16fe7105'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='7c1ab40c' size-in-bits='512' id='cbd91ec1'>
	<subrange length='4' type-id='7359adad' id='16fe7105'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='6d059eaa' size-in-bits='1024' id='729b6ebb'>
	<subrange length='4' type-id='7359adad' id='16fe7105'/>
	</array-type-def>
	<enum-decl name='zio_byteorder_t' naming-typedef-id='595a65ec' id='fc861be0'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZIO_CHECKSUM_NATIVE' value='0'/>
	<enumerator name='ZIO_CHECKSUM_BYTESWAP' value='1'/>
	</enum-decl>
	<typedef-decl name='zio_byteorder_t' type-id='fc861be0' id='595a65ec'/>
	<class-decl name='zio_abd_checksum_data' size-in-bits='256' is-struct='yes' visibility='default' id='4bf4b004'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='acd_byteorder' type-id='595a65ec' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='acd_ctx' type-id='0f7df99e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='acd_zcp' type-id='c24fc2ee' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='acd_private' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zio_abd_checksum_data_t' type-id='4bf4b004' id='74e39470'/>
	<typedef-decl name='zio_abd_checksum_init_t' type-id='a5444274' id='029a8ebe'/>
	<typedef-decl name='zio_abd_checksum_fini_t' type-id='a5444274' id='d6fd5c6c'/>
	<typedef-decl name='zio_abd_checksum_iter_t' type-id='f4a1892e' id='cefa0f4a'/>
	<class-decl name='zio_abd_checksum_func' size-in-bits='192' is-struct='yes' visibility='default' id='aa14691a'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='acf_init' type-id='0bcca125' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='acf_fini' type-id='bfe36153' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='acf_iter' type-id='1e276399' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zio_abd_checksum_func_t' type-id='3f8e8d11' id='c2eb138a'/>
	<class-decl name='zfs_fletcher_superscalar' size-in-bits='256' is-struct='yes' visibility='default' id='28efb250'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='v' type-id='85c64d26' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zfs_fletcher_superscalar_t' type-id='28efb250' id='6d059eaa'/>
	<class-decl name='zfs_fletcher_sse' size-in-bits='128' is-struct='yes' visibility='default' id='acd4019a'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='v' type-id='c1c22e6c' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zfs_fletcher_sse_t' type-id='acd4019a' id='7c1ab40c'/>
	<class-decl name='zfs_fletcher_avx' size-in-bits='256' is-struct='yes' visibility='default' id='8c208dfa'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='v' type-id='85c64d26' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zfs_fletcher_avx_t' type-id='8c208dfa' id='8240361c'/>
	<class-decl name='zfs_fletcher_avx512' size-in-bits='512' is-struct='yes' visibility='default' id='c6d0c382'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='v' type-id='c5d13f42' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zfs_fletcher_avx512_t' type-id='c6d0c382' id='90dbb6d6'/>
	<union-decl name='fletcher_4_ctx' size-in-bits='2048' visibility='default' id='1f951ade'>
	<data-member access='public'>
	<var-decl name='scalar' type-id='39730d0b' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='superscalar' type-id='729b6ebb' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='sse' type-id='cbd91ec1' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='avx' type-id='481f90b1' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='avx512' type-id='16582e69' visibility='default'/>
	</data-member>
	</union-decl>
	<typedef-decl name='fletcher_4_ctx_t' type-id='1f951ade' id='4b675395'/>
	<qualified-type-def type-id='aa14691a' const='yes' id='3f8e8d11'/>
	<pointer-type-def type-id='4b675395' size-in-bits='64' id='0f7df99e'/>
	<qualified-type-def type-id='8f92235e' volatile='yes' id='430e0681'/>
	<pointer-type-def type-id='430e0681' size-in-bits='64' id='3a147f31'/>
	<pointer-type-def type-id='74e39470' size-in-bits='64' id='eefe7427'/>
	<pointer-type-def type-id='d6fd5c6c' size-in-bits='64' id='bfe36153'/>
	<pointer-type-def type-id='029a8ebe' size-in-bits='64' id='0bcca125'/>
	<pointer-type-def type-id='cefa0f4a' size-in-bits='64' id='1e276399'/>
	<var-decl name='fletcher_4_abd_ops' type-id='c2eb138a' mangled-name='fletcher_4_abd_ops' visibility='default' elf-symbol-id='fletcher_4_abd_ops'/>
	<function-decl name='atomic_swap_32' mangled-name='atomic_swap_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_32'>
	<parameter type-id='3a147f31'/>
	<parameter type-id='8f92235e'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='membar_producer' mangled-name='membar_producer' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='membar_producer'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fletcher_init' mangled-name='fletcher_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_init'>
	<parameter type-id='c24fc2ee' name='zcp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fletcher_2_incremental_native' mangled-name='fletcher_2_incremental_native' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_2_incremental_native'>
	<parameter type-id='eaa32e2f' name='buf'/>
	<parameter type-id='b59d7dce' name='size'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='fletcher_2_native' mangled-name='fletcher_2_native' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_2_native'>
	<parameter type-id='eaa32e2f' name='buf'/>
	<parameter type-id='9c313c2d' name='size'/>
	<parameter type-id='eaa32e2f' name='ctx_template'/>
	<parameter type-id='c24fc2ee' name='zcp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fletcher_2_incremental_byteswap' mangled-name='fletcher_2_incremental_byteswap' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_2_incremental_byteswap'>
	<parameter type-id='eaa32e2f' name='buf'/>
	<parameter type-id='b59d7dce' name='size'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='fletcher_2_byteswap' mangled-name='fletcher_2_byteswap' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_2_byteswap'>
	<parameter type-id='eaa32e2f' name='buf'/>
	<parameter type-id='9c313c2d' name='size'/>
	<parameter type-id='eaa32e2f' name='ctx_template'/>
	<parameter type-id='c24fc2ee' name='zcp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fletcher_4_impl_set' mangled-name='fletcher_4_impl_set' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_impl_set'>
	<parameter type-id='80f4b756' name='val'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='fletcher_4_native' mangled-name='fletcher_4_native' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_native'>
	<parameter type-id='eaa32e2f' name='buf'/>
	<parameter type-id='9c313c2d' name='size'/>
	<parameter type-id='eaa32e2f' name='ctx_template'/>
	<parameter type-id='c24fc2ee' name='zcp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fletcher_4_byteswap' mangled-name='fletcher_4_byteswap' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='fletcher_4_byteswap'>
	<parameter type-id='eaa32e2f' name='buf'/>
	<parameter type-id='9c313c2d' name='size'/>
	<parameter type-id='eaa32e2f' name='ctx_template'/>
	<parameter type-id='c24fc2ee' name='zcp'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-type size-in-bits='64' id='f4a1892e'>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='a5444274'>
	<parameter type-id='eefe7427'/>
	<return type-id='48b5725f'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zfs_fletcher_avx512.c' language='LANG_C99'>
	<typedef-decl name='fletcher_4_init_f' type-id='173aa527' id='b9ae1656'/>
	<typedef-decl name='fletcher_4_fini_f' type-id='0ad5b8a8' id='c4c1f4fc'/>
	<typedef-decl name='fletcher_4_compute_f' type-id='38147eff' id='ad1dc4cb'/>
	<class-decl name='fletcher_4_func' size-in-bits='1024' is-struct='yes' visibility='default' id='57f479a0'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='init_native' type-id='b9ae1656' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='fini_native' type-id='c4c1f4fc' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='compute_native' type-id='ad1dc4cb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='init_byteswap' type-id='b9ae1656' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='fini_byteswap' type-id='c4c1f4fc' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='compute_byteswap' type-id='ad1dc4cb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='valid' type-id='297d38bc' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='uses_fpu' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='name' type-id='80f4b756' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='fletcher_4_ops_t' type-id='57f479a0' id='eba91718'/>
	<qualified-type-def type-id='eba91718' const='yes' id='9eeabdc8'/>
	<pointer-type-def type-id='e9e61702' size-in-bits='64' id='297d38bc'/>
	<pointer-type-def type-id='fe40251b' size-in-bits='64' id='173aa527'/>
	<pointer-type-def type-id='17fb1f83' size-in-bits='64' id='38147eff'/>
	<pointer-type-def type-id='fb39e25e' size-in-bits='64' id='0ad5b8a8'/>
	<var-decl name='fletcher_4_avx512f_ops' type-id='9eeabdc8' mangled-name='fletcher_4_avx512f_ops' visibility='default' elf-symbol-id='fletcher_4_avx512f_ops'/>
	<var-decl name='fletcher_4_avx512bw_ops' type-id='9eeabdc8' mangled-name='fletcher_4_avx512bw_ops' visibility='default' elf-symbol-id='fletcher_4_avx512bw_ops'/>
	<function-type size-in-bits='64' id='e9e61702'>
	<return type-id='c19b74c3'/>
	</function-type>
	<function-type size-in-bits='64' id='fe40251b'>
	<parameter type-id='0f7df99e'/>
	<return type-id='48b5725f'/>
	</function-type>
	<function-type size-in-bits='64' id='17fb1f83'>
	<parameter type-id='0f7df99e'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='9c313c2d'/>
	<return type-id='48b5725f'/>
	</function-type>
	<function-type size-in-bits='64' id='fb39e25e'>
	<parameter type-id='0f7df99e'/>
	<parameter type-id='c24fc2ee'/>
	<return type-id='48b5725f'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zfs_fletcher_intel.c' language='LANG_C99'>
	<var-decl name='fletcher_4_avx2_ops' type-id='9eeabdc8' mangled-name='fletcher_4_avx2_ops' visibility='default' elf-symbol-id='fletcher_4_avx2_ops'/>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zfs_fletcher_sse.c' language='LANG_C99'>
	<var-decl name='fletcher_4_sse2_ops' type-id='9eeabdc8' mangled-name='fletcher_4_sse2_ops' visibility='default' elf-symbol-id='fletcher_4_sse2_ops'/>
	<var-decl name='fletcher_4_ssse3_ops' type-id='9eeabdc8' mangled-name='fletcher_4_ssse3_ops' visibility='default' elf-symbol-id='fletcher_4_ssse3_ops'/>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zfs_fletcher_superscalar.c' language='LANG_C99'>
	<var-decl name='fletcher_4_superscalar_ops' type-id='9eeabdc8' mangled-name='fletcher_4_superscalar_ops' visibility='default' elf-symbol-id='fletcher_4_superscalar_ops'/>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zfs_fletcher_superscalar4.c' language='LANG_C99'>
	<var-decl name='fletcher_4_superscalar4_ops' type-id='9eeabdc8' mangled-name='fletcher_4_superscalar4_ops' visibility='default' elf-symbol-id='fletcher_4_superscalar4_ops'/>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zfs_namecheck.c' language='LANG_C99'>
	<var-decl name='zfs_max_dataset_nesting' type-id='95e97e5e' mangled-name='zfs_max_dataset_nesting' visibility='default' elf-symbol-id='zfs_max_dataset_nesting'/>
	<function-decl name='get_dataset_depth' mangled-name='get_dataset_depth' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='get_dataset_depth'>
	<parameter type-id='80f4b756' name='path'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_component_namecheck' mangled-name='zfs_component_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_component_namecheck'>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='053457bd' name='why'/>
	<parameter type-id='26a90f95' name='what'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='dataset_namecheck' mangled-name='dataset_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='dataset_namecheck'>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='053457bd' name='why'/>
	<parameter type-id='26a90f95' name='what'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='bookmark_namecheck' mangled-name='bookmark_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='bookmark_namecheck'>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='053457bd' name='why'/>
	<parameter type-id='26a90f95' name='what'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='snapshot_namecheck' mangled-name='snapshot_namecheck' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='snapshot_namecheck'>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='053457bd' name='why'/>
	<parameter type-id='26a90f95' name='what'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zfs_prop.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='b99c00c9' size-in-bits='768' id='bcc77e38'>
	<subrange length='12' type-id='7359adad' id='84827bdc'/>
	</array-type-def>
	<pointer-type-def type-id='3eee3342' size-in-bits='64' id='73f8e240'/>
	<var-decl name='zfs_userquota_prop_prefixes' type-id='bcc77e38' mangled-name='zfs_userquota_prop_prefixes' visibility='default' elf-symbol-id='zfs_userquota_prop_prefixes'/>
	<function-decl name='zfs_mod_list_supported' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<return type-id='73f8e240'/>
	</function-decl>
	<function-decl name='zfs_mod_list_supported_free' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='73f8e240'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zprop_register_impl' mangled-name='zprop_register_impl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_register_impl'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='31429eff'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='999701cc'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='c19b74c3'/>
	<parameter type-id='c19b74c3'/>
	<parameter type-id='c19b74c3'/>
	<parameter type-id='c8bc397b'/>
	<parameter type-id='a3372543'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zprop_register_string' mangled-name='zprop_register_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_register_string'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='999701cc'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='a3372543'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zprop_register_number' mangled-name='zprop_register_number' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_register_number'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='999701cc'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='c19b74c3'/>
	<parameter type-id='a3372543'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zprop_register_index' mangled-name='zprop_register_index' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_register_index'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='999701cc'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='c8bc397b'/>
	<parameter type-id='a3372543'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zprop_register_hidden' mangled-name='zprop_register_hidden' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_register_hidden'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='31429eff'/>
	<parameter type-id='999701cc'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='c19b74c3'/>
	<parameter type-id='a3372543'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='zprop_index_to_string' mangled-name='zprop_index_to_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_index_to_string'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='7d3cd834'/>
	<parameter type-id='2e45de5d'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zprop_random_value' mangled-name='zprop_random_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_random_value'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='9c313c2d'/>
	<parameter type-id='2e45de5d'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='zprop_valid_char' mangled-name='zprop_valid_char' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zprop_valid_char'>
	<parameter type-id='a84c031d'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_string_to_index' mangled-name='zfs_prop_string_to_index' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_string_to_index'>
	<parameter type-id='58603c44' name='prop'/>
	<parameter type-id='80f4b756' name='string'/>
	<parameter type-id='5d6479ae' name='index'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_random_value' mangled-name='zfs_prop_random_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_random_value'>
	<parameter type-id='58603c44' name='prop'/>
	<parameter type-id='9c313c2d' name='seed'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='zfs_prop_visible' mangled-name='zfs_prop_visible' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_visible'>
	<parameter type-id='58603c44' name='prop'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='zfs_prop_values' mangled-name='zfs_prop_values' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_values'>
	<parameter type-id='58603c44' name='prop'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_prop_is_string' mangled-name='zfs_prop_is_string' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_is_string'>
	<parameter type-id='58603c44' name='prop'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zfs_prop_column_name' mangled-name='zfs_prop_column_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_column_name'>
	<parameter type-id='58603c44' name='prop'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zfs_prop_align_right' mangled-name='zfs_prop_align_right' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zfs_prop_align_right'>
	<parameter type-id='58603c44' name='prop'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zpool_prop.c' language='LANG_C99'>
	<function-decl name='zpool_prop_string_to_index' mangled-name='zpool_prop_string_to_index' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_string_to_index'>
	<parameter type-id='5d0c23fb' name='prop'/>
	<parameter type-id='80f4b756' name='string'/>
	<parameter type-id='5d6479ae' name='index'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='zpool_prop_random_value' mangled-name='zpool_prop_random_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_random_value'>
	<parameter type-id='5d0c23fb' name='prop'/>
	<parameter type-id='9c313c2d' name='seed'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='zpool_prop_values' mangled-name='zpool_prop_values' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_values'>
	<parameter type-id='5d0c23fb' name='prop'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zpool_prop_column_name' mangled-name='zpool_prop_column_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_column_name'>
	<parameter type-id='5d0c23fb' name='prop'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='zpool_prop_align_right' mangled-name='zpool_prop_align_right' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='zpool_prop_align_right'>
	<parameter type-id='5d0c23fb' name='prop'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='vdev_prop_get_table' mangled-name='vdev_prop_get_table' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_get_table'>
	<return type-id='76c8174b'/>
	</function-decl>
	<function-decl name='vdev_prop_string_to_index' mangled-name='vdev_prop_string_to_index' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_string_to_index'>
	<parameter type-id='5aa5c90c' name='prop'/>
	<parameter type-id='80f4b756' name='string'/>
	<parameter type-id='5d6479ae' name='index'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='vdev_prop_random_value' mangled-name='vdev_prop_random_value' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_random_value'>
	<parameter type-id='5aa5c90c' name='prop'/>
	<parameter type-id='9c313c2d' name='seed'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='vdev_prop_values' mangled-name='vdev_prop_values' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_values'>
	<parameter type-id='5aa5c90c' name='prop'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='vdev_prop_column_name' mangled-name='vdev_prop_column_name' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_column_name'>
	<parameter type-id='5aa5c90c' name='prop'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='vdev_prop_align_right' mangled-name='vdev_prop_align_right' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='vdev_prop_align_right'>
	<parameter type-id='5aa5c90c' name='prop'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='module/zcommon/zprop_common.c' language='LANG_C99'>
	<function-decl name='__ctype_tolower_loc' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='24f95ba5'/>
	</function-decl>
	</abi-instr>
	</abi-corpus>
	diff --git a/sys/contrib/openzfs/lib/libzfs/libzfs_mount.c b/sys/contrib/openzfs/lib/libzfs/libzfs_mount.c
	index ec6ebad2f180..42988bf9cbb2 100644
	--- a/sys/contrib/openzfs/lib/libzfs/libzfs_mount.c
	+++ b/sys/contrib/openzfs/lib/libzfs/libzfs_mount.c
	@@ -1,1463 +1,1467 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright 2015 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2014, 2022 by Delphix. All rights reserved.
	* Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>
	* Copyright 2017 RackTop Systems.
	* Copyright (c) 2018 Datto Inc.
	* Copyright 2018 OmniOS Community Edition (OmniOSce) Association.
	*/

	/*
	* Routines to manage ZFS mounts. We separate all the nasty routines that have
	* to deal with the OS. The following functions are the main entry points --
	* they are used by mount and unmount and when changing a filesystem's
	* mountpoint.
	*
	* zfs_is_mounted()
	* zfs_mount()
	* zfs_mount_at()
	* zfs_unmount()
	* zfs_unmountall()
	*
	* This file also contains the functions used to manage sharing filesystems:
	*
	* zfs_is_shared()
	* zfs_share()
	* zfs_unshare()
	* zfs_unshareall()
	* zfs_commit_shares()
	*
	* The following functions are available for pool consumers, and will
	* mount/unmount and share/unshare all datasets within pool:
	*
	* zpool_enable_datasets()
	* zpool_disable_datasets()
	*/

	#include <dirent.h>
	#include <dlfcn.h>
	#include <errno.h>
	#include <fcntl.h>
	#include <libgen.h>
	#include <libintl.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <string.h>
	#include <unistd.h>
	#include <zone.h>
	#include <sys/mntent.h>
	#include <sys/mount.h>
	#include <sys/stat.h>
	#include <sys/vfs.h>
	#include <sys/dsl_crypt.h>

	#include <libzfs.h>
	#include <libzutil.h>

	#include "libzfs_impl.h"
	#include <thread_pool.h>

	#include <libshare.h>
	#include <sys/systeminfo.h>
	#define MAXISALEN 257 /* based on sysinfo(2) man page */

	-static int mount_tp_nthr = 512; /* tpool threads for multi-threaded mounting */
	-
	static void zfs_mount_task(void *);

	static const proto_table_t proto_table[SA_PROTOCOL_COUNT] = {
	[SA_PROTOCOL_NFS] =
	{ZFS_PROP_SHARENFS, EZFS_SHARENFSFAILED, EZFS_UNSHARENFSFAILED},
	[SA_PROTOCOL_SMB] =
	{ZFS_PROP_SHARESMB, EZFS_SHARESMBFAILED, EZFS_UNSHARESMBFAILED},
	};

	static const enum sa_protocol share_all_proto[SA_PROTOCOL_COUNT + 1] = {
	SA_PROTOCOL_NFS,
	SA_PROTOCOL_SMB,
	SA_NO_PROTOCOL
	};



	static boolean_t
	dir_is_empty_stat(const char *dirname)
	{
	struct stat st;

	/*
	* We only want to return false if the given path is a non empty
	* directory, all other errors are handled elsewhere.
	*/
	if (stat(dirname, &st) < 0 \|\| !S_ISDIR(st.st_mode)) {
	return (B_TRUE);
	}

	/*
	* An empty directory will still have two entries in it, one
	* entry for each of "." and "..".
	*/
	if (st.st_size > 2) {
	return (B_FALSE);
	}

	return (B_TRUE);
	}

	static boolean_t
	dir_is_empty_readdir(const char *dirname)
	{
	DIR *dirp;
	struct dirent64 *dp;
	int dirfd;

	if ((dirfd = openat(AT_FDCWD, dirname,
	O_RDONLY \| O_NDELAY \| O_LARGEFILE \| O_CLOEXEC, 0)) < 0) {
	return (B_TRUE);
	}

	if ((dirp = fdopendir(dirfd)) == NULL) {
	(void) close(dirfd);
	return (B_TRUE);
	}

	while ((dp = readdir64(dirp)) != NULL) {

	if (strcmp(dp->d_name, ".") == 0 \|\|
	strcmp(dp->d_name, "..") == 0)
	continue;

	(void) closedir(dirp);
	return (B_FALSE);
	}

	(void) closedir(dirp);
	return (B_TRUE);
	}

	/*
	* Returns true if the specified directory is empty. If we can't open the
	* directory at all, return true so that the mount can fail with a more
	* informative error message.
	*/
	static boolean_t
	dir_is_empty(const char *dirname)
	{
	struct statfs64 st;

	/*
	* If the statvfs call fails or the filesystem is not a ZFS
	* filesystem, fall back to the slow path which uses readdir.
	*/
	if ((statfs64(dirname, &st) != 0) \|\|
	(st.f_type != ZFS_SUPER_MAGIC)) {
	return (dir_is_empty_readdir(dirname));
	}

	/*
	* At this point, we know the provided path is on a ZFS
	* filesystem, so we can use stat instead of readdir to
	* determine if the directory is empty or not. We try to avoid
	* using readdir because that requires opening "dirname"; this
	* open file descriptor can potentially end up in a child
	* process if there's a concurrent fork, thus preventing the
	* zfs_mount() from otherwise succeeding (the open file
	* descriptor inherited by the child process will cause the
	* parent's mount to fail with EBUSY). The performance
	* implications of replacing the open, read, and close with a
	* single stat is nice; but is not the main motivation for the
	* added complexity.
	*/
	return (dir_is_empty_stat(dirname));
	}

	/*
	* Checks to see if the mount is active. If the filesystem is mounted, we fill
	* in 'where' with the current mountpoint, and return 1. Otherwise, we return
	* 0.
	*/
	boolean_t
	is_mounted(libzfs_handle_t zfs_hdl, const char special, char **where)
	{
	struct mnttab entry;

	if (libzfs_mnttab_find(zfs_hdl, special, &entry) != 0)
	return (B_FALSE);

	if (where != NULL)
	*where = zfs_strdup(zfs_hdl, entry.mnt_mountp);

	return (B_TRUE);
	}

	boolean_t
	zfs_is_mounted(zfs_handle_t zhp, char *where)
	{
	return (is_mounted(zhp->zfs_hdl, zfs_get_name(zhp), where));
	}

	/*
	* Checks any higher order concerns about whether the given dataset is
	* mountable, false otherwise. zfs_is_mountable_internal specifically assumes
	* that the caller has verified the sanity of mounting the dataset at
	* its mountpoint to the extent the caller wants.
	*/
	static boolean_t
	zfs_is_mountable_internal(zfs_handle_t *zhp)
	{
	if (zfs_prop_get_int(zhp, ZFS_PROP_ZONED) &&
	getzoneid() == GLOBAL_ZONEID)
	return (B_FALSE);

	return (B_TRUE);
	}

	/*
	* Returns true if the given dataset is mountable, false otherwise. Returns the
	* mountpoint in 'buf'.
	*/
	static boolean_t
	zfs_is_mountable(zfs_handle_t zhp, char buf, size_t buflen,
	zprop_source_t *source, int flags)
	{
	char sourceloc[MAXNAMELEN];
	zprop_source_t sourcetype;

	if (!zfs_prop_valid_for_type(ZFS_PROP_MOUNTPOINT, zhp->zfs_type,
	B_FALSE))
	return (B_FALSE);

	verify(zfs_prop_get(zhp, ZFS_PROP_MOUNTPOINT, buf, buflen,
	&sourcetype, sourceloc, sizeof (sourceloc), B_FALSE) == 0);

	if (strcmp(buf, ZFS_MOUNTPOINT_NONE) == 0 \|\|
	strcmp(buf, ZFS_MOUNTPOINT_LEGACY) == 0)
	return (B_FALSE);

	if (zfs_prop_get_int(zhp, ZFS_PROP_CANMOUNT) == ZFS_CANMOUNT_OFF)
	return (B_FALSE);

	if (!zfs_is_mountable_internal(zhp))
	return (B_FALSE);

	if (zfs_prop_get_int(zhp, ZFS_PROP_REDACTED) && !(flags & MS_FORCE))
	return (B_FALSE);

	if (source)
	*source = sourcetype;

	return (B_TRUE);
	}

	/*
	* The filesystem is mounted by invoking the system mount utility rather
	* than by the system call mount(2). This ensures that the /etc/mtab
	* file is correctly locked for the update. Performing our own locking
	* and /etc/mtab update requires making an unsafe assumption about how
	* the mount utility performs its locking. Unfortunately, this also means
	* in the case of a mount failure we do not have the exact errno. We must
	* make due with return value from the mount process.
	*
	* In the long term a shared library called libmount is under development
	* which provides a common API to address the locking and errno issues.
	* Once the standard mount utility has been updated to use this library
	* we can add an autoconf check to conditionally use it.
	*
	* http://www.kernel.org/pub/linux/utils/util-linux/libmount-docs/index.html
	*/

	static int
	zfs_add_option(zfs_handle_t zhp, char options, int len,
	zfs_prop_t prop, const char on, const char off)
	{
	const char *source;
	uint64_t value;

	/* Skip adding duplicate default options */
	if ((strstr(options, on) != NULL) \|\| (strstr(options, off) != NULL))
	return (0);

	/*
	* zfs_prop_get_int() is not used to ensure our mount options
	* are not influenced by the current /proc/self/mounts contents.
	*/
	value = getprop_uint64(zhp, prop, &source);

	(void) strlcat(options, ",", len);
	(void) strlcat(options, value ? on : off, len);

	return (0);
	}

	static int
	zfs_add_options(zfs_handle_t zhp, char options, int len)
	{
	int error = 0;

	error = zfs_add_option(zhp, options, len,
	ZFS_PROP_ATIME, MNTOPT_ATIME, MNTOPT_NOATIME);
	/*
	* don't add relatime/strictatime when atime=off, otherwise strictatime
	* will force atime=on
	*/
	if (strstr(options, MNTOPT_NOATIME) == NULL) {
	error = zfs_add_option(zhp, options, len,
	ZFS_PROP_RELATIME, MNTOPT_RELATIME, MNTOPT_STRICTATIME);
	}
	error = error ? error : zfs_add_option(zhp, options, len,
	ZFS_PROP_DEVICES, MNTOPT_DEVICES, MNTOPT_NODEVICES);
	error = error ? error : zfs_add_option(zhp, options, len,
	ZFS_PROP_EXEC, MNTOPT_EXEC, MNTOPT_NOEXEC);
	error = error ? error : zfs_add_option(zhp, options, len,
	ZFS_PROP_READONLY, MNTOPT_RO, MNTOPT_RW);
	error = error ? error : zfs_add_option(zhp, options, len,
	ZFS_PROP_SETUID, MNTOPT_SETUID, MNTOPT_NOSETUID);
	error = error ? error : zfs_add_option(zhp, options, len,
	ZFS_PROP_NBMAND, MNTOPT_NBMAND, MNTOPT_NONBMAND);

	return (error);
	}

	int
	zfs_mount(zfs_handle_t zhp, const char options, int flags)
	{
	char mountpoint[ZFS_MAXPROPLEN];

	if (!zfs_is_mountable(zhp, mountpoint, sizeof (mountpoint), NULL,
	flags))
	return (0);

	return (zfs_mount_at(zhp, options, flags, mountpoint));
	}

	/*
	* Mount the given filesystem.
	*/
	int
	zfs_mount_at(zfs_handle_t zhp, const char options, int flags,
	const char *mountpoint)
	{
	struct stat buf;
	char mntopts[MNT_LINE_MAX];
	char overlay[ZFS_MAXPROPLEN];
	char prop_encroot[MAXNAMELEN];
	boolean_t is_encroot;
	zfs_handle_t *encroot_hp = zhp;
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	uint64_t keystatus;
	int remount = 0, rc;

	if (options == NULL) {
	(void) strlcpy(mntopts, MNTOPT_DEFAULTS, sizeof (mntopts));
	} else {
	(void) strlcpy(mntopts, options, sizeof (mntopts));
	}

	if (strstr(mntopts, MNTOPT_REMOUNT) != NULL)
	remount = 1;

	/* Potentially duplicates some checks if invoked by zfs_mount(). */
	if (!zfs_is_mountable_internal(zhp))
	return (0);

	/*
	* If the pool is imported read-only then all mounts must be read-only
	*/
	if (zpool_get_prop_int(zhp->zpool_hdl, ZPOOL_PROP_READONLY, NULL))
	(void) strlcat(mntopts, "," MNTOPT_RO, sizeof (mntopts));

	/*
	* Append default mount options which apply to the mount point.
	* This is done because under Linux (unlike Solaris) multiple mount
	* points may reference a single super block. This means that just
	* given a super block there is no back reference to update the per
	* mount point options.
	*/
	rc = zfs_add_options(zhp, mntopts, sizeof (mntopts));
	if (rc) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"default options unavailable"));
	return (zfs_error_fmt(hdl, EZFS_MOUNTFAILED,
	dgettext(TEXT_DOMAIN, "cannot mount '%s'"),
	mountpoint));
	}

	/*
	* If the filesystem is encrypted the key must be loaded in order to
	* mount. If the key isn't loaded, the MS_CRYPT flag decides whether
	* or not we attempt to load the keys. Note: we must call
	* zfs_refresh_properties() here since some callers of this function
	* (most notably zpool_enable_datasets()) may implicitly load our key
	* by loading the parent's key first.
	*/
	if (zfs_prop_get_int(zhp, ZFS_PROP_ENCRYPTION) != ZIO_CRYPT_OFF) {
	zfs_refresh_properties(zhp);
	keystatus = zfs_prop_get_int(zhp, ZFS_PROP_KEYSTATUS);

	/*
	* If the key is unavailable and MS_CRYPT is set give the
	* user a chance to enter the key. Otherwise just fail
	* immediately.
	*/
	if (keystatus == ZFS_KEYSTATUS_UNAVAILABLE) {
	if (flags & MS_CRYPT) {
	rc = zfs_crypto_get_encryption_root(zhp,
	&is_encroot, prop_encroot);
	if (rc) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"Failed to get encryption root for "
	"'%s'."), zfs_get_name(zhp));
	return (rc);
	}

	if (!is_encroot) {
	encroot_hp = zfs_open(hdl, prop_encroot,
	ZFS_TYPE_DATASET);
	if (encroot_hp == NULL)
	return (hdl->libzfs_error);
	}

	rc = zfs_crypto_load_key(encroot_hp,
	B_FALSE, NULL);

	if (!is_encroot)
	zfs_close(encroot_hp);
	if (rc)
	return (rc);
	} else {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"encryption key not loaded"));
	return (zfs_error_fmt(hdl, EZFS_MOUNTFAILED,
	dgettext(TEXT_DOMAIN, "cannot mount '%s'"),
	mountpoint));
	}
	}

	}

	/*
	* Append zfsutil option so the mount helper allow the mount
	*/
	strlcat(mntopts, "," MNTOPT_ZFSUTIL, sizeof (mntopts));

	/* Create the directory if it doesn't already exist */
	if (lstat(mountpoint, &buf) != 0) {
	if (mkdirp(mountpoint, 0755) != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"failed to create mountpoint: %s"),
	zfs_strerror(errno));
	return (zfs_error_fmt(hdl, EZFS_MOUNTFAILED,
	dgettext(TEXT_DOMAIN, "cannot mount '%s'"),
	mountpoint));
	}
	}

	/*
	* Overlay mounts are enabled by default but may be disabled
	* via the 'overlay' property. The -O flag remains for compatibility.
	*/
	if (!(flags & MS_OVERLAY)) {
	if (zfs_prop_get(zhp, ZFS_PROP_OVERLAY, overlay,
	sizeof (overlay), NULL, NULL, 0, B_FALSE) == 0) {
	if (strcmp(overlay, "on") == 0) {
	flags \|= MS_OVERLAY;
	}
	}
	}

	/*
	* Determine if the mountpoint is empty. If so, refuse to perform the
	* mount. We don't perform this check if 'remount' is
	* specified or if overlay option (-O) is given
	*/
	if ((flags & MS_OVERLAY) == 0 && !remount &&
	!dir_is_empty(mountpoint)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"directory is not empty"));
	return (zfs_error_fmt(hdl, EZFS_MOUNTFAILED,
	dgettext(TEXT_DOMAIN, "cannot mount '%s'"), mountpoint));
	}

	/* perform the mount */
	rc = do_mount(zhp, mountpoint, mntopts, flags);
	if (rc) {
	/*
	* Generic errors are nasty, but there are just way too many
	* from mount(), and they're well-understood. We pick a few
	* common ones to improve upon.
	*/
	if (rc == EBUSY) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"mountpoint or dataset is busy"));
	} else if (rc == EPERM) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"Insufficient privileges"));
	} else if (rc == ENOTSUP) {
	int spa_version;

	VERIFY(zfs_spa_version(zhp, &spa_version) == 0);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"Can't mount a version %llu "
	"file system on a version %d pool. Pool must be"
	" upgraded to mount this file system."),
	(u_longlong_t)zfs_prop_get_int(zhp,
	ZFS_PROP_VERSION), spa_version);
	} else {
	zfs_error_aux(hdl, "%s", zfs_strerror(rc));
	}
	return (zfs_error_fmt(hdl, EZFS_MOUNTFAILED,
	dgettext(TEXT_DOMAIN, "cannot mount '%s'"),
	zhp->zfs_name));
	}

	/* remove the mounted entry before re-adding on remount */
	if (remount)
	libzfs_mnttab_remove(hdl, zhp->zfs_name);

	/* add the mounted entry into our cache */
	libzfs_mnttab_add(hdl, zfs_get_name(zhp), mountpoint, mntopts);
	return (0);
	}

	/*
	* Unmount a single filesystem.
	*/
	static int
	unmount_one(zfs_handle_t zhp, const char mountpoint, int flags)
	{
	int error;

	error = do_unmount(zhp, mountpoint, flags);
	if (error != 0) {
	int libzfs_err;

	switch (error) {
	case EBUSY:
	libzfs_err = EZFS_BUSY;
	break;
	case EIO:
	libzfs_err = EZFS_IO;
	break;
	case ENOENT:
	libzfs_err = EZFS_NOENT;
	break;
	case ENOMEM:
	libzfs_err = EZFS_NOMEM;
	break;
	case EPERM:
	libzfs_err = EZFS_PERM;
	break;
	default:
	libzfs_err = EZFS_UMOUNTFAILED;
	}
	if (zhp) {
	return (zfs_error_fmt(zhp->zfs_hdl, libzfs_err,
	dgettext(TEXT_DOMAIN, "cannot unmount '%s'"),
	mountpoint));
	} else {
	return (-1);
	}
	}

	return (0);
	}

	/*
	* Unmount the given filesystem.
	*/
	int
	zfs_unmount(zfs_handle_t zhp, const char mountpoint, int flags)
	{
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	struct mnttab entry;
	char *mntpt = NULL;
	boolean_t encroot, unmounted = B_FALSE;

	/* check to see if we need to unmount the filesystem */
	if (mountpoint != NULL \|\| ((zfs_get_type(zhp) == ZFS_TYPE_FILESYSTEM) &&
	libzfs_mnttab_find(hdl, zhp->zfs_name, &entry) == 0)) {
	/*
	* mountpoint may have come from a call to
	* getmnt/getmntany if it isn't NULL. If it is NULL,
	* we know it comes from libzfs_mnttab_find which can
	* then get freed later. We strdup it to play it safe.
	*/
	if (mountpoint == NULL)
	mntpt = zfs_strdup(hdl, entry.mnt_mountp);
	else
	mntpt = zfs_strdup(hdl, mountpoint);

	/*
	* Unshare and unmount the filesystem
	*/
	if (zfs_unshare(zhp, mntpt, share_all_proto) != 0) {
	free(mntpt);
	return (-1);
	}
	zfs_commit_shares(NULL);

	if (unmount_one(zhp, mntpt, flags) != 0) {
	free(mntpt);
	(void) zfs_share(zhp, NULL);
	zfs_commit_shares(NULL);
	return (-1);
	}

	libzfs_mnttab_remove(hdl, zhp->zfs_name);
	free(mntpt);
	unmounted = B_TRUE;
	}

	/*
	* If the MS_CRYPT flag is provided we must ensure we attempt to
	* unload the dataset's key regardless of whether we did any work
	* to unmount it. We only do this for encryption roots.
	*/
	if ((flags & MS_CRYPT) != 0 &&
	zfs_prop_get_int(zhp, ZFS_PROP_ENCRYPTION) != ZIO_CRYPT_OFF) {
	zfs_refresh_properties(zhp);

	if (zfs_crypto_get_encryption_root(zhp, &encroot, NULL) != 0 &&
	unmounted) {
	(void) zfs_mount(zhp, NULL, 0);
	return (-1);
	}

	if (encroot && zfs_prop_get_int(zhp, ZFS_PROP_KEYSTATUS) ==
	ZFS_KEYSTATUS_AVAILABLE &&
	zfs_crypto_unload_key(zhp) != 0) {
	(void) zfs_mount(zhp, NULL, 0);
	return (-1);
	}
	}

	zpool_disable_volume_os(zhp->zfs_name);

	return (0);
	}

	/*
	* Unmount this filesystem and any children inheriting the mountpoint property.
	* To do this, just act like we're changing the mountpoint property, but don't
	* remount the filesystems afterwards.
	*/
	int
	zfs_unmountall(zfs_handle_t *zhp, int flags)
	{
	prop_changelist_t *clp;
	int ret;

	clp = changelist_gather(zhp, ZFS_PROP_MOUNTPOINT,
	CL_GATHER_ITER_MOUNTED, flags);
	if (clp == NULL)
	return (-1);

	ret = changelist_prefix(clp);
	changelist_free(clp);

	return (ret);
	}

	/*
	* Unshare a filesystem by mountpoint.
	*/
	static int
	unshare_one(libzfs_handle_t hdl, const char name, const char *mountpoint,
	enum sa_protocol proto)
	{
	int err = sa_disable_share(mountpoint, proto);
	if (err != SA_OK)
	return (zfs_error_fmt(hdl, proto_table[proto].p_unshare_err,
	dgettext(TEXT_DOMAIN, "cannot unshare '%s': %s"),
	name, sa_errorstr(err)));

	return (0);
	}

	/*
	* Share the given filesystem according to the options in the specified
	* protocol specific properties (sharenfs, sharesmb). We rely
	* on "libshare" to do the dirty work for us.
	*/
	int
	zfs_share(zfs_handle_t zhp, const enum sa_protocol proto)
	{
	char mountpoint[ZFS_MAXPROPLEN];
	char shareopts[ZFS_MAXPROPLEN];
	char sourcestr[ZFS_MAXPROPLEN];
	const enum sa_protocol *curr_proto;
	zprop_source_t sourcetype;
	int err = 0;

	if (proto == NULL)
	proto = share_all_proto;

	if (!zfs_is_mountable(zhp, mountpoint, sizeof (mountpoint), NULL, 0))
	return (0);

	for (curr_proto = proto; *curr_proto != SA_NO_PROTOCOL; curr_proto++) {
	/*
	* Return success if there are no share options.
	*/
	if (zfs_prop_get(zhp, proto_table[*curr_proto].p_prop,
	shareopts, sizeof (shareopts), &sourcetype, sourcestr,
	ZFS_MAXPROPLEN, B_FALSE) != 0 \|\|
	strcmp(shareopts, "off") == 0)
	continue;

	/*
	* If the 'zoned' property is set, then zfs_is_mountable()
	* will have already bailed out if we are in the global zone.
	* But local zones cannot be NFS servers, so we ignore it for
	* local zones as well.
	*/
	if (zfs_prop_get_int(zhp, ZFS_PROP_ZONED))
	continue;

	err = sa_enable_share(zfs_get_name(zhp), mountpoint, shareopts,
	*curr_proto);
	if (err != SA_OK) {
	return (zfs_error_fmt(zhp->zfs_hdl,
	proto_table[*curr_proto].p_share_err,
	dgettext(TEXT_DOMAIN, "cannot share '%s: %s'"),
	zfs_get_name(zhp), sa_errorstr(err)));
	}

	}
	return (0);
	}

	/*
	* Check to see if the filesystem is currently shared.
	*/
	boolean_t
	zfs_is_shared(zfs_handle_t zhp, char *where,
	const enum sa_protocol *proto)
	{
	char *mountpoint;
	if (proto == NULL)
	proto = share_all_proto;

	if (ZFS_IS_VOLUME(zhp))
	return (B_FALSE);

	if (!zfs_is_mounted(zhp, &mountpoint))
	return (B_FALSE);

	for (const enum sa_protocol p = proto; p != SA_NO_PROTOCOL; ++p)
	if (sa_is_shared(mountpoint, *p)) {
	if (where != NULL)
	*where = mountpoint;
	else
	free(mountpoint);
	return (B_TRUE);
	}

	free(mountpoint);
	return (B_FALSE);
	}

	void
	zfs_commit_shares(const enum sa_protocol *proto)
	{
	if (proto == NULL)
	proto = share_all_proto;

	for (const enum sa_protocol p = proto; p != SA_NO_PROTOCOL; ++p)
	sa_commit_shares(*p);
	}

	void
	zfs_truncate_shares(const enum sa_protocol *proto)
	{
	if (proto == NULL)
	proto = share_all_proto;

	for (const enum sa_protocol p = proto; p != SA_NO_PROTOCOL; ++p)
	sa_truncate_shares(*p);
	}

	/*
	* Unshare the given filesystem.
	*/
	int
	zfs_unshare(zfs_handle_t zhp, const char mountpoint,
	const enum sa_protocol *proto)
	{
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	struct mnttab entry;

	if (proto == NULL)
	proto = share_all_proto;

	if (mountpoint != NULL \|\| ((zfs_get_type(zhp) == ZFS_TYPE_FILESYSTEM) &&
	libzfs_mnttab_find(hdl, zfs_get_name(zhp), &entry) == 0)) {

	/* check to see if need to unmount the filesystem */
	const char *mntpt = mountpoint ?: entry.mnt_mountp;

	for (const enum sa_protocol *curr_proto = proto;
	*curr_proto != SA_NO_PROTOCOL; curr_proto++)
	if (sa_is_shared(mntpt, *curr_proto) &&
	unshare_one(hdl, zhp->zfs_name,
	mntpt, *curr_proto) != 0)
	return (-1);
	}

	return (0);
	}

	/*
	* Same as zfs_unmountall(), but for NFS and SMB unshares.
	*/
	int
	zfs_unshareall(zfs_handle_t zhp, const enum sa_protocol proto)
	{
	prop_changelist_t *clp;
	int ret;

	if (proto == NULL)
	proto = share_all_proto;

	clp = changelist_gather(zhp, ZFS_PROP_SHARENFS, 0, 0);
	if (clp == NULL)
	return (-1);

	ret = changelist_unshare(clp, proto);
	changelist_free(clp);

	return (ret);
	}

	/*
	* Remove the mountpoint associated with the current dataset, if necessary.
	* We only remove the underlying directory if:
	*
	* - The mountpoint is not 'none' or 'legacy'
	* - The mountpoint is non-empty
	* - The mountpoint is the default or inherited
	* - The 'zoned' property is set, or we're in a local zone
	*
	* Any other directories we leave alone.
	*/
	void
	remove_mountpoint(zfs_handle_t *zhp)
	{
	char mountpoint[ZFS_MAXPROPLEN];
	zprop_source_t source;

	if (!zfs_is_mountable(zhp, mountpoint, sizeof (mountpoint),
	&source, 0))
	return;

	if (source == ZPROP_SRC_DEFAULT \|\|
	source == ZPROP_SRC_INHERITED) {
	/*
	* Try to remove the directory, silently ignoring any errors.
	* The filesystem may have since been removed or moved around,
	* and this error isn't really useful to the administrator in
	* any way.
	*/
	(void) rmdir(mountpoint);
	}
	}

	/*
	* Add the given zfs handle to the cb_handles array, dynamically reallocating
	* the array if it is out of space.
	*/
	void
	libzfs_add_handle(get_all_cb_t cbp, zfs_handle_t zhp)
	{
	if (cbp->cb_alloc == cbp->cb_used) {
	size_t newsz;
	zfs_handle_t **newhandles;

	newsz = cbp->cb_alloc != 0 ? cbp->cb_alloc * 2 : 64;
	newhandles = zfs_realloc(zhp->zfs_hdl,
	cbp->cb_handles, cbp->cb_alloc * sizeof (zfs_handle_t *),
	newsz * sizeof (zfs_handle_t *));
	cbp->cb_handles = newhandles;
	cbp->cb_alloc = newsz;
	}
	cbp->cb_handles[cbp->cb_used++] = zhp;
	}

	/*
	* Recursive helper function used during file system enumeration
	*/
	static int
	zfs_iter_cb(zfs_handle_t zhp, void data)
	{
	get_all_cb_t *cbp = data;

	if (!(zfs_get_type(zhp) & ZFS_TYPE_FILESYSTEM)) {
	zfs_close(zhp);
	return (0);
	}

	if (zfs_prop_get_int(zhp, ZFS_PROP_CANMOUNT) == ZFS_CANMOUNT_NOAUTO) {
	zfs_close(zhp);
	return (0);
	}

	if (zfs_prop_get_int(zhp, ZFS_PROP_KEYSTATUS) ==
	ZFS_KEYSTATUS_UNAVAILABLE) {
	zfs_close(zhp);
	return (0);
	}

	/*
	* If this filesystem is inconsistent and has a receive resume
	* token, we can not mount it.
	*/
	if (zfs_prop_get_int(zhp, ZFS_PROP_INCONSISTENT) &&
	zfs_prop_get(zhp, ZFS_PROP_RECEIVE_RESUME_TOKEN,
	NULL, 0, NULL, NULL, 0, B_TRUE) == 0) {
	zfs_close(zhp);
	return (0);
	}

	libzfs_add_handle(cbp, zhp);
	if (zfs_iter_filesystems_v2(zhp, 0, zfs_iter_cb, cbp) != 0) {
	zfs_close(zhp);
	return (-1);
	}
	return (0);
	}

	/*
	* Sort comparator that compares two mountpoint paths. We sort these paths so
	* that subdirectories immediately follow their parents. This means that we
	* effectively treat the '/' character as the lowest value non-nul char.
	* Since filesystems from non-global zones can have the same mountpoint
	* as other filesystems, the comparator sorts global zone filesystems to
	* the top of the list. This means that the global zone will traverse the
	* filesystem list in the correct order and can stop when it sees the
	* first zoned filesystem. In a non-global zone, only the delegated
	* filesystems are seen.
	*
	* An example sorted list using this comparator would look like:
	*
	* /foo
	* /foo/bar
	* /foo/bar/baz
	* /foo/baz
	* /foo.bar
	* /foo (NGZ1)
	* /foo (NGZ2)
	*
	* The mounting code depends on this ordering to deterministically iterate
	* over filesystems in order to spawn parallel mount tasks.
	*/
	static int
	mountpoint_cmp(const void arga, const void argb)
	{
	zfs_handle_t const zap = arga;
	zfs_handle_t za = zap;
	zfs_handle_t const zbp = argb;
	zfs_handle_t zb = zbp;
	char mounta[MAXPATHLEN];
	char mountb[MAXPATHLEN];
	const char *a = mounta;
	const char *b = mountb;
	boolean_t gota, gotb;
	uint64_t zoneda, zonedb;

	zoneda = zfs_prop_get_int(za, ZFS_PROP_ZONED);
	zonedb = zfs_prop_get_int(zb, ZFS_PROP_ZONED);
	if (zoneda && !zonedb)
	return (1);
	if (!zoneda && zonedb)
	return (-1);

	gota = (zfs_get_type(za) == ZFS_TYPE_FILESYSTEM);
	if (gota) {
	verify(zfs_prop_get(za, ZFS_PROP_MOUNTPOINT, mounta,
	sizeof (mounta), NULL, NULL, 0, B_FALSE) == 0);
	}
	gotb = (zfs_get_type(zb) == ZFS_TYPE_FILESYSTEM);
	if (gotb) {
	verify(zfs_prop_get(zb, ZFS_PROP_MOUNTPOINT, mountb,
	sizeof (mountb), NULL, NULL, 0, B_FALSE) == 0);
	}

	if (gota && gotb) {
	while (a != '\0' && (a == *b)) {
	a++;
	b++;
	}
	if (a == b)
	return (0);
	if (*a == '\0')
	return (-1);
	if (*b == '\0')
	return (1);
	if (*a == '/')
	return (-1);
	if (*b == '/')
	return (1);
	return (a < b ? -1 : a > b);
	}

	if (gota)
	return (-1);
	if (gotb)
	return (1);

	/*
	* If neither filesystem has a mountpoint, revert to sorting by
	* dataset name.
	*/
	return (strcmp(zfs_get_name(za), zfs_get_name(zb)));
	}

	/*
	* Return true if path2 is a child of path1 or path2 equals path1 or
	* path1 is "/" (path2 is always a child of "/").
	*/
	static boolean_t
	libzfs_path_contains(const char path1, const char path2)
	{
	return (strcmp(path1, path2) == 0 \|\| strcmp(path1, "/") == 0 \|\|
	(strstr(path2, path1) == path2 && path2[strlen(path1)] == '/'));
	}

	/*
	* Given a mountpoint specified by idx in the handles array, find the first
	* non-descendent of that mountpoint and return its index. Descendant paths
	* start with the parent's path. This function relies on the ordering
	* enforced by mountpoint_cmp().
	*/
	static int
	non_descendant_idx(zfs_handle_t **handles, size_t num_handles, int idx)
	{
	char parent[ZFS_MAXPROPLEN];
	char child[ZFS_MAXPROPLEN];
	int i;

	verify(zfs_prop_get(handles[idx], ZFS_PROP_MOUNTPOINT, parent,
	sizeof (parent), NULL, NULL, 0, B_FALSE) == 0);

	for (i = idx + 1; i < num_handles; i++) {
	verify(zfs_prop_get(handles[i], ZFS_PROP_MOUNTPOINT, child,
	sizeof (child), NULL, NULL, 0, B_FALSE) == 0);
	if (!libzfs_path_contains(parent, child))
	break;
	}
	return (i);
	}

	typedef struct mnt_param {
	libzfs_handle_t *mnt_hdl;
	tpool_t *mnt_tp;
	zfs_handle_t *mnt_zhps; / filesystems to mount */
	size_t mnt_num_handles;
	int mnt_idx; /* Index of selected entry to mount */
	zfs_iter_f mnt_func;
	void *mnt_data;
	} mnt_param_t;

	/*
	* Allocate and populate the parameter struct for mount function, and
	* schedule mounting of the entry selected by idx.
	*/
	static void
	zfs_dispatch_mount(libzfs_handle_t hdl, zfs_handle_t *handles,
	size_t num_handles, int idx, zfs_iter_f func, void data, tpool_t tp)
	{
	mnt_param_t *mnt_param = zfs_alloc(hdl, sizeof (mnt_param_t));

	mnt_param->mnt_hdl = hdl;
	mnt_param->mnt_tp = tp;
	mnt_param->mnt_zhps = handles;
	mnt_param->mnt_num_handles = num_handles;
	mnt_param->mnt_idx = idx;
	mnt_param->mnt_func = func;
	mnt_param->mnt_data = data;

	- (void) tpool_dispatch(tp, zfs_mount_task, (void*)mnt_param);
	+ if (tpool_dispatch(tp, zfs_mount_task, (void*)mnt_param)) {
	+ /* Could not dispatch to thread pool; execute directly */
	+ zfs_mount_task((void*)mnt_param);
	+ }
	}

	/*
	* This is the structure used to keep state of mounting or sharing operations
	* during a call to zpool_enable_datasets().
	*/
	typedef struct mount_state {
	/*
	* ms_mntstatus is set to -1 if any mount fails. While multiple threads
	* could update this variable concurrently, no synchronization is
	* needed as it's only ever set to -1.
	*/
	int ms_mntstatus;
	int ms_mntflags;
	const char *ms_mntopts;
	} mount_state_t;

	static int
	zfs_mount_one(zfs_handle_t zhp, void arg)
	{
	mount_state_t *ms = arg;
	int ret = 0;

	/*
	* don't attempt to mount encrypted datasets with
	* unloaded keys
	*/
	if (zfs_prop_get_int(zhp, ZFS_PROP_KEYSTATUS) ==
	ZFS_KEYSTATUS_UNAVAILABLE)
	return (0);

	if (zfs_mount(zhp, ms->ms_mntopts, ms->ms_mntflags) != 0)
	ret = ms->ms_mntstatus = -1;
	return (ret);
	}

	static int
	zfs_share_one(zfs_handle_t zhp, void arg)
	{
	mount_state_t *ms = arg;
	int ret = 0;

	if (zfs_share(zhp, NULL) != 0)
	ret = ms->ms_mntstatus = -1;
	return (ret);
	}

	/*
	* Thread pool function to mount one file system. On completion, it finds and
	* schedules its children to be mounted. This depends on the sorting done in
	* zfs_foreach_mountpoint(). Note that the degenerate case (chain of entries
	* each descending from the previous) will have no parallelism since we always
	* have to wait for the parent to finish mounting before we can schedule
	* its children.
	*/
	static void
	zfs_mount_task(void *arg)
	{
	mnt_param_t *mp = arg;
	int idx = mp->mnt_idx;
	zfs_handle_t **handles = mp->mnt_zhps;
	size_t num_handles = mp->mnt_num_handles;
	char mountpoint[ZFS_MAXPROPLEN];

	verify(zfs_prop_get(handles[idx], ZFS_PROP_MOUNTPOINT, mountpoint,
	sizeof (mountpoint), NULL, NULL, 0, B_FALSE) == 0);

	if (mp->mnt_func(handles[idx], mp->mnt_data) != 0)
	goto out;

	/*
	* We dispatch tasks to mount filesystems with mountpoints underneath
	* this one. We do this by dispatching the next filesystem with a
	* descendant mountpoint of the one we just mounted, then skip all of
	* its descendants, dispatch the next descendant mountpoint, and so on.
	* The non_descendant_idx() function skips over filesystems that are
	* descendants of the filesystem we just dispatched.
	*/
	for (int i = idx + 1; i < num_handles;
	i = non_descendant_idx(handles, num_handles, i)) {
	char child[ZFS_MAXPROPLEN];
	verify(zfs_prop_get(handles[i], ZFS_PROP_MOUNTPOINT,
	child, sizeof (child), NULL, NULL, 0, B_FALSE) == 0);

	if (!libzfs_path_contains(mountpoint, child))
	break; /* not a descendant, return */
	zfs_dispatch_mount(mp->mnt_hdl, handles, num_handles, i,
	mp->mnt_func, mp->mnt_data, mp->mnt_tp);
	}

	out:
	free(mp);
	}

	/*
	* Issue the func callback for each ZFS handle contained in the handles
	* array. This function is used to mount all datasets, and so this function
	* guarantees that filesystems for parent mountpoints are called before their
	* children. As such, before issuing any callbacks, we first sort the array
	* of handles by mountpoint.
	*
	* Callbacks are issued in one of two ways:
	*
	- * 1. Sequentially: If the parallel argument is B_FALSE or the ZFS_SERIAL_MOUNT
	+ * 1. Sequentially: If the nthr argument is <= 1 or the ZFS_SERIAL_MOUNT
	* environment variable is set, then we issue callbacks sequentially.
	*
	- * 2. In parallel: If the parallel argument is B_TRUE and the ZFS_SERIAL_MOUNT
	+ * 2. In parallel: If the nthr argument is > 1 and the ZFS_SERIAL_MOUNT
	* environment variable is not set, then we use a tpool to dispatch threads
	* to mount filesystems in parallel. This function dispatches tasks to mount
	* the filesystems at the top-level mountpoints, and these tasks in turn
	* are responsible for recursively mounting filesystems in their children
	- * mountpoints.
	+ * mountpoints. The value of the nthr argument will be the number of worker
	+ * threads for the thread pool.
	*/
	void
	zfs_foreach_mountpoint(libzfs_handle_t hdl, zfs_handle_t *handles,
	- size_t num_handles, zfs_iter_f func, void *data, boolean_t parallel)
	+ size_t num_handles, zfs_iter_f func, void *data, uint_t nthr)
	{
	zoneid_t zoneid = getzoneid();

	/*
	* The ZFS_SERIAL_MOUNT environment variable is an undocumented
	* variable that can be used as a convenience to do a/b comparison
	* of serial vs. parallel mounting.
	*/
	- boolean_t serial_mount = !parallel \|\|
	+ boolean_t serial_mount = nthr <= 1 \|\|
	(getenv("ZFS_SERIAL_MOUNT") != NULL);

	/*
	* Sort the datasets by mountpoint. See mountpoint_cmp for details
	* of how these are sorted.
	*/
	qsort(handles, num_handles, sizeof (zfs_handle_t *), mountpoint_cmp);

	if (serial_mount) {
	for (int i = 0; i < num_handles; i++) {
	func(handles[i], data);
	}
	return;
	}

	/*
	* Issue the callback function for each dataset using a parallel
	* algorithm that uses a thread pool to manage threads.
	*/
	- tpool_t *tp = tpool_create(1, mount_tp_nthr, 0, NULL);
	+ tpool_t *tp = tpool_create(1, nthr, 0, NULL);

	/*
	* There may be multiple "top level" mountpoints outside of the pool's
	* root mountpoint, e.g.: /foo /bar. Dispatch a mount task for each of
	* these.
	*/
	for (int i = 0; i < num_handles;
	i = non_descendant_idx(handles, num_handles, i)) {
	/*
	* Since the mountpoints have been sorted so that the zoned
	* filesystems are at the end, a zoned filesystem seen from
	* the global zone means that we're done.
	*/
	if (zoneid == GLOBAL_ZONEID &&
	zfs_prop_get_int(handles[i], ZFS_PROP_ZONED))
	break;
	zfs_dispatch_mount(hdl, handles, num_handles, i, func, data,
	tp);
	}

	tpool_wait(tp); /* wait for all scheduled mounts to complete */
	tpool_destroy(tp);
	}

	/*
	* Mount and share all datasets within the given pool. This assumes that no
	- * datasets within the pool are currently mounted.
	+ * datasets within the pool are currently mounted. nthr will be number of
	+ * worker threads to use while mounting datasets.
	*/
	int
	-zpool_enable_datasets(zpool_handle_t zhp, const char mntopts, int flags)
	+zpool_enable_datasets(zpool_handle_t zhp, const char mntopts, int flags,
	+ uint_t nthr)
	{
	get_all_cb_t cb = { 0 };
	mount_state_t ms = { 0 };
	zfs_handle_t *zfsp;
	int ret = 0;

	if ((zfsp = zfs_open(zhp->zpool_hdl, zhp->zpool_name,
	ZFS_TYPE_DATASET)) == NULL)
	goto out;

	/*
	* Gather all non-snapshot datasets within the pool. Start by adding
	* the root filesystem for this pool to the list, and then iterate
	* over all child filesystems.
	*/
	libzfs_add_handle(&cb, zfsp);
	if (zfs_iter_filesystems_v2(zfsp, 0, zfs_iter_cb, &cb) != 0)
	goto out;

	/*
	* Mount all filesystems
	*/
	ms.ms_mntopts = mntopts;
	ms.ms_mntflags = flags;
	zfs_foreach_mountpoint(zhp->zpool_hdl, cb.cb_handles, cb.cb_used,
	- zfs_mount_one, &ms, B_TRUE);
	+ zfs_mount_one, &ms, nthr);
	if (ms.ms_mntstatus != 0)
	ret = EZFS_MOUNTFAILED;

	/*
	* Share all filesystems that need to be shared. This needs to be
	* a separate pass because libshare is not mt-safe, and so we need
	* to share serially.
	*/
	ms.ms_mntstatus = 0;
	zfs_foreach_mountpoint(zhp->zpool_hdl, cb.cb_handles, cb.cb_used,
	- zfs_share_one, &ms, B_FALSE);
	+ zfs_share_one, &ms, 1);
	if (ms.ms_mntstatus != 0)
	ret = EZFS_SHAREFAILED;
	else
	zfs_commit_shares(NULL);

	out:
	for (int i = 0; i < cb.cb_used; i++)
	zfs_close(cb.cb_handles[i]);
	free(cb.cb_handles);

	return (ret);
	}

	struct sets_s {
	char *mountpoint;
	zfs_handle_t *dataset;
	};

	static int
	mountpoint_compare(const void a, const void b)
	{
	const struct sets_s mounta = (struct sets_s )a;
	const struct sets_s mountb = (struct sets_s )b;

	return (strcmp(mountb->mountpoint, mounta->mountpoint));
	}

	/*
	* Unshare and unmount all datasets within the given pool. We don't want to
	* rely on traversing the DSL to discover the filesystems within the pool,
	* because this may be expensive (if not all of them are mounted), and can fail
	* arbitrarily (on I/O error, for example). Instead, we walk /proc/self/mounts
	* and gather all the filesystems that are currently mounted.
	*/
	int
	zpool_disable_datasets(zpool_handle_t *zhp, boolean_t force)
	{
	int used, alloc;
	FILE *mnttab;
	struct mnttab entry;
	size_t namelen;
	struct sets_s *sets = NULL;
	libzfs_handle_t *hdl = zhp->zpool_hdl;
	int i;
	int ret = -1;
	int flags = (force ? MS_FORCE : 0);

	namelen = strlen(zhp->zpool_name);

	if ((mnttab = fopen(MNTTAB, "re")) == NULL)
	return (ENOENT);

	used = alloc = 0;
	while (getmntent(mnttab, &entry) == 0) {
	/*
	* Ignore non-ZFS entries.
	*/
	if (entry.mnt_fstype == NULL \|\|
	strcmp(entry.mnt_fstype, MNTTYPE_ZFS) != 0)
	continue;

	/*
	* Ignore filesystems not within this pool.
	*/
	if (entry.mnt_mountp == NULL \|\|
	strncmp(entry.mnt_special, zhp->zpool_name, namelen) != 0 \|\|
	(entry.mnt_special[namelen] != '/' &&
	entry.mnt_special[namelen] != '\0'))
	continue;

	/*
	* At this point we've found a filesystem within our pool. Add
	* it to our growing list.
	*/
	if (used == alloc) {
	if (alloc == 0) {
	sets = zfs_alloc(hdl,
	8 * sizeof (struct sets_s));
	alloc = 8;
	} else {
	sets = zfs_realloc(hdl, sets,
	alloc * sizeof (struct sets_s),
	alloc * 2 * sizeof (struct sets_s));

	alloc *= 2;
	}
	}

	sets[used].mountpoint = zfs_strdup(hdl, entry.mnt_mountp);

	/*
	* This is allowed to fail, in case there is some I/O error. It
	* is only used to determine if we need to remove the underlying
	* mountpoint, so failure is not fatal.
	*/
	sets[used].dataset = make_dataset_handle(hdl,
	entry.mnt_special);

	used++;
	}

	/*
	* At this point, we have the entire list of filesystems, so sort it by
	* mountpoint.
	*/
	if (used != 0)
	qsort(sets, used, sizeof (struct sets_s), mountpoint_compare);

	/*
	* Walk through and first unshare everything.
	*/
	for (i = 0; i < used; i++) {
	for (enum sa_protocol p = 0; p < SA_PROTOCOL_COUNT; ++p) {
	if (sa_is_shared(sets[i].mountpoint, p) &&
	unshare_one(hdl, sets[i].mountpoint,
	sets[i].mountpoint, p) != 0)
	goto out;
	}
	}
	zfs_commit_shares(NULL);

	/*
	* Now unmount everything, removing the underlying directories as
	* appropriate.
	*/
	for (i = 0; i < used; i++) {
	if (unmount_one(sets[i].dataset, sets[i].mountpoint,
	flags) != 0)
	goto out;
	}

	for (i = 0; i < used; i++) {
	if (sets[i].dataset)
	remove_mountpoint(sets[i].dataset);
	}

	zpool_disable_datasets_os(zhp, force);

	ret = 0;
	out:
	(void) fclose(mnttab);
	for (i = 0; i < used; i++) {
	if (sets[i].dataset)
	zfs_close(sets[i].dataset);
	free(sets[i].mountpoint);
	}
	free(sets);

	return (ret);
	}
	diff --git a/sys/contrib/openzfs/lib/libzfs/libzfs_sendrecv.c b/sys/contrib/openzfs/lib/libzfs/libzfs_sendrecv.c
	index 526f57ea403c..0370112c022a 100644
	--- a/sys/contrib/openzfs/lib/libzfs/libzfs_sendrecv.c
	+++ b/sys/contrib/openzfs/lib/libzfs/libzfs_sendrecv.c
	@@ -1,5626 +1,5628 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
	* Copyright (c) 2012, Joyent, Inc. All rights reserved.
	* Copyright (c) 2012 Pawel Jakub Dawidek <pawel@dawidek.net>.
	* All rights reserved
	* Copyright (c) 2013 Steven Hartland. All rights reserved.
	* Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved.
	* Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>
	* Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
	* Copyright (c) 2019 Datto Inc.
	*/

	#include <assert.h>
	#include <ctype.h>
	#include <errno.h>
	#include <libintl.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <string.h>
	#include <unistd.h>
	#include <stddef.h>
	#include <fcntl.h>
	#include <sys/mount.h>
	#include <sys/mntent.h>
	#include <sys/mnttab.h>
	#include <sys/avl.h>
	#include <sys/debug.h>
	#include <sys/stat.h>
	#include <pthread.h>
	#include <umem.h>
	#include <time.h>

	#include <libzfs.h>
	#include <libzfs_core.h>
	#include <libzutil.h>

	#include "zfs_namecheck.h"
	#include "zfs_prop.h"
	#include "zfs_fletcher.h"
	#include "libzfs_impl.h"
	#include <cityhash.h>
	#include <zlib.h>
	#include <sys/zio_checksum.h>
	#include <sys/dsl_crypt.h>
	#include <sys/ddt.h>
	#include <sys/socket.h>
	#include <sys/sha2.h>

	static int zfs_receive_impl(libzfs_handle_t , const char , const char *,
	recvflags_t , int, const char , nvlist_t , avl_tree_t , char **,
	const char , nvlist_t );
	static int guid_to_name_redact_snaps(libzfs_handle_t hdl, const char parent,
	uint64_t guid, boolean_t bookmark_ok, uint64_t *redact_snap_guids,
	uint64_t num_redact_snaps, char *name);
	static int guid_to_name(libzfs_handle_t , const char ,
	uint64_t, boolean_t, char *);

	typedef struct progress_arg {
	zfs_handle_t *pa_zhp;
	int pa_fd;
	boolean_t pa_parsable;
	boolean_t pa_estimate;
	int pa_verbosity;
	boolean_t pa_astitle;
	boolean_t pa_progress;
	uint64_t pa_size;
	} progress_arg_t;

	static int
	dump_record(dmu_replay_record_t drr, void payload, size_t payload_len,
	zio_cksum_t *zc, int outfd)
	{
	ASSERT3U(offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum),
	==, sizeof (dmu_replay_record_t) - sizeof (zio_cksum_t));
	fletcher_4_incremental_native(drr,
	offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum), zc);
	if (drr->drr_type != DRR_BEGIN) {
	ASSERT(ZIO_CHECKSUM_IS_ZERO(&drr->drr_u.
	drr_checksum.drr_checksum));
	drr->drr_u.drr_checksum.drr_checksum = *zc;
	}
	fletcher_4_incremental_native(&drr->drr_u.drr_checksum.drr_checksum,
	sizeof (zio_cksum_t), zc);
	if (write(outfd, drr, sizeof (*drr)) == -1)
	return (errno);
	if (payload_len != 0) {
	fletcher_4_incremental_native(payload, payload_len, zc);
	if (write(outfd, payload, payload_len) == -1)
	return (errno);
	}
	return (0);
	}

	/*
	* Routines for dealing with the AVL tree of fs-nvlists
	*/
	typedef struct fsavl_node {
	avl_node_t fn_node;
	nvlist_t *fn_nvfs;
	const char *fn_snapname;
	uint64_t fn_guid;
	} fsavl_node_t;

	static int
	fsavl_compare(const void arg1, const void arg2)
	{
	const fsavl_node_t fn1 = (const fsavl_node_t )arg1;
	const fsavl_node_t fn2 = (const fsavl_node_t )arg2;

	return (TREE_CMP(fn1->fn_guid, fn2->fn_guid));
	}

	/*
	* Given the GUID of a snapshot, find its containing filesystem and
	* (optionally) name.
	*/
	static nvlist_t *
	fsavl_find(avl_tree_t avl, uint64_t snapguid, const char *snapname)
	{
	fsavl_node_t fn_find;
	fsavl_node_t *fn;

	fn_find.fn_guid = snapguid;

	fn = avl_find(avl, &fn_find, NULL);
	if (fn) {
	if (snapname)
	*snapname = fn->fn_snapname;
	return (fn->fn_nvfs);
	}
	return (NULL);
	}

	static void
	fsavl_destroy(avl_tree_t *avl)
	{
	fsavl_node_t *fn;
	void *cookie;

	if (avl == NULL)
	return;

	cookie = NULL;
	while ((fn = avl_destroy_nodes(avl, &cookie)) != NULL)
	free(fn);
	avl_destroy(avl);
	free(avl);
	}

	/*
	* Given an nvlist, produce an avl tree of snapshots, ordered by guid
	*/
	static avl_tree_t *
	fsavl_create(nvlist_t *fss)
	{
	avl_tree_t *fsavl;
	nvpair_t *fselem = NULL;

	if ((fsavl = malloc(sizeof (avl_tree_t))) == NULL)
	return (NULL);

	avl_create(fsavl, fsavl_compare, sizeof (fsavl_node_t),
	offsetof(fsavl_node_t, fn_node));

	while ((fselem = nvlist_next_nvpair(fss, fselem)) != NULL) {
	nvlist_t nvfs, snaps;
	nvpair_t *snapelem = NULL;

	nvfs = fnvpair_value_nvlist(fselem);
	snaps = fnvlist_lookup_nvlist(nvfs, "snaps");

	while ((snapelem =
	nvlist_next_nvpair(snaps, snapelem)) != NULL) {
	fsavl_node_t *fn;

	if ((fn = malloc(sizeof (fsavl_node_t))) == NULL) {
	fsavl_destroy(fsavl);
	return (NULL);
	}
	fn->fn_nvfs = nvfs;
	fn->fn_snapname = nvpair_name(snapelem);
	fn->fn_guid = fnvpair_value_uint64(snapelem);

	/*
	* Note: if there are multiple snaps with the
	* same GUID, we ignore all but one.
	*/
	avl_index_t where = 0;
	if (avl_find(fsavl, fn, &where) == NULL)
	avl_insert(fsavl, fn, where);
	else
	free(fn);
	}
	}

	return (fsavl);
	}

	/*
	* Routines for dealing with the giant nvlist of fs-nvlists, etc.
	*/
	typedef struct send_data {
	/*
	* assigned inside every recursive call,
	* restored from *_save on return:
	*
	* guid of fromsnap snapshot in parent dataset
	* txg of fromsnap snapshot in current dataset
	* txg of tosnap snapshot in current dataset
	*/

	uint64_t parent_fromsnap_guid;
	uint64_t fromsnap_txg;
	uint64_t tosnap_txg;

	/* the nvlists get accumulated during depth-first traversal */
	nvlist_t *parent_snaps;
	nvlist_t *fss;
	nvlist_t *snapprops;
	nvlist_t snapholds; / user holds */

	/* send-receive configuration, does not change during traversal */
	const char *fsname;
	const char *fromsnap;
	const char *tosnap;
	boolean_t recursive;
	boolean_t raw;
	boolean_t doall;
	boolean_t replicate;
	boolean_t skipmissing;
	boolean_t verbose;
	boolean_t backup;
	boolean_t seenfrom;
	boolean_t seento;
	boolean_t holds; /* were holds requested with send -h */
	boolean_t props;

	/*
	* The header nvlist is of the following format:
	* {
	* "tosnap" -> string
	* "fromsnap" -> string (if incremental)
	* "fss" -> {
	* id -> {
	*
	* "name" -> string (full name; for debugging)
	* "parentfromsnap" -> number (guid of fromsnap in parent)
	*
	* "props" -> { name -> value (only if set here) }
	* "snaps" -> { name (lastname) -> number (guid) }
	* "snapprops" -> { name (lastname) -> { name -> value } }
	* "snapholds" -> { name (lastname) -> { holdname -> crtime } }
	*
	* "origin" -> number (guid) (if clone)
	* "is_encroot" -> boolean
	* "sent" -> boolean (not on-disk)
	* }
	* }
	* }
	*
	*/
	} send_data_t;

	static void
	send_iterate_prop(zfs_handle_t zhp, boolean_t received_only, nvlist_t nv);

	/*
	* Collect guid, valid props, optionally holds, etc. of a snapshot.
	* This interface is intended for use as a zfs_iter_snapshots_v2_sorted visitor.
	*/
	static int
	send_iterate_snap(zfs_handle_t zhp, void arg)
	{
	send_data_t *sd = arg;
	uint64_t guid = zhp->zfs_dmustats.dds_guid;
	uint64_t txg = zhp->zfs_dmustats.dds_creation_txg;
	boolean_t isfromsnap, istosnap, istosnapwithnofrom;
	char *snapname;
	const char *from = sd->fromsnap;
	const char *to = sd->tosnap;

	snapname = strrchr(zhp->zfs_name, '@');
	assert(snapname != NULL);
	++snapname;

	isfromsnap = (from != NULL && strcmp(from, snapname) == 0);
	istosnap = (to != NULL && strcmp(to, snapname) == 0);
	istosnapwithnofrom = (istosnap && from == NULL);

	if (sd->tosnap_txg != 0 && txg > sd->tosnap_txg) {
	if (sd->verbose) {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"skipping snapshot %s because it was created "
	"after the destination snapshot (%s)\n"),
	zhp->zfs_name, to);
	}
	zfs_close(zhp);
	return (0);
	}

	fnvlist_add_uint64(sd->parent_snaps, snapname, guid);

	/*
	* NB: if there is no fromsnap here (it's a newly created fs in
	* an incremental replication), we will substitute the tosnap.
	*/
	if (isfromsnap \|\| (sd->parent_fromsnap_guid == 0 && istosnap))
	sd->parent_fromsnap_guid = guid;

	if (!sd->recursive) {
	/*
	* To allow a doall stream to work properly
	* with a NULL fromsnap
	*/
	if (sd->doall && from == NULL && !sd->seenfrom)
	sd->seenfrom = B_TRUE;

	if (!sd->seenfrom && isfromsnap) {
	sd->seenfrom = B_TRUE;
	zfs_close(zhp);
	return (0);
	}

	if ((sd->seento \|\| !sd->seenfrom) && !istosnapwithnofrom) {
	zfs_close(zhp);
	return (0);
	}

	if (istosnap)
	sd->seento = B_TRUE;
	}

	nvlist_t *nv = fnvlist_alloc();
	send_iterate_prop(zhp, sd->backup, nv);
	fnvlist_add_nvlist(sd->snapprops, snapname, nv);
	fnvlist_free(nv);

	if (sd->holds) {
	nvlist_t *holds;
	if (lzc_get_holds(zhp->zfs_name, &holds) == 0) {
	fnvlist_add_nvlist(sd->snapholds, snapname, holds);
	fnvlist_free(holds);
	}
	}

	zfs_close(zhp);
	return (0);
	}

	/*
	* Collect all valid props from the handle snap into an nvlist.
	*/
	static void
	send_iterate_prop(zfs_handle_t zhp, boolean_t received_only, nvlist_t nv)
	{
	nvlist_t *props;

	if (received_only)
	props = zfs_get_recvd_props(zhp);
	else
	props = zhp->zfs_props;

	nvpair_t *elem = NULL;
	while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
	const char *propname = nvpair_name(elem);
	zfs_prop_t prop = zfs_name_to_prop(propname);

	if (!zfs_prop_user(propname)) {
	/*
	* Realistically, this should never happen. However,
	* we want the ability to add DSL properties without
	* needing to make incompatible version changes. We
	* need to ignore unknown properties to allow older
	* software to still send datasets containing these
	* properties, with the unknown properties elided.
	*/
	if (prop == ZPROP_INVAL)
	continue;

	if (zfs_prop_readonly(prop))
	continue;
	}

	nvlist_t *propnv = fnvpair_value_nvlist(elem);

	boolean_t isspacelimit = (prop == ZFS_PROP_QUOTA \|\|
	prop == ZFS_PROP_RESERVATION \|\|
	prop == ZFS_PROP_REFQUOTA \|\|
	prop == ZFS_PROP_REFRESERVATION);
	if (isspacelimit && zhp->zfs_type == ZFS_TYPE_SNAPSHOT)
	continue;

	const char *source;
	if (nvlist_lookup_string(propnv, ZPROP_SOURCE, &source) == 0) {
	if (strcmp(source, zhp->zfs_name) != 0 &&
	strcmp(source, ZPROP_SOURCE_VAL_RECVD) != 0)
	continue;
	} else {
	/*
	* May have no source before SPA_VERSION_RECVD_PROPS,
	* but is still modifiable.
	*/
	if (!isspacelimit)
	continue;
	}

	if (zfs_prop_user(propname) \|\|
	zfs_prop_get_type(prop) == PROP_TYPE_STRING) {
	const char *value;
	value = fnvlist_lookup_string(propnv, ZPROP_VALUE);
	fnvlist_add_string(nv, propname, value);
	} else {
	uint64_t value;
	value = fnvlist_lookup_uint64(propnv, ZPROP_VALUE);
	fnvlist_add_uint64(nv, propname, value);
	}
	}
	}

	/*
	* returns snapshot guid
	* and returns 0 if the snapshot does not exist
	*/
	static uint64_t
	get_snap_guid(libzfs_handle_t hdl, const char fs, const char *snap)
	{
	char name[MAXPATHLEN + 1];
	uint64_t guid = 0;

	if (fs == NULL \|\| fs[0] == '\0' \|\| snap == NULL \|\| snap[0] == '\0')
	return (guid);

	(void) snprintf(name, sizeof (name), "%s@%s", fs, snap);
	zfs_handle_t *zhp = zfs_open(hdl, name, ZFS_TYPE_SNAPSHOT);
	if (zhp != NULL) {
	guid = zfs_prop_get_int(zhp, ZFS_PROP_GUID);
	zfs_close(zhp);
	}

	return (guid);
	}

	/*
	* returns snapshot creation txg
	* and returns 0 if the snapshot does not exist
	*/
	static uint64_t
	get_snap_txg(libzfs_handle_t hdl, const char fs, const char *snap)
	{
	char name[ZFS_MAX_DATASET_NAME_LEN];
	uint64_t txg = 0;

	if (fs == NULL \|\| fs[0] == '\0' \|\| snap == NULL \|\| snap[0] == '\0')
	return (txg);

	(void) snprintf(name, sizeof (name), "%s@%s", fs, snap);
	if (zfs_dataset_exists(hdl, name, ZFS_TYPE_SNAPSHOT)) {
	zfs_handle_t *zhp = zfs_open(hdl, name, ZFS_TYPE_SNAPSHOT);
	if (zhp != NULL) {
	txg = zfs_prop_get_int(zhp, ZFS_PROP_CREATETXG);
	zfs_close(zhp);
	}
	}

	return (txg);
	}

	/*
	* Recursively generate nvlists describing datasets. See comment
	* for the data structure send_data_t above for description of contents
	* of the nvlist.
	*/
	static int
	send_iterate_fs(zfs_handle_t zhp, void arg)
	{
	send_data_t *sd = arg;
	nvlist_t nvfs = NULL, nv = NULL;
	int rv = 0;
	uint64_t min_txg = 0, max_txg = 0;
	uint64_t txg = zhp->zfs_dmustats.dds_creation_txg;
	uint64_t guid = zhp->zfs_dmustats.dds_guid;
	uint64_t fromsnap_txg, tosnap_txg;
	char guidstring[64];

	/* These fields are restored on return from a recursive call. */
	uint64_t parent_fromsnap_guid_save = sd->parent_fromsnap_guid;
	uint64_t fromsnap_txg_save = sd->fromsnap_txg;
	uint64_t tosnap_txg_save = sd->tosnap_txg;

	fromsnap_txg = get_snap_txg(zhp->zfs_hdl, zhp->zfs_name, sd->fromsnap);
	if (fromsnap_txg != 0)
	sd->fromsnap_txg = fromsnap_txg;

	tosnap_txg = get_snap_txg(zhp->zfs_hdl, zhp->zfs_name, sd->tosnap);
	if (tosnap_txg != 0)
	sd->tosnap_txg = tosnap_txg;

	/*
	* On the send side, if the current dataset does not have tosnap,
	* perform two additional checks:
	*
	* - Skip sending the current dataset if it was created later than
	* the parent tosnap.
	* - Return error if the current dataset was created earlier than
	* the parent tosnap, unless --skip-missing specified. Then
	* just print a warning.
	*/
	if (sd->tosnap != NULL && tosnap_txg == 0) {
	if (sd->tosnap_txg != 0 && txg > sd->tosnap_txg) {
	if (sd->verbose) {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"skipping dataset %s: snapshot %s does "
	"not exist\n"), zhp->zfs_name, sd->tosnap);
	}
	} else if (sd->skipmissing) {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"WARNING: skipping dataset %s and its children:"
	" snapshot %s does not exist\n"),
	zhp->zfs_name, sd->tosnap);
	} else {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"cannot send %s@%s%s: snapshot %s@%s does not "
	"exist\n"), sd->fsname, sd->tosnap, sd->recursive ?
	dgettext(TEXT_DOMAIN, " recursively") : "",
	zhp->zfs_name, sd->tosnap);
	rv = EZFS_NOENT;
	}
	goto out;
	}

	nvfs = fnvlist_alloc();
	fnvlist_add_string(nvfs, "name", zhp->zfs_name);
	fnvlist_add_uint64(nvfs, "parentfromsnap", sd->parent_fromsnap_guid);

	if (zhp->zfs_dmustats.dds_origin[0] != '\0') {
	zfs_handle_t *origin = zfs_open(zhp->zfs_hdl,
	zhp->zfs_dmustats.dds_origin, ZFS_TYPE_SNAPSHOT);
	if (origin == NULL) {
	rv = -1;
	goto out;
	}
	fnvlist_add_uint64(nvfs, "origin",
	origin->zfs_dmustats.dds_guid);
	zfs_close(origin);
	}

	/* Iterate over props. */
	if (sd->props \|\| sd->backup \|\| sd->recursive) {
	nv = fnvlist_alloc();
	send_iterate_prop(zhp, sd->backup, nv);
	fnvlist_add_nvlist(nvfs, "props", nv);
	}
	if (zfs_prop_get_int(zhp, ZFS_PROP_ENCRYPTION) != ZIO_CRYPT_OFF) {
	boolean_t encroot;

	/* Determine if this dataset is an encryption root. */
	if (zfs_crypto_get_encryption_root(zhp, &encroot, NULL) != 0) {
	rv = -1;
	goto out;
	}

	if (encroot)
	fnvlist_add_boolean(nvfs, "is_encroot");

	/*
	* Encrypted datasets can only be sent with properties if
	* the raw flag is specified because the receive side doesn't
	* currently have a mechanism for recursively asking the user
	* for new encryption parameters.
	*/
	if (!sd->raw) {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"cannot send %s@%s: encrypted dataset %s may not "
	"be sent with properties without the raw flag\n"),
	sd->fsname, sd->tosnap, zhp->zfs_name);
	rv = -1;
	goto out;
	}

	}

	/*
	* Iterate over snaps, and set sd->parent_fromsnap_guid.
	*
	* If this is a "doall" send, a replicate send or we're just trying
	* to gather a list of previous snapshots, iterate through all the
	* snaps in the txg range. Otherwise just look at the one we're
	* interested in.
	*/
	sd->parent_fromsnap_guid = 0;
	sd->parent_snaps = fnvlist_alloc();
	sd->snapprops = fnvlist_alloc();
	if (sd->holds)
	sd->snapholds = fnvlist_alloc();
	if (sd->doall \|\| sd->replicate \|\| sd->tosnap == NULL) {
	if (!sd->replicate && fromsnap_txg != 0)
	min_txg = fromsnap_txg;
	if (!sd->replicate && tosnap_txg != 0)
	max_txg = tosnap_txg;
	(void) zfs_iter_snapshots_sorted_v2(zhp, 0, send_iterate_snap,
	sd, min_txg, max_txg);
	} else {
	char snapname[MAXPATHLEN] = { 0 };
	zfs_handle_t *snap;

	(void) snprintf(snapname, sizeof (snapname), "%s@%s",
	zhp->zfs_name, sd->tosnap);
	if (sd->fromsnap != NULL)
	sd->seenfrom = B_TRUE;
	snap = zfs_open(zhp->zfs_hdl, snapname, ZFS_TYPE_SNAPSHOT);
	if (snap != NULL)
	(void) send_iterate_snap(snap, sd);
	}

	fnvlist_add_nvlist(nvfs, "snaps", sd->parent_snaps);
	fnvlist_free(sd->parent_snaps);
	fnvlist_add_nvlist(nvfs, "snapprops", sd->snapprops);
	fnvlist_free(sd->snapprops);
	if (sd->holds) {
	fnvlist_add_nvlist(nvfs, "snapholds", sd->snapholds);
	fnvlist_free(sd->snapholds);
	}

	/* Do not allow the size of the properties list to exceed the limit */
	if ((fnvlist_size(nvfs) + fnvlist_size(sd->fss)) >
	zhp->zfs_hdl->libzfs_max_nvlist) {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"warning: cannot send %s@%s: the size of the list of "
	"snapshots and properties is too large to be received "
	"successfully.\n"
	"Select a smaller number of snapshots to send.\n"),
	zhp->zfs_name, sd->tosnap);
	rv = EZFS_NOSPC;
	goto out;
	}
	/* Add this fs to nvlist. */
	(void) snprintf(guidstring, sizeof (guidstring),
	"0x%llx", (longlong_t)guid);
	fnvlist_add_nvlist(sd->fss, guidstring, nvfs);

	/* Iterate over children. */
	if (sd->recursive)
	rv = zfs_iter_filesystems_v2(zhp, 0, send_iterate_fs, sd);

	out:
	/* Restore saved fields. */
	sd->parent_fromsnap_guid = parent_fromsnap_guid_save;
	sd->fromsnap_txg = fromsnap_txg_save;
	sd->tosnap_txg = tosnap_txg_save;

	fnvlist_free(nv);
	fnvlist_free(nvfs);

	zfs_close(zhp);
	return (rv);
	}

	static int
	gather_nvlist(libzfs_handle_t hdl, const char fsname, const char *fromsnap,
	const char *tosnap, boolean_t recursive, boolean_t raw, boolean_t doall,
	boolean_t replicate, boolean_t skipmissing, boolean_t verbose,
	boolean_t backup, boolean_t holds, boolean_t props, nvlist_t **nvlp,
	avl_tree_t **avlp)
	{
	zfs_handle_t *zhp;
	send_data_t sd = { 0 };
	int error;

	zhp = zfs_open(hdl, fsname, ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME);
	if (zhp == NULL)
	return (EZFS_BADTYPE);

	sd.fss = fnvlist_alloc();
	sd.fsname = fsname;
	sd.fromsnap = fromsnap;
	sd.tosnap = tosnap;
	sd.recursive = recursive;
	sd.raw = raw;
	sd.doall = doall;
	sd.replicate = replicate;
	sd.skipmissing = skipmissing;
	sd.verbose = verbose;
	sd.backup = backup;
	sd.holds = holds;
	sd.props = props;

	if ((error = send_iterate_fs(zhp, &sd)) != 0) {
	fnvlist_free(sd.fss);
	if (avlp != NULL)
	*avlp = NULL;
	*nvlp = NULL;
	return (error);
	}

	if (avlp != NULL && (*avlp = fsavl_create(sd.fss)) == NULL) {
	fnvlist_free(sd.fss);
	*nvlp = NULL;
	return (EZFS_NOMEM);
	}

	*nvlp = sd.fss;
	return (0);
	}

	/*
	* Routines specific to "zfs send"
	*/
	typedef struct send_dump_data {
	/* these are all just the short snapname (the part after the @) */
	const char *fromsnap;
	const char *tosnap;
	char prevsnap[ZFS_MAX_DATASET_NAME_LEN];
	uint64_t prevsnap_obj;
	boolean_t seenfrom, seento, replicate, doall, fromorigin;
	boolean_t dryrun, parsable, progress, embed_data, std_out;
	boolean_t large_block, compress, raw, holds;
	boolean_t progressastitle;
	int outfd;
	boolean_t err;
	nvlist_t *fss;
	nvlist_t *snapholds;
	avl_tree_t *fsavl;
	snapfilter_cb_t *filter_cb;
	void *filter_cb_arg;
	nvlist_t *debugnv;
	char holdtag[ZFS_MAX_DATASET_NAME_LEN];
	int cleanup_fd;
	int verbosity;
	uint64_t size;
	} send_dump_data_t;

	static int
	zfs_send_space(zfs_handle_t zhp, const char snapname, const char *from,
	enum lzc_send_flags flags, uint64_t *spacep)
	{
	assert(snapname != NULL);

	int error = lzc_send_space(snapname, from, flags, spacep);
	if (error == 0)
	return (0);

	char errbuf[ERRBUFLEN];
	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"warning: cannot estimate space for '%s'"), snapname);

	libzfs_handle_t *hdl = zhp->zfs_hdl;
	switch (error) {
	case EXDEV:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"not an earlier snapshot from the same fs"));
	return (zfs_error(hdl, EZFS_CROSSTARGET, errbuf));

	case ENOENT:
	if (zfs_dataset_exists(hdl, snapname,
	ZFS_TYPE_SNAPSHOT)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"incremental source (%s) does not exist"),
	snapname);
	}
	return (zfs_error(hdl, EZFS_NOENT, errbuf));

	case EDQUOT:
	case EFBIG:
	case EIO:
	case ENOLINK:
	case ENOSPC:
	case ENOSTR:
	case ENXIO:
	case EPIPE:
	case ERANGE:
	case EFAULT:
	case EROFS:
	case EINVAL:
	zfs_error_aux(hdl, "%s", zfs_strerror(error));
	return (zfs_error(hdl, EZFS_BADBACKUP, errbuf));

	default:
	return (zfs_standard_error(hdl, error, errbuf));
	}
	}

	/*
	* Dumps a backup of the given snapshot (incremental from fromsnap if it's not
	* NULL) to the file descriptor specified by outfd.
	*/
	static int
	dump_ioctl(zfs_handle_t zhp, const char fromsnap, uint64_t fromsnap_obj,
	boolean_t fromorigin, int outfd, enum lzc_send_flags flags,
	nvlist_t *debugnv)
	{
	zfs_cmd_t zc = {"\0"};
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	nvlist_t *thisdbg;

	assert(zhp->zfs_type == ZFS_TYPE_SNAPSHOT);
	assert(fromsnap_obj == 0 \|\| !fromorigin);

	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
	zc.zc_cookie = outfd;
	zc.zc_obj = fromorigin;
	zc.zc_sendobj = zfs_prop_get_int(zhp, ZFS_PROP_OBJSETID);
	zc.zc_fromobj = fromsnap_obj;
	zc.zc_flags = flags;

	if (debugnv != NULL) {
	thisdbg = fnvlist_alloc();
	if (fromsnap != NULL && fromsnap[0] != '\0')
	fnvlist_add_string(thisdbg, "fromsnap", fromsnap);
	}

	if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_SEND, &zc) != 0) {
	char errbuf[ERRBUFLEN];
	int error = errno;

	(void) snprintf(errbuf, sizeof (errbuf), "%s '%s'",
	dgettext(TEXT_DOMAIN, "warning: cannot send"),
	zhp->zfs_name);

	if (debugnv != NULL) {
	fnvlist_add_uint64(thisdbg, "error", error);
	fnvlist_add_nvlist(debugnv, zhp->zfs_name, thisdbg);
	fnvlist_free(thisdbg);
	}

	switch (error) {
	case EXDEV:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"not an earlier snapshot from the same fs"));
	return (zfs_error(hdl, EZFS_CROSSTARGET, errbuf));

	case EACCES:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"source key must be loaded"));
	return (zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf));

	case ENOENT:
	if (zfs_dataset_exists(hdl, zc.zc_name,
	ZFS_TYPE_SNAPSHOT)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"incremental source (@%s) does not exist"),
	zc.zc_value);
	}
	return (zfs_error(hdl, EZFS_NOENT, errbuf));

	case EDQUOT:
	case EFBIG:
	case EIO:
	case ENOLINK:
	case ENOSPC:
	case ENOSTR:
	case ENXIO:
	case EPIPE:
	case ERANGE:
	case EFAULT:
	case EROFS:
	case EINVAL:
	zfs_error_aux(hdl, "%s", zfs_strerror(errno));
	return (zfs_error(hdl, EZFS_BADBACKUP, errbuf));

	default:
	return (zfs_standard_error(hdl, errno, errbuf));
	}
	}

	if (debugnv != NULL) {
	fnvlist_add_nvlist(debugnv, zhp->zfs_name, thisdbg);
	fnvlist_free(thisdbg);
	}

	return (0);
	}

	static void
	gather_holds(zfs_handle_t zhp, send_dump_data_t sdd)
	{
	assert(zhp->zfs_type == ZFS_TYPE_SNAPSHOT);

	/*
	* zfs_send() only sets snapholds for sends that need them,
	* e.g. replication and doall.
	*/
	if (sdd->snapholds == NULL)
	return;

	fnvlist_add_string(sdd->snapholds, zhp->zfs_name, sdd->holdtag);
	}

	int
	zfs_send_progress(zfs_handle_t zhp, int fd, uint64_t bytes_written,
	uint64_t *blocks_visited)
	{
	zfs_cmd_t zc = {"\0"};

	if (bytes_written != NULL)
	*bytes_written = 0;
	if (blocks_visited != NULL)
	*blocks_visited = 0;
	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
	zc.zc_cookie = fd;
	if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_SEND_PROGRESS, &zc) != 0)
	return (errno);
	if (bytes_written != NULL)
	*bytes_written = zc.zc_cookie;
	if (blocks_visited != NULL)
	*blocks_visited = zc.zc_objset_type;
	return (0);
	}

	static volatile boolean_t send_progress_thread_signal_duetotimer;
	static void
	send_progress_thread_act(int sig, siginfo_t info, void ucontext)
	{
	(void) sig, (void) ucontext;
	send_progress_thread_signal_duetotimer = info->si_code == SI_TIMER;
	}

	struct timer_desirability {
	timer_t timer;
	boolean_t desired;
	};
	static void
	timer_delete_cleanup(void *timer)
	{
	struct timer_desirability *td = timer;
	if (td->desired)
	timer_delete(td->timer);
	}

	#ifdef SIGINFO
	#define SEND_PROGRESS_THREAD_PARENT_BLOCK_SIGINFO sigaddset(&new, SIGINFO)
	#else
	#define SEND_PROGRESS_THREAD_PARENT_BLOCK_SIGINFO
	#endif
	#define SEND_PROGRESS_THREAD_PARENT_BLOCK(old) { \
	sigset_t new; \
	sigemptyset(&new); \
	sigaddset(&new, SIGUSR1); \
	SEND_PROGRESS_THREAD_PARENT_BLOCK_SIGINFO; \
	pthread_sigmask(SIG_BLOCK, &new, old); \
	}

	static void *
	send_progress_thread(void *arg)
	{
	progress_arg_t *pa = arg;
	zfs_handle_t *zhp = pa->pa_zhp;
	uint64_t bytes;
	uint64_t blocks;
	uint64_t total = pa->pa_size / 100;
	char buf[16];
	time_t t;
	struct tm tm;
	int err;

	const struct sigaction signal_action =
	{.sa_sigaction = send_progress_thread_act, .sa_flags = SA_SIGINFO};
	struct sigevent timer_cfg =
	{.sigev_notify = SIGEV_SIGNAL, .sigev_signo = SIGUSR1};
	const struct itimerspec timer_time =
	{.it_value = {.tv_sec = 1}, .it_interval = {.tv_sec = 1}};
	struct timer_desirability timer = {};

	sigaction(SIGUSR1, &signal_action, NULL);
	#ifdef SIGINFO
	sigaction(SIGINFO, &signal_action, NULL);
	#endif

	if ((timer.desired = pa->pa_progress \|\| pa->pa_astitle)) {
	if (timer_create(CLOCK_MONOTONIC, &timer_cfg, &timer.timer))
	return ((void *)(uintptr_t)errno);
	(void) timer_settime(timer.timer, 0, &timer_time, NULL);
	}
	pthread_cleanup_push(timer_delete_cleanup, &timer);

	if (!pa->pa_parsable && pa->pa_progress) {
	(void) fprintf(stderr,
	"TIME %s %sSNAPSHOT %s\n",
	pa->pa_estimate ? "BYTES" : " SENT",
	pa->pa_verbosity >= 2 ? " BLOCKS " : "",
	zhp->zfs_name);
	}

	/*
	* Print the progress from ZFS_IOC_SEND_PROGRESS every second.
	*/
	for (;;) {
	pause();
	if ((err = zfs_send_progress(zhp, pa->pa_fd, &bytes,
	&blocks)) != 0) {
	if (err == EINTR \|\| err == ENOENT)
	err = 0;
	pthread_exit(((void *)(uintptr_t)err));
	}

	(void) time(&t);
	localtime_r(&t, &tm);

	if (pa->pa_astitle) {
	char buf_bytes[16];
	char buf_size[16];
	int pct;
	zfs_nicenum(bytes, buf_bytes, sizeof (buf_bytes));
	zfs_nicenum(pa->pa_size, buf_size, sizeof (buf_size));
	pct = (total > 0) ? bytes / total : 100;
	zfs_setproctitle("sending %s (%d%%: %s/%s)",
	zhp->zfs_name, MIN(pct, 100), buf_bytes, buf_size);
	}

	if (pa->pa_verbosity >= 2 && pa->pa_parsable) {
	(void) fprintf(stderr,
	"%02d:%02d:%02d\t%llu\t%llu\t%s\n",
	tm.tm_hour, tm.tm_min, tm.tm_sec,
	(u_longlong_t)bytes, (u_longlong_t)blocks,
	zhp->zfs_name);
	} else if (pa->pa_verbosity >= 2) {
	zfs_nicenum(bytes, buf, sizeof (buf));
	(void) fprintf(stderr,
	"%02d:%02d:%02d %5s %8llu %s\n",
	tm.tm_hour, tm.tm_min, tm.tm_sec,
	buf, (u_longlong_t)blocks, zhp->zfs_name);
	} else if (pa->pa_parsable) {
	(void) fprintf(stderr, "%02d:%02d:%02d\t%llu\t%s\n",
	tm.tm_hour, tm.tm_min, tm.tm_sec,
	(u_longlong_t)bytes, zhp->zfs_name);
	} else if (pa->pa_progress \|\|
	!send_progress_thread_signal_duetotimer) {
	zfs_nicebytes(bytes, buf, sizeof (buf));
	(void) fprintf(stderr, "%02d:%02d:%02d %5s %s\n",
	tm.tm_hour, tm.tm_min, tm.tm_sec,
	buf, zhp->zfs_name);
	}
	}
	pthread_cleanup_pop(B_TRUE);
	return (NULL);
	}

	static boolean_t
	send_progress_thread_exit(
	libzfs_handle_t hdl, pthread_t ptid, sigset_t oldmask)
	{
	void *status = NULL;
	(void) pthread_cancel(ptid);
	(void) pthread_join(ptid, &status);
	pthread_sigmask(SIG_SETMASK, oldmask, NULL);
	int error = (int)(uintptr_t)status;
	if (error != 0 && status != PTHREAD_CANCELED)
	return (zfs_standard_error(hdl, error,
	dgettext(TEXT_DOMAIN, "progress thread exited nonzero")));
	else
	return (B_FALSE);
	}

	static void
	send_print_verbose(FILE fout, const char tosnap, const char *fromsnap,
	uint64_t size, boolean_t parsable)
	{
	if (parsable) {
	if (fromsnap != NULL) {
	(void) fprintf(fout, dgettext(TEXT_DOMAIN,
	"incremental\t%s\t%s"), fromsnap, tosnap);
	} else {
	/*
	* Workaround for GCC 12+ with UBSan enabled deficencies.
	*
	* GCC 12+ invoked with -fsanitize=undefined incorrectly reports the code
	* below as violating -Wformat-overflow.
	*/
	#if defined(__GNUC__) && !defined(__clang__) && \
	defined(ZFS_UBSAN_ENABLED) && defined(HAVE_FORMAT_OVERFLOW)
	#pragma GCC diagnostic push
	#pragma GCC diagnostic ignored "-Wformat-overflow"
	#endif
	(void) fprintf(fout, dgettext(TEXT_DOMAIN,
	"full\t%s"), tosnap);
	#if defined(__GNUC__) && !defined(__clang__) && \
	defined(ZFS_UBSAN_ENABLED) && defined(HAVE_FORMAT_OVERFLOW)
	#pragma GCC diagnostic pop
	#endif
	}
	(void) fprintf(fout, "\t%llu", (longlong_t)size);
	} else {
	if (fromsnap != NULL) {
	if (strchr(fromsnap, '@') == NULL &&
	strchr(fromsnap, '#') == NULL) {
	(void) fprintf(fout, dgettext(TEXT_DOMAIN,
	"send from @%s to %s"), fromsnap, tosnap);
	} else {
	(void) fprintf(fout, dgettext(TEXT_DOMAIN,
	"send from %s to %s"), fromsnap, tosnap);
	}
	} else {
	(void) fprintf(fout, dgettext(TEXT_DOMAIN,
	"full send of %s"), tosnap);
	}
	if (size != 0) {
	char buf[16];
	zfs_nicebytes(size, buf, sizeof (buf));
	/*
	* Workaround for GCC 12+ with UBSan enabled deficencies.
	*
	* GCC 12+ invoked with -fsanitize=undefined incorrectly reports the code
	* below as violating -Wformat-overflow.
	*/
	#if defined(__GNUC__) && !defined(__clang__) && \
	defined(ZFS_UBSAN_ENABLED) && defined(HAVE_FORMAT_OVERFLOW)
	#pragma GCC diagnostic push
	#pragma GCC diagnostic ignored "-Wformat-overflow"
	#endif
	(void) fprintf(fout, dgettext(TEXT_DOMAIN,
	" estimated size is %s"), buf);
	#if defined(__GNUC__) && !defined(__clang__) && \
	defined(ZFS_UBSAN_ENABLED) && defined(HAVE_FORMAT_OVERFLOW)
	#pragma GCC diagnostic pop
	#endif
	}
	}
	(void) fprintf(fout, "\n");
	}

	/*
	* Send a single filesystem snapshot, updating the send dump data.
	* This interface is intended for use as a zfs_iter_snapshots_v2_sorted visitor.
	*/
	static int
	dump_snapshot(zfs_handle_t zhp, void arg)
	{
	send_dump_data_t *sdd = arg;
	progress_arg_t pa = { 0 };
	pthread_t tid;
	char *thissnap;
	enum lzc_send_flags flags = 0;
	int err;
	boolean_t isfromsnap, istosnap, fromorigin;
	boolean_t exclude = B_FALSE;
	FILE *fout = sdd->std_out ? stdout : stderr;

	err = 0;
	thissnap = strchr(zhp->zfs_name, '@') + 1;
	isfromsnap = (sdd->fromsnap != NULL &&
	strcmp(sdd->fromsnap, thissnap) == 0);

	if (!sdd->seenfrom && isfromsnap) {
	gather_holds(zhp, sdd);
	sdd->seenfrom = B_TRUE;
	(void) strlcpy(sdd->prevsnap, thissnap, sizeof (sdd->prevsnap));
	sdd->prevsnap_obj = zfs_prop_get_int(zhp, ZFS_PROP_OBJSETID);
	zfs_close(zhp);
	return (0);
	}

	if (sdd->seento \|\| !sdd->seenfrom) {
	zfs_close(zhp);
	return (0);
	}

	istosnap = (strcmp(sdd->tosnap, thissnap) == 0);
	if (istosnap)
	sdd->seento = B_TRUE;

	if (sdd->large_block)
	flags \|= LZC_SEND_FLAG_LARGE_BLOCK;
	if (sdd->embed_data)
	flags \|= LZC_SEND_FLAG_EMBED_DATA;
	if (sdd->compress)
	flags \|= LZC_SEND_FLAG_COMPRESS;
	if (sdd->raw)
	flags \|= LZC_SEND_FLAG_RAW;

	if (!sdd->doall && !isfromsnap && !istosnap) {
	if (sdd->replicate) {
	const char *snapname;
	nvlist_t *snapprops;
	/*
	* Filter out all intermediate snapshots except origin
	* snapshots needed to replicate clones.
	*/
	nvlist_t *nvfs = fsavl_find(sdd->fsavl,
	zhp->zfs_dmustats.dds_guid, &snapname);

	if (nvfs != NULL) {
	snapprops = fnvlist_lookup_nvlist(nvfs,
	"snapprops");
	snapprops = fnvlist_lookup_nvlist(snapprops,
	thissnap);
	exclude = !nvlist_exists(snapprops,
	"is_clone_origin");
	}
	} else {
	exclude = B_TRUE;
	}
	}

	/*
	* If a filter function exists, call it to determine whether
	* this snapshot will be sent.
	*/
	if (exclude \|\| (sdd->filter_cb != NULL &&
	sdd->filter_cb(zhp, sdd->filter_cb_arg) == B_FALSE)) {
	/*
	* This snapshot is filtered out. Don't send it, and don't
	* set prevsnap_obj, so it will be as if this snapshot didn't
	* exist, and the next accepted snapshot will be sent as
	* an incremental from the last accepted one, or as the
	* first (and full) snapshot in the case of a replication,
	* non-incremental send.
	*/
	zfs_close(zhp);
	return (0);
	}

	gather_holds(zhp, sdd);
	fromorigin = sdd->prevsnap[0] == '\0' &&
	(sdd->fromorigin \|\| sdd->replicate);

	if (sdd->verbosity != 0) {
	uint64_t size = 0;
	char fromds[ZFS_MAX_DATASET_NAME_LEN];

	if (sdd->prevsnap[0] != '\0') {
	(void) strlcpy(fromds, zhp->zfs_name, sizeof (fromds));
	*(strchr(fromds, '@') + 1) = '\0';
	(void) strlcat(fromds, sdd->prevsnap, sizeof (fromds));
	}
	if (zfs_send_space(zhp, zhp->zfs_name,
	sdd->prevsnap[0] ? fromds : NULL, flags, &size) == 0) {
	send_print_verbose(fout, zhp->zfs_name,
	sdd->prevsnap[0] ? sdd->prevsnap : NULL,
	size, sdd->parsable);
	sdd->size += size;
	}
	}

	if (!sdd->dryrun) {
	/*
	* If progress reporting is requested, spawn a new thread to
	* poll ZFS_IOC_SEND_PROGRESS at a regular interval.
	*/
	sigset_t oldmask;
	{
	pa.pa_zhp = zhp;
	pa.pa_fd = sdd->outfd;
	pa.pa_parsable = sdd->parsable;
	pa.pa_estimate = B_FALSE;
	pa.pa_verbosity = sdd->verbosity;
	pa.pa_size = sdd->size;
	pa.pa_astitle = sdd->progressastitle;
	pa.pa_progress = sdd->progress;

	if ((err = pthread_create(&tid, NULL,
	send_progress_thread, &pa)) != 0) {
	zfs_close(zhp);
	return (err);
	}
	SEND_PROGRESS_THREAD_PARENT_BLOCK(&oldmask);
	}

	err = dump_ioctl(zhp, sdd->prevsnap, sdd->prevsnap_obj,
	fromorigin, sdd->outfd, flags, sdd->debugnv);

	if (send_progress_thread_exit(zhp->zfs_hdl, tid, &oldmask))
	return (-1);
	}

	(void) strlcpy(sdd->prevsnap, thissnap, sizeof (sdd->prevsnap));
	sdd->prevsnap_obj = zfs_prop_get_int(zhp, ZFS_PROP_OBJSETID);
	zfs_close(zhp);
	return (err);
	}

	/*
	* Send all snapshots for a filesystem, updating the send dump data.
	*/
	static int
	dump_filesystem(zfs_handle_t zhp, send_dump_data_t sdd)
	{
	int rv = 0;
	boolean_t missingfrom = B_FALSE;
	zfs_cmd_t zc = {"\0"};
	uint64_t min_txg = 0, max_txg = 0;

	/*
	* Make sure the tosnap exists.
	*/
	(void) snprintf(zc.zc_name, sizeof (zc.zc_name), "%s@%s",
	zhp->zfs_name, sdd->tosnap);
	if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_OBJSET_STATS, &zc) != 0) {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"WARNING: could not send %s@%s: does not exist\n"),
	zhp->zfs_name, sdd->tosnap);
	sdd->err = B_TRUE;
	return (0);
	}

	/*
	* If this fs does not have fromsnap, and we're doing
	* recursive, we need to send a full stream from the
	* beginning (or an incremental from the origin if this
	* is a clone). If we're doing non-recursive, then let
	* them get the error.
	*/
	if (sdd->replicate && sdd->fromsnap) {
	/*
	* Make sure the fromsnap exists.
	*/
	(void) snprintf(zc.zc_name, sizeof (zc.zc_name), "%s@%s",
	zhp->zfs_name, sdd->fromsnap);
	if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_OBJSET_STATS, &zc) != 0)
	missingfrom = B_TRUE;
	}

	sdd->seenfrom = sdd->seento = B_FALSE;
	sdd->prevsnap[0] = '\0';
	sdd->prevsnap_obj = 0;
	if (sdd->fromsnap == NULL \|\| missingfrom)
	sdd->seenfrom = B_TRUE;

	/*
	* Iterate through all snapshots and process the ones we will be
	* sending. If we only have a "from" and "to" snapshot to deal
	* with, we can avoid iterating through all the other snapshots.
	*/
	if (sdd->doall \|\| sdd->replicate \|\| sdd->tosnap == NULL) {
	if (!sdd->replicate) {
	if (sdd->fromsnap != NULL) {
	min_txg = get_snap_txg(zhp->zfs_hdl,
	zhp->zfs_name, sdd->fromsnap);
	}
	if (sdd->tosnap != NULL) {
	max_txg = get_snap_txg(zhp->zfs_hdl,
	zhp->zfs_name, sdd->tosnap);
	}
	}
	rv = zfs_iter_snapshots_sorted_v2(zhp, 0, dump_snapshot, sdd,
	min_txg, max_txg);
	} else {
	char snapname[MAXPATHLEN] = { 0 };
	zfs_handle_t *snap;

	/* Dump fromsnap. */
	if (!sdd->seenfrom) {
	(void) snprintf(snapname, sizeof (snapname),
	"%s@%s", zhp->zfs_name, sdd->fromsnap);
	snap = zfs_open(zhp->zfs_hdl, snapname,
	ZFS_TYPE_SNAPSHOT);
	if (snap != NULL)
	rv = dump_snapshot(snap, sdd);
	else
	rv = errno;
	}

	/* Dump tosnap. */
	if (rv == 0) {
	(void) snprintf(snapname, sizeof (snapname),
	"%s@%s", zhp->zfs_name, sdd->tosnap);
	snap = zfs_open(zhp->zfs_hdl, snapname,
	ZFS_TYPE_SNAPSHOT);
	if (snap != NULL)
	rv = dump_snapshot(snap, sdd);
	else
	rv = errno;
	}
	}

	if (!sdd->seenfrom) {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"WARNING: could not send %s@%s:\n"
	"incremental source (%s@%s) does not exist\n"),
	zhp->zfs_name, sdd->tosnap,
	zhp->zfs_name, sdd->fromsnap);
	sdd->err = B_TRUE;
	} else if (!sdd->seento) {
	if (sdd->fromsnap) {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"WARNING: could not send %s@%s:\n"
	"incremental source (%s@%s) "
	"is not earlier than it\n"),
	zhp->zfs_name, sdd->tosnap,
	zhp->zfs_name, sdd->fromsnap);
	} else {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"WARNING: "
	"could not send %s@%s: does not exist\n"),
	zhp->zfs_name, sdd->tosnap);
	}
	sdd->err = B_TRUE;
	}

	return (rv);
	}

	/*
	* Send all snapshots for all filesystems in sdd.
	*/
	static int
	dump_filesystems(zfs_handle_t rzhp, send_dump_data_t sdd)
	{
	nvpair_t *fspair;
	boolean_t needagain, progress;

	if (!sdd->replicate)
	return (dump_filesystem(rzhp, sdd));

	/* Mark the clone origin snapshots. */
	for (fspair = nvlist_next_nvpair(sdd->fss, NULL); fspair;
	fspair = nvlist_next_nvpair(sdd->fss, fspair)) {
	nvlist_t *nvfs;
	uint64_t origin_guid = 0;

	nvfs = fnvpair_value_nvlist(fspair);
	(void) nvlist_lookup_uint64(nvfs, "origin", &origin_guid);
	if (origin_guid != 0) {
	const char *snapname;
	nvlist_t *origin_nv = fsavl_find(sdd->fsavl,
	origin_guid, &snapname);
	if (origin_nv != NULL) {
	nvlist_t *snapprops;
	snapprops = fnvlist_lookup_nvlist(origin_nv,
	"snapprops");
	snapprops = fnvlist_lookup_nvlist(snapprops,
	snapname);
	fnvlist_add_boolean(snapprops,
	"is_clone_origin");
	}
	}
	}
	again:
	needagain = progress = B_FALSE;
	for (fspair = nvlist_next_nvpair(sdd->fss, NULL); fspair;
	fspair = nvlist_next_nvpair(sdd->fss, fspair)) {
	nvlist_t fslist, parent_nv;
	const char *fsname;
	zfs_handle_t *zhp;
	int err;
	uint64_t origin_guid = 0;
	uint64_t parent_guid = 0;

	fslist = fnvpair_value_nvlist(fspair);
	if (nvlist_lookup_boolean(fslist, "sent") == 0)
	continue;

	fsname = fnvlist_lookup_string(fslist, "name");
	(void) nvlist_lookup_uint64(fslist, "origin", &origin_guid);
	(void) nvlist_lookup_uint64(fslist, "parentfromsnap",
	&parent_guid);

	if (parent_guid != 0) {
	parent_nv = fsavl_find(sdd->fsavl, parent_guid, NULL);
	if (!nvlist_exists(parent_nv, "sent")) {
	/* Parent has not been sent; skip this one. */
	needagain = B_TRUE;
	continue;
	}
	}

	if (origin_guid != 0) {
	nvlist_t *origin_nv = fsavl_find(sdd->fsavl,
	origin_guid, NULL);
	if (origin_nv != NULL &&
	!nvlist_exists(origin_nv, "sent")) {
	/*
	* Origin has not been sent yet;
	* skip this clone.
	*/
	needagain = B_TRUE;
	continue;
	}
	}

	zhp = zfs_open(rzhp->zfs_hdl, fsname, ZFS_TYPE_DATASET);
	if (zhp == NULL)
	return (-1);
	err = dump_filesystem(zhp, sdd);
	fnvlist_add_boolean(fslist, "sent");
	progress = B_TRUE;
	zfs_close(zhp);
	if (err)
	return (err);
	}
	if (needagain) {
	assert(progress);
	goto again;
	}

	/* Clean out the sent flags in case we reuse this fss. */
	for (fspair = nvlist_next_nvpair(sdd->fss, NULL); fspair;
	fspair = nvlist_next_nvpair(sdd->fss, fspair)) {
	nvlist_t *fslist;

	fslist = fnvpair_value_nvlist(fspair);
	(void) nvlist_remove_all(fslist, "sent");
	}

	return (0);
	}

	nvlist_t *
	zfs_send_resume_token_to_nvlist(libzfs_handle_t hdl, const char token)
	{
	unsigned int version;
	int nread, i;
	unsigned long long checksum, packed_len;

	/*
	* Decode token header, which is:
	* <token version>-<checksum of payload>-<uncompressed payload length>
	* Note that the only supported token version is 1.
	*/
	nread = sscanf(token, "%u-%llx-%llx-",
	&version, &checksum, &packed_len);
	if (nread != 3) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"resume token is corrupt (invalid format)"));
	return (NULL);
	}

	if (version != ZFS_SEND_RESUME_TOKEN_VERSION) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"resume token is corrupt (invalid version %u)"),
	version);
	return (NULL);
	}

	/* Convert hexadecimal representation to binary. */
	token = strrchr(token, '-') + 1;
	int len = strlen(token) / 2;
	unsigned char *compressed = zfs_alloc(hdl, len);
	for (i = 0; i < len; i++) {
	nread = sscanf(token + i * 2, "%2hhx", compressed + i);
	if (nread != 1) {
	free(compressed);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"resume token is corrupt "
	"(payload is not hex-encoded)"));
	return (NULL);
	}
	}

	/* Verify checksum. */
	zio_cksum_t cksum;
	fletcher_4_native_varsize(compressed, len, &cksum);
	if (cksum.zc_word[0] != checksum) {
	free(compressed);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"resume token is corrupt (incorrect checksum)"));
	return (NULL);
	}

	/* Uncompress. */
	void *packed = zfs_alloc(hdl, packed_len);
	uLongf packed_len_long = packed_len;
	if (uncompress(packed, &packed_len_long, compressed, len) != Z_OK \|\|
	packed_len_long != packed_len) {
	free(packed);
	free(compressed);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"resume token is corrupt (decompression failed)"));
	return (NULL);
	}

	/* Unpack nvlist. */
	nvlist_t *nv;
	int error = nvlist_unpack(packed, packed_len, &nv, KM_SLEEP);
	free(packed);
	free(compressed);
	if (error != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"resume token is corrupt (nvlist_unpack failed)"));
	return (NULL);
	}
	return (nv);
	}

	static enum lzc_send_flags
	lzc_flags_from_sendflags(const sendflags_t *flags)
	{
	enum lzc_send_flags lzc_flags = 0;

	if (flags->largeblock)
	lzc_flags \|= LZC_SEND_FLAG_LARGE_BLOCK;
	if (flags->embed_data)
	lzc_flags \|= LZC_SEND_FLAG_EMBED_DATA;
	if (flags->compress)
	lzc_flags \|= LZC_SEND_FLAG_COMPRESS;
	if (flags->raw)
	lzc_flags \|= LZC_SEND_FLAG_RAW;
	if (flags->saved)
	lzc_flags \|= LZC_SEND_FLAG_SAVED;

	return (lzc_flags);
	}

	static int
	estimate_size(zfs_handle_t zhp, const char from, int fd, sendflags_t *flags,
	uint64_t resumeobj, uint64_t resumeoff, uint64_t bytes,
	const char redactbook, char errbuf, uint64_t *sizep)
	{
	uint64_t size;
	FILE *fout = flags->dryrun ? stdout : stderr;
	progress_arg_t pa = { 0 };
	int err = 0;
	pthread_t ptid;
	sigset_t oldmask;

	{
	pa.pa_zhp = zhp;
	pa.pa_fd = fd;
	pa.pa_parsable = flags->parsable;
	pa.pa_estimate = B_TRUE;
	pa.pa_verbosity = flags->verbosity;

	err = pthread_create(&ptid, NULL,
	send_progress_thread, &pa);
	if (err != 0) {
	zfs_error_aux(zhp->zfs_hdl, "%s", zfs_strerror(errno));
	return (zfs_error(zhp->zfs_hdl,
	EZFS_THREADCREATEFAILED, errbuf));
	}
	SEND_PROGRESS_THREAD_PARENT_BLOCK(&oldmask);
	}

	err = lzc_send_space_resume_redacted(zhp->zfs_name, from,
	lzc_flags_from_sendflags(flags), resumeobj, resumeoff, bytes,
	redactbook, fd, &size);
	*sizep = size;

	if (send_progress_thread_exit(zhp->zfs_hdl, ptid, &oldmask))
	return (-1);

	if (!flags->progress && !flags->parsable)
	return (err);

	if (err != 0) {
	zfs_error_aux(zhp->zfs_hdl, "%s", zfs_strerror(err));
	return (zfs_error(zhp->zfs_hdl, EZFS_BADBACKUP,
	errbuf));
	}
	send_print_verbose(fout, zhp->zfs_name, from, size,
	flags->parsable);

	if (flags->parsable) {
	(void) fprintf(fout, "size\t%llu\n", (longlong_t)size);
	} else {
	char buf[16];
	zfs_nicenum(size, buf, sizeof (buf));
	(void) fprintf(fout, dgettext(TEXT_DOMAIN,
	"total estimated size is %s\n"), buf);
	}
	return (0);
	}

	static boolean_t
	redact_snaps_contains(const uint64_t *snaps, uint64_t num_snaps, uint64_t guid)
	{
	for (int i = 0; i < num_snaps; i++) {
	if (snaps[i] == guid)
	return (B_TRUE);
	}
	return (B_FALSE);
	}

	static boolean_t
	redact_snaps_equal(const uint64_t *snaps1, uint64_t num_snaps1,
	const uint64_t *snaps2, uint64_t num_snaps2)
	{
	if (num_snaps1 != num_snaps2)
	return (B_FALSE);
	for (int i = 0; i < num_snaps1; i++) {
	if (!redact_snaps_contains(snaps2, num_snaps2, snaps1[i]))
	return (B_FALSE);
	}
	return (B_TRUE);
	}

	static int
	get_bookmarks(const char path, nvlist_t *bmarksp)
	{
	nvlist_t *props = fnvlist_alloc();
	int error;

	fnvlist_add_boolean(props, "redact_complete");
	fnvlist_add_boolean(props, zfs_prop_to_name(ZFS_PROP_REDACT_SNAPS));
	error = lzc_get_bookmarks(path, props, bmarksp);
	fnvlist_free(props);
	return (error);
	}

	static nvpair_t *
	find_redact_pair(nvlist_t bmarks, const uint64_t redact_snap_guids,
	int num_redact_snaps)
	{
	nvpair_t *pair;

	for (pair = nvlist_next_nvpair(bmarks, NULL); pair;
	pair = nvlist_next_nvpair(bmarks, pair)) {

	nvlist_t *bmark = fnvpair_value_nvlist(pair);
	nvlist_t *vallist = fnvlist_lookup_nvlist(bmark,
	zfs_prop_to_name(ZFS_PROP_REDACT_SNAPS));
	uint_t len = 0;
	uint64_t *bmarksnaps = fnvlist_lookup_uint64_array(vallist,
	ZPROP_VALUE, &len);
	if (redact_snaps_equal(redact_snap_guids,
	num_redact_snaps, bmarksnaps, len)) {
	break;
	}
	}
	return (pair);
	}

	static boolean_t
	get_redact_complete(nvpair_t *pair)
	{
	nvlist_t *bmark = fnvpair_value_nvlist(pair);
	nvlist_t *vallist = fnvlist_lookup_nvlist(bmark, "redact_complete");
	boolean_t complete = fnvlist_lookup_boolean_value(vallist,
	ZPROP_VALUE);

	return (complete);
	}

	/*
	* Check that the list of redaction snapshots in the bookmark matches the send
	* we're resuming, and return whether or not it's complete.
	*
	* Note that the caller needs to free the contents of *bookname with free() if
	* this function returns successfully.
	*/
	static int
	find_redact_book(libzfs_handle_t hdl, const char path,
	const uint64_t *redact_snap_guids, int num_redact_snaps,
	char **bookname)
	{
	char errbuf[ERRBUFLEN];
	nvlist_t *bmarks;

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot resume send"));

	int error = get_bookmarks(path, &bmarks);
	if (error != 0) {
	if (error == ESRCH) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"nonexistent redaction bookmark provided"));
	} else if (error == ENOENT) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"dataset to be sent no longer exists"));
	} else {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"unknown error: %s"), zfs_strerror(error));
	}
	return (zfs_error(hdl, EZFS_BADPROP, errbuf));
	}
	nvpair_t *pair = find_redact_pair(bmarks, redact_snap_guids,
	num_redact_snaps);
	if (pair == NULL) {
	fnvlist_free(bmarks);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"no appropriate redaction bookmark exists"));
	return (zfs_error(hdl, EZFS_BADPROP, errbuf));
	}
	boolean_t complete = get_redact_complete(pair);
	if (!complete) {
	fnvlist_free(bmarks);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"incomplete redaction bookmark provided"));
	return (zfs_error(hdl, EZFS_BADPROP, errbuf));
	}
	*bookname = strndup(nvpair_name(pair), ZFS_MAX_DATASET_NAME_LEN);
	ASSERT3P(*bookname, !=, NULL);
	fnvlist_free(bmarks);
	return (0);
	}

	static enum lzc_send_flags
	lzc_flags_from_resume_nvl(nvlist_t *resume_nvl)
	{
	enum lzc_send_flags lzc_flags = 0;

	if (nvlist_exists(resume_nvl, "largeblockok"))
	lzc_flags \|= LZC_SEND_FLAG_LARGE_BLOCK;
	if (nvlist_exists(resume_nvl, "embedok"))
	lzc_flags \|= LZC_SEND_FLAG_EMBED_DATA;
	if (nvlist_exists(resume_nvl, "compressok"))
	lzc_flags \|= LZC_SEND_FLAG_COMPRESS;
	if (nvlist_exists(resume_nvl, "rawok"))
	lzc_flags \|= LZC_SEND_FLAG_RAW;
	if (nvlist_exists(resume_nvl, "savedok"))
	lzc_flags \|= LZC_SEND_FLAG_SAVED;

	return (lzc_flags);
	}

	static int
	zfs_send_resume_impl_cb_impl(libzfs_handle_t hdl, sendflags_t flags,
	int outfd, nvlist_t *resume_nvl)
	{
	char errbuf[ERRBUFLEN];
	const char *toname;
	const char *fromname = NULL;
	uint64_t resumeobj, resumeoff, toguid, fromguid, bytes;
	zfs_handle_t *zhp;
	int error = 0;
	char name[ZFS_MAX_DATASET_NAME_LEN];
	FILE *fout = (flags->verbosity > 0 && flags->dryrun) ? stdout : stderr;
	uint64_t *redact_snap_guids = NULL;
	int num_redact_snaps = 0;
	char *redact_book = NULL;
	uint64_t size = 0;

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot resume send"));

	if (flags->verbosity != 0) {
	(void) fprintf(fout, dgettext(TEXT_DOMAIN,
	"resume token contents:\n"));
	nvlist_print(fout, resume_nvl);
	}

	if (nvlist_lookup_string(resume_nvl, "toname", &toname) != 0 \|\|
	nvlist_lookup_uint64(resume_nvl, "object", &resumeobj) != 0 \|\|
	nvlist_lookup_uint64(resume_nvl, "offset", &resumeoff) != 0 \|\|
	nvlist_lookup_uint64(resume_nvl, "bytes", &bytes) != 0 \|\|
	nvlist_lookup_uint64(resume_nvl, "toguid", &toguid) != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"resume token is corrupt"));
	return (zfs_error(hdl, EZFS_FAULT, errbuf));
	}
	fromguid = 0;
	(void) nvlist_lookup_uint64(resume_nvl, "fromguid", &fromguid);

	if (flags->saved) {
	(void) strlcpy(name, toname, sizeof (name));
	} else {
	error = guid_to_name(hdl, toname, toguid, B_FALSE, name);
	if (error != 0) {
	if (zfs_dataset_exists(hdl, toname, ZFS_TYPE_DATASET)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' is no longer the same snapshot "
	"used in the initial send"), toname);
	} else {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"'%s' used in the initial send no "
	"longer exists"), toname);
	}
	return (zfs_error(hdl, EZFS_BADPATH, errbuf));
	}
	}

	zhp = zfs_open(hdl, name, ZFS_TYPE_DATASET);
	if (zhp == NULL) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"unable to access '%s'"), name);
	return (zfs_error(hdl, EZFS_BADPATH, errbuf));
	}

	if (nvlist_lookup_uint64_array(resume_nvl, "book_redact_snaps",
	&redact_snap_guids, (uint_t *)&num_redact_snaps) != 0) {
	num_redact_snaps = -1;
	}

	if (fromguid != 0) {
	if (guid_to_name_redact_snaps(hdl, toname, fromguid, B_TRUE,
	redact_snap_guids, num_redact_snaps, name) != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"incremental source %#llx no longer exists"),
	(longlong_t)fromguid);
	return (zfs_error(hdl, EZFS_BADPATH, errbuf));
	}
	fromname = name;
	}

	redact_snap_guids = NULL;

	if (nvlist_lookup_uint64_array(resume_nvl,
	zfs_prop_to_name(ZFS_PROP_REDACT_SNAPS), &redact_snap_guids,
	(uint_t *)&num_redact_snaps) == 0) {
	char path[ZFS_MAX_DATASET_NAME_LEN];

	(void) strlcpy(path, toname, sizeof (path));
	char *at = strchr(path, '@');
	ASSERT3P(at, !=, NULL);

	*at = '\0';

	if ((error = find_redact_book(hdl, path, redact_snap_guids,
	num_redact_snaps, &redact_book)) != 0) {
	return (error);
	}
	}

	enum lzc_send_flags lzc_flags = lzc_flags_from_sendflags(flags) \|
	lzc_flags_from_resume_nvl(resume_nvl);

	if (flags->verbosity != 0 \|\| flags->progressastitle) {
	/*
	* Some of these may have come from the resume token, set them
	* here for size estimate purposes.
	*/
	sendflags_t tmpflags = *flags;
	if (lzc_flags & LZC_SEND_FLAG_LARGE_BLOCK)
	tmpflags.largeblock = B_TRUE;
	if (lzc_flags & LZC_SEND_FLAG_COMPRESS)
	tmpflags.compress = B_TRUE;
	if (lzc_flags & LZC_SEND_FLAG_EMBED_DATA)
	tmpflags.embed_data = B_TRUE;
	if (lzc_flags & LZC_SEND_FLAG_RAW)
	tmpflags.raw = B_TRUE;
	if (lzc_flags & LZC_SEND_FLAG_SAVED)
	tmpflags.saved = B_TRUE;
	error = estimate_size(zhp, fromname, outfd, &tmpflags,
	resumeobj, resumeoff, bytes, redact_book, errbuf, &size);
	}

	if (!flags->dryrun) {
	progress_arg_t pa = { 0 };
	pthread_t tid;
	sigset_t oldmask;
	/*
	* If progress reporting is requested, spawn a new thread to
	* poll ZFS_IOC_SEND_PROGRESS at a regular interval.
	*/
	{
	pa.pa_zhp = zhp;
	pa.pa_fd = outfd;
	pa.pa_parsable = flags->parsable;
	pa.pa_estimate = B_FALSE;
	pa.pa_verbosity = flags->verbosity;
	pa.pa_size = size;
	pa.pa_astitle = flags->progressastitle;
	pa.pa_progress = flags->progress;

	error = pthread_create(&tid, NULL,
	send_progress_thread, &pa);
	if (error != 0) {
	if (redact_book != NULL)
	free(redact_book);
	zfs_close(zhp);
	return (error);
	}
	SEND_PROGRESS_THREAD_PARENT_BLOCK(&oldmask);
	}

	error = lzc_send_resume_redacted(zhp->zfs_name, fromname, outfd,
	lzc_flags, resumeobj, resumeoff, redact_book);
	if (redact_book != NULL)
	free(redact_book);

	if (send_progress_thread_exit(hdl, tid, &oldmask)) {
	zfs_close(zhp);
	return (-1);
	}

	char errbuf[ERRBUFLEN];
	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"warning: cannot send '%s'"), zhp->zfs_name);

	zfs_close(zhp);

	switch (error) {
	case 0:
	return (0);
	case EACCES:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"source key must be loaded"));
	return (zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf));
	case ESRCH:
	if (lzc_exists(zhp->zfs_name)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"incremental source could not be found"));
	}
	return (zfs_error(hdl, EZFS_NOENT, errbuf));

	case EXDEV:
	case ENOENT:
	case EDQUOT:
	case EFBIG:
	case EIO:
	case ENOLINK:
	case ENOSPC:
	case ENOSTR:
	case ENXIO:
	case EPIPE:
	case ERANGE:
	case EFAULT:
	case EROFS:
	zfs_error_aux(hdl, "%s", zfs_strerror(errno));
	return (zfs_error(hdl, EZFS_BADBACKUP, errbuf));

	default:
	return (zfs_standard_error(hdl, errno, errbuf));
	}
	} else {
	if (redact_book != NULL)
	free(redact_book);
	}

	zfs_close(zhp);

	return (error);
	}

	struct zfs_send_resume_impl {
	libzfs_handle_t *hdl;
	sendflags_t *flags;
	nvlist_t *resume_nvl;
	};

	static int
	zfs_send_resume_impl_cb(int outfd, void *arg)
	{
	struct zfs_send_resume_impl *zsri = arg;
	return (zfs_send_resume_impl_cb_impl(zsri->hdl, zsri->flags, outfd,
	zsri->resume_nvl));
	}

	static int
	zfs_send_resume_impl(libzfs_handle_t hdl, sendflags_t flags, int outfd,
	nvlist_t *resume_nvl)
	{
	struct zfs_send_resume_impl zsri = {
	.hdl = hdl,
	.flags = flags,
	.resume_nvl = resume_nvl,
	};
	return (lzc_send_wrapper(zfs_send_resume_impl_cb, outfd, &zsri));
	}

	int
	zfs_send_resume(libzfs_handle_t hdl, sendflags_t flags, int outfd,
	const char *resume_token)
	{
	int ret;
	char errbuf[ERRBUFLEN];
	nvlist_t *resume_nvl;

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot resume send"));

	resume_nvl = zfs_send_resume_token_to_nvlist(hdl, resume_token);
	if (resume_nvl == NULL) {
	/*
	* zfs_error_aux has already been set by
	* zfs_send_resume_token_to_nvlist()
	*/
	return (zfs_error(hdl, EZFS_FAULT, errbuf));
	}

	ret = zfs_send_resume_impl(hdl, flags, outfd, resume_nvl);
	fnvlist_free(resume_nvl);

	return (ret);
	}

	int
	zfs_send_saved(zfs_handle_t zhp, sendflags_t flags, int outfd,
	const char *resume_token)
	{
	int ret;
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	nvlist_t saved_nvl = NULL, resume_nvl = NULL;
	uint64_t saved_guid = 0, resume_guid = 0;
	uint64_t obj = 0, off = 0, bytes = 0;
	char token_buf[ZFS_MAXPROPLEN];
	char errbuf[ERRBUFLEN];

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"saved send failed"));

	ret = zfs_prop_get(zhp, ZFS_PROP_RECEIVE_RESUME_TOKEN,
	token_buf, sizeof (token_buf), NULL, NULL, 0, B_TRUE);
	if (ret != 0)
	goto out;

	saved_nvl = zfs_send_resume_token_to_nvlist(hdl, token_buf);
	if (saved_nvl == NULL) {
	/*
	* zfs_error_aux has already been set by
	* zfs_send_resume_token_to_nvlist()
	*/
	ret = zfs_error(hdl, EZFS_FAULT, errbuf);
	goto out;
	}

	/*
	* If a resume token is provided we use the object and offset
	* from that instead of the default, which starts from the
	* beginning.
	*/
	if (resume_token != NULL) {
	resume_nvl = zfs_send_resume_token_to_nvlist(hdl,
	resume_token);
	if (resume_nvl == NULL) {
	ret = zfs_error(hdl, EZFS_FAULT, errbuf);
	goto out;
	}

	if (nvlist_lookup_uint64(resume_nvl, "object", &obj) != 0 \|\|
	nvlist_lookup_uint64(resume_nvl, "offset", &off) != 0 \|\|
	nvlist_lookup_uint64(resume_nvl, "bytes", &bytes) != 0 \|\|
	nvlist_lookup_uint64(resume_nvl, "toguid",
	&resume_guid) != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"provided resume token is corrupt"));
	ret = zfs_error(hdl, EZFS_FAULT, errbuf);
	goto out;
	}

	if (nvlist_lookup_uint64(saved_nvl, "toguid",
	&saved_guid)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"dataset's resume token is corrupt"));
	ret = zfs_error(hdl, EZFS_FAULT, errbuf);
	goto out;
	}

	if (resume_guid != saved_guid) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"provided resume token does not match dataset"));
	ret = zfs_error(hdl, EZFS_BADBACKUP, errbuf);
	goto out;
	}
	}

	(void) nvlist_remove_all(saved_nvl, "object");
	fnvlist_add_uint64(saved_nvl, "object", obj);

	(void) nvlist_remove_all(saved_nvl, "offset");
	fnvlist_add_uint64(saved_nvl, "offset", off);

	(void) nvlist_remove_all(saved_nvl, "bytes");
	fnvlist_add_uint64(saved_nvl, "bytes", bytes);

	(void) nvlist_remove_all(saved_nvl, "toname");
	fnvlist_add_string(saved_nvl, "toname", zhp->zfs_name);

	ret = zfs_send_resume_impl(hdl, flags, outfd, saved_nvl);

	out:
	fnvlist_free(saved_nvl);
	fnvlist_free(resume_nvl);
	return (ret);
	}

	/*
	* This function informs the target system that the recursive send is complete.
	* The record is also expected in the case of a send -p.
	*/
	static int
	send_conclusion_record(int fd, zio_cksum_t *zc)
	{
	- dmu_replay_record_t drr = { 0 };
	+ dmu_replay_record_t drr;
	+ memset(&drr, 0, sizeof (dmu_replay_record_t));
	drr.drr_type = DRR_END;
	if (zc != NULL)
	drr.drr_u.drr_end.drr_checksum = *zc;
	if (write(fd, &drr, sizeof (drr)) == -1) {
	return (errno);
	}
	return (0);
	}

	/*
	* This function is responsible for sending the records that contain the
	* necessary information for the target system's libzfs to be able to set the
	* properties of the filesystem being received, or to be able to prepare for
	* a recursive receive.
	*
	* The "zhp" argument is the handle of the snapshot we are sending
	* (the "tosnap"). The "from" argument is the short snapshot name (the part
	* after the @) of the incremental source.
	*/
	static int
	send_prelim_records(zfs_handle_t zhp, const char from, int fd,
	boolean_t gather_props, boolean_t recursive, boolean_t verbose,
	boolean_t dryrun, boolean_t raw, boolean_t replicate, boolean_t skipmissing,
	boolean_t backup, boolean_t holds, boolean_t props, boolean_t doall,
	nvlist_t fssp, avl_tree_t fsavlp)
	{
	int err = 0;
	char *packbuf = NULL;
	size_t buflen = 0;
	zio_cksum_t zc = { {0} };
	int featureflags = 0;
	/* name of filesystem/volume that contains snapshot we are sending */
	char tofs[ZFS_MAX_DATASET_NAME_LEN];
	/* short name of snap we are sending */
	const char *tosnap = "";

	char errbuf[ERRBUFLEN];
	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"warning: cannot send '%s'"), zhp->zfs_name);
	if (zhp->zfs_type == ZFS_TYPE_FILESYSTEM && zfs_prop_get_int(zhp,
	ZFS_PROP_VERSION) >= ZPL_VERSION_SA) {
	featureflags \|= DMU_BACKUP_FEATURE_SA_SPILL;
	}

	if (holds)
	featureflags \|= DMU_BACKUP_FEATURE_HOLDS;

	(void) strlcpy(tofs, zhp->zfs_name, ZFS_MAX_DATASET_NAME_LEN);
	char *at = strchr(tofs, '@');
	if (at != NULL) {
	*at = '\0';
	tosnap = at + 1;
	}

	if (gather_props) {
	nvlist_t *hdrnv = fnvlist_alloc();
	nvlist_t *fss = NULL;

	if (from != NULL)
	fnvlist_add_string(hdrnv, "fromsnap", from);
	fnvlist_add_string(hdrnv, "tosnap", tosnap);
	if (!recursive)
	fnvlist_add_boolean(hdrnv, "not_recursive");

	if (raw) {
	fnvlist_add_boolean(hdrnv, "raw");
	}

	if (gather_nvlist(zhp->zfs_hdl, tofs,
	from, tosnap, recursive, raw, doall, replicate, skipmissing,
	verbose, backup, holds, props, &fss, fsavlp) != 0) {
	return (zfs_error(zhp->zfs_hdl, EZFS_BADBACKUP,
	errbuf));
	}
	/*
	* Do not allow the size of the properties list to exceed
	* the limit
	*/
	if ((fnvlist_size(fss) + fnvlist_size(hdrnv)) >
	zhp->zfs_hdl->libzfs_max_nvlist) {
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "warning: cannot send '%s': "
	"the size of the list of snapshots and properties "
	"is too large to be received successfully.\n"
	"Select a smaller number of snapshots to send.\n"),
	zhp->zfs_name);
	return (zfs_error(zhp->zfs_hdl, EZFS_NOSPC,
	errbuf));
	}
	fnvlist_add_nvlist(hdrnv, "fss", fss);
	VERIFY0(nvlist_pack(hdrnv, &packbuf, &buflen, NV_ENCODE_XDR,
	0));
	if (fssp != NULL) {
	*fssp = fss;
	} else {
	fnvlist_free(fss);
	}
	fnvlist_free(hdrnv);
	}

	if (!dryrun) {
	- dmu_replay_record_t drr = { 0 };
	+ dmu_replay_record_t drr;
	+ memset(&drr, 0, sizeof (dmu_replay_record_t));
	/* write first begin record */
	drr.drr_type = DRR_BEGIN;
	drr.drr_u.drr_begin.drr_magic = DMU_BACKUP_MAGIC;
	DMU_SET_STREAM_HDRTYPE(drr.drr_u.drr_begin.
	drr_versioninfo, DMU_COMPOUNDSTREAM);
	DMU_SET_FEATUREFLAGS(drr.drr_u.drr_begin.
	drr_versioninfo, featureflags);
	if (snprintf(drr.drr_u.drr_begin.drr_toname,
	sizeof (drr.drr_u.drr_begin.drr_toname), "%s@%s", tofs,
	tosnap) >= sizeof (drr.drr_u.drr_begin.drr_toname)) {
	return (zfs_error(zhp->zfs_hdl, EZFS_BADBACKUP,
	errbuf));
	}
	drr.drr_payloadlen = buflen;

	err = dump_record(&drr, packbuf, buflen, &zc, fd);
	free(packbuf);
	if (err != 0) {
	zfs_error_aux(zhp->zfs_hdl, "%s", zfs_strerror(err));
	return (zfs_error(zhp->zfs_hdl, EZFS_BADBACKUP,
	errbuf));
	}
	err = send_conclusion_record(fd, &zc);
	if (err != 0) {
	zfs_error_aux(zhp->zfs_hdl, "%s", zfs_strerror(err));
	return (zfs_error(zhp->zfs_hdl, EZFS_BADBACKUP,
	errbuf));
	}
	}
	return (0);
	}

	/*
	* Generate a send stream. The "zhp" argument is the filesystem/volume
	* that contains the snapshot to send. The "fromsnap" argument is the
	* short name (the part after the '@') of the snapshot that is the
	* incremental source to send from (if non-NULL). The "tosnap" argument
	* is the short name of the snapshot to send.
	*
	* The content of the send stream is the snapshot identified by
	* 'tosnap'. Incremental streams are requested in two ways:
	* - from the snapshot identified by "fromsnap" (if non-null) or
	* - from the origin of the dataset identified by zhp, which must
	* be a clone. In this case, "fromsnap" is null and "fromorigin"
	* is TRUE.
	*
	* The send stream is recursive (i.e. dumps a hierarchy of snapshots) and
	* uses a special header (with a hdrtype field of DMU_COMPOUNDSTREAM)
	* if "replicate" is set. If "doall" is set, dump all the intermediate
	* snapshots. The DMU_COMPOUNDSTREAM header is used in the "doall"
	* case too. If "props" is set, send properties.
	*
	* Pre-wrapped (cf. lzc_send_wrapper()).
	*/
	static int
	zfs_send_cb_impl(zfs_handle_t zhp, const char fromsnap, const char *tosnap,
	sendflags_t *flags, int outfd, snapfilter_cb_t filter_func,
	void cb_arg, nvlist_t *debugnvp)
	{
	char errbuf[ERRBUFLEN];
	send_dump_data_t sdd = { 0 };
	int err = 0;
	nvlist_t *fss = NULL;
	avl_tree_t *fsavl = NULL;
	static uint64_t holdseq;
	int spa_version;
	FILE *fout;

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot send '%s'"), zhp->zfs_name);

	if (fromsnap && fromsnap[0] == '\0') {
	zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN,
	"zero-length incremental source"));
	return (zfs_error(zhp->zfs_hdl, EZFS_NOENT, errbuf));
	}

	if (fromsnap) {
	char full_fromsnap_name[ZFS_MAX_DATASET_NAME_LEN];
	if (snprintf(full_fromsnap_name, sizeof (full_fromsnap_name),
	"%s@%s", zhp->zfs_name, fromsnap) >=
	sizeof (full_fromsnap_name)) {
	err = EINVAL;
	goto stderr_out;
	}
	zfs_handle_t *fromsnapn = zfs_open(zhp->zfs_hdl,
	full_fromsnap_name, ZFS_TYPE_SNAPSHOT);
	if (fromsnapn == NULL) {
	err = -1;
	goto err_out;
	}
	zfs_close(fromsnapn);
	}

	if (flags->replicate \|\| flags->doall \|\| flags->props \|\|
	flags->holds \|\| flags->backup) {
	char full_tosnap_name[ZFS_MAX_DATASET_NAME_LEN];
	if (snprintf(full_tosnap_name, sizeof (full_tosnap_name),
	"%s@%s", zhp->zfs_name, tosnap) >=
	sizeof (full_tosnap_name)) {
	err = EINVAL;
	goto stderr_out;
	}
	zfs_handle_t *tosnap = zfs_open(zhp->zfs_hdl,
	full_tosnap_name, ZFS_TYPE_SNAPSHOT);
	if (tosnap == NULL) {
	err = -1;
	goto err_out;
	}
	err = send_prelim_records(tosnap, fromsnap, outfd,
	flags->replicate \|\| flags->props \|\| flags->holds,
	flags->replicate, flags->verbosity > 0, flags->dryrun,
	flags->raw, flags->replicate, flags->skipmissing,
	flags->backup, flags->holds, flags->props, flags->doall,
	&fss, &fsavl);
	zfs_close(tosnap);
	if (err != 0)
	goto err_out;
	}

	/* dump each stream */
	sdd.fromsnap = fromsnap;
	sdd.tosnap = tosnap;
	sdd.outfd = outfd;
	sdd.replicate = flags->replicate;
	sdd.doall = flags->doall;
	sdd.fromorigin = flags->fromorigin;
	sdd.fss = fss;
	sdd.fsavl = fsavl;
	sdd.verbosity = flags->verbosity;
	sdd.parsable = flags->parsable;
	sdd.progress = flags->progress;
	sdd.progressastitle = flags->progressastitle;
	sdd.dryrun = flags->dryrun;
	sdd.large_block = flags->largeblock;
	sdd.embed_data = flags->embed_data;
	sdd.compress = flags->compress;
	sdd.raw = flags->raw;
	sdd.holds = flags->holds;
	sdd.filter_cb = filter_func;
	sdd.filter_cb_arg = cb_arg;
	if (debugnvp)
	sdd.debugnv = *debugnvp;
	if (sdd.verbosity != 0 && sdd.dryrun)
	sdd.std_out = B_TRUE;
	fout = sdd.std_out ? stdout : stderr;

	/*
	* Some flags require that we place user holds on the datasets that are
	* being sent so they don't get destroyed during the send. We can skip
	* this step if the pool is imported read-only since the datasets cannot
	* be destroyed.
	*/
	if (!flags->dryrun && !zpool_get_prop_int(zfs_get_pool_handle(zhp),
	ZPOOL_PROP_READONLY, NULL) &&
	zfs_spa_version(zhp, &spa_version) == 0 &&
	spa_version >= SPA_VERSION_USERREFS &&
	(flags->doall \|\| flags->replicate)) {
	++holdseq;
	(void) snprintf(sdd.holdtag, sizeof (sdd.holdtag),
	".send-%d-%llu", getpid(), (u_longlong_t)holdseq);
	sdd.cleanup_fd = open(ZFS_DEV, O_RDWR \| O_CLOEXEC);
	if (sdd.cleanup_fd < 0) {
	err = errno;
	goto stderr_out;
	}
	sdd.snapholds = fnvlist_alloc();
	} else {
	sdd.cleanup_fd = -1;
	sdd.snapholds = NULL;
	}

	if (flags->verbosity != 0 \|\| sdd.snapholds != NULL) {
	/*
	* Do a verbose no-op dry run to get all the verbose output
	* or to gather snapshot hold's before generating any data,
	* then do a non-verbose real run to generate the streams.
	*/
	sdd.dryrun = B_TRUE;
	err = dump_filesystems(zhp, &sdd);

	if (err != 0)
	goto stderr_out;

	if (flags->verbosity != 0) {
	if (flags->parsable) {
	(void) fprintf(fout, "size\t%llu\n",
	(longlong_t)sdd.size);
	} else {
	char buf[16];
	zfs_nicebytes(sdd.size, buf, sizeof (buf));
	(void) fprintf(fout, dgettext(TEXT_DOMAIN,
	"total estimated size is %s\n"), buf);
	}
	}

	/* Ensure no snaps found is treated as an error. */
	if (!sdd.seento) {
	err = ENOENT;
	goto err_out;
	}

	/* Skip the second run if dryrun was requested. */
	if (flags->dryrun)
	goto err_out;

	if (sdd.snapholds != NULL) {
	err = zfs_hold_nvl(zhp, sdd.cleanup_fd, sdd.snapholds);
	if (err != 0)
	goto stderr_out;

	fnvlist_free(sdd.snapholds);
	sdd.snapholds = NULL;
	}

	sdd.dryrun = B_FALSE;
	sdd.verbosity = 0;
	}

	err = dump_filesystems(zhp, &sdd);
	fsavl_destroy(fsavl);
	fnvlist_free(fss);

	/* Ensure no snaps found is treated as an error. */
	if (err == 0 && !sdd.seento)
	err = ENOENT;

	if (sdd.cleanup_fd != -1) {
	VERIFY(0 == close(sdd.cleanup_fd));
	sdd.cleanup_fd = -1;
	}

	if (!flags->dryrun && (flags->replicate \|\| flags->doall \|\|
	flags->props \|\| flags->backup \|\| flags->holds)) {
	/*
	* write final end record. NB: want to do this even if
	* there was some error, because it might not be totally
	* failed.
	*/
	int err2 = send_conclusion_record(outfd, NULL);
	if (err2 != 0)
	return (zfs_standard_error(zhp->zfs_hdl, err2, errbuf));
	}

	return (err \|\| sdd.err);

	stderr_out:
	err = zfs_standard_error(zhp->zfs_hdl, err, errbuf);
	err_out:
	fsavl_destroy(fsavl);
	fnvlist_free(fss);
	fnvlist_free(sdd.snapholds);

	if (sdd.cleanup_fd != -1)
	VERIFY(0 == close(sdd.cleanup_fd));
	return (err);
	}

	struct zfs_send {
	zfs_handle_t *zhp;
	const char *fromsnap;
	const char *tosnap;
	sendflags_t *flags;
	snapfilter_cb_t *filter_func;
	void *cb_arg;
	nvlist_t **debugnvp;
	};

	static int
	zfs_send_cb(int outfd, void *arg)
	{
	struct zfs_send *zs = arg;
	return (zfs_send_cb_impl(zs->zhp, zs->fromsnap, zs->tosnap, zs->flags,
	outfd, zs->filter_func, zs->cb_arg, zs->debugnvp));
	}

	int
	zfs_send(zfs_handle_t zhp, const char fromsnap, const char *tosnap,
	sendflags_t *flags, int outfd, snapfilter_cb_t filter_func,
	void cb_arg, nvlist_t *debugnvp)
	{
	struct zfs_send arg = {
	.zhp = zhp,
	.fromsnap = fromsnap,
	.tosnap = tosnap,
	.flags = flags,
	.filter_func = filter_func,
	.cb_arg = cb_arg,
	.debugnvp = debugnvp,
	};
	return (lzc_send_wrapper(zfs_send_cb, outfd, &arg));
	}


	static zfs_handle_t *
	name_to_dir_handle(libzfs_handle_t hdl, const char snapname)
	{
	char dirname[ZFS_MAX_DATASET_NAME_LEN];
	(void) strlcpy(dirname, snapname, ZFS_MAX_DATASET_NAME_LEN);
	char *c = strchr(dirname, '@');
	if (c != NULL)
	*c = '\0';
	return (zfs_open(hdl, dirname, ZFS_TYPE_DATASET));
	}

	/*
	* Returns B_TRUE if earlier is an earlier snapshot in later's timeline; either
	* an earlier snapshot in the same filesystem, or a snapshot before later's
	* origin, or it's origin's origin, etc.
	*/
	static boolean_t
	snapshot_is_before(zfs_handle_t earlier, zfs_handle_t later)
	{
	boolean_t ret;
	uint64_t later_txg =
	(later->zfs_type == ZFS_TYPE_FILESYSTEM \|\|
	later->zfs_type == ZFS_TYPE_VOLUME ?
	UINT64_MAX : zfs_prop_get_int(later, ZFS_PROP_CREATETXG));
	uint64_t earlier_txg = zfs_prop_get_int(earlier, ZFS_PROP_CREATETXG);

	if (earlier_txg >= later_txg)
	return (B_FALSE);

	zfs_handle_t *earlier_dir = name_to_dir_handle(earlier->zfs_hdl,
	earlier->zfs_name);
	zfs_handle_t *later_dir = name_to_dir_handle(later->zfs_hdl,
	later->zfs_name);

	if (strcmp(earlier_dir->zfs_name, later_dir->zfs_name) == 0) {
	zfs_close(earlier_dir);
	zfs_close(later_dir);
	return (B_TRUE);
	}

	char clonename[ZFS_MAX_DATASET_NAME_LEN];
	if (zfs_prop_get(later_dir, ZFS_PROP_ORIGIN, clonename,
	ZFS_MAX_DATASET_NAME_LEN, NULL, NULL, 0, B_TRUE) != 0) {
	zfs_close(earlier_dir);
	zfs_close(later_dir);
	return (B_FALSE);
	}

	zfs_handle_t *origin = zfs_open(earlier->zfs_hdl, clonename,
	ZFS_TYPE_DATASET);
	uint64_t origin_txg = zfs_prop_get_int(origin, ZFS_PROP_CREATETXG);

	/*
	* If "earlier" is exactly the origin, then
	* snapshot_is_before(earlier, origin) will return false (because
	* they're the same).
	*/
	if (origin_txg == earlier_txg &&
	strcmp(origin->zfs_name, earlier->zfs_name) == 0) {
	zfs_close(earlier_dir);
	zfs_close(later_dir);
	zfs_close(origin);
	return (B_TRUE);
	}
	zfs_close(earlier_dir);
	zfs_close(later_dir);

	ret = snapshot_is_before(earlier, origin);
	zfs_close(origin);
	return (ret);
	}

	/*
	* The "zhp" argument is the handle of the dataset to send (typically a
	* snapshot). The "from" argument is the full name of the snapshot or
	* bookmark that is the incremental source.
	*
	* Pre-wrapped (cf. lzc_send_wrapper()).
	*/
	static int
	zfs_send_one_cb_impl(zfs_handle_t zhp, const char from, int fd,
	sendflags_t flags, const char redactbook)
	{
	int err;
	libzfs_handle_t *hdl = zhp->zfs_hdl;
	char *name = zhp->zfs_name;
	pthread_t ptid;
	progress_arg_t pa = { 0 };
	uint64_t size = 0;

	char errbuf[ERRBUFLEN];
	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"warning: cannot send '%s'"), name);

	if (from != NULL && strchr(from, '@')) {
	zfs_handle_t *from_zhp = zfs_open(hdl, from,
	ZFS_TYPE_DATASET);
	if (from_zhp == NULL)
	return (-1);
	if (!snapshot_is_before(from_zhp, zhp)) {
	zfs_close(from_zhp);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"not an earlier snapshot from the same fs"));
	return (zfs_error(hdl, EZFS_CROSSTARGET, errbuf));
	}
	zfs_close(from_zhp);
	}

	if (redactbook != NULL) {
	char bookname[ZFS_MAX_DATASET_NAME_LEN];
	nvlist_t *redact_snaps;
	zfs_handle_t *book_zhp;
	char at, pound;
	int dsnamelen;

	pound = strchr(redactbook, '#');
	if (pound != NULL)
	redactbook = pound + 1;
	at = strchr(name, '@');
	if (at == NULL) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"cannot do a redacted send to a filesystem"));
	return (zfs_error(hdl, EZFS_BADTYPE, errbuf));
	}
	dsnamelen = at - name;
	if (snprintf(bookname, sizeof (bookname), "%.*s#%s",
	dsnamelen, name, redactbook)
	>= sizeof (bookname)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"invalid bookmark name"));
	return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
	}
	book_zhp = zfs_open(hdl, bookname, ZFS_TYPE_BOOKMARK);
	if (book_zhp == NULL)
	return (-1);
	if (nvlist_lookup_nvlist(book_zhp->zfs_props,
	zfs_prop_to_name(ZFS_PROP_REDACT_SNAPS),
	&redact_snaps) != 0 \|\| redact_snaps == NULL) {
	zfs_close(book_zhp);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"not a redaction bookmark"));
	return (zfs_error(hdl, EZFS_BADTYPE, errbuf));
	}
	zfs_close(book_zhp);
	}

	/*
	* Send fs properties
	*/
	if (flags->props \|\| flags->holds \|\| flags->backup) {
	/*
	* Note: the header generated by send_prelim_records()
	* assumes that the incremental source is in the same
	* filesystem/volume as the target (which is a requirement
	* when doing "zfs send -R"). But that isn't always the
	* case here (e.g. send from snap in origin, or send from
	* bookmark). We pass from=NULL, which will omit this
	* information from the prelim records; it isn't used
	* when receiving this type of stream.
	*/
	err = send_prelim_records(zhp, NULL, fd, B_TRUE, B_FALSE,
	flags->verbosity > 0, flags->dryrun, flags->raw,
	flags->replicate, B_FALSE, flags->backup, flags->holds,
	flags->props, flags->doall, NULL, NULL);
	if (err != 0)
	return (err);
	}

	/*
	* Perform size estimate if verbose was specified.
	*/
	if (flags->verbosity != 0 \|\| flags->progressastitle) {
	err = estimate_size(zhp, from, fd, flags, 0, 0, 0, redactbook,
	errbuf, &size);
	if (err != 0)
	return (err);
	}

	if (flags->dryrun)
	return (0);

	/*
	* If progress reporting is requested, spawn a new thread to poll
	* ZFS_IOC_SEND_PROGRESS at a regular interval.
	*/
	sigset_t oldmask;
	{
	pa.pa_zhp = zhp;
	pa.pa_fd = fd;
	pa.pa_parsable = flags->parsable;
	pa.pa_estimate = B_FALSE;
	pa.pa_verbosity = flags->verbosity;
	pa.pa_size = size;
	pa.pa_astitle = flags->progressastitle;
	pa.pa_progress = flags->progress;

	err = pthread_create(&ptid, NULL,
	send_progress_thread, &pa);
	if (err != 0) {
	zfs_error_aux(zhp->zfs_hdl, "%s", zfs_strerror(errno));
	return (zfs_error(zhp->zfs_hdl,
	EZFS_THREADCREATEFAILED, errbuf));
	}
	SEND_PROGRESS_THREAD_PARENT_BLOCK(&oldmask);
	}

	err = lzc_send_redacted(name, from, fd,
	lzc_flags_from_sendflags(flags), redactbook);

	if (send_progress_thread_exit(hdl, ptid, &oldmask))
	return (-1);

	if (err == 0 && (flags->props \|\| flags->holds \|\| flags->backup)) {
	/* Write the final end record. */
	err = send_conclusion_record(fd, NULL);
	if (err != 0)
	return (zfs_standard_error(hdl, err, errbuf));
	}
	if (err != 0) {
	switch (errno) {
	case EXDEV:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"not an earlier snapshot from the same fs"));
	return (zfs_error(hdl, EZFS_CROSSTARGET, errbuf));

	case ENOENT:
	case ESRCH:
	if (lzc_exists(name)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"incremental source (%s) does not exist"),
	from);
	}
	return (zfs_error(hdl, EZFS_NOENT, errbuf));

	case EACCES:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"dataset key must be loaded"));
	return (zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf));

	case EBUSY:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"target is busy; if a filesystem, "
	"it must not be mounted"));
	return (zfs_error(hdl, EZFS_BUSY, errbuf));

	case EDQUOT:
	case EFAULT:
	case EFBIG:
	case EINVAL:
	case EIO:
	case ENOLINK:
	case ENOSPC:
	case ENOSTR:
	case ENXIO:
	case EPIPE:
	case ERANGE:
	case EROFS:
	zfs_error_aux(hdl, "%s", zfs_strerror(errno));
	return (zfs_error(hdl, EZFS_BADBACKUP, errbuf));

	default:
	return (zfs_standard_error(hdl, errno, errbuf));
	}
	}
	return (err != 0);
	}

	struct zfs_send_one {
	zfs_handle_t *zhp;
	const char *from;
	sendflags_t *flags;
	const char *redactbook;
	};

	static int
	zfs_send_one_cb(int fd, void *arg)
	{
	struct zfs_send_one *zso = arg;
	return (zfs_send_one_cb_impl(zso->zhp, zso->from, fd, zso->flags,
	zso->redactbook));
	}

	int
	zfs_send_one(zfs_handle_t zhp, const char from, int fd, sendflags_t *flags,
	const char *redactbook)
	{
	struct zfs_send_one zso = {
	.zhp = zhp,
	.from = from,
	.flags = flags,
	.redactbook = redactbook,
	};
	return (lzc_send_wrapper(zfs_send_one_cb, fd, &zso));
	}

	/*
	* Routines specific to "zfs recv"
	*/

	static int
	recv_read(libzfs_handle_t hdl, int fd, void buf, int ilen,
	boolean_t byteswap, zio_cksum_t *zc)
	{
	char *cp = buf;
	int rv;
	int len = ilen;

	do {
	rv = read(fd, cp, len);
	cp += rv;
	len -= rv;
	} while (rv > 0);

	if (rv < 0 \|\| len != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"failed to read from stream"));
	return (zfs_error(hdl, EZFS_BADSTREAM, dgettext(TEXT_DOMAIN,
	"cannot receive")));
	}

	if (zc) {
	if (byteswap)
	fletcher_4_incremental_byteswap(buf, ilen, zc);
	else
	fletcher_4_incremental_native(buf, ilen, zc);
	}
	return (0);
	}

	static int
	recv_read_nvlist(libzfs_handle_t hdl, int fd, int len, nvlist_t *nvp,
	boolean_t byteswap, zio_cksum_t *zc)
	{
	char *buf;
	int err;

	buf = zfs_alloc(hdl, len);

	if (len > hdl->libzfs_max_nvlist) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "nvlist too large"));
	free(buf);
	return (ENOMEM);
	}

	err = recv_read(hdl, fd, buf, len, byteswap, zc);
	if (err != 0) {
	free(buf);
	return (err);
	}

	err = nvlist_unpack(buf, len, nvp, 0);
	free(buf);
	if (err != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid "
	"stream (malformed nvlist)"));
	return (EINVAL);
	}
	return (0);
	}

	/*
	* Returns the grand origin (origin of origin of origin...) of a given handle.
	* If this dataset is not a clone, it simply returns a copy of the original
	* handle.
	*/
	static zfs_handle_t *
	recv_open_grand_origin(zfs_handle_t *zhp)
	{
	char origin[ZFS_MAX_DATASET_NAME_LEN];
	zprop_source_t src;
	zfs_handle_t *ozhp = zfs_handle_dup(zhp);

	while (ozhp != NULL) {
	if (zfs_prop_get(ozhp, ZFS_PROP_ORIGIN, origin,
	sizeof (origin), &src, NULL, 0, B_FALSE) != 0)
	break;

	(void) zfs_close(ozhp);
	ozhp = zfs_open(zhp->zfs_hdl, origin, ZFS_TYPE_FILESYSTEM);
	}

	return (ozhp);
	}

	static int
	recv_rename_impl(zfs_handle_t zhp, const char name, const char *newname)
	{
	int err;
	zfs_handle_t *ozhp = NULL;

	/*
	* Attempt to rename the dataset. If it fails with EACCES we have
	* attempted to rename the dataset outside of its encryption root.
	* Force the dataset to become an encryption root and try again.
	*/
	err = lzc_rename(name, newname);
	if (err == EACCES) {
	ozhp = recv_open_grand_origin(zhp);
	if (ozhp == NULL) {
	err = ENOENT;
	goto out;
	}

	err = lzc_change_key(ozhp->zfs_name, DCP_CMD_FORCE_NEW_KEY,
	NULL, NULL, 0);
	if (err != 0)
	goto out;

	err = lzc_rename(name, newname);
	}

	out:
	if (ozhp != NULL)
	zfs_close(ozhp);
	return (err);
	}

	static int
	recv_rename(libzfs_handle_t hdl, const char name, const char *tryname,
	int baselen, char newname, recvflags_t flags)
	{
	static int seq;
	int err;
	prop_changelist_t *clp = NULL;
	zfs_handle_t *zhp = NULL;

	zhp = zfs_open(hdl, name, ZFS_TYPE_DATASET);
	if (zhp == NULL) {
	err = -1;
	goto out;
	}
	clp = changelist_gather(zhp, ZFS_PROP_NAME, 0,
	flags->force ? MS_FORCE : 0);
	if (clp == NULL) {
	err = -1;
	goto out;
	}
	err = changelist_prefix(clp);
	if (err)
	goto out;

	if (tryname) {
	(void) strlcpy(newname, tryname, ZFS_MAX_DATASET_NAME_LEN);
	if (flags->verbose) {
	(void) printf("attempting rename %s to %s\n",
	name, newname);
	}
	err = recv_rename_impl(zhp, name, newname);
	if (err == 0)
	changelist_rename(clp, name, tryname);
	} else {
	err = ENOENT;
	}

	if (err != 0 && strncmp(name + baselen, "recv-", 5) != 0) {
	seq++;

	(void) snprintf(newname, ZFS_MAX_DATASET_NAME_LEN,
	"%.*srecv-%u-%u", baselen, name, getpid(), seq);

	if (flags->verbose) {
	(void) printf("failed - trying rename %s to %s\n",
	name, newname);
	}
	err = recv_rename_impl(zhp, name, newname);
	if (err == 0)
	changelist_rename(clp, name, newname);
	if (err && flags->verbose) {
	(void) printf("failed (%u) - "
	"will try again on next pass\n", errno);
	}
	err = EAGAIN;
	} else if (flags->verbose) {
	if (err == 0)
	(void) printf("success\n");
	else
	(void) printf("failed (%u)\n", errno);
	}

	(void) changelist_postfix(clp);

	out:
	if (clp != NULL)
	changelist_free(clp);
	if (zhp != NULL)
	zfs_close(zhp);

	return (err);
	}

	static int
	recv_promote(libzfs_handle_t hdl, const char fsname,
	const char origin_fsname, recvflags_t flags)
	{
	int err;
	zfs_cmd_t zc = {"\0"};
	zfs_handle_t zhp = NULL, ozhp = NULL;

	if (flags->verbose)
	(void) printf("promoting %s\n", fsname);

	(void) strlcpy(zc.zc_value, origin_fsname, sizeof (zc.zc_value));
	(void) strlcpy(zc.zc_name, fsname, sizeof (zc.zc_name));

	/*
	* Attempt to promote the dataset. If it fails with EACCES the
	* promotion would cause this dataset to leave its encryption root.
	* Force the origin to become an encryption root and try again.
	*/
	err = zfs_ioctl(hdl, ZFS_IOC_PROMOTE, &zc);
	if (err == EACCES) {
	zhp = zfs_open(hdl, fsname, ZFS_TYPE_DATASET);
	if (zhp == NULL) {
	err = -1;
	goto out;
	}

	ozhp = recv_open_grand_origin(zhp);
	if (ozhp == NULL) {
	err = -1;
	goto out;
	}

	err = lzc_change_key(ozhp->zfs_name, DCP_CMD_FORCE_NEW_KEY,
	NULL, NULL, 0);
	if (err != 0)
	goto out;

	err = zfs_ioctl(hdl, ZFS_IOC_PROMOTE, &zc);
	}

	out:
	if (zhp != NULL)
	zfs_close(zhp);
	if (ozhp != NULL)
	zfs_close(ozhp);

	return (err);
	}

	static int
	recv_destroy(libzfs_handle_t hdl, const char name, int baselen,
	char newname, recvflags_t flags)
	{
	int err = 0;
	prop_changelist_t *clp;
	zfs_handle_t *zhp;
	boolean_t defer = B_FALSE;
	int spa_version;

	zhp = zfs_open(hdl, name, ZFS_TYPE_DATASET);
	if (zhp == NULL)
	return (-1);
	zfs_type_t type = zfs_get_type(zhp);
	if (type == ZFS_TYPE_SNAPSHOT &&
	zfs_spa_version(zhp, &spa_version) == 0 &&
	spa_version >= SPA_VERSION_USERREFS)
	defer = B_TRUE;
	clp = changelist_gather(zhp, ZFS_PROP_NAME, 0,
	flags->force ? MS_FORCE : 0);
	zfs_close(zhp);
	if (clp == NULL)
	return (-1);

	err = changelist_prefix(clp);
	if (err)
	return (err);

	if (flags->verbose)
	(void) printf("attempting destroy %s\n", name);
	if (type == ZFS_TYPE_SNAPSHOT) {
	nvlist_t *nv = fnvlist_alloc();
	fnvlist_add_boolean(nv, name);
	err = lzc_destroy_snaps(nv, defer, NULL);
	fnvlist_free(nv);
	} else {
	err = lzc_destroy(name);
	}
	if (err == 0) {
	if (flags->verbose)
	(void) printf("success\n");
	changelist_remove(clp, name);
	}

	(void) changelist_postfix(clp);
	changelist_free(clp);

	/*
	* Deferred destroy might destroy the snapshot or only mark it to be
	* destroyed later, and it returns success in either case.
	*/
	if (err != 0 \|\| (defer && zfs_dataset_exists(hdl, name,
	ZFS_TYPE_SNAPSHOT))) {
	err = recv_rename(hdl, name, NULL, baselen, newname, flags);
	}

	return (err);
	}

	typedef struct guid_to_name_data {
	uint64_t guid;
	boolean_t bookmark_ok;
	char *name;
	char *skip;
	uint64_t *redact_snap_guids;
	uint64_t num_redact_snaps;
	} guid_to_name_data_t;

	static boolean_t
	redact_snaps_match(zfs_handle_t zhp, guid_to_name_data_t gtnd)
	{
	uint64_t *bmark_snaps;
	uint_t bmark_num_snaps;
	nvlist_t *nvl;
	if (zhp->zfs_type != ZFS_TYPE_BOOKMARK)
	return (B_FALSE);

	nvl = fnvlist_lookup_nvlist(zhp->zfs_props,
	zfs_prop_to_name(ZFS_PROP_REDACT_SNAPS));
	bmark_snaps = fnvlist_lookup_uint64_array(nvl, ZPROP_VALUE,
	&bmark_num_snaps);
	if (bmark_num_snaps != gtnd->num_redact_snaps)
	return (B_FALSE);
	int i = 0;
	for (; i < bmark_num_snaps; i++) {
	int j = 0;
	for (; j < bmark_num_snaps; j++) {
	if (bmark_snaps[i] == gtnd->redact_snap_guids[j])
	break;
	}
	if (j == bmark_num_snaps)
	break;
	}
	return (i == bmark_num_snaps);
	}

	static int
	guid_to_name_cb(zfs_handle_t zhp, void arg)
	{
	guid_to_name_data_t *gtnd = arg;
	const char *slash;
	int err;

	if (gtnd->skip != NULL &&
	(slash = strrchr(zhp->zfs_name, '/')) != NULL &&
	strcmp(slash + 1, gtnd->skip) == 0) {
	zfs_close(zhp);
	return (0);
	}

	if (zfs_prop_get_int(zhp, ZFS_PROP_GUID) == gtnd->guid &&
	(gtnd->num_redact_snaps == -1 \|\| redact_snaps_match(zhp, gtnd))) {
	(void) strcpy(gtnd->name, zhp->zfs_name);
	zfs_close(zhp);
	return (EEXIST);
	}

	err = zfs_iter_children_v2(zhp, 0, guid_to_name_cb, gtnd);
	if (err != EEXIST && gtnd->bookmark_ok)
	err = zfs_iter_bookmarks_v2(zhp, 0, guid_to_name_cb, gtnd);
	zfs_close(zhp);
	return (err);
	}

	/*
	* Attempt to find the local dataset associated with this guid. In the case of
	* multiple matches, we attempt to find the "best" match by searching
	* progressively larger portions of the hierarchy. This allows one to send a
	* tree of datasets individually and guarantee that we will find the source
	* guid within that hierarchy, even if there are multiple matches elsewhere.
	*
	* If num_redact_snaps is not -1, we attempt to find a redaction bookmark with
	* the specified number of redaction snapshots. If num_redact_snaps isn't 0 or
	* -1, then redact_snap_guids will be an array of the guids of the snapshots the
	* redaction bookmark was created with. If num_redact_snaps is -1, then we will
	* attempt to find a snapshot or bookmark (if bookmark_ok is passed) with the
	* given guid. Note that a redaction bookmark can be returned if
	* num_redact_snaps == -1.
	*/
	static int
	guid_to_name_redact_snaps(libzfs_handle_t hdl, const char parent,
	uint64_t guid, boolean_t bookmark_ok, uint64_t *redact_snap_guids,
	uint64_t num_redact_snaps, char *name)
	{
	char pname[ZFS_MAX_DATASET_NAME_LEN];
	guid_to_name_data_t gtnd;

	gtnd.guid = guid;
	gtnd.bookmark_ok = bookmark_ok;
	gtnd.name = name;
	gtnd.skip = NULL;
	gtnd.redact_snap_guids = redact_snap_guids;
	gtnd.num_redact_snaps = num_redact_snaps;

	/*
	* Search progressively larger portions of the hierarchy, starting
	* with the filesystem specified by 'parent'. This will
	* select the "most local" version of the origin snapshot in the case
	* that there are multiple matching snapshots in the system.
	*/
	(void) strlcpy(pname, parent, sizeof (pname));
	char *cp = strrchr(pname, '@');
	if (cp == NULL)
	cp = strchr(pname, '\0');
	for (; cp != NULL; cp = strrchr(pname, '/')) {
	/* Chop off the last component and open the parent */
	*cp = '\0';
	zfs_handle_t *zhp = make_dataset_handle(hdl, pname);

	if (zhp == NULL)
	continue;
	int err = guid_to_name_cb(zfs_handle_dup(zhp), &gtnd);
	if (err != EEXIST)
	err = zfs_iter_children_v2(zhp, 0, guid_to_name_cb,
	&gtnd);
	if (err != EEXIST && bookmark_ok)
	err = zfs_iter_bookmarks_v2(zhp, 0, guid_to_name_cb,
	&gtnd);
	zfs_close(zhp);
	if (err == EEXIST)
	return (0);

	/*
	* Remember the last portion of the dataset so we skip it next
	* time through (as we've already searched that portion of the
	* hierarchy).
	*/
	gtnd.skip = strrchr(pname, '/') + 1;
	}

	return (ENOENT);
	}

	static int
	guid_to_name(libzfs_handle_t hdl, const char parent, uint64_t guid,
	boolean_t bookmark_ok, char *name)
	{
	return (guid_to_name_redact_snaps(hdl, parent, guid, bookmark_ok, NULL,
	-1, name));
	}

	/*
	* Return +1 if guid1 is before guid2, 0 if they are the same, and -1 if
	* guid1 is after guid2.
	*/
	static int
	created_before(libzfs_handle_t hdl, avl_tree_t avl,
	uint64_t guid1, uint64_t guid2)
	{
	nvlist_t *nvfs;
	const char fsname = NULL, snapname = NULL;
	char buf[ZFS_MAX_DATASET_NAME_LEN];
	int rv;
	zfs_handle_t guid1hdl, guid2hdl;
	uint64_t create1, create2;

	if (guid2 == 0)
	return (0);
	if (guid1 == 0)
	return (1);

	nvfs = fsavl_find(avl, guid1, &snapname);
	fsname = fnvlist_lookup_string(nvfs, "name");
	(void) snprintf(buf, sizeof (buf), "%s@%s", fsname, snapname);
	guid1hdl = zfs_open(hdl, buf, ZFS_TYPE_SNAPSHOT);
	if (guid1hdl == NULL)
	return (-1);

	nvfs = fsavl_find(avl, guid2, &snapname);
	fsname = fnvlist_lookup_string(nvfs, "name");
	(void) snprintf(buf, sizeof (buf), "%s@%s", fsname, snapname);
	guid2hdl = zfs_open(hdl, buf, ZFS_TYPE_SNAPSHOT);
	if (guid2hdl == NULL) {
	zfs_close(guid1hdl);
	return (-1);
	}

	create1 = zfs_prop_get_int(guid1hdl, ZFS_PROP_CREATETXG);
	create2 = zfs_prop_get_int(guid2hdl, ZFS_PROP_CREATETXG);

	if (create1 < create2)
	rv = -1;
	else if (create1 > create2)
	rv = +1;
	else
	rv = 0;

	zfs_close(guid1hdl);
	zfs_close(guid2hdl);

	return (rv);
	}

	/*
	* This function reestablishes the hierarchy of encryption roots after a
	* recursive incremental receive has completed. This must be done after the
	* second call to recv_incremental_replication() has renamed and promoted all
	* sent datasets to their final locations in the dataset hierarchy.
	*/
	static int
	recv_fix_encryption_hierarchy(libzfs_handle_t hdl, const char top_zfs,
	nvlist_t *stream_nv)
	{
	int err;
	nvpair_t *fselem = NULL;
	nvlist_t *stream_fss;

	stream_fss = fnvlist_lookup_nvlist(stream_nv, "fss");

	while ((fselem = nvlist_next_nvpair(stream_fss, fselem)) != NULL) {
	zfs_handle_t *zhp = NULL;
	uint64_t crypt;
	nvlist_t snaps, props, *stream_nvfs = NULL;
	nvpair_t *snapel = NULL;
	boolean_t is_encroot, is_clone, stream_encroot;
	char *cp;
	const char *stream_keylocation = NULL;
	char keylocation[MAXNAMELEN];
	char fsname[ZFS_MAX_DATASET_NAME_LEN];

	keylocation[0] = '\0';
	stream_nvfs = fnvpair_value_nvlist(fselem);
	snaps = fnvlist_lookup_nvlist(stream_nvfs, "snaps");
	props = fnvlist_lookup_nvlist(stream_nvfs, "props");
	stream_encroot = nvlist_exists(stream_nvfs, "is_encroot");

	/* find a snapshot from the stream that exists locally */
	err = ENOENT;
	while ((snapel = nvlist_next_nvpair(snaps, snapel)) != NULL) {
	uint64_t guid;

	guid = fnvpair_value_uint64(snapel);
	err = guid_to_name(hdl, top_zfs, guid, B_FALSE,
	fsname);
	if (err == 0)
	break;
	}

	if (err != 0)
	continue;

	cp = strchr(fsname, '@');
	if (cp != NULL)
	*cp = '\0';

	zhp = zfs_open(hdl, fsname, ZFS_TYPE_DATASET);
	if (zhp == NULL) {
	err = ENOENT;
	goto error;
	}

	crypt = zfs_prop_get_int(zhp, ZFS_PROP_ENCRYPTION);
	is_clone = zhp->zfs_dmustats.dds_origin[0] != '\0';
	(void) zfs_crypto_get_encryption_root(zhp, &is_encroot, NULL);

	/* we don't need to do anything for unencrypted datasets */
	if (crypt == ZIO_CRYPT_OFF) {
	zfs_close(zhp);
	continue;
	}

	/*
	* If the dataset is flagged as an encryption root, was not
	* received as a clone and is not currently an encryption root,
	* force it to become one. Fixup the keylocation if necessary.
	*/
	if (stream_encroot) {
	if (!is_clone && !is_encroot) {
	err = lzc_change_key(fsname,
	DCP_CMD_FORCE_NEW_KEY, NULL, NULL, 0);
	if (err != 0) {
	zfs_close(zhp);
	goto error;
	}
	}

	stream_keylocation = fnvlist_lookup_string(props,
	zfs_prop_to_name(ZFS_PROP_KEYLOCATION));

	/*
	* Refresh the properties in case the call to
	* lzc_change_key() changed the value.
	*/
	zfs_refresh_properties(zhp);
	err = zfs_prop_get(zhp, ZFS_PROP_KEYLOCATION,
	keylocation, sizeof (keylocation), NULL, NULL,
	0, B_TRUE);
	if (err != 0) {
	zfs_close(zhp);
	goto error;
	}

	if (strcmp(keylocation, stream_keylocation) != 0) {
	err = zfs_prop_set(zhp,
	zfs_prop_to_name(ZFS_PROP_KEYLOCATION),
	stream_keylocation);
	if (err != 0) {
	zfs_close(zhp);
	goto error;
	}
	}
	}

	/*
	* If the dataset is not flagged as an encryption root and is
	* currently an encryption root, force it to inherit from its
	* parent. The root of a raw send should never be
	* force-inherited.
	*/
	if (!stream_encroot && is_encroot &&
	strcmp(top_zfs, fsname) != 0) {
	err = lzc_change_key(fsname, DCP_CMD_FORCE_INHERIT,
	NULL, NULL, 0);
	if (err != 0) {
	zfs_close(zhp);
	goto error;
	}
	}

	zfs_close(zhp);
	}

	return (0);

	error:
	return (err);
	}

	static int
	recv_incremental_replication(libzfs_handle_t hdl, const char tofs,
	recvflags_t flags, nvlist_t stream_nv, avl_tree_t *stream_avl,
	nvlist_t *renamed)
	{
	nvlist_t local_nv, deleted = NULL;
	avl_tree_t *local_avl;
	nvpair_t fselem, nextfselem;
	const char *fromsnap;
	char newname[ZFS_MAX_DATASET_NAME_LEN];
	char guidname[32];
	int error;
	boolean_t needagain, progress, recursive;
	const char s1, s2;

	fromsnap = fnvlist_lookup_string(stream_nv, "fromsnap");

	recursive = (nvlist_lookup_boolean(stream_nv, "not_recursive") ==
	ENOENT);

	if (flags->dryrun)
	return (0);

	again:
	needagain = progress = B_FALSE;

	deleted = fnvlist_alloc();

	if ((error = gather_nvlist(hdl, tofs, fromsnap, NULL,
	recursive, B_TRUE, B_FALSE, recursive, B_FALSE, B_FALSE, B_FALSE,
	B_FALSE, B_TRUE, &local_nv, &local_avl)) != 0)
	return (error);

	/*
	* Process deletes and renames
	*/
	for (fselem = nvlist_next_nvpair(local_nv, NULL);
	fselem; fselem = nextfselem) {
	nvlist_t nvfs, snaps;
	nvlist_t *stream_nvfs = NULL;
	nvpair_t snapelem, nextsnapelem;
	uint64_t fromguid = 0;
	uint64_t originguid = 0;
	uint64_t stream_originguid = 0;
	uint64_t parent_fromsnap_guid, stream_parent_fromsnap_guid;
	const char fsname, stream_fsname;

	nextfselem = nvlist_next_nvpair(local_nv, fselem);

	nvfs = fnvpair_value_nvlist(fselem);
	snaps = fnvlist_lookup_nvlist(nvfs, "snaps");
	fsname = fnvlist_lookup_string(nvfs, "name");
	parent_fromsnap_guid = fnvlist_lookup_uint64(nvfs,
	"parentfromsnap");
	(void) nvlist_lookup_uint64(nvfs, "origin", &originguid);

	/*
	* First find the stream's fs, so we can check for
	* a different origin (due to "zfs promote")
	*/
	for (snapelem = nvlist_next_nvpair(snaps, NULL);
	snapelem; snapelem = nvlist_next_nvpair(snaps, snapelem)) {
	uint64_t thisguid;

	thisguid = fnvpair_value_uint64(snapelem);
	stream_nvfs = fsavl_find(stream_avl, thisguid, NULL);

	if (stream_nvfs != NULL)
	break;
	}

	/* check for promote */
	(void) nvlist_lookup_uint64(stream_nvfs, "origin",
	&stream_originguid);
	if (stream_nvfs && originguid != stream_originguid) {
	switch (created_before(hdl, local_avl,
	stream_originguid, originguid)) {
	case 1: {
	/* promote it! */
	nvlist_t *origin_nvfs;
	const char *origin_fsname;

	origin_nvfs = fsavl_find(local_avl, originguid,
	NULL);
	origin_fsname = fnvlist_lookup_string(
	origin_nvfs, "name");
	error = recv_promote(hdl, fsname, origin_fsname,
	flags);
	if (error == 0)
	progress = B_TRUE;
	break;
	}
	default:
	break;
	case -1:
	fsavl_destroy(local_avl);
	fnvlist_free(local_nv);
	return (-1);
	}
	/*
	* We had/have the wrong origin, therefore our
	* list of snapshots is wrong. Need to handle
	* them on the next pass.
	*/
	needagain = B_TRUE;
	continue;
	}

	for (snapelem = nvlist_next_nvpair(snaps, NULL);
	snapelem; snapelem = nextsnapelem) {
	uint64_t thisguid;
	const char *stream_snapname;
	nvlist_t found, props;

	nextsnapelem = nvlist_next_nvpair(snaps, snapelem);

	thisguid = fnvpair_value_uint64(snapelem);
	found = fsavl_find(stream_avl, thisguid,
	&stream_snapname);

	/* check for delete */
	if (found == NULL) {
	char name[ZFS_MAX_DATASET_NAME_LEN];

	if (!flags->force)
	continue;

	(void) snprintf(name, sizeof (name), "%s@%s",
	fsname, nvpair_name(snapelem));

	error = recv_destroy(hdl, name,
	strlen(fsname)+1, newname, flags);
	if (error)
	needagain = B_TRUE;
	else
	progress = B_TRUE;
	sprintf(guidname, "%llu",
	(u_longlong_t)thisguid);
	nvlist_add_boolean(deleted, guidname);
	continue;
	}

	stream_nvfs = found;

	if (0 == nvlist_lookup_nvlist(stream_nvfs, "snapprops",
	&props) && 0 == nvlist_lookup_nvlist(props,
	stream_snapname, &props)) {
	zfs_cmd_t zc = {"\0"};

	zc.zc_cookie = B_TRUE; /* received */
	(void) snprintf(zc.zc_name, sizeof (zc.zc_name),
	"%s@%s", fsname, nvpair_name(snapelem));
	zcmd_write_src_nvlist(hdl, &zc, props);
	(void) zfs_ioctl(hdl,
	ZFS_IOC_SET_PROP, &zc);
	zcmd_free_nvlists(&zc);
	}

	/* check for different snapname */
	if (strcmp(nvpair_name(snapelem),
	stream_snapname) != 0) {
	char name[ZFS_MAX_DATASET_NAME_LEN];
	char tryname[ZFS_MAX_DATASET_NAME_LEN];

	(void) snprintf(name, sizeof (name), "%s@%s",
	fsname, nvpair_name(snapelem));
	(void) snprintf(tryname, sizeof (name), "%s@%s",
	fsname, stream_snapname);

	error = recv_rename(hdl, name, tryname,
	strlen(fsname)+1, newname, flags);
	if (error)
	needagain = B_TRUE;
	else
	progress = B_TRUE;
	}

	if (strcmp(stream_snapname, fromsnap) == 0)
	fromguid = thisguid;
	}

	/* check for delete */
	if (stream_nvfs == NULL) {
	if (!flags->force)
	continue;

	error = recv_destroy(hdl, fsname, strlen(tofs)+1,
	newname, flags);
	if (error)
	needagain = B_TRUE;
	else
	progress = B_TRUE;
	sprintf(guidname, "%llu",
	(u_longlong_t)parent_fromsnap_guid);
	nvlist_add_boolean(deleted, guidname);
	continue;
	}

	if (fromguid == 0) {
	if (flags->verbose) {
	(void) printf("local fs %s does not have "
	"fromsnap (%s in stream); must have "
	"been deleted locally; ignoring\n",
	fsname, fromsnap);
	}
	continue;
	}

	stream_fsname = fnvlist_lookup_string(stream_nvfs, "name");
	stream_parent_fromsnap_guid = fnvlist_lookup_uint64(
	stream_nvfs, "parentfromsnap");

	s1 = strrchr(fsname, '/');
	s2 = strrchr(stream_fsname, '/');

	/*
	* Check if we're going to rename based on parent guid change
	* and the current parent guid was also deleted. If it was then
	* rename will fail and is likely unneeded, so avoid this and
	* force an early retry to determine the new
	* parent_fromsnap_guid.
	*/
	if (stream_parent_fromsnap_guid != 0 &&
	parent_fromsnap_guid != 0 &&
	stream_parent_fromsnap_guid != parent_fromsnap_guid) {
	sprintf(guidname, "%llu",
	(u_longlong_t)parent_fromsnap_guid);
	if (nvlist_exists(deleted, guidname)) {
	progress = B_TRUE;
	needagain = B_TRUE;
	goto doagain;
	}
	}

	/*
	* Check for rename. If the exact receive path is specified, it
	* does not count as a rename, but we still need to check the
	* datasets beneath it.
	*/
	if ((stream_parent_fromsnap_guid != 0 &&
	parent_fromsnap_guid != 0 &&
	stream_parent_fromsnap_guid != parent_fromsnap_guid) \|\|
	((flags->isprefix \|\| strcmp(tofs, fsname) != 0) &&
	(s1 != NULL) && (s2 != NULL) && strcmp(s1, s2) != 0)) {
	nvlist_t *parent;
	char tryname[ZFS_MAX_DATASET_NAME_LEN];

	parent = fsavl_find(local_avl,
	stream_parent_fromsnap_guid, NULL);
	/*
	* NB: parent might not be found if we used the
	* tosnap for stream_parent_fromsnap_guid,
	* because the parent is a newly-created fs;
	* we'll be able to rename it after we recv the
	* new fs.
	*/
	if (parent != NULL) {
	const char *pname;

	pname = fnvlist_lookup_string(parent, "name");
	(void) snprintf(tryname, sizeof (tryname),
	"%s%s", pname, strrchr(stream_fsname, '/'));
	} else {
	tryname[0] = '\0';
	if (flags->verbose) {
	(void) printf("local fs %s new parent "
	"not found\n", fsname);
	}
	}

	newname[0] = '\0';

	error = recv_rename(hdl, fsname, tryname,
	strlen(tofs)+1, newname, flags);

	if (renamed != NULL && newname[0] != '\0') {
	fnvlist_add_boolean(renamed, newname);
	}

	if (error)
	needagain = B_TRUE;
	else
	progress = B_TRUE;
	}
	}

	doagain:
	fsavl_destroy(local_avl);
	fnvlist_free(local_nv);
	fnvlist_free(deleted);

	if (needagain && progress) {
	/* do another pass to fix up temporary names */
	if (flags->verbose)
	(void) printf("another pass:\n");
	goto again;
	}

	return (needagain \|\| error != 0);
	}

	static int
	zfs_receive_package(libzfs_handle_t hdl, int fd, const char destname,
	recvflags_t flags, dmu_replay_record_t drr, zio_cksum_t *zc,
	char *top_zfs, nvlist_t cmdprops)
	{
	nvlist_t *stream_nv = NULL;
	avl_tree_t *stream_avl = NULL;
	const char *fromsnap = NULL;
	const char *sendsnap = NULL;
	char *cp;
	char tofs[ZFS_MAX_DATASET_NAME_LEN];
	char sendfs[ZFS_MAX_DATASET_NAME_LEN];
	char errbuf[ERRBUFLEN];
	dmu_replay_record_t drre;
	int error;
	boolean_t anyerr = B_FALSE;
	boolean_t softerr = B_FALSE;
	boolean_t recursive, raw;

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot receive"));

	assert(drr->drr_type == DRR_BEGIN);
	assert(drr->drr_u.drr_begin.drr_magic == DMU_BACKUP_MAGIC);
	assert(DMU_GET_STREAM_HDRTYPE(drr->drr_u.drr_begin.drr_versioninfo) ==
	DMU_COMPOUNDSTREAM);

	/*
	* Read in the nvlist from the stream.
	*/
	if (drr->drr_payloadlen != 0) {
	error = recv_read_nvlist(hdl, fd, drr->drr_payloadlen,
	&stream_nv, flags->byteswap, zc);
	if (error) {
	error = zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	goto out;
	}
	}

	recursive = (nvlist_lookup_boolean(stream_nv, "not_recursive") ==
	ENOENT);
	raw = (nvlist_lookup_boolean(stream_nv, "raw") == 0);

	if (recursive && strchr(destname, '@')) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"cannot specify snapshot name for multi-snapshot stream"));
	error = zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	goto out;
	}

	/*
	* Read in the end record and verify checksum.
	*/
	if (0 != (error = recv_read(hdl, fd, &drre, sizeof (drre),
	flags->byteswap, NULL)))
	goto out;
	if (flags->byteswap) {
	drre.drr_type = BSWAP_32(drre.drr_type);
	drre.drr_u.drr_end.drr_checksum.zc_word[0] =
	BSWAP_64(drre.drr_u.drr_end.drr_checksum.zc_word[0]);
	drre.drr_u.drr_end.drr_checksum.zc_word[1] =
	BSWAP_64(drre.drr_u.drr_end.drr_checksum.zc_word[1]);
	drre.drr_u.drr_end.drr_checksum.zc_word[2] =
	BSWAP_64(drre.drr_u.drr_end.drr_checksum.zc_word[2]);
	drre.drr_u.drr_end.drr_checksum.zc_word[3] =
	BSWAP_64(drre.drr_u.drr_end.drr_checksum.zc_word[3]);
	}
	if (drre.drr_type != DRR_END) {
	error = zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	goto out;
	}
	if (!ZIO_CHECKSUM_EQUAL(drre.drr_u.drr_end.drr_checksum, *zc)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"incorrect header checksum"));
	error = zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	goto out;
	}

	(void) nvlist_lookup_string(stream_nv, "fromsnap", &fromsnap);

	if (drr->drr_payloadlen != 0) {
	nvlist_t *stream_fss;

	stream_fss = fnvlist_lookup_nvlist(stream_nv, "fss");
	if ((stream_avl = fsavl_create(stream_fss)) == NULL) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"couldn't allocate avl tree"));
	error = zfs_error(hdl, EZFS_NOMEM, errbuf);
	goto out;
	}

	if (fromsnap != NULL && recursive) {
	nvlist_t *renamed = NULL;
	nvpair_t *pair = NULL;

	(void) strlcpy(tofs, destname, sizeof (tofs));
	if (flags->isprefix) {
	struct drr_begin *drrb = &drr->drr_u.drr_begin;
	int i;

	if (flags->istail) {
	cp = strrchr(drrb->drr_toname, '/');
	if (cp == NULL) {
	(void) strlcat(tofs, "/",
	sizeof (tofs));
	i = 0;
	} else {
	i = (cp - drrb->drr_toname);
	}
	} else {
	i = strcspn(drrb->drr_toname, "/@");
	}
	/* zfs_receive_one() will create_parents() */
	(void) strlcat(tofs, &drrb->drr_toname[i],
	sizeof (tofs));
	*strchr(tofs, '@') = '\0';
	}

	if (!flags->dryrun && !flags->nomount) {
	renamed = fnvlist_alloc();
	}

	softerr = recv_incremental_replication(hdl, tofs, flags,
	stream_nv, stream_avl, renamed);

	/* Unmount renamed filesystems before receiving. */
	while ((pair = nvlist_next_nvpair(renamed,
	pair)) != NULL) {
	zfs_handle_t *zhp;
	prop_changelist_t *clp = NULL;

	zhp = zfs_open(hdl, nvpair_name(pair),
	ZFS_TYPE_FILESYSTEM);
	if (zhp != NULL) {
	clp = changelist_gather(zhp,
	ZFS_PROP_MOUNTPOINT, 0,
	flags->forceunmount ? MS_FORCE : 0);
	zfs_close(zhp);
	if (clp != NULL) {
	softerr \|=
	changelist_prefix(clp);
	changelist_free(clp);
	}
	}
	}

	fnvlist_free(renamed);
	}
	}

	/*
	* Get the fs specified by the first path in the stream (the top level
	* specified by 'zfs send') and pass it to each invocation of
	* zfs_receive_one().
	*/
	(void) strlcpy(sendfs, drr->drr_u.drr_begin.drr_toname,
	sizeof (sendfs));
	if ((cp = strchr(sendfs, '@')) != NULL) {
	*cp = '\0';
	/*
	* Find the "sendsnap", the final snapshot in a replication
	* stream. zfs_receive_one() handles certain errors
	* differently, depending on if the contained stream is the
	* last one or not.
	*/
	sendsnap = (cp + 1);
	}

	/* Finally, receive each contained stream */
	do {
	/*
	* we should figure out if it has a recoverable
	* error, in which case do a recv_skip() and drive on.
	* Note, if we fail due to already having this guid,
	* zfs_receive_one() will take care of it (ie,
	* recv_skip() and return 0).
	*/
	error = zfs_receive_impl(hdl, destname, NULL, flags, fd,
	sendfs, stream_nv, stream_avl, top_zfs, sendsnap, cmdprops);
	if (error == ENODATA) {
	error = 0;
	break;
	}
	anyerr \|= error;
	} while (error == 0);

	if (drr->drr_payloadlen != 0 && recursive && fromsnap != NULL) {
	/*
	* Now that we have the fs's they sent us, try the
	* renames again.
	*/
	softerr = recv_incremental_replication(hdl, tofs, flags,
	stream_nv, stream_avl, NULL);
	}

	if (raw && softerr == 0 && *top_zfs != NULL) {
	softerr = recv_fix_encryption_hierarchy(hdl, *top_zfs,
	stream_nv);
	}

	out:
	fsavl_destroy(stream_avl);
	fnvlist_free(stream_nv);
	if (softerr)
	error = -2;
	if (anyerr)
	error = -1;
	return (error);
	}

	static void
	trunc_prop_errs(int truncated)
	{
	ASSERT(truncated != 0);

	if (truncated == 1)
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"1 more property could not be set\n"));
	else
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"%d more properties could not be set\n"), truncated);
	}

	static int
	recv_skip(libzfs_handle_t *hdl, int fd, boolean_t byteswap)
	{
	dmu_replay_record_t *drr;
	void *buf = zfs_alloc(hdl, SPA_MAXBLOCKSIZE);
	uint64_t payload_size;
	char errbuf[ERRBUFLEN];

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot receive"));

	/* XXX would be great to use lseek if possible... */
	drr = buf;

	while (recv_read(hdl, fd, drr, sizeof (dmu_replay_record_t),
	byteswap, NULL) == 0) {
	if (byteswap)
	drr->drr_type = BSWAP_32(drr->drr_type);

	switch (drr->drr_type) {
	case DRR_BEGIN:
	if (drr->drr_payloadlen != 0) {
	(void) recv_read(hdl, fd, buf,
	drr->drr_payloadlen, B_FALSE, NULL);
	}
	break;

	case DRR_END:
	free(buf);
	return (0);

	case DRR_OBJECT:
	if (byteswap) {
	drr->drr_u.drr_object.drr_bonuslen =
	BSWAP_32(drr->drr_u.drr_object.
	drr_bonuslen);
	drr->drr_u.drr_object.drr_raw_bonuslen =
	BSWAP_32(drr->drr_u.drr_object.
	drr_raw_bonuslen);
	}

	payload_size =
	DRR_OBJECT_PAYLOAD_SIZE(&drr->drr_u.drr_object);
	(void) recv_read(hdl, fd, buf, payload_size,
	B_FALSE, NULL);
	break;

	case DRR_WRITE:
	if (byteswap) {
	drr->drr_u.drr_write.drr_logical_size =
	BSWAP_64(
	drr->drr_u.drr_write.drr_logical_size);
	drr->drr_u.drr_write.drr_compressed_size =
	BSWAP_64(
	drr->drr_u.drr_write.drr_compressed_size);
	}
	payload_size =
	DRR_WRITE_PAYLOAD_SIZE(&drr->drr_u.drr_write);
	assert(payload_size <= SPA_MAXBLOCKSIZE);
	(void) recv_read(hdl, fd, buf,
	payload_size, B_FALSE, NULL);
	break;
	case DRR_SPILL:
	if (byteswap) {
	drr->drr_u.drr_spill.drr_length =
	BSWAP_64(drr->drr_u.drr_spill.drr_length);
	drr->drr_u.drr_spill.drr_compressed_size =
	BSWAP_64(drr->drr_u.drr_spill.
	drr_compressed_size);
	}

	payload_size =
	DRR_SPILL_PAYLOAD_SIZE(&drr->drr_u.drr_spill);
	(void) recv_read(hdl, fd, buf, payload_size,
	B_FALSE, NULL);
	break;
	case DRR_WRITE_EMBEDDED:
	if (byteswap) {
	drr->drr_u.drr_write_embedded.drr_psize =
	BSWAP_32(drr->drr_u.drr_write_embedded.
	drr_psize);
	}
	(void) recv_read(hdl, fd, buf,
	P2ROUNDUP(drr->drr_u.drr_write_embedded.drr_psize,
	8), B_FALSE, NULL);
	break;
	case DRR_OBJECT_RANGE:
	case DRR_WRITE_BYREF:
	case DRR_FREEOBJECTS:
	case DRR_FREE:
	break;

	default:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"invalid record type"));
	free(buf);
	return (zfs_error(hdl, EZFS_BADSTREAM, errbuf));
	}
	}

	free(buf);
	return (-1);
	}

	static void
	recv_ecksum_set_aux(libzfs_handle_t hdl, const char target_snap,
	boolean_t resumable, boolean_t checksum)
	{
	char target_fs[ZFS_MAX_DATASET_NAME_LEN];

	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, (checksum ?
	"checksum mismatch" : "incomplete stream")));

	if (!resumable)
	return;
	(void) strlcpy(target_fs, target_snap, sizeof (target_fs));
	*strchr(target_fs, '@') = '\0';
	zfs_handle_t *zhp = zfs_open(hdl, target_fs,
	ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME);
	if (zhp == NULL)
	return;

	char token_buf[ZFS_MAXPROPLEN];
	int error = zfs_prop_get(zhp, ZFS_PROP_RECEIVE_RESUME_TOKEN,
	token_buf, sizeof (token_buf),
	NULL, NULL, 0, B_TRUE);
	if (error == 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"checksum mismatch or incomplete stream.\n"
	"Partially received snapshot is saved.\n"
	"A resuming stream can be generated on the sending "
	"system by running:\n"
	" zfs send -t %s"),
	token_buf);
	}
	zfs_close(zhp);
	}

	/*
	* Prepare a new nvlist of properties that are to override (-o) or be excluded
	* (-x) from the received dataset
	* recvprops: received properties from the send stream
	* cmdprops: raw input properties from command line
	* origprops: properties, both locally-set and received, currently set on the
	* target dataset if it exists, NULL otherwise.
	* oxprops: valid output override (-o) and excluded (-x) properties
	*/
	static int
	zfs_setup_cmdline_props(libzfs_handle_t *hdl, zfs_type_t type,
	char *fsname, boolean_t zoned, boolean_t recursive, boolean_t newfs,
	boolean_t raw, boolean_t toplevel, nvlist_t recvprops, nvlist_t cmdprops,
	nvlist_t origprops, nvlist_t oxprops, uint8_t *wkeydata_out,
	uint_t wkeylen_out, const char errbuf)
	{
	nvpair_t *nvp;
	nvlist_t oprops, voprops;
	zfs_handle_t *zhp = NULL;
	zpool_handle_t *zpool_hdl = NULL;
	char *cp;
	int ret = 0;
	char namebuf[ZFS_MAX_DATASET_NAME_LEN];

	if (nvlist_empty(cmdprops))
	return (0); /* No properties to override or exclude */

	*oxprops = fnvlist_alloc();
	oprops = fnvlist_alloc();

	strlcpy(namebuf, fsname, ZFS_MAX_DATASET_NAME_LEN);

	/*
	* Get our dataset handle. The target dataset may not exist yet.
	*/
	if (zfs_dataset_exists(hdl, namebuf, ZFS_TYPE_DATASET)) {
	zhp = zfs_open(hdl, namebuf, ZFS_TYPE_DATASET);
	if (zhp == NULL) {
	ret = -1;
	goto error;
	}
	}

	/* open the zpool handle */
	cp = strchr(namebuf, '/');
	if (cp != NULL)
	*cp = '\0';
	zpool_hdl = zpool_open(hdl, namebuf);
	if (zpool_hdl == NULL) {
	ret = -1;
	goto error;
	}

	/* restore namebuf to match fsname for later use */
	if (cp != NULL)
	*cp = '/';

	/*
	* first iteration: process excluded (-x) properties now and gather
	* added (-o) properties to be later processed by zfs_valid_proplist()
	*/
	nvp = NULL;
	while ((nvp = nvlist_next_nvpair(cmdprops, nvp)) != NULL) {
	const char *name = nvpair_name(nvp);
	zfs_prop_t prop = zfs_name_to_prop(name);

	/*
	* It turns out, if we don't normalize "aliased" names
	* e.g. compress= against the "real" names (e.g. compression)
	* here, then setting/excluding them does not work as
	* intended.
	*
	* But since user-defined properties wouldn't have a valid
	* mapping here, we do this conditional dance.
	*/
	const char *newname = name;
	if (prop >= ZFS_PROP_TYPE)
	newname = zfs_prop_to_name(prop);

	/* "origin" is processed separately, don't handle it here */
	if (prop == ZFS_PROP_ORIGIN)
	continue;

	/* raw streams can't override encryption properties */
	if ((zfs_prop_encryption_key_param(prop) \|\|
	prop == ZFS_PROP_ENCRYPTION) && raw) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"encryption property '%s' cannot "
	"be set or excluded for raw streams."), name);
	ret = zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}

	/*
	* For plain replicated send, we can ignore encryption
	* properties other than first stream
	*/
	if ((zfs_prop_encryption_key_param(prop) \|\| prop ==
	ZFS_PROP_ENCRYPTION) && !newfs && recursive && !raw) {
	continue;
	}

	/* incremental streams can only exclude encryption properties */
	if ((zfs_prop_encryption_key_param(prop) \|\|
	prop == ZFS_PROP_ENCRYPTION) && !newfs &&
	nvpair_type(nvp) != DATA_TYPE_BOOLEAN) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"encryption property '%s' cannot "
	"be set for incremental streams."), name);
	ret = zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}

	switch (nvpair_type(nvp)) {
	case DATA_TYPE_BOOLEAN: /* -x property */
	/*
	* DATA_TYPE_BOOLEAN is the way we're asked to "exclude"
	* a property: this is done by forcing an explicit
	* inherit on the destination so the effective value is
	* not the one we received from the send stream.
	*/
	if (!zfs_prop_valid_for_type(prop, type, B_FALSE) &&
	!zfs_prop_user(name)) {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN,
	"Warning: %s: property '%s' does not "
	"apply to datasets of this type\n"),
	fsname, name);
	continue;
	}
	/*
	* We do this only if the property is not already
	* locally-set, in which case its value will take
	* priority over the received anyway.
	*/
	if (nvlist_exists(origprops, newname)) {
	nvlist_t *attrs;
	const char *source = NULL;

	attrs = fnvlist_lookup_nvlist(origprops,
	newname);
	if (nvlist_lookup_string(attrs,
	ZPROP_SOURCE, &source) == 0 &&
	strcmp(source, ZPROP_SOURCE_VAL_RECVD) != 0)
	continue;
	}
	/*
	* We can't force an explicit inherit on non-inheritable
	* properties: if we're asked to exclude this kind of
	* values we remove them from "recvprops" input nvlist.
	*/
	if (!zfs_prop_user(name) && /* can be inherited too */
	!zfs_prop_inheritable(prop) &&
	nvlist_exists(recvprops, newname))
	fnvlist_remove(recvprops, newname);
	else
	fnvlist_add_boolean(*oxprops, newname);
	break;
	case DATA_TYPE_STRING: /* -o property=value */
	/*
	* we're trying to override a property that does not
	* make sense for this type of dataset, but we don't
	* want to fail if the receive is recursive: this comes
	* in handy when the send stream contains, for
	* instance, a child ZVOL and we're trying to receive
	* it with "-o atime=on"
	*/
	if (!zfs_prop_valid_for_type(prop, type, B_FALSE) &&
	!zfs_prop_user(name)) {
	if (recursive)
	continue;
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"property '%s' does not apply to datasets "
	"of this type"), name);
	ret = zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}
	fnvlist_add_string(oprops, newname,
	fnvpair_value_string(nvp));
	break;
	default:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"property '%s' must be a string or boolean"), name);
	ret = zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}
	}

	if (toplevel) {
	/* convert override strings properties to native */
	if ((voprops = zfs_valid_proplist(hdl, ZFS_TYPE_DATASET,
	oprops, zoned, zhp, zpool_hdl, B_FALSE, errbuf)) == NULL) {
	ret = zfs_error(hdl, EZFS_BADPROP, errbuf);
	goto error;
	}

	/*
	* zfs_crypto_create() requires the parent name. Get it
	* by truncating the fsname copy stored in namebuf.
	*/
	cp = strrchr(namebuf, '/');
	if (cp != NULL)
	*cp = '\0';

	if (!raw && !(!newfs && recursive) &&
	zfs_crypto_create(hdl, namebuf, voprops, NULL,
	B_FALSE, wkeydata_out, wkeylen_out) != 0) {
	fnvlist_free(voprops);
	ret = zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf);
	goto error;
	}

	/* second pass: process "-o" properties */
	fnvlist_merge(*oxprops, voprops);
	fnvlist_free(voprops);
	} else {
	/* override props on child dataset are inherited */
	nvp = NULL;
	while ((nvp = nvlist_next_nvpair(oprops, nvp)) != NULL) {
	const char *name = nvpair_name(nvp);
	fnvlist_add_boolean(*oxprops, name);
	}
	}

	error:
	if (zhp != NULL)
	zfs_close(zhp);
	if (zpool_hdl != NULL)
	zpool_close(zpool_hdl);
	fnvlist_free(oprops);
	return (ret);
	}

	/*
	* Restores a backup of tosnap from the file descriptor specified by infd.
	*/
	static int
	zfs_receive_one(libzfs_handle_t hdl, int infd, const char tosnap,
	const char originsnap, recvflags_t flags, dmu_replay_record_t *drr,
	dmu_replay_record_t drr_noswap, const char sendfs, nvlist_t *stream_nv,
	avl_tree_t stream_avl, char *top_zfs,
	const char finalsnap, nvlist_t cmdprops)
	{
	struct timespec begin_time;
	int ioctl_err, ioctl_errno, err;
	char *cp;
	struct drr_begin *drrb = &drr->drr_u.drr_begin;
	char errbuf[ERRBUFLEN];
	const char *chopprefix;
	boolean_t newfs = B_FALSE;
	boolean_t stream_wantsnewfs, stream_resumingnewfs;
	boolean_t newprops = B_FALSE;
	uint64_t read_bytes = 0;
	uint64_t errflags = 0;
	uint64_t parent_snapguid = 0;
	prop_changelist_t *clp = NULL;
	nvlist_t *snapprops_nvlist = NULL;
	nvlist_t *snapholds_nvlist = NULL;
	zprop_errflags_t prop_errflags;
	nvlist_t *prop_errors = NULL;
	boolean_t recursive;
	const char *snapname = NULL;
	char destsnap[MAXPATHLEN * 2];
	char origin[MAXNAMELEN] = {0};
	char name[MAXPATHLEN];
	char tmp_keylocation[MAXNAMELEN] = {0};
	nvlist_t rcvprops = NULL; / props received from the send stream */
	nvlist_t oxprops = NULL; / override (-o) and exclude (-x) props */
	nvlist_t origprops = NULL; / original props (if destination exists) */
	zfs_type_t type = ZFS_TYPE_INVALID;
	boolean_t toplevel = B_FALSE;
	boolean_t zoned = B_FALSE;
	boolean_t hastoken = B_FALSE;
	boolean_t redacted;
	uint8_t *wkeydata = NULL;
	uint_t wkeylen = 0;

	#ifndef CLOCK_MONOTONIC_RAW
	#define CLOCK_MONOTONIC_RAW CLOCK_MONOTONIC
	#endif
	clock_gettime(CLOCK_MONOTONIC_RAW, &begin_time);

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot receive"));

	recursive = (nvlist_lookup_boolean(stream_nv, "not_recursive") ==
	ENOENT);

	/* Did the user request holds be skipped via zfs recv -k? */
	boolean_t holds = flags->holds && !flags->skipholds;

	if (stream_avl != NULL) {
	const char *keylocation = NULL;
	nvlist_t *lookup = NULL;
	nvlist_t *fs = fsavl_find(stream_avl, drrb->drr_toguid,
	&snapname);

	(void) nvlist_lookup_uint64(fs, "parentfromsnap",
	&parent_snapguid);
	err = nvlist_lookup_nvlist(fs, "props", &rcvprops);
	if (err) {
	rcvprops = fnvlist_alloc();
	newprops = B_TRUE;
	}

	/*
	* The keylocation property may only be set on encryption roots,
	* but this dataset might not become an encryption root until
	* recv_fix_encryption_hierarchy() is called. That function
	* will fixup the keylocation anyway, so we temporarily unset
	* the keylocation for now to avoid any errors from the receive
	* ioctl.
	*/
	err = nvlist_lookup_string(rcvprops,
	zfs_prop_to_name(ZFS_PROP_KEYLOCATION), &keylocation);
	if (err == 0) {
	strlcpy(tmp_keylocation, keylocation, MAXNAMELEN);
	(void) nvlist_remove_all(rcvprops,
	zfs_prop_to_name(ZFS_PROP_KEYLOCATION));
	}

	if (flags->canmountoff) {
	fnvlist_add_uint64(rcvprops,
	zfs_prop_to_name(ZFS_PROP_CANMOUNT), 0);
	} else if (newprops) { /* nothing in rcvprops, eliminate it */
	fnvlist_free(rcvprops);
	rcvprops = NULL;
	newprops = B_FALSE;
	}
	if (0 == nvlist_lookup_nvlist(fs, "snapprops", &lookup)) {
	snapprops_nvlist = fnvlist_lookup_nvlist(lookup,
	snapname);
	}
	if (holds) {
	if (0 == nvlist_lookup_nvlist(fs, "snapholds",
	&lookup)) {
	snapholds_nvlist = fnvlist_lookup_nvlist(
	lookup, snapname);
	}
	}
	}

	cp = NULL;

	/*
	* Determine how much of the snapshot name stored in the stream
	* we are going to tack on to the name they specified on the
	* command line, and how much we are going to chop off.
	*
	* If they specified a snapshot, chop the entire name stored in
	* the stream.
	*/
	if (flags->istail) {
	/*
	* A filesystem was specified with -e. We want to tack on only
	* the tail of the sent snapshot path.
	*/
	if (strchr(tosnap, '@')) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid "
	"argument - snapshot not allowed with -e"));
	err = zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
	goto out;
	}

	chopprefix = strrchr(sendfs, '/');

	if (chopprefix == NULL) {
	/*
	* The tail is the poolname, so we need to
	* prepend a path separator.
	*/
	int len = strlen(drrb->drr_toname);
	cp = umem_alloc(len + 2, UMEM_NOFAIL);
	cp[0] = '/';
	(void) strcpy(&cp[1], drrb->drr_toname);
	chopprefix = cp;
	} else {
	chopprefix = drrb->drr_toname + (chopprefix - sendfs);
	}
	} else if (flags->isprefix) {
	/*
	* A filesystem was specified with -d. We want to tack on
	* everything but the first element of the sent snapshot path
	* (all but the pool name).
	*/
	if (strchr(tosnap, '@')) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid "
	"argument - snapshot not allowed with -d"));
	err = zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
	goto out;
	}

	chopprefix = strchr(drrb->drr_toname, '/');
	if (chopprefix == NULL)
	chopprefix = strchr(drrb->drr_toname, '@');
	} else if (strchr(tosnap, '@') == NULL) {
	/*
	* If a filesystem was specified without -d or -e, we want to
	* tack on everything after the fs specified by 'zfs send'.
	*/
	chopprefix = drrb->drr_toname + strlen(sendfs);
	} else {
	/* A snapshot was specified as an exact path (no -d or -e). */
	if (recursive) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"cannot specify snapshot name for multi-snapshot "
	"stream"));
	err = zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	goto out;
	}
	chopprefix = drrb->drr_toname + strlen(drrb->drr_toname);
	}

	ASSERT(strstr(drrb->drr_toname, sendfs) == drrb->drr_toname);
	ASSERT(chopprefix > drrb->drr_toname \|\| strchr(sendfs, '/') == NULL);
	ASSERT(chopprefix <= drrb->drr_toname + strlen(drrb->drr_toname) \|\|
	strchr(sendfs, '/') == NULL);
	ASSERT(chopprefix[0] == '/' \|\| chopprefix[0] == '@' \|\|
	chopprefix[0] == '\0');

	/*
	* Determine name of destination snapshot.
	*/
	(void) strlcpy(destsnap, tosnap, sizeof (destsnap));
	(void) strlcat(destsnap, chopprefix, sizeof (destsnap));
	if (cp != NULL)
	umem_free(cp, strlen(cp) + 1);
	if (!zfs_name_valid(destsnap, ZFS_TYPE_SNAPSHOT)) {
	err = zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
	goto out;
	}

	/*
	* Determine the name of the origin snapshot.
	*/
	if (originsnap) {
	(void) strlcpy(origin, originsnap, sizeof (origin));
	if (flags->verbose)
	(void) printf("using provided clone origin %s\n",
	origin);
	} else if (drrb->drr_flags & DRR_FLAG_CLONE) {
	if (guid_to_name(hdl, destsnap,
	drrb->drr_fromguid, B_FALSE, origin) != 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"local origin for clone %s does not exist"),
	destsnap);
	err = zfs_error(hdl, EZFS_NOENT, errbuf);
	goto out;
	}
	if (flags->verbose)
	(void) printf("found clone origin %s\n", origin);
	}

	if ((DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo) &
	DMU_BACKUP_FEATURE_DEDUP)) {
	(void) fprintf(stderr,
	gettext("ERROR: \"zfs receive\" no longer supports "
	"deduplicated send streams. Use\n"
	"the \"zstream redup\" command to convert this stream "
	"to a regular,\n"
	"non-deduplicated stream.\n"));
	err = zfs_error(hdl, EZFS_NOTSUP, errbuf);
	goto out;
	}

	boolean_t resuming = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo) &
	DMU_BACKUP_FEATURE_RESUMING;
	boolean_t raw = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo) &
	DMU_BACKUP_FEATURE_RAW;
	boolean_t embedded = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo) &
	DMU_BACKUP_FEATURE_EMBED_DATA;
	stream_wantsnewfs = (drrb->drr_fromguid == 0 \|\|
	(drrb->drr_flags & DRR_FLAG_CLONE) \|\| originsnap) && !resuming;
	stream_resumingnewfs = (drrb->drr_fromguid == 0 \|\|
	(drrb->drr_flags & DRR_FLAG_CLONE) \|\| originsnap) && resuming;

	if (stream_wantsnewfs) {
	/*
	* if the parent fs does not exist, look for it based on
	* the parent snap GUID
	*/
	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot receive new filesystem stream"));

	(void) strlcpy(name, destsnap, sizeof (name));
	cp = strrchr(name, '/');
	if (cp)
	*cp = '\0';
	if (cp &&
	!zfs_dataset_exists(hdl, name, ZFS_TYPE_DATASET)) {
	char suffix[ZFS_MAX_DATASET_NAME_LEN];
	(void) strlcpy(suffix, strrchr(destsnap, '/'),
	sizeof (suffix));
	if (guid_to_name(hdl, name, parent_snapguid,
	B_FALSE, destsnap) == 0) {
	*strchr(destsnap, '@') = '\0';
	(void) strlcat(destsnap, suffix,
	sizeof (destsnap));
	}
	}
	} else {
	/*
	* If the fs does not exist, look for it based on the
	* fromsnap GUID.
	*/
	if (resuming) {
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN,
	"cannot receive resume stream"));
	} else {
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN,
	"cannot receive incremental stream"));
	}

	(void) strlcpy(name, destsnap, sizeof (name));
	*strchr(name, '@') = '\0';

	/*
	* If the exact receive path was specified and this is the
	* topmost path in the stream, then if the fs does not exist we
	* should look no further.
	*/
	if ((flags->isprefix \|\| (*(chopprefix = drrb->drr_toname +
	strlen(sendfs)) != '\0' && *chopprefix != '@')) &&
	!zfs_dataset_exists(hdl, name, ZFS_TYPE_DATASET)) {
	char snap[ZFS_MAX_DATASET_NAME_LEN];
	(void) strlcpy(snap, strchr(destsnap, '@'),
	sizeof (snap));
	if (guid_to_name(hdl, name, drrb->drr_fromguid,
	B_FALSE, destsnap) == 0) {
	*strchr(destsnap, '@') = '\0';
	(void) strlcat(destsnap, snap,
	sizeof (destsnap));
	}
	}
	}

	(void) strlcpy(name, destsnap, sizeof (name));
	*strchr(name, '@') = '\0';

	redacted = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo) &
	DMU_BACKUP_FEATURE_REDACTED;

	if (flags->heal) {
	if (flags->isprefix \|\| flags->istail \|\| flags->force \|\|
	flags->canmountoff \|\| flags->resumable \|\| flags->nomount \|\|
	flags->skipholds) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"corrective recv can not be used when combined with"
	" this flag"));
	err = zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
	goto out;
	}
	uint64_t guid =
	get_snap_guid(hdl, name, strchr(destsnap, '@') + 1);
	if (guid == 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"corrective recv must specify an existing snapshot"
	" to heal"));
	err = zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
	goto out;
	} else if (guid != drrb->drr_toguid) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"local snapshot doesn't match the snapshot"
	" in the provided stream"));
	err = zfs_error(hdl, EZFS_WRONG_PARENT, errbuf);
	goto out;
	}
	} else if (zfs_dataset_exists(hdl, name, ZFS_TYPE_DATASET)) {
	zfs_cmd_t zc = {"\0"};
	zfs_handle_t *zhp = NULL;
	boolean_t encrypted;

	(void) strcpy(zc.zc_name, name);

	/*
	* Destination fs exists. It must be one of these cases:
	* - an incremental send stream
	* - the stream specifies a new fs (full stream or clone)
	* and they want us to blow away the existing fs (and
	* have therefore specified -F and removed any snapshots)
	* - we are resuming a failed receive.
	*/
	if (stream_wantsnewfs) {
	boolean_t is_volume = drrb->drr_type == DMU_OST_ZVOL;
	if (!flags->force) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"destination '%s' exists\n"
	"must specify -F to overwrite it"), name);
	err = zfs_error(hdl, EZFS_EXISTS, errbuf);
	goto out;
	}
	if (zfs_ioctl(hdl, ZFS_IOC_SNAPSHOT_LIST_NEXT,
	&zc) == 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"destination has snapshots (eg. %s)\n"
	"must destroy them to overwrite it"),
	zc.zc_name);
	err = zfs_error(hdl, EZFS_EXISTS, errbuf);
	goto out;
	}
	if (is_volume && strrchr(name, '/') == NULL) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"destination %s is the root dataset\n"
	"cannot overwrite with a ZVOL"),
	name);
	err = zfs_error(hdl, EZFS_EXISTS, errbuf);
	goto out;
	}
	if (is_volume &&
	zfs_ioctl(hdl, ZFS_IOC_DATASET_LIST_NEXT,
	&zc) == 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"destination has children (eg. %s)\n"
	"cannot overwrite with a ZVOL"),
	zc.zc_name);
	err = zfs_error(hdl, EZFS_WRONG_PARENT, errbuf);
	goto out;
	}
	}

	if ((zhp = zfs_open(hdl, name,
	ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME)) == NULL) {
	err = -1;
	goto out;
	}

	/*
	* When receiving full/newfs on existing dataset, then it
	* should be done with "-F" flag. Its enforced for initial
	* receive in previous checks in this function.
	* Similarly, on resuming full/newfs recv on existing dataset,
	* it should be done with "-F" flag.
	*
	* When dataset doesn't exist, then full/newfs recv is done on
	* newly created dataset and it's marked INCONSISTENT. But
	* When receiving on existing dataset, recv is first done on
	* %recv and its marked INCONSISTENT. Existing dataset is not
	* marked INCONSISTENT.
	* Resume of full/newfs receive with dataset not INCONSISTENT
	* indicates that its resuming newfs on existing dataset. So,
	* enforce "-F" flag in this case.
	*/
	if (stream_resumingnewfs &&
	!zfs_prop_get_int(zhp, ZFS_PROP_INCONSISTENT) &&
	!flags->force) {
	zfs_close(zhp);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"Resuming recv on existing destination '%s'\n"
	"must specify -F to overwrite it"), name);
	err = zfs_error(hdl, EZFS_RESUME_EXISTS, errbuf);
	goto out;
	}

	if (stream_wantsnewfs &&
	zhp->zfs_dmustats.dds_origin[0]) {
	zfs_close(zhp);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"destination '%s' is a clone\n"
	"must destroy it to overwrite it"), name);
	err = zfs_error(hdl, EZFS_EXISTS, errbuf);
	goto out;
	}

	/*
	* Raw sends can not be performed as an incremental on top
	* of existing unencrypted datasets. zfs recv -F can't be
	* used to blow away an existing encrypted filesystem. This
	* is because it would require the dsl dir to point to the
	* new key (or lack of a key) and the old key at the same
	* time. The -F flag may still be used for deleting
	* intermediate snapshots that would otherwise prevent the
	* receive from working.
	*/
	encrypted = zfs_prop_get_int(zhp, ZFS_PROP_ENCRYPTION) !=
	ZIO_CRYPT_OFF;
	if (!stream_wantsnewfs && !encrypted && raw) {
	zfs_close(zhp);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"cannot perform raw receive on top of "
	"existing unencrypted dataset"));
	err = zfs_error(hdl, EZFS_BADRESTORE, errbuf);
	goto out;
	}

	if (stream_wantsnewfs && flags->force &&
	((raw && !encrypted) \|\| encrypted)) {
	zfs_close(zhp);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"zfs receive -F cannot be used to destroy an "
	"encrypted filesystem or overwrite an "
	"unencrypted one with an encrypted one"));
	err = zfs_error(hdl, EZFS_BADRESTORE, errbuf);
	goto out;
	}

	if (!flags->dryrun && zhp->zfs_type == ZFS_TYPE_FILESYSTEM &&
	(stream_wantsnewfs \|\| stream_resumingnewfs)) {
	/* We can't do online recv in this case */
	clp = changelist_gather(zhp, ZFS_PROP_NAME, 0,
	flags->forceunmount ? MS_FORCE : 0);
	if (clp == NULL) {
	zfs_close(zhp);
	err = -1;
	goto out;
	}
	if (changelist_prefix(clp) != 0) {
	changelist_free(clp);
	zfs_close(zhp);
	err = -1;
	goto out;
	}
	}

	/*
	* If we are resuming a newfs, set newfs here so that we will
	* mount it if the recv succeeds this time. We can tell
	* that it was a newfs on the first recv because the fs
	* itself will be inconsistent (if the fs existed when we
	* did the first recv, we would have received it into
	* .../%recv).
	*/
	if (resuming && zfs_prop_get_int(zhp, ZFS_PROP_INCONSISTENT))
	newfs = B_TRUE;

	/* we want to know if we're zoned when validating -o\|-x props */
	zoned = zfs_prop_get_int(zhp, ZFS_PROP_ZONED);

	/* may need this info later, get it now we have zhp around */
	if (zfs_prop_get(zhp, ZFS_PROP_RECEIVE_RESUME_TOKEN, NULL, 0,
	NULL, NULL, 0, B_TRUE) == 0)
	hastoken = B_TRUE;

	/* gather existing properties on destination */
	origprops = fnvlist_alloc();
	fnvlist_merge(origprops, zhp->zfs_props);
	fnvlist_merge(origprops, zhp->zfs_user_props);

	zfs_close(zhp);
	} else {
	zfs_handle_t *zhp;

	/*
	* Destination filesystem does not exist. Therefore we better
	* be creating a new filesystem (either from a full backup, or
	* a clone). It would therefore be invalid if the user
	* specified only the pool name (i.e. if the destination name
	* contained no slash character).
	*/
	cp = strrchr(name, '/');

	if (!stream_wantsnewfs \|\| cp == NULL) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"destination '%s' does not exist"), name);
	err = zfs_error(hdl, EZFS_NOENT, errbuf);
	goto out;
	}

	/*
	* Trim off the final dataset component so we perform the
	* recvbackup ioctl to the filesystems's parent.
	*/
	*cp = '\0';

	if (flags->isprefix && !flags->istail && !flags->dryrun &&
	create_parents(hdl, destsnap, strlen(tosnap)) != 0) {
	err = zfs_error(hdl, EZFS_BADRESTORE, errbuf);
	goto out;
	}

	/* validate parent */
	zhp = zfs_open(hdl, name, ZFS_TYPE_DATASET);
	if (zhp == NULL) {
	err = zfs_error(hdl, EZFS_BADRESTORE, errbuf);
	goto out;
	}
	if (zfs_get_type(zhp) != ZFS_TYPE_FILESYSTEM) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"parent '%s' is not a filesystem"), name);
	err = zfs_error(hdl, EZFS_WRONG_PARENT, errbuf);
	zfs_close(zhp);
	goto out;
	}

	zfs_close(zhp);

	newfs = B_TRUE;
	*cp = '/';
	}

	if (flags->verbose) {
	(void) printf("%s %s%s stream of %s into %s\n",
	flags->dryrun ? "would receive" : "receiving",
	flags->heal ? " corrective" : "",
	drrb->drr_fromguid ? "incremental" : "full",
	drrb->drr_toname, destsnap);
	(void) fflush(stdout);
	}

	/*
	* If this is the top-level dataset, record it so we can use it
	* for recursive operations later.
	*/
	if (top_zfs != NULL &&
	(top_zfs == NULL \|\| strcmp(top_zfs, name) == 0)) {
	toplevel = B_TRUE;
	if (*top_zfs == NULL)
	*top_zfs = zfs_strdup(hdl, name);
	}

	if (drrb->drr_type == DMU_OST_ZVOL) {
	type = ZFS_TYPE_VOLUME;
	} else if (drrb->drr_type == DMU_OST_ZFS) {
	type = ZFS_TYPE_FILESYSTEM;
	} else {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"invalid record type: 0x%d"), drrb->drr_type);
	err = zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	goto out;
	}
	if ((err = zfs_setup_cmdline_props(hdl, type, name, zoned, recursive,
	stream_wantsnewfs, raw, toplevel, rcvprops, cmdprops, origprops,
	&oxprops, &wkeydata, &wkeylen, errbuf)) != 0)
	goto out;

	/*
	* When sending with properties (zfs send -p), the encryption property
	* is not included because it is a SETONCE property and therefore
	* treated as read only. However, we are always able to determine its
	* value because raw sends will include it in the DRR_BDEGIN payload
	* and non-raw sends with properties are not allowed for encrypted
	* datasets. Therefore, if this is a non-raw properties stream, we can
	* infer that the value should be ZIO_CRYPT_OFF and manually add that
	* to the received properties.
	*/
	if (stream_wantsnewfs && !raw && rcvprops != NULL &&
	!nvlist_exists(cmdprops, zfs_prop_to_name(ZFS_PROP_ENCRYPTION))) {
	if (oxprops == NULL)
	oxprops = fnvlist_alloc();
	fnvlist_add_uint64(oxprops,
	zfs_prop_to_name(ZFS_PROP_ENCRYPTION), ZIO_CRYPT_OFF);
	}

	if (flags->dryrun) {
	void *buf = zfs_alloc(hdl, SPA_MAXBLOCKSIZE);

	/*
	* We have read the DRR_BEGIN record, but we have
	* not yet read the payload. For non-dryrun sends
	* this will be done by the kernel, so we must
	* emulate that here, before attempting to read
	* more records.
	*/
	err = recv_read(hdl, infd, buf, drr->drr_payloadlen,
	flags->byteswap, NULL);
	free(buf);
	if (err != 0)
	goto out;

	err = recv_skip(hdl, infd, flags->byteswap);
	goto out;
	}

	if (flags->heal) {
	err = ioctl_err = lzc_receive_with_heal(destsnap, rcvprops,
	oxprops, wkeydata, wkeylen, origin, flags->force,
	flags->heal, flags->resumable, raw, infd, drr_noswap, -1,
	&read_bytes, &errflags, NULL, &prop_errors);
	} else {
	err = ioctl_err = lzc_receive_with_cmdprops(destsnap, rcvprops,
	oxprops, wkeydata, wkeylen, origin, flags->force,
	flags->resumable, raw, infd, drr_noswap, -1, &read_bytes,
	&errflags, NULL, &prop_errors);
	}
	ioctl_errno = ioctl_err;
	prop_errflags = errflags;

	if (err == 0) {
	nvpair_t *prop_err = NULL;

	while ((prop_err = nvlist_next_nvpair(prop_errors,
	prop_err)) != NULL) {
	char tbuf[1024];
	zfs_prop_t prop;
	int intval;

	prop = zfs_name_to_prop(nvpair_name(prop_err));
	(void) nvpair_value_int32(prop_err, &intval);
	if (strcmp(nvpair_name(prop_err),
	ZPROP_N_MORE_ERRORS) == 0) {
	trunc_prop_errs(intval);
	break;
	} else if (snapname == NULL \|\| finalsnap == NULL \|\|
	strcmp(finalsnap, snapname) == 0 \|\|
	strcmp(nvpair_name(prop_err),
	zfs_prop_to_name(ZFS_PROP_REFQUOTA)) != 0) {
	/*
	* Skip the special case of, for example,
	* "refquota", errors on intermediate
	* snapshots leading up to a final one.
	* That's why we have all of the checks above.
	*
	* See zfs_ioctl.c's extract_delay_props() for
	* a list of props which can fail on
	* intermediate snapshots, but shouldn't
	* affect the overall receive.
	*/
	(void) snprintf(tbuf, sizeof (tbuf),
	dgettext(TEXT_DOMAIN,
	"cannot receive %s property on %s"),
	nvpair_name(prop_err), name);
	zfs_setprop_error(hdl, prop, intval, tbuf);
	}
	}
	}

	if (err == 0 && snapprops_nvlist) {
	zfs_cmd_t zc = {"\0"};

	(void) strlcpy(zc.zc_name, destsnap, sizeof (zc.zc_name));
	zc.zc_cookie = B_TRUE; /* received */
	zcmd_write_src_nvlist(hdl, &zc, snapprops_nvlist);
	(void) zfs_ioctl(hdl, ZFS_IOC_SET_PROP, &zc);
	zcmd_free_nvlists(&zc);
	}
	if (err == 0 && snapholds_nvlist) {
	nvpair_t *pair;
	nvlist_t holds, errors = NULL;
	int cleanup_fd = -1;

	VERIFY(0 == nvlist_alloc(&holds, 0, KM_SLEEP));
	for (pair = nvlist_next_nvpair(snapholds_nvlist, NULL);
	pair != NULL;
	pair = nvlist_next_nvpair(snapholds_nvlist, pair)) {
	fnvlist_add_string(holds, destsnap, nvpair_name(pair));
	}
	(void) lzc_hold(holds, cleanup_fd, &errors);
	fnvlist_free(snapholds_nvlist);
	fnvlist_free(holds);
	}

	if (err && (ioctl_errno == ENOENT \|\| ioctl_errno == EEXIST)) {
	/*
	* It may be that this snapshot already exists,
	* in which case we want to consume & ignore it
	* rather than failing.
	*/
	avl_tree_t *local_avl;
	nvlist_t local_nv, fs;
	cp = strchr(destsnap, '@');

	/*
	* XXX Do this faster by just iterating over snaps in
	* this fs. Also if zc_value does not exist, we will
	* get a strange "does not exist" error message.
	*/
	*cp = '\0';
	if (gather_nvlist(hdl, destsnap, NULL, NULL, B_FALSE, B_TRUE,
	B_FALSE, B_FALSE, B_FALSE, B_FALSE, B_FALSE, B_FALSE,
	B_TRUE, &local_nv, &local_avl) == 0) {
	*cp = '@';
	fs = fsavl_find(local_avl, drrb->drr_toguid, NULL);
	fsavl_destroy(local_avl);
	fnvlist_free(local_nv);

	if (fs != NULL) {
	if (flags->verbose) {
	(void) printf("snap %s already exists; "
	"ignoring\n", destsnap);
	}
	err = ioctl_err = recv_skip(hdl, infd,
	flags->byteswap);
	}
	}
	*cp = '@';
	}

	if (ioctl_err != 0) {
	switch (ioctl_errno) {
	case ENODEV:
	cp = strchr(destsnap, '@');
	*cp = '\0';
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"most recent snapshot of %s does not\n"
	"match incremental source"), destsnap);
	(void) zfs_error(hdl, EZFS_BADRESTORE, errbuf);
	*cp = '@';
	break;
	case ETXTBSY:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"destination %s has been modified\n"
	"since most recent snapshot"), name);
	(void) zfs_error(hdl, EZFS_BADRESTORE, errbuf);
	break;
	case EACCES:
	if (flags->heal) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"key must be loaded to do a non-raw "
	"corrective recv on an encrypted "
	"dataset."));
	} else if (raw && stream_wantsnewfs) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"failed to create encryption key"));
	} else if (raw && !stream_wantsnewfs) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"encryption key does not match "
	"existing key"));
	} else {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"inherited key must be loaded"));
	}
	(void) zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf);
	break;
	case EEXIST:
	cp = strchr(destsnap, '@');
	if (newfs) {
	/* it's the containing fs that exists */
	*cp = '\0';
	}
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"destination already exists"));
	(void) zfs_error_fmt(hdl, EZFS_EXISTS,
	dgettext(TEXT_DOMAIN, "cannot restore to %s"),
	destsnap);
	*cp = '@';
	break;
	case EINVAL:
	if (embedded && !raw) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"incompatible embedded data stream "
	"feature with encrypted receive."));
	} else if (flags->resumable) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"kernel modules must be upgraded to "
	"receive this stream."));
	}
	(void) zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	break;
	case ECKSUM:
	case ZFS_ERR_STREAM_TRUNCATED:
	if (flags->heal)
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"corrective receive was not able to "
	"reconstruct the data needed for "
	"healing."));
	else
	recv_ecksum_set_aux(hdl, destsnap,
	flags->resumable, ioctl_err == ECKSUM);
	(void) zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	break;
	case ZFS_ERR_STREAM_LARGE_BLOCK_MISMATCH:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"incremental send stream requires -L "
	"(--large-block), to match previous receive."));
	(void) zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	break;
	case ENOTSUP:
	if (flags->heal)
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"stream is not compatible with the "
	"data in the pool."));
	else
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"pool must be upgraded to receive this "
	"stream."));
	(void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
	break;
	case ZFS_ERR_CRYPTO_NOTSUP:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"stream uses crypto parameters not compatible with "
	"this pool"));
	(void) zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	break;
	case EDQUOT:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"destination %s space quota exceeded."), name);
	(void) zfs_error(hdl, EZFS_NOSPC, errbuf);
	break;
	case ZFS_ERR_FROM_IVSET_GUID_MISSING:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"IV set guid missing. See errata %u at "
	"https://openzfs.github.io/openzfs-docs/msg/"
	"ZFS-8000-ER."),
	ZPOOL_ERRATA_ZOL_8308_ENCRYPTION);
	(void) zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	break;
	case ZFS_ERR_FROM_IVSET_GUID_MISMATCH:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"IV set guid mismatch. See the 'zfs receive' "
	"man page section\n discussing the limitations "
	"of raw encrypted send streams."));
	(void) zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	break;
	case ZFS_ERR_SPILL_BLOCK_FLAG_MISSING:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"Spill block flag missing for raw send.\n"
	"The zfs software on the sending system must "
	"be updated."));
	(void) zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	break;
	case ZFS_ERR_RESUME_EXISTS:
	cp = strchr(destsnap, '@');
	if (newfs) {
	/* it's the containing fs that exists */
	*cp = '\0';
	}
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"Resuming recv on existing dataset without force"));
	(void) zfs_error_fmt(hdl, EZFS_RESUME_EXISTS,
	dgettext(TEXT_DOMAIN, "cannot resume recv %s"),
	destsnap);
	*cp = '@';
	break;
	case E2BIG:
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"zfs receive required kernel memory allocation "
	"larger than the system can support. Please file "
	"an issue at the OpenZFS issue tracker:\n"
	"https://github.com/openzfs/zfs/issues/new"));
	(void) zfs_error(hdl, EZFS_BADSTREAM, errbuf);
	break;
	case EBUSY:
	if (hastoken) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"destination %s contains "
	"partially-complete state from "
	"\"zfs receive -s\"."), name);
	(void) zfs_error(hdl, EZFS_BUSY, errbuf);
	break;
	}
	zfs_fallthrough;
	default:
	(void) zfs_standard_error(hdl, ioctl_errno, errbuf);
	}
	}

	/*
	* Mount the target filesystem (if created). Also mount any
	* children of the target filesystem if we did a replication
	* receive (indicated by stream_avl being non-NULL).
	*/
	if (clp) {
	if (!flags->nomount)
	err \|= changelist_postfix(clp);
	changelist_free(clp);
	}

	if ((newfs \|\| stream_avl) && type == ZFS_TYPE_FILESYSTEM && !redacted)
	flags->domount = B_TRUE;

	if (prop_errflags & ZPROP_ERR_NOCLEAR) {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Warning: "
	"failed to clear unreceived properties on %s"), name);
	(void) fprintf(stderr, "\n");
	}
	if (prop_errflags & ZPROP_ERR_NORESTORE) {
	(void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Warning: "
	"failed to restore original properties on %s"), name);
	(void) fprintf(stderr, "\n");
	}

	if (err \|\| ioctl_err) {
	err = -1;
	goto out;
	}

	if (flags->verbose) {
	char buf1[64];
	char buf2[64];
	uint64_t bytes = read_bytes;
	struct timespec delta;
	clock_gettime(CLOCK_MONOTONIC_RAW, &delta);
	if (begin_time.tv_nsec > delta.tv_nsec) {
	delta.tv_nsec =
	1000000000 + delta.tv_nsec - begin_time.tv_nsec;
	delta.tv_sec -= 1;
	} else
	delta.tv_nsec -= begin_time.tv_nsec;
	delta.tv_sec -= begin_time.tv_sec;
	if (delta.tv_sec == 0 && delta.tv_nsec == 0)
	delta.tv_nsec = 1;
	double delta_f = delta.tv_sec + (delta.tv_nsec / 1e9);
	zfs_nicebytes(bytes, buf1, sizeof (buf1));
	zfs_nicebytes(bytes / delta_f, buf2, sizeof (buf2));

	(void) printf("received %s stream in %.2f seconds (%s/sec)\n",
	buf1, delta_f, buf2);
	}

	err = 0;
	out:
	if (prop_errors != NULL)
	fnvlist_free(prop_errors);

	if (tmp_keylocation[0] != '\0') {
	fnvlist_add_string(rcvprops,
	zfs_prop_to_name(ZFS_PROP_KEYLOCATION), tmp_keylocation);
	}

	if (newprops)
	fnvlist_free(rcvprops);

	fnvlist_free(oxprops);
	fnvlist_free(origprops);

	return (err);
	}

	/*
	* Check properties we were asked to override (both -o\|-x)
	*/
	static boolean_t
	zfs_receive_checkprops(libzfs_handle_t hdl, nvlist_t props,
	const char *errbuf)
	{
	nvpair_t *nvp = NULL;
	zfs_prop_t prop;
	const char *name;

	while ((nvp = nvlist_next_nvpair(props, nvp)) != NULL) {
	name = nvpair_name(nvp);
	prop = zfs_name_to_prop(name);

	if (prop == ZPROP_USERPROP) {
	if (!zfs_prop_user(name)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"%s: invalid property '%s'"), errbuf, name);
	return (B_FALSE);
	}
	continue;
	}
	/*
	* "origin" is readonly but is used to receive datasets as
	* clones so we don't raise an error here
	*/
	if (prop == ZFS_PROP_ORIGIN)
	continue;

	/* encryption params have their own verification later */
	if (prop == ZFS_PROP_ENCRYPTION \|\|
	zfs_prop_encryption_key_param(prop))
	continue;

	/*
	* cannot override readonly, set-once and other specific
	* settable properties
	*/
	if (zfs_prop_readonly(prop) \|\| prop == ZFS_PROP_VERSION \|\|
	prop == ZFS_PROP_VOLSIZE) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"%s: invalid property '%s'"), errbuf, name);
	return (B_FALSE);
	}
	}

	return (B_TRUE);
	}

	static int
	zfs_receive_impl(libzfs_handle_t hdl, const char tosnap,
	const char originsnap, recvflags_t flags, int infd, const char *sendfs,
	nvlist_t stream_nv, avl_tree_t stream_avl, char **top_zfs,
	const char finalsnap, nvlist_t cmdprops)
	{
	int err;
	dmu_replay_record_t drr, drr_noswap;
	struct drr_begin *drrb = &drr.drr_u.drr_begin;
	char errbuf[ERRBUFLEN];
	zio_cksum_t zcksum = { { 0 } };
	uint64_t featureflags;
	int hdrtype;

	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
	"cannot receive"));

	/* check cmdline props, raise an error if they cannot be received */
	if (!zfs_receive_checkprops(hdl, cmdprops, errbuf))
	return (zfs_error(hdl, EZFS_BADPROP, errbuf));

	if (flags->isprefix &&
	!zfs_dataset_exists(hdl, tosnap, ZFS_TYPE_DATASET)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "specified fs "
	"(%s) does not exist"), tosnap);
	return (zfs_error(hdl, EZFS_NOENT, errbuf));
	}
	if (originsnap &&
	!zfs_dataset_exists(hdl, originsnap, ZFS_TYPE_DATASET)) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "specified origin fs "
	"(%s) does not exist"), originsnap);
	return (zfs_error(hdl, EZFS_NOENT, errbuf));
	}

	/* read in the BEGIN record */
	if (0 != (err = recv_read(hdl, infd, &drr, sizeof (drr), B_FALSE,
	&zcksum)))
	return (err);

	if (drr.drr_type == DRR_END \|\| drr.drr_type == BSWAP_32(DRR_END)) {
	/* It's the double end record at the end of a package */
	return (ENODATA);
	}

	/* the kernel needs the non-byteswapped begin record */
	drr_noswap = drr;

	flags->byteswap = B_FALSE;
	if (drrb->drr_magic == BSWAP_64(DMU_BACKUP_MAGIC)) {
	/*
	* We computed the checksum in the wrong byteorder in
	* recv_read() above; do it again correctly.
	*/
	memset(&zcksum, 0, sizeof (zio_cksum_t));
	fletcher_4_incremental_byteswap(&drr, sizeof (drr), &zcksum);
	flags->byteswap = B_TRUE;

	drr.drr_type = BSWAP_32(drr.drr_type);
	drr.drr_payloadlen = BSWAP_32(drr.drr_payloadlen);
	drrb->drr_magic = BSWAP_64(drrb->drr_magic);
	drrb->drr_versioninfo = BSWAP_64(drrb->drr_versioninfo);
	drrb->drr_creation_time = BSWAP_64(drrb->drr_creation_time);
	drrb->drr_type = BSWAP_32(drrb->drr_type);
	drrb->drr_flags = BSWAP_32(drrb->drr_flags);
	drrb->drr_toguid = BSWAP_64(drrb->drr_toguid);
	drrb->drr_fromguid = BSWAP_64(drrb->drr_fromguid);
	}

	if (drrb->drr_magic != DMU_BACKUP_MAGIC \|\| drr.drr_type != DRR_BEGIN) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid "
	"stream (bad magic number)"));
	return (zfs_error(hdl, EZFS_BADSTREAM, errbuf));
	}

	featureflags = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo);
	hdrtype = DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo);

	if (!DMU_STREAM_SUPPORTED(featureflags) \|\|
	(hdrtype != DMU_SUBSTREAM && hdrtype != DMU_COMPOUNDSTREAM)) {
	/*
	* Let's be explicit about this one, since rather than
	* being a new feature we can't know, it's an old
	* feature we dropped.
	*/
	if (featureflags & DMU_BACKUP_FEATURE_DEDUP) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"stream has deprecated feature: dedup, try "
	"'zstream redup [send in a file] \| zfs recv "
	"[...]'"));
	} else {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
	"stream has unsupported feature, feature flags = "
	"%llx (unknown flags = %llx)"),
	(u_longlong_t)featureflags,
	(u_longlong_t)((featureflags) &
	~DMU_BACKUP_FEATURE_MASK));
	}
	return (zfs_error(hdl, EZFS_BADSTREAM, errbuf));
	}

	/* Holds feature is set once in the compound stream header. */
	if (featureflags & DMU_BACKUP_FEATURE_HOLDS)
	flags->holds = B_TRUE;

	if (strchr(drrb->drr_toname, '@') == NULL) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "invalid "
	"stream (bad snapshot name)"));
	return (zfs_error(hdl, EZFS_BADSTREAM, errbuf));
	}

	if (DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo) == DMU_SUBSTREAM) {
	char nonpackage_sendfs[ZFS_MAX_DATASET_NAME_LEN];
	if (sendfs == NULL) {
	/*
	* We were not called from zfs_receive_package(). Get
	* the fs specified by 'zfs send'.
	*/
	char *cp;
	(void) strlcpy(nonpackage_sendfs,
	drr.drr_u.drr_begin.drr_toname,
	sizeof (nonpackage_sendfs));
	if ((cp = strchr(nonpackage_sendfs, '@')) != NULL)
	*cp = '\0';
	sendfs = nonpackage_sendfs;
	VERIFY(finalsnap == NULL);
	}
	return (zfs_receive_one(hdl, infd, tosnap, originsnap, flags,
	&drr, &drr_noswap, sendfs, stream_nv, stream_avl, top_zfs,
	finalsnap, cmdprops));
	} else {
	assert(DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo) ==
	DMU_COMPOUNDSTREAM);
	return (zfs_receive_package(hdl, infd, tosnap, flags, &drr,
	&zcksum, top_zfs, cmdprops));
	}
	}

	/*
	* Restores a backup of tosnap from the file descriptor specified by infd.
	* Return 0 on total success, -2 if some things couldn't be
	* destroyed/renamed/promoted, -1 if some things couldn't be received.
	* (-1 will override -2, if -1 and the resumable flag was specified the
	* transfer can be resumed if the sending side supports it).
	*/
	int
	zfs_receive(libzfs_handle_t hdl, const char tosnap, nvlist_t *props,
	recvflags_t flags, int infd, avl_tree_t stream_avl)
	{
	char *top_zfs = NULL;
	int err;
	struct stat sb;
	const char *originsnap = NULL;

	/*
	* The only way fstat can fail is if we do not have a valid file
	* descriptor.
	*/
	if (fstat(infd, &sb) == -1) {
	perror("fstat");
	return (-2);
	}

	if (props) {
	err = nvlist_lookup_string(props, "origin", &originsnap);
	if (err && err != ENOENT)
	return (err);
	}

	err = zfs_receive_impl(hdl, tosnap, originsnap, flags, infd, NULL, NULL,
	stream_avl, &top_zfs, NULL, props);

	if (err == 0 && !flags->nomount && flags->domount && top_zfs) {
	zfs_handle_t *zhp = NULL;
	prop_changelist_t *clp = NULL;

	zhp = zfs_open(hdl, top_zfs,
	ZFS_TYPE_FILESYSTEM \| ZFS_TYPE_VOLUME);
	if (zhp == NULL) {
	err = -1;
	goto out;
	} else {
	if (zhp->zfs_type == ZFS_TYPE_VOLUME) {
	zfs_close(zhp);
	goto out;
	}

	clp = changelist_gather(zhp, ZFS_PROP_MOUNTPOINT,
	CL_GATHER_MOUNT_ALWAYS,
	flags->forceunmount ? MS_FORCE : 0);
	zfs_close(zhp);
	if (clp == NULL) {
	err = -1;
	goto out;
	}

	/* mount and share received datasets */
	err = changelist_postfix(clp);
	changelist_free(clp);
	if (err != 0)
	err = -1;
	}
	}

	out:
	if (top_zfs)
	free(top_zfs);

	return (err);
	}
	diff --git a/sys/contrib/openzfs/lib/libzfs/os/linux/libzfs_pool_os.c b/sys/contrib/openzfs/lib/libzfs/os/linux/libzfs_pool_os.c
	index 86eef3255bc2..7b18e31c8674 100644
	--- a/sys/contrib/openzfs/lib/libzfs/os/linux/libzfs_pool_os.c
	+++ b/sys/contrib/openzfs/lib/libzfs/os/linux/libzfs_pool_os.c
	@@ -1,350 +1,351 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright 2015 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2018 by Delphix. All rights reserved.
	* Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>
	* Copyright (c) 2018 Datto Inc.
	* Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
	* Copyright (c) 2017, Intel Corporation.
	* Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>
	*/

	#include <errno.h>
	#include <libintl.h>
	#include <stdio.h>
	#include <stdlib.h>
	#include <string.h>
	#include <unistd.h>
	#include <libgen.h>
	#include <zone.h>
	#include <sys/stat.h>
	#include <sys/efi_partition.h>
	#include <sys/systeminfo.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/vdev_disk.h>
	#include <dlfcn.h>
	#include <libzutil.h>

	#include "zfs_namecheck.h"
	#include "zfs_prop.h"
	#include "../../libzfs_impl.h"
	#include "zfs_comutil.h"
	#include "zfeature_common.h"

	/*
	* If the device has being dynamically expanded then we need to relabel
	* the disk to use the new unallocated space.
	*/
	int
	zpool_relabel_disk(libzfs_handle_t hdl, const char path, const char *msg)
	{
	int fd, error;

	if ((fd = open(path, O_RDWR\|O_DIRECT\|O_CLOEXEC)) < 0) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
	"relabel '%s': unable to open device: %d"), path, errno);
	return (zfs_error(hdl, EZFS_OPENFAILED, msg));
	}

	/*
	* It's possible that we might encounter an error if the device
	* does not have any unallocated space left. If so, we simply
	* ignore that error and continue on.
	*/
	error = efi_use_whole_disk(fd);

	/* Flush the buffers to disk and invalidate the page cache. */
	(void) fsync(fd);
	(void) ioctl(fd, BLKFLSBUF);

	(void) close(fd);
	if (error && error != VT_ENOSPC) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
	"relabel '%s': unable to read disk capacity"), path);
	return (zfs_error(hdl, EZFS_NOCAP, msg));
	}
	return (0);
	}

	/*
	* Read the EFI label from the config, if a label does not exist then
	* pass back the error to the caller. If the caller has passed a non-NULL
	* diskaddr argument then we set it to the starting address of the EFI
	* partition.
	*/
	static int
	read_efi_label(nvlist_t config, diskaddr_t sb)
	{
	const char *path;
	int fd;
	char diskname[MAXPATHLEN];
	int err = -1;

	if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0)
	return (err);

	(void) snprintf(diskname, sizeof (diskname), "%s%s", DISK_ROOT,
	strrchr(path, '/'));
	if ((fd = open(diskname, O_RDONLY\|O_DIRECT\|O_CLOEXEC)) >= 0) {
	struct dk_gpt *vtoc;

	if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) {
	if (sb != NULL)
	*sb = vtoc->efi_parts[0].p_start;
	efi_free(vtoc);
	}
	(void) close(fd);
	}
	return (err);
	}

	/*
	* determine where a partition starts on a disk in the current
	* configuration
	*/
	static diskaddr_t
	find_start_block(nvlist_t *config)
	{
	nvlist_t **child;
	uint_t c, children;
	diskaddr_t sb = MAXOFFSET_T;
	uint64_t wholedisk;

	if (nvlist_lookup_nvlist_array(config,
	ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) {
	if (nvlist_lookup_uint64(config,
	ZPOOL_CONFIG_WHOLE_DISK,
	&wholedisk) != 0 \|\| !wholedisk) {
	return (MAXOFFSET_T);
	}
	if (read_efi_label(config, &sb) < 0)
	sb = MAXOFFSET_T;
	return (sb);
	}

	for (c = 0; c < children; c++) {
	sb = find_start_block(child[c]);
	if (sb != MAXOFFSET_T) {
	return (sb);
	}
	}
	return (MAXOFFSET_T);
	}

	static int
	zpool_label_disk_check(char *path)
	{
	struct dk_gpt *vtoc;
	int fd, err;

	if ((fd = open(path, O_RDONLY\|O_DIRECT\|O_CLOEXEC)) < 0)
	return (errno);

	if ((err = efi_alloc_and_read(fd, &vtoc)) != 0) {
	(void) close(fd);
	return (err);
	}

	if (vtoc->efi_flags & EFI_GPT_PRIMARY_CORRUPT) {
	efi_free(vtoc);
	(void) close(fd);
	return (EIDRM);
	}

	efi_free(vtoc);
	(void) close(fd);
	return (0);
	}

	/*
	* Generate a unique partition name for the ZFS member. Partitions must
	* have unique names to ensure udev will be able to create symlinks under
	* /dev/disk/by-partlabel/ for all pool members. The partition names are
	* of the form <pool>-<unique-id>.
	*/
	static void
	zpool_label_name(char *label_name, int label_size)
	{
	uint64_t id = 0;
	int fd;

	fd = open("/dev/urandom", O_RDONLY\|O_CLOEXEC);
	if (fd >= 0) {
	if (read(fd, &id, sizeof (id)) != sizeof (id))
	id = 0;

	close(fd);
	}

	if (id == 0)
	id = (((uint64_t)rand()) << 32) \| (uint64_t)rand();

	snprintf(label_name, label_size, "zfs-%016llx", (u_longlong_t)id);
	}

	/*
	* Label an individual disk. The name provided is the short name,
	* stripped of any leading /dev path.
	*/
	int
	zpool_label_disk(libzfs_handle_t hdl, zpool_handle_t zhp, const char *name)
	{
	char path[MAXPATHLEN];
	struct dk_gpt *vtoc;
	int rval, fd;
	size_t resv = EFI_MIN_RESV_SIZE;
	uint64_t slice_size;
	diskaddr_t start_block;
	char errbuf[ERRBUFLEN];

	/* prepare an error message just in case */
	(void) snprintf(errbuf, sizeof (errbuf),
	dgettext(TEXT_DOMAIN, "cannot label '%s'"), name);

	if (zhp) {
	nvlist_t *nvroot = fnvlist_lookup_nvlist(zhp->zpool_config,
	ZPOOL_CONFIG_VDEV_TREE);

	if (zhp->zpool_start_block == 0)
	start_block = find_start_block(nvroot);
	else
	start_block = zhp->zpool_start_block;
	zhp->zpool_start_block = start_block;
	} else {
	/* new pool */
	start_block = NEW_START_BLOCK;
	}

	(void) snprintf(path, sizeof (path), "%s/%s", DISK_ROOT, name);

	if ((fd = open(path, O_RDWR\|O_DIRECT\|O_EXCL\|O_CLOEXEC)) < 0) {
	/*
	* This shouldn't happen. We've long since verified that this
	* is a valid device.
	*/
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
	"label '%s': unable to open device: %d"), path, errno);
	return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
	}

	if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) {
	/*
	* The only way this can fail is if we run out of memory, or we
	* were unable to read the disk's capacity
	*/
	if (errno == ENOMEM)
	(void) no_memory(hdl);

	(void) close(fd);
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
	"label '%s': unable to read disk capacity"), path);

	return (zfs_error(hdl, EZFS_NOCAP, errbuf));
	}

	slice_size = vtoc->efi_last_u_lba + 1;
	slice_size -= EFI_MIN_RESV_SIZE;
	if (start_block == MAXOFFSET_T)
	start_block = NEW_START_BLOCK;
	slice_size -= start_block;
	- slice_size = P2ALIGN(slice_size, PARTITION_END_ALIGNMENT);
	+ slice_size = P2ALIGN_TYPED(slice_size, PARTITION_END_ALIGNMENT,
	+ uint64_t);

	vtoc->efi_parts[0].p_start = start_block;
	vtoc->efi_parts[0].p_size = slice_size;

	if (vtoc->efi_parts[0].p_size * vtoc->efi_lbasize < SPA_MINDEVSIZE) {
	(void) close(fd);
	efi_free(vtoc);

	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
	"label '%s': partition would be less than the minimum "
	"device size (64M)"), path);
	return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
	}

	/*
	* Why we use V_USR: V_BACKUP confuses users, and is considered
	* disposable by some EFI utilities (since EFI doesn't have a backup
	* slice). V_UNASSIGNED is supposed to be used only for zero size
	* partitions, and efi_write() will fail if we use it.
	* Other available types were all pretty specific.
	* V_USR is as close to reality as we
	* can get, in the absence of V_OTHER.
	*/
	vtoc->efi_parts[0].p_tag = V_USR;
	zpool_label_name(vtoc->efi_parts[0].p_name, EFI_PART_NAME_LEN);

	vtoc->efi_parts[8].p_start = slice_size + start_block;
	vtoc->efi_parts[8].p_size = resv;
	vtoc->efi_parts[8].p_tag = V_RESERVED;

	rval = efi_write(fd, vtoc);

	/* Flush the buffers to disk and invalidate the page cache. */
	(void) fsync(fd);
	(void) ioctl(fd, BLKFLSBUF);

	if (rval == 0)
	rval = efi_rescan(fd);

	/*
	* Some block drivers (like pcata) may not support EFI GPT labels.
	* Print out a helpful error message directing the user to manually
	* label the disk and give a specific slice.
	*/
	if (rval != 0) {
	(void) close(fd);
	efi_free(vtoc);

	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "try using "
	"parted(8) and then provide a specific slice: %d"), rval);
	return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
	}

	(void) close(fd);
	efi_free(vtoc);

	(void) snprintf(path, sizeof (path), "%s/%s", DISK_ROOT, name);
	(void) zfs_append_partition(path, MAXPATHLEN);

	/* Wait to udev to signal use the device has settled. */
	rval = zpool_label_disk_wait(path, DISK_LABEL_WAIT);
	if (rval) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "failed to "
	"detect device partitions on '%s': %d"), path, rval);
	return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
	}

	/* We can't be to paranoid. Read the label back and verify it. */
	(void) snprintf(path, sizeof (path), "%s/%s", DISK_ROOT, name);
	rval = zpool_label_disk_check(path);
	if (rval) {
	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "freshly written "
	"EFI label on '%s' is damaged. Ensure\nthis device "
	"is not in use, and is functioning properly: %d"),
	path, rval);
	return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
	}
	return (0);
	}
	diff --git a/sys/contrib/openzfs/lib/libzfs_core/libzfs_core.abi b/sys/contrib/openzfs/lib/libzfs_core/libzfs_core.abi
	index 5b95c8f779db..cf9d6bddc9fc 100644
	--- a/sys/contrib/openzfs/lib/libzfs_core/libzfs_core.abi
	+++ b/sys/contrib/openzfs/lib/libzfs_core/libzfs_core.abi
	@@ -1,2930 +1,3003 @@
	<abi-corpus version='2.0' architecture='elf-amd-x86_64' soname='libzfs_core.so.3'>
	<elf-needed>
	<dependency name='libnvpair.so.3'/>
	<dependency name='libc.so.6'/>
	<dependency name='ld-linux-x86-64.so.2'/>
	</elf-needed>
	<elf-function-symbols>
	<elf-symbol name='_sol_getmntent' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_char' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_char_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_int' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_int_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_long' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_long_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_ptr' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_ptr_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_short' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_add_short_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_uchar_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_uint_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_ulong_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_and_ushort_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_ptr' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_cas_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_clear_long_excl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_uchar_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_uint_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_ulong_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_dec_ushort_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_uchar_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_uint_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_ulong_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_inc_ushort_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_uchar_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_uint_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_ulong_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_or_ushort_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_set_long_excl' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_16_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_32_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_64_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_8_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_char' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_char_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_int' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_int_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_long' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_long_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_ptr' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_ptr_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_short' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_sub_short_nv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_16' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_32' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_64' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_8' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_ptr' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_uchar' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_uint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_ulong' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='atomic_swap_ushort' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='get_system_hostid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='getexecname' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='getextmntent' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='getmntany' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='getzoneid' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libspl_assertf' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	+ <elf-symbol name='libspl_backtrace' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libspl_set_assert_ok' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_core_fini' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='libzfs_core_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_destroy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_head' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_insert_after' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_insert_before' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_insert_head' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_insert_tail' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_is_empty' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_link_active' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_link_init' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_link_replace' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_move_tail' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_next' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_prev' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_remove' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_remove_head' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_remove_tail' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='list_tail' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_bookmark' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_change_key' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_channel_program' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_channel_program_nosync' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_clone' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_create' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_destroy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_destroy_bookmarks' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_destroy_snaps' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_exists' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_get_bookmark_props' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_get_bookmarks' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_get_bootenv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_get_holds' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_get_vdev_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_hold' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_initialize' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_ioctl_fd' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_load_key' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_pool_checkpoint' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_pool_checkpoint_discard' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_promote' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_receive' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_receive_one' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_receive_resumable' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_receive_with_cmdprops' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_receive_with_header' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_receive_with_heal' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_redact' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_release' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_rename' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_reopen' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_rollback' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_rollback_to' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_scrub' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_send' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_send_redacted' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_send_resume' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_send_resume_redacted' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_send_space' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_send_space_resume_redacted' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_send_wrapper' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_set_bootenv' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_set_vdev_prop' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_snaprange_space' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_snapshot' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_sync' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_trim' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_unload_key' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_wait' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_wait_fs' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='lzc_wait_tag' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='membar_consumer' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='membar_enter' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='membar_exit' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='membar_producer' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='membar_sync' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='mkdirp' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='print_timestamp' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='spl_pagesize' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='strlcat' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	<elf-symbol name='strlcpy' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
	</elf-function-symbols>
	<abi-instr address-size='64' path='lib/libspl/assert.c' language='LANG_C99'>
	<class-decl name='__va_list_tag' size-in-bits='192' is-struct='yes' visibility='default' id='d5027220'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='gp_offset' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='fp_offset' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='overflow_arg_area' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='reg_save_area' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	</class-decl>
	+ <typedef-decl name='__pid_t' type-id='95e97e5e' id='3629bad8'/>
	<pointer-type-def type-id='d5027220' size-in-bits='64' id='b7f2d5e6'/>
	+ <function-decl name='libspl_backtrace' mangled-name='libspl_backtrace' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libspl_backtrace'>
	+ <parameter type-id='95e97e5e'/>
	+ <return type-id='48b5725f'/>
	+ </function-decl>
	<function-decl name='abort' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='48b5725f'/>
	</function-decl>
	+ <function-decl name='getpid' visibility='default' binding='global' size-in-bits='64'>
	+ <return type-id='3629bad8'/>
	+ </function-decl>
	<function-decl name='__vfprintf_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='e75a27e9'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='9d26089a'/>
	<parameter type-id='b7f2d5e6'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	+ <function-decl name='gettid' visibility='default' binding='global' size-in-bits='64'>
	+ <return type-id='3629bad8'/>
	+ </function-decl>
	+ <function-decl name='prctl' visibility='default' binding='global' size-in-bits='64'>
	+ <parameter type-id='95e97e5e'/>
	+ <parameter is-variadic='yes'/>
	+ <return type-id='95e97e5e'/>
	+ </function-decl>
	<function-decl name='libspl_set_assert_ok' mangled-name='libspl_set_assert_ok' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libspl_set_assert_ok'>
	<parameter type-id='c19b74c3' name='val'/>
	<return type-id='48b5725f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/atomic.c' language='LANG_C99'>
	<typedef-decl name='ulong_t' type-id='7359adad' id='ee1f298e'/>
	<typedef-decl name='int8_t' type-id='2171a512' id='ee31ee44'/>
	<typedef-decl name='uint16_t' type-id='253c2d2a' id='149c6638'/>
	<typedef-decl name='__int8_t' type-id='28577a57' id='2171a512'/>
	<typedef-decl name='__uint16_t' type-id='8efea9e5' id='253c2d2a'/>
	<qualified-type-def type-id='149c6638' volatile='yes' id='5120c5f7'/>
	<pointer-type-def type-id='5120c5f7' size-in-bits='64' id='93977ae7'/>
	<qualified-type-def type-id='8f92235e' volatile='yes' id='430e0681'/>
	<pointer-type-def type-id='430e0681' size-in-bits='64' id='3a147f31'/>
	<qualified-type-def type-id='b96825af' volatile='yes' id='84ff7d66'/>
	<pointer-type-def type-id='84ff7d66' size-in-bits='64' id='aa323ea4'/>
	<qualified-type-def type-id='ee1f298e' volatile='yes' id='6f7e09cb'/>
	<pointer-type-def type-id='6f7e09cb' size-in-bits='64' id='64698d33'/>
	<qualified-type-def type-id='48b5725f' volatile='yes' id='b0b3cbf9'/>
	<pointer-type-def type-id='b0b3cbf9' size-in-bits='64' id='fe09dd29'/>
	<function-decl name='atomic_inc_8' mangled-name='atomic_inc_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_inc_16' mangled-name='atomic_inc_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_16'>
	<parameter type-id='93977ae7' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_inc_32' mangled-name='atomic_inc_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_32'>
	<parameter type-id='3a147f31' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_inc_ulong' mangled-name='atomic_inc_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_dec_8' mangled-name='atomic_dec_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_dec_16' mangled-name='atomic_dec_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_16'>
	<parameter type-id='93977ae7' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_dec_32' mangled-name='atomic_dec_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_32'>
	<parameter type-id='3a147f31' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_dec_ulong' mangled-name='atomic_dec_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_add_ptr' mangled-name='atomic_add_ptr' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_ptr'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='79a0948f' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_add_8' mangled-name='atomic_add_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='ee31ee44' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_add_16' mangled-name='atomic_add_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='23bd8cb5' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_add_32' mangled-name='atomic_add_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_32'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='3ff5601b' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_sub_ptr' mangled-name='atomic_sub_ptr' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_ptr'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='79a0948f' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_sub_8' mangled-name='atomic_sub_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='ee31ee44' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_sub_16' mangled-name='atomic_sub_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='23bd8cb5' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_sub_32' mangled-name='atomic_sub_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_32'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='3ff5601b' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_or_8' mangled-name='atomic_or_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_or_16' mangled-name='atomic_or_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_or_32' mangled-name='atomic_or_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_32'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='8f92235e' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_or_ulong' mangled-name='atomic_or_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_and_8' mangled-name='atomic_and_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_and_16' mangled-name='atomic_and_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_and_32' mangled-name='atomic_and_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_32'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='8f92235e' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_and_ulong' mangled-name='atomic_and_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='bits'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='atomic_inc_8_nv' mangled-name='atomic_inc_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_inc_16_nv' mangled-name='atomic_inc_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_inc_32_nv' mangled-name='atomic_inc_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_inc_ulong_nv' mangled-name='atomic_inc_ulong_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_inc_ulong_nv'>
	<parameter type-id='64698d33' name='target'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_dec_8_nv' mangled-name='atomic_dec_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_dec_16_nv' mangled-name='atomic_dec_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_dec_32_nv' mangled-name='atomic_dec_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_dec_ulong_nv' mangled-name='atomic_dec_ulong_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_dec_ulong_nv'>
	<parameter type-id='64698d33' name='target'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_add_ptr_nv' mangled-name='atomic_add_ptr_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_ptr_nv'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='79a0948f' name='bits'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='atomic_add_8_nv' mangled-name='atomic_add_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='ee31ee44' name='bits'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_add_16_nv' mangled-name='atomic_add_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='23bd8cb5' name='bits'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_add_32_nv' mangled-name='atomic_add_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='3ff5601b' name='bits'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_add_long_nv' mangled-name='atomic_add_long_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_add_long_nv'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='bd54fe1a' name='bits'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_sub_ptr_nv' mangled-name='atomic_sub_ptr_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_ptr_nv'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='79a0948f' name='bits'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='atomic_sub_8_nv' mangled-name='atomic_sub_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='ee31ee44' name='bits'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_sub_16_nv' mangled-name='atomic_sub_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='23bd8cb5' name='bits'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_sub_32_nv' mangled-name='atomic_sub_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='3ff5601b' name='bits'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_sub_long_nv' mangled-name='atomic_sub_long_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_sub_long_nv'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='bd54fe1a' name='bits'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_or_8_nv' mangled-name='atomic_or_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='bits'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_or_16_nv' mangled-name='atomic_or_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='bits'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_or_32_nv' mangled-name='atomic_or_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='8f92235e' name='bits'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_or_ulong_nv' mangled-name='atomic_or_ulong_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_or_ulong_nv'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='bits'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_and_8_nv' mangled-name='atomic_and_8_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_8_nv'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='bits'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_and_16_nv' mangled-name='atomic_and_16_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_16_nv'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='bits'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_and_32_nv' mangled-name='atomic_and_32_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_32_nv'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='8f92235e' name='bits'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_and_ulong_nv' mangled-name='atomic_and_ulong_nv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_and_ulong_nv'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='bits'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_cas_ptr' mangled-name='atomic_cas_ptr' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_ptr'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='eaa32e2f' name='exp'/>
	<parameter type-id='eaa32e2f' name='des'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='atomic_cas_8' mangled-name='atomic_cas_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='exp'/>
	<parameter type-id='b96825af' name='des'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_cas_16' mangled-name='atomic_cas_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='exp'/>
	<parameter type-id='149c6638' name='des'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_cas_32' mangled-name='atomic_cas_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_32'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='8f92235e' name='exp'/>
	<parameter type-id='8f92235e' name='des'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_cas_ulong' mangled-name='atomic_cas_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_cas_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='exp'/>
	<parameter type-id='ee1f298e' name='des'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_swap_8' mangled-name='atomic_swap_8' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_8'>
	<parameter type-id='aa323ea4' name='target'/>
	<parameter type-id='b96825af' name='bits'/>
	<return type-id='b96825af'/>
	</function-decl>
	<function-decl name='atomic_swap_16' mangled-name='atomic_swap_16' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_16'>
	<parameter type-id='93977ae7' name='target'/>
	<parameter type-id='149c6638' name='bits'/>
	<return type-id='149c6638'/>
	</function-decl>
	<function-decl name='atomic_swap_32' mangled-name='atomic_swap_32' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_32'>
	<parameter type-id='3a147f31' name='target'/>
	<parameter type-id='8f92235e' name='bits'/>
	<return type-id='8f92235e'/>
	</function-decl>
	<function-decl name='atomic_swap_ulong' mangled-name='atomic_swap_ulong' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_ulong'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='ee1f298e' name='bits'/>
	<return type-id='ee1f298e'/>
	</function-decl>
	<function-decl name='atomic_swap_ptr' mangled-name='atomic_swap_ptr' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_swap_ptr'>
	<parameter type-id='fe09dd29' name='target'/>
	<parameter type-id='eaa32e2f' name='bits'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='atomic_set_long_excl' mangled-name='atomic_set_long_excl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_set_long_excl'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='3502e3ff' name='value'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='atomic_clear_long_excl' mangled-name='atomic_clear_long_excl' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='atomic_clear_long_excl'>
	<parameter type-id='64698d33' name='target'/>
	<parameter type-id='3502e3ff' name='value'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='membar_enter' mangled-name='membar_enter' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='membar_enter'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='membar_producer' mangled-name='membar_producer' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='membar_producer'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='membar_consumer' mangled-name='membar_consumer' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='membar_consumer'>
	<return type-id='48b5725f'/>
	</function-decl>
	<type-decl name='unsigned long int' size-in-bits='64' id='7359adad'/>
	<type-decl name='unsigned short int' size-in-bits='16' id='8efea9e5'/>
	</abi-instr>
	+ <abi-instr address-size='64' path='lib/libspl/backtrace.c' language='LANG_C99'>
	+ <qualified-type-def type-id='eaa32e2f' const='yes' id='83be723c'/>
	+ <pointer-type-def type-id='83be723c' size-in-bits='64' id='7acd98a2'/>
	+ <function-decl name='backtrace' visibility='default' binding='global' size-in-bits='64'>
	+ <parameter type-id='63e171df'/>
	+ <parameter type-id='95e97e5e'/>
	+ <return type-id='95e97e5e'/>
	+ </function-decl>
	+ <function-decl name='backtrace_symbols_fd' visibility='default' binding='global' size-in-bits='64'>
	+ <parameter type-id='7acd98a2'/>
	+ <parameter type-id='95e97e5e'/>
	+ <parameter type-id='95e97e5e'/>
	+ <return type-id='48b5725f'/>
	+ </function-decl>
	+ <function-decl name='write' visibility='default' binding='global' size-in-bits='64'>
	+ <parameter type-id='95e97e5e'/>
	+ <parameter type-id='eaa32e2f'/>
	+ <parameter type-id='b59d7dce'/>
	+ <return type-id='79a0948f'/>
	+ </function-decl>
	+ </abi-instr>
	<abi-instr address-size='64' path='lib/libspl/getexecname.c' language='LANG_C99'>
	<function-decl name='getexecname' mangled-name='getexecname' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='getexecname'>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='getexecname_impl' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='26a90f95'/>
	<return type-id='79a0948f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/list.c' language='LANG_C99'>
	<typedef-decl name='list_node_t' type-id='b0b5e45e' id='b21843b2'/>
	<typedef-decl name='list_t' type-id='e824dae9' id='0899125f'/>
	<class-decl name='list_node' size-in-bits='128' is-struct='yes' visibility='default' id='b0b5e45e'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='next' type-id='b03eadb4' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='prev' type-id='b03eadb4' visibility='default'/>
	</data-member>
	</class-decl>
	- <class-decl name='list' size-in-bits='256' is-struct='yes' visibility='default' id='e824dae9'>
	+ <class-decl name='list' size-in-bits='192' is-struct='yes' visibility='default' id='e824dae9'>
	<data-member access='public' layout-offset-in-bits='0'>
	- <var-decl name='list_size' type-id='b59d7dce' visibility='default'/>
	- </data-member>
	- <data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='list_offset' type-id='b59d7dce' visibility='default'/>
	</data-member>
	- <data-member access='public' layout-offset-in-bits='128'>
	+ <data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='list_head' type-id='b0b5e45e' visibility='default'/>
	</data-member>
	</class-decl>
	<pointer-type-def type-id='b0b5e45e' size-in-bits='64' id='b03eadb4'/>
	<pointer-type-def type-id='b21843b2' size-in-bits='64' id='ccc38265'/>
	<pointer-type-def type-id='0899125f' size-in-bits='64' id='352ec160'/>
	<function-decl name='list_create' mangled-name='list_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_create'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='b59d7dce' name='size'/>
	<parameter type-id='b59d7dce' name='offset'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_destroy' mangled-name='list_destroy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_destroy'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_insert_after' mangled-name='list_insert_after' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_insert_after'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<parameter type-id='eaa32e2f' name='nobject'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_insert_before' mangled-name='list_insert_before' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_insert_before'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<parameter type-id='eaa32e2f' name='nobject'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_insert_head' mangled-name='list_insert_head' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_insert_head'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_insert_tail' mangled-name='list_insert_tail' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_insert_tail'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_remove' mangled-name='list_remove' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_remove'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_remove_head' mangled-name='list_remove_head' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_remove_head'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_remove_tail' mangled-name='list_remove_tail' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_remove_tail'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_head' mangled-name='list_head' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_head'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_tail' mangled-name='list_tail' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_tail'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_next' mangled-name='list_next' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_next'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_prev' mangled-name='list_prev' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_prev'>
	<parameter type-id='352ec160' name='list'/>
	<parameter type-id='eaa32e2f' name='object'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='list_move_tail' mangled-name='list_move_tail' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_move_tail'>
	<parameter type-id='352ec160' name='dst'/>
	<parameter type-id='352ec160' name='src'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_link_replace' mangled-name='list_link_replace' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_link_replace'>
	<parameter type-id='ccc38265' name='lold'/>
	<parameter type-id='ccc38265' name='lnew'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_link_init' mangled-name='list_link_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_link_init'>
	<parameter type-id='ccc38265' name='ln'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='list_link_active' mangled-name='list_link_active' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_link_active'>
	<parameter type-id='ccc38265' name='ln'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='list_is_empty' mangled-name='list_is_empty' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='list_is_empty'>
	<parameter type-id='352ec160' name='list'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/mkdirp.c' language='LANG_C99'>
	<typedef-decl name='mode_t' type-id='e1c52942' id='d50d396c'/>
	<typedef-decl name='wchar_t' type-id='95e97e5e' id='928221d2'/>
	<qualified-type-def type-id='26a90f95' restrict='yes' id='266fe297'/>
	<qualified-type-def type-id='928221d2' const='yes' id='effb3702'/>
	<pointer-type-def type-id='effb3702' size-in-bits='64' id='f077d3f8'/>
	<qualified-type-def type-id='f077d3f8' restrict='yes' id='598aab80'/>
	<pointer-type-def type-id='928221d2' size-in-bits='64' id='323d93c1'/>
	<qualified-type-def type-id='323d93c1' restrict='yes' id='f1358bc3'/>
	<function-decl name='calloc' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='strdup' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='strlen' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='access' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='__mbstowcs_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='f1358bc3'/>
	<parameter type-id='9d26089a'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='__wcstombs_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='266fe297'/>
	<parameter type-id='598aab80'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='mkdir' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='e1c52942'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='mkdirp' mangled-name='mkdirp' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='mkdirp'>
	<parameter type-id='80f4b756' name='d'/>
	<parameter type-id='d50d396c' name='mode'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/os/linux/getexecname.c' language='LANG_C99'>
	<function-decl name='__readlink_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9d26089a'/>
	<parameter type-id='266fe297'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='79a0948f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/os/linux/gethostid.c' language='LANG_C99'>
	<type-decl name='long long unsigned int' size-in-bits='64' id='3a47d82b'/>
	<pointer-type-def type-id='26a90f95' size-in-bits='64' id='9b23c9ad'/>
	<qualified-type-def type-id='9b23c9ad' restrict='yes' id='8c85230f'/>
	<function-decl name='strtoull' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9d26089a'/>
	<parameter type-id='8c85230f'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='3a47d82b'/>
	</function-decl>
	<function-decl name='get_system_hostid' mangled-name='get_system_hostid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='get_system_hostid'>
	<return type-id='7359adad'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/os/linux/getmntany.c' language='LANG_C99'>
	+ <array-type-def dimensions='1' type-id='38b51b3c' size-in-bits='832' id='02b72c00'>
	+ <subrange length='13' type-id='7359adad' id='487fded1'/>
	+ </array-type-def>
	+ <class-decl name='__locale_data' is-struct='yes' visibility='default' is-declaration-only='yes' id='23de8b96'/>
	+ <array-type-def dimensions='1' type-id='80f4b756' size-in-bits='832' id='39e6f84a'>
	+ <subrange length='13' type-id='7359adad' id='487fded1'/>
	+ </array-type-def>
	<class-decl name='mnttab' size-in-bits='256' is-struct='yes' visibility='default' id='1b055409'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='mnt_special' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='mnt_mountp' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='mnt_fstype' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='mnt_mntopts' type-id='26a90f95' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='extmnttab' size-in-bits='320' is-struct='yes' visibility='default' id='0c544dc0'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='mnt_special' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='mnt_mountp' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='mnt_fstype' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='mnt_mntopts' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='mnt_major' type-id='3502e3ff' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='mnt_minor' type-id='3502e3ff' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='mntent' size-in-bits='320' is-struct='yes' visibility='default' id='56fe4a37'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='mnt_fsname' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='mnt_dir' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='mnt_type' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='mnt_opts' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='mnt_freq' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='mnt_passno' type-id='95e97e5e' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='stat64' size-in-bits='1152' is-struct='yes' visibility='default' id='0bbec9cd'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='st_dev' type-id='35ed8932' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='st_ino' type-id='71288a47' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='st_nlink' type-id='80f0b9df' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='st_mode' type-id='e1c52942' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='224'>
	<var-decl name='st_uid' type-id='cc5fcceb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='st_gid' type-id='d94ec6d9' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='__pad0' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='st_rdev' type-id='35ed8932' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='st_size' type-id='79989e9c' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='st_blksize' type-id='d3f10a7f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='st_blocks' type-id='4e711bf1' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='576'>
	<var-decl name='st_atim' type-id='a9c79a1f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='704'>
	<var-decl name='st_mtim' type-id='a9c79a1f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='832'>
	<var-decl name='st_ctim' type-id='a9c79a1f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='960'>
	<var-decl name='__glibc_reserved' type-id='083f8d58' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='__ino64_t' type-id='7359adad' id='71288a47'/>
	<typedef-decl name='__blkcnt64_t' type-id='bd54fe1a' id='4e711bf1'/>
	+ <class-decl name='__locale_struct' size-in-bits='1856' is-struct='yes' visibility='default' id='90cc1ce3'>
	+ <data-member access='public' layout-offset-in-bits='0'>
	+ <var-decl name='__locales' type-id='02b72c00' visibility='default'/>
	+ </data-member>
	+ <data-member access='public' layout-offset-in-bits='832'>
	+ <var-decl name='__ctype_b' type-id='31347b7a' visibility='default'/>
	+ </data-member>
	+ <data-member access='public' layout-offset-in-bits='896'>
	+ <var-decl name='__ctype_tolower' type-id='6d60f45d' visibility='default'/>
	+ </data-member>
	+ <data-member access='public' layout-offset-in-bits='960'>
	+ <var-decl name='__ctype_toupper' type-id='6d60f45d' visibility='default'/>
	+ </data-member>
	+ <data-member access='public' layout-offset-in-bits='1024'>
	+ <var-decl name='__names' type-id='39e6f84a' visibility='default'/>
	+ </data-member>
	+ </class-decl>
	+ <typedef-decl name='__locale_t' type-id='f01e1813' id='b7ac9b5f'/>
	+ <typedef-decl name='locale_t' type-id='b7ac9b5f' id='973a4f8d'/>
	+ <pointer-type-def type-id='23de8b96' size-in-bits='64' id='38b51b3c'/>
	+ <pointer-type-def type-id='90cc1ce3' size-in-bits='64' id='f01e1813'/>
	+ <qualified-type-def type-id='95e97e5e' const='yes' id='2448a865'/>
	+ <pointer-type-def type-id='2448a865' size-in-bits='64' id='6d60f45d'/>
	+ <qualified-type-def type-id='8efea9e5' const='yes' id='3beb2af4'/>
	+ <pointer-type-def type-id='3beb2af4' size-in-bits='64' id='31347b7a'/>
	<pointer-type-def type-id='0c544dc0' size-in-bits='64' id='394fc496'/>
	<pointer-type-def type-id='56fe4a37' size-in-bits='64' id='b6b61d2f'/>
	<qualified-type-def type-id='b6b61d2f' restrict='yes' id='3cad23cd'/>
	<pointer-type-def type-id='1b055409' size-in-bits='64' id='9d424d31'/>
	<pointer-type-def type-id='0bbec9cd' size-in-bits='64' id='62f7a03d'/>
	<qualified-type-def type-id='62f7a03d' restrict='yes' id='f1cadedf'/>
	+ <class-decl name='__locale_data' is-struct='yes' visibility='default' is-declaration-only='yes' id='23de8b96'/>
	+ <function-decl name='uselocale' visibility='default' binding='global' size-in-bits='64'>
	+ <parameter type-id='973a4f8d'/>
	+ <return type-id='973a4f8d'/>
	+ </function-decl>
	<function-decl name='getmntent_r' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='e75a27e9'/>
	<parameter type-id='3cad23cd'/>
	<parameter type-id='266fe297'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='b6b61d2f'/>
	</function-decl>
	<function-decl name='feof' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='822cd80b'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='strcmp' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	- <function-decl name='strerror' visibility='default' binding='global' size-in-bits='64'>
	+ <function-decl name='strerror_l' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	+ <parameter type-id='973a4f8d'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='__fprintf_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='e75a27e9'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='9d26089a'/>
	<parameter is-variadic='yes'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='stat64' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9d26089a'/>
	<parameter type-id='f1cadedf'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='getmntany' mangled-name='getmntany' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='getmntany'>
	<parameter type-id='822cd80b' name='fp'/>
	<parameter type-id='9d424d31' name='mgetp'/>
	<parameter type-id='9d424d31' name='mrefp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='_sol_getmntent' mangled-name='_sol_getmntent' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='_sol_getmntent'>
	<parameter type-id='822cd80b' name='fp'/>
	<parameter type-id='9d424d31' name='mgetp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='getextmntent' mangled-name='getextmntent' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='getextmntent'>
	<parameter type-id='80f4b756' name='path'/>
	<parameter type-id='394fc496' name='entry'/>
	<parameter type-id='62f7a03d' name='statbuf'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/os/linux/zone.c' language='LANG_C99'>
	<typedef-decl name='zoneid_t' type-id='3502e3ff' id='4da03624'/>
	<function-decl name='strtoul' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='9d26089a'/>
	<parameter type-id='8c85230f'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='7359adad'/>
	</function-decl>
	<function-decl name='getzoneid' mangled-name='getzoneid' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='getzoneid'>
	<return type-id='4da03624'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/page.c' language='LANG_C99'>
	<function-decl name='sysconf' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<return type-id='bd54fe1a'/>
	</function-decl>
	<function-decl name='spl_pagesize' mangled-name='spl_pagesize' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='spl_pagesize'>
	<return type-id='b59d7dce'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/strlcat.c' language='LANG_C99'>
	<function-decl name='strlcat' mangled-name='strlcat' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='strlcat'>
	<parameter type-id='26a90f95' name='dst'/>
	<parameter type-id='80f4b756' name='src'/>
	<parameter type-id='b59d7dce' name='dstsize'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libspl/timestamp.c' language='LANG_C99'>
	<typedef-decl name='nl_item' type-id='95e97e5e' id='03b79a94'/>
	<class-decl name='tm' size-in-bits='448' is-struct='yes' visibility='default' id='dddf6ca2'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='tm_sec' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='tm_min' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='tm_hour' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='tm_mday' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='tm_mon' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='160'>
	<var-decl name='tm_year' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='tm_wday' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='224'>
	<var-decl name='tm_yday' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='tm_isdst' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='tm_gmtoff' type-id='bd54fe1a' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='tm_zone' type-id='80f4b756' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='time_t' type-id='65eda9c0' id='c9d12d66'/>
	<qualified-type-def type-id='c9d12d66' const='yes' id='588b3216'/>
	<pointer-type-def type-id='588b3216' size-in-bits='64' id='9f201474'/>
	<qualified-type-def type-id='9f201474' restrict='yes' id='d6e2847c'/>
	<qualified-type-def type-id='dddf6ca2' const='yes' id='e824a34f'/>
	<pointer-type-def type-id='e824a34f' size-in-bits='64' id='d6ad37ff'/>
	<qualified-type-def type-id='d6ad37ff' restrict='yes' id='f8c6051d'/>
	<pointer-type-def type-id='c9d12d66' size-in-bits='64' id='b2eb2c3f'/>
	<pointer-type-def type-id='dddf6ca2' size-in-bits='64' id='d915a820'/>
	<qualified-type-def type-id='d915a820' restrict='yes' id='f099ad08'/>
	<function-decl name='nl_langinfo' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='03b79a94'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='time' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b2eb2c3f'/>
	<return type-id='c9d12d66'/>
	</function-decl>
	<function-decl name='strftime' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='266fe297'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='9d26089a'/>
	<parameter type-id='f8c6051d'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='localtime_r' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='d6e2847c'/>
	<parameter type-id='f099ad08'/>
	<return type-id='d915a820'/>
	</function-decl>
	<function-decl name='__printf_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<parameter is-variadic='yes'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='print_timestamp' mangled-name='print_timestamp' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='print_timestamp'>
	<parameter type-id='3502e3ff' name='timestamp_fmt'/>
	<return type-id='48b5725f'/>
	</function-decl>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs_core/libzfs_core.c' language='LANG_C99'>
	<array-type-def dimensions='1' type-id='03085adc' size-in-bits='192' id='083f8d58'>
	<subrange length='3' type-id='7359adad' id='56f209d2'/>
	</array-type-def>
	<type-decl name='char' size-in-bits='8' id='a84c031d'/>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='8' id='89feb1ec'>
	<subrange length='1' type-id='7359adad' id='52f813b4'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='160' id='664ac0b7'>
	<subrange length='20' type-id='7359adad' id='fdca39cf'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='2048' id='d1617432'>
	<subrange length='256' type-id='7359adad' id='36e5b9fa'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='32768' id='d16c6df4'>
	<subrange length='4096' type-id='7359adad' id='bc1b5ddc'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='320' id='36c46961'>
	<subrange length='40' type-id='7359adad' id='8f80b239'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='448' id='6093ff7c'>
	<subrange length='56' type-id='7359adad' id='f8137894'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='65536' id='163f6aa5'>
	<subrange length='8192' type-id='7359adad' id='c88f397d'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='a84c031d' size-in-bits='infinite' id='e84913bd'>
	<subrange length='infinite' type-id='7359adad' id='031f2035'/>
	</array-type-def>
	<class-decl name='_IO_codecvt' is-struct='yes' visibility='default' is-declaration-only='yes' id='a4036571'/>
	<class-decl name='_IO_marker' is-struct='yes' visibility='default' is-declaration-only='yes' id='010ae0b9'/>
	<class-decl name='_IO_wide_data' is-struct='yes' visibility='default' is-declaration-only='yes' id='79bd3751'/>
	<type-decl name='int' size-in-bits='32' id='95e97e5e'/>
	<type-decl name='long int' size-in-bits='64' id='bd54fe1a'/>
	<type-decl name='short int' size-in-bits='16' id='a2185560'/>
	<type-decl name='signed char' size-in-bits='8' id='28577a57'/>
	<array-type-def dimensions='1' type-id='9c313c2d' size-in-bits='128' id='c1c22e6c'>
	<subrange length='2' type-id='7359adad' id='52efc4ef'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='9c313c2d' size-in-bits='2176' id='8c2bcad1'>
	<subrange length='34' type-id='7359adad' id='6a6a7e00'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='9c313c2d' size-in-bits='256' id='85c64d26'>
	<subrange length='4' type-id='7359adad' id='16fe7105'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='b96825af' size-in-bits='96' id='fa8ef949'>
	<subrange length='12' type-id='7359adad' id='84827bdc'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='b96825af' size-in-bits='128' id='fa9986a5'>
	<subrange length='16' type-id='7359adad' id='848d0938'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='b96825af' size-in-bits='24' id='d3490169'>
	<subrange length='3' type-id='7359adad' id='56f209d2'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='b96825af' size-in-bits='40' id='0f4ddd0b'>
	<subrange length='5' type-id='7359adad' id='53010e10'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='b96825af' size-in-bits='48' id='0f562bd0'>
	<subrange length='6' type-id='7359adad' id='52fa524b'/>
	</array-type-def>
	<array-type-def dimensions='1' type-id='b96825af' size-in-bits='64' id='13339fda'>
	<subrange length='8' type-id='7359adad' id='56e0c0b1'/>
	</array-type-def>
	<type-decl name='unnamed-enum-underlying-type-32' is-anonymous='yes' size-in-bits='32' alignment-in-bits='32' id='9cac1fee'/>
	<type-decl name='unsigned char' size-in-bits='8' id='002ac4a6'/>
	<type-decl name='unsigned int' size-in-bits='32' id='f0981eeb'/>
	<type-decl name='unsigned long int' size-in-bits='64' id='7359adad'/>
	<type-decl name='unsigned short int' size-in-bits='16' id='8efea9e5'/>
	<type-decl name='variadic parameter type' id='2c1145c5'/>
	<type-decl name='void' id='48b5725f'/>
	<enum-decl name='lzc_dataset_type' id='bc9887f1'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='LZC_DATSET_TYPE_ZFS' value='2'/>
	<enumerator name='LZC_DATSET_TYPE_ZVOL' value='3'/>
	</enum-decl>
	<enum-decl name='lzc_send_flags' id='bfbd3c8e'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='LZC_SEND_FLAG_EMBED_DATA' value='1'/>
	<enumerator name='LZC_SEND_FLAG_LARGE_BLOCK' value='2'/>
	<enumerator name='LZC_SEND_FLAG_COMPRESS' value='4'/>
	<enumerator name='LZC_SEND_FLAG_RAW' value='8'/>
	<enumerator name='LZC_SEND_FLAG_SAVED' value='16'/>
	</enum-decl>
	- <class-decl name='ddt_key' size-in-bits='320' is-struct='yes' visibility='default' id='e0a4a1cb'>
	+ <class-decl name='ddt_key_t' size-in-bits='320' is-struct='yes' naming-typedef-id='67f6d2cf' visibility='default' id='5fae1718'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='ddk_cksum' type-id='39730d0b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='ddk_prop' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	- <typedef-decl name='ddt_key_t' type-id='e0a4a1cb' id='67f6d2cf'/>
	+ <typedef-decl name='ddt_key_t' type-id='5fae1718' id='67f6d2cf'/>
	<enum-decl name='dmu_object_type' id='04b3b0b9'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='DMU_OT_NONE' value='0'/>
	<enumerator name='DMU_OT_OBJECT_DIRECTORY' value='1'/>
	<enumerator name='DMU_OT_OBJECT_ARRAY' value='2'/>
	<enumerator name='DMU_OT_PACKED_NVLIST' value='3'/>
	<enumerator name='DMU_OT_PACKED_NVLIST_SIZE' value='4'/>
	<enumerator name='DMU_OT_BPOBJ' value='5'/>
	<enumerator name='DMU_OT_BPOBJ_HDR' value='6'/>
	<enumerator name='DMU_OT_SPACE_MAP_HEADER' value='7'/>
	<enumerator name='DMU_OT_SPACE_MAP' value='8'/>
	<enumerator name='DMU_OT_INTENT_LOG' value='9'/>
	<enumerator name='DMU_OT_DNODE' value='10'/>
	<enumerator name='DMU_OT_OBJSET' value='11'/>
	<enumerator name='DMU_OT_DSL_DIR' value='12'/>
	<enumerator name='DMU_OT_DSL_DIR_CHILD_MAP' value='13'/>
	<enumerator name='DMU_OT_DSL_DS_SNAP_MAP' value='14'/>
	<enumerator name='DMU_OT_DSL_PROPS' value='15'/>
	<enumerator name='DMU_OT_DSL_DATASET' value='16'/>
	<enumerator name='DMU_OT_ZNODE' value='17'/>
	<enumerator name='DMU_OT_OLDACL' value='18'/>
	<enumerator name='DMU_OT_PLAIN_FILE_CONTENTS' value='19'/>
	<enumerator name='DMU_OT_DIRECTORY_CONTENTS' value='20'/>
	<enumerator name='DMU_OT_MASTER_NODE' value='21'/>
	<enumerator name='DMU_OT_UNLINKED_SET' value='22'/>
	<enumerator name='DMU_OT_ZVOL' value='23'/>
	<enumerator name='DMU_OT_ZVOL_PROP' value='24'/>
	<enumerator name='DMU_OT_PLAIN_OTHER' value='25'/>
	<enumerator name='DMU_OT_UINT64_OTHER' value='26'/>
	<enumerator name='DMU_OT_ZAP_OTHER' value='27'/>
	<enumerator name='DMU_OT_ERROR_LOG' value='28'/>
	<enumerator name='DMU_OT_SPA_HISTORY' value='29'/>
	<enumerator name='DMU_OT_SPA_HISTORY_OFFSETS' value='30'/>
	<enumerator name='DMU_OT_POOL_PROPS' value='31'/>
	<enumerator name='DMU_OT_DSL_PERMS' value='32'/>
	<enumerator name='DMU_OT_ACL' value='33'/>
	<enumerator name='DMU_OT_SYSACL' value='34'/>
	<enumerator name='DMU_OT_FUID' value='35'/>
	<enumerator name='DMU_OT_FUID_SIZE' value='36'/>
	<enumerator name='DMU_OT_NEXT_CLONES' value='37'/>
	<enumerator name='DMU_OT_SCAN_QUEUE' value='38'/>
	<enumerator name='DMU_OT_USERGROUP_USED' value='39'/>
	<enumerator name='DMU_OT_USERGROUP_QUOTA' value='40'/>
	<enumerator name='DMU_OT_USERREFS' value='41'/>
	<enumerator name='DMU_OT_DDT_ZAP' value='42'/>
	<enumerator name='DMU_OT_DDT_STATS' value='43'/>
	<enumerator name='DMU_OT_SA' value='44'/>
	<enumerator name='DMU_OT_SA_MASTER_NODE' value='45'/>
	<enumerator name='DMU_OT_SA_ATTR_REGISTRATION' value='46'/>
	<enumerator name='DMU_OT_SA_ATTR_LAYOUTS' value='47'/>
	<enumerator name='DMU_OT_SCAN_XLATE' value='48'/>
	<enumerator name='DMU_OT_DEDUP' value='49'/>
	<enumerator name='DMU_OT_DEADLIST' value='50'/>
	<enumerator name='DMU_OT_DEADLIST_HDR' value='51'/>
	<enumerator name='DMU_OT_DSL_CLONES' value='52'/>
	<enumerator name='DMU_OT_BPOBJ_SUBOBJ' value='53'/>
	<enumerator name='DMU_OT_NUMTYPES' value='54'/>
	<enumerator name='DMU_OTN_UINT8_DATA' value='128'/>
	<enumerator name='DMU_OTN_UINT8_METADATA' value='192'/>
	<enumerator name='DMU_OTN_UINT16_DATA' value='129'/>
	<enumerator name='DMU_OTN_UINT16_METADATA' value='193'/>
	<enumerator name='DMU_OTN_UINT32_DATA' value='130'/>
	<enumerator name='DMU_OTN_UINT32_METADATA' value='194'/>
	<enumerator name='DMU_OTN_UINT64_DATA' value='131'/>
	<enumerator name='DMU_OTN_UINT64_METADATA' value='195'/>
	<enumerator name='DMU_OTN_ZAP_DATA' value='132'/>
	<enumerator name='DMU_OTN_ZAP_METADATA' value='196'/>
	<enumerator name='DMU_OTN_UINT8_ENC_DATA' value='160'/>
	<enumerator name='DMU_OTN_UINT8_ENC_METADATA' value='224'/>
	<enumerator name='DMU_OTN_UINT16_ENC_DATA' value='161'/>
	<enumerator name='DMU_OTN_UINT16_ENC_METADATA' value='225'/>
	<enumerator name='DMU_OTN_UINT32_ENC_DATA' value='162'/>
	<enumerator name='DMU_OTN_UINT32_ENC_METADATA' value='226'/>
	<enumerator name='DMU_OTN_UINT64_ENC_DATA' value='163'/>
	<enumerator name='DMU_OTN_UINT64_ENC_METADATA' value='227'/>
	<enumerator name='DMU_OTN_ZAP_ENC_DATA' value='164'/>
	<enumerator name='DMU_OTN_ZAP_ENC_METADATA' value='228'/>
	</enum-decl>
	<typedef-decl name='dmu_object_type_t' type-id='04b3b0b9' id='5c9d8906'/>
	<class-decl name='dmu_objset_stats' size-in-bits='2304' is-struct='yes' visibility='default' id='098f0221'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='dds_num_clones' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='dds_creation_txg' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='dds_guid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='dds_type' type-id='230f1e16' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='224'>
	<var-decl name='dds_is_snapshot' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='232'>
	<var-decl name='dds_inconsistent' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='240'>
	<var-decl name='dds_redacted' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='248'>
	<var-decl name='dds_origin' type-id='d1617432' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='dmu_objset_stats_t' type-id='098f0221' id='b2c14f17'/>
	<enum-decl name='dmu_objset_type' id='6b1b19f9'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='DMU_OST_NONE' value='0'/>
	<enumerator name='DMU_OST_META' value='1'/>
	<enumerator name='DMU_OST_ZFS' value='2'/>
	<enumerator name='DMU_OST_ZVOL' value='3'/>
	<enumerator name='DMU_OST_OTHER' value='4'/>
	<enumerator name='DMU_OST_ANY' value='5'/>
	<enumerator name='DMU_OST_NUMTYPES' value='6'/>
	</enum-decl>
	<typedef-decl name='dmu_objset_type_t' type-id='6b1b19f9' id='230f1e16'/>
	<enum-decl name='pool_initialize_func' id='5c246ad4'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='POOL_INITIALIZE_START' value='0'/>
	<enumerator name='POOL_INITIALIZE_CANCEL' value='1'/>
	<enumerator name='POOL_INITIALIZE_SUSPEND' value='2'/>
	<enumerator name='POOL_INITIALIZE_UNINIT' value='3'/>
	<enumerator name='POOL_INITIALIZE_FUNCS' value='4'/>
	</enum-decl>
	<typedef-decl name='pool_initialize_func_t' type-id='5c246ad4' id='7063e1ab'/>
	<enum-decl name='pool_trim_func' id='54ed608a'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='POOL_TRIM_START' value='0'/>
	<enumerator name='POOL_TRIM_CANCEL' value='1'/>
	<enumerator name='POOL_TRIM_SUSPEND' value='2'/>
	<enumerator name='POOL_TRIM_FUNCS' value='3'/>
	</enum-decl>
	<typedef-decl name='pool_trim_func_t' type-id='54ed608a' id='b1146b8d'/>
	<enum-decl name='zfs_ioc' id='12033f13'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZFS_IOC_FIRST' value='23040'/>
	<enumerator name='ZFS_IOC' value='23040'/>
	<enumerator name='ZFS_IOC_POOL_CREATE' value='23040'/>
	<enumerator name='ZFS_IOC_POOL_DESTROY' value='23041'/>
	<enumerator name='ZFS_IOC_POOL_IMPORT' value='23042'/>
	<enumerator name='ZFS_IOC_POOL_EXPORT' value='23043'/>
	<enumerator name='ZFS_IOC_POOL_CONFIGS' value='23044'/>
	<enumerator name='ZFS_IOC_POOL_STATS' value='23045'/>
	<enumerator name='ZFS_IOC_POOL_TRYIMPORT' value='23046'/>
	<enumerator name='ZFS_IOC_POOL_SCAN' value='23047'/>
	<enumerator name='ZFS_IOC_POOL_FREEZE' value='23048'/>
	<enumerator name='ZFS_IOC_POOL_UPGRADE' value='23049'/>
	<enumerator name='ZFS_IOC_POOL_GET_HISTORY' value='23050'/>
	<enumerator name='ZFS_IOC_VDEV_ADD' value='23051'/>
	<enumerator name='ZFS_IOC_VDEV_REMOVE' value='23052'/>
	<enumerator name='ZFS_IOC_VDEV_SET_STATE' value='23053'/>
	<enumerator name='ZFS_IOC_VDEV_ATTACH' value='23054'/>
	<enumerator name='ZFS_IOC_VDEV_DETACH' value='23055'/>
	<enumerator name='ZFS_IOC_VDEV_SETPATH' value='23056'/>
	<enumerator name='ZFS_IOC_VDEV_SETFRU' value='23057'/>
	<enumerator name='ZFS_IOC_OBJSET_STATS' value='23058'/>
	<enumerator name='ZFS_IOC_OBJSET_ZPLPROPS' value='23059'/>
	<enumerator name='ZFS_IOC_DATASET_LIST_NEXT' value='23060'/>
	<enumerator name='ZFS_IOC_SNAPSHOT_LIST_NEXT' value='23061'/>
	<enumerator name='ZFS_IOC_SET_PROP' value='23062'/>
	<enumerator name='ZFS_IOC_CREATE' value='23063'/>
	<enumerator name='ZFS_IOC_DESTROY' value='23064'/>
	<enumerator name='ZFS_IOC_ROLLBACK' value='23065'/>
	<enumerator name='ZFS_IOC_RENAME' value='23066'/>
	<enumerator name='ZFS_IOC_RECV' value='23067'/>
	<enumerator name='ZFS_IOC_SEND' value='23068'/>
	<enumerator name='ZFS_IOC_INJECT_FAULT' value='23069'/>
	<enumerator name='ZFS_IOC_CLEAR_FAULT' value='23070'/>
	<enumerator name='ZFS_IOC_INJECT_LIST_NEXT' value='23071'/>
	<enumerator name='ZFS_IOC_ERROR_LOG' value='23072'/>
	<enumerator name='ZFS_IOC_CLEAR' value='23073'/>
	<enumerator name='ZFS_IOC_PROMOTE' value='23074'/>
	<enumerator name='ZFS_IOC_SNAPSHOT' value='23075'/>
	<enumerator name='ZFS_IOC_DSOBJ_TO_DSNAME' value='23076'/>
	<enumerator name='ZFS_IOC_OBJ_TO_PATH' value='23077'/>
	<enumerator name='ZFS_IOC_POOL_SET_PROPS' value='23078'/>
	<enumerator name='ZFS_IOC_POOL_GET_PROPS' value='23079'/>
	<enumerator name='ZFS_IOC_SET_FSACL' value='23080'/>
	<enumerator name='ZFS_IOC_GET_FSACL' value='23081'/>
	<enumerator name='ZFS_IOC_SHARE' value='23082'/>
	<enumerator name='ZFS_IOC_INHERIT_PROP' value='23083'/>
	<enumerator name='ZFS_IOC_SMB_ACL' value='23084'/>
	<enumerator name='ZFS_IOC_USERSPACE_ONE' value='23085'/>
	<enumerator name='ZFS_IOC_USERSPACE_MANY' value='23086'/>
	<enumerator name='ZFS_IOC_USERSPACE_UPGRADE' value='23087'/>
	<enumerator name='ZFS_IOC_HOLD' value='23088'/>
	<enumerator name='ZFS_IOC_RELEASE' value='23089'/>
	<enumerator name='ZFS_IOC_GET_HOLDS' value='23090'/>
	<enumerator name='ZFS_IOC_OBJSET_RECVD_PROPS' value='23091'/>
	<enumerator name='ZFS_IOC_VDEV_SPLIT' value='23092'/>
	<enumerator name='ZFS_IOC_NEXT_OBJ' value='23093'/>
	<enumerator name='ZFS_IOC_DIFF' value='23094'/>
	<enumerator name='ZFS_IOC_TMP_SNAPSHOT' value='23095'/>
	<enumerator name='ZFS_IOC_OBJ_TO_STATS' value='23096'/>
	<enumerator name='ZFS_IOC_SPACE_WRITTEN' value='23097'/>
	<enumerator name='ZFS_IOC_SPACE_SNAPS' value='23098'/>
	<enumerator name='ZFS_IOC_DESTROY_SNAPS' value='23099'/>
	<enumerator name='ZFS_IOC_POOL_REGUID' value='23100'/>
	<enumerator name='ZFS_IOC_POOL_REOPEN' value='23101'/>
	<enumerator name='ZFS_IOC_SEND_PROGRESS' value='23102'/>
	<enumerator name='ZFS_IOC_LOG_HISTORY' value='23103'/>
	<enumerator name='ZFS_IOC_SEND_NEW' value='23104'/>
	<enumerator name='ZFS_IOC_SEND_SPACE' value='23105'/>
	<enumerator name='ZFS_IOC_CLONE' value='23106'/>
	<enumerator name='ZFS_IOC_BOOKMARK' value='23107'/>
	<enumerator name='ZFS_IOC_GET_BOOKMARKS' value='23108'/>
	<enumerator name='ZFS_IOC_DESTROY_BOOKMARKS' value='23109'/>
	<enumerator name='ZFS_IOC_RECV_NEW' value='23110'/>
	<enumerator name='ZFS_IOC_POOL_SYNC' value='23111'/>
	<enumerator name='ZFS_IOC_CHANNEL_PROGRAM' value='23112'/>
	<enumerator name='ZFS_IOC_LOAD_KEY' value='23113'/>
	<enumerator name='ZFS_IOC_UNLOAD_KEY' value='23114'/>
	<enumerator name='ZFS_IOC_CHANGE_KEY' value='23115'/>
	<enumerator name='ZFS_IOC_REMAP' value='23116'/>
	<enumerator name='ZFS_IOC_POOL_CHECKPOINT' value='23117'/>
	<enumerator name='ZFS_IOC_POOL_DISCARD_CHECKPOINT' value='23118'/>
	<enumerator name='ZFS_IOC_POOL_INITIALIZE' value='23119'/>
	<enumerator name='ZFS_IOC_POOL_TRIM' value='23120'/>
	<enumerator name='ZFS_IOC_REDACT' value='23121'/>
	<enumerator name='ZFS_IOC_GET_BOOKMARK_PROPS' value='23122'/>
	<enumerator name='ZFS_IOC_WAIT' value='23123'/>
	<enumerator name='ZFS_IOC_WAIT_FS' value='23124'/>
	<enumerator name='ZFS_IOC_VDEV_GET_PROPS' value='23125'/>
	<enumerator name='ZFS_IOC_VDEV_SET_PROPS' value='23126'/>
	<enumerator name='ZFS_IOC_POOL_SCRUB' value='23127'/>
	<enumerator name='ZFS_IOC_PLATFORM' value='23168'/>
	<enumerator name='ZFS_IOC_EVENTS_NEXT' value='23169'/>
	<enumerator name='ZFS_IOC_EVENTS_CLEAR' value='23170'/>
	<enumerator name='ZFS_IOC_EVENTS_SEEK' value='23171'/>
	<enumerator name='ZFS_IOC_NEXTBOOT' value='23172'/>
	<enumerator name='ZFS_IOC_JAIL' value='23173'/>
	<enumerator name='ZFS_IOC_USERNS_ATTACH' value='23173'/>
	<enumerator name='ZFS_IOC_UNJAIL' value='23174'/>
	<enumerator name='ZFS_IOC_USERNS_DETACH' value='23174'/>
	<enumerator name='ZFS_IOC_SET_BOOTENV' value='23175'/>
	<enumerator name='ZFS_IOC_GET_BOOTENV' value='23176'/>
	<enumerator name='ZFS_IOC_LAST' value='23177'/>
	</enum-decl>
	<typedef-decl name='zfs_ioc_t' type-id='12033f13' id='5b35941c'/>
	<enum-decl name='zpool_wait_activity_t' naming-typedef-id='73446457' id='849338e3'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZPOOL_WAIT_CKPT_DISCARD' value='0'/>
	<enumerator name='ZPOOL_WAIT_FREE' value='1'/>
	<enumerator name='ZPOOL_WAIT_INITIALIZE' value='2'/>
	<enumerator name='ZPOOL_WAIT_REPLACE' value='3'/>
	<enumerator name='ZPOOL_WAIT_REMOVE' value='4'/>
	<enumerator name='ZPOOL_WAIT_RESILVER' value='5'/>
	<enumerator name='ZPOOL_WAIT_SCRUB' value='6'/>
	<enumerator name='ZPOOL_WAIT_TRIM' value='7'/>
	<enumerator name='ZPOOL_WAIT_RAIDZ_EXPAND' value='8'/>
	<enumerator name='ZPOOL_WAIT_NUM_ACTIVITIES' value='9'/>
	</enum-decl>
	<typedef-decl name='zpool_wait_activity_t' type-id='849338e3' id='73446457'/>
	<enum-decl name='zfs_wait_activity_t' naming-typedef-id='3024501a' id='527d5dc6'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='ZFS_WAIT_DELETEQ' value='0'/>
	<enumerator name='ZFS_WAIT_NUM_ACTIVITIES' value='1'/>
	</enum-decl>
	<typedef-decl name='zfs_wait_activity_t' type-id='527d5dc6' id='3024501a'/>
	<enum-decl name='data_type_t' naming-typedef-id='8d0687d2' id='aeeae136'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='DATA_TYPE_DONTCARE' value='-1'/>
	<enumerator name='DATA_TYPE_UNKNOWN' value='0'/>
	<enumerator name='DATA_TYPE_BOOLEAN' value='1'/>
	<enumerator name='DATA_TYPE_BYTE' value='2'/>
	<enumerator name='DATA_TYPE_INT16' value='3'/>
	<enumerator name='DATA_TYPE_UINT16' value='4'/>
	<enumerator name='DATA_TYPE_INT32' value='5'/>
	<enumerator name='DATA_TYPE_UINT32' value='6'/>
	<enumerator name='DATA_TYPE_INT64' value='7'/>
	<enumerator name='DATA_TYPE_UINT64' value='8'/>
	<enumerator name='DATA_TYPE_STRING' value='9'/>
	<enumerator name='DATA_TYPE_BYTE_ARRAY' value='10'/>
	<enumerator name='DATA_TYPE_INT16_ARRAY' value='11'/>
	<enumerator name='DATA_TYPE_UINT16_ARRAY' value='12'/>
	<enumerator name='DATA_TYPE_INT32_ARRAY' value='13'/>
	<enumerator name='DATA_TYPE_UINT32_ARRAY' value='14'/>
	<enumerator name='DATA_TYPE_INT64_ARRAY' value='15'/>
	<enumerator name='DATA_TYPE_UINT64_ARRAY' value='16'/>
	<enumerator name='DATA_TYPE_STRING_ARRAY' value='17'/>
	<enumerator name='DATA_TYPE_HRTIME' value='18'/>
	<enumerator name='DATA_TYPE_NVLIST' value='19'/>
	<enumerator name='DATA_TYPE_NVLIST_ARRAY' value='20'/>
	<enumerator name='DATA_TYPE_BOOLEAN_VALUE' value='21'/>
	<enumerator name='DATA_TYPE_INT8' value='22'/>
	<enumerator name='DATA_TYPE_UINT8' value='23'/>
	<enumerator name='DATA_TYPE_BOOLEAN_ARRAY' value='24'/>
	<enumerator name='DATA_TYPE_INT8_ARRAY' value='25'/>
	<enumerator name='DATA_TYPE_UINT8_ARRAY' value='26'/>
	<enumerator name='DATA_TYPE_DOUBLE' value='27'/>
	</enum-decl>
	<typedef-decl name='data_type_t' type-id='aeeae136' id='8d0687d2'/>
	<class-decl name='nvpair' size-in-bits='128' is-struct='yes' visibility='default' id='1c34e459'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='nvp_size' type-id='3ff5601b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='nvp_name_sz' type-id='23bd8cb5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='48'>
	<var-decl name='nvp_reserve' type-id='23bd8cb5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='nvp_value_elem' type-id='3ff5601b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='nvp_type' type-id='8d0687d2' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='nvp_name' type-id='e84913bd' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='nvpair_t' type-id='1c34e459' id='57928edf'/>
	<class-decl name='nvlist' size-in-bits='192' is-struct='yes' visibility='default' id='ac266fd9'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='nvl_version' type-id='3ff5601b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='nvl_nvflag' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='nvl_priv' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='nvl_flag' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='160'>
	<var-decl name='nvl_pad' type-id='3ff5601b' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='nvlist_t' type-id='ac266fd9' id='8e8d4be3'/>
	<class-decl name='zio_cksum' size-in-bits='256' is-struct='yes' visibility='default' id='1d53e28b'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='zc_word' type-id='85c64d26' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zio_cksum_t' type-id='1d53e28b' id='39730d0b'/>
	<class-decl name='drr_begin' size-in-bits='2432' is-struct='yes' visibility='default' id='09fcdc01'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_magic' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='drr_versioninfo' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='drr_creation_time' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='drr_type' type-id='230f1e16' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='224'>
	<var-decl name='drr_flags' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='drr_fromguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='drr_toname' type-id='d1617432' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='dmu_replay_record' size-in-bits='2496' is-struct='yes' visibility='default' id='781a52d7'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_type' type-id='08f5ca17' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='drr_payloadlen' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='drr_u' type-id='ac5ab595' visibility='default'/>
	</data-member>
	</class-decl>
	<enum-decl name='__anonymous_enum__' is-anonymous='yes' id='08f5ca17'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='DRR_BEGIN' value='0'/>
	<enumerator name='DRR_OBJECT' value='1'/>
	<enumerator name='DRR_FREEOBJECTS' value='2'/>
	<enumerator name='DRR_WRITE' value='3'/>
	<enumerator name='DRR_FREE' value='4'/>
	<enumerator name='DRR_END' value='5'/>
	<enumerator name='DRR_WRITE_BYREF' value='6'/>
	<enumerator name='DRR_SPILL' value='7'/>
	<enumerator name='DRR_WRITE_EMBEDDED' value='8'/>
	<enumerator name='DRR_OBJECT_RANGE' value='9'/>
	<enumerator name='DRR_REDACT' value='10'/>
	<enumerator name='DRR_NUMTYPES' value='11'/>
	</enum-decl>
	<union-decl name='__anonymous_union__' size-in-bits='2432' is-anonymous='yes' visibility='default' id='ac5ab595'>
	<data-member access='public'>
	<var-decl name='drr_begin' type-id='09fcdc01' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='drr_end' type-id='6ee25631' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='drr_object' type-id='f9ad530b' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='drr_freeobjects' type-id='a27d958e' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='drr_write' type-id='4cc69e4b' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='drr_free' type-id='c836cfd2' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='drr_write_byref' type-id='e511cdce' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='drr_spill' type-id='1e69a80a' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='drr_write_embedded' type-id='98b1345e' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='drr_object_range' type-id='aba1f9e1' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='drr_redact' type-id='50389039' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='drr_checksum' type-id='a5fe3647' visibility='default'/>
	</data-member>
	</union-decl>
	<class-decl name='drr_end' size-in-bits='320' is-struct='yes' visibility='default' id='6ee25631'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_checksum' type-id='39730d0b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='drr_object' size-in-bits='448' is-struct='yes' visibility='default' id='f9ad530b'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_object' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='drr_type' type-id='5c9d8906' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='drr_bonustype' type-id='5c9d8906' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='drr_blksz' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='160'>
	<var-decl name='drr_bonuslen' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='drr_checksumtype' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='200'>
	<var-decl name='drr_compress' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='208'>
	<var-decl name='drr_dn_slots' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='216'>
	<var-decl name='drr_flags' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='224'>
	<var-decl name='drr_raw_bonuslen' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='drr_indblkshift' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='328'>
	<var-decl name='drr_nlevels' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='336'>
	<var-decl name='drr_nblkptr' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='344'>
	<var-decl name='drr_pad' type-id='0f4ddd0b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='drr_maxblkid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='drr_freeobjects' size-in-bits='192' is-struct='yes' visibility='default' id='a27d958e'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_firstobj' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='drr_numobjs' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='drr_write' size-in-bits='1088' is-struct='yes' visibility='default' id='4cc69e4b'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_object' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='drr_type' type-id='5c9d8906' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='drr_pad' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='drr_offset' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='drr_logical_size' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='drr_checksumtype' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='328'>
	<var-decl name='drr_flags' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='336'>
	<var-decl name='drr_compressiontype' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='344'>
	<var-decl name='drr_pad2' type-id='0f4ddd0b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='drr_key' type-id='67f6d2cf' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='704'>
	<var-decl name='drr_compressed_size' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='768'>
	<var-decl name='drr_salt' type-id='13339fda' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='832'>
	<var-decl name='drr_iv' type-id='fa8ef949' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='928'>
	<var-decl name='drr_mac' type-id='fa9986a5' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='drr_free' size-in-bits='256' is-struct='yes' visibility='default' id='c836cfd2'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_object' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='drr_offset' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='drr_length' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='drr_write_byref' size-in-bits='832' is-struct='yes' visibility='default' id='e511cdce'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_object' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='drr_offset' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='drr_length' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='drr_refguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='drr_refobject' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='drr_refoffset' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='drr_checksumtype' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='456'>
	<var-decl name='drr_flags' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='464'>
	<var-decl name='drr_pad2' type-id='0f562bd0' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='drr_key' type-id='67f6d2cf' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='drr_spill' size-in-bits='640' is-struct='yes' visibility='default' id='1e69a80a'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_object' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='drr_length' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='drr_flags' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='200'>
	<var-decl name='drr_compressiontype' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='208'>
	<var-decl name='drr_pad' type-id='0f562bd0' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='drr_compressed_size' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='drr_salt' type-id='13339fda' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='drr_iv' type-id='fa8ef949' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='480'>
	<var-decl name='drr_mac' type-id='fa9986a5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='608'>
	<var-decl name='drr_type' type-id='5c9d8906' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='drr_write_embedded' size-in-bits='384' is-struct='yes' visibility='default' id='98b1345e'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_object' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='drr_offset' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='drr_length' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='drr_compression' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='264'>
	<var-decl name='drr_etype' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='272'>
	<var-decl name='drr_pad' type-id='0f562bd0' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='drr_lsize' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='352'>
	<var-decl name='drr_psize' type-id='8f92235e' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='drr_object_range' size-in-bits='512' is-struct='yes' visibility='default' id='aba1f9e1'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_firstobj' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='drr_numslots' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='drr_salt' type-id='13339fda' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='drr_iv' type-id='fa8ef949' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='352'>
	<var-decl name='drr_mac' type-id='fa9986a5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='480'>
	<var-decl name='drr_flags' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='488'>
	<var-decl name='drr_pad' type-id='d3490169' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='drr_redact' size-in-bits='256' is-struct='yes' visibility='default' id='50389039'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_object' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='drr_offset' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='drr_length' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='drr_toguid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='drr_checksum' size-in-bits='2432' is-struct='yes' visibility='default' id='a5fe3647'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='drr_pad' type-id='8c2bcad1' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2176'>
	<var-decl name='drr_checksum' type-id='39730d0b' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='dmu_replay_record_t' type-id='781a52d7' id='8b8fc893'/>
	<class-decl name='zinject_record' size-in-bits='2816' is-struct='yes' visibility='default' id='3216f820'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='zi_objset' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='zi_object' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='zi_start' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='zi_end' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='zi_guid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='zi_level' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='352'>
	<var-decl name='zi_error' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='zi_type' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='zi_freq' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='480'>
	<var-decl name='zi_failfast' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='zi_func' type-id='d1617432' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2560'>
	<var-decl name='zi_iotype' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2592'>
	<var-decl name='zi_duration' type-id='3ff5601b' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2624'>
	<var-decl name='zi_timer' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2688'>
	<var-decl name='zi_nlanes' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2752'>
	<var-decl name='zi_cmd' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='2784'>
	<var-decl name='zi_dvas' type-id='8f92235e' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zinject_record_t' type-id='3216f820' id='a4301ca6'/>
	<class-decl name='zfs_share' size-in-bits='256' is-struct='yes' visibility='default' id='feb6f2da'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='z_exportdata' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='z_sharedata' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='z_sharetype' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='z_sharemax' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zfs_share_t' type-id='feb6f2da' id='ee5cec36'/>
	<class-decl name='zfs_cmd' size-in-bits='109952' is-struct='yes' visibility='default' id='3522cd69'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='zc_name' type-id='d16c6df4' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32768'>
	<var-decl name='zc_nvlist_src' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32832'>
	<var-decl name='zc_nvlist_src_size' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32896'>
	<var-decl name='zc_nvlist_dst' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32960'>
	<var-decl name='zc_nvlist_dst_size' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='33024'>
	<var-decl name='zc_nvlist_dst_filled' type-id='c19b74c3' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='33056'>
	<var-decl name='zc_pad2' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='33088'>
	<var-decl name='zc_history' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='33152'>
	<var-decl name='zc_value' type-id='163f6aa5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='98688'>
	<var-decl name='zc_string' type-id='d1617432' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='100736'>
	<var-decl name='zc_guid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='100800'>
	<var-decl name='zc_nvlist_conf' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='100864'>
	<var-decl name='zc_nvlist_conf_size' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='100928'>
	<var-decl name='zc_cookie' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='100992'>
	<var-decl name='zc_objset_type' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='101056'>
	<var-decl name='zc_perm_action' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='101120'>
	<var-decl name='zc_history_len' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='101184'>
	<var-decl name='zc_history_offset' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='101248'>
	<var-decl name='zc_obj' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='101312'>
	<var-decl name='zc_iflags' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='101376'>
	<var-decl name='zc_share' type-id='ee5cec36' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='101632'>
	<var-decl name='zc_objset_stats' type-id='b2c14f17' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='103936'>
	<var-decl name='zc_begin_record' type-id='09fcdc01' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='106368'>
	<var-decl name='zc_inject_record' type-id='a4301ca6' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109184'>
	<var-decl name='zc_defer_destroy' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109216'>
	<var-decl name='zc_flags' type-id='8f92235e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109248'>
	<var-decl name='zc_action_handle' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109312'>
	<var-decl name='zc_cleanup_fd' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109344'>
	<var-decl name='zc_simple' type-id='b96825af' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109352'>
	<var-decl name='zc_pad' type-id='d3490169' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109376'>
	<var-decl name='zc_sendobj' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109440'>
	<var-decl name='zc_fromobj' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109504'>
	<var-decl name='zc_createtxg' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109568'>
	<var-decl name='zc_stat' type-id='0371a9c7' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='109888'>
	<var-decl name='zc_zoneid' type-id='9c313c2d' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='zfs_stat' size-in-bits='320' is-struct='yes' visibility='default' id='6417f0b9'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='zs_gen' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='zs_mode' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='zs_links' type-id='9c313c2d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='zs_ctime' type-id='c1c22e6c' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='zfs_stat_t' type-id='6417f0b9' id='0371a9c7'/>
	<enum-decl name='boolean_t' naming-typedef-id='c19b74c3' id='f58c8277'>
	<underlying-type type-id='9cac1fee'/>
	<enumerator name='B_FALSE' value='0'/>
	<enumerator name='B_TRUE' value='1'/>
	</enum-decl>
	<typedef-decl name='boolean_t' type-id='f58c8277' id='c19b74c3'/>
	<typedef-decl name='uchar_t' type-id='002ac4a6' id='d8bf0010'/>
	<typedef-decl name='uint_t' type-id='f0981eeb' id='3502e3ff'/>
	<typedef-decl name='pthread_t' type-id='7359adad' id='4051f5e7'/>
	<union-decl name='pthread_attr_t' size-in-bits='448' visibility='default' id='b63afacd'>
	<data-member access='public'>
	<var-decl name='__size' type-id='6093ff7c' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='__align' type-id='bd54fe1a' visibility='default'/>
	</data-member>
	</union-decl>
	<typedef-decl name='pthread_attr_t' type-id='b63afacd' id='7d8569fd'/>
	<union-decl name='pthread_mutex_t' size-in-bits='320' naming-typedef-id='7a6844eb' visibility='default' id='70681f9b'>
	<data-member access='public'>
	<var-decl name='__data' type-id='4c734837' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='__size' type-id='36c46961' visibility='default'/>
	</data-member>
	<data-member access='public'>
	<var-decl name='__align' type-id='bd54fe1a' visibility='default'/>
	</data-member>
	</union-decl>
	<typedef-decl name='pthread_mutex_t' type-id='70681f9b' id='7a6844eb'/>
	<typedef-decl name='int16_t' type-id='03896e23' id='23bd8cb5'/>
	<typedef-decl name='int32_t' type-id='33f57a65' id='3ff5601b'/>
	<typedef-decl name='uint8_t' type-id='c51d6389' id='b96825af'/>
	<typedef-decl name='uint32_t' type-id='62f1140c' id='8f92235e'/>
	<typedef-decl name='uint64_t' type-id='8910171f' id='9c313c2d'/>
	<class-decl name='__pthread_mutex_s' size-in-bits='320' is-struct='yes' visibility='default' id='4c734837'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='__lock' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='32'>
	<var-decl name='__count' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='__owner' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='96'>
	<var-decl name='__nusers' type-id='f0981eeb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='__kind' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='160'>
	<var-decl name='__spins' type-id='a2185560' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='176'>
	<var-decl name='__elision' type-id='a2185560' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='__list' type-id='518fb49c' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='stat' size-in-bits='1152' is-struct='yes' visibility='default' id='aafc373f'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='st_dev' type-id='35ed8932' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='st_ino' type-id='e43e523d' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='st_nlink' type-id='80f0b9df' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='st_mode' type-id='e1c52942' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='224'>
	<var-decl name='st_uid' type-id='cc5fcceb' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='st_gid' type-id='d94ec6d9' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='288'>
	<var-decl name='__pad0' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='st_rdev' type-id='35ed8932' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='st_size' type-id='79989e9c' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='st_blksize' type-id='d3f10a7f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='st_blocks' type-id='dbc43803' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='576'>
	<var-decl name='st_atim' type-id='a9c79a1f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='704'>
	<var-decl name='st_mtim' type-id='a9c79a1f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='832'>
	<var-decl name='st_ctim' type-id='a9c79a1f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='960'>
	<var-decl name='__glibc_reserved' type-id='083f8d58' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='__pthread_internal_list' size-in-bits='128' is-struct='yes' visibility='default' id='0e01899c'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='__prev' type-id='4d98cd5a' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='__next' type-id='4d98cd5a' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='__pthread_list_t' type-id='0e01899c' id='518fb49c'/>
	<typedef-decl name='__uint8_t' type-id='002ac4a6' id='c51d6389'/>
	<typedef-decl name='__int16_t' type-id='a2185560' id='03896e23'/>
	<typedef-decl name='__int32_t' type-id='95e97e5e' id='33f57a65'/>
	<typedef-decl name='__uint32_t' type-id='f0981eeb' id='62f1140c'/>
	<typedef-decl name='__uint64_t' type-id='7359adad' id='8910171f'/>
	<typedef-decl name='__dev_t' type-id='7359adad' id='35ed8932'/>
	<typedef-decl name='__uid_t' type-id='f0981eeb' id='cc5fcceb'/>
	<typedef-decl name='__gid_t' type-id='f0981eeb' id='d94ec6d9'/>
	<typedef-decl name='__ino_t' type-id='7359adad' id='e43e523d'/>
	<typedef-decl name='__mode_t' type-id='f0981eeb' id='e1c52942'/>
	<typedef-decl name='__nlink_t' type-id='7359adad' id='80f0b9df'/>
	<typedef-decl name='__off_t' type-id='bd54fe1a' id='79989e9c'/>
	<typedef-decl name='__off64_t' type-id='bd54fe1a' id='724e4de6'/>
	<typedef-decl name='__time_t' type-id='bd54fe1a' id='65eda9c0'/>
	<typedef-decl name='__blksize_t' type-id='bd54fe1a' id='d3f10a7f'/>
	<typedef-decl name='__blkcnt_t' type-id='bd54fe1a' id='dbc43803'/>
	<typedef-decl name='__ssize_t' type-id='bd54fe1a' id='41060289'/>
	<typedef-decl name='__syscall_slong_t' type-id='bd54fe1a' id='03085adc'/>
	<typedef-decl name='FILE' type-id='ec1ed955' id='aa12d1ba'/>
	<typedef-decl name='_IO_lock_t' type-id='48b5725f' id='bb4788fa'/>
	<class-decl name='_IO_FILE' size-in-bits='1728' is-struct='yes' visibility='default' id='ec1ed955'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='_flags' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='_IO_read_ptr' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='128'>
	<var-decl name='_IO_read_end' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='192'>
	<var-decl name='_IO_read_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='256'>
	<var-decl name='_IO_write_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='320'>
	<var-decl name='_IO_write_ptr' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='384'>
	<var-decl name='_IO_write_end' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='448'>
	<var-decl name='_IO_buf_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='512'>
	<var-decl name='_IO_buf_end' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='576'>
	<var-decl name='_IO_save_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='640'>
	<var-decl name='_IO_backup_base' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='704'>
	<var-decl name='_IO_save_end' type-id='26a90f95' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='768'>
	<var-decl name='_markers' type-id='e4c6fa61' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='832'>
	<var-decl name='_chain' type-id='dca988a5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='896'>
	<var-decl name='_fileno' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='928'>
	<var-decl name='_flags2' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='960'>
	<var-decl name='_old_offset' type-id='79989e9c' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1024'>
	<var-decl name='_cur_column' type-id='8efea9e5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1040'>
	<var-decl name='_vtable_offset' type-id='28577a57' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1048'>
	<var-decl name='_shortbuf' type-id='89feb1ec' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1088'>
	<var-decl name='_lock' type-id='cecf4ea7' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1152'>
	<var-decl name='_offset' type-id='724e4de6' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1216'>
	<var-decl name='_codecvt' type-id='570f8c59' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1280'>
	<var-decl name='_wide_data' type-id='c65a1f29' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1344'>
	<var-decl name='_freeres_list' type-id='dca988a5' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1408'>
	<var-decl name='_freeres_buf' type-id='eaa32e2f' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1472'>
	<var-decl name='__pad5' type-id='b59d7dce' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1536'>
	<var-decl name='_mode' type-id='95e97e5e' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='1568'>
	<var-decl name='_unused2' type-id='664ac0b7' visibility='default'/>
	</data-member>
	</class-decl>
	<class-decl name='timespec' size-in-bits='128' is-struct='yes' visibility='default' id='a9c79a1f'>
	<data-member access='public' layout-offset-in-bits='0'>
	<var-decl name='tv_sec' type-id='65eda9c0' visibility='default'/>
	</data-member>
	<data-member access='public' layout-offset-in-bits='64'>
	<var-decl name='tv_nsec' type-id='03085adc' visibility='default'/>
	</data-member>
	</class-decl>
	<typedef-decl name='ssize_t' type-id='41060289' id='79a0948f'/>
	<typedef-decl name='size_t' type-id='7359adad' id='b59d7dce'/>
	<pointer-type-def type-id='aa12d1ba' size-in-bits='64' id='822cd80b'/>
	<qualified-type-def type-id='822cd80b' restrict='yes' id='e75a27e9'/>
	<pointer-type-def type-id='ec1ed955' size-in-bits='64' id='dca988a5'/>
	<pointer-type-def type-id='a4036571' size-in-bits='64' id='570f8c59'/>
	<pointer-type-def type-id='bb4788fa' size-in-bits='64' id='cecf4ea7'/>
	<pointer-type-def type-id='010ae0b9' size-in-bits='64' id='e4c6fa61'/>
	<pointer-type-def type-id='79bd3751' size-in-bits='64' id='c65a1f29'/>
	<pointer-type-def type-id='724e4de6' size-in-bits='64' id='ecf845f9'/>
	<pointer-type-def type-id='0e01899c' size-in-bits='64' id='4d98cd5a'/>
	<pointer-type-def type-id='c19b74c3' size-in-bits='64' id='37e3bd22'/>
	<pointer-type-def type-id='a84c031d' size-in-bits='64' id='26a90f95'/>
	<qualified-type-def type-id='a84c031d' const='yes' id='9b45d938'/>
	<pointer-type-def type-id='9b45d938' size-in-bits='64' id='80f4b756'/>
	<qualified-type-def type-id='80f4b756' restrict='yes' id='9d26089a'/>
	<qualified-type-def type-id='8b8fc893' const='yes' id='9623bc03'/>
	<pointer-type-def type-id='9623bc03' size-in-bits='64' id='8341348b'/>
	<qualified-type-def type-id='8e8d4be3' const='yes' id='693c3853'/>
	<pointer-type-def type-id='693c3853' size-in-bits='64' id='22cce67b'/>
	<qualified-type-def type-id='57928edf' const='yes' id='642ee20f'/>
	<pointer-type-def type-id='642ee20f' size-in-bits='64' id='dace003f'/>
	<qualified-type-def type-id='7d8569fd' const='yes' id='e06dee2d'/>
	<pointer-type-def type-id='e06dee2d' size-in-bits='64' id='540db505'/>
	<qualified-type-def type-id='540db505' restrict='yes' id='e1815e87'/>
	<qualified-type-def type-id='d8bf0010' const='yes' id='a9125480'/>
	<pointer-type-def type-id='a9125480' size-in-bits='64' id='d1db479e'/>
	<qualified-type-def type-id='b96825af' const='yes' id='2b61797f'/>
	<pointer-type-def type-id='2b61797f' size-in-bits='64' id='9f7200cf'/>
	<pointer-type-def type-id='c70fa2e8' size-in-bits='64' id='2e711a2a'/>
	<pointer-type-def type-id='95e97e5e' size-in-bits='64' id='7292109c'/>
	<pointer-type-def type-id='8e8d4be3' size-in-bits='64' id='5ce45b60'/>
	<pointer-type-def type-id='5ce45b60' size-in-bits='64' id='857bb57e'/>
	<pointer-type-def type-id='57928edf' size-in-bits='64' id='3fa542f0'/>
	<pointer-type-def type-id='7a6844eb' size-in-bits='64' id='18c91f9e'/>
	<pointer-type-def type-id='4051f5e7' size-in-bits='64' id='e01b5462'/>
	<qualified-type-def type-id='e01b5462' restrict='yes' id='cc338b26'/>
	<pointer-type-def type-id='b59d7dce' size-in-bits='64' id='78c01427'/>
	<pointer-type-def type-id='aafc373f' size-in-bits='64' id='4330df87'/>
	<pointer-type-def type-id='9c313c2d' size-in-bits='64' id='5d6479ae'/>
	<pointer-type-def type-id='b96825af' size-in-bits='64' id='ae3e8ca6'/>
	<pointer-type-def type-id='48b5725f' size-in-bits='64' id='eaa32e2f'/>
	<pointer-type-def type-id='cd5d79f4' size-in-bits='64' id='5ad9edb6'/>
	<qualified-type-def type-id='eaa32e2f' restrict='yes' id='1b7446cd'/>
	<pointer-type-def type-id='eaa32e2f' size-in-bits='64' id='63e171df'/>
	<pointer-type-def type-id='3522cd69' size-in-bits='64' id='b65f7fd1'/>
	<class-decl name='_IO_codecvt' is-struct='yes' visibility='default' is-declaration-only='yes' id='a4036571'/>
	<class-decl name='_IO_marker' is-struct='yes' visibility='default' is-declaration-only='yes' id='010ae0b9'/>
	<class-decl name='_IO_wide_data' is-struct='yes' visibility='default' is-declaration-only='yes' id='79bd3751'/>
	<function-decl name='lzc_ioctl_fd' mangled-name='lzc_ioctl_fd' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_ioctl_fd'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='7359adad'/>
	<parameter type-id='b65f7fd1'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_free' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='nvlist_unpack' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='26a90f95'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='857bb57e'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_uint64' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='22cce67b'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='5d6479ae'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_lookup_nvlist' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='857bb57e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='nvlist_next_nvpair' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='dace003f'/>
	<return type-id='3fa542f0'/>
	</function-decl>
	<function-decl name='nvpair_name' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='dace003f'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='fnvlist_alloc' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='fnvlist_free' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_pack' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='78c01427'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='fnvlist_pack_free' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='26a90f95'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_unpack' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='26a90f95'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='fnvlist_dup' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='22cce67b'/>
	<return type-id='5ce45b60'/>
	</function-decl>
	<function-decl name='fnvlist_add_boolean' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_boolean_value' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='c19b74c3'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_int32' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='3ff5601b'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_uint64' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='9c313c2d'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_string' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_nvlist' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='5ce45b60'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_byte_array' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='d1db479e'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_add_uint8_array' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='5ce45b60'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='9f7200cf'/>
	<parameter type-id='3502e3ff'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_boolean_value' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='22cce67b'/>
	<parameter type-id='80f4b756'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_uint64' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='22cce67b'/>
	<parameter type-id='80f4b756'/>
	<return type-id='9c313c2d'/>
	</function-decl>
	<function-decl name='fnvlist_lookup_string' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='22cce67b'/>
	<parameter type-id='80f4b756'/>
	<return type-id='80f4b756'/>
	</function-decl>
	<function-decl name='libspl_assertf' mangled-name='libspl_assertf' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libspl_assertf'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='80f4b756'/>
	<parameter is-variadic='yes'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='strlcpy' mangled-name='strlcpy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='strlcpy'>
	<parameter type-id='26a90f95'/>
	<parameter type-id='80f4b756'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='__errno_location' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='7292109c'/>
	</function-decl>
	<function-decl name='pthread_create' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='cc338b26'/>
	<parameter type-id='e1815e87'/>
	<parameter type-id='5ad9edb6'/>
	<parameter type-id='1b7446cd'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_join' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='4051f5e7'/>
	<parameter type-id='63e171df'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_mutex_lock' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='18c91f9e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pthread_mutex_unlock' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='18c91f9e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='fclose' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='822cd80b'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='malloc' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='b59d7dce'/>
	<return type-id='eaa32e2f'/>
	</function-decl>
	<function-decl name='free' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='eaa32e2f'/>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='getenv' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='strchr' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='strrchr' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='26a90f95'/>
	</function-decl>
	<function-decl name='strcspn' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='80f4b756'/>
	<parameter type-id='80f4b756'/>
	<return type-id='b59d7dce'/>
	</function-decl>
	<function-decl name='close' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='pipe2' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='7292109c'/>
	<parameter type-id='95e97e5e'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='splice' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='ecf845f9'/>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='ecf845f9'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='f0981eeb'/>
	<return type-id='41060289'/>
	</function-decl>
	<function-decl name='__open_too_many_args' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='__open_missing_mode' visibility='default' binding='global' size-in-bits='64'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='__read_chk' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='eaa32e2f'/>
	<parameter type-id='b59d7dce'/>
	<parameter type-id='b59d7dce'/>
	<return type-id='79a0948f'/>
	</function-decl>
	<function-decl name='libzfs_core_init' mangled-name='libzfs_core_init' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_core_init'>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='libzfs_core_fini' mangled-name='libzfs_core_fini' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='libzfs_core_fini'>
	<return type-id='48b5725f'/>
	</function-decl>
	<function-decl name='lzc_scrub' mangled-name='lzc_scrub' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_scrub'>
	<parameter type-id='5b35941c' name='ioc'/>
	<parameter type-id='80f4b756' name='name'/>
	<parameter type-id='5ce45b60' name='source'/>
	<parameter type-id='857bb57e' name='resultp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_create' mangled-name='lzc_create' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_create'>
	<parameter type-id='80f4b756' name='fsname'/>
	<parameter type-id='bc9887f1' name='type'/>
	<parameter type-id='5ce45b60' name='props'/>
	<parameter type-id='ae3e8ca6' name='wkeydata'/>
	<parameter type-id='3502e3ff' name='wkeylen'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_clone' mangled-name='lzc_clone' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_clone'>
	<parameter type-id='80f4b756' name='fsname'/>
	<parameter type-id='80f4b756' name='origin'/>
	<parameter type-id='5ce45b60' name='props'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_promote' mangled-name='lzc_promote' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_promote'>
	<parameter type-id='80f4b756' name='fsname'/>
	<parameter type-id='26a90f95' name='snapnamebuf'/>
	<parameter type-id='95e97e5e' name='snapnamelen'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_rename' mangled-name='lzc_rename' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_rename'>
	<parameter type-id='80f4b756' name='source'/>
	<parameter type-id='80f4b756' name='target'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_destroy' mangled-name='lzc_destroy' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_destroy'>
	<parameter type-id='80f4b756' name='fsname'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_snapshot' mangled-name='lzc_snapshot' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_snapshot'>
	<parameter type-id='5ce45b60' name='snaps'/>
	<parameter type-id='5ce45b60' name='props'/>
	<parameter type-id='857bb57e' name='errlist'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_destroy_snaps' mangled-name='lzc_destroy_snaps' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_destroy_snaps'>
	<parameter type-id='5ce45b60' name='snaps'/>
	<parameter type-id='c19b74c3' name='defer'/>
	<parameter type-id='857bb57e' name='errlist'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_snaprange_space' mangled-name='lzc_snaprange_space' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_snaprange_space'>
	<parameter type-id='80f4b756' name='firstsnap'/>
	<parameter type-id='80f4b756' name='lastsnap'/>
	<parameter type-id='5d6479ae' name='usedp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_exists' mangled-name='lzc_exists' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_exists'>
	<parameter type-id='80f4b756' name='dataset'/>
	<return type-id='c19b74c3'/>
	</function-decl>
	<function-decl name='lzc_sync' mangled-name='lzc_sync' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_sync'>
	<parameter type-id='80f4b756' name='pool_name'/>
	<parameter type-id='5ce45b60' name='innvl'/>
	<parameter type-id='857bb57e' name='outnvl'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_hold' mangled-name='lzc_hold' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_hold'>
	<parameter type-id='5ce45b60' name='holds'/>
	<parameter type-id='95e97e5e' name='cleanup_fd'/>
	<parameter type-id='857bb57e' name='errlist'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_release' mangled-name='lzc_release' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_release'>
	<parameter type-id='5ce45b60' name='holds'/>
	<parameter type-id='857bb57e' name='errlist'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_get_holds' mangled-name='lzc_get_holds' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_get_holds'>
	<parameter type-id='80f4b756' name='snapname'/>
	<parameter type-id='857bb57e' name='holdsp'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_send_wrapper' mangled-name='lzc_send_wrapper' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_send_wrapper'>
	<parameter type-id='2e711a2a' name='func'/>
	<parameter type-id='95e97e5e' name='orig_fd'/>
	<parameter type-id='eaa32e2f' name='data'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_send' mangled-name='lzc_send' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_send'>
	<parameter type-id='80f4b756' name='snapname'/>
	<parameter type-id='80f4b756' name='from'/>
	<parameter type-id='95e97e5e' name='fd'/>
	<parameter type-id='bfbd3c8e' name='flags'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_send_redacted' mangled-name='lzc_send_redacted' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_send_redacted'>
	<parameter type-id='80f4b756' name='snapname'/>
	<parameter type-id='80f4b756' name='from'/>
	<parameter type-id='95e97e5e' name='fd'/>
	<parameter type-id='bfbd3c8e' name='flags'/>
	<parameter type-id='80f4b756' name='redactbook'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_send_resume' mangled-name='lzc_send_resume' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_send_resume'>
	<parameter type-id='80f4b756' name='snapname'/>
	<parameter type-id='80f4b756' name='from'/>
	<parameter type-id='95e97e5e' name='fd'/>
	<parameter type-id='bfbd3c8e' name='flags'/>
	<parameter type-id='9c313c2d' name='resumeobj'/>
	<parameter type-id='9c313c2d' name='resumeoff'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_send_resume_redacted' mangled-name='lzc_send_resume_redacted' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_send_resume_redacted'>
	<parameter type-id='80f4b756' name='snapname'/>
	<parameter type-id='80f4b756' name='from'/>
	<parameter type-id='95e97e5e' name='fd'/>
	<parameter type-id='bfbd3c8e' name='flags'/>
	<parameter type-id='9c313c2d' name='resumeobj'/>
	<parameter type-id='9c313c2d' name='resumeoff'/>
	<parameter type-id='80f4b756' name='redactbook'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_send_space_resume_redacted' mangled-name='lzc_send_space_resume_redacted' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_send_space_resume_redacted'>
	<parameter type-id='80f4b756' name='snapname'/>
	<parameter type-id='80f4b756' name='from'/>
	<parameter type-id='bfbd3c8e' name='flags'/>
	<parameter type-id='9c313c2d' name='resumeobj'/>
	<parameter type-id='9c313c2d' name='resumeoff'/>
	<parameter type-id='9c313c2d' name='resume_bytes'/>
	<parameter type-id='80f4b756' name='redactbook'/>
	<parameter type-id='95e97e5e' name='fd'/>
	<parameter type-id='5d6479ae' name='spacep'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_send_space' mangled-name='lzc_send_space' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_send_space'>
	<parameter type-id='80f4b756' name='snapname'/>
	<parameter type-id='80f4b756' name='from'/>
	<parameter type-id='bfbd3c8e' name='flags'/>
	<parameter type-id='5d6479ae' name='spacep'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_receive' mangled-name='lzc_receive' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_receive'>
	<parameter type-id='80f4b756' name='snapname'/>
	<parameter type-id='5ce45b60' name='props'/>
	<parameter type-id='80f4b756' name='origin'/>
	<parameter type-id='c19b74c3' name='force'/>
	<parameter type-id='c19b74c3' name='raw'/>
	<parameter type-id='95e97e5e' name='fd'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_receive_resumable' mangled-name='lzc_receive_resumable' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_receive_resumable'>
	<parameter type-id='80f4b756' name='snapname'/>
	<parameter type-id='5ce45b60' name='props'/>
	<parameter type-id='80f4b756' name='origin'/>
	<parameter type-id='c19b74c3' name='force'/>
	<parameter type-id='c19b74c3' name='raw'/>
	<parameter type-id='95e97e5e' name='fd'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_receive_with_header' mangled-name='lzc_receive_with_header' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_receive_with_header'>
	<parameter type-id='80f4b756' name='snapname'/>
	<parameter type-id='5ce45b60' name='props'/>
	<parameter type-id='80f4b756' name='origin'/>
	<parameter type-id='c19b74c3' name='force'/>
	<parameter type-id='c19b74c3' name='resumable'/>
	<parameter type-id='c19b74c3' name='raw'/>
	<parameter type-id='95e97e5e' name='fd'/>
	<parameter type-id='8341348b' name='begin_record'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_receive_one' mangled-name='lzc_receive_one' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_receive_one'>
	<parameter type-id='80f4b756' name='snapname'/>
	<parameter type-id='5ce45b60' name='props'/>
	<parameter type-id='80f4b756' name='origin'/>
	<parameter type-id='c19b74c3' name='force'/>
	<parameter type-id='c19b74c3' name='resumable'/>
	<parameter type-id='c19b74c3' name='raw'/>
	<parameter type-id='95e97e5e' name='input_fd'/>
	<parameter type-id='8341348b' name='begin_record'/>
	<parameter type-id='95e97e5e' name='cleanup_fd'/>
	<parameter type-id='5d6479ae' name='read_bytes'/>
	<parameter type-id='5d6479ae' name='errflags'/>
	<parameter type-id='5d6479ae' name='action_handle'/>
	<parameter type-id='857bb57e' name='errors'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_receive_with_cmdprops' mangled-name='lzc_receive_with_cmdprops' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_receive_with_cmdprops'>
	<parameter type-id='80f4b756' name='snapname'/>
	<parameter type-id='5ce45b60' name='props'/>
	<parameter type-id='5ce45b60' name='cmdprops'/>
	<parameter type-id='ae3e8ca6' name='wkeydata'/>
	<parameter type-id='3502e3ff' name='wkeylen'/>
	<parameter type-id='80f4b756' name='origin'/>
	<parameter type-id='c19b74c3' name='force'/>
	<parameter type-id='c19b74c3' name='resumable'/>
	<parameter type-id='c19b74c3' name='raw'/>
	<parameter type-id='95e97e5e' name='input_fd'/>
	<parameter type-id='8341348b' name='begin_record'/>
	<parameter type-id='95e97e5e' name='cleanup_fd'/>
	<parameter type-id='5d6479ae' name='read_bytes'/>
	<parameter type-id='5d6479ae' name='errflags'/>
	<parameter type-id='5d6479ae' name='action_handle'/>
	<parameter type-id='857bb57e' name='errors'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_receive_with_heal' mangled-name='lzc_receive_with_heal' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_receive_with_heal'>
	<parameter type-id='80f4b756' name='snapname'/>
	<parameter type-id='5ce45b60' name='props'/>
	<parameter type-id='5ce45b60' name='cmdprops'/>
	<parameter type-id='ae3e8ca6' name='wkeydata'/>
	<parameter type-id='3502e3ff' name='wkeylen'/>
	<parameter type-id='80f4b756' name='origin'/>
	<parameter type-id='c19b74c3' name='force'/>
	<parameter type-id='c19b74c3' name='heal'/>
	<parameter type-id='c19b74c3' name='resumable'/>
	<parameter type-id='c19b74c3' name='raw'/>
	<parameter type-id='95e97e5e' name='input_fd'/>
	<parameter type-id='8341348b' name='begin_record'/>
	<parameter type-id='95e97e5e' name='cleanup_fd'/>
	<parameter type-id='5d6479ae' name='read_bytes'/>
	<parameter type-id='5d6479ae' name='errflags'/>
	<parameter type-id='5d6479ae' name='action_handle'/>
	<parameter type-id='857bb57e' name='errors'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_rollback' mangled-name='lzc_rollback' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_rollback'>
	<parameter type-id='80f4b756' name='fsname'/>
	<parameter type-id='26a90f95' name='snapnamebuf'/>
	<parameter type-id='95e97e5e' name='snapnamelen'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_rollback_to' mangled-name='lzc_rollback_to' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_rollback_to'>
	<parameter type-id='80f4b756' name='fsname'/>
	<parameter type-id='80f4b756' name='snapname'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_bookmark' mangled-name='lzc_bookmark' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_bookmark'>
	<parameter type-id='5ce45b60' name='bookmarks'/>
	<parameter type-id='857bb57e' name='errlist'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_get_bookmarks' mangled-name='lzc_get_bookmarks' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_get_bookmarks'>
	<parameter type-id='80f4b756' name='fsname'/>
	<parameter type-id='5ce45b60' name='props'/>
	<parameter type-id='857bb57e' name='bmarks'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_get_bookmark_props' mangled-name='lzc_get_bookmark_props' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_get_bookmark_props'>
	<parameter type-id='80f4b756' name='bookmark'/>
	<parameter type-id='857bb57e' name='props'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_destroy_bookmarks' mangled-name='lzc_destroy_bookmarks' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_destroy_bookmarks'>
	<parameter type-id='5ce45b60' name='bmarks'/>
	<parameter type-id='857bb57e' name='errlist'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_channel_program' mangled-name='lzc_channel_program' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_channel_program'>
	<parameter type-id='80f4b756' name='pool'/>
	<parameter type-id='80f4b756' name='program'/>
	<parameter type-id='9c313c2d' name='instrlimit'/>
	<parameter type-id='9c313c2d' name='memlimit'/>
	<parameter type-id='5ce45b60' name='argnvl'/>
	<parameter type-id='857bb57e' name='outnvl'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_pool_checkpoint' mangled-name='lzc_pool_checkpoint' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_pool_checkpoint'>
	<parameter type-id='80f4b756' name='pool'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_pool_checkpoint_discard' mangled-name='lzc_pool_checkpoint_discard' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_pool_checkpoint_discard'>
	<parameter type-id='80f4b756' name='pool'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_channel_program_nosync' mangled-name='lzc_channel_program_nosync' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_channel_program_nosync'>
	<parameter type-id='80f4b756' name='pool'/>
	<parameter type-id='80f4b756' name='program'/>
	<parameter type-id='9c313c2d' name='timeout'/>
	<parameter type-id='9c313c2d' name='memlimit'/>
	<parameter type-id='5ce45b60' name='argnvl'/>
	<parameter type-id='857bb57e' name='outnvl'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_get_vdev_prop' mangled-name='lzc_get_vdev_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_get_vdev_prop'>
	<parameter type-id='80f4b756' name='poolname'/>
	<parameter type-id='5ce45b60' name='innvl'/>
	<parameter type-id='857bb57e' name='outnvl'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_set_vdev_prop' mangled-name='lzc_set_vdev_prop' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_set_vdev_prop'>
	<parameter type-id='80f4b756' name='poolname'/>
	<parameter type-id='5ce45b60' name='innvl'/>
	<parameter type-id='857bb57e' name='outnvl'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_load_key' mangled-name='lzc_load_key' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_load_key'>
	<parameter type-id='80f4b756' name='fsname'/>
	<parameter type-id='c19b74c3' name='noop'/>
	<parameter type-id='ae3e8ca6' name='wkeydata'/>
	<parameter type-id='3502e3ff' name='wkeylen'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_unload_key' mangled-name='lzc_unload_key' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_unload_key'>
	<parameter type-id='80f4b756' name='fsname'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_change_key' mangled-name='lzc_change_key' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_change_key'>
	<parameter type-id='80f4b756' name='fsname'/>
	<parameter type-id='9c313c2d' name='crypt_cmd'/>
	<parameter type-id='5ce45b60' name='props'/>
	<parameter type-id='ae3e8ca6' name='wkeydata'/>
	<parameter type-id='3502e3ff' name='wkeylen'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_reopen' mangled-name='lzc_reopen' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_reopen'>
	<parameter type-id='80f4b756' name='pool_name'/>
	<parameter type-id='c19b74c3' name='scrub_restart'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_initialize' mangled-name='lzc_initialize' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_initialize'>
	<parameter type-id='80f4b756' name='poolname'/>
	<parameter type-id='7063e1ab' name='cmd_type'/>
	<parameter type-id='5ce45b60' name='vdevs'/>
	<parameter type-id='857bb57e' name='errlist'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_trim' mangled-name='lzc_trim' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_trim'>
	<parameter type-id='80f4b756' name='poolname'/>
	<parameter type-id='b1146b8d' name='cmd_type'/>
	<parameter type-id='9c313c2d' name='rate'/>
	<parameter type-id='c19b74c3' name='secure'/>
	<parameter type-id='5ce45b60' name='vdevs'/>
	<parameter type-id='857bb57e' name='errlist'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_redact' mangled-name='lzc_redact' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_redact'>
	<parameter type-id='80f4b756' name='snapshot'/>
	<parameter type-id='80f4b756' name='bookname'/>
	<parameter type-id='5ce45b60' name='snapnv'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_wait' mangled-name='lzc_wait' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_wait'>
	<parameter type-id='80f4b756' name='pool'/>
	<parameter type-id='73446457' name='activity'/>
	<parameter type-id='37e3bd22' name='waited'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_wait_tag' mangled-name='lzc_wait_tag' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_wait_tag'>
	<parameter type-id='80f4b756' name='pool'/>
	<parameter type-id='73446457' name='activity'/>
	<parameter type-id='9c313c2d' name='tag'/>
	<parameter type-id='37e3bd22' name='waited'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_wait_fs' mangled-name='lzc_wait_fs' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_wait_fs'>
	<parameter type-id='80f4b756' name='fs'/>
	<parameter type-id='3024501a' name='activity'/>
	<parameter type-id='37e3bd22' name='waited'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_set_bootenv' mangled-name='lzc_set_bootenv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_set_bootenv'>
	<parameter type-id='80f4b756' name='pool'/>
	<parameter type-id='22cce67b' name='env'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-decl name='lzc_get_bootenv' mangled-name='lzc_get_bootenv' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='lzc_get_bootenv'>
	<parameter type-id='80f4b756' name='pool'/>
	<parameter type-id='857bb57e' name='outnvl'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	<function-type size-in-bits='64' id='c70fa2e8'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='eaa32e2f'/>
	<return type-id='95e97e5e'/>
	</function-type>
	<function-type size-in-bits='64' id='cd5d79f4'>
	<parameter type-id='eaa32e2f'/>
	<return type-id='eaa32e2f'/>
	</function-type>
	</abi-instr>
	<abi-instr address-size='64' path='lib/libzfs_core/os/linux/libzfs_core_ioctl.c' language='LANG_C99'>
	<function-decl name='ioctl' visibility='default' binding='global' size-in-bits='64'>
	<parameter type-id='95e97e5e'/>
	<parameter type-id='7359adad'/>
	<parameter is-variadic='yes'/>
	<return type-id='95e97e5e'/>
	</function-decl>
	</abi-instr>
	</abi-corpus>
	diff --git a/sys/contrib/openzfs/man/man4/zfs.4 b/sys/contrib/openzfs/man/man4/zfs.4
	index 5edd80659e08..f1d14b4d01a4 100644
	--- a/sys/contrib/openzfs/man/man4/zfs.4
	+++ b/sys/contrib/openzfs/man/man4/zfs.4
	@@ -1,2703 +1,2706 @@
	.\"
	.\" Copyright (c) 2013 by Turbo Fredriksson <turbo@bayour.com>. All rights reserved.
	.\" Copyright (c) 2019, 2021 by Delphix. All rights reserved.
	.\" Copyright (c) 2019 Datto Inc.
	.\" Copyright (c) 2023, 2024 Klara, Inc.
	.\" The contents of this file are subject to the terms of the Common Development
	.\" and Distribution License (the "License"). You may not use this file except
	.\" in compliance with the License. You can obtain a copy of the license at
	.\" usr/src/OPENSOLARIS.LICENSE or https://opensource.org/licenses/CDDL-1.0.
	.\"
	.\" See the License for the specific language governing permissions and
	.\" limitations under the License. When distributing Covered Code, include this
	.\" CDDL HEADER in each file and include the License file at
	.\" usr/src/OPENSOLARIS.LICENSE. If applicable, add the following below this
	.\" CDDL HEADER, with the fields enclosed by brackets "[]" replaced with your
	.\" own identifying information:
	.\" Portions Copyright [yyyy] [name of copyright owner]
	.\"
	.Dd February 14, 2024
	.Dt ZFS 4
	.Os
	.
	.Sh NAME
	.Nm zfs
	.Nd tuning of the ZFS kernel module
	.
	.Sh DESCRIPTION
	The ZFS module supports these parameters:
	.Bl -tag -width Ds
	.It Sy dbuf_cache_max_bytes Ns = Ns Sy UINT64_MAX Ns B Pq u64
	Maximum size in bytes of the dbuf cache.
	The target size is determined by the MIN versus
	.No 1/2^ Ns Sy dbuf_cache_shift Pq 1/32nd
	of the target ARC size.
	The behavior of the dbuf cache and its associated settings
	can be observed via the
	.Pa /proc/spl/kstat/zfs/dbufstats
	kstat.
	.
	.It Sy dbuf_metadata_cache_max_bytes Ns = Ns Sy UINT64_MAX Ns B Pq u64
	Maximum size in bytes of the metadata dbuf cache.
	The target size is determined by the MIN versus
	.No 1/2^ Ns Sy dbuf_metadata_cache_shift Pq 1/64th
	of the target ARC size.
	The behavior of the metadata dbuf cache and its associated settings
	can be observed via the
	.Pa /proc/spl/kstat/zfs/dbufstats
	kstat.
	.
	.It Sy dbuf_cache_hiwater_pct Ns = Ns Sy 10 Ns % Pq uint
	The percentage over
	.Sy dbuf_cache_max_bytes
	when dbufs must be evicted directly.
	.
	.It Sy dbuf_cache_lowater_pct Ns = Ns Sy 10 Ns % Pq uint
	The percentage below
	.Sy dbuf_cache_max_bytes
	when the evict thread stops evicting dbufs.
	.
	.It Sy dbuf_cache_shift Ns = Ns Sy 5 Pq uint
	Set the size of the dbuf cache
	.Pq Sy dbuf_cache_max_bytes
	to a log2 fraction of the target ARC size.
	.
	.It Sy dbuf_metadata_cache_shift Ns = Ns Sy 6 Pq uint
	Set the size of the dbuf metadata cache
	.Pq Sy dbuf_metadata_cache_max_bytes
	to a log2 fraction of the target ARC size.
	.
	.It Sy dbuf_mutex_cache_shift Ns = Ns Sy 0 Pq uint
	Set the size of the mutex array for the dbuf cache.
	When set to
	.Sy 0
	the array is dynamically sized based on total system memory.
	.
	.It Sy dmu_object_alloc_chunk_shift Ns = Ns Sy 7 Po 128 Pc Pq uint
	dnode slots allocated in a single operation as a power of 2.
	The default value minimizes lock contention for the bulk operation performed.
	.
	.It Sy dmu_prefetch_max Ns = Ns Sy 134217728 Ns B Po 128 MiB Pc Pq uint
	Limit the amount we can prefetch with one call to this amount in bytes.
	This helps to limit the amount of memory that can be used by prefetching.
	.
	.It Sy ignore_hole_birth Pq int
	Alias for
	.Sy send_holes_without_birth_time .
	.
	.It Sy l2arc_feed_again Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Turbo L2ARC warm-up.
	When the L2ARC is cold the fill interval will be set as fast as possible.
	.
	.It Sy l2arc_feed_min_ms Ns = Ns Sy 200 Pq u64
	Min feed interval in milliseconds.
	Requires
	.Sy l2arc_feed_again Ns = Ns Ar 1
	and only applicable in related situations.
	.
	.It Sy l2arc_feed_secs Ns = Ns Sy 1 Pq u64
	Seconds between L2ARC writing.
	.
	.It Sy l2arc_headroom Ns = Ns Sy 8 Pq u64
	How far through the ARC lists to search for L2ARC cacheable content,
	expressed as a multiplier of
	.Sy l2arc_write_max .
	ARC persistence across reboots can be achieved with persistent L2ARC
	by setting this parameter to
	.Sy 0 ,
	allowing the full length of ARC lists to be searched for cacheable content.
	.
	.It Sy l2arc_headroom_boost Ns = Ns Sy 200 Ns % Pq u64
	Scales
	.Sy l2arc_headroom
	by this percentage when L2ARC contents are being successfully compressed
	before writing.
	A value of
	.Sy 100
	disables this feature.
	.
	.It Sy l2arc_exclude_special Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Controls whether buffers present on special vdevs are eligible for caching
	into L2ARC.
	If set to 1, exclude dbufs on special vdevs from being cached to L2ARC.
	.
	.It Sy l2arc_mfuonly Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Controls whether only MFU metadata and data are cached from ARC into L2ARC.
	This may be desired to avoid wasting space on L2ARC when reading/writing large
	amounts of data that are not expected to be accessed more than once.
	.Pp
	The default is off,
	meaning both MRU and MFU data and metadata are cached.
	When turning off this feature, some MRU buffers will still be present
	in ARC and eventually cached on L2ARC.
	.No If Sy l2arc_noprefetch Ns = Ns Sy 0 ,
	some prefetched buffers will be cached to L2ARC, and those might later
	transition to MRU, in which case the
	.Sy l2arc_mru_asize No arcstat will not be Sy 0 .
	.Pp
	Regardless of
	.Sy l2arc_noprefetch ,
	some MFU buffers might be evicted from ARC,
	accessed later on as prefetches and transition to MRU as prefetches.
	If accessed again they are counted as MRU and the
	.Sy l2arc_mru_asize No arcstat will not be Sy 0 .
	.Pp
	The ARC status of L2ARC buffers when they were first cached in
	L2ARC can be seen in the
	.Sy l2arc_mru_asize , Sy l2arc_mfu_asize , No and Sy l2arc_prefetch_asize
	arcstats when importing the pool or onlining a cache
	device if persistent L2ARC is enabled.
	.Pp
	The
	.Sy evict_l2_eligible_mru
	arcstat does not take into account if this option is enabled as the information
	provided by the
	.Sy evict_l2_eligible_m[rf]u
	arcstats can be used to decide if toggling this option is appropriate
	for the current workload.
	.
	.It Sy l2arc_meta_percent Ns = Ns Sy 33 Ns % Pq uint
	Percent of ARC size allowed for L2ARC-only headers.
	Since L2ARC buffers are not evicted on memory pressure,
	too many headers on a system with an irrationally large L2ARC
	can render it slow or unusable.
	This parameter limits L2ARC writes and rebuilds to achieve the target.
	.
	.It Sy l2arc_trim_ahead Ns = Ns Sy 0 Ns % Pq u64
	Trims ahead of the current write size
	.Pq Sy l2arc_write_max
	on L2ARC devices by this percentage of write size if we have filled the device.
	If set to
	.Sy 100
	we TRIM twice the space required to accommodate upcoming writes.
	A minimum of
	.Sy 64 MiB
	will be trimmed.
	It also enables TRIM of the whole L2ARC device upon creation
	or addition to an existing pool or if the header of the device is
	invalid upon importing a pool or onlining a cache device.
	A value of
	.Sy 0
	disables TRIM on L2ARC altogether and is the default as it can put significant
	stress on the underlying storage devices.
	This will vary depending of how well the specific device handles these commands.
	.
	.It Sy l2arc_noprefetch Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Do not write buffers to L2ARC if they were prefetched but not used by
	applications.
	In case there are prefetched buffers in L2ARC and this option
	is later set, we do not read the prefetched buffers from L2ARC.
	Unsetting this option is useful for caching sequential reads from the
	disks to L2ARC and serve those reads from L2ARC later on.
	This may be beneficial in case the L2ARC device is significantly faster
	in sequential reads than the disks of the pool.
	.Pp
	Use
	.Sy 1
	to disable and
	.Sy 0
	to enable caching/reading prefetches to/from L2ARC.
	.
	.It Sy l2arc_norw Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	No reads during writes.
	.
	.It Sy l2arc_write_boost Ns = Ns Sy 33554432 Ns B Po 32 MiB Pc Pq u64
	Cold L2ARC devices will have
	.Sy l2arc_write_max
	increased by this amount while they remain cold.
	.
	.It Sy l2arc_write_max Ns = Ns Sy 33554432 Ns B Po 32 MiB Pc Pq u64
	Max write bytes per interval.
	.
	.It Sy l2arc_rebuild_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Rebuild the L2ARC when importing a pool (persistent L2ARC).
	This can be disabled if there are problems importing a pool
	or attaching an L2ARC device (e.g. the L2ARC device is slow
	in reading stored log metadata, or the metadata
	has become somehow fragmented/unusable).
	.
	.It Sy l2arc_rebuild_blocks_min_l2size Ns = Ns Sy 1073741824 Ns B Po 1 GiB Pc Pq u64
	Mininum size of an L2ARC device required in order to write log blocks in it.
	The log blocks are used upon importing the pool to rebuild the persistent L2ARC.
	.Pp
	For L2ARC devices less than 1 GiB, the amount of data
	.Fn l2arc_evict
	evicts is significant compared to the amount of restored L2ARC data.
	In this case, do not write log blocks in L2ARC in order not to waste space.
	.
	.It Sy metaslab_aliquot Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq u64
	Metaslab granularity, in bytes.
	This is roughly similar to what would be referred to as the "stripe size"
	in traditional RAID arrays.
	In normal operation, ZFS will try to write this amount of data to each disk
	before moving on to the next top-level vdev.
	.
	.It Sy metaslab_bias_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Enable metaslab group biasing based on their vdevs' over- or under-utilization
	relative to the pool.
	.
	.It Sy metaslab_force_ganging Ns = Ns Sy 16777217 Ns B Po 16 MiB + 1 B Pc Pq u64
	Make some blocks above a certain size be gang blocks.
	This option is used by the test suite to facilitate testing.
	.
	.It Sy metaslab_force_ganging_pct Ns = Ns Sy 3 Ns % Pq uint
	For blocks that could be forced to be a gang block (due to
	.Sy metaslab_force_ganging ) ,
	force this many of them to be gang blocks.
	.
	.It Sy brt_zap_prefetch Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Controls prefetching BRT records for blocks which are going to be cloned.
	.
	.It Sy brt_zap_default_bs Ns = Ns Sy 12 Po 4 KiB Pc Pq int
	Default BRT ZAP data block size as a power of 2. Note that changing this after
	creating a BRT on the pool will not affect existing BRTs, only newly created
	ones.
	.
	.It Sy brt_zap_default_ibs Ns = Ns Sy 12 Po 4 KiB Pc Pq int
	Default BRT ZAP indirect block size as a power of 2. Note that changing this
	after creating a BRT on the pool will not affect existing BRTs, only newly
	created ones.
	.
	.It Sy ddt_zap_default_bs Ns = Ns Sy 15 Po 32 KiB Pc Pq int
	Default DDT ZAP data block size as a power of 2. Note that changing this after
	creating a DDT on the pool will not affect existing DDTs, only newly created
	ones.
	.
	.It Sy ddt_zap_default_ibs Ns = Ns Sy 15 Po 32 KiB Pc Pq int
	Default DDT ZAP indirect block size as a power of 2. Note that changing this
	after creating a DDT on the pool will not affect existing DDTs, only newly
	created ones.
	.
	.It Sy zfs_default_bs Ns = Ns Sy 9 Po 512 B Pc Pq int
	Default dnode block size as a power of 2.
	.
	.It Sy zfs_default_ibs Ns = Ns Sy 17 Po 128 KiB Pc Pq int
	Default dnode indirect block size as a power of 2.
	.
	.It Sy zfs_history_output_max Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq u64
	When attempting to log an output nvlist of an ioctl in the on-disk history,
	the output will not be stored if it is larger than this size (in bytes).
	This must be less than
	.Sy DMU_MAX_ACCESS Pq 64 MiB .
	This applies primarily to
	.Fn zfs_ioc_channel_program Pq cf. Xr zfs-program 8 .
	.
	.It Sy zfs_keep_log_spacemaps_at_export Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Prevent log spacemaps from being destroyed during pool exports and destroys.
	.
	.It Sy zfs_metaslab_segment_weight_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Enable/disable segment-based metaslab selection.
	.
	.It Sy zfs_metaslab_switch_threshold Ns = Ns Sy 2 Pq int
	When using segment-based metaslab selection, continue allocating
	from the active metaslab until this option's
	worth of buckets have been exhausted.
	.
	.It Sy metaslab_debug_load Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Load all metaslabs during pool import.
	.
	.It Sy metaslab_debug_unload Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Prevent metaslabs from being unloaded.
	.
	.It Sy metaslab_fragmentation_factor_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Enable use of the fragmentation metric in computing metaslab weights.
	.
	.It Sy metaslab_df_max_search Ns = Ns Sy 16777216 Ns B Po 16 MiB Pc Pq uint
	Maximum distance to search forward from the last offset.
	Without this limit, fragmented pools can see
	.Em >100`000
	iterations and
	.Fn metaslab_block_picker
	becomes the performance limiting factor on high-performance storage.
	.Pp
	With the default setting of
	.Sy 16 MiB ,
	we typically see less than
	.Em 500
	iterations, even with very fragmented
	.Sy ashift Ns = Ns Sy 9
	pools.
	The maximum number of iterations possible is
	.Sy metaslab_df_max_search / 2^(ashift+1) .
	With the default setting of
	.Sy 16 MiB
	this is
	.Em 16*1024 Pq with Sy ashift Ns = Ns Sy 9
	or
	.Em 2*1024 Pq with Sy ashift Ns = Ns Sy 12 .
	.
	.It Sy metaslab_df_use_largest_segment Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	If not searching forward (due to
	.Sy metaslab_df_max_search , metaslab_df_free_pct ,
	.No or Sy metaslab_df_alloc_threshold ) ,
	this tunable controls which segment is used.
	If set, we will use the largest free segment.
	If unset, we will use a segment of at least the requested size.
	.
	.It Sy zfs_metaslab_max_size_cache_sec Ns = Ns Sy 3600 Ns s Po 1 hour Pc Pq u64
	When we unload a metaslab, we cache the size of the largest free chunk.
	We use that cached size to determine whether or not to load a metaslab
	for a given allocation.
	As more frees accumulate in that metaslab while it's unloaded,
	the cached max size becomes less and less accurate.
	After a number of seconds controlled by this tunable,
	we stop considering the cached max size and start
	considering only the histogram instead.
	.
	.It Sy zfs_metaslab_mem_limit Ns = Ns Sy 25 Ns % Pq uint
	When we are loading a new metaslab, we check the amount of memory being used
	to store metaslab range trees.
	If it is over a threshold, we attempt to unload the least recently used metaslab
	to prevent the system from clogging all of its memory with range trees.
	This tunable sets the percentage of total system memory that is the threshold.
	.
	.It Sy zfs_metaslab_try_hard_before_gang Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	.Bl -item -compact
	.It
	If unset, we will first try normal allocation.
	.It
	If that fails then we will do a gang allocation.
	.It
	If that fails then we will do a "try hard" gang allocation.
	.It
	If that fails then we will have a multi-layer gang block.
	.El
	.Pp
	.Bl -item -compact
	.It
	If set, we will first try normal allocation.
	.It
	If that fails then we will do a "try hard" allocation.
	.It
	If that fails we will do a gang allocation.
	.It
	If that fails we will do a "try hard" gang allocation.
	.It
	If that fails then we will have a multi-layer gang block.
	.El
	.
	.It Sy zfs_metaslab_find_max_tries Ns = Ns Sy 100 Pq uint
	When not trying hard, we only consider this number of the best metaslabs.
	This improves performance, especially when there are many metaslabs per vdev
	and the allocation can't actually be satisfied
	(so we would otherwise iterate all metaslabs).
	.
	.It Sy zfs_vdev_default_ms_count Ns = Ns Sy 200 Pq uint
	When a vdev is added, target this number of metaslabs per top-level vdev.
	.
	.It Sy zfs_vdev_default_ms_shift Ns = Ns Sy 29 Po 512 MiB Pc Pq uint
	Default lower limit for metaslab size.
	.
	.It Sy zfs_vdev_max_ms_shift Ns = Ns Sy 34 Po 16 GiB Pc Pq uint
	Default upper limit for metaslab size.
	.
	.It Sy zfs_vdev_max_auto_ashift Ns = Ns Sy 14 Pq uint
	Maximum ashift used when optimizing for logical \[->] physical sector size on
	new
	top-level vdevs.
	May be increased up to
	.Sy ASHIFT_MAX Po 16 Pc ,
	but this may negatively impact pool space efficiency.
	.
	.It Sy zfs_vdev_min_auto_ashift Ns = Ns Sy ASHIFT_MIN Po 9 Pc Pq uint
	Minimum ashift used when creating new top-level vdevs.
	.
	.It Sy zfs_vdev_min_ms_count Ns = Ns Sy 16 Pq uint
	Minimum number of metaslabs to create in a top-level vdev.
	.
	.It Sy vdev_validate_skip Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Skip label validation steps during pool import.
	Changing is not recommended unless you know what you're doing
	and are recovering a damaged label.
	.
	.It Sy zfs_vdev_ms_count_limit Ns = Ns Sy 131072 Po 128k Pc Pq uint
	Practical upper limit of total metaslabs per top-level vdev.
	.
	.It Sy metaslab_preload_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Enable metaslab group preloading.
	.
	.It Sy metaslab_preload_limit Ns = Ns Sy 10 Pq uint
	Maximum number of metaslabs per group to preload
	.
	.It Sy metaslab_preload_pct Ns = Ns Sy 50 Pq uint
	Percentage of CPUs to run a metaslab preload taskq
	.
	.It Sy metaslab_lba_weighting_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Give more weight to metaslabs with lower LBAs,
	assuming they have greater bandwidth,
	as is typically the case on a modern constant angular velocity disk drive.
	.
	.It Sy metaslab_unload_delay Ns = Ns Sy 32 Pq uint
	After a metaslab is used, we keep it loaded for this many TXGs, to attempt to
	reduce unnecessary reloading.
	Note that both this many TXGs and
	.Sy metaslab_unload_delay_ms
	milliseconds must pass before unloading will occur.
	.
	.It Sy metaslab_unload_delay_ms Ns = Ns Sy 600000 Ns ms Po 10 min Pc Pq uint
	After a metaslab is used, we keep it loaded for this many milliseconds,
	to attempt to reduce unnecessary reloading.
	Note, that both this many milliseconds and
	.Sy metaslab_unload_delay
	TXGs must pass before unloading will occur.
	.
	.It Sy reference_history Ns = Ns Sy 3 Pq uint
	Maximum reference holders being tracked when reference_tracking_enable is
	active.
	.It Sy raidz_expand_max_copy_bytes Ns = Ns Sy 160MB Pq ulong
	Max amount of memory to use for RAID-Z expansion I/O.
	This limits how much I/O can be outstanding at once.
	.
	.It Sy raidz_expand_max_reflow_bytes Ns = Ns Sy 0 Pq ulong
	For testing, pause RAID-Z expansion when reflow amount reaches this value.
	.
	.It Sy raidz_io_aggregate_rows Ns = Ns Sy 4 Pq ulong
	For expanded RAID-Z, aggregate reads that have more rows than this.
	.
	.It Sy reference_history Ns = Ns Sy 3 Pq int
	Maximum reference holders being tracked when reference_tracking_enable is
	active.
	.
	.It Sy reference_tracking_enable Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Track reference holders to
	.Sy refcount_t
	objects (debug builds only).
	.
	.It Sy send_holes_without_birth_time Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	When set, the
	.Sy hole_birth
	optimization will not be used, and all holes will always be sent during a
	.Nm zfs Cm send .
	This is useful if you suspect your datasets are affected by a bug in
	.Sy hole_birth .
	.
	.It Sy spa_config_path Ns = Ns Pa /etc/zfs/zpool.cache Pq charp
	SPA config file.
	.
	.It Sy spa_asize_inflation Ns = Ns Sy 24 Pq uint
	Multiplication factor used to estimate actual disk consumption from the
	size of data being written.
	The default value is a worst case estimate,
	but lower values may be valid for a given pool depending on its configuration.
	Pool administrators who understand the factors involved
	may wish to specify a more realistic inflation factor,
	particularly if they operate close to quota or capacity limits.
	.
	.It Sy spa_load_print_vdev_tree Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Whether to print the vdev tree in the debugging message buffer during pool
	import.
	.
	.It Sy spa_load_verify_data Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Whether to traverse data blocks during an "extreme rewind"
	.Pq Fl X
	import.
	.Pp
	An extreme rewind import normally performs a full traversal of all
	blocks in the pool for verification.
	If this parameter is unset, the traversal skips non-metadata blocks.
	It can be toggled once the
	import has started to stop or start the traversal of non-metadata blocks.
	.
	.It Sy spa_load_verify_metadata Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Whether to traverse blocks during an "extreme rewind"
	.Pq Fl X
	pool import.
	.Pp
	An extreme rewind import normally performs a full traversal of all
	blocks in the pool for verification.
	If this parameter is unset, the traversal is not performed.
	It can be toggled once the import has started to stop or start the traversal.
	.
	.It Sy spa_load_verify_shift Ns = Ns Sy 4 Po 1/16th Pc Pq uint
	Sets the maximum number of bytes to consume during pool import to the log2
	fraction of the target ARC size.
	.
	.It Sy spa_slop_shift Ns = Ns Sy 5 Po 1/32nd Pc Pq int
	Normally, we don't allow the last
	.Sy 3.2% Pq Sy 1/2^spa_slop_shift
	of space in the pool to be consumed.
	This ensures that we don't run the pool completely out of space,
	due to unaccounted changes (e.g. to the MOS).
	It also limits the worst-case time to allocate space.
	If we have less than this amount of free space,
	most ZPL operations (e.g. write, create) will return
	.Sy ENOSPC .
	.
	.It Sy spa_num_allocators Ns = Ns Sy 4 Pq int
	Determines the number of block alloctators to use per spa instance.
	Capped by the number of actual CPUs in the system via
	.Sy spa_cpus_per_allocator .
	.Pp
	Note that setting this value too high could result in performance
	degredation and/or excess fragmentation.
	Set value only applies to pools imported/created after that.
	.
	.It Sy spa_cpus_per_allocator Ns = Ns Sy 4 Pq int
	Determines the minimum number of CPUs in a system for block alloctator
	per spa instance.
	Set value only applies to pools imported/created after that.
	.
	.It Sy spa_upgrade_errlog_limit Ns = Ns Sy 0 Pq uint
	Limits the number of on-disk error log entries that will be converted to the
	new format when enabling the
	.Sy head_errlog
	feature.
	The default is to convert all log entries.
	.
	.It Sy vdev_removal_max_span Ns = Ns Sy 32768 Ns B Po 32 KiB Pc Pq uint
	During top-level vdev removal, chunks of data are copied from the vdev
	which may include free space in order to trade bandwidth for IOPS.
	This parameter determines the maximum span of free space, in bytes,
	which will be included as "unnecessary" data in a chunk of copied data.
	.Pp
	The default value here was chosen to align with
	.Sy zfs_vdev_read_gap_limit ,
	which is a similar concept when doing
	regular reads (but there's no reason it has to be the same).
	.
	.It Sy vdev_file_logical_ashift Ns = Ns Sy 9 Po 512 B Pc Pq u64
	Logical ashift for file-based devices.
	.
	.It Sy vdev_file_physical_ashift Ns = Ns Sy 9 Po 512 B Pc Pq u64
	Physical ashift for file-based devices.
	.
	.It Sy zap_iterate_prefetch Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	If set, when we start iterating over a ZAP object,
	prefetch the entire object (all leaf blocks).
	However, this is limited by
	.Sy dmu_prefetch_max .
	.
	.It Sy zap_micro_max_size Ns = Ns Sy 131072 Ns B Po 128 KiB Pc Pq int
	Maximum micro ZAP size.
	A micro ZAP is upgraded to a fat ZAP, once it grows beyond the specified size.
	.
	.It Sy zap_shrink_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	If set, adjacent empty ZAP blocks will be collapsed, reducing disk space.
	.
	.It Sy zfetch_min_distance Ns = Ns Sy 4194304 Ns B Po 4 MiB Pc Pq uint
	Min bytes to prefetch per stream.
	Prefetch distance starts from the demand access size and quickly grows to
	this value, doubling on each hit.
	After that it may grow further by 1/8 per hit, but only if some prefetch
	since last time haven't completed in time to satisfy demand request, i.e.
	prefetch depth didn't cover the read latency or the pool got saturated.
	.
	.It Sy zfetch_max_distance Ns = Ns Sy 67108864 Ns B Po 64 MiB Pc Pq uint
	Max bytes to prefetch per stream.
	.
	.It Sy zfetch_max_idistance Ns = Ns Sy 67108864 Ns B Po 64 MiB Pc Pq uint
	Max bytes to prefetch indirects for per stream.
	.
	.It Sy zfetch_max_reorder Ns = Ns Sy 16777216 Ns B Po 16 MiB Pc Pq uint
	Requests within this byte distance from the current prefetch stream position
	are considered parts of the stream, reordered due to parallel processing.
	Such requests do not advance the stream position immediately unless
	.Sy zfetch_hole_shift
	fill threshold is reached, but saved to fill holes in the stream later.
	.
	.It Sy zfetch_max_streams Ns = Ns Sy 8 Pq uint
	Max number of streams per zfetch (prefetch streams per file).
	.
	.It Sy zfetch_min_sec_reap Ns = Ns Sy 1 Pq uint
	Min time before inactive prefetch stream can be reclaimed
	.
	.It Sy zfetch_max_sec_reap Ns = Ns Sy 2 Pq uint
	Max time before inactive prefetch stream can be deleted
	.
	.It Sy zfs_abd_scatter_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Enables ARC from using scatter/gather lists and forces all allocations to be
	linear in kernel memory.
	Disabling can improve performance in some code paths
	at the expense of fragmented kernel memory.
	.
	.It Sy zfs_abd_scatter_max_order Ns = Ns Sy MAX_ORDER\-1 Pq uint
	Maximum number of consecutive memory pages allocated in a single block for
	scatter/gather lists.
	.Pp
	The value of
	.Sy MAX_ORDER
	depends on kernel configuration.
	.
	.It Sy zfs_abd_scatter_min_size Ns = Ns Sy 1536 Ns B Po 1.5 KiB Pc Pq uint
	This is the minimum allocation size that will use scatter (page-based) ABDs.
	Smaller allocations will use linear ABDs.
	.
	.It Sy zfs_arc_dnode_limit Ns = Ns Sy 0 Ns B Pq u64
	When the number of bytes consumed by dnodes in the ARC exceeds this number of
	bytes, try to unpin some of it in response to demand for non-metadata.
	This value acts as a ceiling to the amount of dnode metadata, and defaults to
	.Sy 0 ,
	which indicates that a percent which is based on
	.Sy zfs_arc_dnode_limit_percent
	of the ARC meta buffers that may be used for dnodes.
	.It Sy zfs_arc_dnode_limit_percent Ns = Ns Sy 10 Ns % Pq u64
	Percentage that can be consumed by dnodes of ARC meta buffers.
	.Pp
	See also
	.Sy zfs_arc_dnode_limit ,
	which serves a similar purpose but has a higher priority if nonzero.
	.
	.It Sy zfs_arc_dnode_reduce_percent Ns = Ns Sy 10 Ns % Pq u64
	Percentage of ARC dnodes to try to scan in response to demand for non-metadata
	when the number of bytes consumed by dnodes exceeds
	.Sy zfs_arc_dnode_limit .
	.
	.It Sy zfs_arc_average_blocksize Ns = Ns Sy 8192 Ns B Po 8 KiB Pc Pq uint
	The ARC's buffer hash table is sized based on the assumption of an average
	block size of this value.
	This works out to roughly 1 MiB of hash table per 1 GiB of physical memory
	with 8-byte pointers.
	For configurations with a known larger average block size,
	this value can be increased to reduce the memory footprint.
	.
	.It Sy zfs_arc_eviction_pct Ns = Ns Sy 200 Ns % Pq uint
	When
	.Fn arc_is_overflowing ,
	.Fn arc_get_data_impl
	waits for this percent of the requested amount of data to be evicted.
	For example, by default, for every
	.Em 2 KiB
	that's evicted,
	.Em 1 KiB
	of it may be "reused" by a new allocation.
	Since this is above
	.Sy 100 Ns % ,
	it ensures that progress is made towards getting
	.Sy arc_size No under Sy arc_c .
	Since this is finite, it ensures that allocations can still happen,
	even during the potentially long time that
	.Sy arc_size No is more than Sy arc_c .
	.
	.It Sy zfs_arc_evict_batch_limit Ns = Ns Sy 10 Pq uint
	Number ARC headers to evict per sub-list before proceeding to another sub-list.
	This batch-style operation prevents entire sub-lists from being evicted at once
	but comes at a cost of additional unlocking and locking.
	.
	.It Sy zfs_arc_grow_retry Ns = Ns Sy 0 Ns s Pq uint
	If set to a non zero value, it will replace the
	.Sy arc_grow_retry
	value with this value.
	The
	.Sy arc_grow_retry
	.No value Pq default Sy 5 Ns s
	is the number of seconds the ARC will wait before
	trying to resume growth after a memory pressure event.
	.
	.It Sy zfs_arc_lotsfree_percent Ns = Ns Sy 10 Ns % Pq int
	Throttle I/O when free system memory drops below this percentage of total
	system memory.
	Setting this value to
	.Sy 0
	will disable the throttle.
	.
	.It Sy zfs_arc_max Ns = Ns Sy 0 Ns B Pq u64
	Max size of ARC in bytes.
	If
	.Sy 0 ,
	then the max size of ARC is determined by the amount of system memory installed.
	The larger of
	.Sy all_system_memory No \- Sy 1 GiB
	and
	.Sy 5/8 No \(mu Sy all_system_memory
	will be used as the limit.
	This value must be at least
	.Sy 67108864 Ns B Pq 64 MiB .
	.Pp
	This value can be changed dynamically, with some caveats.
	It cannot be set back to
	.Sy 0
	while running, and reducing it below the current ARC size will not cause
	the ARC to shrink without memory pressure to induce shrinking.
	.
	.It Sy zfs_arc_meta_balance Ns = Ns Sy 500 Pq uint
	Balance between metadata and data on ghost hits.
	Values above 100 increase metadata caching by proportionally reducing effect
	of ghost data hits on target data/metadata rate.
	.
	.It Sy zfs_arc_min Ns = Ns Sy 0 Ns B Pq u64
	Min size of ARC in bytes.
	.No If set to Sy 0 , arc_c_min
	will default to consuming the larger of
	.Sy 32 MiB
	and
	.Sy all_system_memory No / Sy 32 .
	.
	.It Sy zfs_arc_min_prefetch_ms Ns = Ns Sy 0 Ns ms Ns Po Ns ≡ Ns 1s Pc Pq uint
	Minimum time prefetched blocks are locked in the ARC.
	.
	.It Sy zfs_arc_min_prescient_prefetch_ms Ns = Ns Sy 0 Ns ms Ns Po Ns ≡ Ns 6s Pc Pq uint
	Minimum time "prescient prefetched" blocks are locked in the ARC.
	These blocks are meant to be prefetched fairly aggressively ahead of
	the code that may use them.
	.
	.It Sy zfs_arc_prune_task_threads Ns = Ns Sy 1 Pq int
	Number of arc_prune threads.
	.Fx
	does not need more than one.
	Linux may theoretically use one per mount point up to number of CPUs,
	but that was not proven to be useful.
	.
	.It Sy zfs_max_missing_tvds Ns = Ns Sy 0 Pq int
	Number of missing top-level vdevs which will be allowed during
	pool import (only in read-only mode).
	.
	.It Sy zfs_max_nvlist_src_size Ns = Sy 0 Pq u64
	Maximum size in bytes allowed to be passed as
	.Sy zc_nvlist_src_size
	for ioctls on
	.Pa /dev/zfs .
	This prevents a user from causing the kernel to allocate
	an excessive amount of memory.
	When the limit is exceeded, the ioctl fails with
	.Sy EINVAL
	and a description of the error is sent to the
	.Pa zfs-dbgmsg
	log.
	This parameter should not need to be touched under normal circumstances.
	If
	.Sy 0 ,
	equivalent to a quarter of the user-wired memory limit under
	.Fx
	and to
	.Sy 134217728 Ns B Pq 128 MiB
	under Linux.
	.
	.It Sy zfs_multilist_num_sublists Ns = Ns Sy 0 Pq uint
	To allow more fine-grained locking, each ARC state contains a series
	of lists for both data and metadata objects.
	Locking is performed at the level of these "sub-lists".
	This parameters controls the number of sub-lists per ARC state,
	and also applies to other uses of the multilist data structure.
	.Pp
	If
	.Sy 0 ,
	equivalent to the greater of the number of online CPUs and
	.Sy 4 .
	.
	.It Sy zfs_arc_overflow_shift Ns = Ns Sy 8 Pq int
	The ARC size is considered to be overflowing if it exceeds the current
	ARC target size
	.Pq Sy arc_c
	by thresholds determined by this parameter.
	Exceeding by
	.Sy ( arc_c No >> Sy zfs_arc_overflow_shift ) No / Sy 2
	starts ARC reclamation process.
	If that appears insufficient, exceeding by
	.Sy ( arc_c No >> Sy zfs_arc_overflow_shift ) No \(mu Sy 1.5
	blocks new buffer allocation until the reclaim thread catches up.
	Started reclamation process continues till ARC size returns below the
	target size.
	.Pp
	The default value of
	.Sy 8
	causes the ARC to start reclamation if it exceeds the target size by
	.Em 0.2%
	of the target size, and block allocations by
	.Em 0.6% .
	.
	.It Sy zfs_arc_shrink_shift Ns = Ns Sy 0 Pq uint
	If nonzero, this will update
	.Sy arc_shrink_shift Pq default Sy 7
	with the new value.
	.
	.It Sy zfs_arc_pc_percent Ns = Ns Sy 0 Ns % Po off Pc Pq uint
	Percent of pagecache to reclaim ARC to.
	.Pp
	This tunable allows the ZFS ARC to play more nicely
	with the kernel's LRU pagecache.
	It can guarantee that the ARC size won't collapse under scanning
	pressure on the pagecache, yet still allows the ARC to be reclaimed down to
	.Sy zfs_arc_min
	if necessary.
	This value is specified as percent of pagecache size (as measured by
	.Sy NR_FILE_PAGES ) ,
	where that percent may exceed
	.Sy 100 .
	This
	only operates during memory pressure/reclaim.
	.
	.It Sy zfs_arc_shrinker_limit Ns = Ns Sy 10000 Pq int
	This is a limit on how many pages the ARC shrinker makes available for
	eviction in response to one page allocation attempt.
	Note that in practice, the kernel's shrinker can ask us to evict
	up to about four times this for one allocation attempt.
	.Pp
	The default limit of
	.Sy 10000 Pq in practice, Em 160 MiB No per allocation attempt with 4 KiB pages
	limits the amount of time spent attempting to reclaim ARC memory to
	less than 100 ms per allocation attempt,
	even with a small average compressed block size of ~8 KiB.
	.Pp
	The parameter can be set to 0 (zero) to disable the limit,
	and only applies on Linux.
	.
	.It Sy zfs_arc_sys_free Ns = Ns Sy 0 Ns B Pq u64
	The target number of bytes the ARC should leave as free memory on the system.
	If zero, equivalent to the bigger of
	.Sy 512 KiB No and Sy all_system_memory/64 .
	.
	.It Sy zfs_autoimport_disable Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Disable pool import at module load by ignoring the cache file
	.Pq Sy spa_config_path .
	.
	.It Sy zfs_checksum_events_per_second Ns = Ns Sy 20 Ns /s Pq uint
	Rate limit checksum events to this many per second.
	Note that this should not be set below the ZED thresholds
	(currently 10 checksums over 10 seconds)
	or else the daemon may not trigger any action.
	.
	.It Sy zfs_commit_timeout_pct Ns = Ns Sy 10 Ns % Pq uint
	This controls the amount of time that a ZIL block (lwb) will remain "open"
	when it isn't "full", and it has a thread waiting for it to be committed to
	stable storage.
	The timeout is scaled based on a percentage of the last lwb
	latency to avoid significantly impacting the latency of each individual
	transaction record (itx).
	.
	.It Sy zfs_condense_indirect_commit_entry_delay_ms Ns = Ns Sy 0 Ns ms Pq int
	Vdev indirection layer (used for device removal) sleeps for this many
	milliseconds during mapping generation.
	Intended for use with the test suite to throttle vdev removal speed.
	.
	.It Sy zfs_condense_indirect_obsolete_pct Ns = Ns Sy 25 Ns % Pq uint
	Minimum percent of obsolete bytes in vdev mapping required to attempt to
	condense
	.Pq see Sy zfs_condense_indirect_vdevs_enable .
	Intended for use with the test suite
	to facilitate triggering condensing as needed.
	.
	.It Sy zfs_condense_indirect_vdevs_enable Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Enable condensing indirect vdev mappings.
	When set, attempt to condense indirect vdev mappings
	if the mapping uses more than
	.Sy zfs_condense_min_mapping_bytes
	bytes of memory and if the obsolete space map object uses more than
	.Sy zfs_condense_max_obsolete_bytes
	bytes on-disk.
	The condensing process is an attempt to save memory by removing obsolete
	mappings.
	.
	.It Sy zfs_condense_max_obsolete_bytes Ns = Ns Sy 1073741824 Ns B Po 1 GiB Pc Pq u64
	Only attempt to condense indirect vdev mappings if the on-disk size
	of the obsolete space map object is greater than this number of bytes
	.Pq see Sy zfs_condense_indirect_vdevs_enable .
	.
	.It Sy zfs_condense_min_mapping_bytes Ns = Ns Sy 131072 Ns B Po 128 KiB Pc Pq u64
	Minimum size vdev mapping to attempt to condense
	.Pq see Sy zfs_condense_indirect_vdevs_enable .
	.
	.It Sy zfs_dbgmsg_enable Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Internally ZFS keeps a small log to facilitate debugging.
	The log is enabled by default, and can be disabled by unsetting this option.
	The contents of the log can be accessed by reading
	.Pa /proc/spl/kstat/zfs/dbgmsg .
	Writing
	.Sy 0
	to the file clears the log.
	.Pp
	This setting does not influence debug prints due to
	.Sy zfs_flags .
	.
	.It Sy zfs_dbgmsg_maxsize Ns = Ns Sy 4194304 Ns B Po 4 MiB Pc Pq uint
	Maximum size of the internal ZFS debug log.
	.
	.It Sy zfs_dbuf_state_index Ns = Ns Sy 0 Pq int
	Historically used for controlling what reporting was available under
	.Pa /proc/spl/kstat/zfs .
	No effect.
	.
	+.It Sy zfs_deadman_checktime_ms Ns = Ns Sy 60000 Ns ms Po 1 min Pc Pq u64
	+Check time in milliseconds.
	+This defines the frequency at which we check for hung I/O requests
	+and potentially invoke the
	+.Sy zfs_deadman_failmode
	+behavior.
	+.
	.It Sy zfs_deadman_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	When a pool sync operation takes longer than
	.Sy zfs_deadman_synctime_ms ,
	or when an individual I/O operation takes longer than
	.Sy zfs_deadman_ziotime_ms ,
	then the operation is considered to be "hung".
	If
	.Sy zfs_deadman_enabled
	is set, then the deadman behavior is invoked as described by
	.Sy zfs_deadman_failmode .
	By default, the deadman is enabled and set to
	.Sy wait
	which results in "hung" I/O operations only being logged.
	The deadman is automatically disabled when a pool gets suspended.
	.
	+.It Sy zfs_deadman_events_per_second Ns = Ns Sy 1 Ns /s Pq int
	+Rate limit deadman zevents (which report hung I/O operations) to this many per
	+second.
	+.
	.It Sy zfs_deadman_failmode Ns = Ns Sy wait Pq charp
	Controls the failure behavior when the deadman detects a "hung" I/O operation.
	Valid values are:
	.Bl -tag -compact -offset 4n -width "continue"
	.It Sy wait
	Wait for a "hung" operation to complete.
	For each "hung" operation a "deadman" event will be posted
	describing that operation.
	.It Sy continue
	Attempt to recover from a "hung" operation by re-dispatching it
	to the I/O pipeline if possible.
	.It Sy panic
	Panic the system.
	This can be used to facilitate automatic fail-over
	to a properly configured fail-over partner.
	.El
	.
	-.It Sy zfs_deadman_checktime_ms Ns = Ns Sy 60000 Ns ms Po 1 min Pc Pq u64
	-Check time in milliseconds.
	-This defines the frequency at which we check for hung I/O requests
	-and potentially invoke the
	-.Sy zfs_deadman_failmode
	-behavior.
	-.
	.It Sy zfs_deadman_synctime_ms Ns = Ns Sy 600000 Ns ms Po 10 min Pc Pq u64
	Interval in milliseconds after which the deadman is triggered and also
	the interval after which a pool sync operation is considered to be "hung".
	Once this limit is exceeded the deadman will be invoked every
	.Sy zfs_deadman_checktime_ms
	milliseconds until the pool sync completes.
	.
	.It Sy zfs_deadman_ziotime_ms Ns = Ns Sy 300000 Ns ms Po 5 min Pc Pq u64
	Interval in milliseconds after which the deadman is triggered and an
	individual I/O operation is considered to be "hung".
	As long as the operation remains "hung",
	the deadman will be invoked every
	.Sy zfs_deadman_checktime_ms
	milliseconds until the operation completes.
	.
	.It Sy zfs_dedup_prefetch Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Enable prefetching dedup-ed blocks which are going to be freed.
	.
	.It Sy zfs_delay_min_dirty_percent Ns = Ns Sy 60 Ns % Pq uint
	Start to delay each transaction once there is this amount of dirty data,
	expressed as a percentage of
	.Sy zfs_dirty_data_max .
	This value should be at least
	.Sy zfs_vdev_async_write_active_max_dirty_percent .
	.No See Sx ZFS TRANSACTION DELAY .
	.
	.It Sy zfs_delay_scale Ns = Ns Sy 500000 Pq int
	This controls how quickly the transaction delay approaches infinity.
	Larger values cause longer delays for a given amount of dirty data.
	.Pp
	For the smoothest delay, this value should be about 1 billion divided
	by the maximum number of operations per second.
	This will smoothly handle between ten times and a tenth of this number.
	.No See Sx ZFS TRANSACTION DELAY .
	.Pp
	.Sy zfs_delay_scale No \(mu Sy zfs_dirty_data_max Em must No be smaller than Sy 2^64 .
	.
	.It Sy zfs_disable_ivset_guid_check Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Disables requirement for IVset GUIDs to be present and match when doing a raw
	receive of encrypted datasets.
	Intended for users whose pools were created with
	OpenZFS pre-release versions and now have compatibility issues.
	.
	.It Sy zfs_key_max_salt_uses Ns = Ns Sy 400000000 Po 4*10^8 Pc Pq ulong
	Maximum number of uses of a single salt value before generating a new one for
	encrypted datasets.
	The default value is also the maximum.
	.
	.It Sy zfs_object_mutex_size Ns = Ns Sy 64 Pq uint
	Size of the znode hashtable used for holds.
	.Pp
	Due to the need to hold locks on objects that may not exist yet, kernel mutexes
	are not created per-object and instead a hashtable is used where collisions
	will result in objects waiting when there is not actually contention on the
	same object.
	.
	.It Sy zfs_slow_io_events_per_second Ns = Ns Sy 20 Ns /s Pq int
	-Rate limit delay and deadman zevents (which report slow I/O operations) to this
	-many per
	+Rate limit delay zevents (which report slow I/O operations) to this many per
	second.
	.
	.It Sy zfs_unflushed_max_mem_amt Ns = Ns Sy 1073741824 Ns B Po 1 GiB Pc Pq u64
	Upper-bound limit for unflushed metadata changes to be held by the
	log spacemap in memory, in bytes.
	.
	.It Sy zfs_unflushed_max_mem_ppm Ns = Ns Sy 1000 Ns ppm Po 0.1% Pc Pq u64
	Part of overall system memory that ZFS allows to be used
	for unflushed metadata changes by the log spacemap, in millionths.
	.
	.It Sy zfs_unflushed_log_block_max Ns = Ns Sy 131072 Po 128k Pc Pq u64
	Describes the maximum number of log spacemap blocks allowed for each pool.
	The default value means that the space in all the log spacemaps
	can add up to no more than
	.Sy 131072
	blocks (which means
	.Em 16 GiB
	of logical space before compression and ditto blocks,
	assuming that blocksize is
	.Em 128 KiB ) .
	.Pp
	This tunable is important because it involves a trade-off between import
	time after an unclean export and the frequency of flushing metaslabs.
	The higher this number is, the more log blocks we allow when the pool is
	active which means that we flush metaslabs less often and thus decrease
	the number of I/O operations for spacemap updates per TXG.
	At the same time though, that means that in the event of an unclean export,
	there will be more log spacemap blocks for us to read, inducing overhead
	in the import time of the pool.
	The lower the number, the amount of flushing increases, destroying log
	blocks quicker as they become obsolete faster, which leaves less blocks
	to be read during import time after a crash.
	.Pp
	Each log spacemap block existing during pool import leads to approximately
	one extra logical I/O issued.
	This is the reason why this tunable is exposed in terms of blocks rather
	than space used.
	.
	.It Sy zfs_unflushed_log_block_min Ns = Ns Sy 1000 Pq u64
	If the number of metaslabs is small and our incoming rate is high,
	we could get into a situation that we are flushing all our metaslabs every TXG.
	Thus we always allow at least this many log blocks.
	.
	.It Sy zfs_unflushed_log_block_pct Ns = Ns Sy 400 Ns % Pq u64
	Tunable used to determine the number of blocks that can be used for
	the spacemap log, expressed as a percentage of the total number of
	unflushed metaslabs in the pool.
	.
	.It Sy zfs_unflushed_log_txg_max Ns = Ns Sy 1000 Pq u64
	Tunable limiting maximum time in TXGs any metaslab may remain unflushed.
	It effectively limits maximum number of unflushed per-TXG spacemap logs
	that need to be read after unclean pool export.
	.
	.It Sy zfs_unlink_suspend_progress Ns = Ns Sy 0 Ns \| Ns 1 Pq uint
	When enabled, files will not be asynchronously removed from the list of pending
	unlinks and the space they consume will be leaked.
	Once this option has been disabled and the dataset is remounted,
	the pending unlinks will be processed and the freed space returned to the pool.
	This option is used by the test suite.
	.
	.It Sy zfs_delete_blocks Ns = Ns Sy 20480 Pq ulong
	This is the used to define a large file for the purposes of deletion.
	Files containing more than
	.Sy zfs_delete_blocks
	will be deleted asynchronously, while smaller files are deleted synchronously.
	Decreasing this value will reduce the time spent in an
	.Xr unlink 2
	system call, at the expense of a longer delay before the freed space is
	available.
	This only applies on Linux.
	.
	.It Sy zfs_dirty_data_max Ns = Pq int
	Determines the dirty space limit in bytes.
	Once this limit is exceeded, new writes are halted until space frees up.
	This parameter takes precedence over
	.Sy zfs_dirty_data_max_percent .
	.No See Sx ZFS TRANSACTION DELAY .
	.Pp
	Defaults to
	.Sy physical_ram/10 ,
	capped at
	.Sy zfs_dirty_data_max_max .
	.
	.It Sy zfs_dirty_data_max_max Ns = Pq int
	Maximum allowable value of
	.Sy zfs_dirty_data_max ,
	expressed in bytes.
	This limit is only enforced at module load time, and will be ignored if
	.Sy zfs_dirty_data_max
	is later changed.
	This parameter takes precedence over
	.Sy zfs_dirty_data_max_max_percent .
	.No See Sx ZFS TRANSACTION DELAY .
	.Pp
	Defaults to
	.Sy min(physical_ram/4, 4GiB) ,
	or
	.Sy min(physical_ram/4, 1GiB)
	for 32-bit systems.
	.
	.It Sy zfs_dirty_data_max_max_percent Ns = Ns Sy 25 Ns % Pq uint
	Maximum allowable value of
	.Sy zfs_dirty_data_max ,
	expressed as a percentage of physical RAM.
	This limit is only enforced at module load time, and will be ignored if
	.Sy zfs_dirty_data_max
	is later changed.
	The parameter
	.Sy zfs_dirty_data_max_max
	takes precedence over this one.
	.No See Sx ZFS TRANSACTION DELAY .
	.
	.It Sy zfs_dirty_data_max_percent Ns = Ns Sy 10 Ns % Pq uint
	Determines the dirty space limit, expressed as a percentage of all memory.
	Once this limit is exceeded, new writes are halted until space frees up.
	The parameter
	.Sy zfs_dirty_data_max
	takes precedence over this one.
	.No See Sx ZFS TRANSACTION DELAY .
	.Pp
	Subject to
	.Sy zfs_dirty_data_max_max .
	.
	.It Sy zfs_dirty_data_sync_percent Ns = Ns Sy 20 Ns % Pq uint
	Start syncing out a transaction group if there's at least this much dirty data
	.Pq as a percentage of Sy zfs_dirty_data_max .
	This should be less than
	.Sy zfs_vdev_async_write_active_min_dirty_percent .
	.
	.It Sy zfs_wrlog_data_max Ns = Pq int
	The upper limit of write-transaction zil log data size in bytes.
	Write operations are throttled when approaching the limit until log data is
	cleared out after transaction group sync.
	Because of some overhead, it should be set at least 2 times the size of
	.Sy zfs_dirty_data_max
	.No to prevent harming normal write throughput .
	It also should be smaller than the size of the slog device if slog is present.
	.Pp
	Defaults to
	.Sy zfs_dirty_data_max*2
	.
	.It Sy zfs_fallocate_reserve_percent Ns = Ns Sy 110 Ns % Pq uint
	Since ZFS is a copy-on-write filesystem with snapshots, blocks cannot be
	preallocated for a file in order to guarantee that later writes will not
	run out of space.
	Instead,
	.Xr fallocate 2
	space preallocation only checks that sufficient space is currently available
	in the pool or the user's project quota allocation,
	and then creates a sparse file of the requested size.
	The requested space is multiplied by
	.Sy zfs_fallocate_reserve_percent
	to allow additional space for indirect blocks and other internal metadata.
	Setting this to
	.Sy 0
	disables support for
	.Xr fallocate 2
	and causes it to return
	.Sy EOPNOTSUPP .
	.
	.It Sy zfs_fletcher_4_impl Ns = Ns Sy fastest Pq string
	Select a fletcher 4 implementation.
	.Pp
	Supported selectors are:
	.Sy fastest , scalar , sse2 , ssse3 , avx2 , avx512f , avx512bw ,
	.No and Sy aarch64_neon .
	All except
	.Sy fastest No and Sy scalar
	require instruction set extensions to be available,
	and will only appear if ZFS detects that they are present at runtime.
	If multiple implementations of fletcher 4 are available, the
	.Sy fastest
	will be chosen using a micro benchmark.
	Selecting
	.Sy scalar
	results in the original CPU-based calculation being used.
	Selecting any option other than
	.Sy fastest No or Sy scalar
	results in vector instructions
	from the respective CPU instruction set being used.
	.
	.It Sy zfs_bclone_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Enable the experimental block cloning feature.
	If this setting is 0, then even if feature@block_cloning is enabled,
	attempts to clone blocks will act as though the feature is disabled.
	.
	.It Sy zfs_bclone_wait_dirty Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	When set to 1 the FICLONE and FICLONERANGE ioctls wait for dirty data to be
	written to disk.
	This allows the clone operation to reliably succeed when a file is
	modified and then immediately cloned.
	For small files this may be slower than making a copy of the file.
	Therefore, this setting defaults to 0 which causes a clone operation to
	immediately fail when encountering a dirty block.
	.
	.It Sy zfs_blake3_impl Ns = Ns Sy fastest Pq string
	Select a BLAKE3 implementation.
	.Pp
	Supported selectors are:
	.Sy cycle , fastest , generic , sse2 , sse41 , avx2 , avx512 .
	All except
	.Sy cycle , fastest No and Sy generic
	require instruction set extensions to be available,
	and will only appear if ZFS detects that they are present at runtime.
	If multiple implementations of BLAKE3 are available, the
	.Sy fastest will be chosen using a micro benchmark. You can see the
	benchmark results by reading this kstat file:
	.Pa /proc/spl/kstat/zfs/chksum_bench .
	.
	.It Sy zfs_free_bpobj_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Enable/disable the processing of the free_bpobj object.
	.
	.It Sy zfs_async_block_max_blocks Ns = Ns Sy UINT64_MAX Po unlimited Pc Pq u64
	Maximum number of blocks freed in a single TXG.
	.
	.It Sy zfs_max_async_dedup_frees Ns = Ns Sy 100000 Po 10^5 Pc Pq u64
	Maximum number of dedup blocks freed in a single TXG.
	.
	.It Sy zfs_vdev_async_read_max_active Ns = Ns Sy 3 Pq uint
	Maximum asynchronous read I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_async_read_min_active Ns = Ns Sy 1 Pq uint
	Minimum asynchronous read I/O operation active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_async_write_active_max_dirty_percent Ns = Ns Sy 60 Ns % Pq uint
	When the pool has more than this much dirty data, use
	.Sy zfs_vdev_async_write_max_active
	to limit active async writes.
	If the dirty data is between the minimum and maximum,
	the active I/O limit is linearly interpolated.
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_async_write_active_min_dirty_percent Ns = Ns Sy 30 Ns % Pq uint
	When the pool has less than this much dirty data, use
	.Sy zfs_vdev_async_write_min_active
	to limit active async writes.
	If the dirty data is between the minimum and maximum,
	the active I/O limit is linearly
	interpolated.
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_async_write_max_active Ns = Ns Sy 10 Pq uint
	Maximum asynchronous write I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_async_write_min_active Ns = Ns Sy 2 Pq uint
	Minimum asynchronous write I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.Pp
	Lower values are associated with better latency on rotational media but poorer
	resilver performance.
	The default value of
	.Sy 2
	was chosen as a compromise.
	A value of
	.Sy 3
	has been shown to improve resilver performance further at a cost of
	further increasing latency.
	.
	.It Sy zfs_vdev_initializing_max_active Ns = Ns Sy 1 Pq uint
	Maximum initializing I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_initializing_min_active Ns = Ns Sy 1 Pq uint
	Minimum initializing I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_max_active Ns = Ns Sy 1000 Pq uint
	The maximum number of I/O operations active to each device.
	Ideally, this will be at least the sum of each queue's
	.Sy max_active .
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_open_timeout_ms Ns = Ns Sy 1000 Pq uint
	Timeout value to wait before determining a device is missing
	during import.
	This is helpful for transient missing paths due
	to links being briefly removed and recreated in response to
	udev events.
	.
	.It Sy zfs_vdev_rebuild_max_active Ns = Ns Sy 3 Pq uint
	Maximum sequential resilver I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_rebuild_min_active Ns = Ns Sy 1 Pq uint
	Minimum sequential resilver I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_removal_max_active Ns = Ns Sy 2 Pq uint
	Maximum removal I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_removal_min_active Ns = Ns Sy 1 Pq uint
	Minimum removal I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_scrub_max_active Ns = Ns Sy 2 Pq uint
	Maximum scrub I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_scrub_min_active Ns = Ns Sy 1 Pq uint
	Minimum scrub I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_sync_read_max_active Ns = Ns Sy 10 Pq uint
	Maximum synchronous read I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_sync_read_min_active Ns = Ns Sy 10 Pq uint
	Minimum synchronous read I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_sync_write_max_active Ns = Ns Sy 10 Pq uint
	Maximum synchronous write I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_sync_write_min_active Ns = Ns Sy 10 Pq uint
	Minimum synchronous write I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_trim_max_active Ns = Ns Sy 2 Pq uint
	Maximum trim/discard I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_trim_min_active Ns = Ns Sy 1 Pq uint
	Minimum trim/discard I/O operations active to each device.
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_nia_delay Ns = Ns Sy 5 Pq uint
	For non-interactive I/O (scrub, resilver, removal, initialize and rebuild),
	the number of concurrently-active I/O operations is limited to
	.Sy zfs_*_min_active ,
	unless the vdev is "idle".
	When there are no interactive I/O operations active (synchronous or otherwise),
	and
	.Sy zfs_vdev_nia_delay
	operations have completed since the last interactive operation,
	then the vdev is considered to be "idle",
	and the number of concurrently-active non-interactive operations is increased to
	.Sy zfs_*_max_active .
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_nia_credit Ns = Ns Sy 5 Pq uint
	Some HDDs tend to prioritize sequential I/O so strongly, that concurrent
	random I/O latency reaches several seconds.
	On some HDDs this happens even if sequential I/O operations
	are submitted one at a time, and so setting
	.Sy zfs_*_max_active Ns = Sy 1
	does not help.
	To prevent non-interactive I/O, like scrub,
	from monopolizing the device, no more than
	.Sy zfs_vdev_nia_credit operations can be sent
	while there are outstanding incomplete interactive operations.
	This enforced wait ensures the HDD services the interactive I/O
	within a reasonable amount of time.
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_queue_depth_pct Ns = Ns Sy 1000 Ns % Pq uint
	Maximum number of queued allocations per top-level vdev expressed as
	a percentage of
	.Sy zfs_vdev_async_write_max_active ,
	which allows the system to detect devices that are more capable
	of handling allocations and to allocate more blocks to those devices.
	This allows for dynamic allocation distribution when devices are imbalanced,
	as fuller devices will tend to be slower than empty devices.
	.Pp
	Also see
	.Sy zio_dva_throttle_enabled .
	.
	.It Sy zfs_vdev_def_queue_depth Ns = Ns Sy 32 Pq uint
	Default queue depth for each vdev IO allocator.
	Higher values allow for better coalescing of sequential writes before sending
	them to the disk, but can increase transaction commit times.
	.
	.It Sy zfs_vdev_failfast_mask Ns = Ns Sy 1 Pq uint
	Defines if the driver should retire on a given error type.
	The following options may be bitwise-ored together:
	.TS
	box;
	lbz r l l .
	Value Name Description
	_
	1 Device No driver retries on device errors
	2 Transport No driver retries on transport errors.
	4 Driver No driver retries on driver errors.
	.TE
	.
	.It Sy zfs_vdev_disk_max_segs Ns = Ns Sy 0 Pq uint
	Maximum number of segments to add to a BIO (min 4).
	If this is higher than the maximum allowed by the device queue or the kernel
	itself, it will be clamped.
	Setting it to zero will cause the kernel's ideal size to be used.
	This parameter only applies on Linux.
	This parameter is ignored if
	.Sy zfs_vdev_disk_classic Ns = Ns Sy 1 .
	.
	.It Sy zfs_vdev_disk_classic Ns = Ns Sy 0 Ns \| Ns 1 Pq uint
	If set to 1, OpenZFS will submit IO to Linux using the method it used in 2.2
	and earlier.
	This "classic" method has known issues with highly fragmented IO requests and
	is slower on many workloads, but it has been in use for many years and is known
	to be very stable.
	If you set this parameter, please also open a bug report why you did so,
	including the workload involved and any error messages.
	.Pp
	This parameter and the classic submission method will be removed once we have
	total confidence in the new method.
	.Pp
	This parameter only applies on Linux, and can only be set at module load time.
	.
	.It Sy zfs_expire_snapshot Ns = Ns Sy 300 Ns s Pq int
	Time before expiring
	.Pa .zfs/snapshot .
	.
	.It Sy zfs_admin_snapshot Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Allow the creation, removal, or renaming of entries in the
	.Sy .zfs/snapshot
	directory to cause the creation, destruction, or renaming of snapshots.
	When enabled, this functionality works both locally and over NFS exports
	which have the
	.Em no_root_squash
	option set.
	.
	.It Sy zfs_flags Ns = Ns Sy 0 Pq int
	Set additional debugging flags.
	The following flags may be bitwise-ored together:
	.TS
	box;
	lbz r l l .
	Value Name Description
	_
	1 ZFS_DEBUG_DPRINTF Enable dprintf entries in the debug log.
	* 2 ZFS_DEBUG_DBUF_VERIFY Enable extra dbuf verifications.
	* 4 ZFS_DEBUG_DNODE_VERIFY Enable extra dnode verifications.
	8 ZFS_DEBUG_SNAPNAMES Enable snapshot name verification.
	* 16 ZFS_DEBUG_MODIFY Check for illegally modified ARC buffers.
	64 ZFS_DEBUG_ZIO_FREE Enable verification of block frees.
	128 ZFS_DEBUG_HISTOGRAM_VERIFY Enable extra spacemap histogram verifications.
	256 ZFS_DEBUG_METASLAB_VERIFY Verify space accounting on disk matches in-memory \fBrange_trees\fP.
	512 ZFS_DEBUG_SET_ERROR Enable \fBSET_ERROR\fP and dprintf entries in the debug log.
	1024 ZFS_DEBUG_INDIRECT_REMAP Verify split blocks created by device removal.
	2048 ZFS_DEBUG_TRIM Verify TRIM ranges are always within the allocatable range tree.
	4096 ZFS_DEBUG_LOG_SPACEMAP Verify that the log summary is consistent with the spacemap log
	and enable \fBzfs_dbgmsgs\fP for metaslab loading and flushing.
	.TE
	.Sy \& * No Requires debug build .
	.
	.It Sy zfs_btree_verify_intensity Ns = Ns Sy 0 Pq uint
	Enables btree verification.
	The following settings are culminative:
	.TS
	box;
	lbz r l l .
	Value Description

	1 Verify height.
	2 Verify pointers from children to parent.
	3 Verify element counts.
	4 Verify element order. (expensive)
	* 5 Verify unused memory is poisoned. (expensive)
	.TE
	.Sy \& * No Requires debug build .
	.
	.It Sy zfs_free_leak_on_eio Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	If destroy encounters an
	.Sy EIO
	while reading metadata (e.g. indirect blocks),
	space referenced by the missing metadata can not be freed.
	Normally this causes the background destroy to become "stalled",
	as it is unable to make forward progress.
	While in this stalled state, all remaining space to free
	from the error-encountering filesystem is "temporarily leaked".
	Set this flag to cause it to ignore the
	.Sy EIO ,
	permanently leak the space from indirect blocks that can not be read,
	and continue to free everything else that it can.
	.Pp
	The default "stalling" behavior is useful if the storage partially
	fails (i.e. some but not all I/O operations fail), and then later recovers.
	In this case, we will be able to continue pool operations while it is
	partially failed, and when it recovers, we can continue to free the
	space, with no leaks.
	Note, however, that this case is actually fairly rare.
	.Pp
	Typically pools either
	.Bl -enum -compact -offset 4n -width "1."
	.It
	fail completely (but perhaps temporarily,
	e.g. due to a top-level vdev going offline), or
	.It
	have localized, permanent errors (e.g. disk returns the wrong data
	due to bit flip or firmware bug).
	.El
	In the former case, this setting does not matter because the
	pool will be suspended and the sync thread will not be able to make
	forward progress regardless.
	In the latter, because the error is permanent, the best we can do
	is leak the minimum amount of space,
	which is what setting this flag will do.
	It is therefore reasonable for this flag to normally be set,
	but we chose the more conservative approach of not setting it,
	so that there is no possibility of
	leaking space in the "partial temporary" failure case.
	.
	.It Sy zfs_free_min_time_ms Ns = Ns Sy 1000 Ns ms Po 1s Pc Pq uint
	During a
	.Nm zfs Cm destroy
	operation using the
	.Sy async_destroy
	feature,
	a minimum of this much time will be spent working on freeing blocks per TXG.
	.
	.It Sy zfs_obsolete_min_time_ms Ns = Ns Sy 500 Ns ms Pq uint
	Similar to
	.Sy zfs_free_min_time_ms ,
	but for cleanup of old indirection records for removed vdevs.
	.
	.It Sy zfs_immediate_write_sz Ns = Ns Sy 32768 Ns B Po 32 KiB Pc Pq s64
	Largest data block to write to the ZIL.
	Larger blocks will be treated as if the dataset being written to had the
	.Sy logbias Ns = Ns Sy throughput
	property set.
	.
	.It Sy zfs_initialize_value Ns = Ns Sy 16045690984833335022 Po 0xDEADBEEFDEADBEEE Pc Pq u64
	Pattern written to vdev free space by
	.Xr zpool-initialize 8 .
	.
	.It Sy zfs_initialize_chunk_size Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq u64
	Size of writes used by
	.Xr zpool-initialize 8 .
	This option is used by the test suite.
	.
	.It Sy zfs_livelist_max_entries Ns = Ns Sy 500000 Po 5*10^5 Pc Pq u64
	The threshold size (in block pointers) at which we create a new sub-livelist.
	Larger sublists are more costly from a memory perspective but the fewer
	sublists there are, the lower the cost of insertion.
	.
	.It Sy zfs_livelist_min_percent_shared Ns = Ns Sy 75 Ns % Pq int
	If the amount of shared space between a snapshot and its clone drops below
	this threshold, the clone turns off the livelist and reverts to the old
	deletion method.
	This is in place because livelists no long give us a benefit
	once a clone has been overwritten enough.
	.
	.It Sy zfs_livelist_condense_new_alloc Ns = Ns Sy 0 Pq int
	Incremented each time an extra ALLOC blkptr is added to a livelist entry while
	it is being condensed.
	This option is used by the test suite to track race conditions.
	.
	.It Sy zfs_livelist_condense_sync_cancel Ns = Ns Sy 0 Pq int
	Incremented each time livelist condensing is canceled while in
	.Fn spa_livelist_condense_sync .
	This option is used by the test suite to track race conditions.
	.
	.It Sy zfs_livelist_condense_sync_pause Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	When set, the livelist condense process pauses indefinitely before
	executing the synctask \(em
	.Fn spa_livelist_condense_sync .
	This option is used by the test suite to trigger race conditions.
	.
	.It Sy zfs_livelist_condense_zthr_cancel Ns = Ns Sy 0 Pq int
	Incremented each time livelist condensing is canceled while in
	.Fn spa_livelist_condense_cb .
	This option is used by the test suite to track race conditions.
	.
	.It Sy zfs_livelist_condense_zthr_pause Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	When set, the livelist condense process pauses indefinitely before
	executing the open context condensing work in
	.Fn spa_livelist_condense_cb .
	This option is used by the test suite to trigger race conditions.
	.
	.It Sy zfs_lua_max_instrlimit Ns = Ns Sy 100000000 Po 10^8 Pc Pq u64
	The maximum execution time limit that can be set for a ZFS channel program,
	specified as a number of Lua instructions.
	.
	.It Sy zfs_lua_max_memlimit Ns = Ns Sy 104857600 Po 100 MiB Pc Pq u64
	The maximum memory limit that can be set for a ZFS channel program, specified
	in bytes.
	.
	.It Sy zfs_max_dataset_nesting Ns = Ns Sy 50 Pq int
	The maximum depth of nested datasets.
	This value can be tuned temporarily to
	fix existing datasets that exceed the predefined limit.
	.
	.It Sy zfs_max_log_walking Ns = Ns Sy 5 Pq u64
	The number of past TXGs that the flushing algorithm of the log spacemap
	feature uses to estimate incoming log blocks.
	.
	.It Sy zfs_max_logsm_summary_length Ns = Ns Sy 10 Pq u64
	Maximum number of rows allowed in the summary of the spacemap log.
	.
	.It Sy zfs_max_recordsize Ns = Ns Sy 16777216 Po 16 MiB Pc Pq uint
	We currently support block sizes from
	.Em 512 Po 512 B Pc No to Em 16777216 Po 16 MiB Pc .
	The benefits of larger blocks, and thus larger I/O,
	need to be weighed against the cost of COWing a giant block to modify one byte.
	Additionally, very large blocks can have an impact on I/O latency,
	and also potentially on the memory allocator.
	Therefore, we formerly forbade creating blocks larger than 1M.
	Larger blocks could be created by changing it,
	and pools with larger blocks can always be imported and used,
	regardless of this setting.
	.
	.It Sy zfs_allow_redacted_dataset_mount Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Allow datasets received with redacted send/receive to be mounted.
	Normally disabled because these datasets may be missing key data.
	.
	.It Sy zfs_min_metaslabs_to_flush Ns = Ns Sy 1 Pq u64
	Minimum number of metaslabs to flush per dirty TXG.
	.
	.It Sy zfs_metaslab_fragmentation_threshold Ns = Ns Sy 70 Ns % Pq uint
	Allow metaslabs to keep their active state as long as their fragmentation
	percentage is no more than this value.
	An active metaslab that exceeds this threshold
	will no longer keep its active status allowing better metaslabs to be selected.
	.
	.It Sy zfs_mg_fragmentation_threshold Ns = Ns Sy 95 Ns % Pq uint
	Metaslab groups are considered eligible for allocations if their
	fragmentation metric (measured as a percentage) is less than or equal to
	this value.
	If a metaslab group exceeds this threshold then it will be
	skipped unless all metaslab groups within the metaslab class have also
	crossed this threshold.
	.
	.It Sy zfs_mg_noalloc_threshold Ns = Ns Sy 0 Ns % Pq uint
	Defines a threshold at which metaslab groups should be eligible for allocations.
	The value is expressed as a percentage of free space
	beyond which a metaslab group is always eligible for allocations.
	If a metaslab group's free space is less than or equal to the
	threshold, the allocator will avoid allocating to that group
	unless all groups in the pool have reached the threshold.
	Once all groups have reached the threshold, all groups are allowed to accept
	allocations.
	The default value of
	.Sy 0
	disables the feature and causes all metaslab groups to be eligible for
	allocations.
	.Pp
	This parameter allows one to deal with pools having heavily imbalanced
	vdevs such as would be the case when a new vdev has been added.
	Setting the threshold to a non-zero percentage will stop allocations
	from being made to vdevs that aren't filled to the specified percentage
	and allow lesser filled vdevs to acquire more allocations than they
	otherwise would under the old
	.Sy zfs_mg_alloc_failures
	facility.
	.
	.It Sy zfs_ddt_data_is_special Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	If enabled, ZFS will place DDT data into the special allocation class.
	.
	.It Sy zfs_user_indirect_is_special Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	If enabled, ZFS will place user data indirect blocks
	into the special allocation class.
	.
	.It Sy zfs_multihost_history Ns = Ns Sy 0 Pq uint
	Historical statistics for this many latest multihost updates will be available
	in
	.Pa /proc/spl/kstat/zfs/ Ns Ao Ar pool Ac Ns Pa /multihost .
	.
	.It Sy zfs_multihost_interval Ns = Ns Sy 1000 Ns ms Po 1 s Pc Pq u64
	Used to control the frequency of multihost writes which are performed when the
	.Sy multihost
	pool property is on.
	This is one of the factors used to determine the
	length of the activity check during import.
	.Pp
	The multihost write period is
	.Sy zfs_multihost_interval No / Sy leaf-vdevs .
	On average a multihost write will be issued for each leaf vdev
	every
	.Sy zfs_multihost_interval
	milliseconds.
	In practice, the observed period can vary with the I/O load
	and this observed value is the delay which is stored in the uberblock.
	.
	.It Sy zfs_multihost_import_intervals Ns = Ns Sy 20 Pq uint
	Used to control the duration of the activity test on import.
	Smaller values of
	.Sy zfs_multihost_import_intervals
	will reduce the import time but increase
	the risk of failing to detect an active pool.
	The total activity check time is never allowed to drop below one second.
	.Pp
	On import the activity check waits a minimum amount of time determined by
	.Sy zfs_multihost_interval No \(mu Sy zfs_multihost_import_intervals ,
	or the same product computed on the host which last had the pool imported,
	whichever is greater.
	The activity check time may be further extended if the value of MMP
	delay found in the best uberblock indicates actual multihost updates happened
	at longer intervals than
	.Sy zfs_multihost_interval .
	A minimum of
	.Em 100 ms
	is enforced.
	.Pp
	.Sy 0 No is equivalent to Sy 1 .
	.
	.It Sy zfs_multihost_fail_intervals Ns = Ns Sy 10 Pq uint
	Controls the behavior of the pool when multihost write failures or delays are
	detected.
	.Pp
	When
	.Sy 0 ,
	multihost write failures or delays are ignored.
	The failures will still be reported to the ZED which depending on
	its configuration may take action such as suspending the pool or offlining a
	device.
	.Pp
	Otherwise, the pool will be suspended if
	.Sy zfs_multihost_fail_intervals No \(mu Sy zfs_multihost_interval
	milliseconds pass without a successful MMP write.
	This guarantees the activity test will see MMP writes if the pool is imported.
	.Sy 1 No is equivalent to Sy 2 ;
	this is necessary to prevent the pool from being suspended
	due to normal, small I/O latency variations.
	.
	.It Sy zfs_no_scrub_io Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Set to disable scrub I/O.
	This results in scrubs not actually scrubbing data and
	simply doing a metadata crawl of the pool instead.
	.
	.It Sy zfs_no_scrub_prefetch Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Set to disable block prefetching for scrubs.
	.
	.It Sy zfs_nocacheflush Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Disable cache flush operations on disks when writing.
	Setting this will cause pool corruption on power loss
	if a volatile out-of-order write cache is enabled.
	.
	.It Sy zfs_nopwrite_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Allow no-operation writes.
	The occurrence of nopwrites will further depend on other pool properties
	.Pq i.a. the checksumming and compression algorithms .
	.
	.It Sy zfs_dmu_offset_next_sync Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Enable forcing TXG sync to find holes.
	When enabled forces ZFS to sync data when
	.Sy SEEK_HOLE No or Sy SEEK_DATA
	flags are used allowing holes in a file to be accurately reported.
	When disabled holes will not be reported in recently dirtied files.
	.
	.It Sy zfs_pd_bytes_max Ns = Ns Sy 52428800 Ns B Po 50 MiB Pc Pq int
	The number of bytes which should be prefetched during a pool traversal, like
	.Nm zfs Cm send
	or other data crawling operations.
	.
	.It Sy zfs_traverse_indirect_prefetch_limit Ns = Ns Sy 32 Pq uint
	The number of blocks pointed by indirect (non-L0) block which should be
	prefetched during a pool traversal, like
	.Nm zfs Cm send
	or other data crawling operations.
	.
	.It Sy zfs_per_txg_dirty_frees_percent Ns = Ns Sy 30 Ns % Pq u64
	Control percentage of dirtied indirect blocks from frees allowed into one TXG.
	After this threshold is crossed, additional frees will wait until the next TXG.
	.Sy 0 No disables this throttle .
	.
	.It Sy zfs_prefetch_disable Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Disable predictive prefetch.
	Note that it leaves "prescient" prefetch
	.Pq for, e.g., Nm zfs Cm send
	intact.
	Unlike predictive prefetch, prescient prefetch never issues I/O
	that ends up not being needed, so it can't hurt performance.
	.
	.It Sy zfs_qat_checksum_disable Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Disable QAT hardware acceleration for SHA256 checksums.
	May be unset after the ZFS modules have been loaded to initialize the QAT
	hardware as long as support is compiled in and the QAT driver is present.
	.
	.It Sy zfs_qat_compress_disable Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Disable QAT hardware acceleration for gzip compression.
	May be unset after the ZFS modules have been loaded to initialize the QAT
	hardware as long as support is compiled in and the QAT driver is present.
	.
	.It Sy zfs_qat_encrypt_disable Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Disable QAT hardware acceleration for AES-GCM encryption.
	May be unset after the ZFS modules have been loaded to initialize the QAT
	hardware as long as support is compiled in and the QAT driver is present.
	.
	.It Sy zfs_vnops_read_chunk_size Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq u64
	Bytes to read per chunk.
	.
	.It Sy zfs_read_history Ns = Ns Sy 0 Pq uint
	Historical statistics for this many latest reads will be available in
	.Pa /proc/spl/kstat/zfs/ Ns Ao Ar pool Ac Ns Pa /reads .
	.
	.It Sy zfs_read_history_hits Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Include cache hits in read history
	.
	.It Sy zfs_rebuild_max_segment Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq u64
	Maximum read segment size to issue when sequentially resilvering a
	top-level vdev.
	.
	.It Sy zfs_rebuild_scrub_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Automatically start a pool scrub when the last active sequential resilver
	completes in order to verify the checksums of all blocks which have been
	resilvered.
	This is enabled by default and strongly recommended.
	.
	.It Sy zfs_rebuild_vdev_limit Ns = Ns Sy 67108864 Ns B Po 64 MiB Pc Pq u64
	Maximum amount of I/O that can be concurrently issued for a sequential
	resilver per leaf device, given in bytes.
	.
	.It Sy zfs_reconstruct_indirect_combinations_max Ns = Ns Sy 4096 Pq int
	If an indirect split block contains more than this many possible unique
	combinations when being reconstructed, consider it too computationally
	expensive to check them all.
	Instead, try at most this many randomly selected
	combinations each time the block is accessed.
	This allows all segment copies to participate fairly
	in the reconstruction when all combinations
	cannot be checked and prevents repeated use of one bad copy.
	.
	.It Sy zfs_recover Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Set to attempt to recover from fatal errors.
	This should only be used as a last resort,
	as it typically results in leaked space, or worse.
	.
	.It Sy zfs_removal_ignore_errors Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Ignore hard I/O errors during device removal.
	When set, if a device encounters a hard I/O error during the removal process
	the removal will not be cancelled.
	This can result in a normally recoverable block becoming permanently damaged
	and is hence not recommended.
	This should only be used as a last resort when the
	pool cannot be returned to a healthy state prior to removing the device.
	.
	.It Sy zfs_removal_suspend_progress Ns = Ns Sy 0 Ns \| Ns 1 Pq uint
	This is used by the test suite so that it can ensure that certain actions
	happen while in the middle of a removal.
	.
	.It Sy zfs_remove_max_segment Ns = Ns Sy 16777216 Ns B Po 16 MiB Pc Pq uint
	The largest contiguous segment that we will attempt to allocate when removing
	a device.
	If there is a performance problem with attempting to allocate large blocks,
	consider decreasing this.
	The default value is also the maximum.
	.
	.It Sy zfs_resilver_disable_defer Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Ignore the
	.Sy resilver_defer
	feature, causing an operation that would start a resilver to
	immediately restart the one in progress.
	.
	.It Sy zfs_resilver_min_time_ms Ns = Ns Sy 3000 Ns ms Po 3 s Pc Pq uint
	Resilvers are processed by the sync thread.
	While resilvering, it will spend at least this much time
	working on a resilver between TXG flushes.
	.
	.It Sy zfs_scan_ignore_errors Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	If set, remove the DTL (dirty time list) upon completion of a pool scan (scrub),
	even if there were unrepairable errors.
	Intended to be used during pool repair or recovery to
	stop resilvering when the pool is next imported.
	.
	.It Sy zfs_scrub_after_expand Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Automatically start a pool scrub after a RAIDZ expansion completes
	in order to verify the checksums of all blocks which have been
	copied during the expansion.
	This is enabled by default and strongly recommended.
	.
	.It Sy zfs_scrub_min_time_ms Ns = Ns Sy 1000 Ns ms Po 1 s Pc Pq uint
	Scrubs are processed by the sync thread.
	While scrubbing, it will spend at least this much time
	working on a scrub between TXG flushes.
	.
	.It Sy zfs_scrub_error_blocks_per_txg Ns = Ns Sy 4096 Pq uint
	Error blocks to be scrubbed in one txg.
	.
	.It Sy zfs_scan_checkpoint_intval Ns = Ns Sy 7200 Ns s Po 2 hour Pc Pq uint
	To preserve progress across reboots, the sequential scan algorithm periodically
	needs to stop metadata scanning and issue all the verification I/O to disk.
	The frequency of this flushing is determined by this tunable.
	.
	.It Sy zfs_scan_fill_weight Ns = Ns Sy 3 Pq uint
	This tunable affects how scrub and resilver I/O segments are ordered.
	A higher number indicates that we care more about how filled in a segment is,
	while a lower number indicates we care more about the size of the extent without
	considering the gaps within a segment.
	This value is only tunable upon module insertion.
	Changing the value afterwards will have no effect on scrub or resilver
	performance.
	.
	.It Sy zfs_scan_issue_strategy Ns = Ns Sy 0 Pq uint
	Determines the order that data will be verified while scrubbing or resilvering:
	.Bl -tag -compact -offset 4n -width "a"
	.It Sy 1
	Data will be verified as sequentially as possible, given the
	amount of memory reserved for scrubbing
	.Pq see Sy zfs_scan_mem_lim_fact .
	This may improve scrub performance if the pool's data is very fragmented.
	.It Sy 2
	The largest mostly-contiguous chunk of found data will be verified first.
	By deferring scrubbing of small segments, we may later find adjacent data
	to coalesce and increase the segment size.
	.It Sy 0
	.No Use strategy Sy 1 No during normal verification
	.No and strategy Sy 2 No while taking a checkpoint .
	.El
	.
	.It Sy zfs_scan_legacy Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	If unset, indicates that scrubs and resilvers will gather metadata in
	memory before issuing sequential I/O.
	Otherwise indicates that the legacy algorithm will be used,
	where I/O is initiated as soon as it is discovered.
	Unsetting will not affect scrubs or resilvers that are already in progress.
	.
	.It Sy zfs_scan_max_ext_gap Ns = Ns Sy 2097152 Ns B Po 2 MiB Pc Pq int
	Sets the largest gap in bytes between scrub/resilver I/O operations
	that will still be considered sequential for sorting purposes.
	Changing this value will not
	affect scrubs or resilvers that are already in progress.
	.
	.It Sy zfs_scan_mem_lim_fact Ns = Ns Sy 20 Ns ^-1 Pq uint
	Maximum fraction of RAM used for I/O sorting by sequential scan algorithm.
	This tunable determines the hard limit for I/O sorting memory usage.
	When the hard limit is reached we stop scanning metadata and start issuing
	data verification I/O.
	This is done until we get below the soft limit.
	.
	.It Sy zfs_scan_mem_lim_soft_fact Ns = Ns Sy 20 Ns ^-1 Pq uint
	The fraction of the hard limit used to determined the soft limit for I/O sorting
	by the sequential scan algorithm.
	When we cross this limit from below no action is taken.
	When we cross this limit from above it is because we are issuing verification
	I/O.
	In this case (unless the metadata scan is done) we stop issuing verification I/O
	and start scanning metadata again until we get to the hard limit.
	.
	.It Sy zfs_scan_report_txgs Ns = Ns Sy 0 Ns \| Ns 1 Pq uint
	When reporting resilver throughput and estimated completion time use the
	performance observed over roughly the last
	.Sy zfs_scan_report_txgs
	TXGs.
	When set to zero performance is calculated over the time between checkpoints.
	.
	.It Sy zfs_scan_strict_mem_lim Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Enforce tight memory limits on pool scans when a sequential scan is in progress.
	When disabled, the memory limit may be exceeded by fast disks.
	.
	.It Sy zfs_scan_suspend_progress Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Freezes a scrub/resilver in progress without actually pausing it.
	Intended for testing/debugging.
	.
	.It Sy zfs_scan_vdev_limit Ns = Ns Sy 16777216 Ns B Po 16 MiB Pc Pq int
	Maximum amount of data that can be concurrently issued at once for scrubs and
	resilvers per leaf device, given in bytes.
	.
	.It Sy zfs_send_corrupt_data Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Allow sending of corrupt data (ignore read/checksum errors when sending).
	.
	.It Sy zfs_send_unmodified_spill_blocks Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Include unmodified spill blocks in the send stream.
	Under certain circumstances, previous versions of ZFS could incorrectly
	remove the spill block from an existing object.
	Including unmodified copies of the spill blocks creates a backwards-compatible
	stream which will recreate a spill block if it was incorrectly removed.
	.
	.It Sy zfs_send_no_prefetch_queue_ff Ns = Ns Sy 20 Ns ^\-1 Pq uint
	The fill fraction of the
	.Nm zfs Cm send
	internal queues.
	The fill fraction controls the timing with which internal threads are woken up.
	.
	.It Sy zfs_send_no_prefetch_queue_length Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq uint
	The maximum number of bytes allowed in
	.Nm zfs Cm send Ns 's
	internal queues.
	.
	.It Sy zfs_send_queue_ff Ns = Ns Sy 20 Ns ^\-1 Pq uint
	The fill fraction of the
	.Nm zfs Cm send
	prefetch queue.
	The fill fraction controls the timing with which internal threads are woken up.
	.
	.It Sy zfs_send_queue_length Ns = Ns Sy 16777216 Ns B Po 16 MiB Pc Pq uint
	The maximum number of bytes allowed that will be prefetched by
	.Nm zfs Cm send .
	This value must be at least twice the maximum block size in use.
	.
	.It Sy zfs_recv_queue_ff Ns = Ns Sy 20 Ns ^\-1 Pq uint
	The fill fraction of the
	.Nm zfs Cm receive
	queue.
	The fill fraction controls the timing with which internal threads are woken up.
	.
	.It Sy zfs_recv_queue_length Ns = Ns Sy 16777216 Ns B Po 16 MiB Pc Pq uint
	The maximum number of bytes allowed in the
	.Nm zfs Cm receive
	queue.
	This value must be at least twice the maximum block size in use.
	.
	.It Sy zfs_recv_write_batch_size Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq uint
	The maximum amount of data, in bytes, that
	.Nm zfs Cm receive
	will write in one DMU transaction.
	This is the uncompressed size, even when receiving a compressed send stream.
	This setting will not reduce the write size below a single block.
	Capped at a maximum of
	.Sy 32 MiB .
	.
	.It Sy zfs_recv_best_effort_corrective Ns = Ns Sy 0 Pq int
	When this variable is set to non-zero a corrective receive:
	.Bl -enum -compact -offset 4n -width "1."
	.It
	Does not enforce the restriction of source & destination snapshot GUIDs
	matching.
	.It
	If there is an error during healing, the healing receive is not
	terminated instead it moves on to the next record.
	.El
	.
	.It Sy zfs_override_estimate_recordsize Ns = Ns Sy 0 Ns \| Ns 1 Pq uint
	Setting this variable overrides the default logic for estimating block
	sizes when doing a
	.Nm zfs Cm send .
	The default heuristic is that the average block size
	will be the current recordsize.
	Override this value if most data in your dataset is not of that size
	and you require accurate zfs send size estimates.
	.
	.It Sy zfs_sync_pass_deferred_free Ns = Ns Sy 2 Pq uint
	Flushing of data to disk is done in passes.
	Defer frees starting in this pass.
	.
	.It Sy zfs_spa_discard_memory_limit Ns = Ns Sy 16777216 Ns B Po 16 MiB Pc Pq int
	Maximum memory used for prefetching a checkpoint's space map on each
	vdev while discarding the checkpoint.
	.
	.It Sy zfs_special_class_metadata_reserve_pct Ns = Ns Sy 25 Ns % Pq uint
	Only allow small data blocks to be allocated on the special and dedup vdev
	types when the available free space percentage on these vdevs exceeds this
	value.
	This ensures reserved space is available for pool metadata as the
	special vdevs approach capacity.
	.
	.It Sy zfs_sync_pass_dont_compress Ns = Ns Sy 8 Pq uint
	Starting in this sync pass, disable compression (including of metadata).
	With the default setting, in practice, we don't have this many sync passes,
	so this has no effect.
	.Pp
	The original intent was that disabling compression would help the sync passes
	to converge.
	However, in practice, disabling compression increases
	the average number of sync passes; because when we turn compression off,
	many blocks' size will change, and thus we have to re-allocate
	(not overwrite) them.
	It also increases the number of
	.Em 128 KiB
	allocations (e.g. for indirect blocks and spacemaps)
	because these will not be compressed.
	The
	.Em 128 KiB
	allocations are especially detrimental to performance
	on highly fragmented systems, which may have very few free segments of this
	size,
	and may need to load new metaslabs to satisfy these allocations.
	.
	.It Sy zfs_sync_pass_rewrite Ns = Ns Sy 2 Pq uint
	Rewrite new block pointers starting in this pass.
	.
	.It Sy zfs_trim_extent_bytes_max Ns = Ns Sy 134217728 Ns B Po 128 MiB Pc Pq uint
	Maximum size of TRIM command.
	Larger ranges will be split into chunks no larger than this value before
	issuing.
	.
	.It Sy zfs_trim_extent_bytes_min Ns = Ns Sy 32768 Ns B Po 32 KiB Pc Pq uint
	Minimum size of TRIM commands.
	TRIM ranges smaller than this will be skipped,
	unless they're part of a larger range which was chunked.
	This is done because it's common for these small TRIMs
	to negatively impact overall performance.
	.
	.It Sy zfs_trim_metaslab_skip Ns = Ns Sy 0 Ns \| Ns 1 Pq uint
	Skip uninitialized metaslabs during the TRIM process.
	This option is useful for pools constructed from large thinly-provisioned
	devices
	where TRIM operations are slow.
	As a pool ages, an increasing fraction of the pool's metaslabs
	will be initialized, progressively degrading the usefulness of this option.
	This setting is stored when starting a manual TRIM and will
	persist for the duration of the requested TRIM.
	.
	.It Sy zfs_trim_queue_limit Ns = Ns Sy 10 Pq uint
	Maximum number of queued TRIMs outstanding per leaf vdev.
	The number of concurrent TRIM commands issued to the device is controlled by
	.Sy zfs_vdev_trim_min_active No and Sy zfs_vdev_trim_max_active .
	.
	.It Sy zfs_trim_txg_batch Ns = Ns Sy 32 Pq uint
	The number of transaction groups' worth of frees which should be aggregated
	before TRIM operations are issued to the device.
	This setting represents a trade-off between issuing larger,
	more efficient TRIM operations and the delay
	before the recently trimmed space is available for use by the device.
	.Pp
	Increasing this value will allow frees to be aggregated for a longer time.
	This will result is larger TRIM operations and potentially increased memory
	usage.
	Decreasing this value will have the opposite effect.
	The default of
	.Sy 32
	was determined to be a reasonable compromise.
	.
	.It Sy zfs_txg_history Ns = Ns Sy 0 Pq uint
	Historical statistics for this many latest TXGs will be available in
	.Pa /proc/spl/kstat/zfs/ Ns Ao Ar pool Ac Ns Pa /TXGs .
	.
	.It Sy zfs_txg_timeout Ns = Ns Sy 5 Ns s Pq uint
	Flush dirty data to disk at least every this many seconds (maximum TXG
	duration).
	.
	.It Sy zfs_vdev_aggregation_limit Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq uint
	Max vdev I/O aggregation size.
	.
	.It Sy zfs_vdev_aggregation_limit_non_rotating Ns = Ns Sy 131072 Ns B Po 128 KiB Pc Pq uint
	Max vdev I/O aggregation size for non-rotating media.
	.
	.It Sy zfs_vdev_mirror_rotating_inc Ns = Ns Sy 0 Pq int
	A number by which the balancing algorithm increments the load calculation for
	the purpose of selecting the least busy mirror member when an I/O operation
	immediately follows its predecessor on rotational vdevs
	for the purpose of making decisions based on load.
	.
	.It Sy zfs_vdev_mirror_rotating_seek_inc Ns = Ns Sy 5 Pq int
	A number by which the balancing algorithm increments the load calculation for
	the purpose of selecting the least busy mirror member when an I/O operation
	lacks locality as defined by
	.Sy zfs_vdev_mirror_rotating_seek_offset .
	Operations within this that are not immediately following the previous operation
	are incremented by half.
	.
	.It Sy zfs_vdev_mirror_rotating_seek_offset Ns = Ns Sy 1048576 Ns B Po 1 MiB Pc Pq int
	The maximum distance for the last queued I/O operation in which
	the balancing algorithm considers an operation to have locality.
	.No See Sx ZFS I/O SCHEDULER .
	.
	.It Sy zfs_vdev_mirror_non_rotating_inc Ns = Ns Sy 0 Pq int
	A number by which the balancing algorithm increments the load calculation for
	the purpose of selecting the least busy mirror member on non-rotational vdevs
	when I/O operations do not immediately follow one another.
	.
	.It Sy zfs_vdev_mirror_non_rotating_seek_inc Ns = Ns Sy 1 Pq int
	A number by which the balancing algorithm increments the load calculation for
	the purpose of selecting the least busy mirror member when an I/O operation
	lacks
	locality as defined by the
	.Sy zfs_vdev_mirror_rotating_seek_offset .
	Operations within this that are not immediately following the previous operation
	are incremented by half.
	.
	.It Sy zfs_vdev_read_gap_limit Ns = Ns Sy 32768 Ns B Po 32 KiB Pc Pq uint
	Aggregate read I/O operations if the on-disk gap between them is within this
	threshold.
	.
	.It Sy zfs_vdev_write_gap_limit Ns = Ns Sy 4096 Ns B Po 4 KiB Pc Pq uint
	Aggregate write I/O operations if the on-disk gap between them is within this
	threshold.
	.
	.It Sy zfs_vdev_raidz_impl Ns = Ns Sy fastest Pq string
	Select the raidz parity implementation to use.
	.Pp
	Variants that don't depend on CPU-specific features
	may be selected on module load, as they are supported on all systems.
	The remaining options may only be set after the module is loaded,
	as they are available only if the implementations are compiled in
	and supported on the running system.
	.Pp
	Once the module is loaded,
	.Pa /sys/module/zfs/parameters/zfs_vdev_raidz_impl
	will show the available options,
	with the currently selected one enclosed in square brackets.
	.Pp
	.TS
	lb l l .
	fastest selected by built-in benchmark
	original original implementation
	scalar scalar implementation
	sse2 SSE2 instruction set 64-bit x86
	ssse3 SSSE3 instruction set 64-bit x86
	avx2 AVX2 instruction set 64-bit x86
	avx512f AVX512F instruction set 64-bit x86
	avx512bw AVX512F & AVX512BW instruction sets 64-bit x86
	aarch64_neon NEON Aarch64/64-bit ARMv8
	aarch64_neonx2 NEON with more unrolling Aarch64/64-bit ARMv8
	powerpc_altivec Altivec PowerPC
	.TE
	.
	.It Sy zfs_vdev_scheduler Pq charp
	.Sy DEPRECATED .
	Prints warning to kernel log for compatibility.
	.
	.It Sy zfs_zevent_len_max Ns = Ns Sy 512 Pq uint
	Max event queue length.
	Events in the queue can be viewed with
	.Xr zpool-events 8 .
	.
	.It Sy zfs_zevent_retain_max Ns = Ns Sy 2000 Pq int
	Maximum recent zevent records to retain for duplicate checking.
	Setting this to
	.Sy 0
	disables duplicate detection.
	.
	.It Sy zfs_zevent_retain_expire_secs Ns = Ns Sy 900 Ns s Po 15 min Pc Pq int
	Lifespan for a recent ereport that was retained for duplicate checking.
	.
	.It Sy zfs_zil_clean_taskq_maxalloc Ns = Ns Sy 1048576 Pq int
	The maximum number of taskq entries that are allowed to be cached.
	When this limit is exceeded transaction records (itxs)
	will be cleaned synchronously.
	.
	.It Sy zfs_zil_clean_taskq_minalloc Ns = Ns Sy 1024 Pq int
	The number of taskq entries that are pre-populated when the taskq is first
	created and are immediately available for use.
	.
	.It Sy zfs_zil_clean_taskq_nthr_pct Ns = Ns Sy 100 Ns % Pq int
	This controls the number of threads used by
	.Sy dp_zil_clean_taskq .
	The default value of
	.Sy 100%
	will create a maximum of one thread per cpu.
	.
	.It Sy zil_maxblocksize Ns = Ns Sy 131072 Ns B Po 128 KiB Pc Pq uint
	This sets the maximum block size used by the ZIL.
	On very fragmented pools, lowering this
	.Pq typically to Sy 36 KiB
	can improve performance.
	.
	.It Sy zil_maxcopied Ns = Ns Sy 7680 Ns B Po 7.5 KiB Pc Pq uint
	This sets the maximum number of write bytes logged via WR_COPIED.
	It tunes a tradeoff between additional memory copy and possibly worse log
	space efficiency vs additional range lock/unlock.
	.
	.It Sy zil_nocacheflush Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Disable the cache flush commands that are normally sent to disk by
	the ZIL after an LWB write has completed.
	Setting this will cause ZIL corruption on power loss
	if a volatile out-of-order write cache is enabled.
	.
	.It Sy zil_replay_disable Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Disable intent logging replay.
	Can be disabled for recovery from corrupted ZIL.
	.
	.It Sy zil_slog_bulk Ns = Ns Sy 67108864 Ns B Po 64 MiB Pc Pq u64
	Limit SLOG write size per commit executed with synchronous priority.
	Any writes above that will be executed with lower (asynchronous) priority
	to limit potential SLOG device abuse by single active ZIL writer.
	.
	.It Sy zfs_zil_saxattr Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Setting this tunable to zero disables ZIL logging of new
	.Sy xattr Ns = Ns Sy sa
	records if the
	.Sy org.openzfs:zilsaxattr
	feature is enabled on the pool.
	This would only be necessary to work around bugs in the ZIL logging or replay
	code for this record type.
	The tunable has no effect if the feature is disabled.
	.
	.It Sy zfs_embedded_slog_min_ms Ns = Ns Sy 64 Pq uint
	Usually, one metaslab from each normal-class vdev is dedicated for use by
	the ZIL to log synchronous writes.
	However, if there are fewer than
	.Sy zfs_embedded_slog_min_ms
	metaslabs in the vdev, this functionality is disabled.
	This ensures that we don't set aside an unreasonable amount of space for the
	ZIL.
	.
	.It Sy zstd_earlyabort_pass Ns = Ns Sy 1 Pq uint
	Whether heuristic for detection of incompressible data with zstd levels >= 3
	using LZ4 and zstd-1 passes is enabled.
	.
	.It Sy zstd_abort_size Ns = Ns Sy 131072 Pq uint
	Minimal uncompressed size (inclusive) of a record before the early abort
	heuristic will be attempted.
	.
	.It Sy zio_deadman_log_all Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	If non-zero, the zio deadman will produce debugging messages
	.Pq see Sy zfs_dbgmsg_enable
	for all zios, rather than only for leaf zios possessing a vdev.
	This is meant to be used by developers to gain
	diagnostic information for hang conditions which don't involve a mutex
	or other locking primitive: typically conditions in which a thread in
	the zio pipeline is looping indefinitely.
	.
	.It Sy zio_slow_io_ms Ns = Ns Sy 30000 Ns ms Po 30 s Pc Pq int
	When an I/O operation takes more than this much time to complete,
	it's marked as slow.
	Each slow operation causes a delay zevent.
	Slow I/O counters can be seen with
	.Nm zpool Cm status Fl s .
	.
	.It Sy zio_dva_throttle_enabled Ns = Ns Sy 1 Ns \| Ns 0 Pq int
	Throttle block allocations in the I/O pipeline.
	This allows for dynamic allocation distribution when devices are imbalanced.
	When enabled, the maximum number of pending allocations per top-level vdev
	is limited by
	.Sy zfs_vdev_queue_depth_pct .
	.
	.It Sy zfs_xattr_compat Ns = Ns 0 Ns \| Ns 1 Pq int
	Control the naming scheme used when setting new xattrs in the user namespace.
	If
	.Sy 0
	.Pq the default on Linux ,
	user namespace xattr names are prefixed with the namespace, to be backwards
	compatible with previous versions of ZFS on Linux.
	If
	.Sy 1
	.Pq the default on Fx ,
	user namespace xattr names are not prefixed, to be backwards compatible with
	previous versions of ZFS on illumos and
	.Fx .
	.Pp
	Either naming scheme can be read on this and future versions of ZFS, regardless
	of this tunable, but legacy ZFS on illumos or
	.Fx
	are unable to read user namespace xattrs written in the Linux format, and
	legacy versions of ZFS on Linux are unable to read user namespace xattrs written
	in the legacy ZFS format.
	.Pp
	An existing xattr with the alternate naming scheme is removed when overwriting
	the xattr so as to not accumulate duplicates.
	.
	.It Sy zio_requeue_io_start_cut_in_line Ns = Ns Sy 0 Ns \| Ns 1 Pq int
	Prioritize requeued I/O.
	.
	.It Sy zio_taskq_batch_pct Ns = Ns Sy 80 Ns % Pq uint
	Percentage of online CPUs which will run a worker thread for I/O.
	These workers are responsible for I/O work such as compression, encryption,
	checksum and parity calculations.
	Fractional number of CPUs will be rounded down.
	.Pp
	The default value of
	.Sy 80%
	was chosen to avoid using all CPUs which can result in
	latency issues and inconsistent application performance,
	especially when slower compression and/or checksumming is enabled.
	Set value only applies to pools imported/created after that.
	.
	.It Sy zio_taskq_batch_tpq Ns = Ns Sy 0 Pq uint
	Number of worker threads per taskq.
	Higher values improve I/O ordering and CPU utilization,
	while lower reduce lock contention.
	Set value only applies to pools imported/created after that.
	.Pp
	If
	.Sy 0 ,
	generate a system-dependent value close to 6 threads per taskq.
	Set value only applies to pools imported/created after that.
	.
	.It Sy zio_taskq_write_tpq Ns = Ns Sy 16 Pq uint
	Determines the minumum number of threads per write issue taskq.
	Higher values improve CPU utilization on high throughput,
	while lower reduce taskq locks contention on high IOPS.
	Set value only applies to pools imported/created after that.
	.
	.It Sy zio_taskq_read Ns = Ns Sy fixed,1,8 null scale null Pq charp
	Set the queue and thread configuration for the IO read queues.
	This is an advanced debugging parameter.
	Don't change this unless you understand what it does.
	Set values only apply to pools imported/created after that.
	.
	-.It Sy zio_taskq_write Ns = Ns Sy sync fixed,1,5 scale fixed,1,5 Pq charp
	+.It Sy zio_taskq_write Ns = Ns Sy sync null scale null Pq charp
	Set the queue and thread configuration for the IO write queues.
	This is an advanced debugging parameter.
	Don't change this unless you understand what it does.
	Set values only apply to pools imported/created after that.
	.
	.It Sy zvol_inhibit_dev Ns = Ns Sy 0 Ns \| Ns 1 Pq uint
	Do not create zvol device nodes.
	This may slightly improve startup time on
	systems with a very large number of zvols.
	.
	.It Sy zvol_major Ns = Ns Sy 230 Pq uint
	Major number for zvol block devices.
	.
	.It Sy zvol_max_discard_blocks Ns = Ns Sy 16384 Pq long
	Discard (TRIM) operations done on zvols will be done in batches of this
	many blocks, where block size is determined by the
	.Sy volblocksize
	property of a zvol.
	.
	.It Sy zvol_prefetch_bytes Ns = Ns Sy 131072 Ns B Po 128 KiB Pc Pq uint
	When adding a zvol to the system, prefetch this many bytes
	from the start and end of the volume.
	Prefetching these regions of the volume is desirable,
	because they are likely to be accessed immediately by
	.Xr blkid 8
	or the kernel partitioner.
	.
	.It Sy zvol_request_sync Ns = Ns Sy 0 Ns \| Ns 1 Pq uint
	When processing I/O requests for a zvol, submit them synchronously.
	This effectively limits the queue depth to
	.Em 1
	for each I/O submitter.
	When unset, requests are handled asynchronously by a thread pool.
	The number of requests which can be handled concurrently is controlled by
	.Sy zvol_threads .
	.Sy zvol_request_sync
	is ignored when running on a kernel that supports block multiqueue
	.Pq Li blk-mq .
	.
	.It Sy zvol_num_taskqs Ns = Ns Sy 0 Pq uint
	Number of zvol taskqs.
	If
	.Sy 0
	(the default) then scaling is done internally to prefer 6 threads per taskq.
	This only applies on Linux.
	.
	.It Sy zvol_threads Ns = Ns Sy 0 Pq uint
	The number of system wide threads to use for processing zvol block IOs.
	If
	.Sy 0
	(the default) then internally set
	.Sy zvol_threads
	to the number of CPUs present or 32 (whichever is greater).
	.
	.It Sy zvol_blk_mq_threads Ns = Ns Sy 0 Pq uint
	The number of threads per zvol to use for queuing IO requests.
	This parameter will only appear if your kernel supports
	.Li blk-mq
	and is only read and assigned to a zvol at zvol load time.
	If
	.Sy 0
	(the default) then internally set
	.Sy zvol_blk_mq_threads
	to the number of CPUs present.
	.
	.It Sy zvol_use_blk_mq Ns = Ns Sy 0 Ns \| Ns 1 Pq uint
	Set to
	.Sy 1
	to use the
	.Li blk-mq
	API for zvols.
	Set to
	.Sy 0
	(the default) to use the legacy zvol APIs.
	This setting can give better or worse zvol performance depending on
	the workload.
	This parameter will only appear if your kernel supports
	.Li blk-mq
	and is only read and assigned to a zvol at zvol load time.
	.
	.It Sy zvol_blk_mq_blocks_per_thread Ns = Ns Sy 8 Pq uint
	If
	.Sy zvol_use_blk_mq
	is enabled, then process this number of
	.Sy volblocksize Ns -sized blocks per zvol thread.
	This tunable can be use to favor better performance for zvol reads (lower
	values) or writes (higher values).
	If set to
	.Sy 0 ,
	then the zvol layer will process the maximum number of blocks
	per thread that it can.
	This parameter will only appear if your kernel supports
	.Li blk-mq
	and is only applied at each zvol's load time.
	.
	.It Sy zvol_blk_mq_queue_depth Ns = Ns Sy 0 Pq uint
	The queue_depth value for the zvol
	.Li blk-mq
	interface.
	This parameter will only appear if your kernel supports
	.Li blk-mq
	and is only applied at each zvol's load time.
	If
	.Sy 0
	(the default) then use the kernel's default queue depth.
	Values are clamped to the kernel's
	.Dv BLKDEV_MIN_RQ
	and
	.Dv BLKDEV_MAX_RQ Ns / Ns Dv BLKDEV_DEFAULT_RQ
	limits.
	.
	.It Sy zvol_volmode Ns = Ns Sy 1 Pq uint
	Defines zvol block devices behaviour when
	.Sy volmode Ns = Ns Sy default :
	.Bl -tag -compact -offset 4n -width "a"
	.It Sy 1
	.No equivalent to Sy full
	.It Sy 2
	.No equivalent to Sy dev
	.It Sy 3
	.No equivalent to Sy none
	.El
	.
	.It Sy zvol_enforce_quotas Ns = Ns Sy 0 Ns \| Ns 1 Pq uint
	Enable strict ZVOL quota enforcement.
	The strict quota enforcement may have a performance impact.
	.El
	.
	.Sh ZFS I/O SCHEDULER
	ZFS issues I/O operations to leaf vdevs to satisfy and complete I/O operations.
	The scheduler determines when and in what order those operations are issued.
	The scheduler divides operations into five I/O classes,
	prioritized in the following order: sync read, sync write, async read,
	async write, and scrub/resilver.
	Each queue defines the minimum and maximum number of concurrent operations
	that may be issued to the device.
	In addition, the device has an aggregate maximum,
	.Sy zfs_vdev_max_active .
	Note that the sum of the per-queue minima must not exceed the aggregate maximum.
	If the sum of the per-queue maxima exceeds the aggregate maximum,
	then the number of active operations may reach
	.Sy zfs_vdev_max_active ,
	in which case no further operations will be issued,
	regardless of whether all per-queue minima have been met.
	.Pp
	For many physical devices, throughput increases with the number of
	concurrent operations, but latency typically suffers.
	Furthermore, physical devices typically have a limit
	at which more concurrent operations have no
	effect on throughput or can actually cause it to decrease.
	.Pp
	The scheduler selects the next operation to issue by first looking for an
	I/O class whose minimum has not been satisfied.
	Once all are satisfied and the aggregate maximum has not been hit,
	the scheduler looks for classes whose maximum has not been satisfied.
	Iteration through the I/O classes is done in the order specified above.
	No further operations are issued
	if the aggregate maximum number of concurrent operations has been hit,
	or if there are no operations queued for an I/O class that has not hit its
	maximum.
	Every time an I/O operation is queued or an operation completes,
	the scheduler looks for new operations to issue.
	.Pp
	In general, smaller
	.Sy max_active Ns s
	will lead to lower latency of synchronous operations.
	Larger
	.Sy max_active Ns s
	may lead to higher overall throughput, depending on underlying storage.
	.Pp
	The ratio of the queues'
	.Sy max_active Ns s
	determines the balance of performance between reads, writes, and scrubs.
	For example, increasing
	.Sy zfs_vdev_scrub_max_active
	will cause the scrub or resilver to complete more quickly,
	but reads and writes to have higher latency and lower throughput.
	.Pp
	All I/O classes have a fixed maximum number of outstanding operations,
	except for the async write class.
	Asynchronous writes represent the data that is committed to stable storage
	during the syncing stage for transaction groups.
	Transaction groups enter the syncing state periodically,
	so the number of queued async writes will quickly burst up
	and then bleed down to zero.
	Rather than servicing them as quickly as possible,
	the I/O scheduler changes the maximum number of active async write operations
	according to the amount of dirty data in the pool.
	Since both throughput and latency typically increase with the number of
	concurrent operations issued to physical devices, reducing the
	burstiness in the number of simultaneous operations also stabilizes the
	response time of operations from other queues, in particular synchronous ones.
	In broad strokes, the I/O scheduler will issue more concurrent operations
	from the async write queue as there is more dirty data in the pool.
	.
	.Ss Async Writes
	The number of concurrent operations issued for the async write I/O class
	follows a piece-wise linear function defined by a few adjustable points:
	.Bd -literal
	\| o---------\| <-- \fBzfs_vdev_async_write_max_active\fP
	^ \| /^ \|
	\| \| / \| \|
	active \| / \| \|
	I/O \| / \| \|
	count \| / \| \|
	\| / \| \|
	\|-------o \| \| <-- \fBzfs_vdev_async_write_min_active\fP
	0\|_______^______\|_________\|
	0% \| \| 100% of \fBzfs_dirty_data_max\fP
	\| \|
	\| `-- \fBzfs_vdev_async_write_active_max_dirty_percent\fP
	`--------- \fBzfs_vdev_async_write_active_min_dirty_percent\fP
	.Ed
	.Pp
	Until the amount of dirty data exceeds a minimum percentage of the dirty
	data allowed in the pool, the I/O scheduler will limit the number of
	concurrent operations to the minimum.
	As that threshold is crossed, the number of concurrent operations issued
	increases linearly to the maximum at the specified maximum percentage
	of the dirty data allowed in the pool.
	.Pp
	Ideally, the amount of dirty data on a busy pool will stay in the sloped
	part of the function between
	.Sy zfs_vdev_async_write_active_min_dirty_percent
	and
	.Sy zfs_vdev_async_write_active_max_dirty_percent .
	If it exceeds the maximum percentage,
	this indicates that the rate of incoming data is
	greater than the rate that the backend storage can handle.
	In this case, we must further throttle incoming writes,
	as described in the next section.
	.
	.Sh ZFS TRANSACTION DELAY
	We delay transactions when we've determined that the backend storage
	isn't able to accommodate the rate of incoming writes.
	.Pp
	If there is already a transaction waiting, we delay relative to when
	that transaction will finish waiting.
	This way the calculated delay time
	is independent of the number of threads concurrently executing transactions.
	.Pp
	If we are the only waiter, wait relative to when the transaction started,
	rather than the current time.
	This credits the transaction for "time already served",
	e.g. reading indirect blocks.
	.Pp
	The minimum time for a transaction to take is calculated as
	.D1 min_time = min( Ns Sy zfs_delay_scale No \(mu Po Sy dirty No \- Sy min Pc / Po Sy max No \- Sy dirty Pc , 100ms)
	.Pp
	The delay has two degrees of freedom that can be adjusted via tunables.
	The percentage of dirty data at which we start to delay is defined by
	.Sy zfs_delay_min_dirty_percent .
	This should typically be at or above
	.Sy zfs_vdev_async_write_active_max_dirty_percent ,
	so that we only start to delay after writing at full speed
	has failed to keep up with the incoming write rate.
	The scale of the curve is defined by
	.Sy zfs_delay_scale .
	Roughly speaking, this variable determines the amount of delay at the midpoint
	of the curve.
	.Bd -literal
	delay
	10ms +-------------------------------------------------------------*+
	\| *\|
	9ms + *+
	\| *\|
	8ms + *+
	\| * \|
	7ms + * +
	\| * \|
	6ms + * +
	\| * \|
	5ms + * +
	\| * \|
	4ms + * +
	\| * \|
	3ms + * +
	\| * \|
	2ms + (midpoint) * +
	\| \| ** \|
	1ms + v *** +
	\| \fBzfs_delay_scale\fP ----------> ******** \|
	0 +-------------------------------------*********----------------+
	0% <- \fBzfs_dirty_data_max\fP -> 100%
	.Ed
	.Pp
	Note, that since the delay is added to the outstanding time remaining on the
	most recent transaction it's effectively the inverse of IOPS.
	Here, the midpoint of
	.Em 500 us
	translates to
	.Em 2000 IOPS .
	The shape of the curve
	was chosen such that small changes in the amount of accumulated dirty data
	in the first three quarters of the curve yield relatively small differences
	in the amount of delay.
	.Pp
	The effects can be easier to understand when the amount of delay is
	represented on a logarithmic scale:
	.Bd -literal
	delay
	100ms +-------------------------------------------------------------++
	+ +
	\| \|
	+ *+
	10ms + *+
	+ ** +
	\| (midpoint) ** \|
	+ \| ** +
	1ms + v **** +
	+ \fBzfs_delay_scale\fP ----------> ***** +
	\| **** \|
	+ **** +
	100us + ** +
	+ * +
	\| * \|
	+ * +
	10us + * +
	+ +
	\| \|
	+ +
	+--------------------------------------------------------------+
	0% <- \fBzfs_dirty_data_max\fP -> 100%
	.Ed
	.Pp
	Note here that only as the amount of dirty data approaches its limit does
	the delay start to increase rapidly.
	The goal of a properly tuned system should be to keep the amount of dirty data
	out of that range by first ensuring that the appropriate limits are set
	for the I/O scheduler to reach optimal throughput on the back-end storage,
	and then by changing the value of
	.Sy zfs_delay_scale
	to increase the steepness of the curve.
	diff --git a/sys/contrib/openzfs/module/icp/io/aes.c b/sys/contrib/openzfs/module/icp/io/aes.c
	index d6f01304f56b..522c436497bc 100644
	--- a/sys/contrib/openzfs/module/icp/io/aes.c
	+++ b/sys/contrib/openzfs/module/icp/io/aes.c
	@@ -1,1319 +1,1323 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
	*/

	/*
	* AES provider for the Kernel Cryptographic Framework (KCF)
	*/

	#include <sys/zfs_context.h>
	#include <sys/crypto/common.h>
	#include <sys/crypto/impl.h>
	#include <sys/crypto/spi.h>
	#include <sys/crypto/icp.h>
	#include <modes/modes.h>
	#define _AES_IMPL
	#include <aes/aes_impl.h>
	#include <modes/gcm_impl.h>

	/*
	* Mechanism info structure passed to KCF during registration.
	*/
	static const crypto_mech_info_t aes_mech_info_tab[] = {
	/* AES_ECB */
	{SUN_CKM_AES_ECB, AES_ECB_MECH_INFO_TYPE,
	CRYPTO_FG_ENCRYPT \| CRYPTO_FG_ENCRYPT_ATOMIC \|
	CRYPTO_FG_DECRYPT \| CRYPTO_FG_DECRYPT_ATOMIC},
	/* AES_CBC */
	{SUN_CKM_AES_CBC, AES_CBC_MECH_INFO_TYPE,
	CRYPTO_FG_ENCRYPT \| CRYPTO_FG_ENCRYPT_ATOMIC \|
	CRYPTO_FG_DECRYPT \| CRYPTO_FG_DECRYPT_ATOMIC},
	/* AES_CTR */
	{SUN_CKM_AES_CTR, AES_CTR_MECH_INFO_TYPE,
	CRYPTO_FG_ENCRYPT \| CRYPTO_FG_ENCRYPT_ATOMIC \|
	CRYPTO_FG_DECRYPT \| CRYPTO_FG_DECRYPT_ATOMIC},
	/* AES_CCM */
	{SUN_CKM_AES_CCM, AES_CCM_MECH_INFO_TYPE,
	CRYPTO_FG_ENCRYPT \| CRYPTO_FG_ENCRYPT_ATOMIC \|
	CRYPTO_FG_DECRYPT \| CRYPTO_FG_DECRYPT_ATOMIC},
	/* AES_GCM */
	{SUN_CKM_AES_GCM, AES_GCM_MECH_INFO_TYPE,
	CRYPTO_FG_ENCRYPT \| CRYPTO_FG_ENCRYPT_ATOMIC \|
	CRYPTO_FG_DECRYPT \| CRYPTO_FG_DECRYPT_ATOMIC},
	/* AES_GMAC */
	{SUN_CKM_AES_GMAC, AES_GMAC_MECH_INFO_TYPE,
	CRYPTO_FG_ENCRYPT \| CRYPTO_FG_ENCRYPT_ATOMIC \|
	CRYPTO_FG_DECRYPT \| CRYPTO_FG_DECRYPT_ATOMIC \|
	CRYPTO_FG_MAC \| CRYPTO_FG_MAC_ATOMIC},
	};

	static int aes_encrypt_init(crypto_ctx_t , crypto_mechanism_t ,
	crypto_key_t *, crypto_spi_ctx_template_t);
	static int aes_decrypt_init(crypto_ctx_t , crypto_mechanism_t ,
	crypto_key_t *, crypto_spi_ctx_template_t);
	static int aes_common_init(crypto_ctx_t , crypto_mechanism_t ,
	crypto_key_t *, crypto_spi_ctx_template_t, boolean_t);
	static int aes_common_init_ctx(aes_ctx_t , crypto_spi_ctx_template_t ,
	crypto_mechanism_t , crypto_key_t , int, boolean_t);
	static int aes_encrypt_final(crypto_ctx_t , crypto_data_t );
	static int aes_decrypt_final(crypto_ctx_t , crypto_data_t );

	static int aes_encrypt(crypto_ctx_t , crypto_data_t , crypto_data_t *);
	static int aes_encrypt_update(crypto_ctx_t , crypto_data_t ,
	crypto_data_t *);
	static int aes_encrypt_atomic(crypto_mechanism_t , crypto_key_t ,
	crypto_data_t , crypto_data_t , crypto_spi_ctx_template_t);

	static int aes_decrypt(crypto_ctx_t , crypto_data_t , crypto_data_t *);
	static int aes_decrypt_update(crypto_ctx_t , crypto_data_t ,
	crypto_data_t *);
	static int aes_decrypt_atomic(crypto_mechanism_t , crypto_key_t ,
	crypto_data_t , crypto_data_t , crypto_spi_ctx_template_t);

	static const crypto_cipher_ops_t aes_cipher_ops = {
	.encrypt_init = aes_encrypt_init,
	.encrypt = aes_encrypt,
	.encrypt_update = aes_encrypt_update,
	.encrypt_final = aes_encrypt_final,
	.encrypt_atomic = aes_encrypt_atomic,
	.decrypt_init = aes_decrypt_init,
	.decrypt = aes_decrypt,
	.decrypt_update = aes_decrypt_update,
	.decrypt_final = aes_decrypt_final,
	.decrypt_atomic = aes_decrypt_atomic
	};

	static int aes_mac_atomic(crypto_mechanism_t , crypto_key_t , crypto_data_t *,
	crypto_data_t *, crypto_spi_ctx_template_t);
	static int aes_mac_verify_atomic(crypto_mechanism_t , crypto_key_t ,
	crypto_data_t , crypto_data_t , crypto_spi_ctx_template_t);

	static const crypto_mac_ops_t aes_mac_ops = {
	.mac_init = NULL,
	.mac = NULL,
	.mac_update = NULL,
	.mac_final = NULL,
	.mac_atomic = aes_mac_atomic,
	.mac_verify_atomic = aes_mac_verify_atomic
	};

	static int aes_create_ctx_template(crypto_mechanism_t , crypto_key_t ,
	crypto_spi_ctx_template_t , size_t );
	static int aes_free_context(crypto_ctx_t *);

	static const crypto_ctx_ops_t aes_ctx_ops = {
	.create_ctx_template = aes_create_ctx_template,
	.free_context = aes_free_context
	};

	static const crypto_ops_t aes_crypto_ops = {
	NULL,
	&aes_cipher_ops,
	&aes_mac_ops,
	&aes_ctx_ops,
	};

	static const crypto_provider_info_t aes_prov_info = {
	"AES Software Provider",
	&aes_crypto_ops,
	sizeof (aes_mech_info_tab) / sizeof (crypto_mech_info_t),
	aes_mech_info_tab
	};

	static crypto_kcf_provider_handle_t aes_prov_handle = 0;
	static crypto_data_t null_crypto_data = { CRYPTO_DATA_RAW };

	int
	aes_mod_init(void)
	{
	/* Determine the fastest available implementation. */
	aes_impl_init();
	gcm_impl_init();

	/* Register with KCF. If the registration fails, remove the module. */
	if (crypto_register_provider(&aes_prov_info, &aes_prov_handle))
	return (EACCES);

	return (0);
	}

	int
	aes_mod_fini(void)
	{
	/* Unregister from KCF if module is registered */
	if (aes_prov_handle != 0) {
	if (crypto_unregister_provider(aes_prov_handle))
	return (EBUSY);

	aes_prov_handle = 0;
	}

	return (0);
	}

	static int
	aes_check_mech_param(crypto_mechanism_t mechanism, aes_ctx_t *ctx)
	{
	void *p = NULL;
	boolean_t param_required = B_TRUE;
	size_t param_len;
	void (alloc_fun)(int);
	int rv = CRYPTO_SUCCESS;

	switch (mechanism->cm_type) {
	case AES_ECB_MECH_INFO_TYPE:
	param_required = B_FALSE;
	alloc_fun = ecb_alloc_ctx;
	break;
	case AES_CBC_MECH_INFO_TYPE:
	param_len = AES_BLOCK_LEN;
	alloc_fun = cbc_alloc_ctx;
	break;
	case AES_CTR_MECH_INFO_TYPE:
	param_len = sizeof (CK_AES_CTR_PARAMS);
	alloc_fun = ctr_alloc_ctx;
	break;
	case AES_CCM_MECH_INFO_TYPE:
	param_len = sizeof (CK_AES_CCM_PARAMS);
	alloc_fun = ccm_alloc_ctx;
	break;
	case AES_GCM_MECH_INFO_TYPE:
	param_len = sizeof (CK_AES_GCM_PARAMS);
	alloc_fun = gcm_alloc_ctx;
	break;
	case AES_GMAC_MECH_INFO_TYPE:
	param_len = sizeof (CK_AES_GMAC_PARAMS);
	alloc_fun = gmac_alloc_ctx;
	break;
	default:
	rv = CRYPTO_MECHANISM_INVALID;
	return (rv);
	}
	if (param_required && mechanism->cm_param != NULL &&
	mechanism->cm_param_len != param_len) {
	rv = CRYPTO_MECHANISM_PARAM_INVALID;
	}
	if (ctx != NULL) {
	p = (alloc_fun)(KM_SLEEP);
	*ctx = p;
	}
	return (rv);
	}

	/*
	* Initialize key schedules for AES
	*/
	static int
	init_keysched(crypto_key_t key, void newbie)
	{
	if (key->ck_length < AES_MINBITS \|\|
	key->ck_length > AES_MAXBITS) {
	return (CRYPTO_KEY_SIZE_RANGE);
	}

	/* key length must be either 128, 192, or 256 */
	if ((key->ck_length & 63) != 0)
	return (CRYPTO_KEY_SIZE_RANGE);

	aes_init_keysched(key->ck_data, key->ck_length, newbie);
	return (CRYPTO_SUCCESS);
	}

	static int
	aes_encrypt_init(crypto_ctx_t ctx, crypto_mechanism_t mechanism,
	crypto_key_t *key, crypto_spi_ctx_template_t template)
	{
	return (aes_common_init(ctx, mechanism, key, template, B_TRUE));
	}

	static int
	aes_decrypt_init(crypto_ctx_t ctx, crypto_mechanism_t mechanism,
	crypto_key_t *key, crypto_spi_ctx_template_t template)
	{
	return (aes_common_init(ctx, mechanism, key, template, B_FALSE));
	}



	/*
	* KCF software provider encrypt entry points.
	*/
	static int
	aes_common_init(crypto_ctx_t ctx, crypto_mechanism_t mechanism,
	crypto_key_t *key, crypto_spi_ctx_template_t template,
	boolean_t is_encrypt_init)
	{
	aes_ctx_t *aes_ctx;
	int rv;

	if ((rv = aes_check_mech_param(mechanism, &aes_ctx))
	!= CRYPTO_SUCCESS)
	return (rv);

	rv = aes_common_init_ctx(aes_ctx, template, mechanism, key, KM_SLEEP,
	is_encrypt_init);
	if (rv != CRYPTO_SUCCESS) {
	crypto_free_mode_ctx(aes_ctx);
	return (rv);
	}

	ctx->cc_provider_private = aes_ctx;

	return (CRYPTO_SUCCESS);
	}

	static void
	aes_copy_block64(uint8_t in, uint64_t out)
	{
	if (IS_P2ALIGNED(in, sizeof (uint64_t))) {
	/* LINTED: pointer alignment */
	out[0] = (uint64_t )&in[0];
	/* LINTED: pointer alignment */
	out[1] = (uint64_t )&in[8];
	} else {
	uint8_t iv8 = (uint8_t )&out[0];

	AES_COPY_BLOCK(in, iv8);
	}
	}


	static int
	aes_encrypt(crypto_ctx_t ctx, crypto_data_t plaintext,
	crypto_data_t *ciphertext)
	{
	int ret = CRYPTO_FAILED;

	aes_ctx_t *aes_ctx;
	size_t saved_length, saved_offset, length_needed;

	ASSERT(ctx->cc_provider_private != NULL);
	aes_ctx = ctx->cc_provider_private;

	/*
	* For block ciphers, plaintext must be a multiple of AES block size.
	* This test is only valid for ciphers whose blocksize is a power of 2.
	*/
	if (((aes_ctx->ac_flags & (CTR_MODE\|CCM_MODE\|GCM_MODE\|GMAC_MODE))
	== 0) && (plaintext->cd_length & (AES_BLOCK_LEN - 1)) != 0)
	return (CRYPTO_DATA_LEN_RANGE);

	ASSERT(ciphertext != NULL);

	/*
	* We need to just return the length needed to store the output.
	* We should not destroy the context for the following case.
	*/
	switch (aes_ctx->ac_flags & (CCM_MODE\|GCM_MODE\|GMAC_MODE)) {
	case CCM_MODE:
	length_needed = plaintext->cd_length + aes_ctx->ac_mac_len;
	break;
	case GCM_MODE:
	length_needed = plaintext->cd_length + aes_ctx->ac_tag_len;
	break;
	case GMAC_MODE:
	if (plaintext->cd_length != 0)
	return (CRYPTO_ARGUMENTS_BAD);

	length_needed = aes_ctx->ac_tag_len;
	break;
	default:
	length_needed = plaintext->cd_length;
	}

	if (ciphertext->cd_length < length_needed) {
	ciphertext->cd_length = length_needed;
	return (CRYPTO_BUFFER_TOO_SMALL);
	}

	saved_length = ciphertext->cd_length;
	saved_offset = ciphertext->cd_offset;

	/*
	* Do an update on the specified input data.
	*/
	ret = aes_encrypt_update(ctx, plaintext, ciphertext);
	if (ret != CRYPTO_SUCCESS) {
	return (ret);
	}

	/*
	* For CCM mode, aes_ccm_encrypt_final() will take care of any
	* left-over unprocessed data, and compute the MAC
	*/
	if (aes_ctx->ac_flags & CCM_MODE) {
	/*
	* ccm_encrypt_final() will compute the MAC and append
	* it to existing ciphertext. So, need to adjust the left over
	* length value accordingly
	*/

	/* order of following 2 lines MUST not be reversed */
	ciphertext->cd_offset = ciphertext->cd_length;
	ciphertext->cd_length = saved_length - ciphertext->cd_length;
	ret = ccm_encrypt_final((ccm_ctx_t *)aes_ctx, ciphertext,
	AES_BLOCK_LEN, aes_encrypt_block, aes_xor_block);
	if (ret != CRYPTO_SUCCESS) {
	return (ret);
	}

	if (plaintext != ciphertext) {
	ciphertext->cd_length =
	ciphertext->cd_offset - saved_offset;
	}
	ciphertext->cd_offset = saved_offset;
	} else if (aes_ctx->ac_flags & (GCM_MODE\|GMAC_MODE)) {
	/*
	* gcm_encrypt_final() will compute the MAC and append
	* it to existing ciphertext. So, need to adjust the left over
	* length value accordingly
	*/

	/* order of following 2 lines MUST not be reversed */
	ciphertext->cd_offset = ciphertext->cd_length;
	ciphertext->cd_length = saved_length - ciphertext->cd_length;
	ret = gcm_encrypt_final((gcm_ctx_t *)aes_ctx, ciphertext,
	AES_BLOCK_LEN, aes_encrypt_block, aes_copy_block,
	aes_xor_block);
	if (ret != CRYPTO_SUCCESS) {
	return (ret);
	}

	if (plaintext != ciphertext) {
	ciphertext->cd_length =
	ciphertext->cd_offset - saved_offset;
	}
	ciphertext->cd_offset = saved_offset;
	}

	ASSERT(aes_ctx->ac_remainder_len == 0);
	(void) aes_free_context(ctx);

	return (ret);
	}


	static int
	aes_decrypt(crypto_ctx_t ctx, crypto_data_t ciphertext,
	crypto_data_t *plaintext)
	{
	int ret = CRYPTO_FAILED;

	aes_ctx_t *aes_ctx;
	off_t saved_offset;
	size_t saved_length, length_needed;

	ASSERT(ctx->cc_provider_private != NULL);
	aes_ctx = ctx->cc_provider_private;

	/*
	* For block ciphers, plaintext must be a multiple of AES block size.
	* This test is only valid for ciphers whose blocksize is a power of 2.
	*/
	if (((aes_ctx->ac_flags & (CTR_MODE\|CCM_MODE\|GCM_MODE\|GMAC_MODE))
	== 0) && (ciphertext->cd_length & (AES_BLOCK_LEN - 1)) != 0) {
	return (CRYPTO_ENCRYPTED_DATA_LEN_RANGE);
	}

	ASSERT(plaintext != NULL);

	/*
	* Return length needed to store the output.
	* Do not destroy context when plaintext buffer is too small.
	*
	* CCM: plaintext is MAC len smaller than cipher text
	* GCM: plaintext is TAG len smaller than cipher text
	* GMAC: plaintext length must be zero
	*/
	switch (aes_ctx->ac_flags & (CCM_MODE\|GCM_MODE\|GMAC_MODE)) {
	case CCM_MODE:
	length_needed = aes_ctx->ac_processed_data_len;
	break;
	case GCM_MODE:
	length_needed = ciphertext->cd_length - aes_ctx->ac_tag_len;
	break;
	case GMAC_MODE:
	if (plaintext->cd_length != 0)
	return (CRYPTO_ARGUMENTS_BAD);

	length_needed = 0;
	break;
	default:
	length_needed = ciphertext->cd_length;
	}

	if (plaintext->cd_length < length_needed) {
	plaintext->cd_length = length_needed;
	return (CRYPTO_BUFFER_TOO_SMALL);
	}

	saved_offset = plaintext->cd_offset;
	saved_length = plaintext->cd_length;

	/*
	* Do an update on the specified input data.
	*/
	ret = aes_decrypt_update(ctx, ciphertext, plaintext);
	if (ret != CRYPTO_SUCCESS) {
	goto cleanup;
	}

	if (aes_ctx->ac_flags & CCM_MODE) {
	ASSERT(aes_ctx->ac_processed_data_len == aes_ctx->ac_data_len);
	ASSERT(aes_ctx->ac_processed_mac_len == aes_ctx->ac_mac_len);

	/* order of following 2 lines MUST not be reversed */
	plaintext->cd_offset = plaintext->cd_length;
	plaintext->cd_length = saved_length - plaintext->cd_length;

	ret = ccm_decrypt_final((ccm_ctx_t *)aes_ctx, plaintext,
	AES_BLOCK_LEN, aes_encrypt_block, aes_copy_block,
	aes_xor_block);
	if (ret == CRYPTO_SUCCESS) {
	if (plaintext != ciphertext) {
	plaintext->cd_length =
	plaintext->cd_offset - saved_offset;
	}
	} else {
	plaintext->cd_length = saved_length;
	}

	plaintext->cd_offset = saved_offset;
	} else if (aes_ctx->ac_flags & (GCM_MODE\|GMAC_MODE)) {
	/* order of following 2 lines MUST not be reversed */
	plaintext->cd_offset = plaintext->cd_length;
	plaintext->cd_length = saved_length - plaintext->cd_length;

	ret = gcm_decrypt_final((gcm_ctx_t *)aes_ctx, plaintext,
	AES_BLOCK_LEN, aes_encrypt_block, aes_xor_block);
	if (ret == CRYPTO_SUCCESS) {
	if (plaintext != ciphertext) {
	plaintext->cd_length =
	plaintext->cd_offset - saved_offset;
	}
	} else {
	plaintext->cd_length = saved_length;
	}

	plaintext->cd_offset = saved_offset;
	}

	ASSERT(aes_ctx->ac_remainder_len == 0);

	cleanup:
	(void) aes_free_context(ctx);

	return (ret);
	}


	static int
	aes_encrypt_update(crypto_ctx_t ctx, crypto_data_t plaintext,
	crypto_data_t *ciphertext)
	{
	off_t saved_offset;
	size_t saved_length, out_len;
	int ret = CRYPTO_SUCCESS;
	aes_ctx_t *aes_ctx;

	ASSERT(ctx->cc_provider_private != NULL);
	aes_ctx = ctx->cc_provider_private;

	ASSERT(ciphertext != NULL);

	/* compute number of bytes that will hold the ciphertext */
	out_len = aes_ctx->ac_remainder_len;
	out_len += plaintext->cd_length;
	out_len &= ~(AES_BLOCK_LEN - 1);

	/* return length needed to store the output */
	if (ciphertext->cd_length < out_len) {
	ciphertext->cd_length = out_len;
	return (CRYPTO_BUFFER_TOO_SMALL);
	}

	saved_offset = ciphertext->cd_offset;
	saved_length = ciphertext->cd_length;

	/*
	* Do the AES update on the specified input data.
	*/
	switch (plaintext->cd_format) {
	case CRYPTO_DATA_RAW:
	ret = crypto_update_iov(ctx->cc_provider_private,
	plaintext, ciphertext, aes_encrypt_contiguous_blocks);
	break;
	case CRYPTO_DATA_UIO:
	ret = crypto_update_uio(ctx->cc_provider_private,
	plaintext, ciphertext, aes_encrypt_contiguous_blocks);
	break;
	default:
	ret = CRYPTO_ARGUMENTS_BAD;
	}

	/*
	* Since AES counter mode is a stream cipher, we call
	* ctr_mode_final() to pick up any remaining bytes.
	* It is an internal function that does not destroy
	* the context like normal final routines.
	*/
	if ((aes_ctx->ac_flags & CTR_MODE) && (aes_ctx->ac_remainder_len > 0)) {
	ret = ctr_mode_final((ctr_ctx_t *)aes_ctx,
	ciphertext, aes_encrypt_block);
	}

	if (ret == CRYPTO_SUCCESS) {
	if (plaintext != ciphertext)
	ciphertext->cd_length =
	ciphertext->cd_offset - saved_offset;
	} else {
	ciphertext->cd_length = saved_length;
	}
	ciphertext->cd_offset = saved_offset;

	return (ret);
	}


	static int
	aes_decrypt_update(crypto_ctx_t ctx, crypto_data_t ciphertext,
	crypto_data_t *plaintext)
	{
	off_t saved_offset;
	size_t saved_length, out_len;
	int ret = CRYPTO_SUCCESS;
	aes_ctx_t *aes_ctx;

	ASSERT(ctx->cc_provider_private != NULL);
	aes_ctx = ctx->cc_provider_private;

	ASSERT(plaintext != NULL);

	/*
	* Compute number of bytes that will hold the plaintext.
	* This is not necessary for CCM, GCM, and GMAC since these
	* mechanisms never return plaintext for update operations.
	*/
	if ((aes_ctx->ac_flags & (CCM_MODE\|GCM_MODE\|GMAC_MODE)) == 0) {
	out_len = aes_ctx->ac_remainder_len;
	out_len += ciphertext->cd_length;
	out_len &= ~(AES_BLOCK_LEN - 1);

	/* return length needed to store the output */
	if (plaintext->cd_length < out_len) {
	plaintext->cd_length = out_len;
	return (CRYPTO_BUFFER_TOO_SMALL);
	}
	}

	saved_offset = plaintext->cd_offset;
	saved_length = plaintext->cd_length;

	/*
	* Do the AES update on the specified input data.
	*/
	switch (ciphertext->cd_format) {
	case CRYPTO_DATA_RAW:
	ret = crypto_update_iov(ctx->cc_provider_private,
	ciphertext, plaintext, aes_decrypt_contiguous_blocks);
	break;
	case CRYPTO_DATA_UIO:
	ret = crypto_update_uio(ctx->cc_provider_private,
	ciphertext, plaintext, aes_decrypt_contiguous_blocks);
	break;
	default:
	ret = CRYPTO_ARGUMENTS_BAD;
	}

	/*
	* Since AES counter mode is a stream cipher, we call
	* ctr_mode_final() to pick up any remaining bytes.
	* It is an internal function that does not destroy
	* the context like normal final routines.
	*/
	if ((aes_ctx->ac_flags & CTR_MODE) && (aes_ctx->ac_remainder_len > 0)) {
	ret = ctr_mode_final((ctr_ctx_t *)aes_ctx, plaintext,
	aes_encrypt_block);
	if (ret == CRYPTO_DATA_LEN_RANGE)
	ret = CRYPTO_ENCRYPTED_DATA_LEN_RANGE;
	}

	if (ret == CRYPTO_SUCCESS) {
	if (ciphertext != plaintext)
	plaintext->cd_length =
	plaintext->cd_offset - saved_offset;
	} else {
	plaintext->cd_length = saved_length;
	}
	plaintext->cd_offset = saved_offset;


	return (ret);
	}

	static int
	aes_encrypt_final(crypto_ctx_t ctx, crypto_data_t data)
	{
	aes_ctx_t *aes_ctx;
	int ret;

	ASSERT(ctx->cc_provider_private != NULL);
	aes_ctx = ctx->cc_provider_private;

	if (data->cd_format != CRYPTO_DATA_RAW &&
	data->cd_format != CRYPTO_DATA_UIO) {
	return (CRYPTO_ARGUMENTS_BAD);
	}

	if (aes_ctx->ac_flags & CTR_MODE) {
	if (aes_ctx->ac_remainder_len > 0) {
	ret = ctr_mode_final((ctr_ctx_t *)aes_ctx, data,
	aes_encrypt_block);
	if (ret != CRYPTO_SUCCESS)
	return (ret);
	}
	} else if (aes_ctx->ac_flags & CCM_MODE) {
	ret = ccm_encrypt_final((ccm_ctx_t *)aes_ctx, data,
	AES_BLOCK_LEN, aes_encrypt_block, aes_xor_block);
	if (ret != CRYPTO_SUCCESS) {
	return (ret);
	}
	} else if (aes_ctx->ac_flags & (GCM_MODE\|GMAC_MODE)) {
	size_t saved_offset = data->cd_offset;

	ret = gcm_encrypt_final((gcm_ctx_t *)aes_ctx, data,
	AES_BLOCK_LEN, aes_encrypt_block, aes_copy_block,
	aes_xor_block);
	if (ret != CRYPTO_SUCCESS) {
	return (ret);
	}
	data->cd_length = data->cd_offset - saved_offset;
	data->cd_offset = saved_offset;
	} else {
	/*
	* There must be no unprocessed plaintext.
	* This happens if the length of the last data is
	* not a multiple of the AES block length.
	*/
	if (aes_ctx->ac_remainder_len > 0) {
	return (CRYPTO_DATA_LEN_RANGE);
	}
	data->cd_length = 0;
	}

	(void) aes_free_context(ctx);

	return (CRYPTO_SUCCESS);
	}

	static int
	aes_decrypt_final(crypto_ctx_t ctx, crypto_data_t data)
	{
	aes_ctx_t *aes_ctx;
	int ret;
	off_t saved_offset;
	size_t saved_length;

	ASSERT(ctx->cc_provider_private != NULL);
	aes_ctx = ctx->cc_provider_private;

	if (data->cd_format != CRYPTO_DATA_RAW &&
	data->cd_format != CRYPTO_DATA_UIO) {
	return (CRYPTO_ARGUMENTS_BAD);
	}

	/*
	* There must be no unprocessed ciphertext.
	* This happens if the length of the last ciphertext is
	* not a multiple of the AES block length.
	*/
	if (aes_ctx->ac_remainder_len > 0) {
	if ((aes_ctx->ac_flags & CTR_MODE) == 0)
	return (CRYPTO_ENCRYPTED_DATA_LEN_RANGE);
	else {
	ret = ctr_mode_final((ctr_ctx_t *)aes_ctx, data,
	aes_encrypt_block);
	if (ret == CRYPTO_DATA_LEN_RANGE)
	ret = CRYPTO_ENCRYPTED_DATA_LEN_RANGE;
	if (ret != CRYPTO_SUCCESS)
	return (ret);
	}
	}

	if (aes_ctx->ac_flags & CCM_MODE) {
	/*
	* This is where all the plaintext is returned, make sure
	* the plaintext buffer is big enough
	*/
	size_t pt_len = aes_ctx->ac_data_len;
	if (data->cd_length < pt_len) {
	data->cd_length = pt_len;
	return (CRYPTO_BUFFER_TOO_SMALL);
	}

	ASSERT(aes_ctx->ac_processed_data_len == pt_len);
	ASSERT(aes_ctx->ac_processed_mac_len == aes_ctx->ac_mac_len);
	saved_offset = data->cd_offset;
	saved_length = data->cd_length;
	ret = ccm_decrypt_final((ccm_ctx_t *)aes_ctx, data,
	AES_BLOCK_LEN, aes_encrypt_block, aes_copy_block,
	aes_xor_block);
	if (ret == CRYPTO_SUCCESS) {
	data->cd_length = data->cd_offset - saved_offset;
	} else {
	data->cd_length = saved_length;
	}

	data->cd_offset = saved_offset;
	if (ret != CRYPTO_SUCCESS) {
	return (ret);
	}
	} else if (aes_ctx->ac_flags & (GCM_MODE\|GMAC_MODE)) {
	/*
	* This is where all the plaintext is returned, make sure
	* the plaintext buffer is big enough
	*/
	gcm_ctx_t ctx = (gcm_ctx_t )aes_ctx;
	size_t pt_len = ctx->gcm_processed_data_len - ctx->gcm_tag_len;

	if (data->cd_length < pt_len) {
	data->cd_length = pt_len;
	return (CRYPTO_BUFFER_TOO_SMALL);
	}

	saved_offset = data->cd_offset;
	saved_length = data->cd_length;
	ret = gcm_decrypt_final((gcm_ctx_t *)aes_ctx, data,
	AES_BLOCK_LEN, aes_encrypt_block, aes_xor_block);
	if (ret == CRYPTO_SUCCESS) {
	data->cd_length = data->cd_offset - saved_offset;
	} else {
	data->cd_length = saved_length;
	}

	data->cd_offset = saved_offset;
	if (ret != CRYPTO_SUCCESS) {
	return (ret);
	}
	}


	if ((aes_ctx->ac_flags & (CTR_MODE\|CCM_MODE\|GCM_MODE\|GMAC_MODE)) == 0) {
	data->cd_length = 0;
	}

	(void) aes_free_context(ctx);

	return (CRYPTO_SUCCESS);
	}

	static int
	aes_encrypt_atomic(crypto_mechanism_t *mechanism,
	crypto_key_t key, crypto_data_t plaintext, crypto_data_t *ciphertext,
	crypto_spi_ctx_template_t template)
	{
	- aes_ctx_t aes_ctx = {{{{0}}}};
	+ aes_ctx_t aes_ctx;
	off_t saved_offset;
	size_t saved_length;
	size_t length_needed;
	int ret;

	+ memset(&aes_ctx, 0, sizeof (aes_ctx_t));
	+
	ASSERT(ciphertext != NULL);

	/*
	* CTR, CCM, GCM, and GMAC modes do not require that plaintext
	* be a multiple of AES block size.
	*/
	switch (mechanism->cm_type) {
	case AES_CTR_MECH_INFO_TYPE:
	case AES_CCM_MECH_INFO_TYPE:
	case AES_GCM_MECH_INFO_TYPE:
	case AES_GMAC_MECH_INFO_TYPE:
	break;
	default:
	if ((plaintext->cd_length & (AES_BLOCK_LEN - 1)) != 0)
	return (CRYPTO_DATA_LEN_RANGE);
	}

	if ((ret = aes_check_mech_param(mechanism, NULL)) != CRYPTO_SUCCESS)
	return (ret);

	ret = aes_common_init_ctx(&aes_ctx, template, mechanism, key,
	KM_SLEEP, B_TRUE);
	if (ret != CRYPTO_SUCCESS)
	return (ret);

	switch (mechanism->cm_type) {
	case AES_CCM_MECH_INFO_TYPE:
	length_needed = plaintext->cd_length + aes_ctx.ac_mac_len;
	break;
	case AES_GMAC_MECH_INFO_TYPE:
	if (plaintext->cd_length != 0)
	return (CRYPTO_ARGUMENTS_BAD);
	zfs_fallthrough;
	case AES_GCM_MECH_INFO_TYPE:
	length_needed = plaintext->cd_length + aes_ctx.ac_tag_len;
	break;
	default:
	length_needed = plaintext->cd_length;
	}

	/* return size of buffer needed to store output */
	if (ciphertext->cd_length < length_needed) {
	ciphertext->cd_length = length_needed;
	ret = CRYPTO_BUFFER_TOO_SMALL;
	goto out;
	}

	saved_offset = ciphertext->cd_offset;
	saved_length = ciphertext->cd_length;

	/*
	* Do an update on the specified input data.
	*/
	switch (plaintext->cd_format) {
	case CRYPTO_DATA_RAW:
	ret = crypto_update_iov(&aes_ctx, plaintext, ciphertext,
	aes_encrypt_contiguous_blocks);
	break;
	case CRYPTO_DATA_UIO:
	ret = crypto_update_uio(&aes_ctx, plaintext, ciphertext,
	aes_encrypt_contiguous_blocks);
	break;
	default:
	ret = CRYPTO_ARGUMENTS_BAD;
	}

	if (ret == CRYPTO_SUCCESS) {
	if (mechanism->cm_type == AES_CCM_MECH_INFO_TYPE) {
	ret = ccm_encrypt_final((ccm_ctx_t *)&aes_ctx,
	ciphertext, AES_BLOCK_LEN, aes_encrypt_block,
	aes_xor_block);
	if (ret != CRYPTO_SUCCESS)
	goto out;
	ASSERT(aes_ctx.ac_remainder_len == 0);
	} else if (mechanism->cm_type == AES_GCM_MECH_INFO_TYPE \|\|
	mechanism->cm_type == AES_GMAC_MECH_INFO_TYPE) {
	ret = gcm_encrypt_final((gcm_ctx_t *)&aes_ctx,
	ciphertext, AES_BLOCK_LEN, aes_encrypt_block,
	aes_copy_block, aes_xor_block);
	if (ret != CRYPTO_SUCCESS)
	goto out;
	ASSERT(aes_ctx.ac_remainder_len == 0);
	} else if (mechanism->cm_type == AES_CTR_MECH_INFO_TYPE) {
	if (aes_ctx.ac_remainder_len > 0) {
	ret = ctr_mode_final((ctr_ctx_t *)&aes_ctx,
	ciphertext, aes_encrypt_block);
	if (ret != CRYPTO_SUCCESS)
	goto out;
	}
	} else {
	ASSERT(aes_ctx.ac_remainder_len == 0);
	}

	if (plaintext != ciphertext) {
	ciphertext->cd_length =
	ciphertext->cd_offset - saved_offset;
	}
	} else {
	ciphertext->cd_length = saved_length;
	}
	ciphertext->cd_offset = saved_offset;

	out:
	if (aes_ctx.ac_flags & PROVIDER_OWNS_KEY_SCHEDULE) {
	memset(aes_ctx.ac_keysched, 0, aes_ctx.ac_keysched_len);
	kmem_free(aes_ctx.ac_keysched, aes_ctx.ac_keysched_len);
	}
	if (aes_ctx.ac_flags & (GCM_MODE\|GMAC_MODE)) {
	gcm_clear_ctx((gcm_ctx_t *)&aes_ctx);
	}
	return (ret);
	}

	static int
	aes_decrypt_atomic(crypto_mechanism_t *mechanism,
	crypto_key_t key, crypto_data_t ciphertext, crypto_data_t *plaintext,
	crypto_spi_ctx_template_t template)
	{
	- aes_ctx_t aes_ctx = {{{{0}}}};
	+ aes_ctx_t aes_ctx;
	off_t saved_offset;
	size_t saved_length;
	size_t length_needed;
	int ret;

	+ memset(&aes_ctx, 0, sizeof (aes_ctx_t));
	+
	ASSERT(plaintext != NULL);

	/*
	* CCM, GCM, CTR, and GMAC modes do not require that ciphertext
	* be a multiple of AES block size.
	*/
	switch (mechanism->cm_type) {
	case AES_CTR_MECH_INFO_TYPE:
	case AES_CCM_MECH_INFO_TYPE:
	case AES_GCM_MECH_INFO_TYPE:
	case AES_GMAC_MECH_INFO_TYPE:
	break;
	default:
	if ((ciphertext->cd_length & (AES_BLOCK_LEN - 1)) != 0)
	return (CRYPTO_ENCRYPTED_DATA_LEN_RANGE);
	}

	if ((ret = aes_check_mech_param(mechanism, NULL)) != CRYPTO_SUCCESS)
	return (ret);

	ret = aes_common_init_ctx(&aes_ctx, template, mechanism, key,
	KM_SLEEP, B_FALSE);
	if (ret != CRYPTO_SUCCESS)
	return (ret);

	switch (mechanism->cm_type) {
	case AES_CCM_MECH_INFO_TYPE:
	length_needed = aes_ctx.ac_data_len;
	break;
	case AES_GCM_MECH_INFO_TYPE:
	length_needed = ciphertext->cd_length - aes_ctx.ac_tag_len;
	break;
	case AES_GMAC_MECH_INFO_TYPE:
	if (plaintext->cd_length != 0)
	return (CRYPTO_ARGUMENTS_BAD);
	length_needed = 0;
	break;
	default:
	length_needed = ciphertext->cd_length;
	}

	/* return size of buffer needed to store output */
	if (plaintext->cd_length < length_needed) {
	plaintext->cd_length = length_needed;
	ret = CRYPTO_BUFFER_TOO_SMALL;
	goto out;
	}

	saved_offset = plaintext->cd_offset;
	saved_length = plaintext->cd_length;

	/*
	* Do an update on the specified input data.
	*/
	switch (ciphertext->cd_format) {
	case CRYPTO_DATA_RAW:
	ret = crypto_update_iov(&aes_ctx, ciphertext, plaintext,
	aes_decrypt_contiguous_blocks);
	break;
	case CRYPTO_DATA_UIO:
	ret = crypto_update_uio(&aes_ctx, ciphertext, plaintext,
	aes_decrypt_contiguous_blocks);
	break;
	default:
	ret = CRYPTO_ARGUMENTS_BAD;
	}

	if (ret == CRYPTO_SUCCESS) {
	if (mechanism->cm_type == AES_CCM_MECH_INFO_TYPE) {
	ASSERT(aes_ctx.ac_processed_data_len
	== aes_ctx.ac_data_len);
	ASSERT(aes_ctx.ac_processed_mac_len
	== aes_ctx.ac_mac_len);
	ret = ccm_decrypt_final((ccm_ctx_t *)&aes_ctx,
	plaintext, AES_BLOCK_LEN, aes_encrypt_block,
	aes_copy_block, aes_xor_block);
	ASSERT(aes_ctx.ac_remainder_len == 0);
	if ((ret == CRYPTO_SUCCESS) &&
	(ciphertext != plaintext)) {
	plaintext->cd_length =
	plaintext->cd_offset - saved_offset;
	} else {
	plaintext->cd_length = saved_length;
	}
	} else if (mechanism->cm_type == AES_GCM_MECH_INFO_TYPE \|\|
	mechanism->cm_type == AES_GMAC_MECH_INFO_TYPE) {
	ret = gcm_decrypt_final((gcm_ctx_t *)&aes_ctx,
	plaintext, AES_BLOCK_LEN, aes_encrypt_block,
	aes_xor_block);
	ASSERT(aes_ctx.ac_remainder_len == 0);
	if ((ret == CRYPTO_SUCCESS) &&
	(ciphertext != plaintext)) {
	plaintext->cd_length =
	plaintext->cd_offset - saved_offset;
	} else {
	plaintext->cd_length = saved_length;
	}
	} else if (mechanism->cm_type != AES_CTR_MECH_INFO_TYPE) {
	ASSERT(aes_ctx.ac_remainder_len == 0);
	if (ciphertext != plaintext)
	plaintext->cd_length =
	plaintext->cd_offset - saved_offset;
	} else {
	if (aes_ctx.ac_remainder_len > 0) {
	ret = ctr_mode_final((ctr_ctx_t *)&aes_ctx,
	plaintext, aes_encrypt_block);
	if (ret == CRYPTO_DATA_LEN_RANGE)
	ret = CRYPTO_ENCRYPTED_DATA_LEN_RANGE;
	if (ret != CRYPTO_SUCCESS)
	goto out;
	}
	if (ciphertext != plaintext)
	plaintext->cd_length =
	plaintext->cd_offset - saved_offset;
	}
	} else {
	plaintext->cd_length = saved_length;
	}
	plaintext->cd_offset = saved_offset;

	out:
	if (aes_ctx.ac_flags & PROVIDER_OWNS_KEY_SCHEDULE) {
	memset(aes_ctx.ac_keysched, 0, aes_ctx.ac_keysched_len);
	kmem_free(aes_ctx.ac_keysched, aes_ctx.ac_keysched_len);
	}

	if (aes_ctx.ac_flags & CCM_MODE) {
	if (aes_ctx.ac_pt_buf != NULL) {
	vmem_free(aes_ctx.ac_pt_buf, aes_ctx.ac_data_len);
	}
	} else if (aes_ctx.ac_flags & (GCM_MODE\|GMAC_MODE)) {
	gcm_clear_ctx((gcm_ctx_t *)&aes_ctx);
	}

	return (ret);
	}

	/*
	* KCF software provider context template entry points.
	*/
	static int
	aes_create_ctx_template(crypto_mechanism_t mechanism, crypto_key_t key,
	crypto_spi_ctx_template_t tmpl, size_t tmpl_size)
	{
	void *keysched;
	size_t size;
	int rv;

	if (mechanism->cm_type != AES_ECB_MECH_INFO_TYPE &&
	mechanism->cm_type != AES_CBC_MECH_INFO_TYPE &&
	mechanism->cm_type != AES_CTR_MECH_INFO_TYPE &&
	mechanism->cm_type != AES_CCM_MECH_INFO_TYPE &&
	mechanism->cm_type != AES_GCM_MECH_INFO_TYPE &&
	mechanism->cm_type != AES_GMAC_MECH_INFO_TYPE)
	return (CRYPTO_MECHANISM_INVALID);

	if ((keysched = aes_alloc_keysched(&size, KM_SLEEP)) == NULL) {
	return (CRYPTO_HOST_MEMORY);
	}

	/*
	* Initialize key schedule. Key length information is stored
	* in the key.
	*/
	if ((rv = init_keysched(key, keysched)) != CRYPTO_SUCCESS) {
	memset(keysched, 0, size);
	kmem_free(keysched, size);
	return (rv);
	}

	*tmpl = keysched;
	*tmpl_size = size;

	return (CRYPTO_SUCCESS);
	}


	static int
	aes_free_context(crypto_ctx_t *ctx)
	{
	aes_ctx_t *aes_ctx = ctx->cc_provider_private;

	if (aes_ctx != NULL) {
	if (aes_ctx->ac_flags & PROVIDER_OWNS_KEY_SCHEDULE) {
	ASSERT(aes_ctx->ac_keysched_len != 0);
	memset(aes_ctx->ac_keysched, 0,
	aes_ctx->ac_keysched_len);
	kmem_free(aes_ctx->ac_keysched,
	aes_ctx->ac_keysched_len);
	}
	crypto_free_mode_ctx(aes_ctx);
	ctx->cc_provider_private = NULL;
	}

	return (CRYPTO_SUCCESS);
	}


	static int
	aes_common_init_ctx(aes_ctx_t aes_ctx, crypto_spi_ctx_template_t template,
	crypto_mechanism_t mechanism, crypto_key_t key, int kmflag,
	boolean_t is_encrypt_init)
	{
	int rv = CRYPTO_SUCCESS;
	void *keysched;
	size_t size = 0;

	if (template == NULL) {
	if ((keysched = aes_alloc_keysched(&size, kmflag)) == NULL)
	return (CRYPTO_HOST_MEMORY);
	/*
	* Initialize key schedule.
	* Key length is stored in the key.
	*/
	if ((rv = init_keysched(key, keysched)) != CRYPTO_SUCCESS) {
	kmem_free(keysched, size);
	return (rv);
	}

	aes_ctx->ac_flags \|= PROVIDER_OWNS_KEY_SCHEDULE;
	aes_ctx->ac_keysched_len = size;
	} else {
	keysched = template;
	}
	aes_ctx->ac_keysched = keysched;

	switch (mechanism->cm_type) {
	case AES_CBC_MECH_INFO_TYPE:
	rv = cbc_init_ctx((cbc_ctx_t *)aes_ctx, mechanism->cm_param,
	mechanism->cm_param_len, AES_BLOCK_LEN, aes_copy_block64);
	break;
	case AES_CTR_MECH_INFO_TYPE: {
	CK_AES_CTR_PARAMS *pp;

	if (mechanism->cm_param == NULL \|\|
	mechanism->cm_param_len != sizeof (CK_AES_CTR_PARAMS)) {
	return (CRYPTO_MECHANISM_PARAM_INVALID);
	}
	pp = (CK_AES_CTR_PARAMS )(void )mechanism->cm_param;
	rv = ctr_init_ctx((ctr_ctx_t *)aes_ctx, pp->ulCounterBits,
	pp->cb, aes_copy_block);
	break;
	}
	case AES_CCM_MECH_INFO_TYPE:
	if (mechanism->cm_param == NULL \|\|
	mechanism->cm_param_len != sizeof (CK_AES_CCM_PARAMS)) {
	return (CRYPTO_MECHANISM_PARAM_INVALID);
	}
	rv = ccm_init_ctx((ccm_ctx_t *)aes_ctx, mechanism->cm_param,
	kmflag, is_encrypt_init, AES_BLOCK_LEN, aes_encrypt_block,
	aes_xor_block);
	break;
	case AES_GCM_MECH_INFO_TYPE:
	if (mechanism->cm_param == NULL \|\|
	mechanism->cm_param_len != sizeof (CK_AES_GCM_PARAMS)) {
	return (CRYPTO_MECHANISM_PARAM_INVALID);
	}
	rv = gcm_init_ctx((gcm_ctx_t *)aes_ctx, mechanism->cm_param,
	AES_BLOCK_LEN, aes_encrypt_block, aes_copy_block,
	aes_xor_block);
	break;
	case AES_GMAC_MECH_INFO_TYPE:
	if (mechanism->cm_param == NULL \|\|
	mechanism->cm_param_len != sizeof (CK_AES_GMAC_PARAMS)) {
	return (CRYPTO_MECHANISM_PARAM_INVALID);
	}
	rv = gmac_init_ctx((gcm_ctx_t *)aes_ctx, mechanism->cm_param,
	AES_BLOCK_LEN, aes_encrypt_block, aes_copy_block,
	aes_xor_block);
	break;
	case AES_ECB_MECH_INFO_TYPE:
	aes_ctx->ac_flags \|= ECB_MODE;
	}

	if (rv != CRYPTO_SUCCESS) {
	if (aes_ctx->ac_flags & PROVIDER_OWNS_KEY_SCHEDULE) {
	memset(keysched, 0, size);
	kmem_free(keysched, size);
	}
	}

	return (rv);
	}

	static int
	process_gmac_mech(crypto_mechanism_t mech, crypto_data_t data,
	CK_AES_GCM_PARAMS *gcm_params)
	{
	/* LINTED: pointer alignment */
	CK_AES_GMAC_PARAMS params = (CK_AES_GMAC_PARAMS )mech->cm_param;

	if (mech->cm_type != AES_GMAC_MECH_INFO_TYPE)
	return (CRYPTO_MECHANISM_INVALID);

	if (mech->cm_param_len != sizeof (CK_AES_GMAC_PARAMS))
	return (CRYPTO_MECHANISM_PARAM_INVALID);

	if (params->pIv == NULL)
	return (CRYPTO_MECHANISM_PARAM_INVALID);

	gcm_params->pIv = params->pIv;
	gcm_params->ulIvLen = AES_GMAC_IV_LEN;
	gcm_params->ulTagBits = AES_GMAC_TAG_BITS;

	if (data == NULL)
	return (CRYPTO_SUCCESS);

	if (data->cd_format != CRYPTO_DATA_RAW)
	return (CRYPTO_ARGUMENTS_BAD);

	gcm_params->pAAD = (uchar_t *)data->cd_raw.iov_base;
	gcm_params->ulAADLen = data->cd_length;
	return (CRYPTO_SUCCESS);
	}

	static int
	aes_mac_atomic(crypto_mechanism_t *mechanism,
	crypto_key_t key, crypto_data_t data, crypto_data_t *mac,
	crypto_spi_ctx_template_t template)
	{
	CK_AES_GCM_PARAMS gcm_params;
	crypto_mechanism_t gcm_mech;
	int rv;

	if ((rv = process_gmac_mech(mechanism, data, &gcm_params))
	!= CRYPTO_SUCCESS)
	return (rv);

	gcm_mech.cm_type = AES_GCM_MECH_INFO_TYPE;
	gcm_mech.cm_param_len = sizeof (CK_AES_GCM_PARAMS);
	gcm_mech.cm_param = (char *)&gcm_params;

	return (aes_encrypt_atomic(&gcm_mech,
	key, &null_crypto_data, mac, template));
	}

	static int
	aes_mac_verify_atomic(crypto_mechanism_t mechanism, crypto_key_t key,
	crypto_data_t data, crypto_data_t mac, crypto_spi_ctx_template_t template)
	{
	CK_AES_GCM_PARAMS gcm_params;
	crypto_mechanism_t gcm_mech;
	int rv;

	if ((rv = process_gmac_mech(mechanism, data, &gcm_params))
	!= CRYPTO_SUCCESS)
	return (rv);

	gcm_mech.cm_type = AES_GCM_MECH_INFO_TYPE;
	gcm_mech.cm_param_len = sizeof (CK_AES_GCM_PARAMS);
	gcm_mech.cm_param = (char *)&gcm_params;

	return (aes_decrypt_atomic(&gcm_mech,
	key, mac, &null_crypto_data, template));
	}
	diff --git a/sys/contrib/openzfs/module/os/freebsd/spl/spl_misc.c b/sys/contrib/openzfs/module/os/freebsd/spl/spl_misc.c
	index a5fc996b6550..2d0821417ad9 100644
	--- a/sys/contrib/openzfs/module/os/freebsd/spl/spl_misc.c
	+++ b/sys/contrib/openzfs/module/os/freebsd/spl/spl_misc.c
	@@ -1,106 +1,105 @@
	/*
	* Copyright (c) 2007 Pawel Jakub Dawidek <pjd@FreeBSD.org>
	* All rights reserved.
	*
	* Redistribution and use in source and binary forms, with or without
	* modification, are permitted provided that the following conditions
	* are met:
	* 1. Redistributions of source code must retain the above copyright
	* notice, this list of conditions and the following disclaimer.
	* 2. Redistributions in binary form must reproduce the above copyright
	* notice, this list of conditions and the following disclaimer in the
	* documentation and/or other materials provided with the distribution.
	*
	* THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
	* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
	* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
	* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
	* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
	* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
	* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
	* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
	* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
	* SUCH DAMAGE.
	*/

	#include <sys/types.h>
	#include <sys/param.h>
	#include <sys/jail.h>
	#include <sys/kernel.h>
	#include <sys/libkern.h>
	#include <sys/limits.h>
	#include <sys/misc.h>
	#include <sys/sysctl.h>
	#include <sys/vnode.h>

	#include <sys/zfs_context.h>

	static struct opensolaris_utsname hw_utsname = {
	+ .sysname = ostype,
	+ .nodename = prison0.pr_hostname,
	+ .release = osrelease,
	.machine = MACHINE
	};

	utsname_t *
	utsname(void)
	{
	return (&hw_utsname);
	}

	static void
	opensolaris_utsname_init(void *arg)
	{
	-
	- hw_utsname.sysname = ostype;
	- hw_utsname.nodename = prison0.pr_hostname;
	- hw_utsname.release = osrelease;
	snprintf(hw_utsname.version, sizeof (hw_utsname.version),
	"%d", osreldate);
	}

	char *
	kmem_strdup(const char *s)
	{
	char *buf;

	buf = kmem_alloc(strlen(s) + 1, KM_SLEEP);
	strcpy(buf, s);
	return (buf);
	}

	int
	ddi_copyin(const void from, void to, size_t len, int flags)
	{
	/* Fake ioctl() issued by kernel, 'from' is a kernel address */
	if (flags & FKIOCTL) {
	memcpy(to, from, len);
	return (0);
	}

	return (copyin(from, to, len));
	}

	int
	ddi_copyout(const void from, void to, size_t len, int flags)
	{
	/* Fake ioctl() issued by kernel, 'from' is a kernel address */
	if (flags & FKIOCTL) {
	memcpy(to, from, len);
	return (0);
	}

	return (copyout(from, to, len));
	}

	void
	spl_panic(const char file, const char func, int line, const char *fmt, ...)
	{
	va_list ap;

	va_start(ap, fmt);
	vpanic(fmt, ap);
	va_end(ap);
	}


	SYSINIT(opensolaris_utsname_init, SI_SUB_TUNABLES, SI_ORDER_ANY,
	opensolaris_utsname_init, NULL);
	diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/vdev_geom.c b/sys/contrib/openzfs/module/os/freebsd/zfs/vdev_geom.c
	index 9d88971919db..38c1d8e9e464 100644
	--- a/sys/contrib/openzfs/module/os/freebsd/zfs/vdev_geom.c
	+++ b/sys/contrib/openzfs/module/os/freebsd/zfs/vdev_geom.c
	@@ -1,1315 +1,1315 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2006 Pawel Jakub Dawidek <pjd@FreeBSD.org>
	* All rights reserved.
	*
	* Portions Copyright (c) 2012 Martin Matuska <mm@FreeBSD.org>
	*/

	#include <sys/zfs_context.h>
	#include <sys/param.h>
	#include <sys/kernel.h>
	#include <sys/bio.h>
	#include <sys/buf.h>
	#include <sys/file.h>
	#include <sys/spa.h>
	#include <sys/spa_impl.h>
	#include <sys/vdev_impl.h>
	#include <sys/vdev_os.h>
	#include <sys/fs/zfs.h>
	#include <sys/zio.h>
	#include <vm/vm_page.h>
	#include <geom/geom.h>
	#include <geom/geom_disk.h>
	#include <geom/geom_int.h>

	#ifndef g_topology_locked
	#define g_topology_locked() sx_xlocked(&topology_lock)
	#endif

	/*
	* Virtual device vector for GEOM.
	*/

	static g_attrchanged_t vdev_geom_attrchanged;
	struct g_class zfs_vdev_class = {
	.name = "ZFS::VDEV",
	.version = G_VERSION,
	.attrchanged = vdev_geom_attrchanged,
	};

	struct consumer_vdev_elem {
	SLIST_ENTRY(consumer_vdev_elem) elems;
	vdev_t *vd;
	};

	SLIST_HEAD(consumer_priv_t, consumer_vdev_elem);
	_Static_assert(
	sizeof (((struct g_consumer *)NULL)->private) ==
	sizeof (struct consumer_priv_t *),
	"consumer_priv_t* can't be stored in g_consumer.private");

	DECLARE_GEOM_CLASS(zfs_vdev_class, zfs_vdev);

	SYSCTL_DECL(_vfs_zfs_vdev);
	/* Don't send BIO_FLUSH. */
	static int vdev_geom_bio_flush_disable;
	SYSCTL_INT(_vfs_zfs_vdev, OID_AUTO, bio_flush_disable, CTLFLAG_RWTUN,
	&vdev_geom_bio_flush_disable, 0, "Disable BIO_FLUSH");
	/* Don't send BIO_DELETE. */
	static int vdev_geom_bio_delete_disable;
	SYSCTL_INT(_vfs_zfs_vdev, OID_AUTO, bio_delete_disable, CTLFLAG_RWTUN,
	&vdev_geom_bio_delete_disable, 0, "Disable BIO_DELETE");

	/* Declare local functions */
	static void vdev_geom_detach(struct g_consumer *cp, boolean_t open_for_read);

	/*
	* Thread local storage used to indicate when a thread is probing geoms
	* for their guids. If NULL, this thread is not tasting geoms. If non NULL,
	* it is looking for a replacement for the vdev_t* that is its value.
	*/
	uint_t zfs_geom_probe_vdev_key;

	static void
	vdev_geom_set_physpath(vdev_t vd, struct g_consumer cp,
	boolean_t do_null_update)
	{
	boolean_t needs_update = B_FALSE;
	char *physpath;
	int error, physpath_len;

	physpath_len = MAXPATHLEN;
	physpath = g_malloc(physpath_len, M_WAITOK\|M_ZERO);
	error = g_io_getattr("GEOM::physpath", cp, &physpath_len, physpath);
	if (error == 0) {
	char *old_physpath;

	/* g_topology lock ensures that vdev has not been closed */
	g_topology_assert();
	old_physpath = vd->vdev_physpath;
	vd->vdev_physpath = spa_strdup(physpath);

	if (old_physpath != NULL) {
	needs_update = (strcmp(old_physpath,
	vd->vdev_physpath) != 0);
	spa_strfree(old_physpath);
	} else
	needs_update = do_null_update;
	}
	g_free(physpath);

	/*
	* If the physical path changed, update the config.
	* Only request an update for previously unset physpaths if
	* requested by the caller.
	*/
	if (needs_update)
	spa_async_request(vd->vdev_spa, SPA_ASYNC_CONFIG_UPDATE);

	}

	static void
	vdev_geom_attrchanged(struct g_consumer cp, const char attr)
	{
	struct consumer_priv_t *priv;
	struct consumer_vdev_elem *elem;

	priv = (struct consumer_priv_t *)&cp->private;
	if (SLIST_EMPTY(priv))
	return;

	SLIST_FOREACH(elem, priv, elems) {
	vdev_t *vd = elem->vd;
	if (strcmp(attr, "GEOM::physpath") == 0) {
	vdev_geom_set_physpath(vd, cp, /* null_update */B_TRUE);
	return;
	}
	}
	}

	static void
	vdev_geom_resize(struct g_consumer *cp)
	{
	struct consumer_priv_t *priv;
	struct consumer_vdev_elem *elem;
	spa_t *spa;
	vdev_t *vd;

	priv = (struct consumer_priv_t *)&cp->private;
	if (SLIST_EMPTY(priv))
	return;

	SLIST_FOREACH(elem, priv, elems) {
	vd = elem->vd;
	if (vd->vdev_state != VDEV_STATE_HEALTHY)
	continue;
	spa = vd->vdev_spa;
	if (!spa->spa_autoexpand)
	continue;
	vdev_online(spa, vd->vdev_guid, ZFS_ONLINE_EXPAND, NULL);
	}
	}

	static void
	vdev_geom_orphan(struct g_consumer *cp)
	{
	struct consumer_priv_t *priv;
	// cppcheck-suppress uninitvar
	struct consumer_vdev_elem *elem;

	g_topology_assert();

	priv = (struct consumer_priv_t *)&cp->private;
	if (SLIST_EMPTY(priv))
	/* Vdev close in progress. Ignore the event. */
	return;

	/*
	* Orphan callbacks occur from the GEOM event thread.
	* Concurrent with this call, new I/O requests may be
	* working their way through GEOM about to find out
	* (only once executed by the g_down thread) that we've
	* been orphaned from our disk provider. These I/Os
	* must be retired before we can detach our consumer.
	* This is most easily achieved by acquiring the
	* SPA ZIO configuration lock as a writer, but doing
	* so with the GEOM topology lock held would cause
	* a lock order reversal. Instead, rely on the SPA's
	* async removal support to invoke a close on this
	* vdev once it is safe to do so.
	*/
	SLIST_FOREACH(elem, priv, elems) {
	// cppcheck-suppress uninitvar
	vdev_t *vd = elem->vd;

	vd->vdev_remove_wanted = B_TRUE;
	spa_async_request(vd->vdev_spa, SPA_ASYNC_REMOVE);
	}
	}

	static struct g_consumer *
	vdev_geom_attach(struct g_provider pp, vdev_t vd, boolean_t sanity)
	{
	struct g_geom *gp;
	struct g_consumer *cp;
	int error;

	g_topology_assert();

	ZFS_LOG(1, "Attaching to %s.", pp->name);

	if (sanity) {
	if (pp->sectorsize > VDEV_PAD_SIZE \|\| !ISP2(pp->sectorsize)) {
	ZFS_LOG(1, "Failing attach of %s. "
	"Incompatible sectorsize %d\n",
	pp->name, pp->sectorsize);
	return (NULL);
	} else if (pp->mediasize < SPA_MINDEVSIZE) {
	ZFS_LOG(1, "Failing attach of %s. "
	"Incompatible mediasize %ju\n",
	pp->name, pp->mediasize);
	return (NULL);
	}
	}

	/* Do we have geom already? No? Create one. */
	LIST_FOREACH(gp, &zfs_vdev_class.geom, geom) {
	if (gp->flags & G_GEOM_WITHER)
	continue;
	if (strcmp(gp->name, "zfs::vdev") != 0)
	continue;
	break;
	}
	if (gp == NULL) {
	gp = g_new_geomf(&zfs_vdev_class, "zfs::vdev");
	gp->orphan = vdev_geom_orphan;
	gp->attrchanged = vdev_geom_attrchanged;
	gp->resize = vdev_geom_resize;
	cp = g_new_consumer(gp);
	error = g_attach(cp, pp);
	if (error != 0) {
	ZFS_LOG(1, "%s(%d): g_attach failed: %d\n", __func__,
	__LINE__, error);
	vdev_geom_detach(cp, B_FALSE);
	return (NULL);
	}
	error = g_access(cp, 1, 0, 1);
	if (error != 0) {
	ZFS_LOG(1, "%s(%d): g_access failed: %d\n", __func__,
	__LINE__, error);
	vdev_geom_detach(cp, B_FALSE);
	return (NULL);
	}
	ZFS_LOG(1, "Created geom and consumer for %s.", pp->name);
	} else {
	/* Check if we are already connected to this provider. */
	LIST_FOREACH(cp, &gp->consumer, consumer) {
	if (cp->provider == pp) {
	ZFS_LOG(1, "Found consumer for %s.", pp->name);
	break;
	}
	}
	if (cp == NULL) {
	cp = g_new_consumer(gp);
	error = g_attach(cp, pp);
	if (error != 0) {
	ZFS_LOG(1, "%s(%d): g_attach failed: %d\n",
	__func__, __LINE__, error);
	vdev_geom_detach(cp, B_FALSE);
	return (NULL);
	}
	error = g_access(cp, 1, 0, 1);
	if (error != 0) {
	ZFS_LOG(1, "%s(%d): g_access failed: %d\n",
	__func__, __LINE__, error);
	vdev_geom_detach(cp, B_FALSE);
	return (NULL);
	}
	ZFS_LOG(1, "Created consumer for %s.", pp->name);
	} else {
	error = g_access(cp, 1, 0, 1);
	if (error != 0) {
	ZFS_LOG(1, "%s(%d): g_access failed: %d\n",
	__func__, __LINE__, error);
	return (NULL);
	}
	ZFS_LOG(1, "Used existing consumer for %s.", pp->name);
	}
	}

	if (vd != NULL)
	vd->vdev_tsd = cp;

	cp->flags \|= G_CF_DIRECT_SEND \| G_CF_DIRECT_RECEIVE;
	return (cp);
	}

	static void
	vdev_geom_detach(struct g_consumer *cp, boolean_t open_for_read)
	{
	struct g_geom *gp;

	g_topology_assert();

	ZFS_LOG(1, "Detaching from %s.",
	cp->provider && cp->provider->name ? cp->provider->name : "NULL");

	gp = cp->geom;
	if (open_for_read)
	g_access(cp, -1, 0, -1);
	/* Destroy consumer on last close. */
	if (cp->acr == 0 && cp->ace == 0) {
	if (cp->acw > 0)
	g_access(cp, 0, -cp->acw, 0);
	if (cp->provider != NULL) {
	ZFS_LOG(1, "Destroying consumer for %s.",
	cp->provider->name ? cp->provider->name : "NULL");
	g_detach(cp);
	}
	g_destroy_consumer(cp);
	}
	/* Destroy geom if there are no consumers left. */
	if (LIST_EMPTY(&gp->consumer)) {
	ZFS_LOG(1, "Destroyed geom %s.", gp->name);
	g_wither_geom(gp, ENXIO);
	}
	}

	static void
	vdev_geom_close_locked(vdev_t *vd)
	{
	struct g_consumer *cp;
	struct consumer_priv_t *priv;
	struct consumer_vdev_elem elem, elem_temp;

	g_topology_assert();

	cp = vd->vdev_tsd;
	vd->vdev_delayed_close = B_FALSE;
	if (cp == NULL)
	return;

	ZFS_LOG(1, "Closing access to %s.", cp->provider->name);
	KASSERT(cp->private != NULL, ("%s: cp->private is NULL", __func__));
	priv = (struct consumer_priv_t *)&cp->private;
	vd->vdev_tsd = NULL;
	SLIST_FOREACH_SAFE(elem, priv, elems, elem_temp) {
	if (elem->vd == vd) {
	SLIST_REMOVE(priv, elem, consumer_vdev_elem, elems);
	g_free(elem);
	}
	}

	vdev_geom_detach(cp, B_TRUE);
	}

	/*
	* Issue one or more bios to the vdev in parallel
	* cmds, datas, offsets, errors, and sizes are arrays of length ncmds. Each IO
	* operation is described by parallel entries from each array. There may be
	* more bios actually issued than entries in the array
	*/
	static void
	vdev_geom_io(struct g_consumer cp, int cmds, void *datas, off_t offsets,
	off_t sizes, int errors, int ncmds)
	{
	struct bio **bios;
	uint8_t *p;
	off_t off, maxio, s, end;
	int i, n_bios, j;
	size_t bios_size;

	#if __FreeBSD_version > 1300130
	maxio = maxphys - (maxphys % cp->provider->sectorsize);
	#else
	maxio = MAXPHYS - (MAXPHYS % cp->provider->sectorsize);
	#endif
	n_bios = 0;

	/* How many bios are required for all commands ? */
	for (i = 0; i < ncmds; i++)
	n_bios += (sizes[i] + maxio - 1) / maxio;

	/* Allocate memory for the bios */
	bios_size = n_bios * sizeof (struct bio *);
	bios = kmem_zalloc(bios_size, KM_SLEEP);

	/* Prepare and issue all of the bios */
	for (i = j = 0; i < ncmds; i++) {
	off = offsets[i];
	p = datas[i];
	s = sizes[i];
	end = off + s;
	ASSERT0(off % cp->provider->sectorsize);
	ASSERT0(s % cp->provider->sectorsize);

	for (; off < end; off += maxio, p += maxio, s -= maxio, j++) {
	bios[j] = g_alloc_bio();
	bios[j]->bio_cmd = cmds[i];
	bios[j]->bio_done = NULL;
	bios[j]->bio_offset = off;
	bios[j]->bio_length = MIN(s, maxio);
	bios[j]->bio_data = (caddr_t)p;
	g_io_request(bios[j], cp);
	}
	}
	ASSERT3S(j, ==, n_bios);

	/* Wait for all of the bios to complete, and clean them up */
	for (i = j = 0; i < ncmds; i++) {
	off = offsets[i];
	s = sizes[i];
	end = off + s;

	for (; off < end; off += maxio, s -= maxio, j++) {
	errors[i] = biowait(bios[j], "vdev_geom_io") \|\|
	errors[i];
	g_destroy_bio(bios[j]);
	}
	}
	kmem_free(bios, bios_size);
	}

	/*
	* Read the vdev config from a device. Return the number of valid labels that
	* were found. The vdev config will be returned in config if and only if at
	* least one valid label was found.
	*/
	static int
	vdev_geom_read_config(struct g_consumer cp, nvlist_t *configp)
	{
	struct g_provider *pp;
	nvlist_t *config;
	vdev_phys_t *vdev_lists[VDEV_LABELS];
	char *buf;
	size_t buflen;
	uint64_t psize, state, txg;
	off_t offsets[VDEV_LABELS];
	off_t size;
	off_t sizes[VDEV_LABELS];
	int cmds[VDEV_LABELS];
	int errors[VDEV_LABELS];
	int l, nlabels;

	g_topology_assert_not();

	pp = cp->provider;
	ZFS_LOG(1, "Reading config from %s...", pp->name);

	psize = pp->mediasize;
	- psize = P2ALIGN(psize, (uint64_t)sizeof (vdev_label_t));
	+ psize = P2ALIGN_TYPED(psize, sizeof (vdev_label_t), uint64_t);

	size = sizeof (*vdev_lists[0]) + pp->sectorsize -
	((sizeof (*vdev_lists[0]) - 1) % pp->sectorsize) - 1;

	buflen = sizeof (vdev_lists[0]->vp_nvlist);

	/* Create all of the IO requests */
	for (l = 0; l < VDEV_LABELS; l++) {
	cmds[l] = BIO_READ;
	vdev_lists[l] = kmem_alloc(size, KM_SLEEP);
	offsets[l] = vdev_label_offset(psize, l, 0) + VDEV_SKIP_SIZE;
	sizes[l] = size;
	errors[l] = 0;
	ASSERT0(offsets[l] % pp->sectorsize);
	}

	/* Issue the IO requests */
	vdev_geom_io(cp, cmds, (void**)vdev_lists, offsets, sizes, errors,
	VDEV_LABELS);

	/* Parse the labels */
	config = *configp = NULL;
	nlabels = 0;
	for (l = 0; l < VDEV_LABELS; l++) {
	if (errors[l] != 0)
	continue;

	buf = vdev_lists[l]->vp_nvlist;

	if (nvlist_unpack(buf, buflen, &config, 0) != 0)
	continue;

	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
	&state) != 0 \|\| state > POOL_STATE_L2CACHE) {
	nvlist_free(config);
	continue;
	}

	if (state != POOL_STATE_SPARE &&
	state != POOL_STATE_L2CACHE &&
	(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG,
	&txg) != 0 \|\| txg == 0)) {
	nvlist_free(config);
	continue;
	}

	if (*configp != NULL)
	nvlist_free(*configp);
	*configp = config;
	nlabels++;
	}

	/* Free the label storage */
	for (l = 0; l < VDEV_LABELS; l++)
	kmem_free(vdev_lists[l], size);

	return (nlabels);
	}

	static void
	resize_configs(nvlist_t **configs, uint64_t count, uint64_t id)
	{
	nvlist_t **new_configs;
	uint64_t i;

	if (id < *count)
	return;
	new_configs = kmem_zalloc((id + 1) * sizeof (nvlist_t *),
	KM_SLEEP);
	for (i = 0; i < *count; i++)
	new_configs[i] = (*configs)[i];
	if (*configs != NULL)
	kmem_free(configs, count * sizeof (void *));
	*configs = new_configs;
	*count = id + 1;
	}

	static void
	process_vdev_config(nvlist_t **configs, uint64_t count, nvlist_t *cfg,
	const char name, uint64_t known_pool_guid)
	{
	nvlist_t *vdev_tree;
	uint64_t pool_guid;
	uint64_t vdev_guid;
	uint64_t id, txg, known_txg;
	const char *pname;

	if (nvlist_lookup_string(cfg, ZPOOL_CONFIG_POOL_NAME, &pname) != 0 \|\|
	strcmp(pname, name) != 0)
	goto ignore;

	if (nvlist_lookup_uint64(cfg, ZPOOL_CONFIG_POOL_GUID, &pool_guid) != 0)
	goto ignore;

	if (nvlist_lookup_uint64(cfg, ZPOOL_CONFIG_TOP_GUID, &vdev_guid) != 0)
	goto ignore;

	if (nvlist_lookup_nvlist(cfg, ZPOOL_CONFIG_VDEV_TREE, &vdev_tree) != 0)
	goto ignore;

	if (nvlist_lookup_uint64(vdev_tree, ZPOOL_CONFIG_ID, &id) != 0)
	goto ignore;

	txg = fnvlist_lookup_uint64(cfg, ZPOOL_CONFIG_POOL_TXG);

	if (*known_pool_guid != 0) {
	if (pool_guid != *known_pool_guid)
	goto ignore;
	} else
	*known_pool_guid = pool_guid;

	resize_configs(configs, count, id);

	if ((*configs)[id] != NULL) {
	known_txg = fnvlist_lookup_uint64((*configs)[id],
	ZPOOL_CONFIG_POOL_TXG);
	if (txg <= known_txg)
	goto ignore;
	nvlist_free((*configs)[id]);
	}

	(*configs)[id] = cfg;
	return;

	ignore:
	nvlist_free(cfg);
	}

	int
	vdev_geom_read_pool_label(const char *name,
	nvlist_t **configs, uint64_t count)
	{
	struct g_class *mp;
	struct g_geom *gp;
	struct g_provider *pp;
	struct g_consumer *zcp;
	nvlist_t *vdev_cfg;
	uint64_t pool_guid;
	int nlabels;

	DROP_GIANT();
	g_topology_lock();

	*configs = NULL;
	*count = 0;
	pool_guid = 0;
	LIST_FOREACH(mp, &g_classes, class) {
	if (mp == &zfs_vdev_class)
	continue;
	LIST_FOREACH(gp, &mp->geom, geom) {
	if (gp->flags & G_GEOM_WITHER)
	continue;
	LIST_FOREACH(pp, &gp->provider, provider) {
	if (pp->flags & G_PF_WITHER)
	continue;
	zcp = vdev_geom_attach(pp, NULL, B_TRUE);
	if (zcp == NULL)
	continue;
	g_topology_unlock();
	nlabels = vdev_geom_read_config(zcp, &vdev_cfg);
	g_topology_lock();
	vdev_geom_detach(zcp, B_TRUE);
	if (nlabels == 0)
	continue;
	ZFS_LOG(1, "successfully read vdev config");

	process_vdev_config(configs, count,
	vdev_cfg, name, &pool_guid);
	}
	}
	}
	g_topology_unlock();
	PICKUP_GIANT();

	return (*count > 0 ? 0 : ENOENT);
	}

	enum match {
	NO_MATCH = 0, /* No matching labels found */
	TOPGUID_MATCH = 1, /* Labels match top guid, not vdev guid */
	ZERO_MATCH = 1, /* Should never be returned */
	ONE_MATCH = 2, /* 1 label matching the vdev_guid */
	TWO_MATCH = 3, /* 2 label matching the vdev_guid */
	THREE_MATCH = 4, /* 3 label matching the vdev_guid */
	FULL_MATCH = 5 /* all labels match the vdev_guid */
	};

	static enum match
	vdev_attach_ok(vdev_t vd, struct g_provider pp)
	{
	nvlist_t *config;
	uint64_t pool_guid, top_guid, vdev_guid;
	struct g_consumer *cp;
	int nlabels;

	cp = vdev_geom_attach(pp, NULL, B_TRUE);
	if (cp == NULL) {
	ZFS_LOG(1, "Unable to attach tasting instance to %s.",
	pp->name);
	return (NO_MATCH);
	}
	g_topology_unlock();
	nlabels = vdev_geom_read_config(cp, &config);
	g_topology_lock();
	vdev_geom_detach(cp, B_TRUE);
	if (nlabels == 0) {
	ZFS_LOG(1, "Unable to read config from %s.", pp->name);
	return (NO_MATCH);
	}

	pool_guid = 0;
	(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &pool_guid);
	top_guid = 0;
	(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_TOP_GUID, &top_guid);
	vdev_guid = 0;
	(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, &vdev_guid);
	nvlist_free(config);

	/*
	* Check that the label's pool guid matches the desired guid.
	* Inactive spares and L2ARCs do not have any pool guid in the label.
	*/
	if (pool_guid != 0 && pool_guid != spa_guid(vd->vdev_spa)) {
	ZFS_LOG(1, "pool guid mismatch for provider %s: %ju != %ju.",
	pp->name,
	(uintmax_t)spa_guid(vd->vdev_spa), (uintmax_t)pool_guid);
	return (NO_MATCH);
	}

	/*
	* Check that the label's vdev guid matches the desired guid.
	* The second condition handles possible race on vdev detach, when
	* remaining vdev receives GUID of destroyed top level mirror vdev.
	*/
	if (vdev_guid == vd->vdev_guid) {
	ZFS_LOG(1, "guids match for provider %s.", pp->name);
	return (ZERO_MATCH + nlabels);
	} else if (top_guid == vd->vdev_guid && vd == vd->vdev_top) {
	ZFS_LOG(1, "top vdev guid match for provider %s.", pp->name);
	return (TOPGUID_MATCH);
	}
	ZFS_LOG(1, "vdev guid mismatch for provider %s: %ju != %ju.",
	pp->name, (uintmax_t)vd->vdev_guid, (uintmax_t)vdev_guid);
	return (NO_MATCH);
	}

	static struct g_consumer *
	vdev_geom_attach_by_guids(vdev_t *vd)
	{
	struct g_class *mp;
	struct g_geom *gp;
	struct g_provider pp, best_pp;
	struct g_consumer *cp;
	const char *vdpath;
	enum match match, best_match;

	g_topology_assert();

	vdpath = vd->vdev_path + sizeof ("/dev/") - 1;
	cp = NULL;
	best_pp = NULL;
	best_match = NO_MATCH;
	LIST_FOREACH(mp, &g_classes, class) {
	if (mp == &zfs_vdev_class)
	continue;
	LIST_FOREACH(gp, &mp->geom, geom) {
	if (gp->flags & G_GEOM_WITHER)
	continue;
	LIST_FOREACH(pp, &gp->provider, provider) {
	match = vdev_attach_ok(vd, pp);
	if (match > best_match) {
	best_match = match;
	best_pp = pp;
	} else if (match == best_match) {
	if (strcmp(pp->name, vdpath) == 0) {
	best_pp = pp;
	}
	}
	if (match == FULL_MATCH)
	goto out;
	}
	}
	}

	out:
	if (best_pp) {
	cp = vdev_geom_attach(best_pp, vd, B_TRUE);
	if (cp == NULL) {
	printf("ZFS WARNING: Unable to attach to %s.\n",
	best_pp->name);
	}
	}
	return (cp);
	}

	static struct g_consumer *
	vdev_geom_open_by_guids(vdev_t *vd)
	{
	struct g_consumer *cp;
	char *buf;
	size_t len;

	g_topology_assert();

	ZFS_LOG(1, "Searching by guids [%ju:%ju].",
	(uintmax_t)spa_guid(vd->vdev_spa), (uintmax_t)vd->vdev_guid);
	cp = vdev_geom_attach_by_guids(vd);
	if (cp != NULL) {
	len = strlen(cp->provider->name) + strlen("/dev/") + 1;
	buf = kmem_alloc(len, KM_SLEEP);

	snprintf(buf, len, "/dev/%s", cp->provider->name);
	spa_strfree(vd->vdev_path);
	vd->vdev_path = buf;

	ZFS_LOG(1, "Attach by guid [%ju:%ju] succeeded, provider %s.",
	(uintmax_t)spa_guid(vd->vdev_spa),
	(uintmax_t)vd->vdev_guid, cp->provider->name);
	} else {
	ZFS_LOG(1, "Search by guid [%ju:%ju] failed.",
	(uintmax_t)spa_guid(vd->vdev_spa),
	(uintmax_t)vd->vdev_guid);
	}

	return (cp);
	}

	static struct g_consumer *
	vdev_geom_open_by_path(vdev_t *vd, int check_guid)
	{
	struct g_provider *pp;
	struct g_consumer *cp;

	g_topology_assert();

	cp = NULL;
	pp = g_provider_by_name(vd->vdev_path + sizeof ("/dev/") - 1);
	if (pp != NULL) {
	ZFS_LOG(1, "Found provider by name %s.", vd->vdev_path);
	if (!check_guid \|\| vdev_attach_ok(vd, pp) == FULL_MATCH)
	cp = vdev_geom_attach(pp, vd, B_FALSE);
	}

	return (cp);
	}

	static int
	vdev_geom_open(vdev_t vd, uint64_t psize, uint64_t *max_psize,
	uint64_t logical_ashift, uint64_t physical_ashift)
	{
	struct g_provider *pp;
	struct g_consumer *cp;
	int error, has_trim;
	uint16_t rate;

	/*
	* Set the TLS to indicate downstack that we
	* should not access zvols
	*/
	VERIFY0(tsd_set(zfs_geom_probe_vdev_key, vd));

	/*
	* We must have a pathname, and it must be absolute.
	*/
	if (vd->vdev_path == NULL \|\| strncmp(vd->vdev_path, "/dev/", 5) != 0) {
	vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
	return (EINVAL);
	}

	/*
	* Reopen the device if it's not currently open. Otherwise,
	* just update the physical size of the device.
	*/
	if ((cp = vd->vdev_tsd) != NULL) {
	ASSERT(vd->vdev_reopening);
	goto skip_open;
	}

	DROP_GIANT();
	g_topology_lock();
	error = 0;

	if (vd->vdev_spa->spa_is_splitting \|\|
	((vd->vdev_prevstate == VDEV_STATE_UNKNOWN &&
	(vd->vdev_spa->spa_load_state == SPA_LOAD_NONE \|\|
	vd->vdev_spa->spa_load_state == SPA_LOAD_CREATE)))) {
	/*
	* We are dealing with a vdev that hasn't been previously
	* opened (since boot), and we are not loading an
	* existing pool configuration. This looks like a
	* vdev add operation to a new or existing pool.
	* Assume the user really wants to do this, and find
	* GEOM provider by its name, ignoring GUID mismatches.
	*
	* XXPOLICY: It would be safer to only allow a device
	* that is unlabeled or labeled but missing
	* GUID information to be opened in this fashion,
	* unless we are doing a split, in which case we
	* should allow any guid.
	*/
	cp = vdev_geom_open_by_path(vd, 0);
	} else {
	/*
	* Try using the recorded path for this device, but only
	* accept it if its label data contains the expected GUIDs.
	*/
	cp = vdev_geom_open_by_path(vd, 1);
	if (cp == NULL) {
	/*
	* The device at vd->vdev_path doesn't have the
	* expected GUIDs. The disks might have merely
	* moved around so try all other GEOM providers
	* to find one with the right GUIDs.
	*/
	cp = vdev_geom_open_by_guids(vd);
	}
	}

	/* Clear the TLS now that tasting is done */
	VERIFY0(tsd_set(zfs_geom_probe_vdev_key, NULL));

	if (cp == NULL) {
	ZFS_LOG(1, "Vdev %s not found.", vd->vdev_path);
	error = ENOENT;
	} else {
	struct consumer_priv_t *priv;
	struct consumer_vdev_elem *elem;
	int spamode;

	priv = (struct consumer_priv_t *)&cp->private;
	if (cp->private == NULL)
	SLIST_INIT(priv);
	elem = g_malloc(sizeof (*elem), M_WAITOK\|M_ZERO);
	elem->vd = vd;
	SLIST_INSERT_HEAD(priv, elem, elems);

	spamode = spa_mode(vd->vdev_spa);
	if (cp->provider->sectorsize > VDEV_PAD_SIZE \|\|
	!ISP2(cp->provider->sectorsize)) {
	ZFS_LOG(1, "Provider %s has unsupported sectorsize.",
	cp->provider->name);

	vdev_geom_close_locked(vd);
	error = EINVAL;
	cp = NULL;
	} else if (cp->acw == 0 && (spamode & FWRITE) != 0) {
	int i;

	for (i = 0; i < 5; i++) {
	error = g_access(cp, 0, 1, 0);
	if (error == 0)
	break;
	g_topology_unlock();
	tsleep(vd, 0, "vdev", hz / 2);
	g_topology_lock();
	}
	if (error != 0) {
	printf("ZFS WARNING: Unable to open %s for "
	"writing (error=%d).\n",
	cp->provider->name, error);
	vdev_geom_close_locked(vd);
	cp = NULL;
	}
	}
	}

	/* Fetch initial physical path information for this device. */
	if (cp != NULL) {
	vdev_geom_attrchanged(cp, "GEOM::physpath");

	/* Set other GEOM characteristics */
	vdev_geom_set_physpath(vd, cp, /* do_null_update */B_FALSE);
	}

	g_topology_unlock();
	PICKUP_GIANT();
	if (cp == NULL) {
	vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
	vdev_dbgmsg(vd, "vdev_geom_open: failed to open [error=%d]",
	error);
	return (error);
	}
	skip_open:
	pp = cp->provider;

	/*
	* Determine the actual size of the device.
	*/
	max_psize = psize = pp->mediasize;

	/*
	* Determine the device's minimum transfer size and preferred
	* transfer size.
	*/
	*logical_ashift = highbit(MAX(pp->sectorsize, SPA_MINBLOCKSIZE)) - 1;
	*physical_ashift = 0;
	if (pp->stripesize && pp->stripesize > (1 << *logical_ashift) &&
	ISP2(pp->stripesize) && pp->stripeoffset == 0)
	*physical_ashift = highbit(pp->stripesize) - 1;

	/*
	* Clear the nowritecache settings, so that on a vdev_reopen()
	* we will try again.
	*/
	vd->vdev_nowritecache = B_FALSE;

	/* Inform the ZIO pipeline that we are non-rotational. */
	error = g_getattr("GEOM::rotation_rate", cp, &rate);
	if (error == 0 && rate == DISK_RR_NON_ROTATING)
	vd->vdev_nonrot = B_TRUE;
	else
	vd->vdev_nonrot = B_FALSE;

	/* Set when device reports it supports TRIM. */
	error = g_getattr("GEOM::candelete", cp, &has_trim);
	vd->vdev_has_trim = (error == 0 && has_trim);

	/* Set when device reports it supports secure TRIM. */
	/* unavailable on FreeBSD */
	vd->vdev_has_securetrim = B_FALSE;

	return (0);
	}

	static void
	vdev_geom_close(vdev_t *vd)
	{
	struct g_consumer *cp;
	boolean_t locked;

	cp = vd->vdev_tsd;

	DROP_GIANT();
	locked = g_topology_locked();
	if (!locked)
	g_topology_lock();

	if (!vd->vdev_reopening \|\|
	(cp != NULL && ((cp->flags & G_CF_ORPHAN) != 0 \|\|
	(cp->provider != NULL && cp->provider->error != 0))))
	vdev_geom_close_locked(vd);

	if (!locked)
	g_topology_unlock();
	PICKUP_GIANT();
	}

	static void
	vdev_geom_io_intr(struct bio *bp)
	{
	vdev_t *vd;
	zio_t *zio;

	zio = bp->bio_caller1;
	vd = zio->io_vd;
	zio->io_error = bp->bio_error;
	if (zio->io_error == 0 && bp->bio_resid != 0)
	zio->io_error = SET_ERROR(EIO);

	switch (zio->io_error) {
	case ENOTSUP:
	/*
	* If we get ENOTSUP for BIO_FLUSH or BIO_DELETE we know
	* that future attempts will never succeed. In this case
	* we set a persistent flag so that we don't bother with
	* requests in the future.
	*/
	switch (bp->bio_cmd) {
	case BIO_FLUSH:
	vd->vdev_nowritecache = B_TRUE;
	break;
	case BIO_DELETE:
	break;
	}
	break;
	case ENXIO:
	if (!vd->vdev_remove_wanted) {
	/*
	* If provider's error is set we assume it is being
	* removed.
	*/
	if (bp->bio_to->error != 0) {
	vd->vdev_remove_wanted = B_TRUE;
	spa_async_request(zio->io_spa,
	SPA_ASYNC_REMOVE);
	} else if (!vd->vdev_delayed_close) {
	vd->vdev_delayed_close = B_TRUE;
	}
	}
	break;
	}

	/*
	* We have to split bio freeing into two parts, because the ABD code
	* cannot be called in this context and vdev_op_io_done is not called
	* for ZIO_TYPE_FLUSH zio-s.
	*/
	if (zio->io_type != ZIO_TYPE_READ && zio->io_type != ZIO_TYPE_WRITE) {
	g_destroy_bio(bp);
	zio->io_bio = NULL;
	}
	zio_delay_interrupt(zio);
	}

	struct vdev_geom_check_unmapped_cb_state {
	int pages;
	uint_t end;
	};

	/*
	* Callback to check the ABD segment size/alignment and count the pages.
	* GEOM requires data buffer to look virtually contiguous. It means only
	* the first page of the buffer may not start and only the last may not
	* end on a page boundary. All other physical pages must be full.
	*/
	static int
	vdev_geom_check_unmapped_cb(void buf, size_t len, void priv)
	{
	struct vdev_geom_check_unmapped_cb_state *s = priv;
	vm_offset_t off = (vm_offset_t)buf & PAGE_MASK;

	if (s->pages != 0 && off != 0)
	return (1);
	if (s->end != 0)
	return (1);
	s->end = (off + len) & PAGE_MASK;
	s->pages += (off + len + PAGE_MASK) >> PAGE_SHIFT;
	return (0);
	}

	/*
	* Check whether we can use unmapped I/O for this ZIO on this device to
	* avoid data copying between scattered and/or gang ABD buffer and linear.
	*/
	static int
	vdev_geom_check_unmapped(zio_t zio, struct g_consumer cp)
	{
	struct vdev_geom_check_unmapped_cb_state s;

	/* If unmapped I/O is administratively disabled, respect that. */
	if (!unmapped_buf_allowed)
	return (0);

	/* If the buffer is already linear, then nothing to do here. */
	if (abd_is_linear(zio->io_abd))
	return (0);

	/*
	* If unmapped I/O is not supported by the GEOM provider,
	* then we can't do anything and have to copy the data.
	*/
	if ((cp->provider->flags & G_PF_ACCEPT_UNMAPPED) == 0)
	return (0);

	/* Check the buffer chunks sizes/alignments and count pages. */
	s.pages = s.end = 0;
	if (abd_iterate_func(zio->io_abd, 0, zio->io_size,
	vdev_geom_check_unmapped_cb, &s))
	return (0);
	return (s.pages);
	}

	/*
	* Callback to translate the ABD segment into array of physical pages.
	*/
	static int
	vdev_geom_fill_unmap_cb(void buf, size_t len, void priv)
	{
	struct bio *bp = priv;
	vm_offset_t addr = (vm_offset_t)buf;
	vm_offset_t end = addr + len;

	if (bp->bio_ma_n == 0) {
	bp->bio_ma_offset = addr & PAGE_MASK;
	addr &= ~PAGE_MASK;
	} else {
	ASSERT0(P2PHASE(addr, PAGE_SIZE));
	}
	do {
	bp->bio_ma[bp->bio_ma_n++] =
	PHYS_TO_VM_PAGE(pmap_kextract(addr));
	addr += PAGE_SIZE;
	} while (addr < end);
	return (0);
	}

	static void
	vdev_geom_io_start(zio_t *zio)
	{
	vdev_t *vd;
	struct g_consumer *cp;
	struct bio *bp;

	vd = zio->io_vd;

	if (zio->io_type == ZIO_TYPE_FLUSH) {
	/* XXPOLICY */
	if (!vdev_readable(vd)) {
	zio->io_error = SET_ERROR(ENXIO);
	zio_interrupt(zio);
	return;
	}

	if (zfs_nocacheflush \|\| vdev_geom_bio_flush_disable) {
	zio_execute(zio);
	return;
	}

	if (vd->vdev_nowritecache) {
	zio->io_error = SET_ERROR(ENOTSUP);
	zio_execute(zio);
	return;
	}
	} else if (zio->io_type == ZIO_TYPE_TRIM) {
	if (vdev_geom_bio_delete_disable) {
	zio_execute(zio);
	return;
	}
	}

	ASSERT(zio->io_type == ZIO_TYPE_READ \|\|
	zio->io_type == ZIO_TYPE_WRITE \|\|
	zio->io_type == ZIO_TYPE_TRIM \|\|
	zio->io_type == ZIO_TYPE_FLUSH);

	cp = vd->vdev_tsd;
	if (cp == NULL) {
	zio->io_error = SET_ERROR(ENXIO);
	zio_interrupt(zio);
	return;
	}
	bp = g_alloc_bio();
	bp->bio_caller1 = zio;
	switch (zio->io_type) {
	case ZIO_TYPE_READ:
	case ZIO_TYPE_WRITE:
	zio->io_target_timestamp = zio_handle_io_delay(zio);
	bp->bio_offset = zio->io_offset;
	bp->bio_length = zio->io_size;
	if (zio->io_type == ZIO_TYPE_READ)
	bp->bio_cmd = BIO_READ;
	else
	bp->bio_cmd = BIO_WRITE;

	/*
	* If possible, represent scattered and/or gang ABD buffer to
	* GEOM as an array of physical pages. It allows to satisfy
	* requirement of virtually contiguous buffer without copying.
	*/
	int pgs = vdev_geom_check_unmapped(zio, cp);
	if (pgs > 0) {
	bp->bio_ma = malloc(sizeof (struct vm_page ) pgs,
	M_DEVBUF, M_WAITOK);
	bp->bio_ma_n = 0;
	bp->bio_ma_offset = 0;
	abd_iterate_func(zio->io_abd, 0, zio->io_size,
	vdev_geom_fill_unmap_cb, bp);
	bp->bio_data = unmapped_buf;
	bp->bio_flags \|= BIO_UNMAPPED;
	} else {
	if (zio->io_type == ZIO_TYPE_READ) {
	bp->bio_data = abd_borrow_buf(zio->io_abd,
	zio->io_size);
	} else {
	bp->bio_data = abd_borrow_buf_copy(zio->io_abd,
	zio->io_size);
	}
	}
	break;
	case ZIO_TYPE_TRIM:
	bp->bio_cmd = BIO_DELETE;
	bp->bio_data = NULL;
	bp->bio_offset = zio->io_offset;
	bp->bio_length = zio->io_size;
	break;
	case ZIO_TYPE_FLUSH:
	bp->bio_cmd = BIO_FLUSH;
	bp->bio_data = NULL;
	bp->bio_offset = cp->provider->mediasize;
	bp->bio_length = 0;
	break;
	default:
	panic("invalid zio->io_type: %d\n", zio->io_type);
	}
	bp->bio_done = vdev_geom_io_intr;
	zio->io_bio = bp;

	g_io_request(bp, cp);
	}

	static void
	vdev_geom_io_done(zio_t *zio)
	{
	struct bio *bp = zio->io_bio;

	if (zio->io_type != ZIO_TYPE_READ && zio->io_type != ZIO_TYPE_WRITE) {
	ASSERT3P(bp, ==, NULL);
	return;
	}

	if (bp == NULL) {
	ASSERT3S(zio->io_error, ==, ENXIO);
	return;
	}

	if (bp->bio_ma != NULL) {
	free(bp->bio_ma, M_DEVBUF);
	} else {
	if (zio->io_type == ZIO_TYPE_READ) {
	abd_return_buf_copy(zio->io_abd, bp->bio_data,
	zio->io_size);
	} else {
	abd_return_buf(zio->io_abd, bp->bio_data,
	zio->io_size);
	}
	}

	g_destroy_bio(bp);
	zio->io_bio = NULL;
	}

	static void
	vdev_geom_hold(vdev_t *vd)
	{
	}

	static void
	vdev_geom_rele(vdev_t *vd)
	{
	}

	vdev_ops_t vdev_disk_ops = {
	.vdev_op_init = NULL,
	.vdev_op_fini = NULL,
	.vdev_op_open = vdev_geom_open,
	.vdev_op_close = vdev_geom_close,
	.vdev_op_asize = vdev_default_asize,
	.vdev_op_min_asize = vdev_default_min_asize,
	.vdev_op_min_alloc = NULL,
	.vdev_op_io_start = vdev_geom_io_start,
	.vdev_op_io_done = vdev_geom_io_done,
	.vdev_op_state_change = NULL,
	.vdev_op_need_resilver = NULL,
	.vdev_op_hold = vdev_geom_hold,
	.vdev_op_rele = vdev_geom_rele,
	.vdev_op_remap = NULL,
	.vdev_op_xlate = vdev_default_xlate,
	.vdev_op_rebuild_asize = NULL,
	.vdev_op_metaslab_init = NULL,
	.vdev_op_config_generate = NULL,
	.vdev_op_nparity = NULL,
	.vdev_op_ndisks = NULL,
	.vdev_op_type = VDEV_TYPE_DISK, /* name of this vdev type */
	.vdev_op_leaf = B_TRUE /* leaf vdev */
	};
	diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_debug.c b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_debug.c
	index 78d50c6fd8b7..c4cebe102075 100644
	--- a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_debug.c
	+++ b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_debug.c
	@@ -1,252 +1,267 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2014 by Delphix. All rights reserved.
	*/

	#include <sys/zfs_context.h>
	#include <sys/kstat.h>

	typedef struct zfs_dbgmsg {
	list_node_t zdm_node;
	time_t zdm_timestamp;
	uint_t zdm_size;
	char zdm_msg[];
	} zfs_dbgmsg_t;

	static list_t zfs_dbgmsgs;
	static uint_t zfs_dbgmsg_size = 0;
	static kmutex_t zfs_dbgmsgs_lock;
	uint_t zfs_dbgmsg_maxsize = 4<<20; /* 4MB */
	static kstat_t *zfs_dbgmsg_kstat;

	/*
	* Internal ZFS debug messages are enabled by default.
	*
	* # Print debug messages as they're logged
	* dtrace -n 'zfs-dbgmsg { print(stringof(arg0)); }'
	*
	* # Print all logged dbgmsg entries
	* sysctl kstat.zfs.misc.dbgmsg
	*
	* # Disable the kernel debug message log.
	* sysctl vfs.zfs.dbgmsg_enable=0
	*/
	int zfs_dbgmsg_enable = B_TRUE;

	static int
	zfs_dbgmsg_headers(char *buf, size_t size)
	{
	(void) snprintf(buf, size, "%-12s %-8s\n", "timestamp", "message");

	return (0);
	}

	static int
	zfs_dbgmsg_data(char buf, size_t size, void data)
	{
	zfs_dbgmsg_t zdm = (zfs_dbgmsg_t )data;

	(void) snprintf(buf, size, "%-12llu %-s\n",
	(u_longlong_t)zdm->zdm_timestamp, zdm->zdm_msg);

	return (0);
	}

	static void *
	zfs_dbgmsg_addr(kstat_t *ksp, loff_t n)
	{
	zfs_dbgmsg_t zdm = (zfs_dbgmsg_t )ksp->ks_private;

	ASSERT(MUTEX_HELD(&zfs_dbgmsgs_lock));

	if (n == 0)
	ksp->ks_private = list_head(&zfs_dbgmsgs);
	else if (zdm)
	ksp->ks_private = list_next(&zfs_dbgmsgs, zdm);

	return (ksp->ks_private);
	}

	static void
	zfs_dbgmsg_purge(uint_t max_size)
	{
	zfs_dbgmsg_t *zdm;
	uint_t size;

	ASSERT(MUTEX_HELD(&zfs_dbgmsgs_lock));

	while (zfs_dbgmsg_size > max_size) {
	zdm = list_remove_head(&zfs_dbgmsgs);
	if (zdm == NULL)
	return;

	size = zdm->zdm_size;
	kmem_free(zdm, size);
	zfs_dbgmsg_size -= size;
	}
	}

	static int
	zfs_dbgmsg_update(kstat_t *ksp, int rw)
	{
	if (rw == KSTAT_WRITE)
	zfs_dbgmsg_purge(0);

	return (0);
	}

	void
	zfs_dbgmsg_init(void)
	{
	list_create(&zfs_dbgmsgs, sizeof (zfs_dbgmsg_t),
	offsetof(zfs_dbgmsg_t, zdm_node));
	mutex_init(&zfs_dbgmsgs_lock, NULL, MUTEX_DEFAULT, NULL);

	zfs_dbgmsg_kstat = kstat_create("zfs", 0, "dbgmsg", "misc",
	KSTAT_TYPE_RAW, 0, KSTAT_FLAG_VIRTUAL);
	if (zfs_dbgmsg_kstat) {
	zfs_dbgmsg_kstat->ks_lock = &zfs_dbgmsgs_lock;
	zfs_dbgmsg_kstat->ks_ndata = UINT32_MAX;
	zfs_dbgmsg_kstat->ks_private = NULL;
	zfs_dbgmsg_kstat->ks_update = zfs_dbgmsg_update;
	kstat_set_raw_ops(zfs_dbgmsg_kstat, zfs_dbgmsg_headers,
	zfs_dbgmsg_data, zfs_dbgmsg_addr);
	kstat_install(zfs_dbgmsg_kstat);
	}
	}

	void
	zfs_dbgmsg_fini(void)
	{
	if (zfs_dbgmsg_kstat)
	kstat_delete(zfs_dbgmsg_kstat);
	/*
	* TODO - decide how to make this permanent
	*/
	#ifdef _KERNEL
	mutex_enter(&zfs_dbgmsgs_lock);
	zfs_dbgmsg_purge(0);
	mutex_exit(&zfs_dbgmsgs_lock);
	mutex_destroy(&zfs_dbgmsgs_lock);
	#endif
	}

	void
	__zfs_dbgmsg(char *buf)
	{
	zfs_dbgmsg_t *zdm;
	uint_t size;

	DTRACE_PROBE1(zfs__dbgmsg, char *, buf);

	size = sizeof (zfs_dbgmsg_t) + strlen(buf) + 1;
	zdm = kmem_zalloc(size, KM_SLEEP);
	zdm->zdm_size = size;
	zdm->zdm_timestamp = gethrestime_sec();
	strcpy(zdm->zdm_msg, buf);

	mutex_enter(&zfs_dbgmsgs_lock);
	list_insert_tail(&zfs_dbgmsgs, zdm);
	zfs_dbgmsg_size += size;
	zfs_dbgmsg_purge(zfs_dbgmsg_maxsize);
	mutex_exit(&zfs_dbgmsgs_lock);
	}

	void
	__set_error(const char file, const char func, int line, int err)
	{
	/*
	* To enable this:
	*
	* $ echo 512 >/sys/module/zfs/parameters/zfs_flags
	*/
	if (zfs_flags & ZFS_DEBUG_SET_ERROR)
	__dprintf(B_FALSE, file, func, line, "error %lu", (ulong_t)err);
	}

	#ifdef _KERNEL
	void
	__dprintf(boolean_t dprint, const char file, const char func,
	int line, const char *fmt, ...)
	{
	const char *newfile;
	va_list adx;
	size_t size;
	char *buf;
	char *nl;
	int i;

	size = 1024;
	buf = kmem_alloc(size, KM_SLEEP);

	/*
	* Get rid of annoying prefix to filename.
	*/
	newfile = strrchr(file, '/');
	if (newfile != NULL) {
	newfile = newfile + 1; /* Get rid of leading / */
	} else {
	newfile = file;
	}

	i = snprintf(buf, size, "%s:%d:%s(): ", newfile, line, func);

	if (i < size) {
	va_start(adx, fmt);
	(void) vsnprintf(buf + i, size - i, fmt, adx);
	va_end(adx);
	}

	/*
	* Get rid of trailing newline.
	*/
	nl = strrchr(buf, '\n');
	if (nl != NULL)
	*nl = '\0';

	__zfs_dbgmsg(buf);

	kmem_free(buf, size);
	}

	#else

	void
	-zfs_dbgmsg_print(const char *tag)
	+zfs_dbgmsg_print(int fd, const char *tag)
	{
	- zfs_dbgmsg_t *zdm;
	+ ssize_t ret __attribute__((unused));
	+
	+ /*
	+ * We use write() in this function instead of printf()
	+ * so it is safe to call from a signal handler.
	+ */
	+ ret = write(fd, "ZFS_DBGMSG(", 11);
	+ ret = write(fd, tag, strlen(tag));
	+ ret = write(fd, ") START:\n", 9);

	- (void) printf("ZFS_DBGMSG(%s):\n", tag);
	mutex_enter(&zfs_dbgmsgs_lock);
	- for (zdm = list_head(&zfs_dbgmsgs); zdm;
	+
	+ for (zfs_dbgmsg_t *zdm = list_head(&zfs_dbgmsgs); zdm != NULL;
	zdm = list_next(&zfs_dbgmsgs, zdm))
	- (void) printf("%s\n", zdm->zdm_msg);
	+ ret = write(fd, zdm->zdm_msg, strlen(zdm->zdm_msg));
	+ ret = write(fd, "\n", 1);
	+ }
	+
	+ ret = write(fd, "ZFS_DBGMSG(", 11);
	+ ret = write(fd, tag, strlen(tag));
	+ ret = write(fd, ") END\n", 6);
	+
	mutex_exit(&zfs_dbgmsgs_lock);
	}
	#endif /* _KERNEL */

	ZFS_MODULE_PARAM(zfs, zfs_, dbgmsg_enable, INT, ZMOD_RW,
	"Enable ZFS debug message log");

	ZFS_MODULE_PARAM(zfs, zfs_, dbgmsg_maxsize, UINT, ZMOD_RW,
	"Maximum ZFS debug log size");
	diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_dir.c b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_dir.c
	index 948df8e50de1..3cdb94d6cd53 100644
	--- a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_dir.c
	+++ b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_dir.c
	@@ -1,969 +1,970 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2013, 2016 by Delphix. All rights reserved.
	* Copyright 2017 Nexenta Systems, Inc.
	*/

	#include <sys/types.h>
	#include <sys/param.h>
	#include <sys/time.h>
	#include <sys/systm.h>
	#include <sys/sysmacros.h>
	#include <sys/resource.h>
	#include <sys/vfs.h>
	#include <sys/vnode.h>
	#include <sys/file.h>
	#include <sys/kmem.h>
	#include <sys/uio.h>
	#include <sys/cmn_err.h>
	#include <sys/errno.h>
	#include <sys/stat.h>
	#include <sys/unistd.h>
	#include <sys/sunddi.h>
	#include <sys/random.h>
	#include <sys/policy.h>
	#include <sys/condvar.h>
	#include <sys/callb.h>
	#include <sys/smp.h>
	#include <sys/zfs_dir.h>
	#include <sys/zfs_acl.h>
	#include <sys/fs/zfs.h>
	#include <sys/zap.h>
	#include <sys/dmu.h>
	#include <sys/atomic.h>
	#include <sys/zfs_ctldir.h>
	#include <sys/zfs_fuid.h>
	#include <sys/sa.h>
	#include <sys/zfs_sa.h>
	#include <sys/dmu_objset.h>
	#include <sys/dsl_dir.h>

	#include <sys/ccompat.h>

	/*
	* zfs_match_find() is used by zfs_dirent_lookup() to perform zap lookups
	* of names after deciding which is the appropriate lookup interface.
	*/
	static int
	zfs_match_find(zfsvfs_t zfsvfs, znode_t dzp, const char *name,
	matchtype_t mt, uint64_t *zoid)
	{
	int error;

	if (zfsvfs->z_norm) {

	/*
	* In the non-mixed case we only expect there would ever
	* be one match, but we need to use the normalizing lookup.
	*/
	error = zap_lookup_norm(zfsvfs->z_os, dzp->z_id, name, 8, 1,
	zoid, mt, NULL, 0, NULL);
	} else {
	error = zap_lookup(zfsvfs->z_os, dzp->z_id, name, 8, 1, zoid);
	}
	zoid = ZFS_DIRENT_OBJ(zoid);

	return (error);
	}

	/*
	* Look up a directory entry under a locked vnode.
	* dvp being locked gives us a guarantee that there are no concurrent
	* modification of the directory and, thus, if a node can be found in
	* the directory, then it must not be unlinked.
	*
	* Input arguments:
	* dzp - znode for directory
	* name - name of entry to lock
	* flag - ZNEW: if the entry already exists, fail with EEXIST.
	* ZEXISTS: if the entry does not exist, fail with ENOENT.
	* ZXATTR: we want dzp's xattr directory
	*
	* Output arguments:
	* zpp - pointer to the znode for the entry (NULL if there isn't one)
	*
	* Return value: 0 on success or errno on failure.
	*
	* NOTE: Always checks for, and rejects, '.' and '..'.
	*/
	int
	zfs_dirent_lookup(znode_t dzp, const char name, znode_t **zpp, int flag)
	{
	zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
	znode_t *zp;
	matchtype_t mt = 0;
	uint64_t zoid;
	int error = 0;

	if (zfsvfs->z_replay == B_FALSE)
	ASSERT_VOP_LOCKED(ZTOV(dzp), __func__);

	*zpp = NULL;

	/*
	* Verify that we are not trying to lock '.', '..', or '.zfs'
	*/
	if (name[0] == '.' &&
	(((name[1] == '\0') \|\| (name[1] == '.' && name[2] == '\0')) \|\|
	(zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0)))
	return (SET_ERROR(EEXIST));

	/*
	* Case sensitivity and normalization preferences are set when
	* the file system is created. These are stored in the
	* zfsvfs->z_case and zfsvfs->z_norm fields. These choices
	* affect how we perform zap lookups.
	*
	* When matching we may need to normalize & change case according to
	* FS settings.
	*
	* Note that a normalized match is necessary for a case insensitive
	* filesystem when the lookup request is not exact because normalization
	* can fold case independent of normalizing code point sequences.
	*
	* See the table above zfs_dropname().
	*/
	if (zfsvfs->z_norm != 0) {
	mt = MT_NORMALIZE;

	/*
	* Determine if the match needs to honor the case specified in
	* lookup, and if so keep track of that so that during
	* normalization we don't fold case.
	*/
	if (zfsvfs->z_case == ZFS_CASE_MIXED) {
	mt \|= MT_MATCH_CASE;
	}
	}

	/*
	* Only look in or update the DNLC if we are looking for the
	* name on a file system that does not require normalization
	* or case folding. We can also look there if we happen to be
	* on a non-normalizing, mixed sensitivity file system IF we
	* are looking for the exact name.
	*
	* NB: we do not need to worry about this flag for ZFS_CASE_SENSITIVE
	* because in that case MT_EXACT and MT_FIRST should produce exactly
	* the same result.
	*/

	if (dzp->z_unlinked && !(flag & ZXATTR))
	return (ENOENT);
	if (flag & ZXATTR) {
	error = sa_lookup(dzp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &zoid,
	sizeof (zoid));
	if (error == 0)
	error = (zoid == 0 ? ENOENT : 0);
	} else {
	error = zfs_match_find(zfsvfs, dzp, name, mt, &zoid);
	}
	if (error) {
	if (error != ENOENT \|\| (flag & ZEXISTS)) {
	return (error);
	}
	} else {
	if (flag & ZNEW) {
	return (SET_ERROR(EEXIST));
	}
	error = zfs_zget(zfsvfs, zoid, &zp);
	if (error)
	return (error);
	ASSERT(!zp->z_unlinked);
	*zpp = zp;
	}

	return (0);
	}

	static int
	zfs_dd_lookup(znode_t dzp, znode_t *zpp)
	{
	zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
	znode_t *zp;
	uint64_t parent;
	int error;

	#ifdef ZFS_DEBUG
	if (zfsvfs->z_replay == B_FALSE)
	ASSERT_VOP_LOCKED(ZTOV(dzp), __func__);
	#endif
	if (dzp->z_unlinked)
	return (ENOENT);

	if ((error = sa_lookup(dzp->z_sa_hdl,
	SA_ZPL_PARENT(zfsvfs), &parent, sizeof (parent))) != 0)
	return (error);

	error = zfs_zget(zfsvfs, parent, &zp);
	if (error == 0)
	*zpp = zp;
	return (error);
	}

	int
	zfs_dirlook(znode_t dzp, const char name, znode_t **zpp)
	{
	zfsvfs_t *zfsvfs __unused = dzp->z_zfsvfs;
	znode_t *zp = NULL;
	int error = 0;

	#ifdef ZFS_DEBUG
	if (zfsvfs->z_replay == B_FALSE)
	ASSERT_VOP_LOCKED(ZTOV(dzp), __func__);
	#endif
	if (dzp->z_unlinked)
	return (SET_ERROR(ENOENT));

	if (name[0] == 0 \|\| (name[0] == '.' && name[1] == 0)) {
	*zpp = dzp;
	} else if (name[0] == '.' && name[1] == '.' && name[2] == 0) {
	error = zfs_dd_lookup(dzp, &zp);
	if (error == 0)
	*zpp = zp;
	} else {
	error = zfs_dirent_lookup(dzp, name, &zp, ZEXISTS);
	if (error == 0) {
	dzp->z_zn_prefetch = B_TRUE; /* enable prefetching */
	*zpp = zp;
	}
	}
	return (error);
	}

	/*
	* unlinked Set (formerly known as the "delete queue") Error Handling
	*
	* When dealing with the unlinked set, we dmu_tx_hold_zap(), but we
	* don't specify the name of the entry that we will be manipulating. We
	* also fib and say that we won't be adding any new entries to the
	* unlinked set, even though we might (this is to lower the minimum file
	* size that can be deleted in a full filesystem). So on the small
	* chance that the nlink list is using a fat zap (ie. has more than
	* 2000 entries), we may not pre-read a block that's needed.
	* Therefore it is remotely possible for some of the assertions
	* regarding the unlinked set below to fail due to i/o error. On a
	* nondebug system, this will result in the space being leaked.
	*/
	void
	zfs_unlinked_add(znode_t zp, dmu_tx_t tx)
	{
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;

	ASSERT(zp->z_unlinked);
	ASSERT3U(zp->z_links, ==, 0);

	VERIFY0(zap_add_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx));

	dataset_kstats_update_nunlinks_kstat(&zfsvfs->z_kstat, 1);
	}

	/*
	* Clean up any znodes that had no links when we either crashed or
	* (force) umounted the file system.
	*/
	void
	zfs_unlinked_drain(zfsvfs_t *zfsvfs)
	{
	zap_cursor_t zc;
	zap_attribute_t zap;
	dmu_object_info_t doi;
	znode_t *zp;
	dmu_tx_t *tx;
	int error;

	/*
	* Iterate over the contents of the unlinked set.
	*/
	for (zap_cursor_init(&zc, zfsvfs->z_os, zfsvfs->z_unlinkedobj);
	zap_cursor_retrieve(&zc, &zap) == 0;
	zap_cursor_advance(&zc)) {

	/*
	* See what kind of object we have in list
	*/

	error = dmu_object_info(zfsvfs->z_os,
	zap.za_first_integer, &doi);
	if (error != 0)
	continue;

	ASSERT((doi.doi_type == DMU_OT_PLAIN_FILE_CONTENTS) \|\|
	(doi.doi_type == DMU_OT_DIRECTORY_CONTENTS));
	/*
	* We need to re-mark these list entries for deletion,
	* so we pull them back into core and set zp->z_unlinked.
	*/
	error = zfs_zget(zfsvfs, zap.za_first_integer, &zp);

	/*
	* We may pick up znodes that are already marked for deletion.
	* This could happen during the purge of an extended attribute
	* directory. All we need to do is skip over them, since they
	* are already in the system marked z_unlinked.
	*/
	if (error != 0)
	continue;

	vn_lock(ZTOV(zp), LK_EXCLUSIVE \| LK_RETRY);

	/*
	* Due to changes in zfs_rmnode we need to make sure the
	* link count is set to zero here.
	*/
	if (zp->z_links != 0) {
	tx = dmu_tx_create(zfsvfs->z_os);
	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error != 0) {
	dmu_tx_abort(tx);
	vput(ZTOV(zp));
	continue;
	}
	zp->z_links = 0;
	VERIFY0(sa_update(zp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs),
	&zp->z_links, sizeof (zp->z_links), tx));
	dmu_tx_commit(tx);
	}

	zp->z_unlinked = B_TRUE;
	vput(ZTOV(zp));
	}
	zap_cursor_fini(&zc);
	}

	/*
	* Delete the entire contents of a directory. Return a count
	* of the number of entries that could not be deleted. If we encounter
	* an error, return a count of at least one so that the directory stays
	* in the unlinked set.
	*
	* NOTE: this function assumes that the directory is inactive,
	* so there is no need to lock its entries before deletion.
	* Also, it assumes the directory contents is only regular
	* files.
	*/
	static int
	zfs_purgedir(znode_t *dzp)
	{
	zap_cursor_t zc;
	zap_attribute_t zap;
	znode_t *xzp;
	dmu_tx_t *tx;
	zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
	int skipped = 0;
	int error;

	for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
	(error = zap_cursor_retrieve(&zc, &zap)) == 0;
	zap_cursor_advance(&zc)) {
	error = zfs_zget(zfsvfs,
	ZFS_DIRENT_OBJ(zap.za_first_integer), &xzp);
	if (error) {
	skipped += 1;
	continue;
	}

	vn_lock(ZTOV(xzp), LK_EXCLUSIVE \| LK_RETRY);
	ASSERT((ZTOV(xzp)->v_type == VREG) \|\|
	(ZTOV(xzp)->v_type == VLNK));

	tx = dmu_tx_create(zfsvfs->z_os);
	dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE);
	dmu_tx_hold_zap(tx, dzp->z_id, FALSE, zap.za_name);
	dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE);
	dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
	/* Is this really needed ? */
	zfs_sa_upgrade_txholds(tx, xzp);
	dmu_tx_mark_netfree(tx);
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error) {
	dmu_tx_abort(tx);
	vput(ZTOV(xzp));
	skipped += 1;
	continue;
	}

	error = zfs_link_destroy(dzp, zap.za_name, xzp, tx, 0, NULL);
	if (error)
	skipped += 1;
	dmu_tx_commit(tx);

	vput(ZTOV(xzp));
	}
	zap_cursor_fini(&zc);
	if (error != ENOENT)
	skipped += 1;
	return (skipped);
	}

	extern taskq_t *zfsvfs_taskq;

	void
	zfs_rmnode(znode_t *zp)
	{
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	objset_t *os = zfsvfs->z_os;
	dmu_tx_t *tx;
	uint64_t z_id = zp->z_id;
	uint64_t acl_obj;
	uint64_t xattr_obj;
	uint64_t count;
	int error;

	ASSERT3U(zp->z_links, ==, 0);
	if (zfsvfs->z_replay == B_FALSE)
	ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);

	/*
	* If this is an attribute directory, purge its contents.
	*/
	if (ZTOV(zp) != NULL && ZTOV(zp)->v_type == VDIR &&
	(zp->z_pflags & ZFS_XATTR)) {
	if (zfs_purgedir(zp) != 0) {
	/*
	* Not enough space to delete some xattrs.
	* Leave it in the unlinked set.
	*/
	ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id);
	zfs_znode_dmu_fini(zp);
	zfs_znode_free(zp);
	ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
	return;
	}
	} else {
	/*
	* Free up all the data in the file. We don't do this for
	* XATTR directories because we need truncate and remove to be
	* in the same tx, like in zfs_znode_delete(). Otherwise, if
	* we crash here we'll end up with an inconsistent truncated
	* zap object in the delete queue. Note a truncated file is
	* harmless since it only contains user data.
	*/
	error = dmu_free_long_range(os, zp->z_id, 0, DMU_OBJECT_END);
	if (error) {
	/*
	* Not enough space or we were interrupted by unmount.
	* Leave the file in the unlinked set.
	*/
	ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id);
	zfs_znode_dmu_fini(zp);
	zfs_znode_free(zp);
	ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
	return;
	}
	}

	/*
	* If the file has extended attributes, we're going to unlink
	* the xattr dir.
	*/
	error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
	&xattr_obj, sizeof (xattr_obj));
	if (error)
	xattr_obj = 0;

	acl_obj = zfs_external_acl(zp);

	/*
	* Set up the final transaction.
	*/
	tx = dmu_tx_create(os);
	dmu_tx_hold_free(tx, zp->z_id, 0, DMU_OBJECT_END);
	dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
	if (xattr_obj)
	dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, TRUE, NULL);
	if (acl_obj)
	dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);

	zfs_sa_upgrade_txholds(tx, zp);
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error) {
	/*
	* Not enough space to delete the file. Leave it in the
	* unlinked set, leaking it until the fs is remounted (at
	* which point we'll call zfs_unlinked_drain() to process it).
	*/
	dmu_tx_abort(tx);
	ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id);
	zfs_znode_dmu_fini(zp);
	zfs_znode_free(zp);
	ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
	return;
	}

	/*
	* FreeBSD's implementation of zfs_zget requires a vnode to back it.
	* This means that we could end up calling into getnewvnode while
	* calling zfs_rmnode as a result of a prior call to getnewvnode
	* trying to clear vnodes out of the cache. If this repeats we can
	* recurse enough that we overflow our stack. To avoid this, we
	* avoid calling zfs_zget on the xattr znode and instead simply add
	* it to the unlinked set and schedule a call to zfs_unlinked_drain.
	*/
	if (xattr_obj) {
	/* Add extended attribute directory to the unlinked set. */
	VERIFY3U(0, ==,
	zap_add_int(os, zfsvfs->z_unlinkedobj, xattr_obj, tx));
	}

	mutex_enter(&os->os_dsl_dataset->ds_dir->dd_activity_lock);

	/* Remove this znode from the unlinked set */
	VERIFY3U(0, ==,
	zap_remove_int(os, zfsvfs->z_unlinkedobj, zp->z_id, tx));

	if (zap_count(os, zfsvfs->z_unlinkedobj, &count) == 0 && count == 0) {
	cv_broadcast(&os->os_dsl_dataset->ds_dir->dd_activity_cv);
	}

	mutex_exit(&os->os_dsl_dataset->ds_dir->dd_activity_lock);

	dataset_kstats_update_nunlinked_kstat(&zfsvfs->z_kstat, 1);

	zfs_znode_delete(zp, tx);
	+ zfs_znode_free(zp);

	dmu_tx_commit(tx);

	if (xattr_obj) {
	/*
	* We're using the FreeBSD taskqueue API here instead of
	* the Solaris taskq API since the FreeBSD API allows for a
	* task to be enqueued multiple times but executed once.
	*/
	taskqueue_enqueue(zfsvfs_taskq->tq_queue,
	&zfsvfs->z_unlinked_drain_task);
	}
	}

	static uint64_t
	zfs_dirent(znode_t *zp, uint64_t mode)
	{
	uint64_t de = zp->z_id;

	if (zp->z_zfsvfs->z_version >= ZPL_VERSION_DIRENT_TYPE)
	de \|= IFTODT(mode) << 60;
	return (de);
	}

	/*
	* Link zp into dzp. Can only fail if zp has been unlinked.
	*/
	int
	zfs_link_create(znode_t dzp, const char name, znode_t zp, dmu_tx_t tx,
	int flag)
	{
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	vnode_t *vp = ZTOV(zp);
	uint64_t value;
	int zp_is_dir = (vp->v_type == VDIR);
	sa_bulk_attr_t bulk[5];
	uint64_t mtime[2], ctime[2];
	int count = 0;
	int error;

	if (zfsvfs->z_replay == B_FALSE) {
	ASSERT_VOP_ELOCKED(ZTOV(dzp), __func__);
	ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
	}
	if (zp_is_dir) {
	if (dzp->z_links >= ZFS_LINK_MAX)
	return (SET_ERROR(EMLINK));
	}
	if (!(flag & ZRENAMING)) {
	if (zp->z_unlinked) { /* no new links to unlinked zp */
	ASSERT(!(flag & (ZNEW \| ZEXISTS)));
	return (SET_ERROR(ENOENT));
	}
	if (zp->z_links >= ZFS_LINK_MAX - zp_is_dir) {
	return (SET_ERROR(EMLINK));
	}
	zp->z_links++;
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
	&zp->z_links, sizeof (zp->z_links));

	} else {
	ASSERT(!zp->z_unlinked);
	}
	value = zfs_dirent(zp, zp->z_mode);
	error = zap_add(zp->z_zfsvfs->z_os, dzp->z_id, name,
	8, 1, &value, tx);

	/*
	* zap_add could fail to add the entry if it exceeds the capacity of the
	* leaf-block and zap_leaf_split() failed to help.
	* The caller of this routine is responsible for failing the transaction
	* which will rollback the SA updates done above.
	*/
	if (error != 0) {
	if (!(flag & ZRENAMING) && !(flag & ZNEW))
	zp->z_links--;
	return (error);
	}

	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL,
	&dzp->z_id, sizeof (dzp->z_id));
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
	&zp->z_pflags, sizeof (zp->z_pflags));

	if (!(flag & ZNEW)) {
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
	ctime, sizeof (ctime));
	zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime,
	ctime);
	}
	error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
	ASSERT0(error);

	dzp->z_size++;
	dzp->z_links += zp_is_dir;
	count = 0;
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
	&dzp->z_size, sizeof (dzp->z_size));
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
	&dzp->z_links, sizeof (dzp->z_links));
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
	mtime, sizeof (mtime));
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
	ctime, sizeof (ctime));
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
	&dzp->z_pflags, sizeof (dzp->z_pflags));
	zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime);
	error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx);
	ASSERT0(error);
	return (0);
	}

	/*
	* The match type in the code for this function should conform to:
	*
	* ------------------------------------------------------------------------
	* fs type \| z_norm \| lookup type \| match type
	* ---------\|-------------\|-------------\|----------------------------------
	* CS !norm \| 0 \| 0 \| 0 (exact)
	* CS norm \| formX \| 0 \| MT_NORMALIZE
	* CI !norm \| upper \| !ZCIEXACT \| MT_NORMALIZE
	* CI !norm \| upper \| ZCIEXACT \| MT_NORMALIZE \| MT_MATCH_CASE
	* CI norm \| upper\|formX \| !ZCIEXACT \| MT_NORMALIZE
	* CI norm \| upper\|formX \| ZCIEXACT \| MT_NORMALIZE \| MT_MATCH_CASE
	* CM !norm \| upper \| !ZCILOOK \| MT_NORMALIZE \| MT_MATCH_CASE
	* CM !norm \| upper \| ZCILOOK \| MT_NORMALIZE
	* CM norm \| upper\|formX \| !ZCILOOK \| MT_NORMALIZE \| MT_MATCH_CASE
	* CM norm \| upper\|formX \| ZCILOOK \| MT_NORMALIZE
	*
	* Abbreviations:
	* CS = Case Sensitive, CI = Case Insensitive, CM = Case Mixed
	* upper = case folding set by fs type on creation (U8_TEXTPREP_TOUPPER)
	* formX = unicode normalization form set on fs creation
	*/
	static int
	zfs_dropname(znode_t dzp, const char name, znode_t zp, dmu_tx_t tx,
	int flag)
	{
	int error;

	if (zp->z_zfsvfs->z_norm) {
	matchtype_t mt = MT_NORMALIZE;

	if (zp->z_zfsvfs->z_case == ZFS_CASE_MIXED) {
	mt \|= MT_MATCH_CASE;
	}

	error = zap_remove_norm(zp->z_zfsvfs->z_os, dzp->z_id,
	name, mt, tx);
	} else {
	error = zap_remove(zp->z_zfsvfs->z_os, dzp->z_id, name, tx);
	}

	return (error);
	}

	/*
	* Unlink zp from dzp, and mark zp for deletion if this was the last link.
	* Can fail if zp is a mount point (EBUSY) or a non-empty directory (EEXIST).
	* If 'unlinkedp' is NULL, we put unlinked znodes on the unlinked list.
	* If it's non-NULL, we use it to indicate whether the znode needs deletion,
	* and it's the caller's job to do it.
	*/
	int
	zfs_link_destroy(znode_t dzp, const char name, znode_t zp, dmu_tx_t tx,
	int flag, boolean_t *unlinkedp)
	{
	zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
	vnode_t *vp = ZTOV(zp);
	int zp_is_dir = (vp->v_type == VDIR);
	boolean_t unlinked = B_FALSE;
	sa_bulk_attr_t bulk[5];
	uint64_t mtime[2], ctime[2];
	int count = 0;
	int error;

	if (zfsvfs->z_replay == B_FALSE) {
	ASSERT_VOP_ELOCKED(ZTOV(dzp), __func__);
	ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
	}
	if (!(flag & ZRENAMING)) {

	if (zp_is_dir && !zfs_dirempty(zp))
	return (SET_ERROR(ENOTEMPTY));

	/*
	* If we get here, we are going to try to remove the object.
	* First try removing the name from the directory; if that
	* fails, return the error.
	*/
	error = zfs_dropname(dzp, name, zp, tx, flag);
	if (error != 0) {
	return (error);
	}

	if (zp->z_links <= zp_is_dir) {
	zfs_panic_recover("zfs: link count on vnode %p is %u, "
	"should be at least %u", zp->z_vnode,
	(int)zp->z_links,
	zp_is_dir + 1);
	zp->z_links = zp_is_dir + 1;
	}
	if (--zp->z_links == zp_is_dir) {
	zp->z_unlinked = B_TRUE;
	zp->z_links = 0;
	unlinked = B_TRUE;
	} else {
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs),
	NULL, &ctime, sizeof (ctime));
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
	NULL, &zp->z_pflags, sizeof (zp->z_pflags));
	zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime,
	ctime);
	}
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs),
	NULL, &zp->z_links, sizeof (zp->z_links));
	error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
	count = 0;
	ASSERT0(error);
	} else {
	ASSERT(!zp->z_unlinked);
	error = zfs_dropname(dzp, name, zp, tx, flag);
	if (error != 0)
	return (error);
	}

	dzp->z_size--; /* one dirent removed */
	dzp->z_links -= zp_is_dir; /* ".." link from zp */
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs),
	NULL, &dzp->z_links, sizeof (dzp->z_links));
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
	NULL, &dzp->z_size, sizeof (dzp->z_size));
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs),
	NULL, ctime, sizeof (ctime));
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs),
	NULL, mtime, sizeof (mtime));
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
	NULL, &dzp->z_pflags, sizeof (dzp->z_pflags));
	zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime);
	error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx);
	ASSERT0(error);

	if (unlinkedp != NULL)
	*unlinkedp = unlinked;
	else if (unlinked)
	zfs_unlinked_add(zp, tx);

	return (0);
	}

	/*
	* Indicate whether the directory is empty.
	*/
	boolean_t
	zfs_dirempty(znode_t *dzp)
	{
	return (dzp->z_size == 2);
	}

	int
	zfs_make_xattrdir(znode_t zp, vattr_t vap, znode_t *xvpp, cred_t cr)
	{
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	znode_t *xzp;
	dmu_tx_t *tx;
	int error;
	zfs_acl_ids_t acl_ids;
	boolean_t fuid_dirtied;
	uint64_t parent __maybe_unused;

	*xvpp = NULL;

	if ((error = zfs_acl_ids_create(zp, IS_XATTR, vap, cr, NULL,
	&acl_ids, NULL)) != 0)
	return (error);
	if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, 0)) {
	zfs_acl_ids_free(&acl_ids);
	return (SET_ERROR(EDQUOT));
	}

	getnewvnode_reserve_();

	tx = dmu_tx_create(zfsvfs->z_os);
	dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
	ZFS_SA_BASE_ATTR_SIZE);
	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
	dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
	fuid_dirtied = zfsvfs->z_fuid_dirty;
	if (fuid_dirtied)
	zfs_fuid_txhold(zfsvfs, tx);
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error) {
	zfs_acl_ids_free(&acl_ids);
	dmu_tx_abort(tx);
	getnewvnode_drop_reserve();
	return (error);
	}
	zfs_mknode(zp, vap, tx, cr, IS_XATTR, &xzp, &acl_ids);

	if (fuid_dirtied)
	zfs_fuid_sync(zfsvfs, tx);

	ASSERT0(sa_lookup(xzp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), &parent,
	sizeof (parent)));
	ASSERT3U(parent, ==, zp->z_id);

	VERIFY0(sa_update(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &xzp->z_id,
	sizeof (xzp->z_id), tx));

	zfs_log_create(zfsvfs->z_log, tx, TX_MKXATTR, zp, xzp, "", NULL,
	acl_ids.z_fuidp, vap);

	zfs_acl_ids_free(&acl_ids);
	dmu_tx_commit(tx);

	getnewvnode_drop_reserve();

	*xvpp = xzp;

	return (0);
	}

	/*
	* Return a znode for the extended attribute directory for zp.
	* If the directory does not already exist, it is created
	*
	* IN: zp - znode to obtain attribute directory from
	* cr - credentials of caller
	* flags - flags from the VOP_LOOKUP call
	*
	* OUT: xzpp - pointer to extended attribute znode
	*
	* RETURN: 0 on success
	* error number on failure
	*/
	int
	zfs_get_xattrdir(znode_t zp, znode_t xzpp, cred_t cr, int flags)
	{
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	znode_t *xzp;
	vattr_t va;
	int error;
	top:
	error = zfs_dirent_lookup(zp, "", &xzp, ZXATTR);
	if (error)
	return (error);

	if (xzp != NULL) {
	*xzpp = xzp;
	return (0);
	}


	if (!(flags & CREATE_XATTR_DIR))
	return (SET_ERROR(ENOATTR));

	if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) {
	return (SET_ERROR(EROFS));
	}

	/*
	* The ability to 'create' files in an attribute
	* directory comes from the write_xattr permission on the base file.
	*
	* The ability to 'search' an attribute directory requires
	* read_xattr permission on the base file.
	*
	* Once in a directory the ability to read/write attributes
	* is controlled by the permissions on the attribute file.
	*/
	va.va_mask = AT_MODE \| AT_UID \| AT_GID;
	va.va_type = VDIR;
	va.va_mode = S_IFDIR \| S_ISVTX \| 0777;
	zfs_fuid_map_ids(zp, cr, &va.va_uid, &va.va_gid);

	error = zfs_make_xattrdir(zp, &va, xzpp, cr);

	if (error == ERESTART) {
	/* NB: we already did dmu_tx_wait() if necessary */
	goto top;
	}
	if (error == 0)
	VOP_UNLOCK1(ZTOV(*xzpp));

	return (error);
	}

	/*
	* Decide whether it is okay to remove within a sticky directory.
	*
	* In sticky directories, write access is not sufficient;
	* you can remove entries from a directory only if:
	*
	* you own the directory,
	* you own the entry,
	* the entry is a plain file and you have write access,
	* or you are privileged (checked in secpolicy...).
	*
	* The function returns 0 if remove access is granted.
	*/
	int
	zfs_sticky_remove_access(znode_t zdp, znode_t zp, cred_t *cr)
	{
	uid_t uid;
	uid_t downer;
	uid_t fowner;
	zfsvfs_t *zfsvfs = zdp->z_zfsvfs;

	if (zdp->z_zfsvfs->z_replay)
	return (0);

	if ((zdp->z_mode & S_ISVTX) == 0)
	return (0);

	downer = zfs_fuid_map_id(zfsvfs, zdp->z_uid, cr, ZFS_OWNER);
	fowner = zfs_fuid_map_id(zfsvfs, zp->z_uid, cr, ZFS_OWNER);

	if ((uid = crgetuid(cr)) == downer \|\| uid == fowner \|\|
	(ZTOV(zp)->v_type == VREG &&
	zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr, NULL) == 0))
	return (0);
	else
	return (secpolicy_vnode_remove(ZTOV(zp), cr));
	}
	diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vnops_os.c b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vnops_os.c
	index 7f4f8e2e2f78..b9d4da3de084 100644
	--- a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vnops_os.c
	+++ b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_vnops_os.c
	@@ -1,6451 +1,6481 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2015 by Delphix. All rights reserved.
	* Copyright (c) 2014 Integros [integros.com]
	* Copyright 2017 Nexenta Systems, Inc.
	*/

	/* Portions Copyright 2007 Jeremy Teo */
	/* Portions Copyright 2010 Robert Milkowski */

	#include <sys/param.h>
	#include <sys/time.h>
	#include <sys/systm.h>
	#include <sys/sysmacros.h>
	#include <sys/resource.h>
	#include <security/mac/mac_framework.h>
	#include <sys/vfs.h>
	#include <sys/endian.h>
	#include <sys/vm.h>
	#include <sys/vnode.h>
	#if __FreeBSD_version >= 1300102
	#include <sys/smr.h>
	#endif
	#include <sys/dirent.h>
	#include <sys/file.h>
	#include <sys/stat.h>
	#include <sys/kmem.h>
	#include <sys/taskq.h>
	#include <sys/uio.h>
	#include <sys/atomic.h>
	#include <sys/namei.h>
	#include <sys/mman.h>
	#include <sys/cmn_err.h>
	#include <sys/kdb.h>
	#include <sys/sysproto.h>
	#include <sys/errno.h>
	#include <sys/unistd.h>
	#include <sys/zfs_dir.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/fs/zfs.h>
	#include <sys/dmu.h>
	#include <sys/dmu_objset.h>
	#include <sys/spa.h>
	#include <sys/txg.h>
	#include <sys/dbuf.h>
	#include <sys/zap.h>
	#include <sys/sa.h>
	#include <sys/policy.h>
	#include <sys/sunddi.h>
	#include <sys/filio.h>
	#include <sys/sid.h>
	#include <sys/zfs_ctldir.h>
	#include <sys/zfs_fuid.h>
	#include <sys/zfs_quota.h>
	#include <sys/zfs_sa.h>
	#include <sys/zfs_rlock.h>
	#include <sys/bio.h>
	#include <sys/buf.h>
	#include <sys/sched.h>
	#include <sys/acl.h>
	#include <sys/vmmeter.h>
	#include <vm/vm_param.h>
	#include <sys/zil.h>
	#include <sys/zfs_vnops.h>
	#include <sys/module.h>
	#include <sys/sysent.h>
	#include <sys/dmu_impl.h>
	#include <sys/brt.h>
	#include <sys/zfeature.h>

	#include <vm/vm_object.h>

	#include <sys/extattr.h>
	#include <sys/priv.h>

	#ifndef VN_OPEN_INVFS
	#define VN_OPEN_INVFS 0x0
	#endif

	VFS_SMR_DECLARE;

	#if __FreeBSD_version < 1300103
	#define NDFREE_PNBUF(ndp) NDFREE((ndp), NDF_ONLY_PNBUF)
	#endif

	#if __FreeBSD_version >= 1300047
	#define vm_page_wire_lock(pp)
	#define vm_page_wire_unlock(pp)
	#else
	#define vm_page_wire_lock(pp) vm_page_lock(pp)
	#define vm_page_wire_unlock(pp) vm_page_unlock(pp)
	#endif

	#ifdef DEBUG_VFS_LOCKS
	#define VNCHECKREF(vp) \
	VNASSERT((vp)->v_holdcnt > 0 && (vp)->v_usecount > 0, vp, \
	("%s: wrong ref counts", __func__));
	#else
	#define VNCHECKREF(vp)
	#endif

	#if __FreeBSD_version >= 1400045
	typedef uint64_t cookie_t;
	#else
	typedef ulong_t cookie_t;
	#endif

	/*
	* Programming rules.
	*
	* Each vnode op performs some logical unit of work. To do this, the ZPL must
	* properly lock its in-core state, create a DMU transaction, do the work,
	* record this work in the intent log (ZIL), commit the DMU transaction,
	* and wait for the intent log to commit if it is a synchronous operation.
	* Moreover, the vnode ops must work in both normal and log replay context.
	* The ordering of events is important to avoid deadlocks and references
	* to freed memory. The example below illustrates the following Big Rules:
	*
	* (1) A check must be made in each zfs thread for a mounted file system.
	* This is done avoiding races using zfs_enter(zfsvfs).
	* A zfs_exit(zfsvfs) is needed before all returns. Any znodes
	* must be checked with zfs_verify_zp(zp). Both of these macros
	* can return EIO from the calling function.
	*
	* (2) VN_RELE() should always be the last thing except for zil_commit()
	* (if necessary) and zfs_exit(). This is for 3 reasons:
	* First, if it's the last reference, the vnode/znode
	* can be freed, so the zp may point to freed memory. Second, the last
	* reference will call zfs_zinactive(), which may induce a lot of work --
	* pushing cached pages (which acquires range locks) and syncing out
	* cached atime changes. Third, zfs_zinactive() may require a new tx,
	* which could deadlock the system if you were already holding one.
	* If you must call VN_RELE() within a tx then use VN_RELE_ASYNC().
	*
	* (3) All range locks must be grabbed before calling dmu_tx_assign(),
	* as they can span dmu_tx_assign() calls.
	*
	* (4) If ZPL locks are held, pass TXG_NOWAIT as the second argument to
	* dmu_tx_assign(). This is critical because we don't want to block
	* while holding locks.
	*
	* If no ZPL locks are held (aside from zfs_enter()), use TXG_WAIT. This
	* reduces lock contention and CPU usage when we must wait (note that if
	* throughput is constrained by the storage, nearly every transaction
	* must wait).
	*
	* Note, in particular, that if a lock is sometimes acquired before
	* the tx assigns, and sometimes after (e.g. z_lock), then failing
	* to use a non-blocking assign can deadlock the system. The scenario:
	*
	* Thread A has grabbed a lock before calling dmu_tx_assign().
	* Thread B is in an already-assigned tx, and blocks for this lock.
	* Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open()
	* forever, because the previous txg can't quiesce until B's tx commits.
	*
	* If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT,
	* then drop all locks, call dmu_tx_wait(), and try again. On subsequent
	* calls to dmu_tx_assign(), pass TXG_NOTHROTTLE in addition to TXG_NOWAIT,
	* to indicate that this operation has already called dmu_tx_wait().
	* This will ensure that we don't retry forever, waiting a short bit
	* each time.
	*
	* (5) If the operation succeeded, generate the intent log entry for it
	* before dropping locks. This ensures that the ordering of events
	* in the intent log matches the order in which they actually occurred.
	* During ZIL replay the zfs_log_* functions will update the sequence
	* number to indicate the zil transaction has replayed.
	*
	* (6) At the end of each vnode op, the DMU tx must always commit,
	* regardless of whether there were any errors.
	*
	* (7) After dropping all locks, invoke zil_commit(zilog, foid)
	* to ensure that synchronous semantics are provided when necessary.
	*
	* In general, this is how things should be ordered in each vnode op:
	*
	* zfs_enter(zfsvfs); // exit if unmounted
	* top:
	* zfs_dirent_lookup(&dl, ...) // lock directory entry (may VN_HOLD())
	* rw_enter(...); // grab any other locks you need
	* tx = dmu_tx_create(...); // get DMU tx
	* dmu_tx_hold_*(); // hold each object you might modify
	* error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) \| TXG_NOWAIT);
	* if (error) {
	* rw_exit(...); // drop locks
	* zfs_dirent_unlock(dl); // unlock directory entry
	* VN_RELE(...); // release held vnodes
	* if (error == ERESTART) {
	* waited = B_TRUE;
	* dmu_tx_wait(tx);
	* dmu_tx_abort(tx);
	* goto top;
	* }
	* dmu_tx_abort(tx); // abort DMU tx
	* zfs_exit(zfsvfs); // finished in zfs
	* return (error); // really out of space
	* }
	* error = do_real_work(); // do whatever this VOP does
	* if (error == 0)
	* zfs_log_*(...); // on success, make ZIL entry
	* dmu_tx_commit(tx); // commit DMU tx -- error or not
	* rw_exit(...); // drop locks
	* zfs_dirent_unlock(dl); // unlock directory entry
	* VN_RELE(...); // release held vnodes
	* zil_commit(zilog, foid); // synchronous when necessary
	* zfs_exit(zfsvfs); // finished in zfs
	* return (error); // done, report error
	*/
	static int
	zfs_open(vnode_t *vpp, int flag, cred_t cr)
	{
	(void) cr;
	znode_t zp = VTOZ(vpp);
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	int error;

	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (error);

	if ((flag & FWRITE) && (zp->z_pflags & ZFS_APPENDONLY) &&
	((flag & FAPPEND) == 0)) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EPERM));
	}

	/*
	* Keep a count of the synchronous opens in the znode. On first
	* synchronous open we must convert all previous async transactions
	* into sync to keep correct ordering.
	*/
	if (flag & O_SYNC) {
	if (atomic_inc_32_nv(&zp->z_sync_cnt) == 1)
	zil_async_to_sync(zfsvfs->z_log, zp->z_id);
	}

	zfs_exit(zfsvfs, FTAG);
	return (0);
	}

	static int
	zfs_close(vnode_t vp, int flag, int count, offset_t offset, cred_t cr)
	{
	(void) offset, (void) cr;
	znode_t *zp = VTOZ(vp);
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	int error;

	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (error);

	/* Decrement the synchronous opens in the znode */
	if ((flag & O_SYNC) && (count == 1))
	atomic_dec_32(&zp->z_sync_cnt);

	zfs_exit(zfsvfs, FTAG);
	return (0);
	}

	static int
	zfs_ioctl(vnode_t vp, ulong_t com, intptr_t data, int flag, cred_t cred,
	int *rvalp)
	{
	(void) flag, (void) cred, (void) rvalp;
	loff_t off;
	int error;

	switch (com) {
	case _FIOFFS:
	{
	return (0);

	/*
	* The following two ioctls are used by bfu. Faking out,
	* necessary to avoid bfu errors.
	*/
	}
	case _FIOGDIO:
	case _FIOSDIO:
	{
	return (0);
	}

	case F_SEEK_DATA:
	case F_SEEK_HOLE:
	{
	off = (offset_t )data;
	/* offset parameter is in/out */
	error = zfs_holey(VTOZ(vp), com, &off);
	if (error)
	return (error);
	(offset_t )data = off;
	return (0);
	}
	}
	return (SET_ERROR(ENOTTY));
	}

	static vm_page_t
	page_busy(vnode_t *vp, int64_t start, int64_t off, int64_t nbytes)
	{
	vm_object_t obj;
	vm_page_t pp;
	int64_t end;

	/*
	* At present vm_page_clear_dirty extends the cleared range to DEV_BSIZE
	* aligned boundaries, if the range is not aligned. As a result a
	* DEV_BSIZE subrange with partially dirty data may get marked as clean.
	* It may happen that all DEV_BSIZE subranges are marked clean and thus
	* the whole page would be considered clean despite have some
	* dirty data.
	* For this reason we should shrink the range to DEV_BSIZE aligned
	* boundaries before calling vm_page_clear_dirty.
	*/
	end = rounddown2(off + nbytes, DEV_BSIZE);
	off = roundup2(off, DEV_BSIZE);
	nbytes = end - off;

	obj = vp->v_object;
	zfs_vmobject_assert_wlocked_12(obj);
	#if __FreeBSD_version < 1300050
	for (;;) {
	if ((pp = vm_page_lookup(obj, OFF_TO_IDX(start))) != NULL &&
	pp->valid) {
	if (vm_page_xbusied(pp)) {
	/*
	* Reference the page before unlocking and
	* sleeping so that the page daemon is less
	* likely to reclaim it.
	*/
	vm_page_reference(pp);
	vm_page_lock(pp);
	zfs_vmobject_wunlock(obj);
	vm_page_busy_sleep(pp, "zfsmwb", true);
	zfs_vmobject_wlock(obj);
	continue;
	}
	vm_page_sbusy(pp);
	} else if (pp != NULL) {
	ASSERT(!pp->valid);
	pp = NULL;
	}
	if (pp != NULL) {
	ASSERT3U(pp->valid, ==, VM_PAGE_BITS_ALL);
	vm_object_pip_add(obj, 1);
	pmap_remove_write(pp);
	if (nbytes != 0)
	vm_page_clear_dirty(pp, off, nbytes);
	}
	break;
	}
	#else
	vm_page_grab_valid_unlocked(&pp, obj, OFF_TO_IDX(start),
	VM_ALLOC_NOCREAT \| VM_ALLOC_SBUSY \| VM_ALLOC_NORMAL \|
	VM_ALLOC_IGN_SBUSY);
	if (pp != NULL) {
	ASSERT3U(pp->valid, ==, VM_PAGE_BITS_ALL);
	vm_object_pip_add(obj, 1);
	pmap_remove_write(pp);
	if (nbytes != 0)
	vm_page_clear_dirty(pp, off, nbytes);
	}
	#endif
	return (pp);
	}

	static void
	page_unbusy(vm_page_t pp)
	{

	vm_page_sunbusy(pp);
	#if __FreeBSD_version >= 1300041
	vm_object_pip_wakeup(pp->object);
	#else
	vm_object_pip_subtract(pp->object, 1);
	#endif
	}

	#if __FreeBSD_version > 1300051
	static vm_page_t
	page_hold(vnode_t *vp, int64_t start)
	{
	vm_object_t obj;
	vm_page_t m;

	obj = vp->v_object;
	vm_page_grab_valid_unlocked(&m, obj, OFF_TO_IDX(start),
	VM_ALLOC_NOCREAT \| VM_ALLOC_WIRED \| VM_ALLOC_IGN_SBUSY \|
	VM_ALLOC_NOBUSY);
	return (m);
	}
	#else
	static vm_page_t
	page_hold(vnode_t *vp, int64_t start)
	{
	vm_object_t obj;
	vm_page_t pp;

	obj = vp->v_object;
	zfs_vmobject_assert_wlocked(obj);

	for (;;) {
	if ((pp = vm_page_lookup(obj, OFF_TO_IDX(start))) != NULL &&
	pp->valid) {
	if (vm_page_xbusied(pp)) {
	/*
	* Reference the page before unlocking and
	* sleeping so that the page daemon is less
	* likely to reclaim it.
	*/
	vm_page_reference(pp);
	vm_page_lock(pp);
	zfs_vmobject_wunlock(obj);
	vm_page_busy_sleep(pp, "zfsmwb", true);
	zfs_vmobject_wlock(obj);
	continue;
	}

	ASSERT3U(pp->valid, ==, VM_PAGE_BITS_ALL);
	vm_page_wire_lock(pp);
	vm_page_hold(pp);
	vm_page_wire_unlock(pp);

	} else
	pp = NULL;
	break;
	}
	return (pp);
	}
	#endif

	static void
	page_unhold(vm_page_t pp)
	{

	vm_page_wire_lock(pp);
	#if __FreeBSD_version >= 1300035
	vm_page_unwire(pp, PQ_ACTIVE);
	#else
	vm_page_unhold(pp);
	#endif
	vm_page_wire_unlock(pp);
	}

	/*
	* When a file is memory mapped, we must keep the IO data synchronized
	* between the DMU cache and the memory mapped pages. What this means:
	*
	* On Write: If we find a memory mapped page, we write to both
	* the page and the dmu buffer.
	*/
	void
	update_pages(znode_t zp, int64_t start, int len, objset_t os)
	{
	vm_object_t obj;
	struct sf_buf *sf;
	vnode_t *vp = ZTOV(zp);
	caddr_t va;
	int off;

	ASSERT3P(vp->v_mount, !=, NULL);
	obj = vp->v_object;
	ASSERT3P(obj, !=, NULL);

	off = start & PAGEOFFSET;
	zfs_vmobject_wlock_12(obj);
	#if __FreeBSD_version >= 1300041
	vm_object_pip_add(obj, 1);
	#endif
	for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
	vm_page_t pp;
	int nbytes = imin(PAGESIZE - off, len);

	if ((pp = page_busy(vp, start, off, nbytes)) != NULL) {
	zfs_vmobject_wunlock_12(obj);

	va = zfs_map_page(pp, &sf);
	(void) dmu_read(os, zp->z_id, start + off, nbytes,
	va + off, DMU_READ_PREFETCH);
	zfs_unmap_page(sf);

	zfs_vmobject_wlock_12(obj);
	page_unbusy(pp);
	}
	len -= nbytes;
	off = 0;
	}
	#if __FreeBSD_version >= 1300041
	vm_object_pip_wakeup(obj);
	#else
	vm_object_pip_wakeupn(obj, 0);
	#endif
	zfs_vmobject_wunlock_12(obj);
	}

	/*
	* Read with UIO_NOCOPY flag means that sendfile(2) requests
	* ZFS to populate a range of page cache pages with data.
	*
	* NOTE: this function could be optimized to pre-allocate
	* all pages in advance, drain exclusive busy on all of them,
	* map them into contiguous KVA region and populate them
	* in one single dmu_read() call.
	*/
	int
	mappedread_sf(znode_t zp, int nbytes, zfs_uio_t uio)
	{
	vnode_t *vp = ZTOV(zp);
	objset_t *os = zp->z_zfsvfs->z_os;
	struct sf_buf *sf;
	vm_object_t obj;
	vm_page_t pp;
	int64_t start;
	caddr_t va;
	int len = nbytes;
	int error = 0;

	ASSERT3U(zfs_uio_segflg(uio), ==, UIO_NOCOPY);
	ASSERT3P(vp->v_mount, !=, NULL);
	obj = vp->v_object;
	ASSERT3P(obj, !=, NULL);
	ASSERT0(zfs_uio_offset(uio) & PAGEOFFSET);

	zfs_vmobject_wlock_12(obj);
	for (start = zfs_uio_offset(uio); len > 0; start += PAGESIZE) {
	int bytes = MIN(PAGESIZE, len);

	pp = vm_page_grab_unlocked(obj, OFF_TO_IDX(start),
	VM_ALLOC_SBUSY \| VM_ALLOC_NORMAL \| VM_ALLOC_IGN_SBUSY);
	if (vm_page_none_valid(pp)) {
	zfs_vmobject_wunlock_12(obj);
	va = zfs_map_page(pp, &sf);
	error = dmu_read(os, zp->z_id, start, bytes, va,
	DMU_READ_PREFETCH);
	if (bytes != PAGESIZE && error == 0)
	memset(va + bytes, 0, PAGESIZE - bytes);
	zfs_unmap_page(sf);
	zfs_vmobject_wlock_12(obj);
	#if __FreeBSD_version >= 1300081
	if (error == 0) {
	vm_page_valid(pp);
	vm_page_activate(pp);
	vm_page_do_sunbusy(pp);
	} else {
	zfs_vmobject_wlock(obj);
	if (!vm_page_wired(pp) && pp->valid == 0 &&
	vm_page_busy_tryupgrade(pp))
	vm_page_free(pp);
	else
	vm_page_sunbusy(pp);
	zfs_vmobject_wunlock(obj);
	}
	#else
	vm_page_do_sunbusy(pp);
	vm_page_lock(pp);
	if (error) {
	if (pp->wire_count == 0 && pp->valid == 0 &&
	!vm_page_busied(pp))
	vm_page_free(pp);
	} else {
	pp->valid = VM_PAGE_BITS_ALL;
	vm_page_activate(pp);
	}
	vm_page_unlock(pp);
	#endif
	} else {
	ASSERT3U(pp->valid, ==, VM_PAGE_BITS_ALL);
	vm_page_do_sunbusy(pp);
	}
	if (error)
	break;
	zfs_uio_advance(uio, bytes);
	len -= bytes;
	}
	zfs_vmobject_wunlock_12(obj);
	return (error);
	}

	/*
	* When a file is memory mapped, we must keep the IO data synchronized
	* between the DMU cache and the memory mapped pages. What this means:
	*
	* On Read: We "read" preferentially from memory mapped pages,
	* else we default from the dmu buffer.
	*
	* NOTE: We will always "break up" the IO into PAGESIZE uiomoves when
	* the file is memory mapped.
	*/
	int
	mappedread(znode_t zp, int nbytes, zfs_uio_t uio)
	{
	vnode_t *vp = ZTOV(zp);
	vm_object_t obj;
	int64_t start;
	int len = nbytes;
	int off;
	int error = 0;

	ASSERT3P(vp->v_mount, !=, NULL);
	obj = vp->v_object;
	ASSERT3P(obj, !=, NULL);

	start = zfs_uio_offset(uio);
	off = start & PAGEOFFSET;
	zfs_vmobject_wlock_12(obj);
	for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
	vm_page_t pp;
	uint64_t bytes = MIN(PAGESIZE - off, len);

	if ((pp = page_hold(vp, start))) {
	struct sf_buf *sf;
	caddr_t va;

	zfs_vmobject_wunlock_12(obj);
	va = zfs_map_page(pp, &sf);
	error = vn_io_fault_uiomove(va + off, bytes,
	GET_UIO_STRUCT(uio));
	zfs_unmap_page(sf);
	zfs_vmobject_wlock_12(obj);
	page_unhold(pp);
	} else {
	zfs_vmobject_wunlock_12(obj);
	error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl),
	uio, bytes);
	zfs_vmobject_wlock_12(obj);
	}
	len -= bytes;
	off = 0;
	if (error)
	break;
	}
	zfs_vmobject_wunlock_12(obj);
	return (error);
	}

	int
	zfs_write_simple(znode_t zp, const void data, size_t len,
	loff_t pos, size_t *presid)
	{
	int error = 0;
	ssize_t resid;

	error = vn_rdwr(UIO_WRITE, ZTOV(zp), __DECONST(void *, data), len, pos,
	UIO_SYSSPACE, IO_SYNC, kcred, NOCRED, &resid, curthread);

	if (error) {
	return (SET_ERROR(error));
	} else if (presid == NULL) {
	if (resid != 0) {
	error = SET_ERROR(EIO);
	}
	} else {
	*presid = resid;
	}
	return (error);
	}

	void
	zfs_zrele_async(znode_t *zp)
	{
	vnode_t *vp = ZTOV(zp);
	objset_t *os = ITOZSB(vp)->z_os;

	VN_RELE_ASYNC(vp, dsl_pool_zrele_taskq(dmu_objset_pool(os)));
	}

	static int
	zfs_dd_callback(struct mount mp, void arg, int lkflags, struct vnode **vpp)
	{
	int error;

	*vpp = arg;
	error = vn_lock(*vpp, lkflags);
	if (error != 0)
	vrele(*vpp);
	return (error);
	}

	static int
	zfs_lookup_lock(vnode_t dvp, vnode_t vp, const char *name, int lkflags)
	{
	znode_t *zdp = VTOZ(dvp);
	zfsvfs_t *zfsvfs __unused = zdp->z_zfsvfs;
	int error;
	int ltype;

	if (zfsvfs->z_replay == B_FALSE)
	ASSERT_VOP_LOCKED(dvp, __func__);

	if (name[0] == 0 \|\| (name[0] == '.' && name[1] == 0)) {
	ASSERT3P(dvp, ==, vp);
	vref(dvp);
	ltype = lkflags & LK_TYPE_MASK;
	if (ltype != VOP_ISLOCKED(dvp)) {
	if (ltype == LK_EXCLUSIVE)
	vn_lock(dvp, LK_UPGRADE \| LK_RETRY);
	else /* if (ltype == LK_SHARED) */
	vn_lock(dvp, LK_DOWNGRADE \| LK_RETRY);

	/*
	* Relock for the "." case could leave us with
	* reclaimed vnode.
	*/
	if (VN_IS_DOOMED(dvp)) {
	vrele(dvp);
	return (SET_ERROR(ENOENT));
	}
	}
	return (0);
	} else if (name[0] == '.' && name[1] == '.' && name[2] == 0) {
	/*
	* Note that in this case, dvp is the child vnode, and we
	* are looking up the parent vnode - exactly reverse from
	* normal operation. Unlocking dvp requires some rather
	* tricky unlock/relock dance to prevent mp from being freed;
	* use vn_vget_ino_gen() which takes care of all that.
	*
	* XXX Note that there is a time window when both vnodes are
	* unlocked. It is possible, although highly unlikely, that
	* during that window the parent-child relationship between
	* the vnodes may change, for example, get reversed.
	* In that case we would have a wrong lock order for the vnodes.
	* All other filesystems seem to ignore this problem, so we
	* do the same here.
	* A potential solution could be implemented as follows:
	* - using LK_NOWAIT when locking the second vnode and retrying
	* if necessary
	* - checking that the parent-child relationship still holds
	* after locking both vnodes and retrying if it doesn't
	*/
	error = vn_vget_ino_gen(dvp, zfs_dd_callback, vp, lkflags, &vp);
	return (error);
	} else {
	error = vn_lock(vp, lkflags);
	if (error != 0)
	vrele(vp);
	return (error);
	}
	}

	/*
	* Lookup an entry in a directory, or an extended attribute directory.
	* If it exists, return a held vnode reference for it.
	*
	* IN: dvp - vnode of directory to search.
	* nm - name of entry to lookup.
	* pnp - full pathname to lookup [UNUSED].
	* flags - LOOKUP_XATTR set if looking for an attribute.
	* rdir - root directory vnode [UNUSED].
	* cr - credentials of caller.
	* ct - caller context
	*
	* OUT: vpp - vnode of located entry, NULL if not found.
	*
	* RETURN: 0 on success, error code on failure.
	*
	* Timestamps:
	* NA
	*/
	static int
	zfs_lookup(vnode_t dvp, const char nm, vnode_t **vpp,
	struct componentname cnp, int nameiop, cred_t cr, int flags,
	boolean_t cached)
	{
	znode_t *zdp = VTOZ(dvp);
	znode_t *zp;
	zfsvfs_t *zfsvfs = zdp->z_zfsvfs;
	#if __FreeBSD_version > 1300124
	seqc_t dvp_seqc;
	#endif
	int error = 0;

	/*
	* Fast path lookup, however we must skip DNLC lookup
	* for case folding or normalizing lookups because the
	* DNLC code only stores the passed in name. This means
	* creating 'a' and removing 'A' on a case insensitive
	* file system would work, but DNLC still thinks 'a'
	* exists and won't let you create it again on the next
	* pass through fast path.
	*/
	if (!(flags & LOOKUP_XATTR)) {
	if (dvp->v_type != VDIR) {
	return (SET_ERROR(ENOTDIR));
	} else if (zdp->z_sa_hdl == NULL) {
	return (SET_ERROR(EIO));
	}
	}

	DTRACE_PROBE2(zfs__fastpath__lookup__miss, vnode_t *, dvp,
	const char *, nm);

	if ((error = zfs_enter_verify_zp(zfsvfs, zdp, FTAG)) != 0)
	return (error);

	#if __FreeBSD_version > 1300124
	dvp_seqc = vn_seqc_read_notmodify(dvp);
	#endif

	*vpp = NULL;

	if (flags & LOOKUP_XATTR) {
	/*
	* If the xattr property is off, refuse the lookup request.
	*/
	if (!(zfsvfs->z_flags & ZSB_XATTR)) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EOPNOTSUPP));
	}

	/*
	* We don't allow recursive attributes..
	* Maybe someday we will.
	*/
	if (zdp->z_pflags & ZFS_XATTR) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EINVAL));
	}

	if ((error = zfs_get_xattrdir(VTOZ(dvp), &zp, cr, flags))) {
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}
	*vpp = ZTOV(zp);

	/*
	* Do we have permission to get into attribute directory?
	*/
	error = zfs_zaccess(zp, ACE_EXECUTE, 0, B_FALSE, cr, NULL);
	if (error) {
	vrele(ZTOV(zp));
	}

	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	/*
	* Check accessibility of directory if we're not coming in via
	* VOP_CACHEDLOOKUP.
	*/
	if (!cached) {
	#ifdef NOEXECCHECK
	if ((cnp->cn_flags & NOEXECCHECK) != 0) {
	cnp->cn_flags &= ~NOEXECCHECK;
	} else
	#endif
	if ((error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr,
	NULL))) {
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}
	}

	if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm),
	NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EILSEQ));
	}


	/*
	* First handle the special cases.
	*/
	if ((cnp->cn_flags & ISDOTDOT) != 0) {
	/*
	* If we are a snapshot mounted under .zfs, return
	* the vp for the snapshot directory.
	*/
	if (zdp->z_id == zfsvfs->z_root && zfsvfs->z_parent != zfsvfs) {
	struct componentname cn;
	vnode_t *zfsctl_vp;
	int ltype;

	zfs_exit(zfsvfs, FTAG);
	ltype = VOP_ISLOCKED(dvp);
	VOP_UNLOCK1(dvp);
	error = zfsctl_root(zfsvfs->z_parent, LK_SHARED,
	&zfsctl_vp);
	if (error == 0) {
	cn.cn_nameptr = "snapshot";
	cn.cn_namelen = strlen(cn.cn_nameptr);
	cn.cn_nameiop = cnp->cn_nameiop;
	cn.cn_flags = cnp->cn_flags & ~ISDOTDOT;
	cn.cn_lkflags = cnp->cn_lkflags;
	error = VOP_LOOKUP(zfsctl_vp, vpp, &cn);
	vput(zfsctl_vp);
	}
	vn_lock(dvp, ltype \| LK_RETRY);
	return (error);
	}
	}
	if (zfs_has_ctldir(zdp) && strcmp(nm, ZFS_CTLDIR_NAME) == 0) {
	zfs_exit(zfsvfs, FTAG);
	if ((cnp->cn_flags & ISLASTCN) != 0 && nameiop != LOOKUP)
	return (SET_ERROR(ENOTSUP));
	error = zfsctl_root(zfsvfs, cnp->cn_lkflags, vpp);
	return (error);
	}

	/*
	* The loop is retry the lookup if the parent-child relationship
	* changes during the dot-dot locking complexities.
	*/
	for (;;) {
	uint64_t parent;

	error = zfs_dirlook(zdp, nm, &zp);
	if (error == 0)
	*vpp = ZTOV(zp);

	zfs_exit(zfsvfs, FTAG);
	if (error != 0)
	break;

	error = zfs_lookup_lock(dvp, *vpp, nm, cnp->cn_lkflags);
	if (error != 0) {
	/*
	* If we've got a locking error, then the vnode
	* got reclaimed because of a force unmount.
	* We never enter doomed vnodes into the name cache.
	*/
	*vpp = NULL;
	return (error);
	}

	if ((cnp->cn_flags & ISDOTDOT) == 0)
	break;

	if ((error = zfs_enter(zfsvfs, FTAG)) != 0) {
	vput(ZTOV(zp));
	*vpp = NULL;
	return (error);
	}
	if (zdp->z_sa_hdl == NULL) {
	error = SET_ERROR(EIO);
	} else {
	error = sa_lookup(zdp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
	&parent, sizeof (parent));
	}
	if (error != 0) {
	zfs_exit(zfsvfs, FTAG);
	vput(ZTOV(zp));
	break;
	}
	if (zp->z_id == parent) {
	zfs_exit(zfsvfs, FTAG);
	break;
	}
	vput(ZTOV(zp));
	}

	if (error != 0)
	*vpp = NULL;

	/* Translate errors and add SAVENAME when needed. */
	if (cnp->cn_flags & ISLASTCN) {
	switch (nameiop) {
	case CREATE:
	case RENAME:
	if (error == ENOENT) {
	error = EJUSTRETURN;
	#if __FreeBSD_version < 1400068
	cnp->cn_flags \|= SAVENAME;
	#endif
	break;
	}
	zfs_fallthrough;
	case DELETE:
	#if __FreeBSD_version < 1400068
	if (error == 0)
	cnp->cn_flags \|= SAVENAME;
	#endif
	break;
	}
	}

	#if __FreeBSD_version > 1300124
	if ((cnp->cn_flags & ISDOTDOT) != 0) {
	/*
	* FIXME: zfs_lookup_lock relocks vnodes and does nothing to
	* handle races. In particular different callers may end up
	* with different vnodes and will try to add conflicting
	* entries to the namecache.
	*
	* While finding different result may be acceptable in face
	* of concurrent modification, adding conflicting entries
	* trips over an assert in the namecache.
	*
	* Ultimately let an entry through once everything settles.
	*/
	if (!vn_seqc_consistent(dvp, dvp_seqc)) {
	cnp->cn_flags &= ~MAKEENTRY;
	}
	}
	#endif

	/* Insert name into cache (as non-existent) if appropriate. */
	if (zfsvfs->z_use_namecache && !zfsvfs->z_replay &&
	error == ENOENT && (cnp->cn_flags & MAKEENTRY) != 0)
	cache_enter(dvp, NULL, cnp);

	/* Insert name into cache if appropriate. */
	if (zfsvfs->z_use_namecache && !zfsvfs->z_replay &&
	error == 0 && (cnp->cn_flags & MAKEENTRY)) {
	if (!(cnp->cn_flags & ISLASTCN) \|\|
	(nameiop != DELETE && nameiop != RENAME)) {
	cache_enter(dvp, *vpp, cnp);
	}
	}

	return (error);
	}

	/*
	* Attempt to create a new entry in a directory. If the entry
	* already exists, truncate the file if permissible, else return
	* an error. Return the vp of the created or trunc'd file.
	*
	* IN: dvp - vnode of directory to put new file entry in.
	* name - name of new file entry.
	* vap - attributes of new file.
	* excl - flag indicating exclusive or non-exclusive mode.
	* mode - mode to open file with.
	* cr - credentials of caller.
	* flag - large file flag [UNUSED].
	* ct - caller context
	* vsecp - ACL to be set
	* mnt_ns - Unused on FreeBSD
	*
	* OUT: vpp - vnode of created or trunc'd entry.
	*
	* RETURN: 0 on success, error code on failure.
	*
	* Timestamps:
	* dvp - ctime\|mtime updated if new entry created
	* vp - ctime\|mtime always, atime if new
	*/
	int
	zfs_create(znode_t dzp, const char name, vattr_t *vap, int excl, int mode,
	znode_t *zpp, cred_t cr, int flag, vsecattr_t vsecp, zidmap_t mnt_ns)
	{
	(void) excl, (void) mode, (void) flag;
	znode_t *zp;
	zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
	zilog_t *zilog;
	objset_t *os;
	dmu_tx_t *tx;
	int error;
	uid_t uid = crgetuid(cr);
	gid_t gid = crgetgid(cr);
	uint64_t projid = ZFS_DEFAULT_PROJID;
	zfs_acl_ids_t acl_ids;
	boolean_t fuid_dirtied;
	uint64_t txtype;
	#ifdef DEBUG_VFS_LOCKS
	vnode_t *dvp = ZTOV(dzp);
	#endif

	/*
	* If we have an ephemeral id, ACL, or XVATTR then
	* make sure file system is at proper version
	*/
	if (zfsvfs->z_use_fuids == B_FALSE &&
	(vsecp \|\| (vap->va_mask & AT_XVATTR) \|\|
	IS_EPHEMERAL(uid) \|\| IS_EPHEMERAL(gid)))
	return (SET_ERROR(EINVAL));

	if ((error = zfs_enter_verify_zp(zfsvfs, dzp, FTAG)) != 0)
	return (error);
	os = zfsvfs->z_os;
	zilog = zfsvfs->z_log;

	if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
	NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EILSEQ));
	}

	if (vap->va_mask & AT_XVATTR) {
	if ((error = secpolicy_xvattr(ZTOV(dzp), (xvattr_t *)vap,
	crgetuid(cr), cr, vap->va_type)) != 0) {
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}
	}

	*zpp = NULL;

	if ((vap->va_mode & S_ISVTX) && secpolicy_vnode_stky_modify(cr))
	vap->va_mode &= ~S_ISVTX;

	error = zfs_dirent_lookup(dzp, name, &zp, ZNEW);
	if (error) {
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}
	ASSERT3P(zp, ==, NULL);

	/*
	* Create a new file object and update the directory
	* to reference it.
	*/
	if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr, mnt_ns))) {
	goto out;
	}

	/*
	* We only support the creation of regular files in
	* extended attribute directories.
	*/

	if ((dzp->z_pflags & ZFS_XATTR) &&
	(vap->va_type != VREG)) {
	error = SET_ERROR(EINVAL);
	goto out;
	}

	if ((error = zfs_acl_ids_create(dzp, 0, vap,
	cr, vsecp, &acl_ids, NULL)) != 0)
	goto out;

	if (S_ISREG(vap->va_mode) \|\| S_ISDIR(vap->va_mode))
	projid = zfs_inherit_projid(dzp);
	if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, projid)) {
	zfs_acl_ids_free(&acl_ids);
	error = SET_ERROR(EDQUOT);
	goto out;
	}

	getnewvnode_reserve_();

	tx = dmu_tx_create(os);

	dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
	ZFS_SA_BASE_ATTR_SIZE);

	fuid_dirtied = zfsvfs->z_fuid_dirty;
	if (fuid_dirtied)
	zfs_fuid_txhold(zfsvfs, tx);
	dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
	dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE);
	if (!zfsvfs->z_use_sa &&
	acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
	dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
	0, acl_ids.z_aclp->z_acl_bytes);
	}
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error) {
	zfs_acl_ids_free(&acl_ids);
	dmu_tx_abort(tx);
	getnewvnode_drop_reserve();
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}
	zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids);
	+
	+ error = zfs_link_create(dzp, name, zp, tx, ZNEW);
	+ if (error != 0) {
	+ /*
	+ * Since, we failed to add the directory entry for it,
	+ * delete the newly created dnode.
	+ */
	+ zfs_znode_delete(zp, tx);
	+ VOP_UNLOCK1(ZTOV(zp));
	+ zrele(zp);
	+ zfs_acl_ids_free(&acl_ids);
	+ dmu_tx_commit(tx);
	+ getnewvnode_drop_reserve();
	+ goto out;
	+ }
	+
	if (fuid_dirtied)
	zfs_fuid_sync(zfsvfs, tx);

	- (void) zfs_link_create(dzp, name, zp, tx, ZNEW);
	txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap);
	zfs_log_create(zilog, tx, txtype, dzp, zp, name,
	vsecp, acl_ids.z_fuidp, vap);
	zfs_acl_ids_free(&acl_ids);
	dmu_tx_commit(tx);

	getnewvnode_drop_reserve();

	out:
	VNCHECKREF(dvp);
	if (error == 0) {
	*zpp = zp;
	}

	if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
	zil_commit(zilog, 0);

	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	/*
	* Remove an entry from a directory.
	*
	* IN: dvp - vnode of directory to remove entry from.
	* name - name of entry to remove.
	* cr - credentials of caller.
	* ct - caller context
	* flags - case flags
	*
	* RETURN: 0 on success, error code on failure.
	*
	* Timestamps:
	* dvp - ctime\|mtime
	* vp - ctime (if nlink > 0)
	*/
	static int
	zfs_remove_(vnode_t dvp, vnode_t vp, const char name, cred_t cr)
	{
	znode_t *dzp = VTOZ(dvp);
	znode_t *zp;
	znode_t *xzp;
	zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
	zilog_t *zilog;
	uint64_t xattr_obj;
	uint64_t obj = 0;
	dmu_tx_t *tx;
	boolean_t unlinked;
	uint64_t txtype;
	int error;


	if ((error = zfs_enter_verify_zp(zfsvfs, dzp, FTAG)) != 0)
	return (error);
	zp = VTOZ(vp);
	if ((error = zfs_verify_zp(zp)) != 0) {
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}
	zilog = zfsvfs->z_log;

	xattr_obj = 0;
	xzp = NULL;

	if ((error = zfs_zaccess_delete(dzp, zp, cr, NULL))) {
	goto out;
	}

	/*
	* Need to use rmdir for removing directories.
	*/
	if (vp->v_type == VDIR) {
	error = SET_ERROR(EPERM);
	goto out;
	}

	vnevent_remove(vp, dvp, name, ct);

	obj = zp->z_id;

	/* are there any extended attributes? */
	error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
	&xattr_obj, sizeof (xattr_obj));
	if (error == 0 && xattr_obj) {
	error = zfs_zget(zfsvfs, xattr_obj, &xzp);
	ASSERT0(error);
	}

	/*
	* We may delete the znode now, or we may put it in the unlinked set;
	* it depends on whether we're the last link, and on whether there are
	* other holds on the vnode. So we dmu_tx_hold() the right things to
	* allow for either case.
	*/
	tx = dmu_tx_create(zfsvfs->z_os);
	dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
	zfs_sa_upgrade_txholds(tx, zp);
	zfs_sa_upgrade_txholds(tx, dzp);

	if (xzp) {
	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
	dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE);
	}

	/* charge as an update -- would be nice not to charge at all */
	dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);

	/*
	* Mark this transaction as typically resulting in a net free of space
	*/
	dmu_tx_mark_netfree(tx);

	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error) {
	dmu_tx_abort(tx);
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	/*
	* Remove the directory entry.
	*/
	error = zfs_link_destroy(dzp, name, zp, tx, ZEXISTS, &unlinked);

	if (error) {
	dmu_tx_commit(tx);
	goto out;
	}

	if (unlinked) {
	zfs_unlinked_add(zp, tx);
	vp->v_vflag \|= VV_NOSYNC;
	}
	/* XXX check changes to linux vnops */
	txtype = TX_REMOVE;
	zfs_log_remove(zilog, tx, txtype, dzp, name, obj, unlinked);

	dmu_tx_commit(tx);
	out:

	if (xzp)
	vrele(ZTOV(xzp));

	if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
	zil_commit(zilog, 0);


	zfs_exit(zfsvfs, FTAG);
	return (error);
	}


	static int
	zfs_lookup_internal(znode_t dzp, const char name, vnode_t **vpp,
	struct componentname *cnp, int nameiop)
	{
	zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
	int error;

	cnp->cn_nameptr = __DECONST(char *, name);
	cnp->cn_namelen = strlen(name);
	cnp->cn_nameiop = nameiop;
	cnp->cn_flags = ISLASTCN;
	#if __FreeBSD_version < 1400068
	cnp->cn_flags \|= SAVENAME;
	#endif
	cnp->cn_lkflags = LK_EXCLUSIVE \| LK_RETRY;
	cnp->cn_cred = kcred;
	#if __FreeBSD_version < 1400037
	cnp->cn_thread = curthread;
	#endif

	if (zfsvfs->z_use_namecache && !zfsvfs->z_replay) {
	struct vop_lookup_args a;

	a.a_gen.a_desc = &vop_lookup_desc;
	a.a_dvp = ZTOV(dzp);
	a.a_vpp = vpp;
	a.a_cnp = cnp;
	error = vfs_cache_lookup(&a);
	} else {
	error = zfs_lookup(ZTOV(dzp), name, vpp, cnp, nameiop, kcred, 0,
	B_FALSE);
	}
	#ifdef ZFS_DEBUG
	if (error) {
	printf("got error %d on name %s on op %d\n", error, name,
	nameiop);
	kdb_backtrace();
	}
	#endif
	return (error);
	}

	int
	zfs_remove(znode_t dzp, const char name, cred_t *cr, int flags)
	{
	vnode_t *vp;
	int error;
	struct componentname cn;

	if ((error = zfs_lookup_internal(dzp, name, &vp, &cn, DELETE)))
	return (error);

	error = zfs_remove_(ZTOV(dzp), vp, name, cr);
	vput(vp);
	return (error);
	}
	/*
	* Create a new directory and insert it into dvp using the name
	* provided. Return a pointer to the inserted directory.
	*
	* IN: dvp - vnode of directory to add subdir to.
	* dirname - name of new directory.
	* vap - attributes of new directory.
	* cr - credentials of caller.
	* ct - caller context
	* flags - case flags
	* vsecp - ACL to be set
	* mnt_ns - Unused on FreeBSD
	*
	* OUT: vpp - vnode of created directory.
	*
	* RETURN: 0 on success, error code on failure.
	*
	* Timestamps:
	* dvp - ctime\|mtime updated
	* vp - ctime\|mtime\|atime updated
	*/
	int
	zfs_mkdir(znode_t dzp, const char dirname, vattr_t vap, znode_t *zpp,
	cred_t cr, int flags, vsecattr_t vsecp, zidmap_t *mnt_ns)
	{
	(void) flags, (void) vsecp;
	znode_t *zp;
	zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
	zilog_t *zilog;
	uint64_t txtype;
	dmu_tx_t *tx;
	int error;
	uid_t uid = crgetuid(cr);
	gid_t gid = crgetgid(cr);
	zfs_acl_ids_t acl_ids;
	boolean_t fuid_dirtied;

	ASSERT3U(vap->va_type, ==, VDIR);

	/*
	* If we have an ephemeral id, ACL, or XVATTR then
	* make sure file system is at proper version
	*/
	if (zfsvfs->z_use_fuids == B_FALSE &&
	((vap->va_mask & AT_XVATTR) \|\|
	IS_EPHEMERAL(uid) \|\| IS_EPHEMERAL(gid)))
	return (SET_ERROR(EINVAL));

	if ((error = zfs_enter_verify_zp(zfsvfs, dzp, FTAG)) != 0)
	return (error);
	zilog = zfsvfs->z_log;

	if (dzp->z_pflags & ZFS_XATTR) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EINVAL));
	}

	if (zfsvfs->z_utf8 && u8_validate(dirname,
	strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EILSEQ));
	}

	if (vap->va_mask & AT_XVATTR) {
	if ((error = secpolicy_xvattr(ZTOV(dzp), (xvattr_t *)vap,
	crgetuid(cr), cr, vap->va_type)) != 0) {
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}
	}

	if ((error = zfs_acl_ids_create(dzp, 0, vap, cr,
	NULL, &acl_ids, NULL)) != 0) {
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	/*
	* First make sure the new directory doesn't exist.
	*
	* Existence is checked first to make sure we don't return
	* EACCES instead of EEXIST which can cause some applications
	* to fail.
	*/
	*zpp = NULL;

	if ((error = zfs_dirent_lookup(dzp, dirname, &zp, ZNEW))) {
	zfs_acl_ids_free(&acl_ids);
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}
	ASSERT3P(zp, ==, NULL);

	if ((error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr,
	mnt_ns))) {
	zfs_acl_ids_free(&acl_ids);
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, zfs_inherit_projid(dzp))) {
	zfs_acl_ids_free(&acl_ids);
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EDQUOT));
	}

	/*
	* Add a new entry to the directory.
	*/
	getnewvnode_reserve_();
	tx = dmu_tx_create(zfsvfs->z_os);
	dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname);
	dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
	fuid_dirtied = zfsvfs->z_fuid_dirty;
	if (fuid_dirtied)
	zfs_fuid_txhold(zfsvfs, tx);
	if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
	dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
	acl_ids.z_aclp->z_acl_bytes);
	}

	dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
	ZFS_SA_BASE_ATTR_SIZE);

	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error) {
	zfs_acl_ids_free(&acl_ids);
	dmu_tx_abort(tx);
	getnewvnode_drop_reserve();
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	/*
	* Create new node.
	*/
	zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids);

	- if (fuid_dirtied)
	- zfs_fuid_sync(zfsvfs, tx);
	-
	/*
	* Now put new name in parent dir.
	*/
	- (void) zfs_link_create(dzp, dirname, zp, tx, ZNEW);
	+ error = zfs_link_create(dzp, dirname, zp, tx, ZNEW);
	+ if (error != 0) {
	+ zfs_znode_delete(zp, tx);
	+ VOP_UNLOCK1(ZTOV(zp));
	+ zrele(zp);
	+ goto out;
	+ }
	+
	+ if (fuid_dirtied)
	+ zfs_fuid_sync(zfsvfs, tx);

	*zpp = zp;

	txtype = zfs_log_create_txtype(Z_DIR, NULL, vap);
	zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, NULL,
	acl_ids.z_fuidp, vap);

	+out:
	zfs_acl_ids_free(&acl_ids);

	dmu_tx_commit(tx);

	getnewvnode_drop_reserve();

	if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
	zil_commit(zilog, 0);

	zfs_exit(zfsvfs, FTAG);
	- return (0);
	+ return (error);
	}

	#if __FreeBSD_version < 1300124
	static void
	cache_vop_rmdir(struct vnode dvp, struct vnode vp)
	{

	cache_purge(dvp);
	cache_purge(vp);
	}
	#endif

	/*
	* Remove a directory subdir entry. If the current working
	* directory is the same as the subdir to be removed, the
	* remove will fail.
	*
	* IN: dvp - vnode of directory to remove from.
	* name - name of directory to be removed.
	* cwd - vnode of current working directory.
	* cr - credentials of caller.
	* ct - caller context
	* flags - case flags
	*
	* RETURN: 0 on success, error code on failure.
	*
	* Timestamps:
	* dvp - ctime\|mtime updated
	*/
	static int
	zfs_rmdir_(vnode_t dvp, vnode_t vp, const char name, cred_t cr)
	{
	znode_t *dzp = VTOZ(dvp);
	znode_t *zp = VTOZ(vp);
	zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
	zilog_t *zilog;
	dmu_tx_t *tx;
	int error;

	if ((error = zfs_enter_verify_zp(zfsvfs, dzp, FTAG)) != 0)
	return (error);
	if ((error = zfs_verify_zp(zp)) != 0) {
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}
	zilog = zfsvfs->z_log;


	if ((error = zfs_zaccess_delete(dzp, zp, cr, NULL))) {
	goto out;
	}

	if (vp->v_type != VDIR) {
	error = SET_ERROR(ENOTDIR);
	goto out;
	}

	vnevent_rmdir(vp, dvp, name, ct);

	tx = dmu_tx_create(zfsvfs->z_os);
	dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name);
	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
	dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
	zfs_sa_upgrade_txholds(tx, zp);
	zfs_sa_upgrade_txholds(tx, dzp);
	dmu_tx_mark_netfree(tx);
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error) {
	dmu_tx_abort(tx);
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	error = zfs_link_destroy(dzp, name, zp, tx, ZEXISTS, NULL);

	if (error == 0) {
	uint64_t txtype = TX_RMDIR;
	zfs_log_remove(zilog, tx, txtype, dzp, name,
	ZFS_NO_OBJECT, B_FALSE);
	}

	dmu_tx_commit(tx);

	if (zfsvfs->z_use_namecache)
	cache_vop_rmdir(dvp, vp);
	out:
	if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
	zil_commit(zilog, 0);

	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	int
	zfs_rmdir(znode_t dzp, const char name, znode_t cwd, cred_t cr, int flags)
	{
	struct componentname cn;
	vnode_t *vp;
	int error;

	if ((error = zfs_lookup_internal(dzp, name, &vp, &cn, DELETE)))
	return (error);

	error = zfs_rmdir_(ZTOV(dzp), vp, name, cr);
	vput(vp);
	return (error);
	}

	/*
	* Read as many directory entries as will fit into the provided
	* buffer from the given directory cursor position (specified in
	* the uio structure).
	*
	* IN: vp - vnode of directory to read.
	* uio - structure supplying read location, range info,
	* and return buffer.
	* cr - credentials of caller.
	* ct - caller context
	*
	* OUT: uio - updated offset and range, buffer filled.
	* eofp - set to true if end-of-file detected.
	* ncookies- number of entries in cookies
	* cookies - offsets to directory entries
	*
	* RETURN: 0 on success, error code on failure.
	*
	* Timestamps:
	* vp - atime updated
	*
	* Note that the low 4 bits of the cookie returned by zap is always zero.
	* This allows us to use the low range for "special" directory entries:
	* We use 0 for '.', and 1 for '..'. If this is the root of the filesystem,
	* we use the offset 2 for the '.zfs' directory.
	*/
	static int
	zfs_readdir(vnode_t vp, zfs_uio_t uio, cred_t cr, int eofp,
	int ncookies, cookie_t *cookies)
	{
	znode_t *zp = VTOZ(vp);
	iovec_t *iovp;
	dirent64_t *odp;
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	objset_t *os;
	caddr_t outbuf;
	size_t bufsize;
	zap_cursor_t zc;
	zap_attribute_t zap;
	uint_t bytes_wanted;
	uint64_t offset; /* must be unsigned; checks for < 1 */
	uint64_t parent;
	int local_eof;
	int outcount;
	int error;
	uint8_t prefetch;
	uint8_t type;
	int ncooks;
	cookie_t *cooks = NULL;

	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (error);

	if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
	&parent, sizeof (parent))) != 0) {
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	/*
	* If we are not given an eof variable,
	* use a local one.
	*/
	if (eofp == NULL)
	eofp = &local_eof;

	/*
	* Check for valid iov_len.
	*/
	if (GET_UIO_STRUCT(uio)->uio_iov->iov_len <= 0) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EINVAL));
	}

	/*
	* Quit if directory has been removed (posix)
	*/
	if ((*eofp = zp->z_unlinked) != 0) {
	zfs_exit(zfsvfs, FTAG);
	return (0);
	}

	error = 0;
	os = zfsvfs->z_os;
	offset = zfs_uio_offset(uio);
	prefetch = zp->z_zn_prefetch;

	/*
	* Initialize the iterator cursor.
	*/
	if (offset <= 3) {
	/*
	* Start iteration from the beginning of the directory.
	*/
	zap_cursor_init(&zc, os, zp->z_id);
	} else {
	/*
	* The offset is a serialized cursor.
	*/
	zap_cursor_init_serialized(&zc, os, zp->z_id, offset);
	}

	/*
	* Get space to change directory entries into fs independent format.
	*/
	iovp = GET_UIO_STRUCT(uio)->uio_iov;
	bytes_wanted = iovp->iov_len;
	if (zfs_uio_segflg(uio) != UIO_SYSSPACE \|\| zfs_uio_iovcnt(uio) != 1) {
	bufsize = bytes_wanted;
	outbuf = kmem_alloc(bufsize, KM_SLEEP);
	odp = (struct dirent64 *)outbuf;
	} else {
	bufsize = bytes_wanted;
	outbuf = NULL;
	odp = (struct dirent64 *)iovp->iov_base;
	}

	if (ncookies != NULL) {
	/*
	* Minimum entry size is dirent size and 1 byte for a file name.
	*/
	ncooks = zfs_uio_resid(uio) / (sizeof (struct dirent) -
	sizeof (((struct dirent *)NULL)->d_name) + 1);
	cooks = malloc(ncooks * sizeof (*cooks), M_TEMP, M_WAITOK);
	*cookies = cooks;
	*ncookies = ncooks;
	}

	/*
	* Transform to file-system independent format
	*/
	outcount = 0;
	while (outcount < bytes_wanted) {
	ino64_t objnum;
	ushort_t reclen;
	off64_t *next = NULL;

	/*
	* Special case `.', `..', and `.zfs'.
	*/
	if (offset == 0) {
	(void) strcpy(zap.za_name, ".");
	zap.za_normalization_conflict = 0;
	objnum = zp->z_id;
	type = DT_DIR;
	} else if (offset == 1) {
	(void) strcpy(zap.za_name, "..");
	zap.za_normalization_conflict = 0;
	objnum = parent;
	type = DT_DIR;
	} else if (offset == 2 && zfs_show_ctldir(zp)) {
	(void) strcpy(zap.za_name, ZFS_CTLDIR_NAME);
	zap.za_normalization_conflict = 0;
	objnum = ZFSCTL_INO_ROOT;
	type = DT_DIR;
	} else {
	/*
	* Grab next entry.
	*/
	if ((error = zap_cursor_retrieve(&zc, &zap))) {
	if ((*eofp = (error == ENOENT)) != 0)
	break;
	else
	goto update;
	}

	if (zap.za_integer_length != 8 \|\|
	zap.za_num_integers != 1) {
	cmn_err(CE_WARN, "zap_readdir: bad directory "
	"entry, obj = %lld, offset = %lld\n",
	(u_longlong_t)zp->z_id,
	(u_longlong_t)offset);
	error = SET_ERROR(ENXIO);
	goto update;
	}

	objnum = ZFS_DIRENT_OBJ(zap.za_first_integer);
	/*
	* MacOS X can extract the object type here such as:
	* uint8_t type = ZFS_DIRENT_TYPE(zap.za_first_integer);
	*/
	type = ZFS_DIRENT_TYPE(zap.za_first_integer);
	}

	reclen = DIRENT64_RECLEN(strlen(zap.za_name));

	/*
	* Will this entry fit in the buffer?
	*/
	if (outcount + reclen > bufsize) {
	/*
	* Did we manage to fit anything in the buffer?
	*/
	if (!outcount) {
	error = SET_ERROR(EINVAL);
	goto update;
	}
	break;
	}
	/*
	* Add normal entry:
	*/
	odp->d_ino = objnum;
	odp->d_reclen = reclen;
	odp->d_namlen = strlen(zap.za_name);
	/* NOTE: d_off is the offset for the next entry. */
	next = &odp->d_off;
	strlcpy(odp->d_name, zap.za_name, odp->d_namlen + 1);
	odp->d_type = type;
	dirent_terminate(odp);
	odp = (dirent64_t *)((intptr_t)odp + reclen);

	outcount += reclen;

	ASSERT3S(outcount, <=, bufsize);

	if (prefetch)
	dmu_prefetch_dnode(os, objnum, ZIO_PRIORITY_SYNC_READ);

	/*
	* Move to the next entry, fill in the previous offset.
	*/
	if (offset > 2 \|\| (offset == 2 && !zfs_show_ctldir(zp))) {
	zap_cursor_advance(&zc);
	offset = zap_cursor_serialize(&zc);
	} else {
	offset += 1;
	}

	/* Fill the offset right after advancing the cursor. */
	if (next != NULL)
	*next = offset;
	if (cooks != NULL) {
	*cooks++ = offset;
	ncooks--;
	KASSERT(ncooks >= 0, ("ncookies=%d", ncooks));
	}
	}
	zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */

	/* Subtract unused cookies */
	if (ncookies != NULL)
	*ncookies -= ncooks;

	if (zfs_uio_segflg(uio) == UIO_SYSSPACE && zfs_uio_iovcnt(uio) == 1) {
	iovp->iov_base += outcount;
	iovp->iov_len -= outcount;
	zfs_uio_resid(uio) -= outcount;
	} else if ((error =
	zfs_uiomove(outbuf, (long)outcount, UIO_READ, uio))) {
	/*
	* Reset the pointer.
	*/
	offset = zfs_uio_offset(uio);
	}

	update:
	zap_cursor_fini(&zc);
	if (zfs_uio_segflg(uio) != UIO_SYSSPACE \|\| zfs_uio_iovcnt(uio) != 1)
	kmem_free(outbuf, bufsize);

	if (error == ENOENT)
	error = 0;

	ZFS_ACCESSTIME_STAMP(zfsvfs, zp);

	zfs_uio_setoffset(uio, offset);
	zfs_exit(zfsvfs, FTAG);
	if (error != 0 && cookies != NULL) {
	free(*cookies, M_TEMP);
	*cookies = NULL;
	*ncookies = 0;
	}
	return (error);
	}

	/*
	* Get the requested file attributes and place them in the provided
	* vattr structure.
	*
	* IN: vp - vnode of file.
	* vap - va_mask identifies requested attributes.
	* If AT_XVATTR set, then optional attrs are requested
	* flags - ATTR_NOACLCHECK (CIFS server context)
	* cr - credentials of caller.
	*
	* OUT: vap - attribute values.
	*
	* RETURN: 0 (always succeeds).
	*/
	static int
	zfs_getattr(vnode_t vp, vattr_t vap, int flags, cred_t *cr)
	{
	znode_t *zp = VTOZ(vp);
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	int error = 0;
	uint32_t blksize;
	u_longlong_t nblocks;
	uint64_t mtime[2], ctime[2], crtime[2], rdev;
	xvattr_t xvap = (xvattr_t )vap; /* vap may be an xvattr_t * */
	xoptattr_t *xoap = NULL;
	boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
	sa_bulk_attr_t bulk[4];
	int count = 0;

	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (error);

	zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid);

	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CRTIME(zfsvfs), NULL, &crtime, 16);
	if (vp->v_type == VBLK \|\| vp->v_type == VCHR)
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zfsvfs), NULL,
	&rdev, 8);

	if ((error = sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) != 0) {
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	/*
	* If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES.
	* Also, if we are the owner don't bother, since owner should
	* always be allowed to read basic attributes of file.
	*/
	if (!(zp->z_pflags & ZFS_ACL_TRIVIAL) &&
	(vap->va_uid != crgetuid(cr))) {
	if ((error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0,
	skipaclchk, cr, NULL))) {
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}
	}

	/*
	* Return all attributes. It's cheaper to provide the answer
	* than to determine whether we were asked the question.
	*/

	vap->va_type = IFTOVT(zp->z_mode);
	vap->va_mode = zp->z_mode & ~S_IFMT;
	vn_fsid(vp, vap);
	vap->va_nodeid = zp->z_id;
	vap->va_nlink = zp->z_links;
	if ((vp->v_flag & VROOT) && zfs_show_ctldir(zp) &&
	zp->z_links < ZFS_LINK_MAX)
	vap->va_nlink++;
	vap->va_size = zp->z_size;
	if (vp->v_type == VBLK \|\| vp->v_type == VCHR)
	vap->va_rdev = zfs_cmpldev(rdev);
	else
	vap->va_rdev = 0;
	vap->va_gen = zp->z_gen;
	vap->va_flags = 0; /* FreeBSD: Reset chflags(2) flags. */
	vap->va_filerev = zp->z_seq;

	/*
	* Add in any requested optional attributes and the create time.
	* Also set the corresponding bits in the returned attribute bitmap.
	*/
	if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) {
	if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
	xoap->xoa_archive =
	((zp->z_pflags & ZFS_ARCHIVE) != 0);
	XVA_SET_RTN(xvap, XAT_ARCHIVE);
	}

	if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
	xoap->xoa_readonly =
	((zp->z_pflags & ZFS_READONLY) != 0);
	XVA_SET_RTN(xvap, XAT_READONLY);
	}

	if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
	xoap->xoa_system =
	((zp->z_pflags & ZFS_SYSTEM) != 0);
	XVA_SET_RTN(xvap, XAT_SYSTEM);
	}

	if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
	xoap->xoa_hidden =
	((zp->z_pflags & ZFS_HIDDEN) != 0);
	XVA_SET_RTN(xvap, XAT_HIDDEN);
	}

	if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
	xoap->xoa_nounlink =
	((zp->z_pflags & ZFS_NOUNLINK) != 0);
	XVA_SET_RTN(xvap, XAT_NOUNLINK);
	}

	if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
	xoap->xoa_immutable =
	((zp->z_pflags & ZFS_IMMUTABLE) != 0);
	XVA_SET_RTN(xvap, XAT_IMMUTABLE);
	}

	if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
	xoap->xoa_appendonly =
	((zp->z_pflags & ZFS_APPENDONLY) != 0);
	XVA_SET_RTN(xvap, XAT_APPENDONLY);
	}

	if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
	xoap->xoa_nodump =
	((zp->z_pflags & ZFS_NODUMP) != 0);
	XVA_SET_RTN(xvap, XAT_NODUMP);
	}

	if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
	xoap->xoa_opaque =
	((zp->z_pflags & ZFS_OPAQUE) != 0);
	XVA_SET_RTN(xvap, XAT_OPAQUE);
	}

	if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
	xoap->xoa_av_quarantined =
	((zp->z_pflags & ZFS_AV_QUARANTINED) != 0);
	XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
	}

	if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
	xoap->xoa_av_modified =
	((zp->z_pflags & ZFS_AV_MODIFIED) != 0);
	XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
	}

	if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) &&
	vp->v_type == VREG) {
	zfs_sa_get_scanstamp(zp, xvap);
	}

	if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
	xoap->xoa_reparse = ((zp->z_pflags & ZFS_REPARSE) != 0);
	XVA_SET_RTN(xvap, XAT_REPARSE);
	}
	if (XVA_ISSET_REQ(xvap, XAT_GEN)) {
	xoap->xoa_generation = zp->z_gen;
	XVA_SET_RTN(xvap, XAT_GEN);
	}

	if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
	xoap->xoa_offline =
	((zp->z_pflags & ZFS_OFFLINE) != 0);
	XVA_SET_RTN(xvap, XAT_OFFLINE);
	}

	if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
	xoap->xoa_sparse =
	((zp->z_pflags & ZFS_SPARSE) != 0);
	XVA_SET_RTN(xvap, XAT_SPARSE);
	}

	if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) {
	xoap->xoa_projinherit =
	((zp->z_pflags & ZFS_PROJINHERIT) != 0);
	XVA_SET_RTN(xvap, XAT_PROJINHERIT);
	}

	if (XVA_ISSET_REQ(xvap, XAT_PROJID)) {
	xoap->xoa_projid = zp->z_projid;
	XVA_SET_RTN(xvap, XAT_PROJID);
	}
	}

	ZFS_TIME_DECODE(&vap->va_atime, zp->z_atime);
	ZFS_TIME_DECODE(&vap->va_mtime, mtime);
	ZFS_TIME_DECODE(&vap->va_ctime, ctime);
	ZFS_TIME_DECODE(&vap->va_birthtime, crtime);


	sa_object_size(zp->z_sa_hdl, &blksize, &nblocks);
	vap->va_blksize = blksize;
	vap->va_bytes = nblocks << 9; /* nblocks * 512 */

	if (zp->z_blksz == 0) {
	/*
	* Block size hasn't been set; suggest maximal I/O transfers.
	*/
	vap->va_blksize = zfsvfs->z_max_blksz;
	}

	zfs_exit(zfsvfs, FTAG);
	return (0);
	}

	/*
	* Set the file attributes to the values contained in the
	* vattr structure.
	*
	* IN: zp - znode of file to be modified.
	* vap - new attribute values.
	* If AT_XVATTR set, then optional attrs are being set
	* flags - ATTR_UTIME set if non-default time values provided.
	* - ATTR_NOACLCHECK (CIFS context only).
	* cr - credentials of caller.
	* mnt_ns - Unused on FreeBSD
	*
	* RETURN: 0 on success, error code on failure.
	*
	* Timestamps:
	* vp - ctime updated, mtime updated if size changed.
	*/
	int
	zfs_setattr(znode_t zp, vattr_t vap, int flags, cred_t cr, zidmap_t mnt_ns)
	{
	vnode_t *vp = ZTOV(zp);
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	objset_t *os;
	zilog_t *zilog;
	dmu_tx_t *tx;
	vattr_t oldva;
	xvattr_t tmpxvattr;
	uint_t mask = vap->va_mask;
	uint_t saved_mask = 0;
	uint64_t saved_mode;
	int trim_mask = 0;
	uint64_t new_mode;
	uint64_t new_uid, new_gid;
	uint64_t xattr_obj;
	uint64_t mtime[2], ctime[2];
	uint64_t projid = ZFS_INVALID_PROJID;
	znode_t *attrzp;
	int need_policy = FALSE;
	int err, err2;
	zfs_fuid_info_t *fuidp = NULL;
	xvattr_t xvap = (xvattr_t )vap; /* vap may be an xvattr_t * */
	xoptattr_t *xoap;
	zfs_acl_t *aclp;
	boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
	boolean_t fuid_dirtied = B_FALSE;
	sa_bulk_attr_t bulk[7], xattr_bulk[7];
	int count = 0, xattr_count = 0;

	if (mask == 0)
	return (0);

	if (mask & AT_NOSET)
	return (SET_ERROR(EINVAL));

	if ((err = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (err);

	os = zfsvfs->z_os;
	zilog = zfsvfs->z_log;

	/*
	* Make sure that if we have ephemeral uid/gid or xvattr specified
	* that file system is at proper version level
	*/

	if (zfsvfs->z_use_fuids == B_FALSE &&
	(((mask & AT_UID) && IS_EPHEMERAL(vap->va_uid)) \|\|
	((mask & AT_GID) && IS_EPHEMERAL(vap->va_gid)) \|\|
	(mask & AT_XVATTR))) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EINVAL));
	}

	if (mask & AT_SIZE && vp->v_type == VDIR) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EISDIR));
	}

	if (mask & AT_SIZE && vp->v_type != VREG && vp->v_type != VFIFO) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EINVAL));
	}

	/*
	* If this is an xvattr_t, then get a pointer to the structure of
	* optional attributes. If this is NULL, then we have a vattr_t.
	*/
	xoap = xva_getxoptattr(xvap);

	xva_init(&tmpxvattr);

	/*
	* Immutable files can only alter immutable bit and atime
	*/
	if ((zp->z_pflags & ZFS_IMMUTABLE) &&
	((mask & (AT_SIZE\|AT_UID\|AT_GID\|AT_MTIME\|AT_MODE)) \|\|
	((mask & AT_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EPERM));
	}

	/*
	* Note: ZFS_READONLY is handled in zfs_zaccess_common.
	*/

	/*
	* Verify timestamps doesn't overflow 32 bits.
	* ZFS can handle large timestamps, but 32bit syscalls can't
	* handle times greater than 2039. This check should be removed
	* once large timestamps are fully supported.
	*/
	if (mask & (AT_ATIME \| AT_MTIME)) {
	if (((mask & AT_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) \|\|
	((mask & AT_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EOVERFLOW));
	}
	}
	if (xoap != NULL && (mask & AT_XVATTR)) {
	if (XVA_ISSET_REQ(xvap, XAT_CREATETIME) &&
	TIMESPEC_OVERFLOW(&vap->va_birthtime)) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EOVERFLOW));
	}

	if (XVA_ISSET_REQ(xvap, XAT_PROJID)) {
	if (!dmu_objset_projectquota_enabled(os) \|\|
	(!S_ISREG(zp->z_mode) && !S_ISDIR(zp->z_mode))) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EOPNOTSUPP));
	}

	projid = xoap->xoa_projid;
	if (unlikely(projid == ZFS_INVALID_PROJID)) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EINVAL));
	}

	if (projid == zp->z_projid && zp->z_pflags & ZFS_PROJID)
	projid = ZFS_INVALID_PROJID;
	else
	need_policy = TRUE;
	}

	if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT) &&
	(xoap->xoa_projinherit !=
	((zp->z_pflags & ZFS_PROJINHERIT) != 0)) &&
	(!dmu_objset_projectquota_enabled(os) \|\|
	(!S_ISREG(zp->z_mode) && !S_ISDIR(zp->z_mode)))) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EOPNOTSUPP));
	}
	}

	attrzp = NULL;
	aclp = NULL;

	if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EROFS));
	}

	/*
	* First validate permissions
	*/

	if (mask & AT_SIZE) {
	/*
	* XXX - Note, we are not providing any open
	* mode flags here (like FNDELAY), so we may
	* block if there are locks present... this
	* should be addressed in openat().
	*/
	/* XXX - would it be OK to generate a log record here? */
	err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE);
	if (err) {
	zfs_exit(zfsvfs, FTAG);
	return (err);
	}
	}

	if (mask & (AT_ATIME\|AT_MTIME) \|\|
	((mask & AT_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) \|\|
	XVA_ISSET_REQ(xvap, XAT_READONLY) \|\|
	XVA_ISSET_REQ(xvap, XAT_ARCHIVE) \|\|
	XVA_ISSET_REQ(xvap, XAT_OFFLINE) \|\|
	XVA_ISSET_REQ(xvap, XAT_SPARSE) \|\|
	XVA_ISSET_REQ(xvap, XAT_CREATETIME) \|\|
	XVA_ISSET_REQ(xvap, XAT_SYSTEM)))) {
	need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0,
	skipaclchk, cr, mnt_ns);
	}

	if (mask & (AT_UID\|AT_GID)) {
	int idmask = (mask & (AT_UID\|AT_GID));
	int take_owner;
	int take_group;

	/*
	* NOTE: even if a new mode is being set,
	* we may clear S_ISUID/S_ISGID bits.
	*/

	if (!(mask & AT_MODE))
	vap->va_mode = zp->z_mode;

	/*
	* Take ownership or chgrp to group we are a member of
	*/

	take_owner = (mask & AT_UID) && (vap->va_uid == crgetuid(cr));
	take_group = (mask & AT_GID) &&
	zfs_groupmember(zfsvfs, vap->va_gid, cr);

	/*
	* If both AT_UID and AT_GID are set then take_owner and
	* take_group must both be set in order to allow taking
	* ownership.
	*
	* Otherwise, send the check through secpolicy_vnode_setattr()
	*
	*/

	if (((idmask == (AT_UID\|AT_GID)) && take_owner && take_group) \|\|
	((idmask == AT_UID) && take_owner) \|\|
	((idmask == AT_GID) && take_group)) {
	if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0,
	skipaclchk, cr, mnt_ns) == 0) {
	/*
	* Remove setuid/setgid for non-privileged users
	*/
	secpolicy_setid_clear(vap, vp, cr);
	trim_mask = (mask & (AT_UID\|AT_GID));
	} else {
	need_policy = TRUE;
	}
	} else {
	need_policy = TRUE;
	}
	}

	oldva.va_mode = zp->z_mode;
	zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid);
	if (mask & AT_XVATTR) {
	/*
	* Update xvattr mask to include only those attributes
	* that are actually changing.
	*
	* the bits will be restored prior to actually setting
	* the attributes so the caller thinks they were set.
	*/
	if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
	if (xoap->xoa_appendonly !=
	((zp->z_pflags & ZFS_APPENDONLY) != 0)) {
	need_policy = TRUE;
	} else {
	XVA_CLR_REQ(xvap, XAT_APPENDONLY);
	XVA_SET_REQ(&tmpxvattr, XAT_APPENDONLY);
	}
	}

	if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) {
	if (xoap->xoa_projinherit !=
	((zp->z_pflags & ZFS_PROJINHERIT) != 0)) {
	need_policy = TRUE;
	} else {
	XVA_CLR_REQ(xvap, XAT_PROJINHERIT);
	XVA_SET_REQ(&tmpxvattr, XAT_PROJINHERIT);
	}
	}

	if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
	if (xoap->xoa_nounlink !=
	((zp->z_pflags & ZFS_NOUNLINK) != 0)) {
	need_policy = TRUE;
	} else {
	XVA_CLR_REQ(xvap, XAT_NOUNLINK);
	XVA_SET_REQ(&tmpxvattr, XAT_NOUNLINK);
	}
	}

	if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
	if (xoap->xoa_immutable !=
	((zp->z_pflags & ZFS_IMMUTABLE) != 0)) {
	need_policy = TRUE;
	} else {
	XVA_CLR_REQ(xvap, XAT_IMMUTABLE);
	XVA_SET_REQ(&tmpxvattr, XAT_IMMUTABLE);
	}
	}

	if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
	if (xoap->xoa_nodump !=
	((zp->z_pflags & ZFS_NODUMP) != 0)) {
	need_policy = TRUE;
	} else {
	XVA_CLR_REQ(xvap, XAT_NODUMP);
	XVA_SET_REQ(&tmpxvattr, XAT_NODUMP);
	}
	}

	if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
	if (xoap->xoa_av_modified !=
	((zp->z_pflags & ZFS_AV_MODIFIED) != 0)) {
	need_policy = TRUE;
	} else {
	XVA_CLR_REQ(xvap, XAT_AV_MODIFIED);
	XVA_SET_REQ(&tmpxvattr, XAT_AV_MODIFIED);
	}
	}

	if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
	if ((vp->v_type != VREG &&
	xoap->xoa_av_quarantined) \|\|
	xoap->xoa_av_quarantined !=
	((zp->z_pflags & ZFS_AV_QUARANTINED) != 0)) {
	need_policy = TRUE;
	} else {
	XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED);
	XVA_SET_REQ(&tmpxvattr, XAT_AV_QUARANTINED);
	}
	}

	if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EPERM));
	}

	if (need_policy == FALSE &&
	(XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) \|\|
	XVA_ISSET_REQ(xvap, XAT_OPAQUE))) {
	need_policy = TRUE;
	}
	}

	if (mask & AT_MODE) {
	if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr,
	mnt_ns) == 0) {
	err = secpolicy_setid_setsticky_clear(vp, vap,
	&oldva, cr);
	if (err) {
	zfs_exit(zfsvfs, FTAG);
	return (err);
	}
	trim_mask \|= AT_MODE;
	} else {
	need_policy = TRUE;
	}
	}

	if (need_policy) {
	/*
	* If trim_mask is set then take ownership
	* has been granted or write_acl is present and user
	* has the ability to modify mode. In that case remove
	* UID\|GID and or MODE from mask so that
	* secpolicy_vnode_setattr() doesn't revoke it.
	*/

	if (trim_mask) {
	saved_mask = vap->va_mask;
	vap->va_mask &= ~trim_mask;
	if (trim_mask & AT_MODE) {
	/*
	* Save the mode, as secpolicy_vnode_setattr()
	* will overwrite it with ova.va_mode.
	*/
	saved_mode = vap->va_mode;
	}
	}
	err = secpolicy_vnode_setattr(cr, vp, vap, &oldva, flags,
	(int ()(void , int, cred_t *))zfs_zaccess_unix, zp);
	if (err) {
	zfs_exit(zfsvfs, FTAG);
	return (err);
	}

	if (trim_mask) {
	vap->va_mask \|= saved_mask;
	if (trim_mask & AT_MODE) {
	/*
	* Recover the mode after
	* secpolicy_vnode_setattr().
	*/
	vap->va_mode = saved_mode;
	}
	}
	}

	/*
	* secpolicy_vnode_setattr, or take ownership may have
	* changed va_mask
	*/
	mask = vap->va_mask;

	if ((mask & (AT_UID \| AT_GID)) \|\| projid != ZFS_INVALID_PROJID) {
	err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
	&xattr_obj, sizeof (xattr_obj));

	if (err == 0 && xattr_obj) {
	err = zfs_zget(zp->z_zfsvfs, xattr_obj, &attrzp);
	if (err == 0) {
	err = vn_lock(ZTOV(attrzp), LK_EXCLUSIVE);
	if (err != 0)
	vrele(ZTOV(attrzp));
	}
	if (err)
	goto out2;
	}
	if (mask & AT_UID) {
	new_uid = zfs_fuid_create(zfsvfs,
	(uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp);
	if (new_uid != zp->z_uid &&
	zfs_id_overquota(zfsvfs, DMU_USERUSED_OBJECT,
	new_uid)) {
	if (attrzp)
	vput(ZTOV(attrzp));
	err = SET_ERROR(EDQUOT);
	goto out2;
	}
	}

	if (mask & AT_GID) {
	new_gid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_gid,
	cr, ZFS_GROUP, &fuidp);
	if (new_gid != zp->z_gid &&
	zfs_id_overquota(zfsvfs, DMU_GROUPUSED_OBJECT,
	new_gid)) {
	if (attrzp)
	vput(ZTOV(attrzp));
	err = SET_ERROR(EDQUOT);
	goto out2;
	}
	}

	if (projid != ZFS_INVALID_PROJID &&
	zfs_id_overquota(zfsvfs, DMU_PROJECTUSED_OBJECT, projid)) {
	if (attrzp)
	vput(ZTOV(attrzp));
	err = SET_ERROR(EDQUOT);
	goto out2;
	}
	}
	tx = dmu_tx_create(os);

	if (mask & AT_MODE) {
	uint64_t pmode = zp->z_mode;
	uint64_t acl_obj;
	new_mode = (pmode & S_IFMT) \| (vap->va_mode & ~S_IFMT);

	if (zp->z_zfsvfs->z_acl_mode == ZFS_ACL_RESTRICTED &&
	!(zp->z_pflags & ZFS_ACL_TRIVIAL)) {
	err = SET_ERROR(EPERM);
	goto out;
	}

	if ((err = zfs_acl_chmod_setattr(zp, &aclp, new_mode)))
	goto out;

	if (!zp->z_is_sa && ((acl_obj = zfs_external_acl(zp)) != 0)) {
	/*
	* Are we upgrading ACL from old V0 format
	* to V1 format?
	*/
	if (zfsvfs->z_version >= ZPL_VERSION_FUID &&
	zfs_znode_acl_version(zp) ==
	ZFS_ACL_VERSION_INITIAL) {
	dmu_tx_hold_free(tx, acl_obj, 0,
	DMU_OBJECT_END);
	dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
	0, aclp->z_acl_bytes);
	} else {
	dmu_tx_hold_write(tx, acl_obj, 0,
	aclp->z_acl_bytes);
	}
	} else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) {
	dmu_tx_hold_write(tx, DMU_NEW_OBJECT,
	0, aclp->z_acl_bytes);
	}
	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
	} else {
	if (((mask & AT_XVATTR) &&
	XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) \|\|
	(projid != ZFS_INVALID_PROJID &&
	!(zp->z_pflags & ZFS_PROJID)))
	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
	else
	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
	}

	if (attrzp) {
	dmu_tx_hold_sa(tx, attrzp->z_sa_hdl, B_FALSE);
	}

	fuid_dirtied = zfsvfs->z_fuid_dirty;
	if (fuid_dirtied)
	zfs_fuid_txhold(zfsvfs, tx);

	zfs_sa_upgrade_txholds(tx, zp);

	err = dmu_tx_assign(tx, TXG_WAIT);
	if (err)
	goto out;

	count = 0;
	/*
	* Set each attribute requested.
	* We group settings according to the locks they need to acquire.
	*
	* Note: you cannot set ctime directly, although it will be
	* updated as a side-effect of calling this function.
	*/

	if (projid != ZFS_INVALID_PROJID && !(zp->z_pflags & ZFS_PROJID)) {
	/*
	* For the existed object that is upgraded from old system,
	* its on-disk layout has no slot for the project ID attribute.
	* But quota accounting logic needs to access related slots by
	* offset directly. So we need to adjust old objects' layout
	* to make the project ID to some unified and fixed offset.
	*/
	if (attrzp)
	err = sa_add_projid(attrzp->z_sa_hdl, tx, projid);
	if (err == 0)
	err = sa_add_projid(zp->z_sa_hdl, tx, projid);

	if (unlikely(err == EEXIST))
	err = 0;
	else if (err != 0)
	goto out;
	else
	projid = ZFS_INVALID_PROJID;
	}

	if (mask & (AT_UID\|AT_GID\|AT_MODE))
	mutex_enter(&zp->z_acl_lock);

	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
	&zp->z_pflags, sizeof (zp->z_pflags));

	if (attrzp) {
	if (mask & (AT_UID\|AT_GID\|AT_MODE))
	mutex_enter(&attrzp->z_acl_lock);
	SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
	SA_ZPL_FLAGS(zfsvfs), NULL, &attrzp->z_pflags,
	sizeof (attrzp->z_pflags));
	if (projid != ZFS_INVALID_PROJID) {
	attrzp->z_projid = projid;
	SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
	SA_ZPL_PROJID(zfsvfs), NULL, &attrzp->z_projid,
	sizeof (attrzp->z_projid));
	}
	}

	if (mask & (AT_UID\|AT_GID)) {

	if (mask & AT_UID) {
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
	&new_uid, sizeof (new_uid));
	zp->z_uid = new_uid;
	if (attrzp) {
	SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
	SA_ZPL_UID(zfsvfs), NULL, &new_uid,
	sizeof (new_uid));
	attrzp->z_uid = new_uid;
	}
	}

	if (mask & AT_GID) {
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs),
	NULL, &new_gid, sizeof (new_gid));
	zp->z_gid = new_gid;
	if (attrzp) {
	SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
	SA_ZPL_GID(zfsvfs), NULL, &new_gid,
	sizeof (new_gid));
	attrzp->z_gid = new_gid;
	}
	}
	if (!(mask & AT_MODE)) {
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs),
	NULL, &new_mode, sizeof (new_mode));
	new_mode = zp->z_mode;
	}
	err = zfs_acl_chown_setattr(zp);
	ASSERT0(err);
	if (attrzp) {
	vn_seqc_write_begin(ZTOV(attrzp));
	err = zfs_acl_chown_setattr(attrzp);
	vn_seqc_write_end(ZTOV(attrzp));
	ASSERT0(err);
	}
	}

	if (mask & AT_MODE) {
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
	&new_mode, sizeof (new_mode));
	zp->z_mode = new_mode;
	ASSERT3P(aclp, !=, NULL);
	err = zfs_aclset_common(zp, aclp, cr, tx);
	ASSERT0(err);
	if (zp->z_acl_cached)
	zfs_acl_free(zp->z_acl_cached);
	zp->z_acl_cached = aclp;
	aclp = NULL;
	}


	if (mask & AT_ATIME) {
	ZFS_TIME_ENCODE(&vap->va_atime, zp->z_atime);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
	&zp->z_atime, sizeof (zp->z_atime));
	}

	if (mask & AT_MTIME) {
	ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
	mtime, sizeof (mtime));
	}

	if (projid != ZFS_INVALID_PROJID) {
	zp->z_projid = projid;
	SA_ADD_BULK_ATTR(bulk, count,
	SA_ZPL_PROJID(zfsvfs), NULL, &zp->z_projid,
	sizeof (zp->z_projid));
	}

	/* XXX - shouldn't this be done before the ATIME/MTIME checks? */
	if (mask & AT_SIZE && !(mask & AT_MTIME)) {
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs),
	NULL, mtime, sizeof (mtime));
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
	&ctime, sizeof (ctime));
	zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime);
	} else if (mask != 0) {
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
	&ctime, sizeof (ctime));
	zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime);
	if (attrzp) {
	SA_ADD_BULK_ATTR(xattr_bulk, xattr_count,
	SA_ZPL_CTIME(zfsvfs), NULL,
	&ctime, sizeof (ctime));
	zfs_tstamp_update_setup(attrzp, STATE_CHANGED,
	mtime, ctime);
	}
	}

	/*
	* Do this after setting timestamps to prevent timestamp
	* update from toggling bit
	*/

	if (xoap && (mask & AT_XVATTR)) {

	if (XVA_ISSET_REQ(xvap, XAT_CREATETIME))
	xoap->xoa_createtime = vap->va_birthtime;
	/*
	* restore trimmed off masks
	* so that return masks can be set for caller.
	*/

	if (XVA_ISSET_REQ(&tmpxvattr, XAT_APPENDONLY)) {
	XVA_SET_REQ(xvap, XAT_APPENDONLY);
	}
	if (XVA_ISSET_REQ(&tmpxvattr, XAT_NOUNLINK)) {
	XVA_SET_REQ(xvap, XAT_NOUNLINK);
	}
	if (XVA_ISSET_REQ(&tmpxvattr, XAT_IMMUTABLE)) {
	XVA_SET_REQ(xvap, XAT_IMMUTABLE);
	}
	if (XVA_ISSET_REQ(&tmpxvattr, XAT_NODUMP)) {
	XVA_SET_REQ(xvap, XAT_NODUMP);
	}
	if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_MODIFIED)) {
	XVA_SET_REQ(xvap, XAT_AV_MODIFIED);
	}
	if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_QUARANTINED)) {
	XVA_SET_REQ(xvap, XAT_AV_QUARANTINED);
	}
	if (XVA_ISSET_REQ(&tmpxvattr, XAT_PROJINHERIT)) {
	XVA_SET_REQ(xvap, XAT_PROJINHERIT);
	}

	if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP))
	ASSERT3S(vp->v_type, ==, VREG);

	zfs_xvattr_set(zp, xvap, tx);
	}

	if (fuid_dirtied)
	zfs_fuid_sync(zfsvfs, tx);

	if (mask != 0)
	zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp);

	if (mask & (AT_UID\|AT_GID\|AT_MODE))
	mutex_exit(&zp->z_acl_lock);

	if (attrzp) {
	if (mask & (AT_UID\|AT_GID\|AT_MODE))
	mutex_exit(&attrzp->z_acl_lock);
	}
	out:
	if (err == 0 && attrzp) {
	err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk,
	xattr_count, tx);
	ASSERT0(err2);
	}

	if (attrzp)
	vput(ZTOV(attrzp));

	if (aclp)
	zfs_acl_free(aclp);

	if (fuidp) {
	zfs_fuid_info_free(fuidp);
	fuidp = NULL;
	}

	if (err) {
	dmu_tx_abort(tx);
	} else {
	err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
	dmu_tx_commit(tx);
	}

	out2:
	if (os->os_sync == ZFS_SYNC_ALWAYS)
	zil_commit(zilog, 0);

	zfs_exit(zfsvfs, FTAG);
	return (err);
	}

	/*
	* Look up the directory entries corresponding to the source and target
	* directory/name pairs.
	*/
	static int
	zfs_rename_relock_lookup(znode_t sdzp, const struct componentname scnp,
	znode_t *szpp, znode_t tdzp, const struct componentname *tcnp,
	znode_t **tzpp)
	{
	zfsvfs_t *zfsvfs;
	znode_t szp, tzp;
	int error;

	/*
	* Before using sdzp and tdzp we must ensure that they are live.
	* As a porting legacy from illumos we have two things to worry
	* about. One is typical for FreeBSD and it is that the vnode is
	* not reclaimed (doomed). The other is that the znode is live.
	* The current code can invalidate the znode without acquiring the
	* corresponding vnode lock if the object represented by the znode
	* and vnode is no longer valid after a rollback or receive operation.
	* z_teardown_lock hidden behind zfs_enter and zfs_exit is the lock
	* that protects the znodes from the invalidation.
	*/
	zfsvfs = sdzp->z_zfsvfs;
	ASSERT3P(zfsvfs, ==, tdzp->z_zfsvfs);
	if ((error = zfs_enter_verify_zp(zfsvfs, sdzp, FTAG)) != 0)
	return (error);
	if ((error = zfs_verify_zp(tdzp)) != 0) {
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	/*
	* Re-resolve svp to be certain it still exists and fetch the
	* correct vnode.
	*/
	error = zfs_dirent_lookup(sdzp, scnp->cn_nameptr, &szp, ZEXISTS);
	if (error != 0) {
	/* Source entry invalid or not there. */
	if ((scnp->cn_flags & ISDOTDOT) != 0 \|\|
	(scnp->cn_namelen == 1 && scnp->cn_nameptr[0] == '.'))
	error = SET_ERROR(EINVAL);
	goto out;
	}
	*szpp = szp;

	/*
	* Re-resolve tvp, if it disappeared we just carry on.
	*/
	error = zfs_dirent_lookup(tdzp, tcnp->cn_nameptr, &tzp, 0);
	if (error != 0) {
	vrele(ZTOV(szp));
	if ((tcnp->cn_flags & ISDOTDOT) != 0)
	error = SET_ERROR(EINVAL);
	goto out;
	}
	*tzpp = tzp;
	out:
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	/*
	* We acquire all but fdvp locks using non-blocking acquisitions. If we
	* fail to acquire any lock in the path we will drop all held locks,
	* acquire the new lock in a blocking fashion, and then release it and
	* restart the rename. This acquire/release step ensures that we do not
	* spin on a lock waiting for release. On error release all vnode locks
	* and decrement references the way tmpfs_rename() would do.
	*/
	static int
	zfs_rename_relock(struct vnode sdvp, struct vnode *svpp,
	struct vnode tdvp, struct vnode *tvpp,
	const struct componentname scnp, const struct componentname tcnp)
	{
	struct vnode nvp, svp, *tvp;
	znode_t sdzp, tdzp, szp, tzp;
	int error;

	VOP_UNLOCK1(tdvp);
	if (tvpp != NULL && tvpp != tdvp)
	VOP_UNLOCK1(*tvpp);

	relock:
	error = vn_lock(sdvp, LK_EXCLUSIVE);
	if (error)
	goto out;
	error = vn_lock(tdvp, LK_EXCLUSIVE \| LK_NOWAIT);
	if (error != 0) {
	VOP_UNLOCK1(sdvp);
	if (error != EBUSY)
	goto out;
	error = vn_lock(tdvp, LK_EXCLUSIVE);
	if (error)
	goto out;
	VOP_UNLOCK1(tdvp);
	goto relock;
	}
	tdzp = VTOZ(tdvp);
	sdzp = VTOZ(sdvp);

	error = zfs_rename_relock_lookup(sdzp, scnp, &szp, tdzp, tcnp, &tzp);
	if (error != 0) {
	VOP_UNLOCK1(sdvp);
	VOP_UNLOCK1(tdvp);
	goto out;
	}
	svp = ZTOV(szp);
	tvp = tzp != NULL ? ZTOV(tzp) : NULL;

	/*
	* Now try acquire locks on svp and tvp.
	*/
	nvp = svp;
	error = vn_lock(nvp, LK_EXCLUSIVE \| LK_NOWAIT);
	if (error != 0) {
	VOP_UNLOCK1(sdvp);
	VOP_UNLOCK1(tdvp);
	if (tvp != NULL)
	vrele(tvp);
	if (error != EBUSY) {
	vrele(nvp);
	goto out;
	}
	error = vn_lock(nvp, LK_EXCLUSIVE);
	if (error != 0) {
	vrele(nvp);
	goto out;
	}
	VOP_UNLOCK1(nvp);
	/*
	* Concurrent rename race.
	* XXX ?
	*/
	if (nvp == tdvp) {
	vrele(nvp);
	error = SET_ERROR(EINVAL);
	goto out;
	}
	vrele(*svpp);
	*svpp = nvp;
	goto relock;
	}
	vrele(*svpp);
	*svpp = nvp;

	if (*tvpp != NULL)
	vrele(*tvpp);
	*tvpp = NULL;
	if (tvp != NULL) {
	nvp = tvp;
	error = vn_lock(nvp, LK_EXCLUSIVE \| LK_NOWAIT);
	if (error != 0) {
	VOP_UNLOCK1(sdvp);
	VOP_UNLOCK1(tdvp);
	VOP_UNLOCK1(*svpp);
	if (error != EBUSY) {
	vrele(nvp);
	goto out;
	}
	error = vn_lock(nvp, LK_EXCLUSIVE);
	if (error != 0) {
	vrele(nvp);
	goto out;
	}
	vput(nvp);
	goto relock;
	}
	*tvpp = nvp;
	}

	return (0);

	out:
	return (error);
	}

	/*
	* Note that we must use VRELE_ASYNC in this function as it walks
	* up the directory tree and vrele may need to acquire an exclusive
	* lock if a last reference to a vnode is dropped.
	*/
	static int
	zfs_rename_check(znode_t szp, znode_t sdzp, znode_t *tdzp)
	{
	zfsvfs_t *zfsvfs;
	znode_t zp, zp1;
	uint64_t parent;
	int error;

	zfsvfs = tdzp->z_zfsvfs;
	if (tdzp == szp)
	return (SET_ERROR(EINVAL));
	if (tdzp == sdzp)
	return (0);
	if (tdzp->z_id == zfsvfs->z_root)
	return (0);
	zp = tdzp;
	for (;;) {
	ASSERT(!zp->z_unlinked);
	if ((error = sa_lookup(zp->z_sa_hdl,
	SA_ZPL_PARENT(zfsvfs), &parent, sizeof (parent))) != 0)
	break;

	if (parent == szp->z_id) {
	error = SET_ERROR(EINVAL);
	break;
	}
	if (parent == zfsvfs->z_root)
	break;
	if (parent == sdzp->z_id)
	break;

	error = zfs_zget(zfsvfs, parent, &zp1);
	if (error != 0)
	break;

	if (zp != tdzp)
	VN_RELE_ASYNC(ZTOV(zp),
	dsl_pool_zrele_taskq(
	dmu_objset_pool(zfsvfs->z_os)));
	zp = zp1;
	}

	if (error == ENOTDIR)
	panic("checkpath: .. not a directory\n");
	if (zp != tdzp)
	VN_RELE_ASYNC(ZTOV(zp),
	dsl_pool_zrele_taskq(dmu_objset_pool(zfsvfs->z_os)));
	return (error);
	}

	#if __FreeBSD_version < 1300124
	static void
	cache_vop_rename(struct vnode fdvp, struct vnode fvp, struct vnode *tdvp,
	struct vnode tvp, struct componentname fcnp, struct componentname *tcnp)
	{

	cache_purge(fvp);
	if (tvp != NULL)
	cache_purge(tvp);
	cache_purge_negative(tdvp);
	}
	#endif

	static int
	zfs_do_rename_impl(vnode_t sdvp, vnode_t svpp, struct componentname scnp,
	vnode_t tdvp, vnode_t tvpp, struct componentname tcnp,
	cred_t *cr);

	/*
	* Move an entry from the provided source directory to the target
	* directory. Change the entry name as indicated.
	*
	* IN: sdvp - Source directory containing the "old entry".
	* scnp - Old entry name.
	* tdvp - Target directory to contain the "new entry".
	* tcnp - New entry name.
	* cr - credentials of caller.
	* INOUT: svpp - Source file
	* tvpp - Target file, may point to NULL initially
	*
	* RETURN: 0 on success, error code on failure.
	*
	* Timestamps:
	* sdvp,tdvp - ctime\|mtime updated
	*/
	static int
	zfs_do_rename(vnode_t sdvp, vnode_t svpp, struct componentname scnp,
	vnode_t tdvp, vnode_t tvpp, struct componentname tcnp,
	cred_t *cr)
	{
	int error;

	ASSERT_VOP_ELOCKED(tdvp, __func__);
	if (*tvpp != NULL)
	ASSERT_VOP_ELOCKED(*tvpp, __func__);

	/* Reject renames across filesystems. */
	if ((*svpp)->v_mount != tdvp->v_mount \|\|
	((tvpp) != NULL && (svpp)->v_mount != (*tvpp)->v_mount)) {
	error = SET_ERROR(EXDEV);
	goto out;
	}

	if (zfsctl_is_node(tdvp)) {
	error = SET_ERROR(EXDEV);
	goto out;
	}

	/*
	* Lock all four vnodes to ensure safety and semantics of renaming.
	*/
	error = zfs_rename_relock(sdvp, svpp, tdvp, tvpp, scnp, tcnp);
	if (error != 0) {
	/* no vnodes are locked in the case of error here */
	return (error);
	}

	error = zfs_do_rename_impl(sdvp, svpp, scnp, tdvp, tvpp, tcnp, cr);
	VOP_UNLOCK1(sdvp);
	VOP_UNLOCK1(*svpp);
	out:
	if (*tvpp != NULL)
	VOP_UNLOCK1(*tvpp);
	if (tdvp != *tvpp)
	VOP_UNLOCK1(tdvp);

	return (error);
	}

	static int
	zfs_do_rename_impl(vnode_t sdvp, vnode_t svpp, struct componentname scnp,
	vnode_t tdvp, vnode_t tvpp, struct componentname tcnp,
	cred_t *cr)
	{
	dmu_tx_t *tx;
	zfsvfs_t *zfsvfs;
	zilog_t *zilog;
	znode_t tdzp, sdzp, tzp, szp;
	const char *snm = scnp->cn_nameptr;
	const char *tnm = tcnp->cn_nameptr;
	int error;

	tdzp = VTOZ(tdvp);
	sdzp = VTOZ(sdvp);
	zfsvfs = tdzp->z_zfsvfs;

	if ((error = zfs_enter_verify_zp(zfsvfs, tdzp, FTAG)) != 0)
	return (error);
	if ((error = zfs_verify_zp(sdzp)) != 0) {
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}
	zilog = zfsvfs->z_log;

	if (zfsvfs->z_utf8 && u8_validate(tnm,
	strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
	error = SET_ERROR(EILSEQ);
	goto out;
	}

	/* If source and target are the same file, there is nothing to do. */
	if ((svpp) == (tvpp)) {
	error = 0;
	goto out;
	}

	if (((svpp)->v_type == VDIR && (svpp)->v_mountedhere != NULL) \|\|
	((tvpp) != NULL && (tvpp)->v_type == VDIR &&
	(*tvpp)->v_mountedhere != NULL)) {
	error = SET_ERROR(EXDEV);
	goto out;
	}

	szp = VTOZ(*svpp);
	if ((error = zfs_verify_zp(szp)) != 0) {
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}
	tzp = tvpp == NULL ? NULL : VTOZ(tvpp);
	if (tzp != NULL) {
	if ((error = zfs_verify_zp(tzp)) != 0) {
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}
	}

	/*
	* This is to prevent the creation of links into attribute space
	* by renaming a linked file into/outof an attribute directory.
	* See the comment in zfs_link() for why this is considered bad.
	*/
	if ((tdzp->z_pflags & ZFS_XATTR) != (sdzp->z_pflags & ZFS_XATTR)) {
	error = SET_ERROR(EINVAL);
	goto out;
	}

	/*
	* If we are using project inheritance, means if the directory has
	* ZFS_PROJINHERIT set, then its descendant directories will inherit
	* not only the project ID, but also the ZFS_PROJINHERIT flag. Under
	* such case, we only allow renames into our tree when the project
	* IDs are the same.
	*/
	if (tdzp->z_pflags & ZFS_PROJINHERIT &&
	tdzp->z_projid != szp->z_projid) {
	error = SET_ERROR(EXDEV);
	goto out;
	}

	/*
	* Must have write access at the source to remove the old entry
	* and write access at the target to create the new entry.
	* Note that if target and source are the same, this can be
	* done in a single check.
	*/
	if ((error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr, NULL)))
	goto out;

	if ((*svpp)->v_type == VDIR) {
	/*
	* Avoid ".", "..", and aliases of "." for obvious reasons.
	*/
	if ((scnp->cn_namelen == 1 && scnp->cn_nameptr[0] == '.') \|\|
	sdzp == szp \|\|
	(scnp->cn_flags \| tcnp->cn_flags) & ISDOTDOT) {
	error = EINVAL;
	goto out;
	}

	/*
	* Check to make sure rename is valid.
	* Can't do a move like this: /usr/a/b to /usr/a/b/c/d
	*/
	if ((error = zfs_rename_check(szp, sdzp, tdzp)))
	goto out;
	}

	/*
	* Does target exist?
	*/
	if (tzp) {
	/*
	* Source and target must be the same type.
	*/
	if ((*svpp)->v_type == VDIR) {
	if ((*tvpp)->v_type != VDIR) {
	error = SET_ERROR(ENOTDIR);
	goto out;
	} else {
	cache_purge(tdvp);
	if (sdvp != tdvp)
	cache_purge(sdvp);
	}
	} else {
	if ((*tvpp)->v_type == VDIR) {
	error = SET_ERROR(EISDIR);
	goto out;
	}
	}
	}

	vn_seqc_write_begin(*svpp);
	vn_seqc_write_begin(sdvp);
	if (*tvpp != NULL)
	vn_seqc_write_begin(*tvpp);
	if (tdvp != *tvpp)
	vn_seqc_write_begin(tdvp);

	vnevent_rename_src(*svpp, sdvp, scnp->cn_nameptr, ct);
	if (tzp)
	vnevent_rename_dest(*tvpp, tdvp, tnm, ct);

	/*
	* notify the target directory if it is not the same
	* as source directory.
	*/
	if (tdvp != sdvp) {
	vnevent_rename_dest_dir(tdvp, ct);
	}

	tx = dmu_tx_create(zfsvfs->z_os);
	dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE);
	dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE);
	dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm);
	dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm);
	if (sdzp != tdzp) {
	dmu_tx_hold_sa(tx, tdzp->z_sa_hdl, B_FALSE);
	zfs_sa_upgrade_txholds(tx, tdzp);
	}
	if (tzp) {
	dmu_tx_hold_sa(tx, tzp->z_sa_hdl, B_FALSE);
	zfs_sa_upgrade_txholds(tx, tzp);
	}

	zfs_sa_upgrade_txholds(tx, szp);
	dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error) {
	dmu_tx_abort(tx);
	goto out_seq;
	}

	if (tzp) /* Attempt to remove the existing target */
	error = zfs_link_destroy(tdzp, tnm, tzp, tx, 0, NULL);

	if (error == 0) {
	error = zfs_link_create(tdzp, tnm, szp, tx, ZRENAMING);
	if (error == 0) {
	szp->z_pflags \|= ZFS_AV_MODIFIED;

	error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs),
	(void *)&szp->z_pflags, sizeof (uint64_t), tx);
	ASSERT0(error);

	error = zfs_link_destroy(sdzp, snm, szp, tx, ZRENAMING,
	NULL);
	if (error == 0) {
	zfs_log_rename(zilog, tx, TX_RENAME, sdzp,
	snm, tdzp, tnm, szp);
	} else {
	/*
	* At this point, we have successfully created
	* the target name, but have failed to remove
	* the source name. Since the create was done
	* with the ZRENAMING flag, there are
	* complications; for one, the link count is
	* wrong. The easiest way to deal with this
	* is to remove the newly created target, and
	* return the original error. This must
	* succeed; fortunately, it is very unlikely to
	* fail, since we just created it.
	*/
	VERIFY0(zfs_link_destroy(tdzp, tnm, szp, tx,
	ZRENAMING, NULL));
	}
	}
	if (error == 0) {
	cache_vop_rename(sdvp, svpp, tdvp, tvpp, scnp, tcnp);
	}
	}

	dmu_tx_commit(tx);

	out_seq:
	vn_seqc_write_end(*svpp);
	vn_seqc_write_end(sdvp);
	if (*tvpp != NULL)
	vn_seqc_write_end(*tvpp);
	if (tdvp != *tvpp)
	vn_seqc_write_end(tdvp);

	out:
	if (error == 0 && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
	zil_commit(zilog, 0);
	zfs_exit(zfsvfs, FTAG);

	return (error);
	}

	int
	zfs_rename(znode_t sdzp, const char sname, znode_t tdzp, const char tname,
	cred_t cr, int flags, uint64_t rflags, vattr_t wo_vap, zidmap_t *mnt_ns)
	{
	struct componentname scn, tcn;
	vnode_t sdvp, tdvp;
	vnode_t svp, tvp;
	int error;
	svp = tvp = NULL;

	if (rflags != 0 \|\| wo_vap != NULL)
	return (SET_ERROR(EINVAL));

	sdvp = ZTOV(sdzp);
	tdvp = ZTOV(tdzp);
	error = zfs_lookup_internal(sdzp, sname, &svp, &scn, DELETE);
	if (sdzp->z_zfsvfs->z_replay == B_FALSE)
	VOP_UNLOCK1(sdvp);
	if (error != 0)
	goto fail;
	VOP_UNLOCK1(svp);

	vn_lock(tdvp, LK_EXCLUSIVE \| LK_RETRY);
	error = zfs_lookup_internal(tdzp, tname, &tvp, &tcn, RENAME);
	if (error == EJUSTRETURN)
	tvp = NULL;
	else if (error != 0) {
	VOP_UNLOCK1(tdvp);
	goto fail;
	}

	error = zfs_do_rename(sdvp, &svp, &scn, tdvp, &tvp, &tcn, cr);
	fail:
	if (svp != NULL)
	vrele(svp);
	if (tvp != NULL)
	vrele(tvp);

	return (error);
	}

	/*
	* Insert the indicated symbolic reference entry into the directory.
	*
	* IN: dvp - Directory to contain new symbolic link.
	* link - Name for new symlink entry.
	* vap - Attributes of new entry.
	* cr - credentials of caller.
	* ct - caller context
	* flags - case flags
	* mnt_ns - Unused on FreeBSD
	*
	* RETURN: 0 on success, error code on failure.
	*
	* Timestamps:
	* dvp - ctime\|mtime updated
	*/
	int
	zfs_symlink(znode_t dzp, const char name, vattr_t *vap,
	const char link, znode_t zpp, cred_t cr, int flags, zidmap_t *mnt_ns)
	{
	(void) flags;
	znode_t *zp;
	dmu_tx_t *tx;
	zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
	zilog_t *zilog;
	uint64_t len = strlen(link);
	int error;
	zfs_acl_ids_t acl_ids;
	boolean_t fuid_dirtied;
	uint64_t txtype = TX_SYMLINK;

	ASSERT3S(vap->va_type, ==, VLNK);

	if ((error = zfs_enter_verify_zp(zfsvfs, dzp, FTAG)) != 0)
	return (error);
	zilog = zfsvfs->z_log;

	if (zfsvfs->z_utf8 && u8_validate(name, strlen(name),
	NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EILSEQ));
	}

	if (len > MAXPATHLEN) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(ENAMETOOLONG));
	}

	if ((error = zfs_acl_ids_create(dzp, 0,
	vap, cr, NULL, &acl_ids, NULL)) != 0) {
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	/*
	* Attempt to lock directory; fail if entry already exists.
	*/
	error = zfs_dirent_lookup(dzp, name, &zp, ZNEW);
	if (error) {
	zfs_acl_ids_free(&acl_ids);
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr, mnt_ns))) {
	zfs_acl_ids_free(&acl_ids);
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	if (zfs_acl_ids_overquota(zfsvfs, &acl_ids,
	0 /* projid */)) {
	zfs_acl_ids_free(&acl_ids);
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EDQUOT));
	}

	getnewvnode_reserve_();
	tx = dmu_tx_create(zfsvfs->z_os);
	fuid_dirtied = zfsvfs->z_fuid_dirty;
	dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len));
	dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name);
	dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
	ZFS_SA_BASE_ATTR_SIZE + len);
	dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE);
	if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) {
	dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0,
	acl_ids.z_aclp->z_acl_bytes);
	}
	if (fuid_dirtied)
	zfs_fuid_txhold(zfsvfs, tx);
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error) {
	zfs_acl_ids_free(&acl_ids);
	dmu_tx_abort(tx);
	getnewvnode_drop_reserve();
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	/*
	* Create a new object for the symlink.
	* for version 4 ZPL datasets the symlink will be an SA attribute
	*/
	zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids);

	if (fuid_dirtied)
	zfs_fuid_sync(zfsvfs, tx);

	if (zp->z_is_sa)
	error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zfsvfs),
	__DECONST(void *, link), len, tx);
	else
	zfs_sa_symlink(zp, __DECONST(char *, link), len, tx);

	zp->z_size = len;
	(void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs),
	&zp->z_size, sizeof (zp->z_size), tx);
	/*
	* Insert the new object into the directory.
	*/
	- (void) zfs_link_create(dzp, name, zp, tx, ZNEW);
	-
	- zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link);
	- *zpp = zp;
	+ error = zfs_link_create(dzp, name, zp, tx, ZNEW);
	+ if (error != 0) {
	+ zfs_znode_delete(zp, tx);
	+ VOP_UNLOCK1(ZTOV(zp));
	+ zrele(zp);
	+ } else {
	+ zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link);
	+ }

	zfs_acl_ids_free(&acl_ids);

	dmu_tx_commit(tx);

	getnewvnode_drop_reserve();

	- if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
	- zil_commit(zilog, 0);
	+ if (error == 0) {
	+ *zpp = zp;
	+
	+ if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
	+ zil_commit(zilog, 0);
	+ }

	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	/*
	* Return, in the buffer contained in the provided uio structure,
	* the symbolic path referred to by vp.
	*
	* IN: vp - vnode of symbolic link.
	* uio - structure to contain the link path.
	* cr - credentials of caller.
	* ct - caller context
	*
	* OUT: uio - structure containing the link path.
	*
	* RETURN: 0 on success, error code on failure.
	*
	* Timestamps:
	* vp - atime updated
	*/
	static int
	zfs_readlink(vnode_t vp, zfs_uio_t uio, cred_t cr, caller_context_t ct)
	{
	(void) cr, (void) ct;
	znode_t *zp = VTOZ(vp);
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	int error;

	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (error);

	if (zp->z_is_sa)
	error = sa_lookup_uio(zp->z_sa_hdl,
	SA_ZPL_SYMLINK(zfsvfs), uio);
	else
	error = zfs_sa_readlink(zp, uio);

	ZFS_ACCESSTIME_STAMP(zfsvfs, zp);

	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	/*
	* Insert a new entry into directory tdvp referencing svp.
	*
	* IN: tdvp - Directory to contain new entry.
	* svp - vnode of new entry.
	* name - name of new entry.
	* cr - credentials of caller.
	*
	* RETURN: 0 on success, error code on failure.
	*
	* Timestamps:
	* tdvp - ctime\|mtime updated
	* svp - ctime updated
	*/
	int
	zfs_link(znode_t tdzp, znode_t szp, const char name, cred_t cr,
	int flags)
	{
	(void) flags;
	znode_t *tzp;
	zfsvfs_t *zfsvfs = tdzp->z_zfsvfs;
	zilog_t *zilog;
	dmu_tx_t *tx;
	int error;
	uint64_t parent;
	uid_t owner;

	ASSERT3S(ZTOV(tdzp)->v_type, ==, VDIR);

	if ((error = zfs_enter_verify_zp(zfsvfs, tdzp, FTAG)) != 0)
	return (error);
	zilog = zfsvfs->z_log;

	/*
	* POSIX dictates that we return EPERM here.
	* Better choices include ENOTSUP or EISDIR.
	*/
	if (ZTOV(szp)->v_type == VDIR) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EPERM));
	}

	if ((error = zfs_verify_zp(szp)) != 0) {
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	/*
	* If we are using project inheritance, means if the directory has
	* ZFS_PROJINHERIT set, then its descendant directories will inherit
	* not only the project ID, but also the ZFS_PROJINHERIT flag. Under
	* such case, we only allow hard link creation in our tree when the
	* project IDs are the same.
	*/
	if (tdzp->z_pflags & ZFS_PROJINHERIT &&
	tdzp->z_projid != szp->z_projid) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EXDEV));
	}

	if (szp->z_pflags & (ZFS_APPENDONLY \|
	ZFS_IMMUTABLE \| ZFS_READONLY)) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EPERM));
	}

	/* Prevent links to .zfs/shares files */

	if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
	&parent, sizeof (uint64_t))) != 0) {
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}
	if (parent == zfsvfs->z_shares_dir) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EPERM));
	}

	if (zfsvfs->z_utf8 && u8_validate(name,
	strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EILSEQ));
	}

	/*
	* We do not support links between attributes and non-attributes
	* because of the potential security risk of creating links
	* into "normal" file space in order to circumvent restrictions
	* imposed in attribute space.
	*/
	if ((szp->z_pflags & ZFS_XATTR) != (tdzp->z_pflags & ZFS_XATTR)) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EINVAL));
	}


	owner = zfs_fuid_map_id(zfsvfs, szp->z_uid, cr, ZFS_OWNER);
	if (owner != crgetuid(cr) && secpolicy_basic_link(ZTOV(szp), cr) != 0) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EPERM));
	}

	if ((error = zfs_zaccess(tdzp, ACE_ADD_FILE, 0, B_FALSE, cr, NULL))) {
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	/*
	* Attempt to lock directory; fail if entry already exists.
	*/
	error = zfs_dirent_lookup(tdzp, name, &tzp, ZNEW);
	if (error) {
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	tx = dmu_tx_create(zfsvfs->z_os);
	dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE);
	dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, name);
	zfs_sa_upgrade_txholds(tx, szp);
	zfs_sa_upgrade_txholds(tx, tdzp);
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error) {
	dmu_tx_abort(tx);
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	error = zfs_link_create(tdzp, name, szp, tx, 0);

	if (error == 0) {
	uint64_t txtype = TX_LINK;
	zfs_log_link(zilog, tx, txtype, tdzp, szp, name);
	}

	dmu_tx_commit(tx);

	if (error == 0) {
	vnevent_link(ZTOV(szp), ct);
	}

	if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
	zil_commit(zilog, 0);

	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	/*
	* Free or allocate space in a file. Currently, this function only
	* supports the `F_FREESP' command. However, this command is somewhat
	* misnamed, as its functionality includes the ability to allocate as
	* well as free space.
	*
	* IN: ip - inode of file to free data in.
	* cmd - action to take (only F_FREESP supported).
	* bfp - section of file to free/alloc.
	* flag - current file open mode flags.
	* offset - current file offset.
	* cr - credentials of caller.
	*
	* RETURN: 0 on success, error code on failure.
	*
	* Timestamps:
	* ip - ctime\|mtime updated
	*/
	int
	zfs_space(znode_t zp, int cmd, flock64_t bfp, int flag,
	offset_t offset, cred_t *cr)
	{
	(void) offset;
	zfsvfs_t *zfsvfs = ZTOZSB(zp);
	uint64_t off, len;
	int error;

	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (error);

	if (cmd != F_FREESP) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EINVAL));
	}

	/*
	* Callers might not be able to detect properly that we are read-only,
	* so check it explicitly here.
	*/
	if (zfs_is_readonly(zfsvfs)) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EROFS));
	}

	if (bfp->l_len < 0) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EINVAL));
	}

	/*
	* Permissions aren't checked on Solaris because on this OS
	* zfs_space() can only be called with an opened file handle.
	* On Linux we can get here through truncate_range() which
	* operates directly on inodes, so we need to check access rights.
	*/
	if ((error = zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr, NULL))) {
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	off = bfp->l_start;
	len = bfp->l_len; /* 0 means from off to end of file */

	error = zfs_freesp(zp, off, len, flag, TRUE);

	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	static void
	zfs_inactive(vnode_t vp, cred_t cr, caller_context_t *ct)
	{
	(void) cr, (void) ct;
	znode_t *zp = VTOZ(vp);
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	int error;

	ZFS_TEARDOWN_INACTIVE_ENTER_READ(zfsvfs);
	if (zp->z_sa_hdl == NULL) {
	/*
	* The fs has been unmounted, or we did a
	* suspend/resume and this file no longer exists.
	*/
	ZFS_TEARDOWN_INACTIVE_EXIT_READ(zfsvfs);
	vrecycle(vp);
	return;
	}

	if (zp->z_unlinked) {
	/*
	* Fast path to recycle a vnode of a removed file.
	*/
	ZFS_TEARDOWN_INACTIVE_EXIT_READ(zfsvfs);
	vrecycle(vp);
	return;
	}

	if (zp->z_atime_dirty && zp->z_unlinked == 0) {
	dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os);

	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
	zfs_sa_upgrade_txholds(tx, zp);
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error) {
	dmu_tx_abort(tx);
	} else {
	(void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs),
	(void *)&zp->z_atime, sizeof (zp->z_atime), tx);
	zp->z_atime_dirty = 0;
	dmu_tx_commit(tx);
	}
	}
	ZFS_TEARDOWN_INACTIVE_EXIT_READ(zfsvfs);
	}


	_Static_assert(sizeof (struct zfid_short) <= sizeof (struct fid),
	"struct zfid_short bigger than struct fid");
	_Static_assert(sizeof (struct zfid_long) <= sizeof (struct fid),
	"struct zfid_long bigger than struct fid");

	static int
	zfs_fid(vnode_t vp, fid_t fidp, caller_context_t *ct)
	{
	(void) ct;
	znode_t *zp = VTOZ(vp);
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	uint32_t gen;
	uint64_t gen64;
	uint64_t object = zp->z_id;
	zfid_short_t *zfid;
	int size, i, error;

	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (error);

	if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs),
	&gen64, sizeof (uint64_t))) != 0) {
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	gen = (uint32_t)gen64;

	size = (zfsvfs->z_parent != zfsvfs) ? LONG_FID_LEN : SHORT_FID_LEN;
	fidp->fid_len = size;

	zfid = (zfid_short_t *)fidp;

	zfid->zf_len = size;

	for (i = 0; i < sizeof (zfid->zf_object); i++)
	zfid->zf_object[i] = (uint8_t)(object >> (8 * i));

	/* Must have a non-zero generation number to distinguish from .zfs */
	if (gen == 0)
	gen = 1;
	for (i = 0; i < sizeof (zfid->zf_gen); i++)
	zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i));

	if (size == LONG_FID_LEN) {
	uint64_t objsetid = dmu_objset_id(zfsvfs->z_os);
	zfid_long_t *zlfid;

	zlfid = (zfid_long_t *)fidp;

	for (i = 0; i < sizeof (zlfid->zf_setid); i++)
	zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i));

	/* XXX - this should be the generation number for the objset */
	for (i = 0; i < sizeof (zlfid->zf_setgen); i++)
	zlfid->zf_setgen[i] = 0;
	}

	zfs_exit(zfsvfs, FTAG);
	return (0);
	}

	static int
	zfs_pathconf(vnode_t vp, int cmd, ulong_t valp, cred_t *cr,
	caller_context_t *ct)
	{
	znode_t *zp;
	zfsvfs_t *zfsvfs;
	int error;

	switch (cmd) {
	case _PC_LINK_MAX:
	*valp = MIN(LONG_MAX, ZFS_LINK_MAX);
	return (0);

	case _PC_FILESIZEBITS:
	*valp = 64;
	return (0);
	case _PC_MIN_HOLE_SIZE:
	*valp = (int)SPA_MINBLOCKSIZE;
	return (0);
	case _PC_ACL_EXTENDED:
	#if 0 /* POSIX ACLs are not implemented for ZFS on FreeBSD yet. */
	zp = VTOZ(vp);
	zfsvfs = zp->z_zfsvfs;
	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (error);
	*valp = zfsvfs->z_acl_type == ZFSACLTYPE_POSIX ? 1 : 0;
	zfs_exit(zfsvfs, FTAG);
	#else
	*valp = 0;
	#endif
	return (0);

	case _PC_ACL_NFS4:
	zp = VTOZ(vp);
	zfsvfs = zp->z_zfsvfs;
	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (error);
	*valp = zfsvfs->z_acl_type == ZFS_ACLTYPE_NFSV4 ? 1 : 0;
	zfs_exit(zfsvfs, FTAG);
	return (0);

	case _PC_ACL_PATH_MAX:
	*valp = ACL_MAX_ENTRIES;
	return (0);

	default:
	return (EOPNOTSUPP);
	}
	}

	static int
	zfs_getpages(struct vnode vp, vm_page_t ma, int count, int *rbehind,
	int *rahead)
	{
	znode_t *zp = VTOZ(vp);
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	zfs_locked_range_t *lr;
	vm_object_t object;
	off_t start, end, obj_size;
	uint_t blksz;
	int pgsin_b, pgsin_a;
	int error;

	if (zfs_enter_verify_zp(zfsvfs, zp, FTAG) != 0)
	return (zfs_vm_pagerret_error);

	start = IDX_TO_OFF(ma[0]->pindex);
	end = IDX_TO_OFF(ma[count - 1]->pindex + 1);

	/*
	* Lock a range covering all required and optional pages.
	* Note that we need to handle the case of the block size growing.
	*/
	for (;;) {
	blksz = zp->z_blksz;
	lr = zfs_rangelock_tryenter(&zp->z_rangelock,
	rounddown(start, blksz),
	roundup(end, blksz) - rounddown(start, blksz), RL_READER);
	if (lr == NULL) {
	if (rahead != NULL) {
	*rahead = 0;
	rahead = NULL;
	}
	if (rbehind != NULL) {
	*rbehind = 0;
	rbehind = NULL;
	}
	break;
	}
	if (blksz == zp->z_blksz)
	break;
	zfs_rangelock_exit(lr);
	}

	object = ma[0]->object;
	zfs_vmobject_wlock(object);
	obj_size = object->un_pager.vnp.vnp_size;
	zfs_vmobject_wunlock(object);
	if (IDX_TO_OFF(ma[count - 1]->pindex) >= obj_size) {
	if (lr != NULL)
	zfs_rangelock_exit(lr);
	zfs_exit(zfsvfs, FTAG);
	return (zfs_vm_pagerret_bad);
	}

	pgsin_b = 0;
	if (rbehind != NULL) {
	pgsin_b = OFF_TO_IDX(start - rounddown(start, blksz));
	pgsin_b = MIN(*rbehind, pgsin_b);
	}

	pgsin_a = 0;
	if (rahead != NULL) {
	pgsin_a = OFF_TO_IDX(roundup(end, blksz) - end);
	if (end + IDX_TO_OFF(pgsin_a) >= obj_size)
	pgsin_a = OFF_TO_IDX(round_page(obj_size) - end);
	pgsin_a = MIN(*rahead, pgsin_a);
	}

	/*
	* NB: we need to pass the exact byte size of the data that we expect
	* to read after accounting for the file size. This is required because
	* ZFS will panic if we request DMU to read beyond the end of the last
	* allocated block.
	*/
	error = dmu_read_pages(zfsvfs->z_os, zp->z_id, ma, count, &pgsin_b,
	&pgsin_a, MIN(end, obj_size) - (end - PAGE_SIZE));

	if (lr != NULL)
	zfs_rangelock_exit(lr);
	ZFS_ACCESSTIME_STAMP(zfsvfs, zp);

	dataset_kstats_update_read_kstats(&zfsvfs->z_kstat, count*PAGE_SIZE);

	zfs_exit(zfsvfs, FTAG);

	if (error != 0)
	return (zfs_vm_pagerret_error);

	VM_CNT_INC(v_vnodein);
	VM_CNT_ADD(v_vnodepgsin, count + pgsin_b + pgsin_a);
	if (rbehind != NULL)
	*rbehind = pgsin_b;
	if (rahead != NULL)
	*rahead = pgsin_a;
	return (zfs_vm_pagerret_ok);
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_getpages_args {
	struct vnode *a_vp;
	vm_page_t *a_m;
	int a_count;
	int *a_rbehind;
	int *a_rahead;
	};
	#endif

	static int
	zfs_freebsd_getpages(struct vop_getpages_args *ap)
	{

	return (zfs_getpages(ap->a_vp, ap->a_m, ap->a_count, ap->a_rbehind,
	ap->a_rahead));
	}

	static int
	zfs_putpages(struct vnode vp, vm_page_t ma, size_t len, int flags,
	int *rtvals)
	{
	znode_t *zp = VTOZ(vp);
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	zfs_locked_range_t *lr;
	dmu_tx_t *tx;
	struct sf_buf *sf;
	vm_object_t object;
	vm_page_t m;
	caddr_t va;
	size_t tocopy;
	size_t lo_len;
	vm_ooffset_t lo_off;
	vm_ooffset_t off;
	uint_t blksz;
	int ncount;
	int pcount;
	int err;
	int i;

	object = vp->v_object;
	KASSERT(ma[0]->object == object, ("mismatching object"));
	KASSERT(len > 0 && (len & PAGE_MASK) == 0, ("unexpected length"));

	pcount = btoc(len);
	ncount = pcount;
	for (i = 0; i < pcount; i++)
	rtvals[i] = zfs_vm_pagerret_error;

	if (zfs_enter_verify_zp(zfsvfs, zp, FTAG) != 0)
	return (zfs_vm_pagerret_error);

	off = IDX_TO_OFF(ma[0]->pindex);
	blksz = zp->z_blksz;
	lo_off = rounddown(off, blksz);
	lo_len = roundup(len + (off - lo_off), blksz);
	lr = zfs_rangelock_enter(&zp->z_rangelock, lo_off, lo_len, RL_WRITER);

	zfs_vmobject_wlock(object);
	if (len + off > object->un_pager.vnp.vnp_size) {
	if (object->un_pager.vnp.vnp_size > off) {
	int pgoff;

	len = object->un_pager.vnp.vnp_size - off;
	ncount = btoc(len);
	if ((pgoff = (int)len & PAGE_MASK) != 0) {
	/*
	* If the object is locked and the following
	* conditions hold, then the page's dirty
	* field cannot be concurrently changed by a
	* pmap operation.
	*/
	m = ma[ncount - 1];
	vm_page_assert_sbusied(m);
	KASSERT(!pmap_page_is_write_mapped(m),
	("zfs_putpages: page %p is not read-only",
	m));
	vm_page_clear_dirty(m, pgoff, PAGE_SIZE -
	pgoff);
	}
	} else {
	len = 0;
	ncount = 0;
	}
	if (ncount < pcount) {
	for (i = ncount; i < pcount; i++) {
	rtvals[i] = zfs_vm_pagerret_bad;
	}
	}
	}
	zfs_vmobject_wunlock(object);

	boolean_t commit = (flags & (zfs_vm_pagerput_sync \|
	zfs_vm_pagerput_inval)) != 0 \|\|
	zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS;

	if (ncount == 0)
	goto out;

	if (zfs_id_overblockquota(zfsvfs, DMU_USERUSED_OBJECT, zp->z_uid) \|\|
	zfs_id_overblockquota(zfsvfs, DMU_GROUPUSED_OBJECT, zp->z_gid) \|\|
	(zp->z_projid != ZFS_DEFAULT_PROJID &&
	zfs_id_overblockquota(zfsvfs, DMU_PROJECTUSED_OBJECT,
	zp->z_projid))) {
	goto out;
	}

	tx = dmu_tx_create(zfsvfs->z_os);
	dmu_tx_hold_write(tx, zp->z_id, off, len);

	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
	zfs_sa_upgrade_txholds(tx, zp);
	err = dmu_tx_assign(tx, TXG_WAIT);
	if (err != 0) {
	dmu_tx_abort(tx);
	goto out;
	}

	if (zp->z_blksz < PAGE_SIZE) {
	for (i = 0; len > 0; off += tocopy, len -= tocopy, i++) {
	tocopy = len > PAGE_SIZE ? PAGE_SIZE : len;
	va = zfs_map_page(ma[i], &sf);
	dmu_write(zfsvfs->z_os, zp->z_id, off, tocopy, va, tx);
	zfs_unmap_page(sf);
	}
	} else {
	err = dmu_write_pages(zfsvfs->z_os, zp->z_id, off, len, ma, tx);
	}

	if (err == 0) {
	uint64_t mtime[2], ctime[2];
	sa_bulk_attr_t bulk[3];
	int count = 0;

	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
	&mtime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
	&ctime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
	&zp->z_pflags, 8);
	zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime);
	err = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
	ASSERT0(err);
	/*
	* XXX we should be passing a callback to undirty
	* but that would make the locking messier
	*/
	zfs_log_write(zfsvfs->z_log, tx, TX_WRITE, zp, off,
	len, commit, NULL, NULL);

	zfs_vmobject_wlock(object);
	for (i = 0; i < ncount; i++) {
	rtvals[i] = zfs_vm_pagerret_ok;
	vm_page_undirty(ma[i]);
	}
	zfs_vmobject_wunlock(object);
	VM_CNT_INC(v_vnodeout);
	VM_CNT_ADD(v_vnodepgsout, ncount);
	}
	dmu_tx_commit(tx);

	out:
	zfs_rangelock_exit(lr);
	if (commit)
	zil_commit(zfsvfs->z_log, zp->z_id);

	dataset_kstats_update_write_kstats(&zfsvfs->z_kstat, len);

	zfs_exit(zfsvfs, FTAG);
	return (rtvals[0]);
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_putpages_args {
	struct vnode *a_vp;
	vm_page_t *a_m;
	int a_count;
	int a_sync;
	int *a_rtvals;
	};
	#endif

	static int
	zfs_freebsd_putpages(struct vop_putpages_args *ap)
	{

	return (zfs_putpages(ap->a_vp, ap->a_m, ap->a_count, ap->a_sync,
	ap->a_rtvals));
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_bmap_args {
	struct vnode *a_vp;
	daddr_t a_bn;
	struct bufobj **a_bop;
	daddr_t *a_bnp;
	int *a_runp;
	int *a_runb;
	};
	#endif

	static int
	zfs_freebsd_bmap(struct vop_bmap_args *ap)
	{

	if (ap->a_bop != NULL)
	*ap->a_bop = &ap->a_vp->v_bufobj;
	if (ap->a_bnp != NULL)
	*ap->a_bnp = ap->a_bn;
	if (ap->a_runp != NULL)
	*ap->a_runp = 0;
	if (ap->a_runb != NULL)
	*ap->a_runb = 0;

	return (0);
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_open_args {
	struct vnode *a_vp;
	int a_mode;
	struct ucred *a_cred;
	struct thread *a_td;
	};
	#endif

	static int
	zfs_freebsd_open(struct vop_open_args *ap)
	{
	vnode_t *vp = ap->a_vp;
	znode_t *zp = VTOZ(vp);
	int error;

	error = zfs_open(&vp, ap->a_mode, ap->a_cred);
	if (error == 0)
	vnode_create_vobject(vp, zp->z_size, ap->a_td);
	return (error);
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_close_args {
	struct vnode *a_vp;
	int a_fflag;
	struct ucred *a_cred;
	struct thread *a_td;
	};
	#endif

	static int
	zfs_freebsd_close(struct vop_close_args *ap)
	{

	return (zfs_close(ap->a_vp, ap->a_fflag, 1, 0, ap->a_cred));
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_ioctl_args {
	struct vnode *a_vp;
	ulong_t a_command;
	caddr_t a_data;
	int a_fflag;
	struct ucred *cred;
	struct thread *td;
	};
	#endif

	static int
	zfs_freebsd_ioctl(struct vop_ioctl_args *ap)
	{

	return (zfs_ioctl(ap->a_vp, ap->a_command, (intptr_t)ap->a_data,
	ap->a_fflag, ap->a_cred, NULL));
	}

	static int
	ioflags(int ioflags)
	{
	int flags = 0;

	if (ioflags & IO_APPEND)
	flags \|= O_APPEND;
	if (ioflags & IO_NDELAY)
	flags \|= O_NONBLOCK;
	if (ioflags & IO_SYNC)
	flags \|= O_SYNC;

	return (flags);
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_read_args {
	struct vnode *a_vp;
	struct uio *a_uio;
	int a_ioflag;
	struct ucred *a_cred;
	};
	#endif

	static int
	zfs_freebsd_read(struct vop_read_args *ap)
	{
	zfs_uio_t uio;
	zfs_uio_init(&uio, ap->a_uio);
	return (zfs_read(VTOZ(ap->a_vp), &uio, ioflags(ap->a_ioflag),
	ap->a_cred));
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_write_args {
	struct vnode *a_vp;
	struct uio *a_uio;
	int a_ioflag;
	struct ucred *a_cred;
	};
	#endif

	static int
	zfs_freebsd_write(struct vop_write_args *ap)
	{
	zfs_uio_t uio;
	zfs_uio_init(&uio, ap->a_uio);
	return (zfs_write(VTOZ(ap->a_vp), &uio, ioflags(ap->a_ioflag),
	ap->a_cred));
	}

	#if __FreeBSD_version >= 1300102
	/*
	* VOP_FPLOOKUP_VEXEC routines are subject to special circumstances, see
	* the comment above cache_fplookup for details.
	*/
	static int
	zfs_freebsd_fplookup_vexec(struct vop_fplookup_vexec_args *v)
	{
	vnode_t *vp;
	znode_t *zp;
	uint64_t pflags;

	vp = v->a_vp;
	zp = VTOZ_SMR(vp);
	if (__predict_false(zp == NULL))
	return (EAGAIN);
	pflags = atomic_load_64(&zp->z_pflags);
	if (pflags & ZFS_AV_QUARANTINED)
	return (EAGAIN);
	if (pflags & ZFS_XATTR)
	return (EAGAIN);
	if ((pflags & ZFS_NO_EXECS_DENIED) == 0)
	return (EAGAIN);
	return (0);
	}
	#endif

	#if __FreeBSD_version >= 1300139
	static int
	zfs_freebsd_fplookup_symlink(struct vop_fplookup_symlink_args *v)
	{
	vnode_t *vp;
	znode_t *zp;
	char *target;

	vp = v->a_vp;
	zp = VTOZ_SMR(vp);
	if (__predict_false(zp == NULL)) {
	return (EAGAIN);
	}

	target = atomic_load_consume_ptr(&zp->z_cached_symlink);
	if (target == NULL) {
	return (EAGAIN);
	}
	return (cache_symlink_resolve(v->a_fpl, target, strlen(target)));
	}
	#endif

	#ifndef _SYS_SYSPROTO_H_
	struct vop_access_args {
	struct vnode *a_vp;
	accmode_t a_accmode;
	struct ucred *a_cred;
	struct thread *a_td;
	};
	#endif

	static int
	zfs_freebsd_access(struct vop_access_args *ap)
	{
	vnode_t *vp = ap->a_vp;
	znode_t *zp = VTOZ(vp);
	accmode_t accmode;
	int error = 0;


	if (ap->a_accmode == VEXEC) {
	if (zfs_fastaccesschk_execute(zp, ap->a_cred) == 0)
	return (0);
	}

	/*
	* ZFS itself only knowns about VREAD, VWRITE, VEXEC and VAPPEND,
	*/
	accmode = ap->a_accmode & (VREAD\|VWRITE\|VEXEC\|VAPPEND);
	if (accmode != 0)
	error = zfs_access(zp, accmode, 0, ap->a_cred);

	/*
	* VADMIN has to be handled by vaccess().
	*/
	if (error == 0) {
	accmode = ap->a_accmode & ~(VREAD\|VWRITE\|VEXEC\|VAPPEND);
	if (accmode != 0) {
	#if __FreeBSD_version >= 1300105
	error = vaccess(vp->v_type, zp->z_mode, zp->z_uid,
	zp->z_gid, accmode, ap->a_cred);
	#else
	error = vaccess(vp->v_type, zp->z_mode, zp->z_uid,
	zp->z_gid, accmode, ap->a_cred, NULL);
	#endif
	}
	}

	/*
	* For VEXEC, ensure that at least one execute bit is set for
	* non-directories.
	*/
	if (error == 0 && (ap->a_accmode & VEXEC) != 0 && vp->v_type != VDIR &&
	(zp->z_mode & (S_IXUSR \| S_IXGRP \| S_IXOTH)) == 0) {
	error = EACCES;
	}

	return (error);
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_lookup_args {
	struct vnode *a_dvp;
	struct vnode **a_vpp;
	struct componentname *a_cnp;
	};
	#endif

	static int
	zfs_freebsd_lookup(struct vop_lookup_args *ap, boolean_t cached)
	{
	struct componentname *cnp = ap->a_cnp;
	char nm[NAME_MAX + 1];

	ASSERT3U(cnp->cn_namelen, <, sizeof (nm));
	strlcpy(nm, cnp->cn_nameptr, MIN(cnp->cn_namelen + 1, sizeof (nm)));

	return (zfs_lookup(ap->a_dvp, nm, ap->a_vpp, cnp, cnp->cn_nameiop,
	cnp->cn_cred, 0, cached));
	}

	static int
	zfs_freebsd_cachedlookup(struct vop_cachedlookup_args *ap)
	{

	return (zfs_freebsd_lookup((struct vop_lookup_args *)ap, B_TRUE));
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_lookup_args {
	struct vnode *a_dvp;
	struct vnode **a_vpp;
	struct componentname *a_cnp;
	};
	#endif

	static int
	zfs_cache_lookup(struct vop_lookup_args *ap)
	{
	zfsvfs_t *zfsvfs;

	zfsvfs = ap->a_dvp->v_mount->mnt_data;
	if (zfsvfs->z_use_namecache)
	return (vfs_cache_lookup(ap));
	else
	return (zfs_freebsd_lookup(ap, B_FALSE));
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_create_args {
	struct vnode *a_dvp;
	struct vnode **a_vpp;
	struct componentname *a_cnp;
	struct vattr *a_vap;
	};
	#endif

	static int
	zfs_freebsd_create(struct vop_create_args *ap)
	{
	zfsvfs_t *zfsvfs;
	struct componentname *cnp = ap->a_cnp;
	vattr_t *vap = ap->a_vap;
	znode_t *zp = NULL;
	int rc, mode;

	#if __FreeBSD_version < 1400068
	ASSERT(cnp->cn_flags & SAVENAME);
	#endif

	vattr_init_mask(vap);
	mode = vap->va_mode & ALLPERMS;
	zfsvfs = ap->a_dvp->v_mount->mnt_data;
	*ap->a_vpp = NULL;

	rc = zfs_create(VTOZ(ap->a_dvp), cnp->cn_nameptr, vap, 0, mode,
	&zp, cnp->cn_cred, 0 /* flag /, NULL / vsecattr */, NULL);
	if (rc == 0)
	*ap->a_vpp = ZTOV(zp);
	if (zfsvfs->z_use_namecache &&
	rc == 0 && (cnp->cn_flags & MAKEENTRY) != 0)
	cache_enter(ap->a_dvp, *ap->a_vpp, cnp);

	return (rc);
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_remove_args {
	struct vnode *a_dvp;
	struct vnode *a_vp;
	struct componentname *a_cnp;
	};
	#endif

	static int
	zfs_freebsd_remove(struct vop_remove_args *ap)
	{

	#if __FreeBSD_version < 1400068
	ASSERT(ap->a_cnp->cn_flags & SAVENAME);
	#endif

	return (zfs_remove_(ap->a_dvp, ap->a_vp, ap->a_cnp->cn_nameptr,
	ap->a_cnp->cn_cred));
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_mkdir_args {
	struct vnode *a_dvp;
	struct vnode **a_vpp;
	struct componentname *a_cnp;
	struct vattr *a_vap;
	};
	#endif

	static int
	zfs_freebsd_mkdir(struct vop_mkdir_args *ap)
	{
	vattr_t *vap = ap->a_vap;
	znode_t *zp = NULL;
	int rc;

	#if __FreeBSD_version < 1400068
	ASSERT(ap->a_cnp->cn_flags & SAVENAME);
	#endif

	vattr_init_mask(vap);
	*ap->a_vpp = NULL;

	rc = zfs_mkdir(VTOZ(ap->a_dvp), ap->a_cnp->cn_nameptr, vap, &zp,
	ap->a_cnp->cn_cred, 0, NULL, NULL);

	if (rc == 0)
	*ap->a_vpp = ZTOV(zp);
	return (rc);
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_rmdir_args {
	struct vnode *a_dvp;
	struct vnode *a_vp;
	struct componentname *a_cnp;
	};
	#endif

	static int
	zfs_freebsd_rmdir(struct vop_rmdir_args *ap)
	{
	struct componentname *cnp = ap->a_cnp;

	#if __FreeBSD_version < 1400068
	ASSERT(cnp->cn_flags & SAVENAME);
	#endif

	return (zfs_rmdir_(ap->a_dvp, ap->a_vp, cnp->cn_nameptr, cnp->cn_cred));
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_readdir_args {
	struct vnode *a_vp;
	struct uio *a_uio;
	struct ucred *a_cred;
	int *a_eofflag;
	int *a_ncookies;
	cookie_t **a_cookies;
	};
	#endif

	static int
	zfs_freebsd_readdir(struct vop_readdir_args *ap)
	{
	zfs_uio_t uio;
	zfs_uio_init(&uio, ap->a_uio);
	return (zfs_readdir(ap->a_vp, &uio, ap->a_cred, ap->a_eofflag,
	ap->a_ncookies, ap->a_cookies));
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_fsync_args {
	struct vnode *a_vp;
	int a_waitfor;
	struct thread *a_td;
	};
	#endif

	static int
	zfs_freebsd_fsync(struct vop_fsync_args *ap)
	{

	return (zfs_fsync(VTOZ(ap->a_vp), 0, ap->a_td->td_ucred));
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_getattr_args {
	struct vnode *a_vp;
	struct vattr *a_vap;
	struct ucred *a_cred;
	};
	#endif

	static int
	zfs_freebsd_getattr(struct vop_getattr_args *ap)
	{
	vattr_t *vap = ap->a_vap;
	xvattr_t xvap;
	ulong_t fflags = 0;
	int error;

	xva_init(&xvap);
	xvap.xva_vattr = *vap;
	xvap.xva_vattr.va_mask \|= AT_XVATTR;

	/* Convert chflags into ZFS-type flags. */
	/* XXX: what about SF_SETTABLE?. */
	XVA_SET_REQ(&xvap, XAT_IMMUTABLE);
	XVA_SET_REQ(&xvap, XAT_APPENDONLY);
	XVA_SET_REQ(&xvap, XAT_NOUNLINK);
	XVA_SET_REQ(&xvap, XAT_NODUMP);
	XVA_SET_REQ(&xvap, XAT_READONLY);
	XVA_SET_REQ(&xvap, XAT_ARCHIVE);
	XVA_SET_REQ(&xvap, XAT_SYSTEM);
	XVA_SET_REQ(&xvap, XAT_HIDDEN);
	XVA_SET_REQ(&xvap, XAT_REPARSE);
	XVA_SET_REQ(&xvap, XAT_OFFLINE);
	XVA_SET_REQ(&xvap, XAT_SPARSE);

	error = zfs_getattr(ap->a_vp, (vattr_t *)&xvap, 0, ap->a_cred);
	if (error != 0)
	return (error);

	/* Convert ZFS xattr into chflags. */
	#define FLAG_CHECK(fflag, xflag, xfield) do { \
	if (XVA_ISSET_RTN(&xvap, (xflag)) && (xfield) != 0) \
	fflags \|= (fflag); \
	} while (0)
	FLAG_CHECK(SF_IMMUTABLE, XAT_IMMUTABLE,
	xvap.xva_xoptattrs.xoa_immutable);
	FLAG_CHECK(SF_APPEND, XAT_APPENDONLY,
	xvap.xva_xoptattrs.xoa_appendonly);
	FLAG_CHECK(SF_NOUNLINK, XAT_NOUNLINK,
	xvap.xva_xoptattrs.xoa_nounlink);
	FLAG_CHECK(UF_ARCHIVE, XAT_ARCHIVE,
	xvap.xva_xoptattrs.xoa_archive);
	FLAG_CHECK(UF_NODUMP, XAT_NODUMP,
	xvap.xva_xoptattrs.xoa_nodump);
	FLAG_CHECK(UF_READONLY, XAT_READONLY,
	xvap.xva_xoptattrs.xoa_readonly);
	FLAG_CHECK(UF_SYSTEM, XAT_SYSTEM,
	xvap.xva_xoptattrs.xoa_system);
	FLAG_CHECK(UF_HIDDEN, XAT_HIDDEN,
	xvap.xva_xoptattrs.xoa_hidden);
	FLAG_CHECK(UF_REPARSE, XAT_REPARSE,
	xvap.xva_xoptattrs.xoa_reparse);
	FLAG_CHECK(UF_OFFLINE, XAT_OFFLINE,
	xvap.xva_xoptattrs.xoa_offline);
	FLAG_CHECK(UF_SPARSE, XAT_SPARSE,
	xvap.xva_xoptattrs.xoa_sparse);

	#undef FLAG_CHECK
	*vap = xvap.xva_vattr;
	vap->va_flags = fflags;
	return (0);
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_setattr_args {
	struct vnode *a_vp;
	struct vattr *a_vap;
	struct ucred *a_cred;
	};
	#endif

	static int
	zfs_freebsd_setattr(struct vop_setattr_args *ap)
	{
	vnode_t *vp = ap->a_vp;
	vattr_t *vap = ap->a_vap;
	cred_t *cred = ap->a_cred;
	xvattr_t xvap;
	ulong_t fflags;
	uint64_t zflags;

	vattr_init_mask(vap);
	vap->va_mask &= ~AT_NOSET;

	xva_init(&xvap);
	xvap.xva_vattr = *vap;

	zflags = VTOZ(vp)->z_pflags;

	if (vap->va_flags != VNOVAL) {
	zfsvfs_t *zfsvfs = VTOZ(vp)->z_zfsvfs;
	int error;

	if (zfsvfs->z_use_fuids == B_FALSE)
	return (EOPNOTSUPP);

	fflags = vap->va_flags;
	/*
	* XXX KDM
	* We need to figure out whether it makes sense to allow
	* UF_REPARSE through, since we don't really have other
	* facilities to handle reparse points and zfs_setattr()
	* doesn't currently allow setting that attribute anyway.
	*/
	if ((fflags & ~(SF_IMMUTABLE\|SF_APPEND\|SF_NOUNLINK\|UF_ARCHIVE\|
	UF_NODUMP\|UF_SYSTEM\|UF_HIDDEN\|UF_READONLY\|UF_REPARSE\|
	UF_OFFLINE\|UF_SPARSE)) != 0)
	return (EOPNOTSUPP);
	/*
	* Unprivileged processes are not permitted to unset system
	* flags, or modify flags if any system flags are set.
	* Privileged non-jail processes may not modify system flags
	* if securelevel > 0 and any existing system flags are set.
	* Privileged jail processes behave like privileged non-jail
	* processes if the PR_ALLOW_CHFLAGS permission bit is set;
	* otherwise, they behave like unprivileged processes.
	*/
	if (secpolicy_fs_owner(vp->v_mount, cred) == 0 \|\|
	spl_priv_check_cred(cred, PRIV_VFS_SYSFLAGS) == 0) {
	if (zflags &
	(ZFS_IMMUTABLE \| ZFS_APPENDONLY \| ZFS_NOUNLINK)) {
	error = securelevel_gt(cred, 0);
	if (error != 0)
	return (error);
	}
	} else {
	/*
	* Callers may only modify the file flags on
	* objects they have VADMIN rights for.
	*/
	if ((error = VOP_ACCESS(vp, VADMIN, cred,
	curthread)) != 0)
	return (error);
	if (zflags &
	(ZFS_IMMUTABLE \| ZFS_APPENDONLY \|
	ZFS_NOUNLINK)) {
	return (EPERM);
	}
	if (fflags &
	(SF_IMMUTABLE \| SF_APPEND \| SF_NOUNLINK)) {
	return (EPERM);
	}
	}

	#define FLAG_CHANGE(fflag, zflag, xflag, xfield) do { \
	if (((fflags & (fflag)) && !(zflags & (zflag))) \|\| \
	((zflags & (zflag)) && !(fflags & (fflag)))) { \
	XVA_SET_REQ(&xvap, (xflag)); \
	(xfield) = ((fflags & (fflag)) != 0); \
	} \
	} while (0)
	/* Convert chflags into ZFS-type flags. */
	/* XXX: what about SF_SETTABLE?. */
	FLAG_CHANGE(SF_IMMUTABLE, ZFS_IMMUTABLE, XAT_IMMUTABLE,
	xvap.xva_xoptattrs.xoa_immutable);
	FLAG_CHANGE(SF_APPEND, ZFS_APPENDONLY, XAT_APPENDONLY,
	xvap.xva_xoptattrs.xoa_appendonly);
	FLAG_CHANGE(SF_NOUNLINK, ZFS_NOUNLINK, XAT_NOUNLINK,
	xvap.xva_xoptattrs.xoa_nounlink);
	FLAG_CHANGE(UF_ARCHIVE, ZFS_ARCHIVE, XAT_ARCHIVE,
	xvap.xva_xoptattrs.xoa_archive);
	FLAG_CHANGE(UF_NODUMP, ZFS_NODUMP, XAT_NODUMP,
	xvap.xva_xoptattrs.xoa_nodump);
	FLAG_CHANGE(UF_READONLY, ZFS_READONLY, XAT_READONLY,
	xvap.xva_xoptattrs.xoa_readonly);
	FLAG_CHANGE(UF_SYSTEM, ZFS_SYSTEM, XAT_SYSTEM,
	xvap.xva_xoptattrs.xoa_system);
	FLAG_CHANGE(UF_HIDDEN, ZFS_HIDDEN, XAT_HIDDEN,
	xvap.xva_xoptattrs.xoa_hidden);
	FLAG_CHANGE(UF_REPARSE, ZFS_REPARSE, XAT_REPARSE,
	xvap.xva_xoptattrs.xoa_reparse);
	FLAG_CHANGE(UF_OFFLINE, ZFS_OFFLINE, XAT_OFFLINE,
	xvap.xva_xoptattrs.xoa_offline);
	FLAG_CHANGE(UF_SPARSE, ZFS_SPARSE, XAT_SPARSE,
	xvap.xva_xoptattrs.xoa_sparse);
	#undef FLAG_CHANGE
	}
	if (vap->va_birthtime.tv_sec != VNOVAL) {
	xvap.xva_vattr.va_mask \|= AT_XVATTR;
	XVA_SET_REQ(&xvap, XAT_CREATETIME);
	}
	return (zfs_setattr(VTOZ(vp), (vattr_t *)&xvap, 0, cred, NULL));
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_rename_args {
	struct vnode *a_fdvp;
	struct vnode *a_fvp;
	struct componentname *a_fcnp;
	struct vnode *a_tdvp;
	struct vnode *a_tvp;
	struct componentname *a_tcnp;
	};
	#endif

	static int
	zfs_freebsd_rename(struct vop_rename_args *ap)
	{
	vnode_t *fdvp = ap->a_fdvp;
	vnode_t *fvp = ap->a_fvp;
	vnode_t *tdvp = ap->a_tdvp;
	vnode_t *tvp = ap->a_tvp;
	int error;

	#if __FreeBSD_version < 1400068
	ASSERT(ap->a_fcnp->cn_flags & (SAVENAME\|SAVESTART));
	ASSERT(ap->a_tcnp->cn_flags & (SAVENAME\|SAVESTART));
	#endif

	error = zfs_do_rename(fdvp, &fvp, ap->a_fcnp, tdvp, &tvp,
	ap->a_tcnp, ap->a_fcnp->cn_cred);

	vrele(fdvp);
	vrele(fvp);
	vrele(tdvp);
	if (tvp != NULL)
	vrele(tvp);

	return (error);
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_symlink_args {
	struct vnode *a_dvp;
	struct vnode **a_vpp;
	struct componentname *a_cnp;
	struct vattr *a_vap;
	char *a_target;
	};
	#endif

	static int
	zfs_freebsd_symlink(struct vop_symlink_args *ap)
	{
	struct componentname *cnp = ap->a_cnp;
	vattr_t *vap = ap->a_vap;
	znode_t *zp = NULL;
	#if __FreeBSD_version >= 1300139
	char *symlink;
	size_t symlink_len;
	#endif
	int rc;

	#if __FreeBSD_version < 1400068
	ASSERT(cnp->cn_flags & SAVENAME);
	#endif

	vap->va_type = VLNK; /* FreeBSD: Syscall only sets va_mode. */
	vattr_init_mask(vap);
	*ap->a_vpp = NULL;

	rc = zfs_symlink(VTOZ(ap->a_dvp), cnp->cn_nameptr, vap,
	ap->a_target, &zp, cnp->cn_cred, 0 /* flags */, NULL);
	if (rc == 0) {
	*ap->a_vpp = ZTOV(zp);
	ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
	#if __FreeBSD_version >= 1300139
	MPASS(zp->z_cached_symlink == NULL);
	symlink_len = strlen(ap->a_target);
	symlink = cache_symlink_alloc(symlink_len + 1, M_WAITOK);
	if (symlink != NULL) {
	memcpy(symlink, ap->a_target, symlink_len);
	symlink[symlink_len] = '\0';
	atomic_store_rel_ptr((uintptr_t *)&zp->z_cached_symlink,
	(uintptr_t)symlink);
	}
	#endif
	}
	return (rc);
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_readlink_args {
	struct vnode *a_vp;
	struct uio *a_uio;
	struct ucred *a_cred;
	};
	#endif

	static int
	zfs_freebsd_readlink(struct vop_readlink_args *ap)
	{
	zfs_uio_t uio;
	int error;
	#if __FreeBSD_version >= 1300139
	znode_t *zp = VTOZ(ap->a_vp);
	char symlink, base;
	size_t symlink_len;
	bool trycache;
	#endif

	zfs_uio_init(&uio, ap->a_uio);
	#if __FreeBSD_version >= 1300139
	trycache = false;
	if (zfs_uio_segflg(&uio) == UIO_SYSSPACE &&
	zfs_uio_iovcnt(&uio) == 1) {
	base = zfs_uio_iovbase(&uio, 0);
	symlink_len = zfs_uio_iovlen(&uio, 0);
	trycache = true;
	}
	#endif
	error = zfs_readlink(ap->a_vp, &uio, ap->a_cred, NULL);
	#if __FreeBSD_version >= 1300139
	if (atomic_load_ptr(&zp->z_cached_symlink) != NULL \|\|
	error != 0 \|\| !trycache) {
	return (error);
	}
	symlink_len -= zfs_uio_resid(&uio);
	symlink = cache_symlink_alloc(symlink_len + 1, M_WAITOK);
	if (symlink != NULL) {
	memcpy(symlink, base, symlink_len);
	symlink[symlink_len] = '\0';
	if (!atomic_cmpset_rel_ptr((uintptr_t *)&zp->z_cached_symlink,
	(uintptr_t)NULL, (uintptr_t)symlink)) {
	cache_symlink_free(symlink, symlink_len + 1);
	}
	}
	#endif
	return (error);
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_link_args {
	struct vnode *a_tdvp;
	struct vnode *a_vp;
	struct componentname *a_cnp;
	};
	#endif

	static int
	zfs_freebsd_link(struct vop_link_args *ap)
	{
	struct componentname *cnp = ap->a_cnp;
	vnode_t *vp = ap->a_vp;
	vnode_t *tdvp = ap->a_tdvp;

	if (tdvp->v_mount != vp->v_mount)
	return (EXDEV);

	#if __FreeBSD_version < 1400068
	ASSERT(cnp->cn_flags & SAVENAME);
	#endif

	return (zfs_link(VTOZ(tdvp), VTOZ(vp),
	cnp->cn_nameptr, cnp->cn_cred, 0));
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_inactive_args {
	struct vnode *a_vp;
	struct thread *a_td;
	};
	#endif

	static int
	zfs_freebsd_inactive(struct vop_inactive_args *ap)
	{
	vnode_t *vp = ap->a_vp;

	#if __FreeBSD_version >= 1300123
	zfs_inactive(vp, curthread->td_ucred, NULL);
	#else
	zfs_inactive(vp, ap->a_td->td_ucred, NULL);
	#endif
	return (0);
	}

	#if __FreeBSD_version >= 1300042
	#ifndef _SYS_SYSPROTO_H_
	struct vop_need_inactive_args {
	struct vnode *a_vp;
	struct thread *a_td;
	};
	#endif

	static int
	zfs_freebsd_need_inactive(struct vop_need_inactive_args *ap)
	{
	vnode_t *vp = ap->a_vp;
	znode_t *zp = VTOZ(vp);
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	int need;

	if (vn_need_pageq_flush(vp))
	return (1);

	if (!ZFS_TEARDOWN_INACTIVE_TRY_ENTER_READ(zfsvfs))
	return (1);
	need = (zp->z_sa_hdl == NULL \|\| zp->z_unlinked \|\| zp->z_atime_dirty);
	ZFS_TEARDOWN_INACTIVE_EXIT_READ(zfsvfs);

	return (need);
	}
	#endif

	#ifndef _SYS_SYSPROTO_H_
	struct vop_reclaim_args {
	struct vnode *a_vp;
	struct thread *a_td;
	};
	#endif

	static int
	zfs_freebsd_reclaim(struct vop_reclaim_args *ap)
	{
	vnode_t *vp = ap->a_vp;
	znode_t *zp = VTOZ(vp);
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;

	ASSERT3P(zp, !=, NULL);

	#if __FreeBSD_version < 1300042
	/* Destroy the vm object and flush associated pages. */
	vnode_destroy_vobject(vp);
	#endif
	/*
	* z_teardown_inactive_lock protects from a race with
	* zfs_znode_dmu_fini in zfsvfs_teardown during
	* force unmount.
	*/
	ZFS_TEARDOWN_INACTIVE_ENTER_READ(zfsvfs);
	if (zp->z_sa_hdl == NULL)
	zfs_znode_free(zp);
	else
	zfs_zinactive(zp);
	ZFS_TEARDOWN_INACTIVE_EXIT_READ(zfsvfs);

	vp->v_data = NULL;
	return (0);
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_fid_args {
	struct vnode *a_vp;
	struct fid *a_fid;
	};
	#endif

	static int
	zfs_freebsd_fid(struct vop_fid_args *ap)
	{

	return (zfs_fid(ap->a_vp, (void *)ap->a_fid, NULL));
	}


	#ifndef _SYS_SYSPROTO_H_
	struct vop_pathconf_args {
	struct vnode *a_vp;
	int a_name;
	register_t *a_retval;
	} *ap;
	#endif

	static int
	zfs_freebsd_pathconf(struct vop_pathconf_args *ap)
	{
	ulong_t val;
	int error;

	error = zfs_pathconf(ap->a_vp, ap->a_name, &val,
	curthread->td_ucred, NULL);
	if (error == 0) {
	*ap->a_retval = val;
	return (error);
	}
	if (error != EOPNOTSUPP)
	return (error);

	switch (ap->a_name) {
	case _PC_NAME_MAX:
	*ap->a_retval = NAME_MAX;
	return (0);
	#if __FreeBSD_version >= 1400032
	case _PC_DEALLOC_PRESENT:
	*ap->a_retval = 1;
	return (0);
	#endif
	case _PC_PIPE_BUF:
	if (ap->a_vp->v_type == VDIR \|\| ap->a_vp->v_type == VFIFO) {
	*ap->a_retval = PIPE_BUF;
	return (0);
	}
	return (EINVAL);
	default:
	return (vop_stdpathconf(ap));
	}
	}

	static int zfs_xattr_compat = 1;

	static int
	zfs_check_attrname(const char *name)
	{
	/* We don't allow '/' character in attribute name. */
	if (strchr(name, '/') != NULL)
	return (SET_ERROR(EINVAL));
	/* We don't allow attribute names that start with a namespace prefix. */
	if (ZFS_XA_NS_PREFIX_FORBIDDEN(name))
	return (SET_ERROR(EINVAL));
	return (0);
	}

	/*
	* FreeBSD's extended attributes namespace defines file name prefix for ZFS'
	* extended attribute name:
	*
	* NAMESPACE XATTR_COMPAT PREFIX
	* system * freebsd:system:
	* user 1 (none, can be used to access ZFS
	* fsattr(5) attributes created on Solaris)
	* user 0 user.
	*/
	static int
	zfs_create_attrname(int attrnamespace, const char name, char attrname,
	size_t size, boolean_t compat)
	{
	const char namespace, prefix, *suffix;

	memset(attrname, 0, size);

	switch (attrnamespace) {
	case EXTATTR_NAMESPACE_USER:
	if (compat) {
	/*
	* This is the default namespace by which we can access
	* all attributes created on Solaris.
	*/
	prefix = namespace = suffix = "";
	} else {
	/*
	* This is compatible with the user namespace encoding
	* on Linux prior to xattr_compat, but nothing
	* else.
	*/
	prefix = "";
	namespace = "user";
	suffix = ".";
	}
	break;
	case EXTATTR_NAMESPACE_SYSTEM:
	prefix = "freebsd:";
	namespace = EXTATTR_NAMESPACE_SYSTEM_STRING;
	suffix = ":";
	break;
	case EXTATTR_NAMESPACE_EMPTY:
	default:
	return (SET_ERROR(EINVAL));
	}
	if (snprintf(attrname, size, "%s%s%s%s", prefix, namespace, suffix,
	name) >= size) {
	return (SET_ERROR(ENAMETOOLONG));
	}
	return (0);
	}

	static int
	zfs_ensure_xattr_cached(znode_t *zp)
	{
	int error = 0;

	ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock));

	if (zp->z_xattr_cached != NULL)
	return (0);

	if (rw_write_held(&zp->z_xattr_lock))
	return (zfs_sa_get_xattr(zp));

	if (!rw_tryupgrade(&zp->z_xattr_lock)) {
	rw_exit(&zp->z_xattr_lock);
	rw_enter(&zp->z_xattr_lock, RW_WRITER);
	}
	if (zp->z_xattr_cached == NULL)
	error = zfs_sa_get_xattr(zp);
	rw_downgrade(&zp->z_xattr_lock);
	return (error);
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_getextattr {
	IN struct vnode *a_vp;
	IN int a_attrnamespace;
	IN const char *a_name;
	INOUT struct uio *a_uio;
	OUT size_t *a_size;
	IN struct ucred *a_cred;
	IN struct thread *a_td;
	};
	#endif

	static int
	zfs_getextattr_dir(struct vop_getextattr_args ap, const char attrname)
	{
	struct thread *td = ap->a_td;
	struct nameidata nd;
	struct vattr va;
	vnode_t xvp = NULL, vp;
	int error, flags;

	error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred,
	LOOKUP_XATTR, B_FALSE);
	if (error != 0)
	return (error);

	flags = FREAD;
	#if __FreeBSD_version < 1400043
	NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, attrname,
	xvp, td);
	#else
	NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, attrname, xvp);
	#endif
	error = vn_open_cred(&nd, &flags, 0, VN_OPEN_INVFS, ap->a_cred, NULL);
	if (error != 0)
	return (SET_ERROR(error));
	vp = nd.ni_vp;
	NDFREE_PNBUF(&nd);

	if (ap->a_size != NULL) {
	error = VOP_GETATTR(vp, &va, ap->a_cred);
	if (error == 0)
	*ap->a_size = (size_t)va.va_size;
	} else if (ap->a_uio != NULL)
	error = VOP_READ(vp, ap->a_uio, IO_UNIT, ap->a_cred);

	VOP_UNLOCK1(vp);
	vn_close(vp, flags, ap->a_cred, td);
	return (error);
	}

	static int
	zfs_getextattr_sa(struct vop_getextattr_args ap, const char attrname)
	{
	znode_t *zp = VTOZ(ap->a_vp);
	uchar_t *nv_value;
	uint_t nv_size;
	int error;

	error = zfs_ensure_xattr_cached(zp);
	if (error != 0)
	return (error);

	ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock));
	ASSERT3P(zp->z_xattr_cached, !=, NULL);

	error = nvlist_lookup_byte_array(zp->z_xattr_cached, attrname,
	&nv_value, &nv_size);
	if (error != 0)
	return (SET_ERROR(error));

	if (ap->a_size != NULL)
	*ap->a_size = nv_size;
	else if (ap->a_uio != NULL)
	error = uiomove(nv_value, nv_size, ap->a_uio);
	if (error != 0)
	return (SET_ERROR(error));

	return (0);
	}

	static int
	zfs_getextattr_impl(struct vop_getextattr_args *ap, boolean_t compat)
	{
	znode_t *zp = VTOZ(ap->a_vp);
	zfsvfs_t *zfsvfs = ZTOZSB(zp);
	char attrname[EXTATTR_MAXNAMELEN+1];
	int error;

	error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
	sizeof (attrname), compat);
	if (error != 0)
	return (error);

	error = ENOENT;
	if (zfsvfs->z_use_sa && zp->z_is_sa)
	error = zfs_getextattr_sa(ap, attrname);
	if (error == ENOENT)
	error = zfs_getextattr_dir(ap, attrname);
	return (error);
	}

	/*
	* Vnode operation to retrieve a named extended attribute.
	*/
	static int
	zfs_getextattr(struct vop_getextattr_args *ap)
	{
	znode_t *zp = VTOZ(ap->a_vp);
	zfsvfs_t *zfsvfs = ZTOZSB(zp);
	int error;

	/*
	* If the xattr property is off, refuse the request.
	*/
	if (!(zfsvfs->z_flags & ZSB_XATTR))
	return (SET_ERROR(EOPNOTSUPP));

	error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
	ap->a_cred, ap->a_td, VREAD);
	if (error != 0)
	return (SET_ERROR(error));

	error = zfs_check_attrname(ap->a_name);
	if (error != 0)
	return (error);

	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (error);
	error = ENOENT;
	rw_enter(&zp->z_xattr_lock, RW_READER);

	error = zfs_getextattr_impl(ap, zfs_xattr_compat);
	if ((error == ENOENT \|\| error == ENOATTR) &&
	ap->a_attrnamespace == EXTATTR_NAMESPACE_USER) {
	/*
	* Fall back to the alternate namespace format if we failed to
	* find a user xattr.
	*/
	error = zfs_getextattr_impl(ap, !zfs_xattr_compat);
	}

	rw_exit(&zp->z_xattr_lock);
	zfs_exit(zfsvfs, FTAG);
	if (error == ENOENT)
	error = SET_ERROR(ENOATTR);
	return (error);
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_deleteextattr {
	IN struct vnode *a_vp;
	IN int a_attrnamespace;
	IN const char *a_name;
	IN struct ucred *a_cred;
	IN struct thread *a_td;
	};
	#endif

	static int
	zfs_deleteextattr_dir(struct vop_deleteextattr_args ap, const char attrname)
	{
	struct nameidata nd;
	vnode_t xvp = NULL, vp;
	int error;

	error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred,
	LOOKUP_XATTR, B_FALSE);
	if (error != 0)
	return (error);

	#if __FreeBSD_version < 1400043
	NDINIT_ATVP(&nd, DELETE, NOFOLLOW \| LOCKPARENT \| LOCKLEAF,
	UIO_SYSSPACE, attrname, xvp, ap->a_td);
	#else
	NDINIT_ATVP(&nd, DELETE, NOFOLLOW \| LOCKPARENT \| LOCKLEAF,
	UIO_SYSSPACE, attrname, xvp);
	#endif
	error = namei(&nd);
	if (error != 0)
	return (SET_ERROR(error));

	vp = nd.ni_vp;
	error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
	NDFREE_PNBUF(&nd);

	vput(nd.ni_dvp);
	if (vp == nd.ni_dvp)
	vrele(vp);
	else
	vput(vp);

	return (error);
	}

	static int
	zfs_deleteextattr_sa(struct vop_deleteextattr_args ap, const char attrname)
	{
	znode_t *zp = VTOZ(ap->a_vp);
	nvlist_t *nvl;
	int error;

	error = zfs_ensure_xattr_cached(zp);
	if (error != 0)
	return (error);

	ASSERT(RW_WRITE_HELD(&zp->z_xattr_lock));
	ASSERT3P(zp->z_xattr_cached, !=, NULL);

	nvl = zp->z_xattr_cached;
	error = nvlist_remove(nvl, attrname, DATA_TYPE_BYTE_ARRAY);
	if (error != 0)
	error = SET_ERROR(error);
	else
	error = zfs_sa_set_xattr(zp, attrname, NULL, 0);
	if (error != 0) {
	zp->z_xattr_cached = NULL;
	nvlist_free(nvl);
	}
	return (error);
	}

	static int
	zfs_deleteextattr_impl(struct vop_deleteextattr_args *ap, boolean_t compat)
	{
	znode_t *zp = VTOZ(ap->a_vp);
	zfsvfs_t *zfsvfs = ZTOZSB(zp);
	char attrname[EXTATTR_MAXNAMELEN+1];
	int error;

	error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
	sizeof (attrname), compat);
	if (error != 0)
	return (error);

	error = ENOENT;
	if (zfsvfs->z_use_sa && zp->z_is_sa)
	error = zfs_deleteextattr_sa(ap, attrname);
	if (error == ENOENT)
	error = zfs_deleteextattr_dir(ap, attrname);
	return (error);
	}

	/*
	* Vnode operation to remove a named attribute.
	*/
	static int
	zfs_deleteextattr(struct vop_deleteextattr_args *ap)
	{
	znode_t *zp = VTOZ(ap->a_vp);
	zfsvfs_t *zfsvfs = ZTOZSB(zp);
	int error;

	/*
	* If the xattr property is off, refuse the request.
	*/
	if (!(zfsvfs->z_flags & ZSB_XATTR))
	return (SET_ERROR(EOPNOTSUPP));

	error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
	ap->a_cred, ap->a_td, VWRITE);
	if (error != 0)
	return (SET_ERROR(error));

	error = zfs_check_attrname(ap->a_name);
	if (error != 0)
	return (error);

	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (error);
	rw_enter(&zp->z_xattr_lock, RW_WRITER);

	error = zfs_deleteextattr_impl(ap, zfs_xattr_compat);
	if ((error == ENOENT \|\| error == ENOATTR) &&
	ap->a_attrnamespace == EXTATTR_NAMESPACE_USER) {
	/*
	* Fall back to the alternate namespace format if we failed to
	* find a user xattr.
	*/
	error = zfs_deleteextattr_impl(ap, !zfs_xattr_compat);
	}

	rw_exit(&zp->z_xattr_lock);
	zfs_exit(zfsvfs, FTAG);
	if (error == ENOENT)
	error = SET_ERROR(ENOATTR);
	return (error);
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_setextattr {
	IN struct vnode *a_vp;
	IN int a_attrnamespace;
	IN const char *a_name;
	INOUT struct uio *a_uio;
	IN struct ucred *a_cred;
	IN struct thread *a_td;
	};
	#endif

	static int
	zfs_setextattr_dir(struct vop_setextattr_args ap, const char attrname)
	{
	struct thread *td = ap->a_td;
	struct nameidata nd;
	struct vattr va;
	vnode_t xvp = NULL, vp;
	int error, flags;

	error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred,
	LOOKUP_XATTR \| CREATE_XATTR_DIR, B_FALSE);
	if (error != 0)
	return (error);

	flags = FFLAGS(O_WRONLY \| O_CREAT);
	#if __FreeBSD_version < 1400043
	NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, attrname, xvp, td);
	#else
	NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, attrname, xvp);
	#endif
	error = vn_open_cred(&nd, &flags, 0600, VN_OPEN_INVFS, ap->a_cred,
	NULL);
	if (error != 0)
	return (SET_ERROR(error));
	vp = nd.ni_vp;
	NDFREE_PNBUF(&nd);

	VATTR_NULL(&va);
	va.va_size = 0;
	error = VOP_SETATTR(vp, &va, ap->a_cred);
	if (error == 0)
	VOP_WRITE(vp, ap->a_uio, IO_UNIT, ap->a_cred);

	VOP_UNLOCK1(vp);
	vn_close(vp, flags, ap->a_cred, td);
	return (error);
	}

	static int
	zfs_setextattr_sa(struct vop_setextattr_args ap, const char attrname)
	{
	znode_t *zp = VTOZ(ap->a_vp);
	nvlist_t *nvl;
	size_t sa_size;
	int error;

	error = zfs_ensure_xattr_cached(zp);
	if (error != 0)
	return (error);

	ASSERT(RW_WRITE_HELD(&zp->z_xattr_lock));
	ASSERT3P(zp->z_xattr_cached, !=, NULL);

	nvl = zp->z_xattr_cached;
	size_t entry_size = ap->a_uio->uio_resid;
	if (entry_size > DXATTR_MAX_ENTRY_SIZE)
	return (SET_ERROR(EFBIG));
	error = nvlist_size(nvl, &sa_size, NV_ENCODE_XDR);
	if (error != 0)
	return (SET_ERROR(error));
	if (sa_size > DXATTR_MAX_SA_SIZE)
	return (SET_ERROR(EFBIG));
	uchar_t *buf = kmem_alloc(entry_size, KM_SLEEP);
	error = uiomove(buf, entry_size, ap->a_uio);
	if (error != 0) {
	error = SET_ERROR(error);
	} else {
	error = nvlist_add_byte_array(nvl, attrname, buf, entry_size);
	if (error != 0)
	error = SET_ERROR(error);
	}
	if (error == 0)
	error = zfs_sa_set_xattr(zp, attrname, buf, entry_size);
	kmem_free(buf, entry_size);
	if (error != 0) {
	zp->z_xattr_cached = NULL;
	nvlist_free(nvl);
	}
	return (error);
	}

	static int
	zfs_setextattr_impl(struct vop_setextattr_args *ap, boolean_t compat)
	{
	znode_t *zp = VTOZ(ap->a_vp);
	zfsvfs_t *zfsvfs = ZTOZSB(zp);
	char attrname[EXTATTR_MAXNAMELEN+1];
	int error;

	error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname,
	sizeof (attrname), compat);
	if (error != 0)
	return (error);

	struct vop_deleteextattr_args vda = {
	.a_vp = ap->a_vp,
	.a_attrnamespace = ap->a_attrnamespace,
	.a_name = ap->a_name,
	.a_cred = ap->a_cred,
	.a_td = ap->a_td,
	};
	error = ENOENT;
	if (zfsvfs->z_use_sa && zp->z_is_sa && zfsvfs->z_xattr_sa) {
	error = zfs_setextattr_sa(ap, attrname);
	if (error == 0) {
	/*
	* Successfully put into SA, we need to clear the one
	* in dir if present.
	*/
	zfs_deleteextattr_dir(&vda, attrname);
	}
	}
	if (error != 0) {
	error = zfs_setextattr_dir(ap, attrname);
	if (error == 0 && zp->z_is_sa) {
	/*
	* Successfully put into dir, we need to clear the one
	* in SA if present.
	*/
	zfs_deleteextattr_sa(&vda, attrname);
	}
	}
	if (error == 0 && ap->a_attrnamespace == EXTATTR_NAMESPACE_USER) {
	/*
	* Also clear all versions of the alternate compat name.
	*/
	zfs_deleteextattr_impl(&vda, !compat);
	}
	return (error);
	}

	/*
	* Vnode operation to set a named attribute.
	*/
	static int
	zfs_setextattr(struct vop_setextattr_args *ap)
	{
	znode_t *zp = VTOZ(ap->a_vp);
	zfsvfs_t *zfsvfs = ZTOZSB(zp);
	int error;

	/*
	* If the xattr property is off, refuse the request.
	*/
	if (!(zfsvfs->z_flags & ZSB_XATTR))
	return (SET_ERROR(EOPNOTSUPP));

	error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
	ap->a_cred, ap->a_td, VWRITE);
	if (error != 0)
	return (SET_ERROR(error));

	error = zfs_check_attrname(ap->a_name);
	if (error != 0)
	return (error);

	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (error);
	rw_enter(&zp->z_xattr_lock, RW_WRITER);

	error = zfs_setextattr_impl(ap, zfs_xattr_compat);

	rw_exit(&zp->z_xattr_lock);
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_listextattr {
	IN struct vnode *a_vp;
	IN int a_attrnamespace;
	INOUT struct uio *a_uio;
	OUT size_t *a_size;
	IN struct ucred *a_cred;
	IN struct thread *a_td;
	};
	#endif

	static int
	zfs_listextattr_dir(struct vop_listextattr_args ap, const char attrprefix)
	{
	struct thread *td = ap->a_td;
	struct nameidata nd;
	uint8_t dirbuf[sizeof (struct dirent)];
	struct iovec aiov;
	struct uio auio;
	vnode_t xvp = NULL, vp;
	int error, eof;

	error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred,
	LOOKUP_XATTR, B_FALSE);
	if (error != 0) {
	/*
	* ENOATTR means that the EA directory does not yet exist,
	* i.e. there are no extended attributes there.
	*/
	if (error == ENOATTR)
	error = 0;
	return (error);
	}

	#if __FreeBSD_version < 1400043
	NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW \| LOCKLEAF \| LOCKSHARED,
	UIO_SYSSPACE, ".", xvp, td);
	#else
	NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW \| LOCKLEAF \| LOCKSHARED,
	UIO_SYSSPACE, ".", xvp);
	#endif
	error = namei(&nd);
	if (error != 0)
	return (SET_ERROR(error));
	vp = nd.ni_vp;
	NDFREE_PNBUF(&nd);

	auio.uio_iov = &aiov;
	auio.uio_iovcnt = 1;
	auio.uio_segflg = UIO_SYSSPACE;
	auio.uio_td = td;
	auio.uio_rw = UIO_READ;
	auio.uio_offset = 0;

	size_t plen = strlen(attrprefix);

	do {
	aiov.iov_base = (void *)dirbuf;
	aiov.iov_len = sizeof (dirbuf);
	auio.uio_resid = sizeof (dirbuf);
	error = VOP_READDIR(vp, &auio, ap->a_cred, &eof, NULL, NULL);
	if (error != 0)
	break;
	int done = sizeof (dirbuf) - auio.uio_resid;
	for (int pos = 0; pos < done; ) {
	struct dirent dp = (struct dirent )(dirbuf + pos);
	pos += dp->d_reclen;
	/*
	* XXX: Temporarily we also accept DT_UNKNOWN, as this
	* is what we get when attribute was created on Solaris.
	*/
	if (dp->d_type != DT_REG && dp->d_type != DT_UNKNOWN)
	continue;
	else if (plen == 0 &&
	ZFS_XA_NS_PREFIX_FORBIDDEN(dp->d_name))
	continue;
	else if (strncmp(dp->d_name, attrprefix, plen) != 0)
	continue;
	uint8_t nlen = dp->d_namlen - plen;
	if (ap->a_size != NULL) {
	*ap->a_size += 1 + nlen;
	} else if (ap->a_uio != NULL) {
	/*
	* Format of extattr name entry is one byte for
	* length and the rest for name.
	*/
	error = uiomove(&nlen, 1, ap->a_uio);
	if (error == 0) {
	char *namep = dp->d_name + plen;
	error = uiomove(namep, nlen, ap->a_uio);
	}
	if (error != 0) {
	error = SET_ERROR(error);
	break;
	}
	}
	}
	} while (!eof && error == 0);

	vput(vp);
	return (error);
	}

	static int
	zfs_listextattr_sa(struct vop_listextattr_args ap, const char attrprefix)
	{
	znode_t *zp = VTOZ(ap->a_vp);
	int error;

	error = zfs_ensure_xattr_cached(zp);
	if (error != 0)
	return (error);

	ASSERT(RW_LOCK_HELD(&zp->z_xattr_lock));
	ASSERT3P(zp->z_xattr_cached, !=, NULL);

	size_t plen = strlen(attrprefix);
	nvpair_t *nvp = NULL;
	while ((nvp = nvlist_next_nvpair(zp->z_xattr_cached, nvp)) != NULL) {
	ASSERT3U(nvpair_type(nvp), ==, DATA_TYPE_BYTE_ARRAY);

	const char *name = nvpair_name(nvp);
	if (plen == 0 && ZFS_XA_NS_PREFIX_FORBIDDEN(name))
	continue;
	else if (strncmp(name, attrprefix, plen) != 0)
	continue;
	uint8_t nlen = strlen(name) - plen;
	if (ap->a_size != NULL) {
	*ap->a_size += 1 + nlen;
	} else if (ap->a_uio != NULL) {
	/*
	* Format of extattr name entry is one byte for
	* length and the rest for name.
	*/
	error = uiomove(&nlen, 1, ap->a_uio);
	if (error == 0) {
	char namep = __DECONST(char , name) + plen;
	error = uiomove(namep, nlen, ap->a_uio);
	}
	if (error != 0) {
	error = SET_ERROR(error);
	break;
	}
	}
	}

	return (error);
	}

	static int
	zfs_listextattr_impl(struct vop_listextattr_args *ap, boolean_t compat)
	{
	znode_t *zp = VTOZ(ap->a_vp);
	zfsvfs_t *zfsvfs = ZTOZSB(zp);
	char attrprefix[16];
	int error;

	error = zfs_create_attrname(ap->a_attrnamespace, "", attrprefix,
	sizeof (attrprefix), compat);
	if (error != 0)
	return (error);

	if (zfsvfs->z_use_sa && zp->z_is_sa)
	error = zfs_listextattr_sa(ap, attrprefix);
	if (error == 0)
	error = zfs_listextattr_dir(ap, attrprefix);
	return (error);
	}

	/*
	* Vnode operation to retrieve extended attributes on a vnode.
	*/
	static int
	zfs_listextattr(struct vop_listextattr_args *ap)
	{
	znode_t *zp = VTOZ(ap->a_vp);
	zfsvfs_t *zfsvfs = ZTOZSB(zp);
	int error;

	if (ap->a_size != NULL)
	*ap->a_size = 0;

	/*
	* If the xattr property is off, refuse the request.
	*/
	if (!(zfsvfs->z_flags & ZSB_XATTR))
	return (SET_ERROR(EOPNOTSUPP));

	error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace,
	ap->a_cred, ap->a_td, VREAD);
	if (error != 0)
	return (SET_ERROR(error));

	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (error);
	rw_enter(&zp->z_xattr_lock, RW_READER);

	error = zfs_listextattr_impl(ap, zfs_xattr_compat);
	if (error == 0 && ap->a_attrnamespace == EXTATTR_NAMESPACE_USER) {
	/* Also list user xattrs with the alternate format. */
	error = zfs_listextattr_impl(ap, !zfs_xattr_compat);
	}

	rw_exit(&zp->z_xattr_lock);
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_getacl_args {
	struct vnode *vp;
	acl_type_t type;
	struct acl *aclp;
	struct ucred *cred;
	struct thread *td;
	};
	#endif

	static int
	zfs_freebsd_getacl(struct vop_getacl_args *ap)
	{
	int error;
	vsecattr_t vsecattr;

	if (ap->a_type != ACL_TYPE_NFS4)
	return (EINVAL);

	vsecattr.vsa_mask = VSA_ACE \| VSA_ACECNT;
	if ((error = zfs_getsecattr(VTOZ(ap->a_vp),
	&vsecattr, 0, ap->a_cred)))
	return (error);

	error = acl_from_aces(ap->a_aclp, vsecattr.vsa_aclentp,
	vsecattr.vsa_aclcnt);
	if (vsecattr.vsa_aclentp != NULL)
	kmem_free(vsecattr.vsa_aclentp, vsecattr.vsa_aclentsz);

	return (error);
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_setacl_args {
	struct vnode *vp;
	acl_type_t type;
	struct acl *aclp;
	struct ucred *cred;
	struct thread *td;
	};
	#endif

	static int
	zfs_freebsd_setacl(struct vop_setacl_args *ap)
	{
	int error;
	vsecattr_t vsecattr;
	int aclbsize; /* size of acl list in bytes */
	aclent_t *aaclp;

	if (ap->a_type != ACL_TYPE_NFS4)
	return (EINVAL);

	if (ap->a_aclp == NULL)
	return (EINVAL);

	if (ap->a_aclp->acl_cnt < 1 \|\| ap->a_aclp->acl_cnt > MAX_ACL_ENTRIES)
	return (EINVAL);

	/*
	* With NFSv4 ACLs, chmod(2) may need to add additional entries,
	* splitting every entry into two and appending "canonical six"
	* entries at the end. Don't allow for setting an ACL that would
	* cause chmod(2) to run out of ACL entries.
	*/
	if (ap->a_aclp->acl_cnt * 2 + 6 > ACL_MAX_ENTRIES)
	return (ENOSPC);

	error = acl_nfs4_check(ap->a_aclp, ap->a_vp->v_type == VDIR);
	if (error != 0)
	return (error);

	vsecattr.vsa_mask = VSA_ACE;
	aclbsize = ap->a_aclp->acl_cnt * sizeof (ace_t);
	vsecattr.vsa_aclentp = kmem_alloc(aclbsize, KM_SLEEP);
	aaclp = vsecattr.vsa_aclentp;
	vsecattr.vsa_aclentsz = aclbsize;

	aces_from_acl(vsecattr.vsa_aclentp, &vsecattr.vsa_aclcnt, ap->a_aclp);
	error = zfs_setsecattr(VTOZ(ap->a_vp), &vsecattr, 0, ap->a_cred);
	kmem_free(aaclp, aclbsize);

	return (error);
	}

	#ifndef _SYS_SYSPROTO_H_
	struct vop_aclcheck_args {
	struct vnode *vp;
	acl_type_t type;
	struct acl *aclp;
	struct ucred *cred;
	struct thread *td;
	};
	#endif

	static int
	zfs_freebsd_aclcheck(struct vop_aclcheck_args *ap)
	{

	return (EOPNOTSUPP);
	}

	static int
	zfs_vptocnp(struct vop_vptocnp_args *ap)
	{
	vnode_t *covered_vp;
	vnode_t *vp = ap->a_vp;
	zfsvfs_t *zfsvfs = vp->v_vfsp->vfs_data;
	znode_t *zp = VTOZ(vp);
	int ltype;
	int error;

	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (error);

	/*
	* If we are a snapshot mounted under .zfs, run the operation
	* on the covered vnode.
	*/
	if (zp->z_id != zfsvfs->z_root \|\| zfsvfs->z_parent == zfsvfs) {
	char name[MAXNAMLEN + 1];
	znode_t *dzp;
	size_t len;

	error = zfs_znode_parent_and_name(zp, &dzp, name);
	if (error == 0) {
	len = strlen(name);
	if (*ap->a_buflen < len)
	error = SET_ERROR(ENOMEM);
	}
	if (error == 0) {
	*ap->a_buflen -= len;
	memcpy(ap->a_buf + *ap->a_buflen, name, len);
	*ap->a_vpp = ZTOV(dzp);
	}
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}
	zfs_exit(zfsvfs, FTAG);

	covered_vp = vp->v_mount->mnt_vnodecovered;
	#if __FreeBSD_version >= 1300045
	enum vgetstate vs = vget_prep(covered_vp);
	#else
	vhold(covered_vp);
	#endif
	ltype = VOP_ISLOCKED(vp);
	VOP_UNLOCK1(vp);
	#if __FreeBSD_version >= 1300045
	error = vget_finish(covered_vp, LK_SHARED, vs);
	#else
	error = vget(covered_vp, LK_SHARED \| LK_VNHELD, curthread);
	#endif
	if (error == 0) {
	#if __FreeBSD_version >= 1300123
	error = VOP_VPTOCNP(covered_vp, ap->a_vpp, ap->a_buf,
	ap->a_buflen);
	#else
	error = VOP_VPTOCNP(covered_vp, ap->a_vpp, ap->a_cred,
	ap->a_buf, ap->a_buflen);
	#endif
	vput(covered_vp);
	}
	vn_lock(vp, ltype \| LK_RETRY);
	if (VN_IS_DOOMED(vp))
	error = SET_ERROR(ENOENT);
	return (error);
	}

	#if __FreeBSD_version >= 1400032
	static int
	zfs_deallocate(struct vop_deallocate_args *ap)
	{
	znode_t *zp = VTOZ(ap->a_vp);
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	zilog_t *zilog;
	off_t off, len, file_sz;
	int error;

	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (error);

	/*
	* Callers might not be able to detect properly that we are read-only,
	* so check it explicitly here.
	*/
	if (zfs_is_readonly(zfsvfs)) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EROFS));
	}

	zilog = zfsvfs->z_log;
	off = *ap->a_offset;
	len = *ap->a_len;
	file_sz = zp->z_size;
	if (off + len > file_sz)
	len = file_sz - off;
	/* Fast path for out-of-range request. */
	if (len <= 0) {
	*ap->a_len = 0;
	zfs_exit(zfsvfs, FTAG);
	return (0);
	}

	error = zfs_freesp(zp, off, len, O_RDWR, TRUE);
	if (error == 0) {
	if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS \|\|
	(ap->a_ioflag & IO_SYNC) != 0)
	zil_commit(zilog, zp->z_id);
	*ap->a_offset = off + len;
	*ap->a_len = 0;
	}

	zfs_exit(zfsvfs, FTAG);
	return (error);
	}
	#endif

	#if __FreeBSD_version >= 1300039
	#ifndef _SYS_SYSPROTO_H_
	struct vop_copy_file_range_args {
	struct vnode *a_invp;
	off_t *a_inoffp;
	struct vnode *a_outvp;
	off_t *a_outoffp;
	size_t *a_lenp;
	unsigned int a_flags;
	struct ucred *a_incred;
	struct ucred *a_outcred;
	struct thread *a_fsizetd;
	}
	#endif
	/*
	* TODO: FreeBSD will only call file system-specific copy_file_range() if both
	* files resides under the same mountpoint. In case of ZFS we want to be called
	* even is files are in different datasets (but on the same pools, but we need
	* to check that ourselves).
	*/
	static int
	zfs_freebsd_copy_file_range(struct vop_copy_file_range_args *ap)
	{
	zfsvfs_t *outzfsvfs;
	struct vnode *invp = ap->a_invp;
	struct vnode *outvp = ap->a_outvp;
	struct mount *mp;
	struct uio io;
	int error;
	uint64_t len = *ap->a_lenp;

	if (!zfs_bclone_enabled) {
	mp = NULL;
	goto bad_write_fallback;
	}

	/*
	* TODO: If offset/length is not aligned to recordsize, use
	* vn_generic_copy_file_range() on this fragment.
	* It would be better to do this after we lock the vnodes, but then we
	* need something else than vn_generic_copy_file_range().
	*/

	vn_start_write(outvp, &mp, V_WAIT);
	if (__predict_true(mp == outvp->v_mount)) {
	outzfsvfs = (zfsvfs_t *)mp->mnt_data;
	if (!spa_feature_is_enabled(dmu_objset_spa(outzfsvfs->z_os),
	SPA_FEATURE_BLOCK_CLONING)) {
	goto bad_write_fallback;
	}
	}
	if (invp == outvp) {
	if (vn_lock(outvp, LK_EXCLUSIVE) != 0) {
	goto bad_write_fallback;
	}
	} else {
	#if (__FreeBSD_version >= 1302506 && __FreeBSD_version < 1400000) \|\| \
	__FreeBSD_version >= 1400086
	vn_lock_pair(invp, false, LK_EXCLUSIVE, outvp, false,
	LK_EXCLUSIVE);
	#else
	vn_lock_pair(invp, false, outvp, false);
	#endif
	if (VN_IS_DOOMED(invp) \|\| VN_IS_DOOMED(outvp)) {
	goto bad_locked_fallback;
	}
	}

	#ifdef MAC
	error = mac_vnode_check_write(curthread->td_ucred, ap->a_outcred,
	outvp);
	if (error != 0)
	goto out_locked;
	#endif

	io.uio_offset = *ap->a_outoffp;
	io.uio_resid = *ap->a_lenp;
	error = vn_rlimit_fsize(outvp, &io, ap->a_fsizetd);
	if (error != 0)
	goto out_locked;

	error = zfs_clone_range(VTOZ(invp), ap->a_inoffp, VTOZ(outvp),
	ap->a_outoffp, &len, ap->a_outcred);
	if (error == EXDEV \|\| error == EAGAIN \|\| error == EINVAL \|\|
	error == EOPNOTSUPP)
	goto bad_locked_fallback;
	*ap->a_lenp = (size_t)len;
	out_locked:
	if (invp != outvp)
	VOP_UNLOCK(invp);
	VOP_UNLOCK(outvp);
	if (mp != NULL)
	vn_finished_write(mp);
	return (error);

	bad_locked_fallback:
	if (invp != outvp)
	VOP_UNLOCK(invp);
	VOP_UNLOCK(outvp);
	bad_write_fallback:
	if (mp != NULL)
	vn_finished_write(mp);
	error = ENOSYS;
	return (error);
	}
	#endif

	struct vop_vector zfs_vnodeops;
	struct vop_vector zfs_fifoops;
	struct vop_vector zfs_shareops;

	struct vop_vector zfs_vnodeops = {
	.vop_default = &default_vnodeops,
	.vop_inactive = zfs_freebsd_inactive,
	#if __FreeBSD_version >= 1300042
	.vop_need_inactive = zfs_freebsd_need_inactive,
	#endif
	.vop_reclaim = zfs_freebsd_reclaim,
	#if __FreeBSD_version >= 1300102
	.vop_fplookup_vexec = zfs_freebsd_fplookup_vexec,
	#endif
	#if __FreeBSD_version >= 1300139
	.vop_fplookup_symlink = zfs_freebsd_fplookup_symlink,
	#endif
	.vop_access = zfs_freebsd_access,
	.vop_allocate = VOP_EINVAL,
	#if __FreeBSD_version >= 1400032
	.vop_deallocate = zfs_deallocate,
	#endif
	.vop_lookup = zfs_cache_lookup,
	.vop_cachedlookup = zfs_freebsd_cachedlookup,
	.vop_getattr = zfs_freebsd_getattr,
	.vop_setattr = zfs_freebsd_setattr,
	.vop_create = zfs_freebsd_create,
	.vop_mknod = (vop_mknod_t *)zfs_freebsd_create,
	.vop_mkdir = zfs_freebsd_mkdir,
	.vop_readdir = zfs_freebsd_readdir,
	.vop_fsync = zfs_freebsd_fsync,
	.vop_open = zfs_freebsd_open,
	.vop_close = zfs_freebsd_close,
	.vop_rmdir = zfs_freebsd_rmdir,
	.vop_ioctl = zfs_freebsd_ioctl,
	.vop_link = zfs_freebsd_link,
	.vop_symlink = zfs_freebsd_symlink,
	.vop_readlink = zfs_freebsd_readlink,
	.vop_read = zfs_freebsd_read,
	.vop_write = zfs_freebsd_write,
	.vop_remove = zfs_freebsd_remove,
	.vop_rename = zfs_freebsd_rename,
	.vop_pathconf = zfs_freebsd_pathconf,
	.vop_bmap = zfs_freebsd_bmap,
	.vop_fid = zfs_freebsd_fid,
	.vop_getextattr = zfs_getextattr,
	.vop_deleteextattr = zfs_deleteextattr,
	.vop_setextattr = zfs_setextattr,
	.vop_listextattr = zfs_listextattr,
	.vop_getacl = zfs_freebsd_getacl,
	.vop_setacl = zfs_freebsd_setacl,
	.vop_aclcheck = zfs_freebsd_aclcheck,
	.vop_getpages = zfs_freebsd_getpages,
	.vop_putpages = zfs_freebsd_putpages,
	.vop_vptocnp = zfs_vptocnp,
	#if __FreeBSD_version >= 1300064
	.vop_lock1 = vop_lock,
	.vop_unlock = vop_unlock,
	.vop_islocked = vop_islocked,
	#endif
	#if __FreeBSD_version >= 1400043
	.vop_add_writecount = vop_stdadd_writecount_nomsync,
	#endif
	#if __FreeBSD_version >= 1300039
	.vop_copy_file_range = zfs_freebsd_copy_file_range,
	#endif
	};
	VFS_VOP_VECTOR_REGISTER(zfs_vnodeops);

	struct vop_vector zfs_fifoops = {
	.vop_default = &fifo_specops,
	.vop_fsync = zfs_freebsd_fsync,
	#if __FreeBSD_version >= 1300102
	.vop_fplookup_vexec = zfs_freebsd_fplookup_vexec,
	#endif
	#if __FreeBSD_version >= 1300139
	.vop_fplookup_symlink = zfs_freebsd_fplookup_symlink,
	#endif
	.vop_access = zfs_freebsd_access,
	.vop_getattr = zfs_freebsd_getattr,
	.vop_inactive = zfs_freebsd_inactive,
	.vop_read = VOP_PANIC,
	.vop_reclaim = zfs_freebsd_reclaim,
	.vop_setattr = zfs_freebsd_setattr,
	.vop_write = VOP_PANIC,
	.vop_pathconf = zfs_freebsd_pathconf,
	.vop_fid = zfs_freebsd_fid,
	.vop_getacl = zfs_freebsd_getacl,
	.vop_setacl = zfs_freebsd_setacl,
	.vop_aclcheck = zfs_freebsd_aclcheck,
	#if __FreeBSD_version >= 1400043
	.vop_add_writecount = vop_stdadd_writecount_nomsync,
	#endif
	};
	VFS_VOP_VECTOR_REGISTER(zfs_fifoops);

	/*
	* special share hidden files vnode operations template
	*/
	struct vop_vector zfs_shareops = {
	.vop_default = &default_vnodeops,
	#if __FreeBSD_version >= 1300121
	.vop_fplookup_vexec = VOP_EAGAIN,
	#endif
	#if __FreeBSD_version >= 1300139
	.vop_fplookup_symlink = VOP_EAGAIN,
	#endif
	.vop_access = zfs_freebsd_access,
	.vop_inactive = zfs_freebsd_inactive,
	.vop_reclaim = zfs_freebsd_reclaim,
	.vop_fid = zfs_freebsd_fid,
	.vop_pathconf = zfs_freebsd_pathconf,
	#if __FreeBSD_version >= 1400043
	.vop_add_writecount = vop_stdadd_writecount_nomsync,
	#endif
	};
	VFS_VOP_VECTOR_REGISTER(zfs_shareops);

	ZFS_MODULE_PARAM(zfs, zfs_, xattr_compat, INT, ZMOD_RW,
	"Use legacy ZFS xattr naming for writing new user namespace xattrs");
	diff --git a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_znode.c b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_znode.c
	index 0d4c94555c6b..0eea2a849416 100644
	--- a/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_znode.c
	+++ b/sys/contrib/openzfs/module/os/freebsd/zfs/zfs_znode.c
	@@ -1,2226 +1,2225 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2014 by Delphix. All rights reserved.
	* Copyright (c) 2014 Integros [integros.com]
	*/

	/* Portions Copyright 2007 Jeremy Teo */
	/* Portions Copyright 2011 Martin Matuska <mm@FreeBSD.org> */

	#ifdef _KERNEL
	#include <sys/types.h>
	#include <sys/param.h>
	#include <sys/time.h>
	#include <sys/systm.h>
	#include <sys/sysmacros.h>
	#include <sys/resource.h>
	#include <sys/resourcevar.h>
	#include <sys/mntent.h>
	#include <sys/u8_textprep.h>
	#include <sys/dsl_dataset.h>
	#include <sys/vfs.h>
	#include <sys/vnode.h>
	#include <sys/file.h>
	#include <sys/kmem.h>
	#include <sys/errno.h>
	#include <sys/unistd.h>
	#include <sys/atomic.h>
	#include <sys/zfs_dir.h>
	#include <sys/zfs_acl.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/zfs_rlock.h>
	#include <sys/zfs_fuid.h>
	#include <sys/dnode.h>
	#include <sys/fs/zfs.h>
	#endif /* _KERNEL */

	#include <sys/dmu.h>
	#include <sys/dmu_objset.h>
	#include <sys/dmu_tx.h>
	#include <sys/zfs_refcount.h>
	#include <sys/stat.h>
	#include <sys/zap.h>
	#include <sys/zfs_znode.h>
	#include <sys/sa.h>
	#include <sys/zfs_sa.h>
	#include <sys/zfs_stat.h>

	#include "zfs_prop.h"
	#include "zfs_comutil.h"

	/* Used by fstat(1). */
	SYSCTL_INT(_debug_sizeof, OID_AUTO, znode, CTLFLAG_RD,
	SYSCTL_NULL_INT_PTR, sizeof (znode_t), "sizeof(znode_t)");

	/*
	* Define ZNODE_STATS to turn on statistic gathering. By default, it is only
	* turned on when DEBUG is also defined.
	*/
	#ifdef ZFS_DEBUG
	#define ZNODE_STATS
	#endif /* DEBUG */

	#ifdef ZNODE_STATS
	#define ZNODE_STAT_ADD(stat) ((stat)++)
	#else
	#define ZNODE_STAT_ADD(stat) /* nothing */
	#endif /* ZNODE_STATS */

	/*
	* Functions needed for userland (ie: libzpool) are not put under
	* #ifdef_KERNEL; the rest of the functions have dependencies
	* (such as VFS logic) that will not compile easily in userland.
	*/
	#ifdef _KERNEL
	#if !defined(KMEM_DEBUG) && __FreeBSD_version >= 1300102
	#define _ZFS_USE_SMR
	static uma_zone_t znode_uma_zone;
	#else
	static kmem_cache_t *znode_cache = NULL;
	#endif

	extern struct vop_vector zfs_vnodeops;
	extern struct vop_vector zfs_fifoops;
	extern struct vop_vector zfs_shareops;


	/*
	* This callback is invoked when acquiring a RL_WRITER or RL_APPEND lock on
	* z_rangelock. It will modify the offset and length of the lock to reflect
	* znode-specific information, and convert RL_APPEND to RL_WRITER. This is
	* called with the rangelock_t's rl_lock held, which avoids races.
	*/
	static void
	zfs_rangelock_cb(zfs_locked_range_t new, void arg)
	{
	znode_t *zp = arg;

	/*
	* If in append mode, convert to writer and lock starting at the
	* current end of file.
	*/
	if (new->lr_type == RL_APPEND) {
	new->lr_offset = zp->z_size;
	new->lr_type = RL_WRITER;
	}

	/*
	* If we need to grow the block size then lock the whole file range.
	*/
	uint64_t end_size = MAX(zp->z_size, new->lr_offset + new->lr_length);
	if (end_size > zp->z_blksz && (!ISP2(zp->z_blksz) \|\|
	zp->z_blksz < ZTOZSB(zp)->z_max_blksz)) {
	new->lr_offset = 0;
	new->lr_length = UINT64_MAX;
	}
	}

	static int
	zfs_znode_cache_constructor(void buf, void arg, int kmflags)
	{
	znode_t *zp = buf;

	POINTER_INVALIDATE(&zp->z_zfsvfs);

	list_link_init(&zp->z_link_node);

	mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL);
	rw_init(&zp->z_xattr_lock, NULL, RW_DEFAULT, NULL);

	zfs_rangelock_init(&zp->z_rangelock, zfs_rangelock_cb, zp);

	zp->z_acl_cached = NULL;
	zp->z_xattr_cached = NULL;
	zp->z_xattr_parent = 0;
	zp->z_vnode = NULL;
	zp->z_sync_writes_cnt = 0;
	zp->z_async_writes_cnt = 0;

	return (0);
	}

	static void
	zfs_znode_cache_destructor(void buf, void arg)
	{
	(void) arg;
	znode_t *zp = buf;

	ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
	ASSERT3P(zp->z_vnode, ==, NULL);
	ASSERT(!list_link_active(&zp->z_link_node));
	mutex_destroy(&zp->z_lock);
	mutex_destroy(&zp->z_acl_lock);
	rw_destroy(&zp->z_xattr_lock);
	zfs_rangelock_fini(&zp->z_rangelock);

	ASSERT3P(zp->z_acl_cached, ==, NULL);
	ASSERT3P(zp->z_xattr_cached, ==, NULL);

	ASSERT0(atomic_load_32(&zp->z_sync_writes_cnt));
	ASSERT0(atomic_load_32(&zp->z_async_writes_cnt));
	}


	#ifdef _ZFS_USE_SMR
	VFS_SMR_DECLARE;

	static int
	zfs_znode_cache_constructor_smr(void mem, int size __unused, void private,
	int flags)
	{
	return (zfs_znode_cache_constructor(mem, private, flags));
	}

	static void
	zfs_znode_cache_destructor_smr(void mem, int size __unused, void private)
	{
	zfs_znode_cache_destructor(mem, private);
	}

	void
	zfs_znode_init(void)
	{
	/*
	* Initialize zcache
	*/
	ASSERT3P(znode_uma_zone, ==, NULL);
	znode_uma_zone = uma_zcreate("zfs_znode_cache",
	sizeof (znode_t), zfs_znode_cache_constructor_smr,
	zfs_znode_cache_destructor_smr, NULL, NULL, 0, 0);
	VFS_SMR_ZONE_SET(znode_uma_zone);
	}

	static znode_t *
	zfs_znode_alloc_kmem(int flags)
	{
	return (uma_zalloc_smr(znode_uma_zone, flags));
	}

	static void
	zfs_znode_free_kmem(znode_t *zp)
	{
	if (zp->z_xattr_cached) {
	nvlist_free(zp->z_xattr_cached);
	zp->z_xattr_cached = NULL;
	}
	uma_zfree_smr(znode_uma_zone, zp);
	}
	#else
	void
	zfs_znode_init(void)
	{
	/*
	* Initialize zcache
	*/
	ASSERT3P(znode_cache, ==, NULL);
	znode_cache = kmem_cache_create("zfs_znode_cache",
	sizeof (znode_t), 0, zfs_znode_cache_constructor,
	zfs_znode_cache_destructor, NULL, NULL, NULL, 0);
	}

	static znode_t *
	zfs_znode_alloc_kmem(int flags)
	{
	return (kmem_cache_alloc(znode_cache, flags));
	}

	static void
	zfs_znode_free_kmem(znode_t *zp)
	{
	if (zp->z_xattr_cached) {
	nvlist_free(zp->z_xattr_cached);
	zp->z_xattr_cached = NULL;
	}
	kmem_cache_free(znode_cache, zp);
	}
	#endif

	void
	zfs_znode_fini(void)
	{
	/*
	* Cleanup zcache
	*/
	#ifdef _ZFS_USE_SMR
	if (znode_uma_zone) {
	uma_zdestroy(znode_uma_zone);
	znode_uma_zone = NULL;
	}
	#else
	if (znode_cache) {
	kmem_cache_destroy(znode_cache);
	znode_cache = NULL;
	}
	#endif
	}


	static int
	zfs_create_share_dir(zfsvfs_t zfsvfs, dmu_tx_t tx)
	{
	zfs_acl_ids_t acl_ids;
	vattr_t vattr;
	znode_t *sharezp;
	znode_t *zp;
	int error;

	vattr.va_mask = AT_MODE\|AT_UID\|AT_GID;
	vattr.va_type = VDIR;
	vattr.va_mode = S_IFDIR\|0555;
	vattr.va_uid = crgetuid(kcred);
	vattr.va_gid = crgetgid(kcred);

	sharezp = zfs_znode_alloc_kmem(KM_SLEEP);
	ASSERT(!POINTER_IS_VALID(sharezp->z_zfsvfs));
	sharezp->z_unlinked = 0;
	sharezp->z_atime_dirty = 0;
	sharezp->z_zfsvfs = zfsvfs;
	sharezp->z_is_sa = zfsvfs->z_use_sa;

	VERIFY0(zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr,
	kcred, NULL, &acl_ids, NULL));
	zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids);
	ASSERT3P(zp, ==, sharezp);
	POINTER_INVALIDATE(&sharezp->z_zfsvfs);
	error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
	ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx);
	zfsvfs->z_shares_dir = sharezp->z_id;

	zfs_acl_ids_free(&acl_ids);
	sa_handle_destroy(sharezp->z_sa_hdl);
	zfs_znode_free_kmem(sharezp);

	return (error);
	}

	/*
	* define a couple of values we need available
	* for both 64 and 32 bit environments.
	*/
	#ifndef NBITSMINOR64
	#define NBITSMINOR64 32
	#endif
	#ifndef MAXMAJ64
	#define MAXMAJ64 0xffffffffUL
	#endif
	#ifndef MAXMIN64
	#define MAXMIN64 0xffffffffUL
	#endif

	/*
	* Create special expldev for ZFS private use.
	* Can't use standard expldev since it doesn't do
	* what we want. The standard expldev() takes a
	* dev32_t in LP64 and expands it to a long dev_t.
	* We need an interface that takes a dev32_t in ILP32
	* and expands it to a long dev_t.
	*/
	static uint64_t
	zfs_expldev(dev_t dev)
	{
	return (((uint64_t)major(dev) << NBITSMINOR64) \| minor(dev));
	}
	/*
	* Special cmpldev for ZFS private use.
	* Can't use standard cmpldev since it takes
	* a long dev_t and compresses it to dev32_t in
	* LP64. We need to do a compaction of a long dev_t
	* to a dev32_t in ILP32.
	*/
	dev_t
	zfs_cmpldev(uint64_t dev)
	{
	return (makedev((dev >> NBITSMINOR64), (dev & MAXMIN64)));
	}

	static void
	zfs_znode_sa_init(zfsvfs_t zfsvfs, znode_t zp,
	dmu_buf_t db, dmu_object_type_t obj_type, sa_handle_t sa_hdl)
	{
	ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) \|\| (zfsvfs == zp->z_zfsvfs));
	ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)));

	ASSERT3P(zp->z_sa_hdl, ==, NULL);
	ASSERT3P(zp->z_acl_cached, ==, NULL);
	if (sa_hdl == NULL) {
	VERIFY0(sa_handle_get_from_db(zfsvfs->z_os, db, zp,
	SA_HDL_SHARED, &zp->z_sa_hdl));
	} else {
	zp->z_sa_hdl = sa_hdl;
	sa_set_userp(sa_hdl, zp);
	}

	zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE;

	/*
	* Slap on VROOT if we are the root znode unless we are the root
	* node of a snapshot mounted under .zfs.
	*/
	if (zp->z_id == zfsvfs->z_root && zfsvfs->z_parent == zfsvfs)
	ZTOV(zp)->v_flag \|= VROOT;

	vn_exists(ZTOV(zp));
	}

	void
	zfs_znode_dmu_fini(znode_t *zp)
	{
	ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) \|\|
	ZFS_TEARDOWN_INACTIVE_WRITE_HELD(zp->z_zfsvfs));

	sa_handle_destroy(zp->z_sa_hdl);
	zp->z_sa_hdl = NULL;
	}

	static void
	zfs_vnode_forget(vnode_t *vp)
	{

	/* copied from insmntque_stddtr */
	vp->v_data = NULL;
	vp->v_op = &dead_vnodeops;
	vgone(vp);
	vput(vp);
	}

	/*
	* Construct a new znode/vnode and initialize.
	*
	* This does not do a call to dmu_set_user() that is
	* up to the caller to do, in case you don't want to
	* return the znode
	*/
	static znode_t *
	zfs_znode_alloc(zfsvfs_t zfsvfs, dmu_buf_t db, int blksz,
	dmu_object_type_t obj_type, sa_handle_t *hdl)
	{
	znode_t *zp;
	vnode_t *vp;
	uint64_t mode;
	uint64_t parent;
	#ifdef notyet
	uint64_t mtime[2], ctime[2];
	#endif
	uint64_t projid = ZFS_DEFAULT_PROJID;
	sa_bulk_attr_t bulk[9];
	int count = 0;
	int error;

	zp = zfs_znode_alloc_kmem(KM_SLEEP);

	#ifndef _ZFS_USE_SMR
	KASSERT((zfsvfs->z_parent->z_vfs->mnt_kern_flag & MNTK_FPLOOKUP) == 0,
	("%s: fast path lookup enabled without smr", __func__));
	#endif

	#if __FreeBSD_version >= 1300076
	KASSERT(curthread->td_vp_reserved != NULL,
	("zfs_znode_alloc: getnewvnode without any vnodes reserved"));
	#else
	KASSERT(curthread->td_vp_reserv > 0,
	("zfs_znode_alloc: getnewvnode without any vnodes reserved"));
	#endif
	error = getnewvnode("zfs", zfsvfs->z_parent->z_vfs, &zfs_vnodeops, &vp);
	if (error != 0) {
	zfs_znode_free_kmem(zp);
	return (NULL);
	}
	zp->z_vnode = vp;
	vp->v_data = zp;

	/*
	* Acquire the vnode lock before any possible interaction with the
	* outside world. Specifically, there is an error path that calls
	* zfs_vnode_forget() and the vnode should be exclusively locked.
	*/
	vn_lock(vp, LK_EXCLUSIVE \| LK_RETRY);

	ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));

	zp->z_sa_hdl = NULL;
	zp->z_unlinked = 0;
	zp->z_atime_dirty = 0;
	zp->z_mapcnt = 0;
	zp->z_id = db->db_object;
	zp->z_blksz = blksz;
	zp->z_seq = 0x7A4653;
	zp->z_sync_cnt = 0;
	zp->z_sync_writes_cnt = 0;
	zp->z_async_writes_cnt = 0;
	#if __FreeBSD_version >= 1300139
	atomic_store_ptr(&zp->z_cached_symlink, NULL);
	#endif

	zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl);

	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &zp->z_gen, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
	&zp->z_size, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
	&zp->z_links, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
	&zp->z_pflags, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
	&zp->z_atime, 16);
	#ifdef notyet
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
	&mtime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
	&ctime, 16);
	#endif
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
	&zp->z_uid, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
	&zp->z_gid, 8);

	if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 \|\| zp->z_gen == 0 \|\|
	(dmu_objset_projectquota_enabled(zfsvfs->z_os) &&
	(zp->z_pflags & ZFS_PROJID) &&
	sa_lookup(zp->z_sa_hdl, SA_ZPL_PROJID(zfsvfs), &projid, 8) != 0)) {
	if (hdl == NULL)
	sa_handle_destroy(zp->z_sa_hdl);
	zfs_vnode_forget(vp);
	zp->z_vnode = NULL;
	zfs_znode_free_kmem(zp);
	return (NULL);
	}

	zp->z_projid = projid;
	zp->z_mode = mode;

	/* Cache the xattr parent id */
	if (zp->z_pflags & ZFS_XATTR)
	zp->z_xattr_parent = parent;

	vp->v_type = IFTOVT((mode_t)mode);

	switch (vp->v_type) {
	case VDIR:
	zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */
	break;
	case VFIFO:
	vp->v_op = &zfs_fifoops;
	break;
	case VREG:
	if (parent == zfsvfs->z_shares_dir) {
	ASSERT0(zp->z_uid);
	ASSERT0(zp->z_gid);
	vp->v_op = &zfs_shareops;
	}
	break;
	default:
	break;
	}

	mutex_enter(&zfsvfs->z_znodes_lock);
	list_insert_tail(&zfsvfs->z_all_znodes, zp);
	zp->z_zfsvfs = zfsvfs;
	mutex_exit(&zfsvfs->z_znodes_lock);

	#if __FreeBSD_version >= 1400077
	vn_set_state(vp, VSTATE_CONSTRUCTED);
	#endif
	VN_LOCK_AREC(vp);
	if (vp->v_type != VFIFO)
	VN_LOCK_ASHARE(vp);

	return (zp);
	}

	static uint64_t empty_xattr;
	static uint64_t pad[4];
	static zfs_acl_phys_t acl_phys;
	/*
	* Create a new DMU object to hold a zfs znode.
	*
	* IN: dzp - parent directory for new znode
	* vap - file attributes for new znode
	* tx - dmu transaction id for zap operations
	* cr - credentials of caller
	* flag - flags:
	* IS_ROOT_NODE - new object will be root
	* IS_XATTR - new object is an attribute
	* bonuslen - length of bonus buffer
	* setaclp - File/Dir initial ACL
	* fuidp - Tracks fuid allocation.
	*
	* OUT: zpp - allocated znode
	*
	*/
	void
	zfs_mknode(znode_t dzp, vattr_t vap, dmu_tx_t tx, cred_t cr,
	uint_t flag, znode_t *zpp, zfs_acl_ids_t acl_ids)
	{
	uint64_t crtime[2], atime[2], mtime[2], ctime[2];
	uint64_t mode, size, links, parent, pflags;
	uint64_t dzp_pflags = 0;
	uint64_t rdev = 0;
	zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
	dmu_buf_t *db;
	timestruc_t now;
	uint64_t gen, obj;
	int bonuslen;
	int dnodesize;
	sa_handle_t *sa_hdl;
	dmu_object_type_t obj_type;
	sa_bulk_attr_t *sa_attrs;
	int cnt = 0;
	zfs_acl_locator_cb_t locate = { 0 };

	ASSERT3P(vap, !=, NULL);
	ASSERT3U((vap->va_mask & AT_MODE), ==, AT_MODE);

	if (zfsvfs->z_replay) {
	obj = vap->va_nodeid;
	now = vap->va_ctime; /* see zfs_replay_create() */
	gen = vap->va_nblocks; /* ditto */
	dnodesize = vap->va_fsid; /* ditto */
	} else {
	obj = 0;
	vfs_timestamp(&now);
	gen = dmu_tx_get_txg(tx);
	dnodesize = dmu_objset_dnodesize(zfsvfs->z_os);
	}

	if (dnodesize == 0)
	dnodesize = DNODE_MIN_SIZE;

	obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
	bonuslen = (obj_type == DMU_OT_SA) ?
	DN_BONUS_SIZE(dnodesize) : ZFS_OLD_ZNODE_PHYS_SIZE;

	/*
	* Create a new DMU object.
	*/
	/*
	* There's currently no mechanism for pre-reading the blocks that will
	* be needed to allocate a new object, so we accept the small chance
	* that there will be an i/o error and we will fail one of the
	* assertions below.
	*/
	if (vap->va_type == VDIR) {
	if (zfsvfs->z_replay) {
	VERIFY0(zap_create_claim_norm_dnsize(zfsvfs->z_os, obj,
	zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
	obj_type, bonuslen, dnodesize, tx));
	} else {
	obj = zap_create_norm_dnsize(zfsvfs->z_os,
	zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
	obj_type, bonuslen, dnodesize, tx);
	}
	} else {
	if (zfsvfs->z_replay) {
	VERIFY0(dmu_object_claim_dnsize(zfsvfs->z_os, obj,
	DMU_OT_PLAIN_FILE_CONTENTS, 0,
	obj_type, bonuslen, dnodesize, tx));
	} else {
	obj = dmu_object_alloc_dnsize(zfsvfs->z_os,
	DMU_OT_PLAIN_FILE_CONTENTS, 0,
	obj_type, bonuslen, dnodesize, tx);
	}
	}

	ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
	VERIFY0(sa_buf_hold(zfsvfs->z_os, obj, NULL, &db));

	/*
	* If this is the root, fix up the half-initialized parent pointer
	* to reference the just-allocated physical data area.
	*/
	if (flag & IS_ROOT_NODE) {
	dzp->z_id = obj;
	} else {
	dzp_pflags = dzp->z_pflags;
	}

	/*
	* If parent is an xattr, so am I.
	*/
	if (dzp_pflags & ZFS_XATTR) {
	flag \|= IS_XATTR;
	}

	if (zfsvfs->z_use_fuids)
	pflags = ZFS_ARCHIVE \| ZFS_AV_MODIFIED;
	else
	pflags = 0;

	if (vap->va_type == VDIR) {
	size = 2; /* contents ("." and "..") */
	links = (flag & (IS_ROOT_NODE \| IS_XATTR)) ? 2 : 1;
	} else {
	size = links = 0;
	}

	if (vap->va_type == VBLK \|\| vap->va_type == VCHR) {
	rdev = zfs_expldev(vap->va_rdev);
	}

	parent = dzp->z_id;
	mode = acl_ids->z_mode;
	if (flag & IS_XATTR)
	pflags \|= ZFS_XATTR;

	/*
	* No execs denied will be determined when zfs_mode_compute() is called.
	*/
	pflags \|= acl_ids->z_aclp->z_hints &
	(ZFS_ACL_TRIVIAL\|ZFS_INHERIT_ACE\|ZFS_ACL_AUTO_INHERIT\|
	ZFS_ACL_DEFAULTED\|ZFS_ACL_PROTECTED);

	ZFS_TIME_ENCODE(&now, crtime);
	ZFS_TIME_ENCODE(&now, ctime);

	if (vap->va_mask & AT_ATIME) {
	ZFS_TIME_ENCODE(&vap->va_atime, atime);
	} else {
	ZFS_TIME_ENCODE(&now, atime);
	}

	if (vap->va_mask & AT_MTIME) {
	ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
	} else {
	ZFS_TIME_ENCODE(&now, mtime);
	}

	/* Now add in all of the "SA" attributes */
	VERIFY0(sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED,
	&sa_hdl));

	/*
	* Setup the array of attributes to be replaced/set on the new file
	*
	* order for DMU_OT_ZNODE is critical since it needs to be constructed
	* in the old znode_phys_t format. Don't change this ordering
	*/
	sa_attrs = kmem_alloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP);

	if (obj_type == DMU_OT_ZNODE) {
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
	NULL, &atime, 16);
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
	NULL, &mtime, 16);
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
	NULL, &ctime, 16);
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
	NULL, &crtime, 16);
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
	NULL, &gen, 8);
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
	NULL, &mode, 8);
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
	NULL, &size, 8);
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
	NULL, &parent, 8);
	} else {
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
	NULL, &mode, 8);
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
	NULL, &size, 8);
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
	NULL, &gen, 8);
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs),
	NULL, &acl_ids->z_fuid, 8);
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs),
	NULL, &acl_ids->z_fgid, 8);
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
	NULL, &parent, 8);
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
	NULL, &pflags, 8);
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
	NULL, &atime, 16);
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
	NULL, &mtime, 16);
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
	NULL, &ctime, 16);
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
	NULL, &crtime, 16);
	}

	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8);

	if (obj_type == DMU_OT_ZNODE) {
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL,
	&empty_xattr, 8);
	}
	if (obj_type == DMU_OT_ZNODE \|\|
	(vap->va_type == VBLK \|\| vap->va_type == VCHR)) {
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs),
	NULL, &rdev, 8);

	}
	if (obj_type == DMU_OT_ZNODE) {
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
	NULL, &pflags, 8);
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
	&acl_ids->z_fuid, 8);
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
	&acl_ids->z_fgid, 8);
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad,
	sizeof (uint64_t) * 4);
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
	&acl_phys, sizeof (zfs_acl_phys_t));
	} else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
	&acl_ids->z_aclp->z_acl_count, 8);
	locate.cb_aclp = acl_ids->z_aclp;
	SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs),
	zfs_acl_data_locator, &locate,
	acl_ids->z_aclp->z_acl_bytes);
	mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
	acl_ids->z_fuid, acl_ids->z_fgid);
	}

	VERIFY0(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx));

	if (!(flag & IS_ROOT_NODE)) {
	*zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl);
	ASSERT3P(*zpp, !=, NULL);
	} else {
	/*
	* If we are creating the root node, the "parent" we
	* passed in is the znode for the root.
	*/
	*zpp = dzp;

	(*zpp)->z_sa_hdl = sa_hdl;
	}

	(*zpp)->z_pflags = pflags;
	(*zpp)->z_mode = mode;
	(*zpp)->z_dnodesize = dnodesize;

	if (vap->va_mask & AT_XVATTR)
	zfs_xvattr_set(zpp, (xvattr_t )vap, tx);

	if (obj_type == DMU_OT_ZNODE \|\|
	acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
	VERIFY0(zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx));
	}
	if (!(flag & IS_ROOT_NODE)) {
	vnode_t vp = ZTOV(zpp);
	vp->v_vflag \|= VV_FORCEINSMQ;
	int err = insmntque(vp, zfsvfs->z_vfs);
	vp->v_vflag &= ~VV_FORCEINSMQ;
	(void) err;
	KASSERT(err == 0, ("insmntque() failed: error %d", err));
	}
	kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * ZPL_END);
	ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
	}

	/*
	* Update in-core attributes. It is assumed the caller will be doing an
	* sa_bulk_update to push the changes out.
	*/
	void
	zfs_xvattr_set(znode_t zp, xvattr_t xvap, dmu_tx_t *tx)
	{
	xoptattr_t *xoap;

	xoap = xva_getxoptattr(xvap);
	ASSERT3P(xoap, !=, NULL);

	if (zp->z_zfsvfs->z_replay == B_FALSE) {
	ASSERT_VOP_IN_SEQC(ZTOV(zp));
	}

	if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
	uint64_t times[2];
	ZFS_TIME_ENCODE(&xoap->xoa_createtime, times);
	(void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
	&times, sizeof (times), tx);
	XVA_SET_RTN(xvap, XAT_CREATETIME);
	}
	if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
	ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly,
	zp->z_pflags, tx);
	XVA_SET_RTN(xvap, XAT_READONLY);
	}
	if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
	ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden,
	zp->z_pflags, tx);
	XVA_SET_RTN(xvap, XAT_HIDDEN);
	}
	if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
	ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system,
	zp->z_pflags, tx);
	XVA_SET_RTN(xvap, XAT_SYSTEM);
	}
	if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
	ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive,
	zp->z_pflags, tx);
	XVA_SET_RTN(xvap, XAT_ARCHIVE);
	}
	if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
	ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable,
	zp->z_pflags, tx);
	XVA_SET_RTN(xvap, XAT_IMMUTABLE);
	}
	if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
	ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink,
	zp->z_pflags, tx);
	XVA_SET_RTN(xvap, XAT_NOUNLINK);
	}
	if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
	ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly,
	zp->z_pflags, tx);
	XVA_SET_RTN(xvap, XAT_APPENDONLY);
	}
	if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
	ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump,
	zp->z_pflags, tx);
	XVA_SET_RTN(xvap, XAT_NODUMP);
	}
	if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
	ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque,
	zp->z_pflags, tx);
	XVA_SET_RTN(xvap, XAT_OPAQUE);
	}
	if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
	ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
	xoap->xoa_av_quarantined, zp->z_pflags, tx);
	XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
	}
	if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
	ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified,
	zp->z_pflags, tx);
	XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
	}
	if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
	zfs_sa_set_scanstamp(zp, xvap, tx);
	XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
	}
	if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
	ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse,
	zp->z_pflags, tx);
	XVA_SET_RTN(xvap, XAT_REPARSE);
	}
	if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
	ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline,
	zp->z_pflags, tx);
	XVA_SET_RTN(xvap, XAT_OFFLINE);
	}
	if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
	ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse,
	zp->z_pflags, tx);
	XVA_SET_RTN(xvap, XAT_SPARSE);
	}
	}

	int
	zfs_zget(zfsvfs_t zfsvfs, uint64_t obj_num, znode_t *zpp)
	{
	dmu_object_info_t doi;
	dmu_buf_t *db;
	znode_t *zp;
	vnode_t *vp;
	sa_handle_t *hdl;
	int locked;
	int err;

	getnewvnode_reserve_();
	again:
	*zpp = NULL;
	ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);

	err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
	if (err) {
	ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
	getnewvnode_drop_reserve();
	return (err);
	}

	dmu_object_info_from_db(db, &doi);
	if (doi.doi_bonus_type != DMU_OT_SA &&
	(doi.doi_bonus_type != DMU_OT_ZNODE \|\|
	(doi.doi_bonus_type == DMU_OT_ZNODE &&
	doi.doi_bonus_size < sizeof (znode_phys_t)))) {
	sa_buf_rele(db, NULL);
	ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
	getnewvnode_drop_reserve();
	return (SET_ERROR(EINVAL));
	}

	hdl = dmu_buf_get_user(db);
	if (hdl != NULL) {
	zp = sa_get_userdata(hdl);

	/*
	* Since "SA" does immediate eviction we
	* should never find a sa handle that doesn't
	* know about the znode.
	*/
	ASSERT3P(zp, !=, NULL);
	ASSERT3U(zp->z_id, ==, obj_num);
	if (zp->z_unlinked) {
	err = SET_ERROR(ENOENT);
	} else {
	vp = ZTOV(zp);
	/*
	* Don't let the vnode disappear after
	* ZFS_OBJ_HOLD_EXIT.
	*/
	VN_HOLD(vp);
	*zpp = zp;
	err = 0;
	}

	sa_buf_rele(db, NULL);
	ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);

	if (err) {
	getnewvnode_drop_reserve();
	return (err);
	}

	locked = VOP_ISLOCKED(vp);
	VI_LOCK(vp);
	if (VN_IS_DOOMED(vp) && locked != LK_EXCLUSIVE) {
	/*
	* The vnode is doomed and this thread doesn't
	* hold the exclusive lock on it, so the vnode
	* must be being reclaimed by another thread.
	* Otherwise the doomed vnode is being reclaimed
	* by this thread and zfs_zget is called from
	* ZIL internals.
	*/
	VI_UNLOCK(vp);

	/*
	* XXX vrele() locks the vnode when the last reference
	* is dropped. Although in this case the vnode is
	* doomed / dead and so no inactivation is required,
	* the vnode lock is still acquired. That could result
	* in a LOR with z_teardown_lock if another thread holds
	* the vnode's lock and tries to take z_teardown_lock.
	* But that is only possible if the other thread peforms
	* a ZFS vnode operation on the vnode. That either
	* should not happen if the vnode is dead or the thread
	* should also have a reference to the vnode and thus
	* our reference is not last.
	*/
	VN_RELE(vp);
	goto again;
	}
	VI_UNLOCK(vp);
	getnewvnode_drop_reserve();
	return (err);
	}

	/*
	* Not found create new znode/vnode
	* but only if file exists.
	*
	* There is a small window where zfs_vget() could
	* find this object while a file create is still in
	* progress. This is checked for in zfs_znode_alloc()
	*
	* if zfs_znode_alloc() fails it will drop the hold on the
	* bonus buffer.
	*/
	zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size,
	doi.doi_bonus_type, NULL);
	if (zp == NULL) {
	err = SET_ERROR(ENOENT);
	} else {
	*zpp = zp;
	}
	if (err == 0) {
	vnode_t *vp = ZTOV(zp);

	err = insmntque(vp, zfsvfs->z_vfs);
	if (err == 0) {
	vp->v_hash = obj_num;
	VOP_UNLOCK1(vp);
	} else {
	zp->z_vnode = NULL;
	zfs_znode_dmu_fini(zp);
	zfs_znode_free(zp);
	*zpp = NULL;
	}
	}
	ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
	getnewvnode_drop_reserve();
	return (err);
	}

	int
	zfs_rezget(znode_t *zp)
	{
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	dmu_object_info_t doi;
	dmu_buf_t *db;
	vnode_t *vp;
	uint64_t obj_num = zp->z_id;
	uint64_t mode, size;
	sa_bulk_attr_t bulk[8];
	int err;
	int count = 0;
	uint64_t gen;

	/*
	* Remove cached pages before reloading the znode, so that they are not
	* lingering after we run into any error. Ideally, we should vgone()
	* the vnode in case of error, but currently we cannot do that
	* because of the LOR between the vnode lock and z_teardown_lock.
	* So, instead, we have to "doom" the znode in the illumos style.
	*
	* Ignore invalid pages during the scan. This is to avoid deadlocks
	* between page busying and the teardown lock, as pages are busied prior
	* to a VOP_GETPAGES operation, which acquires the teardown read lock.
	* Such pages will be invalid and can safely be skipped here.
	*/
	vp = ZTOV(zp);
	#if __FreeBSD_version >= 1400042
	vn_pages_remove_valid(vp, 0, 0);
	#else
	vn_pages_remove(vp, 0, 0);
	#endif

	ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);

	mutex_enter(&zp->z_acl_lock);
	if (zp->z_acl_cached) {
	zfs_acl_free(zp->z_acl_cached);
	zp->z_acl_cached = NULL;
	}
	mutex_exit(&zp->z_acl_lock);

	rw_enter(&zp->z_xattr_lock, RW_WRITER);
	if (zp->z_xattr_cached) {
	nvlist_free(zp->z_xattr_cached);
	zp->z_xattr_cached = NULL;
	}
	rw_exit(&zp->z_xattr_lock);

	ASSERT3P(zp->z_sa_hdl, ==, NULL);
	err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
	if (err) {
	ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
	return (err);
	}

	dmu_object_info_from_db(db, &doi);
	if (doi.doi_bonus_type != DMU_OT_SA &&
	(doi.doi_bonus_type != DMU_OT_ZNODE \|\|
	(doi.doi_bonus_type == DMU_OT_ZNODE &&
	doi.doi_bonus_size < sizeof (znode_phys_t)))) {
	sa_buf_rele(db, NULL);
	ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
	return (SET_ERROR(EINVAL));
	}

	zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL);
	size = zp->z_size;

	/* reload cached values */
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
	&gen, sizeof (gen));
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
	&zp->z_size, sizeof (zp->z_size));
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
	&zp->z_links, sizeof (zp->z_links));
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
	&zp->z_pflags, sizeof (zp->z_pflags));
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
	&zp->z_atime, sizeof (zp->z_atime));
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
	&zp->z_uid, sizeof (zp->z_uid));
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
	&zp->z_gid, sizeof (zp->z_gid));
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
	&mode, sizeof (mode));

	if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
	zfs_znode_dmu_fini(zp);
	ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
	return (SET_ERROR(EIO));
	}

	zp->z_mode = mode;

	if (gen != zp->z_gen) {
	zfs_znode_dmu_fini(zp);
	ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
	return (SET_ERROR(EIO));
	}

	/*
	* It is highly improbable but still quite possible that two
	* objects in different datasets are created with the same
	* object numbers and in transaction groups with the same
	* numbers. znodes corresponding to those objects would
	* have the same z_id and z_gen, but their other attributes
	* may be different.
	* zfs recv -F may replace one of such objects with the other.
	* As a result file properties recorded in the replaced
	* object's vnode may no longer match the received object's
	* properties. At present the only cached property is the
	* files type recorded in v_type.
	* So, handle this case by leaving the old vnode and znode
	* disassociated from the actual object. A new vnode and a
	* znode will be created if the object is accessed
	* (e.g. via a look-up). The old vnode and znode will be
	* recycled when the last vnode reference is dropped.
	*/
	if (vp->v_type != IFTOVT((mode_t)zp->z_mode)) {
	zfs_znode_dmu_fini(zp);
	ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
	return (SET_ERROR(EIO));
	}

	/*
	* If the file has zero links, then it has been unlinked on the send
	* side and it must be in the received unlinked set.
	* We call zfs_znode_dmu_fini() now to prevent any accesses to the
	* stale data and to prevent automatically removal of the file in
	* zfs_zinactive(). The file will be removed either when it is removed
	* on the send side and the next incremental stream is received or
	* when the unlinked set gets processed.
	*/
	zp->z_unlinked = (zp->z_links == 0);
	if (zp->z_unlinked) {
	zfs_znode_dmu_fini(zp);
	ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
	return (0);
	}

	zp->z_blksz = doi.doi_data_block_size;
	if (zp->z_size != size)
	vnode_pager_setsize(vp, zp->z_size);

	ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);

	return (0);
	}

	void
	zfs_znode_delete(znode_t zp, dmu_tx_t tx)
	{
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	objset_t *os = zfsvfs->z_os;
	uint64_t obj = zp->z_id;
	uint64_t acl_obj = zfs_external_acl(zp);

	ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
	if (acl_obj) {
	VERIFY(!zp->z_is_sa);
	VERIFY0(dmu_object_free(os, acl_obj, tx));
	}
	VERIFY0(dmu_object_free(os, obj, tx));
	zfs_znode_dmu_fini(zp);
	ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
	- zfs_znode_free(zp);
	}

	void
	zfs_zinactive(znode_t *zp)
	{
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	uint64_t z_id = zp->z_id;

	ASSERT3P(zp->z_sa_hdl, !=, NULL);

	/*
	* Don't allow a zfs_zget() while were trying to release this znode
	*/
	ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id);

	/*
	* If this was the last reference to a file with no links, remove
	* the file from the file system unless the file system is mounted
	* read-only. That can happen, for example, if the file system was
	* originally read-write, the file was opened, then unlinked and
	* the file system was made read-only before the file was finally
	* closed. The file will remain in the unlinked set.
	*/
	if (zp->z_unlinked) {
	ASSERT(!zfsvfs->z_issnap);
	if ((zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) == 0) {
	ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
	zfs_rmnode(zp);
	return;
	}
	}

	zfs_znode_dmu_fini(zp);
	ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
	zfs_znode_free(zp);
	}

	void
	zfs_znode_free(znode_t *zp)
	{
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	#if __FreeBSD_version >= 1300139
	char *symlink;
	#endif

	ASSERT3P(zp->z_sa_hdl, ==, NULL);
	zp->z_vnode = NULL;
	mutex_enter(&zfsvfs->z_znodes_lock);
	POINTER_INVALIDATE(&zp->z_zfsvfs);
	list_remove(&zfsvfs->z_all_znodes, zp);
	mutex_exit(&zfsvfs->z_znodes_lock);

	#if __FreeBSD_version >= 1300139
	symlink = atomic_load_ptr(&zp->z_cached_symlink);
	if (symlink != NULL) {
	atomic_store_rel_ptr((uintptr_t *)&zp->z_cached_symlink,
	(uintptr_t)NULL);
	cache_symlink_free(symlink, strlen(symlink) + 1);
	}
	#endif

	if (zp->z_acl_cached) {
	zfs_acl_free(zp->z_acl_cached);
	zp->z_acl_cached = NULL;
	}

	zfs_znode_free_kmem(zp);
	}

	void
	zfs_tstamp_update_setup_ext(znode_t *zp, uint_t flag, uint64_t mtime[2],
	uint64_t ctime[2], boolean_t have_tx)
	{
	timestruc_t now;

	vfs_timestamp(&now);

	if (have_tx) { /* will sa_bulk_update happen really soon? */
	zp->z_atime_dirty = 0;
	zp->z_seq++;
	} else {
	zp->z_atime_dirty = 1;
	}

	if (flag & AT_ATIME) {
	ZFS_TIME_ENCODE(&now, zp->z_atime);
	}

	if (flag & AT_MTIME) {
	ZFS_TIME_ENCODE(&now, mtime);
	if (zp->z_zfsvfs->z_use_fuids) {
	zp->z_pflags \|= (ZFS_ARCHIVE \|
	ZFS_AV_MODIFIED);
	}
	}

	if (flag & AT_CTIME) {
	ZFS_TIME_ENCODE(&now, ctime);
	if (zp->z_zfsvfs->z_use_fuids)
	zp->z_pflags \|= ZFS_ARCHIVE;
	}
	}


	void
	zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
	uint64_t ctime[2])
	{
	zfs_tstamp_update_setup_ext(zp, flag, mtime, ctime, B_TRUE);
	}
	/*
	* Grow the block size for a file.
	*
	* IN: zp - znode of file to free data in.
	* size - requested block size
	* tx - open transaction.
	*
	* NOTE: this function assumes that the znode is write locked.
	*/
	void
	zfs_grow_blocksize(znode_t zp, uint64_t size, dmu_tx_t tx)
	{
	int error;
	u_longlong_t dummy;

	if (size <= zp->z_blksz)
	return;
	/*
	* If the file size is already greater than the current blocksize,
	* we will not grow. If there is more than one block in a file,
	* the blocksize cannot change.
	*/
	if (zp->z_blksz && zp->z_size > zp->z_blksz)
	return;

	error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id,
	size, 0, tx);

	if (error == ENOTSUP)
	return;
	ASSERT0(error);

	/* What blocksize did we actually get? */
	dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
	}

	/*
	* Increase the file length
	*
	* IN: zp - znode of file to free data in.
	* end - new end-of-file
	*
	* RETURN: 0 on success, error code on failure
	*/
	static int
	zfs_extend(znode_t *zp, uint64_t end)
	{
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	dmu_tx_t *tx;
	zfs_locked_range_t *lr;
	uint64_t newblksz;
	int error;

	/*
	* We will change zp_size, lock the whole file.
	*/
	lr = zfs_rangelock_enter(&zp->z_rangelock, 0, UINT64_MAX, RL_WRITER);

	/*
	* Nothing to do if file already at desired length.
	*/
	if (end <= zp->z_size) {
	zfs_rangelock_exit(lr);
	return (0);
	}
	tx = dmu_tx_create(zfsvfs->z_os);
	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
	zfs_sa_upgrade_txholds(tx, zp);
	if (end > zp->z_blksz &&
	(!ISP2(zp->z_blksz) \|\| zp->z_blksz < zfsvfs->z_max_blksz)) {
	/*
	* We are growing the file past the current block size.
	*/
	if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) {
	/*
	* File's blocksize is already larger than the
	* "recordsize" property. Only let it grow to
	* the next power of 2.
	*/
	ASSERT(!ISP2(zp->z_blksz));
	newblksz = MIN(end, 1 << highbit64(zp->z_blksz));
	} else {
	newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz);
	}
	dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
	} else {
	newblksz = 0;
	}

	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error) {
	dmu_tx_abort(tx);
	zfs_rangelock_exit(lr);
	return (error);
	}

	if (newblksz)
	zfs_grow_blocksize(zp, newblksz, tx);

	zp->z_size = end;

	VERIFY0(sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zp->z_zfsvfs),
	&zp->z_size, sizeof (zp->z_size), tx));

	vnode_pager_setsize(ZTOV(zp), end);

	zfs_rangelock_exit(lr);

	dmu_tx_commit(tx);

	return (0);
	}

	/*
	* Free space in a file.
	*
	* IN: zp - znode of file to free data in.
	* off - start of section to free.
	* len - length of section to free.
	*
	* RETURN: 0 on success, error code on failure
	*/
	static int
	zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
	{
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	zfs_locked_range_t *lr;
	int error;

	/*
	* Lock the range being freed.
	*/
	lr = zfs_rangelock_enter(&zp->z_rangelock, off, len, RL_WRITER);

	/*
	* Nothing to do if file already at desired length.
	*/
	if (off >= zp->z_size) {
	zfs_rangelock_exit(lr);
	return (0);
	}

	if (off + len > zp->z_size)
	len = zp->z_size - off;

	error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len);

	if (error == 0) {
	#if __FreeBSD_version >= 1400032
	vnode_pager_purge_range(ZTOV(zp), off, off + len);
	#else
	/*
	* Before __FreeBSD_version 1400032 we cannot free block in the
	* middle of a file, but only at the end of a file, so this code
	* path should never happen.
	*/
	vnode_pager_setsize(ZTOV(zp), off);
	#endif
	}

	zfs_rangelock_exit(lr);

	return (error);
	}

	/*
	* Truncate a file
	*
	* IN: zp - znode of file to free data in.
	* end - new end-of-file.
	*
	* RETURN: 0 on success, error code on failure
	*/
	static int
	zfs_trunc(znode_t *zp, uint64_t end)
	{
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	vnode_t *vp = ZTOV(zp);
	dmu_tx_t *tx;
	zfs_locked_range_t *lr;
	int error;
	sa_bulk_attr_t bulk[2];
	int count = 0;

	/*
	* We will change zp_size, lock the whole file.
	*/
	lr = zfs_rangelock_enter(&zp->z_rangelock, 0, UINT64_MAX, RL_WRITER);

	/*
	* Nothing to do if file already at desired length.
	*/
	if (end >= zp->z_size) {
	zfs_rangelock_exit(lr);
	return (0);
	}

	error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end,
	DMU_OBJECT_END);
	if (error) {
	zfs_rangelock_exit(lr);
	return (error);
	}
	tx = dmu_tx_create(zfsvfs->z_os);
	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
	zfs_sa_upgrade_txholds(tx, zp);
	dmu_tx_mark_netfree(tx);
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error) {
	dmu_tx_abort(tx);
	zfs_rangelock_exit(lr);
	return (error);
	}

	zp->z_size = end;
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
	NULL, &zp->z_size, sizeof (zp->z_size));

	if (end == 0) {
	zp->z_pflags &= ~ZFS_SPARSE;
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
	NULL, &zp->z_pflags, 8);
	}
	VERIFY0(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx));

	dmu_tx_commit(tx);

	/*
	* Clear any mapped pages in the truncated region. This has to
	* happen outside of the transaction to avoid the possibility of
	* a deadlock with someone trying to push a page that we are
	* about to invalidate.
	*/
	vnode_pager_setsize(vp, end);

	zfs_rangelock_exit(lr);

	return (0);
	}

	/*
	* Free space in a file
	*
	* IN: zp - znode of file to free data in.
	* off - start of range
	* len - end of range (0 => EOF)
	* flag - current file open mode flags.
	* log - TRUE if this action should be logged
	*
	* RETURN: 0 on success, error code on failure
	*/
	int
	zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
	{
	dmu_tx_t *tx;
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	zilog_t *zilog = zfsvfs->z_log;
	uint64_t mode;
	uint64_t mtime[2], ctime[2];
	sa_bulk_attr_t bulk[3];
	int count = 0;
	int error;

	if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode,
	sizeof (mode))) != 0)
	return (error);

	if (off > zp->z_size) {
	error = zfs_extend(zp, off+len);
	if (error == 0 && log)
	goto log;
	else
	return (error);
	}

	if (len == 0) {
	error = zfs_trunc(zp, off);
	} else {
	if ((error = zfs_free_range(zp, off, len)) == 0 &&
	off + len > zp->z_size)
	error = zfs_extend(zp, off+len);
	}
	if (error \|\| !log)
	return (error);
	log:
	tx = dmu_tx_create(zfsvfs->z_os);
	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
	zfs_sa_upgrade_txholds(tx, zp);
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error) {
	dmu_tx_abort(tx);
	return (error);
	}

	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
	NULL, &zp->z_pflags, 8);
	zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime);
	error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
	ASSERT0(error);

	zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);

	dmu_tx_commit(tx);
	return (0);
	}

	void
	zfs_create_fs(objset_t os, cred_t cr, nvlist_t zplprops, dmu_tx_t tx)
	{
	uint64_t moid, obj, sa_obj, version;
	uint64_t sense = ZFS_CASE_SENSITIVE;
	uint64_t norm = 0;
	nvpair_t *elem;
	int error;
	int i;
	znode_t *rootzp = NULL;
	zfsvfs_t *zfsvfs;
	vattr_t vattr;
	znode_t *zp;
	zfs_acl_ids_t acl_ids;

	/*
	* First attempt to create master node.
	*/
	/*
	* In an empty objset, there are no blocks to read and thus
	* there can be no i/o errors (which we assert below).
	*/
	moid = MASTER_NODE_OBJ;
	error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
	DMU_OT_NONE, 0, tx);
	ASSERT0(error);

	/*
	* Set starting attributes.
	*/
	version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
	elem = NULL;
	while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
	/* For the moment we expect all zpl props to be uint64_ts */
	uint64_t val;
	const char *name;

	ASSERT3S(nvpair_type(elem), ==, DATA_TYPE_UINT64);
	val = fnvpair_value_uint64(elem);
	name = nvpair_name(elem);
	if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
	if (val < version)
	version = val;
	} else {
	error = zap_update(os, moid, name, 8, 1, &val, tx);
	}
	ASSERT0(error);
	if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
	norm = val;
	else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0)
	sense = val;
	}
	ASSERT3U(version, !=, 0);
	error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
	ASSERT0(error);

	/*
	* Create zap object used for SA attribute registration
	*/

	if (version >= ZPL_VERSION_SA) {
	sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
	DMU_OT_NONE, 0, tx);
	error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
	ASSERT0(error);
	} else {
	sa_obj = 0;
	}
	/*
	* Create a delete queue.
	*/
	obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);

	error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
	ASSERT0(error);

	/*
	* Create root znode. Create minimal znode/vnode/zfsvfs
	* to allow zfs_mknode to work.
	*/
	VATTR_NULL(&vattr);
	vattr.va_mask = AT_MODE\|AT_UID\|AT_GID;
	vattr.va_type = VDIR;
	vattr.va_mode = S_IFDIR\|0755;
	vattr.va_uid = crgetuid(cr);
	vattr.va_gid = crgetgid(cr);

	zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP);

	rootzp = zfs_znode_alloc_kmem(KM_SLEEP);
	ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs));
	rootzp->z_unlinked = 0;
	rootzp->z_atime_dirty = 0;
	rootzp->z_is_sa = USE_SA(version, os);

	zfsvfs->z_os = os;
	zfsvfs->z_parent = zfsvfs;
	zfsvfs->z_version = version;
	zfsvfs->z_use_fuids = USE_FUIDS(version, os);
	zfsvfs->z_use_sa = USE_SA(version, os);
	zfsvfs->z_norm = norm;

	error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
	&zfsvfs->z_attr_table);

	ASSERT0(error);

	/*
	* Fold case on file systems that are always or sometimes case
	* insensitive.
	*/
	if (sense == ZFS_CASE_INSENSITIVE \|\| sense == ZFS_CASE_MIXED)
	zfsvfs->z_norm \|= U8_TEXTPREP_TOUPPER;

	mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
	list_create(&zfsvfs->z_all_znodes, sizeof (znode_t),
	offsetof(znode_t, z_link_node));

	for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
	mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);

	rootzp->z_zfsvfs = zfsvfs;
	VERIFY0(zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr,
	cr, NULL, &acl_ids, NULL));
	zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids);
	ASSERT3P(zp, ==, rootzp);
	error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx);
	ASSERT0(error);
	zfs_acl_ids_free(&acl_ids);
	POINTER_INVALIDATE(&rootzp->z_zfsvfs);

	sa_handle_destroy(rootzp->z_sa_hdl);
	zfs_znode_free_kmem(rootzp);

	/*
	* Create shares directory
	*/

	error = zfs_create_share_dir(zfsvfs, tx);

	ASSERT0(error);

	for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
	mutex_destroy(&zfsvfs->z_hold_mtx[i]);
	kmem_free(zfsvfs, sizeof (zfsvfs_t));
	}
	#endif /* _KERNEL */

	static int
	zfs_sa_setup(objset_t osp, sa_attr_type_t *sa_table)
	{
	uint64_t sa_obj = 0;
	int error;

	error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj);
	if (error != 0 && error != ENOENT)
	return (error);

	error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table);
	return (error);
	}

	static int
	zfs_grab_sa_handle(objset_t osp, uint64_t obj, sa_handle_t *hdlp,
	dmu_buf_t *db, const void tag)
	{
	dmu_object_info_t doi;
	int error;

	if ((error = sa_buf_hold(osp, obj, tag, db)) != 0)
	return (error);

	dmu_object_info_from_db(*db, &doi);
	if ((doi.doi_bonus_type != DMU_OT_SA &&
	doi.doi_bonus_type != DMU_OT_ZNODE) \|\|
	(doi.doi_bonus_type == DMU_OT_ZNODE &&
	doi.doi_bonus_size < sizeof (znode_phys_t))) {
	sa_buf_rele(*db, tag);
	return (SET_ERROR(ENOTSUP));
	}

	error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp);
	if (error != 0) {
	sa_buf_rele(*db, tag);
	return (error);
	}

	return (0);
	}

	static void
	zfs_release_sa_handle(sa_handle_t hdl, dmu_buf_t db, const void *tag)
	{
	sa_handle_destroy(hdl);
	sa_buf_rele(db, tag);
	}

	/*
	* Given an object number, return its parent object number and whether
	* or not the object is an extended attribute directory.
	*/
	static int
	zfs_obj_to_pobj(objset_t osp, sa_handle_t hdl, sa_attr_type_t *sa_table,
	uint64_t pobjp, int is_xattrdir)
	{
	uint64_t parent;
	uint64_t pflags;
	uint64_t mode;
	uint64_t parent_mode;
	sa_bulk_attr_t bulk[3];
	sa_handle_t *sa_hdl;
	dmu_buf_t *sa_db;
	int count = 0;
	int error;

	SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL,
	&parent, sizeof (parent));
	SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL,
	&pflags, sizeof (pflags));
	SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
	&mode, sizeof (mode));

	if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0)
	return (error);

	/*
	* When a link is removed its parent pointer is not changed and will
	* be invalid. There are two cases where a link is removed but the
	* file stays around, when it goes to the delete queue and when there
	* are additional links.
	*/
	error = zfs_grab_sa_handle(osp, parent, &sa_hdl, &sa_db, FTAG);
	if (error != 0)
	return (error);

	error = sa_lookup(sa_hdl, ZPL_MODE, &parent_mode, sizeof (parent_mode));
	zfs_release_sa_handle(sa_hdl, sa_db, FTAG);
	if (error != 0)
	return (error);

	*is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode);

	/*
	* Extended attributes can be applied to files, directories, etc.
	* Otherwise the parent must be a directory.
	*/
	if (!*is_xattrdir && !S_ISDIR(parent_mode))
	return (SET_ERROR(EINVAL));

	*pobjp = parent;

	return (0);
	}

	/*
	* Given an object number, return some zpl level statistics
	*/
	static int
	zfs_obj_to_stats_impl(sa_handle_t hdl, sa_attr_type_t sa_table,
	zfs_stat_t *sb)
	{
	sa_bulk_attr_t bulk[4];
	int count = 0;

	SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
	&sb->zs_mode, sizeof (sb->zs_mode));
	SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL,
	&sb->zs_gen, sizeof (sb->zs_gen));
	SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL,
	&sb->zs_links, sizeof (sb->zs_links));
	SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL,
	&sb->zs_ctime, sizeof (sb->zs_ctime));

	return (sa_bulk_lookup(hdl, bulk, count));
	}

	static int
	zfs_obj_to_path_impl(objset_t osp, uint64_t obj, sa_handle_t hdl,
	sa_attr_type_t sa_table, char buf, int len)
	{
	sa_handle_t *sa_hdl;
	sa_handle_t *prevhdl = NULL;
	dmu_buf_t *prevdb = NULL;
	dmu_buf_t *sa_db = NULL;
	char *path = buf + len - 1;
	int error;

	*path = '\0';
	sa_hdl = hdl;

	uint64_t deleteq_obj;
	VERIFY0(zap_lookup(osp, MASTER_NODE_OBJ,
	ZFS_UNLINKED_SET, sizeof (uint64_t), 1, &deleteq_obj));
	error = zap_lookup_int(osp, deleteq_obj, obj);
	if (error == 0) {
	return (ESTALE);
	} else if (error != ENOENT) {
	return (error);
	}

	for (;;) {
	uint64_t pobj;
	char component[MAXNAMELEN + 2];
	size_t complen;
	int is_xattrdir;

	if (prevdb) {
	ASSERT3P(prevhdl, !=, NULL);
	zfs_release_sa_handle(prevhdl, prevdb, FTAG);
	}

	if ((error = zfs_obj_to_pobj(osp, sa_hdl, sa_table, &pobj,
	&is_xattrdir)) != 0)
	break;

	if (pobj == obj) {
	if (path[0] != '/')
	*--path = '/';
	break;
	}

	component[0] = '/';
	if (is_xattrdir) {
	(void) sprintf(component + 1, "<xattrdir>");
	} else {
	error = zap_value_search(osp, pobj, obj,
	ZFS_DIRENT_OBJ(-1ULL), component + 1);
	if (error != 0)
	break;
	}

	complen = strlen(component);
	path -= complen;
	ASSERT3P(path, >=, buf);
	memcpy(path, component, complen);
	obj = pobj;

	if (sa_hdl != hdl) {
	prevhdl = sa_hdl;
	prevdb = sa_db;
	}
	error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG);
	if (error != 0) {
	sa_hdl = prevhdl;
	sa_db = prevdb;
	break;
	}
	}

	if (sa_hdl != NULL && sa_hdl != hdl) {
	ASSERT3P(sa_db, !=, NULL);
	zfs_release_sa_handle(sa_hdl, sa_db, FTAG);
	}

	if (error == 0)
	(void) memmove(buf, path, buf + len - path);

	return (error);
	}

	int
	zfs_obj_to_path(objset_t osp, uint64_t obj, char buf, int len)
	{
	sa_attr_type_t *sa_table;
	sa_handle_t *hdl;
	dmu_buf_t *db;
	int error;

	error = zfs_sa_setup(osp, &sa_table);
	if (error != 0)
	return (error);

	error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG);
	if (error != 0)
	return (error);

	error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);

	zfs_release_sa_handle(hdl, db, FTAG);
	return (error);
	}

	int
	zfs_obj_to_stats(objset_t osp, uint64_t obj, zfs_stat_t sb,
	char *buf, int len)
	{
	char *path = buf + len - 1;
	sa_attr_type_t *sa_table;
	sa_handle_t *hdl;
	dmu_buf_t *db;
	int error;

	*path = '\0';

	error = zfs_sa_setup(osp, &sa_table);
	if (error != 0)
	return (error);

	error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG);
	if (error != 0)
	return (error);

	error = zfs_obj_to_stats_impl(hdl, sa_table, sb);
	if (error != 0) {
	zfs_release_sa_handle(hdl, db, FTAG);
	return (error);
	}

	error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);

	zfs_release_sa_handle(hdl, db, FTAG);
	return (error);
	}

	/*
	* Read a property stored within the master node.
	*/
	int
	zfs_get_zplprop(objset_t os, zfs_prop_t prop, uint64_t value)
	{
	uint64_t *cached_copy = NULL;

	/*
	* Figure out where in the objset_t the cached copy would live, if it
	* is available for the requested property.
	*/
	if (os != NULL) {
	switch (prop) {
	case ZFS_PROP_VERSION:
	cached_copy = &os->os_version;
	break;
	case ZFS_PROP_NORMALIZE:
	cached_copy = &os->os_normalization;
	break;
	case ZFS_PROP_UTF8ONLY:
	cached_copy = &os->os_utf8only;
	break;
	case ZFS_PROP_CASE:
	cached_copy = &os->os_casesensitivity;
	break;
	default:
	break;
	}
	}
	if (cached_copy != NULL && *cached_copy != OBJSET_PROP_UNINITIALIZED) {
	value = cached_copy;
	return (0);
	}

	/*
	* If the property wasn't cached, look up the file system's value for
	* the property. For the version property, we look up a slightly
	* different string.
	*/
	const char *pname;
	int error = ENOENT;
	if (prop == ZFS_PROP_VERSION) {
	pname = ZPL_VERSION_STR;
	} else {
	pname = zfs_prop_to_name(prop);
	}

	if (os != NULL) {
	ASSERT3U(os->os_phys->os_type, ==, DMU_OST_ZFS);
	error = zap_lookup(os, MASTER_NODE_OBJ, pname, 8, 1, value);
	}

	if (error == ENOENT) {
	/* No value set, use the default value */
	switch (prop) {
	case ZFS_PROP_VERSION:
	*value = ZPL_VERSION;
	break;
	case ZFS_PROP_NORMALIZE:
	case ZFS_PROP_UTF8ONLY:
	*value = 0;
	break;
	case ZFS_PROP_CASE:
	*value = ZFS_CASE_SENSITIVE;
	break;
	case ZFS_PROP_ACLTYPE:
	*value = ZFS_ACLTYPE_NFSV4;
	break;
	default:
	return (error);
	}
	error = 0;
	}

	/*
	* If one of the methods for getting the property value above worked,
	* copy it into the objset_t's cache.
	*/
	if (error == 0 && cached_copy != NULL) {
	cached_copy = value;
	}

	return (error);
	}



	void
	zfs_znode_update_vfs(znode_t *zp)
	{
	vm_object_t object;

	if ((object = ZTOV(zp)->v_object) == NULL \|\|
	zp->z_size == object->un_pager.vnp.vnp_size)
	return;

	vnode_pager_setsize(ZTOV(zp), zp->z_size);
	}


	#ifdef _KERNEL
	int
	zfs_znode_parent_and_name(znode_t zp, znode_t dzpp, char buf)
	{
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;
	uint64_t parent;
	int is_xattrdir;
	int err;

	/* Extended attributes should not be visible as regular files. */
	if ((zp->z_pflags & ZFS_XATTR) != 0)
	return (SET_ERROR(EINVAL));

	err = zfs_obj_to_pobj(zfsvfs->z_os, zp->z_sa_hdl, zfsvfs->z_attr_table,
	&parent, &is_xattrdir);
	if (err != 0)
	return (err);
	ASSERT0(is_xattrdir);

	/* No name as this is a root object. */
	if (parent == zp->z_id)
	return (SET_ERROR(EINVAL));

	err = zap_value_search(zfsvfs->z_os, parent, zp->z_id,
	ZFS_DIRENT_OBJ(-1ULL), buf);
	if (err != 0)
	return (err);
	err = zfs_zget(zfsvfs, parent, dzpp);
	return (err);
	}
	#endif /* _KERNEL */

	#ifdef _KERNEL
	int
	zfs_rlimit_fsize(off_t fsize)
	{
	struct thread *td = curthread;
	off_t lim;

	if (td == NULL)
	return (0);

	lim = lim_cur(td, RLIMIT_FSIZE);
	if (__predict_true((uoff_t)fsize <= lim))
	return (0);

	/*
	* The limit is reached.
	*/
	PROC_LOCK(td->td_proc);
	kern_psignal(td->td_proc, SIGXFSZ);
	PROC_UNLOCK(td->td_proc);

	return (EFBIG);
	}
	#endif /* _KERNEL */
	diff --git a/sys/contrib/openzfs/module/os/linux/spl/spl-procfs-list.c b/sys/contrib/openzfs/module/os/linux/spl/spl-procfs-list.c
	index 5e073950d61a..91840ed2ca22 100644
	--- a/sys/contrib/openzfs/module/os/linux/spl/spl-procfs-list.c
	+++ b/sys/contrib/openzfs/module/os/linux/spl/spl-procfs-list.c
	@@ -1,284 +1,284 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2018 by Delphix. All rights reserved.
	*/

	#include <sys/list.h>
	#include <sys/procfs_list.h>
	#include <linux/proc_fs.h>
	#include <sys/mutex.h>

	/*
	* A procfs_list is a wrapper around a linked list which implements the seq_file
	* interface, allowing the contents of the list to be exposed through procfs.
	* The kernel already has some utilities to help implement the seq_file
	* interface for linked lists (seq_list_*), but they aren't appropriate for use
	* with lists that have many entries, because seq_list_start walks the list at
	* the start of each read syscall to find where it left off, so reading a file
	* ends up being quadratic in the number of entries in the list.
	*
	* This implementation avoids this penalty by maintaining a separate cursor into
	* the list per instance of the file that is open. It also maintains some extra
	* information in each node of the list to prevent reads of entries that have
	* been dropped from the list.
	*
	* Callers should only add elements to the list using procfs_list_add, which
	* adds an element to the tail of the list. Other operations can be performed
	* directly on the wrapped list using the normal list manipulation functions,
	* but elements should only be removed from the head of the list.
	*/

	#define NODE_ID(procfs_list, obj) \
	(((procfs_list_node_t )(((char )obj) + \
	(procfs_list)->pl_node_offset))->pln_id)

	typedef struct procfs_list_cursor {
	procfs_list_t procfs_list; / List into which this cursor points */
	void cached_node; / Most recently accessed node */
	loff_t cached_pos; /* Position of cached_node */
	} procfs_list_cursor_t;

	static int
	procfs_list_seq_show(struct seq_file f, void p)
	{
	procfs_list_cursor_t *cursor = f->private;
	procfs_list_t *procfs_list = cursor->procfs_list;

	ASSERT(MUTEX_HELD(&procfs_list->pl_lock));
	if (p == SEQ_START_TOKEN) {
	if (procfs_list->pl_show_header != NULL)
	return (procfs_list->pl_show_header(f));
	else
	return (0);
	}
	return (procfs_list->pl_show(f, p));
	}

	static void *
	procfs_list_next_node(procfs_list_cursor_t cursor, loff_t pos)
	{
	void *next_node;
	procfs_list_t *procfs_list = cursor->procfs_list;

	if (cursor->cached_node == SEQ_START_TOKEN)
	next_node = list_head(&procfs_list->pl_list);
	else
	next_node = list_next(&procfs_list->pl_list,
	cursor->cached_node);

	if (next_node != NULL) {
	cursor->cached_node = next_node;
	cursor->cached_pos = NODE_ID(procfs_list, cursor->cached_node);
	*pos = cursor->cached_pos;
	} else {
	/*
	* seq_read() expects ->next() to update the position even
	* when there are no more entries. Advance the position to
	* prevent a warning from being logged.
	*/
	cursor->cached_node = NULL;
	cursor->cached_pos++;
	*pos = cursor->cached_pos;
	}

	return (next_node);
	}

	static void *
	procfs_list_seq_start(struct seq_file f, loff_t pos)
	{
	procfs_list_cursor_t *cursor = f->private;
	procfs_list_t *procfs_list = cursor->procfs_list;

	mutex_enter(&procfs_list->pl_lock);

	if (*pos == 0) {
	cursor->cached_node = SEQ_START_TOKEN;
	cursor->cached_pos = 0;
	return (SEQ_START_TOKEN);
	} else if (cursor->cached_node == NULL) {
	return (NULL);
	}

	/*
	* Check if our cached pointer has become stale, which happens if the
	* the message where we left off has been dropped from the list since
	* the last read syscall completed.
	*/
	void *oldest_node = list_head(&procfs_list->pl_list);
	if (cursor->cached_node != SEQ_START_TOKEN && (oldest_node == NULL \|\|
	NODE_ID(procfs_list, oldest_node) > cursor->cached_pos))
	return (ERR_PTR(-EIO));

	/*
	* If it isn't starting from the beginning of the file, the seq_file
	* code will either pick up at the same position it visited last or the
	* following one.
	*/
	if (*pos == cursor->cached_pos) {
	return (cursor->cached_node);
	} else {
	ASSERT3U(*pos, ==, cursor->cached_pos + 1);
	return (procfs_list_next_node(cursor, pos));
	}
	}

	static void *
	procfs_list_seq_next(struct seq_file f, void p, loff_t *pos)
	{
	procfs_list_cursor_t *cursor = f->private;
	ASSERT(MUTEX_HELD(&cursor->procfs_list->pl_lock));
	return (procfs_list_next_node(cursor, pos));
	}

	static void
	procfs_list_seq_stop(struct seq_file f, void p)
	{
	procfs_list_cursor_t *cursor = f->private;
	procfs_list_t *procfs_list = cursor->procfs_list;
	mutex_exit(&procfs_list->pl_lock);
	}

	static const struct seq_operations procfs_list_seq_ops = {
	.show = procfs_list_seq_show,
	.start = procfs_list_seq_start,
	.next = procfs_list_seq_next,
	.stop = procfs_list_seq_stop,
	};

	static int
	procfs_list_open(struct inode inode, struct file filp)
	{
	int rc = seq_open_private(filp, &procfs_list_seq_ops,
	sizeof (procfs_list_cursor_t));
	if (rc != 0)
	return (rc);

	struct seq_file *f = filp->private_data;
	procfs_list_cursor_t *cursor = f->private;
	cursor->procfs_list = SPL_PDE_DATA(inode);
	cursor->cached_node = NULL;
	cursor->cached_pos = 0;

	return (0);
	}

	static ssize_t
	procfs_list_write(struct file filp, const char __user buf, size_t len,
	loff_t *ppos)
	{
	struct seq_file *f = filp->private_data;
	procfs_list_cursor_t *cursor = f->private;
	procfs_list_t *procfs_list = cursor->procfs_list;
	int rc;

	if (procfs_list->pl_clear != NULL &&
	(rc = procfs_list->pl_clear(procfs_list)) != 0)
	return (-rc);
	return (len);
	}

	static const kstat_proc_op_t procfs_list_operations = {
	#ifdef HAVE_PROC_OPS_STRUCT
	.proc_open = procfs_list_open,
	.proc_write = procfs_list_write,
	.proc_read = seq_read,
	.proc_lseek = seq_lseek,
	.proc_release = seq_release_private,
	#else
	.open = procfs_list_open,
	.write = procfs_list_write,
	.read = seq_read,
	.llseek = seq_lseek,
	.release = seq_release_private,
	#endif
	};

	/*
	* Initialize a procfs_list and create a file for it in the proc filesystem
	* under the kstat namespace.
	*/
	void
	procfs_list_install(const char *module,
	const char *submodule,
	const char *name,
	mode_t mode,
	procfs_list_t *procfs_list,
	int (show)(struct seq_file f, void *p),
	int (show_header)(struct seq_file f),
	int (clear)(procfs_list_t procfs_list),
	size_t procfs_list_node_off)
	{
	char *modulestr;

	if (submodule != NULL)
	modulestr = kmem_asprintf("%s/%s", module, submodule);
	else
	modulestr = kmem_asprintf("%s", module);
	- mutex_init(&procfs_list->pl_lock, NULL, MUTEX_DEFAULT, NULL);
	+ mutex_init(&procfs_list->pl_lock, NULL, MUTEX_NOLOCKDEP, NULL);
	list_create(&procfs_list->pl_list,
	procfs_list_node_off + sizeof (procfs_list_node_t),
	procfs_list_node_off + offsetof(procfs_list_node_t, pln_link));
	procfs_list->pl_next_id = 1; /* Save id 0 for SEQ_START_TOKEN */
	procfs_list->pl_show = show;
	procfs_list->pl_show_header = show_header;
	procfs_list->pl_clear = clear;
	procfs_list->pl_node_offset = procfs_list_node_off;

	kstat_proc_entry_init(&procfs_list->pl_kstat_entry, modulestr, name);
	kstat_proc_entry_install(&procfs_list->pl_kstat_entry, mode,
	&procfs_list_operations, procfs_list);
	kmem_strfree(modulestr);
	}
	EXPORT_SYMBOL(procfs_list_install);

	/* Remove the proc filesystem file corresponding to the given list */
	void
	procfs_list_uninstall(procfs_list_t *procfs_list)
	{
	kstat_proc_entry_delete(&procfs_list->pl_kstat_entry);
	}
	EXPORT_SYMBOL(procfs_list_uninstall);

	void
	procfs_list_destroy(procfs_list_t *procfs_list)
	{
	ASSERT(list_is_empty(&procfs_list->pl_list));
	list_destroy(&procfs_list->pl_list);
	mutex_destroy(&procfs_list->pl_lock);
	}
	EXPORT_SYMBOL(procfs_list_destroy);

	/*
	* Add a new node to the tail of the list. While the standard list manipulation
	* functions can be use for all other operation, adding elements to the list
	* should only be done using this helper so that the id of the new node is set
	* correctly.
	*/
	void
	procfs_list_add(procfs_list_t procfs_list, void p)
	{
	ASSERT(MUTEX_HELD(&procfs_list->pl_lock));
	NODE_ID(procfs_list, p) = procfs_list->pl_next_id++;
	list_insert_tail(&procfs_list->pl_list, p);
	}
	EXPORT_SYMBOL(procfs_list_add);
	diff --git a/sys/contrib/openzfs/module/os/linux/spl/spl-taskq.c b/sys/contrib/openzfs/module/os/linux/spl/spl-taskq.c
	index c384b7b378c3..e7b812c3b5b5 100644
	--- a/sys/contrib/openzfs/module/os/linux/spl/spl-taskq.c
	+++ b/sys/contrib/openzfs/module/os/linux/spl/spl-taskq.c
	@@ -1,1475 +1,1475 @@
	/*
	* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
	* Copyright (C) 2007 The Regents of the University of California.
	* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
	* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
	* UCRL-CODE-235197
	*
	* This file is part of the SPL, Solaris Porting Layer.
	*
	* The SPL is free software; you can redistribute it and/or modify it
	* under the terms of the GNU General Public License as published by the
	* Free Software Foundation; either version 2 of the License, or (at your
	* option) any later version.
	*
	* The SPL is distributed in the hope that it will be useful, but WITHOUT
	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
	* for more details.
	*
	* You should have received a copy of the GNU General Public License along
	* with the SPL. If not, see <http://www.gnu.org/licenses/>.
	*
	* Solaris Porting Layer (SPL) Task Queue Implementation.
	*/

	#include <sys/timer.h>
	#include <sys/taskq.h>
	#include <sys/kmem.h>
	#include <sys/tsd.h>
	#include <sys/trace_spl.h>
	#ifdef HAVE_CPU_HOTPLUG
	#include <linux/cpuhotplug.h>
	#endif

	static int spl_taskq_thread_bind = 0;
	module_param(spl_taskq_thread_bind, int, 0644);
	MODULE_PARM_DESC(spl_taskq_thread_bind, "Bind taskq thread to CPU by default");

	static uint_t spl_taskq_thread_timeout_ms = 5000;
	/* BEGIN CSTYLED */
	module_param(spl_taskq_thread_timeout_ms, uint, 0644);
	/* END CSTYLED */
	MODULE_PARM_DESC(spl_taskq_thread_timeout_ms,
	"Minimum idle threads exit interval for dynamic taskqs");

	static int spl_taskq_thread_dynamic = 1;
	module_param(spl_taskq_thread_dynamic, int, 0444);
	MODULE_PARM_DESC(spl_taskq_thread_dynamic, "Allow dynamic taskq threads");

	static int spl_taskq_thread_priority = 1;
	module_param(spl_taskq_thread_priority, int, 0644);
	MODULE_PARM_DESC(spl_taskq_thread_priority,
	"Allow non-default priority for taskq threads");

	static uint_t spl_taskq_thread_sequential = 4;
	/* BEGIN CSTYLED */
	module_param(spl_taskq_thread_sequential, uint, 0644);
	/* END CSTYLED */
	MODULE_PARM_DESC(spl_taskq_thread_sequential,
	"Create new taskq threads after N sequential tasks");

	/*
	* Global system-wide dynamic task queue available for all consumers. This
	* taskq is not intended for long-running tasks; instead, a dedicated taskq
	* should be created.
	*/
	taskq_t *system_taskq;
	EXPORT_SYMBOL(system_taskq);
	/* Global dynamic task queue for long delay */
	taskq_t *system_delay_taskq;
	EXPORT_SYMBOL(system_delay_taskq);

	/* Private dedicated taskq for creating new taskq threads on demand. */
	static taskq_t *dynamic_taskq;
	static taskq_thread_t taskq_thread_create(taskq_t );

	#ifdef HAVE_CPU_HOTPLUG
	/* Multi-callback id for cpu hotplugging. */
	static int spl_taskq_cpuhp_state;
	#endif

	/* List of all taskqs */
	LIST_HEAD(tq_list);
	struct rw_semaphore tq_list_sem;
	static uint_t taskq_tsd;

	static int
	task_km_flags(uint_t flags)
	{
	if (flags & TQ_NOSLEEP)
	return (KM_NOSLEEP);

	if (flags & TQ_PUSHPAGE)
	return (KM_PUSHPAGE);

	return (KM_SLEEP);
	}

	/*
	* taskq_find_by_name - Find the largest instance number of a named taskq.
	*/
	static int
	taskq_find_by_name(const char *name)
	{
	struct list_head *tql = NULL;
	taskq_t *tq;

	list_for_each_prev(tql, &tq_list) {
	tq = list_entry(tql, taskq_t, tq_taskqs);
	if (strcmp(name, tq->tq_name) == 0)
	return (tq->tq_instance);
	}
	return (-1);
	}

	/*
	* NOTE: Must be called with tq->tq_lock held, returns a list_t which
	* is not attached to the free, work, or pending taskq lists.
	*/
	static taskq_ent_t *
	task_alloc(taskq_t tq, uint_t flags, unsigned long irqflags)
	{
	taskq_ent_t *t;
	int count = 0;

	ASSERT(tq);
	retry:
	/* Acquire taskq_ent_t's from free list if available */
	if (!list_empty(&tq->tq_free_list) && !(flags & TQ_NEW)) {
	t = list_entry(tq->tq_free_list.next, taskq_ent_t, tqent_list);

	ASSERT(!(t->tqent_flags & TQENT_FLAG_PREALLOC));
	ASSERT(!(t->tqent_flags & TQENT_FLAG_CANCEL));
	ASSERT(!timer_pending(&t->tqent_timer));

	list_del_init(&t->tqent_list);
	return (t);
	}

	/* Free list is empty and memory allocations are prohibited */
	if (flags & TQ_NOALLOC)
	return (NULL);

	/* Hit maximum taskq_ent_t pool size */
	if (tq->tq_nalloc >= tq->tq_maxalloc) {
	if (flags & TQ_NOSLEEP)
	return (NULL);

	/*
	* Sleep periodically polling the free list for an available
	* taskq_ent_t. Dispatching with TQ_SLEEP should always succeed
	* but we cannot block forever waiting for an taskq_ent_t to
	* show up in the free list, otherwise a deadlock can happen.
	*
	* Therefore, we need to allocate a new task even if the number
	* of allocated tasks is above tq->tq_maxalloc, but we still
	* end up delaying the task allocation by one second, thereby
	* throttling the task dispatch rate.
	*/
	spin_unlock_irqrestore(&tq->tq_lock, *irqflags);
	- schedule_timeout(HZ / 100);
	+ schedule_timeout_interruptible(HZ / 100);
	spin_lock_irqsave_nested(&tq->tq_lock, *irqflags,
	tq->tq_lock_class);
	if (count < 100) {
	count++;
	goto retry;
	}
	}

	spin_unlock_irqrestore(&tq->tq_lock, *irqflags);
	t = kmem_alloc(sizeof (taskq_ent_t), task_km_flags(flags));
	spin_lock_irqsave_nested(&tq->tq_lock, *irqflags, tq->tq_lock_class);

	if (t) {
	taskq_init_ent(t);
	tq->tq_nalloc++;
	}

	return (t);
	}

	/*
	* NOTE: Must be called with tq->tq_lock held, expects the taskq_ent_t
	* to already be removed from the free, work, or pending taskq lists.
	*/
	static void
	task_free(taskq_t tq, taskq_ent_t t)
	{
	ASSERT(tq);
	ASSERT(t);
	ASSERT(list_empty(&t->tqent_list));
	ASSERT(!timer_pending(&t->tqent_timer));

	kmem_free(t, sizeof (taskq_ent_t));
	tq->tq_nalloc--;
	}

	/*
	* NOTE: Must be called with tq->tq_lock held, either destroys the
	* taskq_ent_t if too many exist or moves it to the free list for later use.
	*/
	static void
	task_done(taskq_t tq, taskq_ent_t t)
	{
	ASSERT(tq);
	ASSERT(t);

	/* Wake tasks blocked in taskq_wait_id() */
	wake_up_all(&t->tqent_waitq);

	list_del_init(&t->tqent_list);

	if (tq->tq_nalloc <= tq->tq_minalloc) {
	t->tqent_id = TASKQID_INVALID;
	t->tqent_func = NULL;
	t->tqent_arg = NULL;
	t->tqent_flags = 0;

	list_add_tail(&t->tqent_list, &tq->tq_free_list);
	} else {
	task_free(tq, t);
	}
	}

	/*
	* When a delayed task timer expires remove it from the delay list and
	* add it to the priority list in order for immediate processing.
	*/
	static void
	task_expire_impl(taskq_ent_t *t)
	{
	taskq_ent_t *w;
	taskq_t *tq = t->tqent_taskq;
	struct list_head *l = NULL;
	unsigned long flags;

	spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);

	if (t->tqent_flags & TQENT_FLAG_CANCEL) {
	ASSERT(list_empty(&t->tqent_list));
	spin_unlock_irqrestore(&tq->tq_lock, flags);
	return;
	}

	t->tqent_birth = jiffies;
	DTRACE_PROBE1(taskq_ent__birth, taskq_ent_t *, t);

	/*
	* The priority list must be maintained in strict task id order
	* from lowest to highest for lowest_id to be easily calculable.
	*/
	list_del(&t->tqent_list);
	list_for_each_prev(l, &tq->tq_prio_list) {
	w = list_entry(l, taskq_ent_t, tqent_list);
	if (w->tqent_id < t->tqent_id) {
	list_add(&t->tqent_list, l);
	break;
	}
	}
	if (l == &tq->tq_prio_list)
	list_add(&t->tqent_list, &tq->tq_prio_list);

	spin_unlock_irqrestore(&tq->tq_lock, flags);

	wake_up(&tq->tq_work_waitq);
	}

	static void
	task_expire(spl_timer_list_t tl)
	{
	struct timer_list tmr = (struct timer_list )tl;
	taskq_ent_t *t = from_timer(t, tmr, tqent_timer);
	task_expire_impl(t);
	}

	/*
	* Returns the lowest incomplete taskqid_t. The taskqid_t may
	* be queued on the pending list, on the priority list, on the
	* delay list, or on the work list currently being handled, but
	* it is not 100% complete yet.
	*/
	static taskqid_t
	taskq_lowest_id(taskq_t *tq)
	{
	taskqid_t lowest_id = tq->tq_next_id;
	taskq_ent_t *t;
	taskq_thread_t *tqt;

	if (!list_empty(&tq->tq_pend_list)) {
	t = list_entry(tq->tq_pend_list.next, taskq_ent_t, tqent_list);
	lowest_id = MIN(lowest_id, t->tqent_id);
	}

	if (!list_empty(&tq->tq_prio_list)) {
	t = list_entry(tq->tq_prio_list.next, taskq_ent_t, tqent_list);
	lowest_id = MIN(lowest_id, t->tqent_id);
	}

	if (!list_empty(&tq->tq_delay_list)) {
	t = list_entry(tq->tq_delay_list.next, taskq_ent_t, tqent_list);
	lowest_id = MIN(lowest_id, t->tqent_id);
	}

	if (!list_empty(&tq->tq_active_list)) {
	tqt = list_entry(tq->tq_active_list.next, taskq_thread_t,
	tqt_active_list);
	ASSERT(tqt->tqt_id != TASKQID_INVALID);
	lowest_id = MIN(lowest_id, tqt->tqt_id);
	}

	return (lowest_id);
	}

	/*
	* Insert a task into a list keeping the list sorted by increasing taskqid.
	*/
	static void
	taskq_insert_in_order(taskq_t tq, taskq_thread_t tqt)
	{
	taskq_thread_t *w;
	struct list_head *l = NULL;

	ASSERT(tq);
	ASSERT(tqt);

	list_for_each_prev(l, &tq->tq_active_list) {
	w = list_entry(l, taskq_thread_t, tqt_active_list);
	if (w->tqt_id < tqt->tqt_id) {
	list_add(&tqt->tqt_active_list, l);
	break;
	}
	}
	if (l == &tq->tq_active_list)
	list_add(&tqt->tqt_active_list, &tq->tq_active_list);
	}

	/*
	* Find and return a task from the given list if it exists. The list
	* must be in lowest to highest task id order.
	*/
	static taskq_ent_t *
	taskq_find_list(taskq_t tq, struct list_head lh, taskqid_t id)
	{
	struct list_head *l = NULL;
	taskq_ent_t *t;

	list_for_each(l, lh) {
	t = list_entry(l, taskq_ent_t, tqent_list);

	if (t->tqent_id == id)
	return (t);

	if (t->tqent_id > id)
	break;
	}

	return (NULL);
	}

	/*
	* Find an already dispatched task given the task id regardless of what
	* state it is in. If a task is still pending it will be returned.
	* If a task is executing, then -EBUSY will be returned instead.
	* If the task has already been run then NULL is returned.
	*/
	static taskq_ent_t *
	taskq_find(taskq_t *tq, taskqid_t id)
	{
	taskq_thread_t *tqt;
	struct list_head *l = NULL;
	taskq_ent_t *t;

	t = taskq_find_list(tq, &tq->tq_delay_list, id);
	if (t)
	return (t);

	t = taskq_find_list(tq, &tq->tq_prio_list, id);
	if (t)
	return (t);

	t = taskq_find_list(tq, &tq->tq_pend_list, id);
	if (t)
	return (t);

	list_for_each(l, &tq->tq_active_list) {
	tqt = list_entry(l, taskq_thread_t, tqt_active_list);
	if (tqt->tqt_id == id) {
	/*
	* Instead of returning tqt_task, we just return a non
	* NULL value to prevent misuse, since tqt_task only
	* has two valid fields.
	*/
	return (ERR_PTR(-EBUSY));
	}
	}

	return (NULL);
	}

	/*
	* Theory for the taskq_wait_id(), taskq_wait_outstanding(), and
	* taskq_wait() functions below.
	*
	* Taskq waiting is accomplished by tracking the lowest outstanding task
	* id and the next available task id. As tasks are dispatched they are
	* added to the tail of the pending, priority, or delay lists. As worker
	* threads become available the tasks are removed from the heads of these
	* lists and linked to the worker threads. This ensures the lists are
	* kept sorted by lowest to highest task id.
	*
	* Therefore the lowest outstanding task id can be quickly determined by
	* checking the head item from all of these lists. This value is stored
	* with the taskq as the lowest id. It only needs to be recalculated when
	* either the task with the current lowest id completes or is canceled.
	*
	* By blocking until the lowest task id exceeds the passed task id the
	* taskq_wait_outstanding() function can be easily implemented. Similarly,
	* by blocking until the lowest task id matches the next task id taskq_wait()
	* can be implemented.
	*
	* Callers should be aware that when there are multiple worked threads it
	* is possible for larger task ids to complete before smaller ones. Also
	* when the taskq contains delay tasks with small task ids callers may
	* block for a considerable length of time waiting for them to expire and
	* execute.
	*/
	static int
	taskq_wait_id_check(taskq_t *tq, taskqid_t id)
	{
	int rc;
	unsigned long flags;

	spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
	rc = (taskq_find(tq, id) == NULL);
	spin_unlock_irqrestore(&tq->tq_lock, flags);

	return (rc);
	}

	/*
	* The taskq_wait_id() function blocks until the passed task id completes.
	* This does not guarantee that all lower task ids have completed.
	*/
	void
	taskq_wait_id(taskq_t *tq, taskqid_t id)
	{
	wait_event(tq->tq_wait_waitq, taskq_wait_id_check(tq, id));
	}
	EXPORT_SYMBOL(taskq_wait_id);

	static int
	taskq_wait_outstanding_check(taskq_t *tq, taskqid_t id)
	{
	int rc;
	unsigned long flags;

	spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
	rc = (id < tq->tq_lowest_id);
	spin_unlock_irqrestore(&tq->tq_lock, flags);

	return (rc);
	}

	/*
	* The taskq_wait_outstanding() function will block until all tasks with a
	* lower taskqid than the passed 'id' have been completed. Note that all
	* task id's are assigned monotonically at dispatch time. Zero may be
	* passed for the id to indicate all tasks dispatch up to this point,
	* but not after, should be waited for.
	*/
	void
	taskq_wait_outstanding(taskq_t *tq, taskqid_t id)
	{
	id = id ? id : tq->tq_next_id - 1;
	wait_event(tq->tq_wait_waitq, taskq_wait_outstanding_check(tq, id));
	}
	EXPORT_SYMBOL(taskq_wait_outstanding);

	static int
	taskq_wait_check(taskq_t *tq)
	{
	int rc;
	unsigned long flags;

	spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
	rc = (tq->tq_lowest_id == tq->tq_next_id);
	spin_unlock_irqrestore(&tq->tq_lock, flags);

	return (rc);
	}

	/*
	* The taskq_wait() function will block until the taskq is empty.
	* This means that if a taskq re-dispatches work to itself taskq_wait()
	* callers will block indefinitely.
	*/
	void
	taskq_wait(taskq_t *tq)
	{
	wait_event(tq->tq_wait_waitq, taskq_wait_check(tq));
	}
	EXPORT_SYMBOL(taskq_wait);

	int
	taskq_member(taskq_t tq, kthread_t t)
	{
	return (tq == (taskq_t *)tsd_get_by_thread(taskq_tsd, t));
	}
	EXPORT_SYMBOL(taskq_member);

	taskq_t *
	taskq_of_curthread(void)
	{
	return (tsd_get(taskq_tsd));
	}
	EXPORT_SYMBOL(taskq_of_curthread);

	/*
	* Cancel an already dispatched task given the task id. Still pending tasks
	* will be immediately canceled, and if the task is active the function will
	* block until it completes. Preallocated tasks which are canceled must be
	* freed by the caller.
	*/
	int
	taskq_cancel_id(taskq_t *tq, taskqid_t id)
	{
	taskq_ent_t *t;
	int rc = ENOENT;
	unsigned long flags;

	ASSERT(tq);

	spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
	t = taskq_find(tq, id);
	if (t && t != ERR_PTR(-EBUSY)) {
	list_del_init(&t->tqent_list);
	t->tqent_flags \|= TQENT_FLAG_CANCEL;

	/*
	* When canceling the lowest outstanding task id we
	* must recalculate the new lowest outstanding id.
	*/
	if (tq->tq_lowest_id == t->tqent_id) {
	tq->tq_lowest_id = taskq_lowest_id(tq);
	ASSERT3S(tq->tq_lowest_id, >, t->tqent_id);
	}

	/*
	* The task_expire() function takes the tq->tq_lock so drop
	* drop the lock before synchronously cancelling the timer.
	*/
	if (timer_pending(&t->tqent_timer)) {
	spin_unlock_irqrestore(&tq->tq_lock, flags);
	del_timer_sync(&t->tqent_timer);
	spin_lock_irqsave_nested(&tq->tq_lock, flags,
	tq->tq_lock_class);
	}

	if (!(t->tqent_flags & TQENT_FLAG_PREALLOC))
	task_done(tq, t);

	rc = 0;
	}
	spin_unlock_irqrestore(&tq->tq_lock, flags);

	if (t == ERR_PTR(-EBUSY)) {
	taskq_wait_id(tq, id);
	rc = EBUSY;
	}

	return (rc);
	}
	EXPORT_SYMBOL(taskq_cancel_id);

	static int taskq_thread_spawn(taskq_t *tq);

	taskqid_t
	taskq_dispatch(taskq_t tq, task_func_t func, void arg, uint_t flags)
	{
	taskq_ent_t *t;
	taskqid_t rc = TASKQID_INVALID;
	unsigned long irqflags;

	ASSERT(tq);
	ASSERT(func);

	spin_lock_irqsave_nested(&tq->tq_lock, irqflags, tq->tq_lock_class);

	/* Taskq being destroyed and all tasks drained */
	if (!(tq->tq_flags & TASKQ_ACTIVE))
	goto out;

	/* Do not queue the task unless there is idle thread for it */
	ASSERT(tq->tq_nactive <= tq->tq_nthreads);
	if ((flags & TQ_NOQUEUE) && (tq->tq_nactive == tq->tq_nthreads)) {
	/* Dynamic taskq may be able to spawn another thread */
	if (taskq_thread_spawn(tq) == 0)
	goto out;
	}

	if ((t = task_alloc(tq, flags, &irqflags)) == NULL)
	goto out;

	spin_lock(&t->tqent_lock);

	/* Queue to the front of the list to enforce TQ_NOQUEUE semantics */
	if (flags & TQ_NOQUEUE)
	list_add(&t->tqent_list, &tq->tq_prio_list);
	/* Queue to the priority list instead of the pending list */
	else if (flags & TQ_FRONT)
	list_add_tail(&t->tqent_list, &tq->tq_prio_list);
	else
	list_add_tail(&t->tqent_list, &tq->tq_pend_list);

	t->tqent_id = rc = tq->tq_next_id;
	tq->tq_next_id++;
	t->tqent_func = func;
	t->tqent_arg = arg;
	t->tqent_taskq = tq;
	t->tqent_timer.function = NULL;
	t->tqent_timer.expires = 0;

	t->tqent_birth = jiffies;
	DTRACE_PROBE1(taskq_ent__birth, taskq_ent_t *, t);

	ASSERT(!(t->tqent_flags & TQENT_FLAG_PREALLOC));

	spin_unlock(&t->tqent_lock);

	wake_up(&tq->tq_work_waitq);

	/* Spawn additional taskq threads if required. */
	if (!(flags & TQ_NOQUEUE) && tq->tq_nactive == tq->tq_nthreads)
	(void) taskq_thread_spawn(tq);
	out:
	spin_unlock_irqrestore(&tq->tq_lock, irqflags);
	return (rc);
	}
	EXPORT_SYMBOL(taskq_dispatch);

	taskqid_t
	taskq_dispatch_delay(taskq_t tq, task_func_t func, void arg,
	uint_t flags, clock_t expire_time)
	{
	taskqid_t rc = TASKQID_INVALID;
	taskq_ent_t *t;
	unsigned long irqflags;

	ASSERT(tq);
	ASSERT(func);

	spin_lock_irqsave_nested(&tq->tq_lock, irqflags, tq->tq_lock_class);

	/* Taskq being destroyed and all tasks drained */
	if (!(tq->tq_flags & TASKQ_ACTIVE))
	goto out;

	if ((t = task_alloc(tq, flags, &irqflags)) == NULL)
	goto out;

	spin_lock(&t->tqent_lock);

	/* Queue to the delay list for subsequent execution */
	list_add_tail(&t->tqent_list, &tq->tq_delay_list);

	t->tqent_id = rc = tq->tq_next_id;
	tq->tq_next_id++;
	t->tqent_func = func;
	t->tqent_arg = arg;
	t->tqent_taskq = tq;
	t->tqent_timer.function = task_expire;
	t->tqent_timer.expires = (unsigned long)expire_time;
	add_timer(&t->tqent_timer);

	ASSERT(!(t->tqent_flags & TQENT_FLAG_PREALLOC));

	spin_unlock(&t->tqent_lock);

	/* Spawn additional taskq threads if required. */
	if (tq->tq_nactive == tq->tq_nthreads)
	(void) taskq_thread_spawn(tq);
	out:
	spin_unlock_irqrestore(&tq->tq_lock, irqflags);
	return (rc);
	}
	EXPORT_SYMBOL(taskq_dispatch_delay);

	void
	taskq_dispatch_ent(taskq_t tq, task_func_t func, void arg, uint_t flags,
	taskq_ent_t *t)
	{
	unsigned long irqflags;
	ASSERT(tq);
	ASSERT(func);

	spin_lock_irqsave_nested(&tq->tq_lock, irqflags,
	tq->tq_lock_class);

	/* Taskq being destroyed and all tasks drained */
	if (!(tq->tq_flags & TASKQ_ACTIVE)) {
	t->tqent_id = TASKQID_INVALID;
	goto out;
	}

	if ((flags & TQ_NOQUEUE) && (tq->tq_nactive == tq->tq_nthreads)) {
	/* Dynamic taskq may be able to spawn another thread */
	if (taskq_thread_spawn(tq) == 0)
	goto out;
	flags \|= TQ_FRONT;
	}

	spin_lock(&t->tqent_lock);

	/*
	* Make sure the entry is not on some other taskq; it is important to
	* ASSERT() under lock
	*/
	ASSERT(taskq_empty_ent(t));

	/*
	* Mark it as a prealloc'd task. This is important
	* to ensure that we don't free it later.
	*/
	t->tqent_flags \|= TQENT_FLAG_PREALLOC;

	/* Queue to the priority list instead of the pending list */
	if (flags & TQ_FRONT)
	list_add_tail(&t->tqent_list, &tq->tq_prio_list);
	else
	list_add_tail(&t->tqent_list, &tq->tq_pend_list);

	t->tqent_id = tq->tq_next_id;
	tq->tq_next_id++;
	t->tqent_func = func;
	t->tqent_arg = arg;
	t->tqent_taskq = tq;

	t->tqent_birth = jiffies;
	DTRACE_PROBE1(taskq_ent__birth, taskq_ent_t *, t);

	spin_unlock(&t->tqent_lock);

	wake_up(&tq->tq_work_waitq);

	/* Spawn additional taskq threads if required. */
	if (tq->tq_nactive == tq->tq_nthreads)
	(void) taskq_thread_spawn(tq);
	out:
	spin_unlock_irqrestore(&tq->tq_lock, irqflags);
	}
	EXPORT_SYMBOL(taskq_dispatch_ent);

	int
	taskq_empty_ent(taskq_ent_t *t)
	{
	return (list_empty(&t->tqent_list));
	}
	EXPORT_SYMBOL(taskq_empty_ent);

	void
	taskq_init_ent(taskq_ent_t *t)
	{
	spin_lock_init(&t->tqent_lock);
	init_waitqueue_head(&t->tqent_waitq);
	timer_setup(&t->tqent_timer, NULL, 0);
	INIT_LIST_HEAD(&t->tqent_list);
	t->tqent_id = 0;
	t->tqent_func = NULL;
	t->tqent_arg = NULL;
	t->tqent_flags = 0;
	t->tqent_taskq = NULL;
	}
	EXPORT_SYMBOL(taskq_init_ent);

	/*
	* Return the next pending task, preference is given to tasks on the
	* priority list which were dispatched with TQ_FRONT.
	*/
	static taskq_ent_t *
	taskq_next_ent(taskq_t *tq)
	{
	struct list_head *list;

	if (!list_empty(&tq->tq_prio_list))
	list = &tq->tq_prio_list;
	else if (!list_empty(&tq->tq_pend_list))
	list = &tq->tq_pend_list;
	else
	return (NULL);

	return (list_entry(list->next, taskq_ent_t, tqent_list));
	}

	/*
	* Spawns a new thread for the specified taskq.
	*/
	static void
	taskq_thread_spawn_task(void *arg)
	{
	taskq_t tq = (taskq_t )arg;
	unsigned long flags;

	if (taskq_thread_create(tq) == NULL) {
	/* restore spawning count if failed */
	spin_lock_irqsave_nested(&tq->tq_lock, flags,
	tq->tq_lock_class);
	tq->tq_nspawn--;
	spin_unlock_irqrestore(&tq->tq_lock, flags);
	}
	}

	/*
	* Spawn addition threads for dynamic taskqs (TASKQ_DYNAMIC) the current
	* number of threads is insufficient to handle the pending tasks. These
	* new threads must be created by the dedicated dynamic_taskq to avoid
	* deadlocks between thread creation and memory reclaim. The system_taskq
	* which is also a dynamic taskq cannot be safely used for this.
	*/
	static int
	taskq_thread_spawn(taskq_t *tq)
	{
	int spawning = 0;

	if (!(tq->tq_flags & TASKQ_DYNAMIC))
	return (0);

	tq->lastspawnstop = jiffies;
	if ((tq->tq_nthreads + tq->tq_nspawn < tq->tq_maxthreads) &&
	(tq->tq_flags & TASKQ_ACTIVE)) {
	spawning = (++tq->tq_nspawn);
	taskq_dispatch(dynamic_taskq, taskq_thread_spawn_task,
	tq, TQ_NOSLEEP);
	}

	return (spawning);
	}

	/*
	* Threads in a dynamic taskq may exit once there is no more work to do.
	* To prevent threads from being created and destroyed too often limit
	* the exit rate to one per spl_taskq_thread_timeout_ms.
	*
	* The first thread is the thread list is treated as the primary thread.
	* There is nothing special about the primary thread but in order to avoid
	* all the taskq pids from changing we opt to make it long running.
	*/
	static int
	taskq_thread_should_stop(taskq_t tq, taskq_thread_t tqt)
	{
	ASSERT(!taskq_next_ent(tq));
	if (!(tq->tq_flags & TASKQ_DYNAMIC) \|\| !spl_taskq_thread_dynamic)
	return (0);
	if (!(tq->tq_flags & TASKQ_ACTIVE))
	return (1);
	if (list_first_entry(&(tq->tq_thread_list), taskq_thread_t,
	tqt_thread_list) == tqt)
	return (0);
	ASSERT3U(tq->tq_nthreads, >, 1);
	if (tq->tq_nspawn != 0)
	return (0);
	if (time_before(jiffies, tq->lastspawnstop +
	msecs_to_jiffies(spl_taskq_thread_timeout_ms)))
	return (0);
	tq->lastspawnstop = jiffies;
	return (1);
	}

	static int
	taskq_thread(void *args)
	{
	DECLARE_WAITQUEUE(wait, current);
	sigset_t blocked;
	taskq_thread_t *tqt = args;
	taskq_t *tq;
	taskq_ent_t *t;
	int seq_tasks = 0;
	unsigned long flags;
	taskq_ent_t dup_task = {};

	ASSERT(tqt);
	ASSERT(tqt->tqt_tq);
	tq = tqt->tqt_tq;
	current->flags \|= PF_NOFREEZE;

	(void) spl_fstrans_mark();

	sigfillset(&blocked);
	sigprocmask(SIG_BLOCK, &blocked, NULL);
	flush_signals(current);

	tsd_set(taskq_tsd, tq);
	spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
	/*
	* If we are dynamically spawned, decrease spawning count. Note that
	* we could be created during taskq_create, in which case we shouldn't
	* do the decrement. But it's fine because taskq_create will reset
	* tq_nspawn later.
	*/
	if (tq->tq_flags & TASKQ_DYNAMIC)
	tq->tq_nspawn--;

	/* Immediately exit if more threads than allowed were created. */
	if (tq->tq_nthreads >= tq->tq_maxthreads)
	goto error;

	tq->tq_nthreads++;
	list_add_tail(&tqt->tqt_thread_list, &tq->tq_thread_list);
	wake_up(&tq->tq_wait_waitq);
	set_current_state(TASK_INTERRUPTIBLE);

	while (!kthread_should_stop()) {

	if (list_empty(&tq->tq_pend_list) &&
	list_empty(&tq->tq_prio_list)) {

	if (taskq_thread_should_stop(tq, tqt))
	break;

	add_wait_queue_exclusive(&tq->tq_work_waitq, &wait);
	spin_unlock_irqrestore(&tq->tq_lock, flags);

	schedule();
	seq_tasks = 0;

	spin_lock_irqsave_nested(&tq->tq_lock, flags,
	tq->tq_lock_class);
	remove_wait_queue(&tq->tq_work_waitq, &wait);
	} else {
	__set_current_state(TASK_RUNNING);
	}

	if ((t = taskq_next_ent(tq)) != NULL) {
	list_del_init(&t->tqent_list);

	/*
	* A TQENT_FLAG_PREALLOC task may be reused or freed
	* during the task function call. Store tqent_id and
	* tqent_flags here.
	*
	* Also use an on stack taskq_ent_t for tqt_task
	* assignment in this case; we want to make sure
	* to duplicate all fields, so the values are
	* correct when it's accessed via DTRACE_PROBE*.
	*/
	tqt->tqt_id = t->tqent_id;
	tqt->tqt_flags = t->tqent_flags;

	if (t->tqent_flags & TQENT_FLAG_PREALLOC) {
	dup_task = *t;
	t = &dup_task;
	}
	tqt->tqt_task = t;

	taskq_insert_in_order(tq, tqt);
	tq->tq_nactive++;
	spin_unlock_irqrestore(&tq->tq_lock, flags);

	DTRACE_PROBE1(taskq_ent__start, taskq_ent_t *, t);

	/* Perform the requested task */
	t->tqent_func(t->tqent_arg);

	DTRACE_PROBE1(taskq_ent__finish, taskq_ent_t *, t);

	spin_lock_irqsave_nested(&tq->tq_lock, flags,
	tq->tq_lock_class);
	tq->tq_nactive--;
	list_del_init(&tqt->tqt_active_list);
	tqt->tqt_task = NULL;

	/* For prealloc'd tasks, we don't free anything. */
	if (!(tqt->tqt_flags & TQENT_FLAG_PREALLOC))
	task_done(tq, t);

	/*
	* When the current lowest outstanding taskqid is
	* done calculate the new lowest outstanding id
	*/
	if (tq->tq_lowest_id == tqt->tqt_id) {
	tq->tq_lowest_id = taskq_lowest_id(tq);
	ASSERT3S(tq->tq_lowest_id, >, tqt->tqt_id);
	}

	/* Spawn additional taskq threads if required. */
	if ((++seq_tasks) > spl_taskq_thread_sequential &&
	taskq_thread_spawn(tq))
	seq_tasks = 0;

	tqt->tqt_id = TASKQID_INVALID;
	tqt->tqt_flags = 0;
	wake_up_all(&tq->tq_wait_waitq);
	}

	set_current_state(TASK_INTERRUPTIBLE);

	}

	__set_current_state(TASK_RUNNING);
	tq->tq_nthreads--;
	list_del_init(&tqt->tqt_thread_list);
	error:
	kmem_free(tqt, sizeof (taskq_thread_t));
	spin_unlock_irqrestore(&tq->tq_lock, flags);

	tsd_set(taskq_tsd, NULL);
	thread_exit();

	return (0);
	}

	static taskq_thread_t *
	taskq_thread_create(taskq_t *tq)
	{
	static int last_used_cpu = 0;
	taskq_thread_t *tqt;

	tqt = kmem_alloc(sizeof (*tqt), KM_PUSHPAGE);
	INIT_LIST_HEAD(&tqt->tqt_thread_list);
	INIT_LIST_HEAD(&tqt->tqt_active_list);
	tqt->tqt_tq = tq;
	tqt->tqt_id = TASKQID_INVALID;

	tqt->tqt_thread = spl_kthread_create(taskq_thread, tqt,
	"%s", tq->tq_name);
	if (tqt->tqt_thread == NULL) {
	kmem_free(tqt, sizeof (taskq_thread_t));
	return (NULL);
	}

	if (spl_taskq_thread_bind) {
	last_used_cpu = (last_used_cpu + 1) % num_online_cpus();
	kthread_bind(tqt->tqt_thread, last_used_cpu);
	}

	if (spl_taskq_thread_priority)
	set_user_nice(tqt->tqt_thread, PRIO_TO_NICE(tq->tq_pri));

	wake_up_process(tqt->tqt_thread);

	return (tqt);
	}

	taskq_t *
	taskq_create(const char *name, int threads_arg, pri_t pri,
	int minalloc, int maxalloc, uint_t flags)
	{
	taskq_t *tq;
	taskq_thread_t *tqt;
	int count = 0, rc = 0, i;
	unsigned long irqflags;
	int nthreads = threads_arg;

	ASSERT(name != NULL);
	ASSERT(minalloc >= 0);
	ASSERT(!(flags & (TASKQ_CPR_SAFE))); /* Unsupported */

	/* Scale the number of threads using nthreads as a percentage */
	if (flags & TASKQ_THREADS_CPU_PCT) {
	ASSERT(nthreads <= 100);
	ASSERT(nthreads >= 0);
	nthreads = MIN(threads_arg, 100);
	nthreads = MAX(nthreads, 0);
	nthreads = MAX((num_online_cpus() * nthreads) /100, 1);
	}

	tq = kmem_alloc(sizeof (*tq), KM_PUSHPAGE);
	if (tq == NULL)
	return (NULL);

	tq->tq_hp_support = B_FALSE;
	#ifdef HAVE_CPU_HOTPLUG
	if (flags & TASKQ_THREADS_CPU_PCT) {
	tq->tq_hp_support = B_TRUE;
	if (cpuhp_state_add_instance_nocalls(spl_taskq_cpuhp_state,
	&tq->tq_hp_cb_node) != 0) {
	kmem_free(tq, sizeof (*tq));
	return (NULL);
	}
	}
	#endif

	spin_lock_init(&tq->tq_lock);
	INIT_LIST_HEAD(&tq->tq_thread_list);
	INIT_LIST_HEAD(&tq->tq_active_list);
	tq->tq_name = kmem_strdup(name);
	tq->tq_nactive = 0;
	tq->tq_nthreads = 0;
	tq->tq_nspawn = 0;
	tq->tq_maxthreads = nthreads;
	tq->tq_cpu_pct = threads_arg;
	tq->tq_pri = pri;
	tq->tq_minalloc = minalloc;
	tq->tq_maxalloc = maxalloc;
	tq->tq_nalloc = 0;
	tq->tq_flags = (flags \| TASKQ_ACTIVE);
	tq->tq_next_id = TASKQID_INITIAL;
	tq->tq_lowest_id = TASKQID_INITIAL;
	tq->lastspawnstop = jiffies;
	INIT_LIST_HEAD(&tq->tq_free_list);
	INIT_LIST_HEAD(&tq->tq_pend_list);
	INIT_LIST_HEAD(&tq->tq_prio_list);
	INIT_LIST_HEAD(&tq->tq_delay_list);
	init_waitqueue_head(&tq->tq_work_waitq);
	init_waitqueue_head(&tq->tq_wait_waitq);
	tq->tq_lock_class = TQ_LOCK_GENERAL;
	INIT_LIST_HEAD(&tq->tq_taskqs);

	if (flags & TASKQ_PREPOPULATE) {
	spin_lock_irqsave_nested(&tq->tq_lock, irqflags,
	tq->tq_lock_class);

	for (i = 0; i < minalloc; i++)
	task_done(tq, task_alloc(tq, TQ_PUSHPAGE \| TQ_NEW,
	&irqflags));

	spin_unlock_irqrestore(&tq->tq_lock, irqflags);
	}

	if ((flags & TASKQ_DYNAMIC) && spl_taskq_thread_dynamic)
	nthreads = 1;

	for (i = 0; i < nthreads; i++) {
	tqt = taskq_thread_create(tq);
	if (tqt == NULL)
	rc = 1;
	else
	count++;
	}

	/* Wait for all threads to be started before potential destroy */
	wait_event(tq->tq_wait_waitq, tq->tq_nthreads == count);
	/*
	* taskq_thread might have touched nspawn, but we don't want them to
	* because they're not dynamically spawned. So we reset it to 0
	*/
	tq->tq_nspawn = 0;

	if (rc) {
	taskq_destroy(tq);
	tq = NULL;
	} else {
	down_write(&tq_list_sem);
	tq->tq_instance = taskq_find_by_name(name) + 1;
	list_add_tail(&tq->tq_taskqs, &tq_list);
	up_write(&tq_list_sem);
	}

	return (tq);
	}
	EXPORT_SYMBOL(taskq_create);

	void
	taskq_destroy(taskq_t *tq)
	{
	struct task_struct *thread;
	taskq_thread_t *tqt;
	taskq_ent_t *t;
	unsigned long flags;

	ASSERT(tq);
	spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
	tq->tq_flags &= ~TASKQ_ACTIVE;
	spin_unlock_irqrestore(&tq->tq_lock, flags);

	#ifdef HAVE_CPU_HOTPLUG
	if (tq->tq_hp_support) {
	VERIFY0(cpuhp_state_remove_instance_nocalls(
	spl_taskq_cpuhp_state, &tq->tq_hp_cb_node));
	}
	#endif
	/*
	* When TASKQ_ACTIVE is clear new tasks may not be added nor may
	* new worker threads be spawned for dynamic taskq.
	*/
	if (dynamic_taskq != NULL)
	taskq_wait_outstanding(dynamic_taskq, 0);

	taskq_wait(tq);

	/* remove taskq from global list used by the kstats */
	down_write(&tq_list_sem);
	list_del(&tq->tq_taskqs);
	up_write(&tq_list_sem);

	spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
	/* wait for spawning threads to insert themselves to the list */
	while (tq->tq_nspawn) {
	spin_unlock_irqrestore(&tq->tq_lock, flags);
	schedule_timeout_interruptible(1);
	spin_lock_irqsave_nested(&tq->tq_lock, flags,
	tq->tq_lock_class);
	}

	/*
	* Signal each thread to exit and block until it does. Each thread
	* is responsible for removing itself from the list and freeing its
	* taskq_thread_t. This allows for idle threads to opt to remove
	* themselves from the taskq. They can be recreated as needed.
	*/
	while (!list_empty(&tq->tq_thread_list)) {
	tqt = list_entry(tq->tq_thread_list.next,
	taskq_thread_t, tqt_thread_list);
	thread = tqt->tqt_thread;
	spin_unlock_irqrestore(&tq->tq_lock, flags);

	kthread_stop(thread);

	spin_lock_irqsave_nested(&tq->tq_lock, flags,
	tq->tq_lock_class);
	}

	while (!list_empty(&tq->tq_free_list)) {
	t = list_entry(tq->tq_free_list.next, taskq_ent_t, tqent_list);

	ASSERT(!(t->tqent_flags & TQENT_FLAG_PREALLOC));

	list_del_init(&t->tqent_list);
	task_free(tq, t);
	}

	ASSERT0(tq->tq_nthreads);
	ASSERT0(tq->tq_nalloc);
	ASSERT0(tq->tq_nspawn);
	ASSERT(list_empty(&tq->tq_thread_list));
	ASSERT(list_empty(&tq->tq_active_list));
	ASSERT(list_empty(&tq->tq_free_list));
	ASSERT(list_empty(&tq->tq_pend_list));
	ASSERT(list_empty(&tq->tq_prio_list));
	ASSERT(list_empty(&tq->tq_delay_list));

	spin_unlock_irqrestore(&tq->tq_lock, flags);

	kmem_strfree(tq->tq_name);
	kmem_free(tq, sizeof (taskq_t));
	}
	EXPORT_SYMBOL(taskq_destroy);

	/*
	* Create a taskq with a specified number of pool threads. Allocate
	* and return an array of nthreads kthread_t pointers, one for each
	* thread in the pool. The array is not ordered and must be freed
	* by the caller.
	*/
	taskq_t *
	taskq_create_synced(const char *name, int nthreads, pri_t pri,
	int minalloc, int maxalloc, uint_t flags, kthread_t ***ktpp)
	{
	taskq_t *tq;
	taskq_thread_t *tqt;
	int i = 0;
	kthread_t *kthreads = kmem_zalloc(sizeof (kthreads) * nthreads,
	KM_SLEEP);

	flags &= ~(TASKQ_DYNAMIC \| TASKQ_THREADS_CPU_PCT \| TASKQ_DC_BATCH);

	/* taskq_create spawns all the threads before returning */
	tq = taskq_create(name, nthreads, minclsyspri, nthreads, INT_MAX,
	flags \| TASKQ_PREPOPULATE);
	VERIFY(tq != NULL);
	VERIFY(tq->tq_nthreads == nthreads);

	list_for_each_entry(tqt, &tq->tq_thread_list, tqt_thread_list) {
	kthreads[i] = tqt->tqt_thread;
	i++;
	}

	ASSERT3S(i, ==, nthreads);
	*ktpp = kthreads;

	return (tq);
	}
	EXPORT_SYMBOL(taskq_create_synced);

	static unsigned int spl_taskq_kick = 0;

	/*
	* 2.6.36 API Change
	* module_param_cb is introduced to take kernel_param_ops and
	* module_param_call is marked as obsolete. Also set and get operations
	* were changed to take a 'const struct kernel_param *'.
	*/
	static int
	#ifdef module_param_cb
	param_set_taskq_kick(const char val, const struct kernel_param kp)
	#else
	param_set_taskq_kick(const char val, struct kernel_param kp)
	#endif
	{
	int ret;
	taskq_t *tq = NULL;
	taskq_ent_t *t;
	unsigned long flags;

	ret = param_set_uint(val, kp);
	if (ret < 0 \|\| !spl_taskq_kick)
	return (ret);
	/* reset value */
	spl_taskq_kick = 0;

	down_read(&tq_list_sem);
	list_for_each_entry(tq, &tq_list, tq_taskqs) {
	spin_lock_irqsave_nested(&tq->tq_lock, flags,
	tq->tq_lock_class);
	/* Check if the first pending is older than 5 seconds */
	t = taskq_next_ent(tq);
	if (t && time_after(jiffies, t->tqent_birth + 5*HZ)) {
	(void) taskq_thread_spawn(tq);
	printk(KERN_INFO "spl: Kicked taskq %s/%d\n",
	tq->tq_name, tq->tq_instance);
	}
	spin_unlock_irqrestore(&tq->tq_lock, flags);
	}
	up_read(&tq_list_sem);
	return (ret);
	}

	#ifdef module_param_cb
	static const struct kernel_param_ops param_ops_taskq_kick = {
	.set = param_set_taskq_kick,
	.get = param_get_uint,
	};
	module_param_cb(spl_taskq_kick, &param_ops_taskq_kick, &spl_taskq_kick, 0644);
	#else
	module_param_call(spl_taskq_kick, param_set_taskq_kick, param_get_uint,
	&spl_taskq_kick, 0644);
	#endif
	MODULE_PARM_DESC(spl_taskq_kick,
	"Write nonzero to kick stuck taskqs to spawn more threads");

	#ifdef HAVE_CPU_HOTPLUG
	/*
	* This callback will be called exactly once for each core that comes online,
	* for each dynamic taskq. We attempt to expand taskqs that have
	* TASKQ_THREADS_CPU_PCT set. We need to redo the percentage calculation every
	* time, to correctly determine whether or not to add a thread.
	*/
	static int
	spl_taskq_expand(unsigned int cpu, struct hlist_node *node)
	{
	taskq_t *tq = list_entry(node, taskq_t, tq_hp_cb_node);
	unsigned long flags;
	int err = 0;

	ASSERT(tq);
	spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);

	if (!(tq->tq_flags & TASKQ_ACTIVE)) {
	spin_unlock_irqrestore(&tq->tq_lock, flags);
	return (err);
	}

	ASSERT(tq->tq_flags & TASKQ_THREADS_CPU_PCT);
	int nthreads = MIN(tq->tq_cpu_pct, 100);
	nthreads = MAX(((num_online_cpus() + 1) * nthreads) / 100, 1);
	tq->tq_maxthreads = nthreads;

	if (!((tq->tq_flags & TASKQ_DYNAMIC) && spl_taskq_thread_dynamic) &&
	tq->tq_maxthreads > tq->tq_nthreads) {
	spin_unlock_irqrestore(&tq->tq_lock, flags);
	taskq_thread_t *tqt = taskq_thread_create(tq);
	if (tqt == NULL)
	err = -1;
	return (err);
	}
	spin_unlock_irqrestore(&tq->tq_lock, flags);
	return (err);
	}

	/*
	* While we don't support offlining CPUs, it is possible that CPUs will fail
	* to online successfully. We do need to be able to handle this case
	* gracefully.
	*/
	static int
	spl_taskq_prepare_down(unsigned int cpu, struct hlist_node *node)
	{
	taskq_t *tq = list_entry(node, taskq_t, tq_hp_cb_node);
	unsigned long flags;

	ASSERT(tq);
	spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);

	if (!(tq->tq_flags & TASKQ_ACTIVE))
	goto out;

	ASSERT(tq->tq_flags & TASKQ_THREADS_CPU_PCT);
	int nthreads = MIN(tq->tq_cpu_pct, 100);
	nthreads = MAX(((num_online_cpus()) * nthreads) / 100, 1);
	tq->tq_maxthreads = nthreads;

	if (!((tq->tq_flags & TASKQ_DYNAMIC) && spl_taskq_thread_dynamic) &&
	tq->tq_maxthreads < tq->tq_nthreads) {
	ASSERT3U(tq->tq_maxthreads, ==, tq->tq_nthreads - 1);
	taskq_thread_t *tqt = list_entry(tq->tq_thread_list.next,
	taskq_thread_t, tqt_thread_list);
	struct task_struct *thread = tqt->tqt_thread;
	spin_unlock_irqrestore(&tq->tq_lock, flags);

	kthread_stop(thread);

	return (0);
	}

	out:
	spin_unlock_irqrestore(&tq->tq_lock, flags);
	return (0);
	}
	#endif

	int
	spl_taskq_init(void)
	{
	init_rwsem(&tq_list_sem);
	tsd_create(&taskq_tsd, NULL);

	#ifdef HAVE_CPU_HOTPLUG
	spl_taskq_cpuhp_state = cpuhp_setup_state_multi(CPUHP_AP_ONLINE_DYN,
	"fs/spl_taskq:online", spl_taskq_expand, spl_taskq_prepare_down);
	#endif

	system_taskq = taskq_create("spl_system_taskq", MAX(boot_ncpus, 64),
	maxclsyspri, boot_ncpus, INT_MAX, TASKQ_PREPOPULATE\|TASKQ_DYNAMIC);
	if (system_taskq == NULL)
	return (-ENOMEM);

	system_delay_taskq = taskq_create("spl_delay_taskq", MAX(boot_ncpus, 4),
	maxclsyspri, boot_ncpus, INT_MAX, TASKQ_PREPOPULATE\|TASKQ_DYNAMIC);
	if (system_delay_taskq == NULL) {
	#ifdef HAVE_CPU_HOTPLUG
	cpuhp_remove_multi_state(spl_taskq_cpuhp_state);
	#endif
	taskq_destroy(system_taskq);
	return (-ENOMEM);
	}

	dynamic_taskq = taskq_create("spl_dynamic_taskq", 1,
	maxclsyspri, boot_ncpus, INT_MAX, TASKQ_PREPOPULATE);
	if (dynamic_taskq == NULL) {
	#ifdef HAVE_CPU_HOTPLUG
	cpuhp_remove_multi_state(spl_taskq_cpuhp_state);
	#endif
	taskq_destroy(system_taskq);
	taskq_destroy(system_delay_taskq);
	return (-ENOMEM);
	}

	/*
	* This is used to annotate tq_lock, so
	* taskq_dispatch -> taskq_thread_spawn -> taskq_dispatch
	* does not trigger a lockdep warning re: possible recursive locking
	*/
	dynamic_taskq->tq_lock_class = TQ_LOCK_DYNAMIC;

	return (0);
	}

	void
	spl_taskq_fini(void)
	{
	taskq_destroy(dynamic_taskq);
	dynamic_taskq = NULL;

	taskq_destroy(system_delay_taskq);
	system_delay_taskq = NULL;

	taskq_destroy(system_taskq);
	system_taskq = NULL;

	tsd_destroy(&taskq_tsd);

	#ifdef HAVE_CPU_HOTPLUG
	cpuhp_remove_multi_state(spl_taskq_cpuhp_state);
	spl_taskq_cpuhp_state = 0;
	#endif
	}
	diff --git a/sys/contrib/openzfs/module/os/linux/spl/spl-thread.c b/sys/contrib/openzfs/module/os/linux/spl/spl-thread.c
	index ee3eb4690c3a..dbb8eefa7ec4 100644
	--- a/sys/contrib/openzfs/module/os/linux/spl/spl-thread.c
	+++ b/sys/contrib/openzfs/module/os/linux/spl/spl-thread.c
	@@ -1,207 +1,197 @@
	/*
	* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
	* Copyright (C) 2007 The Regents of the University of California.
	* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
	* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
	* UCRL-CODE-235197
	*
	* This file is part of the SPL, Solaris Porting Layer.
	*
	* The SPL is free software; you can redistribute it and/or modify it
	* under the terms of the GNU General Public License as published by the
	* Free Software Foundation; either version 2 of the License, or (at your
	* option) any later version.
	*
	* The SPL is distributed in the hope that it will be useful, but WITHOUT
	* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
	* for more details.
	*
	* You should have received a copy of the GNU General Public License along
	* with the SPL. If not, see <http://www.gnu.org/licenses/>.
	*
	* Solaris Porting Layer (SPL) Thread Implementation.
	*/

	#include <sys/thread.h>
	#include <sys/kmem.h>
	#include <sys/tsd.h>
	#include <sys/string.h>

	/*
	* Thread interfaces
	*/
	typedef struct thread_priv_s {
	unsigned long tp_magic; /* Magic */
	int tp_name_size; /* Name size */
	char tp_name; / Name (without _thread suffix) */
	void (tp_func)(void ); /* Registered function */
	void tp_args; / Args to be passed to function */
	size_t tp_len; /* Len to be passed to function */
	int tp_state; /* State to start thread at */
	pri_t tp_pri; /* Priority to start threat at */
	} thread_priv_t;

	static int
	thread_generic_wrapper(void *arg)
	{
	thread_priv_t tp = (thread_priv_t )arg;
	void (func)(void );
	void *args;

	ASSERT(tp->tp_magic == TP_MAGIC);
	func = tp->tp_func;
	args = tp->tp_args;
	set_current_state(tp->tp_state);
	set_user_nice((kthread_t *)current, PRIO_TO_NICE(tp->tp_pri));
	kmem_free(tp->tp_name, tp->tp_name_size);
	kmem_free(tp, sizeof (thread_priv_t));

	if (func)
	func(args);

	return (0);
	}

	/*
	* thread_create() may block forever if it cannot create a thread or
	* allocate memory. This is preferable to returning a NULL which Solaris
	* style callers likely never check for... since it can't fail.
	*/
	kthread_t *
	__thread_create(caddr_t stk, size_t stksize, thread_func_t func,
	const char name, void args, size_t len, proc_t *pp, int state, pri_t pri)
	{
	thread_priv_t *tp;
	struct task_struct *tsk;
	char *p;

	/* Option pp is simply ignored */
	/* Variable stack size unsupported */
	ASSERT(stk == NULL);

	tp = kmem_alloc(sizeof (thread_priv_t), KM_PUSHPAGE);
	if (tp == NULL)
	return (NULL);

	tp->tp_magic = TP_MAGIC;
	tp->tp_name_size = strlen(name) + 1;

	tp->tp_name = kmem_alloc(tp->tp_name_size, KM_PUSHPAGE);
	if (tp->tp_name == NULL) {
	kmem_free(tp, sizeof (thread_priv_t));
	return (NULL);
	}

	strlcpy(tp->tp_name, name, tp->tp_name_size);

	/*
	* Strip trailing "_thread" from passed name which will be the func
	* name since the exposed API has no parameter for passing a name.
	*/
	p = strstr(tp->tp_name, "_thread");
	if (p)
	p[0] = '\0';

	tp->tp_func = func;
	tp->tp_args = args;
	tp->tp_len = len;
	tp->tp_state = state;
	tp->tp_pri = pri;

	tsk = spl_kthread_create(thread_generic_wrapper, (void *)tp,
	"%s", tp->tp_name);
	if (IS_ERR(tsk))
	return (NULL);

	wake_up_process(tsk);
	return ((kthread_t *)tsk);
	}
	EXPORT_SYMBOL(__thread_create);

	/*
	* spl_kthread_create - Wrapper providing pre-3.13 semantics for
	* kthread_create() in which it is not killable and less likely
	* to return -ENOMEM.
	*/
	struct task_struct *
	spl_kthread_create(int (func)(void ), void *data, const char namefmt[], ...)
	{
	struct task_struct *tsk;
	va_list args;
	char name[TASK_COMM_LEN];

	va_start(args, namefmt);
	vsnprintf(name, sizeof (name), namefmt, args);
	va_end(args);
	do {
	tsk = kthread_create(func, data, "%s", name);
	if (IS_ERR(tsk)) {
	if (signal_pending(current)) {
	clear_thread_flag(TIF_SIGPENDING);
	continue;
	}
	if (PTR_ERR(tsk) == -ENOMEM)
	continue;
	return (NULL);
	} else {
	return (tsk);
	}
	} while (1);
	}
	EXPORT_SYMBOL(spl_kthread_create);

	/*
	- * The "why" argument indicates the allowable side-effects of the call:
	- *
	- * FORREAL: Extract the next pending signal from p_sig into p_cursig;
	- * stop the process if a stop has been requested or if a traced signal
	- * is pending.
	- *
	- * JUSTLOOKING: Don't stop the process, just indicate whether or not
	- * a signal might be pending (FORREAL is needed to tell for sure).
	+ * Extract the next pending signal from p_sig into p_cursig; stop the process
	+ * if a stop has been requested or if a traced signal is pending.
	*/
	int
	-issig(int why)
	+issig(void)
	{
	- ASSERT(why == FORREAL \|\| why == JUSTLOOKING);

	if (!signal_pending(current))
	return (0);

	- if (why != FORREAL)
	- return (1);
	-
	struct task_struct *task = current;
	spl_kernel_siginfo_t __info;
	sigset_t set;
	siginitsetinv(&set, 1ULL << (SIGSTOP - 1) \| 1ULL << (SIGTSTP - 1));
	sigorsets(&set, &task->blocked, &set);

	spin_lock_irq(&task->sighand->siglock);
	#ifdef HAVE_DEQUEUE_SIGNAL_4ARG
	enum pid_type __type;
	if (dequeue_signal(task, &set, &__info, &__type) != 0) {
	#else
	if (dequeue_signal(task, &set, &__info) != 0) {
	#endif
	#ifdef HAVE_SIGNAL_STOP
	spin_unlock_irq(&task->sighand->siglock);
	kernel_signal_stop();
	#else
	if (current->jobctl & JOBCTL_STOP_DEQUEUED)
	spl_set_special_state(TASK_STOPPED);

	spin_unlock_irq(&current->sighand->siglock);

	schedule();
	#endif
	return (0);
	}

	spin_unlock_irq(&task->sighand->siglock);

	return (1);
	}

	EXPORT_SYMBOL(issig);
	diff --git a/sys/contrib/openzfs/module/os/linux/zfs/vdev_disk.c b/sys/contrib/openzfs/module/os/linux/zfs/vdev_disk.c
	index 463c5f705102..7284b922b3bf 100644
	--- a/sys/contrib/openzfs/module/os/linux/zfs/vdev_disk.c
	+++ b/sys/contrib/openzfs/module/os/linux/zfs/vdev_disk.c
	@@ -1,1661 +1,1661 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (C) 2008-2010 Lawrence Livermore National Security, LLC.
	* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
	* Rewritten for Linux by Brian Behlendorf <behlendorf1@llnl.gov>.
	* LLNL-CODE-403049.
	* Copyright (c) 2012, 2019 by Delphix. All rights reserved.
	* Copyright (c) 2023, 2024, Klara Inc.
	*/

	#include <sys/zfs_context.h>
	#include <sys/spa_impl.h>
	#include <sys/vdev_disk.h>
	#include <sys/vdev_impl.h>
	#include <sys/vdev_trim.h>
	#include <sys/abd.h>
	#include <sys/fs/zfs.h>
	#include <sys/zio.h>
	#include <linux/blkpg.h>
	#include <linux/msdos_fs.h>
	#include <linux/vfs_compat.h>
	#ifdef HAVE_LINUX_BLK_CGROUP_HEADER
	#include <linux/blk-cgroup.h>
	#endif

	/*
	* Linux 6.8.x uses a bdev_handle as an instance/refcount for an underlying
	* block_device. Since it carries the block_device inside, its convenient to
	* just use the handle as a proxy.
	*
	* Linux 6.9.x uses a file for the same purpose.
	*
	* For pre-6.8, we just emulate this with a cast, since we don't need any of
	* the other fields inside the handle.
	*/
	#if defined(HAVE_BDEV_OPEN_BY_PATH)
	typedef struct bdev_handle zfs_bdev_handle_t;
	#define BDH_BDEV(bdh) ((bdh)->bdev)
	#define BDH_IS_ERR(bdh) (IS_ERR(bdh))
	#define BDH_PTR_ERR(bdh) (PTR_ERR(bdh))
	#define BDH_ERR_PTR(err) (ERR_PTR(err))
	#elif defined(HAVE_BDEV_FILE_OPEN_BY_PATH)
	typedef struct file zfs_bdev_handle_t;
	#define BDH_BDEV(bdh) (file_bdev(bdh))
	#define BDH_IS_ERR(bdh) (IS_ERR(bdh))
	#define BDH_PTR_ERR(bdh) (PTR_ERR(bdh))
	#define BDH_ERR_PTR(err) (ERR_PTR(err))
	#else
	typedef void zfs_bdev_handle_t;
	#define BDH_BDEV(bdh) ((struct block_device *)bdh)
	#define BDH_IS_ERR(bdh) (IS_ERR(BDH_BDEV(bdh)))
	#define BDH_PTR_ERR(bdh) (PTR_ERR(BDH_BDEV(bdh)))
	#define BDH_ERR_PTR(err) (ERR_PTR(err))
	#endif

	typedef struct vdev_disk {
	zfs_bdev_handle_t *vd_bdh;
	krwlock_t vd_lock;
	} vdev_disk_t;

	/*
	* Maximum number of segments to add to a bio (min 4). If this is higher than
	* the maximum allowed by the device queue or the kernel itself, it will be
	* clamped. Setting it to zero will cause the kernel's ideal size to be used.
	*/
	uint_t zfs_vdev_disk_max_segs = 0;

	/*
	* Unique identifier for the exclusive vdev holder.
	*/
	static void *zfs_vdev_holder = VDEV_HOLDER;

	/*
	* Wait up to zfs_vdev_open_timeout_ms milliseconds before determining the
	* device is missing. The missing path may be transient since the links
	* can be briefly removed and recreated in response to udev events.
	*/
	static uint_t zfs_vdev_open_timeout_ms = 1000;

	/*
	* Size of the "reserved" partition, in blocks.
	*/
	#define EFI_MIN_RESV_SIZE (16 * 1024)

	/*
	* BIO request failfast mask.
	*/

	static unsigned int zfs_vdev_failfast_mask = 1;

	/*
	* Convert SPA mode flags into bdev open mode flags.
	*/
	#ifdef HAVE_BLK_MODE_T
	typedef blk_mode_t vdev_bdev_mode_t;
	#define VDEV_BDEV_MODE_READ BLK_OPEN_READ
	#define VDEV_BDEV_MODE_WRITE BLK_OPEN_WRITE
	#define VDEV_BDEV_MODE_EXCL BLK_OPEN_EXCL
	#define VDEV_BDEV_MODE_MASK (BLK_OPEN_READ\|BLK_OPEN_WRITE\|BLK_OPEN_EXCL)
	#else
	typedef fmode_t vdev_bdev_mode_t;
	#define VDEV_BDEV_MODE_READ FMODE_READ
	#define VDEV_BDEV_MODE_WRITE FMODE_WRITE
	#define VDEV_BDEV_MODE_EXCL FMODE_EXCL
	#define VDEV_BDEV_MODE_MASK (FMODE_READ\|FMODE_WRITE\|FMODE_EXCL)
	#endif

	static vdev_bdev_mode_t
	vdev_bdev_mode(spa_mode_t smode)
	{
	ASSERT3U(smode, !=, SPA_MODE_UNINIT);
	ASSERT0(smode & ~(SPA_MODE_READ\|SPA_MODE_WRITE));

	vdev_bdev_mode_t bmode = VDEV_BDEV_MODE_EXCL;

	if (smode & SPA_MODE_READ)
	bmode \|= VDEV_BDEV_MODE_READ;

	if (smode & SPA_MODE_WRITE)
	bmode \|= VDEV_BDEV_MODE_WRITE;

	ASSERT(bmode & VDEV_BDEV_MODE_MASK);
	ASSERT0(bmode & ~VDEV_BDEV_MODE_MASK);

	return (bmode);
	}

	/*
	* Returns the usable capacity (in bytes) for the partition or disk.
	*/
	static uint64_t
	bdev_capacity(struct block_device *bdev)
	{
	return (i_size_read(bdev->bd_inode));
	}

	#if !defined(HAVE_BDEV_WHOLE)
	static inline struct block_device *
	bdev_whole(struct block_device *bdev)
	{
	return (bdev->bd_contains);
	}
	#endif

	#if defined(HAVE_BDEVNAME)
	#define vdev_bdevname(bdev, name) bdevname(bdev, name)
	#else
	static inline void
	vdev_bdevname(struct block_device bdev, char name)
	{
	snprintf(name, BDEVNAME_SIZE, "%pg", bdev);
	}
	#endif

	/*
	* Returns the maximum expansion capacity of the block device (in bytes).
	*
	* It is possible to expand a vdev when it has been created as a wholedisk
	* and the containing block device has increased in capacity. Or when the
	* partition containing the pool has been manually increased in size.
	*
	* This function is only responsible for calculating the potential expansion
	* size so it can be reported by 'zpool list'. The efi_use_whole_disk() is
	* responsible for verifying the expected partition layout in the wholedisk
	* case, and updating the partition table if appropriate. Once the partition
	* size has been increased the additional capacity will be visible using
	* bdev_capacity().
	*
	* The returned maximum expansion capacity is always expected to be larger, or
	* at the very least equal, to its usable capacity to prevent overestimating
	* the pool expandsize.
	*/
	static uint64_t
	bdev_max_capacity(struct block_device *bdev, uint64_t wholedisk)
	{
	uint64_t psize;
	int64_t available;

	if (wholedisk && bdev != bdev_whole(bdev)) {
	/*
	* When reporting maximum expansion capacity for a wholedisk
	* deduct any capacity which is expected to be lost due to
	* alignment restrictions. Over reporting this value isn't
	* harmful and would only result in slightly less capacity
	* than expected post expansion.
	* The estimated available space may be slightly smaller than
	* bdev_capacity() for devices where the number of sectors is
	* not a multiple of the alignment size and the partition layout
	* is keeping less than PARTITION_END_ALIGNMENT bytes after the
	* "reserved" EFI partition: in such cases return the device
	* usable capacity.
	*/
	available = i_size_read(bdev_whole(bdev)->bd_inode) -
	((EFI_MIN_RESV_SIZE + NEW_START_BLOCK +
	PARTITION_END_ALIGNMENT) << SECTOR_BITS);
	psize = MAX(available, bdev_capacity(bdev));
	} else {
	psize = bdev_capacity(bdev);
	}

	return (psize);
	}

	static void
	vdev_disk_error(zio_t *zio)
	{
	/*
	* This function can be called in interrupt context, for instance while
	* handling IRQs coming from a misbehaving disk device; use printk()
	* which is safe from any context.
	*/
	printk(KERN_WARNING "zio pool=%s vdev=%s error=%d type=%d "
	"offset=%llu size=%llu flags=%llu\n", spa_name(zio->io_spa),
	zio->io_vd->vdev_path, zio->io_error, zio->io_type,
	(u_longlong_t)zio->io_offset, (u_longlong_t)zio->io_size,
	zio->io_flags);
	}

	static void
	vdev_disk_kobj_evt_post(vdev_t *v)
	{
	vdev_disk_t *vd = v->vdev_tsd;
	if (vd && vd->vd_bdh) {
	spl_signal_kobj_evt(BDH_BDEV(vd->vd_bdh));
	} else {
	vdev_dbgmsg(v, "vdev_disk_t is NULL for VDEV:%s\n",
	v->vdev_path);
	}
	}

	static zfs_bdev_handle_t *
	vdev_blkdev_get_by_path(const char path, spa_mode_t smode, void holder)
	{
	vdev_bdev_mode_t bmode = vdev_bdev_mode(smode);

	#if defined(HAVE_BDEV_FILE_OPEN_BY_PATH)
	return (bdev_file_open_by_path(path, bmode, holder, NULL));
	#elif defined(HAVE_BDEV_OPEN_BY_PATH)
	return (bdev_open_by_path(path, bmode, holder, NULL));
	#elif defined(HAVE_BLKDEV_GET_BY_PATH_4ARG)
	return (blkdev_get_by_path(path, bmode, holder, NULL));
	#else
	return (blkdev_get_by_path(path, bmode, holder));
	#endif
	}

	static void
	vdev_blkdev_put(zfs_bdev_handle_t bdh, spa_mode_t smode, void holder)
	{
	#if defined(HAVE_BDEV_RELEASE)
	return (bdev_release(bdh));
	#elif defined(HAVE_BLKDEV_PUT_HOLDER)
	return (blkdev_put(BDH_BDEV(bdh), holder));
	#elif defined(HAVE_BLKDEV_PUT)
	return (blkdev_put(BDH_BDEV(bdh), vdev_bdev_mode(smode)));
	#else
	fput(bdh);
	#endif
	}

	static int
	vdev_disk_open(vdev_t v, uint64_t psize, uint64_t *max_psize,
	uint64_t logical_ashift, uint64_t physical_ashift)
	{
	zfs_bdev_handle_t *bdh;
	spa_mode_t smode = spa_mode(v->vdev_spa);
	hrtime_t timeout = MSEC2NSEC(zfs_vdev_open_timeout_ms);
	vdev_disk_t *vd;

	/* Must have a pathname and it must be absolute. */
	if (v->vdev_path == NULL \|\| v->vdev_path[0] != '/') {
	v->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
	vdev_dbgmsg(v, "invalid vdev_path");
	return (SET_ERROR(EINVAL));
	}

	/*
	* Reopen the device if it is currently open. When expanding a
	* partition force re-scanning the partition table if userland
	* did not take care of this already. We need to do this while closed
	* in order to get an accurate updated block device size. Then
	* since udev may need to recreate the device links increase the
	* open retry timeout before reporting the device as unavailable.
	*/
	vd = v->vdev_tsd;
	if (vd) {
	char disk_name[BDEVNAME_SIZE + 6] = "/dev/";
	boolean_t reread_part = B_FALSE;

	rw_enter(&vd->vd_lock, RW_WRITER);
	bdh = vd->vd_bdh;
	vd->vd_bdh = NULL;

	if (bdh) {
	struct block_device *bdev = BDH_BDEV(bdh);
	if (v->vdev_expanding && bdev != bdev_whole(bdev)) {
	vdev_bdevname(bdev_whole(bdev), disk_name + 5);
	/*
	* If userland has BLKPG_RESIZE_PARTITION,
	* then it should have updated the partition
	* table already. We can detect this by
	* comparing our current physical size
	* with that of the device. If they are
	* the same, then we must not have
	* BLKPG_RESIZE_PARTITION or it failed to
	* update the partition table online. We
	* fallback to rescanning the partition
	* table from the kernel below. However,
	* if the capacity already reflects the
	* updated partition, then we skip
	* rescanning the partition table here.
	*/
	if (v->vdev_psize == bdev_capacity(bdev))
	reread_part = B_TRUE;
	}

	vdev_blkdev_put(bdh, smode, zfs_vdev_holder);
	}

	if (reread_part) {
	bdh = vdev_blkdev_get_by_path(disk_name, smode,
	zfs_vdev_holder);
	if (!BDH_IS_ERR(bdh)) {
	int error =
	vdev_bdev_reread_part(BDH_BDEV(bdh));
	vdev_blkdev_put(bdh, smode, zfs_vdev_holder);
	if (error == 0) {
	timeout = MSEC2NSEC(
	zfs_vdev_open_timeout_ms * 2);
	}
	}
	}
	} else {
	vd = kmem_zalloc(sizeof (vdev_disk_t), KM_SLEEP);

	rw_init(&vd->vd_lock, NULL, RW_DEFAULT, NULL);
	rw_enter(&vd->vd_lock, RW_WRITER);
	}

	/*
	* Devices are always opened by the path provided at configuration
	* time. This means that if the provided path is a udev by-id path
	* then drives may be re-cabled without an issue. If the provided
	* path is a udev by-path path, then the physical location information
	* will be preserved. This can be critical for more complicated
	* configurations where drives are located in specific physical
	* locations to maximize the systems tolerance to component failure.
	*
	* Alternatively, you can provide your own udev rule to flexibly map
	* the drives as you see fit. It is not advised that you use the
	* /dev/[hd]d devices which may be reordered due to probing order.
	* Devices in the wrong locations will be detected by the higher
	* level vdev validation.
	*
	* The specified paths may be briefly removed and recreated in
	* response to udev events. This should be exceptionally unlikely
	* because the zpool command makes every effort to verify these paths
	* have already settled prior to reaching this point. Therefore,
	* a ENOENT failure at this point is highly likely to be transient
	* and it is reasonable to sleep and retry before giving up. In
	* practice delays have been observed to be on the order of 100ms.
	*
	* When ERESTARTSYS is returned it indicates the block device is
	* a zvol which could not be opened due to the deadlock detection
	* logic in zvol_open(). Extend the timeout and retry the open
	* subsequent attempts are expected to eventually succeed.
	*/
	hrtime_t start = gethrtime();
	bdh = BDH_ERR_PTR(-ENXIO);
	while (BDH_IS_ERR(bdh) && ((gethrtime() - start) < timeout)) {
	bdh = vdev_blkdev_get_by_path(v->vdev_path, smode,
	zfs_vdev_holder);
	if (unlikely(BDH_PTR_ERR(bdh) == -ENOENT)) {
	/*
	* There is no point of waiting since device is removed
	* explicitly
	*/
	if (v->vdev_removed)
	break;

	- schedule_timeout(MSEC_TO_TICK(10));
	+ schedule_timeout_interruptible(MSEC_TO_TICK(10));
	} else if (unlikely(BDH_PTR_ERR(bdh) == -ERESTARTSYS)) {
	timeout = MSEC2NSEC(zfs_vdev_open_timeout_ms * 10);
	continue;
	} else if (BDH_IS_ERR(bdh)) {
	break;
	}
	}

	if (BDH_IS_ERR(bdh)) {
	int error = -BDH_PTR_ERR(bdh);
	vdev_dbgmsg(v, "open error=%d timeout=%llu/%llu", error,
	(u_longlong_t)(gethrtime() - start),
	(u_longlong_t)timeout);
	vd->vd_bdh = NULL;
	v->vdev_tsd = vd;
	rw_exit(&vd->vd_lock);
	return (SET_ERROR(error));
	} else {
	vd->vd_bdh = bdh;
	v->vdev_tsd = vd;
	rw_exit(&vd->vd_lock);
	}

	struct block_device *bdev = BDH_BDEV(vd->vd_bdh);

	/* Determine the physical block size */
	int physical_block_size = bdev_physical_block_size(bdev);

	/* Determine the logical block size */
	int logical_block_size = bdev_logical_block_size(bdev);

	/*
	* If the device has a write cache, clear the nowritecache flag,
	* so that we start issuing flush requests again.
	*/
	v->vdev_nowritecache = !zfs_bdev_has_write_cache(bdev);

	/* Set when device reports it supports TRIM. */
	v->vdev_has_trim = bdev_discard_supported(bdev);

	/* Set when device reports it supports secure TRIM. */
	v->vdev_has_securetrim = bdev_secure_discard_supported(bdev);

	/* Inform the ZIO pipeline that we are non-rotational */
	v->vdev_nonrot = blk_queue_nonrot(bdev_get_queue(bdev));

	/* Physical volume size in bytes for the partition */
	*psize = bdev_capacity(bdev);

	/* Physical volume size in bytes including possible expansion space */
	*max_psize = bdev_max_capacity(bdev, v->vdev_wholedisk);

	/* Based on the minimum sector size set the block size */
	*physical_ashift = highbit64(MAX(physical_block_size,
	SPA_MINBLOCKSIZE)) - 1;

	*logical_ashift = highbit64(MAX(logical_block_size,
	SPA_MINBLOCKSIZE)) - 1;

	return (0);
	}

	static void
	vdev_disk_close(vdev_t *v)
	{
	vdev_disk_t *vd = v->vdev_tsd;

	if (v->vdev_reopening \|\| vd == NULL)
	return;

	if (vd->vd_bdh != NULL)
	vdev_blkdev_put(vd->vd_bdh, spa_mode(v->vdev_spa),
	zfs_vdev_holder);

	rw_destroy(&vd->vd_lock);
	kmem_free(vd, sizeof (vdev_disk_t));
	v->vdev_tsd = NULL;
	}

	static inline void
	vdev_submit_bio_impl(struct bio *bio)
	{
	#ifdef HAVE_1ARG_SUBMIT_BIO
	(void) submit_bio(bio);
	#else
	(void) submit_bio(bio_data_dir(bio), bio);
	#endif
	}

	/*
	* preempt_schedule_notrace is GPL-only which breaks the ZFS build, so
	* replace it with preempt_schedule under the following condition:
	*/
	#if defined(CONFIG_ARM64) && \
	defined(CONFIG_PREEMPTION) && \
	defined(CONFIG_BLK_CGROUP)
	#define preempt_schedule_notrace(x) preempt_schedule(x)
	#endif

	/*
	* As for the Linux 5.18 kernel bio_alloc() expects a block_device struct
	* as an argument removing the need to set it with bio_set_dev(). This
	* removes the need for all of the following compatibility code.
	*/
	#if !defined(HAVE_BIO_ALLOC_4ARG)

	#ifdef HAVE_BIO_SET_DEV
	#if defined(CONFIG_BLK_CGROUP) && defined(HAVE_BIO_SET_DEV_GPL_ONLY)
	/*
	* The Linux 5.5 kernel updated percpu_ref_tryget() which is inlined by
	* blkg_tryget() to use rcu_read_lock() instead of rcu_read_lock_sched().
	* As a side effect the function was converted to GPL-only. Define our
	* own version when needed which uses rcu_read_lock_sched().
	*
	* The Linux 5.17 kernel split linux/blk-cgroup.h into a private and a public
	* part, moving blkg_tryget into the private one. Define our own version.
	*/
	#if defined(HAVE_BLKG_TRYGET_GPL_ONLY) \|\| !defined(HAVE_BLKG_TRYGET)
	static inline bool
	vdev_blkg_tryget(struct blkcg_gq *blkg)
	{
	struct percpu_ref *ref = &blkg->refcnt;
	unsigned long __percpu *count;
	bool rc;

	rcu_read_lock_sched();

	if (__ref_is_percpu(ref, &count)) {
	this_cpu_inc(*count);
	rc = true;
	} else {
	#ifdef ZFS_PERCPU_REF_COUNT_IN_DATA
	rc = atomic_long_inc_not_zero(&ref->data->count);
	#else
	rc = atomic_long_inc_not_zero(&ref->count);
	#endif
	}

	rcu_read_unlock_sched();

	return (rc);
	}
	#else
	#define vdev_blkg_tryget(bg) blkg_tryget(bg)
	#endif
	#ifdef HAVE_BIO_SET_DEV_MACRO
	/*
	* The Linux 5.0 kernel updated the bio_set_dev() macro so it calls the
	* GPL-only bio_associate_blkg() symbol thus inadvertently converting
	* the entire macro. Provide a minimal version which always assigns the
	* request queue's root_blkg to the bio.
	*/
	static inline void
	vdev_bio_associate_blkg(struct bio *bio)
	{
	#if defined(HAVE_BIO_BDEV_DISK)
	struct request_queue *q = bio->bi_bdev->bd_disk->queue;
	#else
	struct request_queue *q = bio->bi_disk->queue;
	#endif

	ASSERT3P(q, !=, NULL);
	ASSERT3P(bio->bi_blkg, ==, NULL);

	if (q->root_blkg && vdev_blkg_tryget(q->root_blkg))
	bio->bi_blkg = q->root_blkg;
	}

	#define bio_associate_blkg vdev_bio_associate_blkg
	#else
	static inline void
	vdev_bio_set_dev(struct bio bio, struct block_device bdev)
	{
	#if defined(HAVE_BIO_BDEV_DISK)
	struct request_queue *q = bdev->bd_disk->queue;
	#else
	struct request_queue *q = bio->bi_disk->queue;
	#endif
	bio_clear_flag(bio, BIO_REMAPPED);
	if (bio->bi_bdev != bdev)
	bio_clear_flag(bio, BIO_THROTTLED);
	bio->bi_bdev = bdev;

	ASSERT3P(q, !=, NULL);
	ASSERT3P(bio->bi_blkg, ==, NULL);

	if (q->root_blkg && vdev_blkg_tryget(q->root_blkg))
	bio->bi_blkg = q->root_blkg;
	}
	#define bio_set_dev vdev_bio_set_dev
	#endif
	#endif
	#else
	/*
	* Provide a bio_set_dev() helper macro for pre-Linux 4.14 kernels.
	*/
	static inline void
	bio_set_dev(struct bio bio, struct block_device bdev)
	{
	bio->bi_bdev = bdev;
	}
	#endif /* HAVE_BIO_SET_DEV */
	#endif /* !HAVE_BIO_ALLOC_4ARG */

	static inline void
	vdev_submit_bio(struct bio *bio)
	{
	struct bio_list *bio_list = current->bio_list;
	current->bio_list = NULL;
	vdev_submit_bio_impl(bio);
	current->bio_list = bio_list;
	}

	static inline struct bio *
	vdev_bio_alloc(struct block_device *bdev, gfp_t gfp_mask,
	unsigned short nr_vecs)
	{
	struct bio *bio;

	#ifdef HAVE_BIO_ALLOC_4ARG
	bio = bio_alloc(bdev, nr_vecs, 0, gfp_mask);
	#else
	bio = bio_alloc(gfp_mask, nr_vecs);
	if (likely(bio != NULL))
	bio_set_dev(bio, bdev);
	#endif

	return (bio);
	}

	static inline uint_t
	vdev_bio_max_segs(struct block_device *bdev)
	{
	/*
	* Smallest of the device max segs and the tuneable max segs. Minimum
	* 4, so there's room to finish split pages if they come up.
	*/
	const uint_t dev_max_segs = queue_max_segments(bdev_get_queue(bdev));
	const uint_t tune_max_segs = (zfs_vdev_disk_max_segs > 0) ?
	MAX(4, zfs_vdev_disk_max_segs) : dev_max_segs;
	const uint_t max_segs = MIN(tune_max_segs, dev_max_segs);

	#ifdef HAVE_BIO_MAX_SEGS
	return (bio_max_segs(max_segs));
	#else
	return (MIN(max_segs, BIO_MAX_PAGES));
	#endif
	}

	static inline uint_t
	vdev_bio_max_bytes(struct block_device *bdev)
	{
	return (queue_max_sectors(bdev_get_queue(bdev)) << 9);
	}


	/*
	* Virtual block IO object (VBIO)
	*
	* Linux block IO (BIO) objects have a limit on how many data segments (pages)
	* they can hold. Depending on how they're allocated and structured, a large
	* ZIO can require more than one BIO to be submitted to the kernel, which then
	* all have to complete before we can return the completed ZIO back to ZFS.
	*
	* A VBIO is a wrapper around multiple BIOs, carrying everything needed to
	* translate a ZIO down into the kernel block layer and back again.
	*
	* Note that these are only used for data ZIOs (read/write). Meta-operations
	* (flush/trim) don't need multiple BIOs and so can just make the call
	* directly.
	*/
	typedef struct {
	zio_t vbio_zio; / parent zio */

	struct block_device vbio_bdev; / blockdev to submit bios to */

	abd_t vbio_abd; / abd carrying borrowed linear buf */

	uint_t vbio_max_segs; /* max segs per bio */

	uint_t vbio_max_bytes; /* max bytes per bio */
	uint_t vbio_lbs_mask; /* logical block size mask */

	uint64_t vbio_offset; /* start offset of next bio */

	struct bio vbio_bio; / pointer to the current bio */
	int vbio_flags; /* bio flags */
	} vbio_t;

	static vbio_t *
	vbio_alloc(zio_t zio, struct block_device bdev, int flags)
	{
	vbio_t *vbio = kmem_zalloc(sizeof (vbio_t), KM_SLEEP);

	vbio->vbio_zio = zio;
	vbio->vbio_bdev = bdev;
	vbio->vbio_abd = NULL;
	vbio->vbio_max_segs = vdev_bio_max_segs(bdev);
	vbio->vbio_max_bytes = vdev_bio_max_bytes(bdev);
	vbio->vbio_lbs_mask = ~(bdev_logical_block_size(bdev)-1);
	vbio->vbio_offset = zio->io_offset;
	vbio->vbio_bio = NULL;
	vbio->vbio_flags = flags;

	return (vbio);
	}

	BIO_END_IO_PROTO(vbio_completion, bio, error);

	static int
	vbio_add_page(vbio_t vbio, struct page page, uint_t size, uint_t offset)
	{
	struct bio *bio = vbio->vbio_bio;
	uint_t ssize;

	while (size > 0) {
	if (bio == NULL) {
	/* New BIO, allocate and set up */
	bio = vdev_bio_alloc(vbio->vbio_bdev, GFP_NOIO,
	vbio->vbio_max_segs);
	VERIFY(bio);

	BIO_BI_SECTOR(bio) = vbio->vbio_offset >> 9;
	bio_set_op_attrs(bio,
	vbio->vbio_zio->io_type == ZIO_TYPE_WRITE ?
	WRITE : READ, vbio->vbio_flags);

	if (vbio->vbio_bio) {
	bio_chain(vbio->vbio_bio, bio);
	vdev_submit_bio(vbio->vbio_bio);
	}
	vbio->vbio_bio = bio;
	}

	/*
	* Only load as much of the current page data as will fit in
	* the space left in the BIO, respecting lbs alignment. Older
	* kernels will error if we try to overfill the BIO, while
	* newer ones will accept it and split the BIO. This ensures
	* everything works on older kernels, and avoids an additional
	* overhead on the new.
	*/
	ssize = MIN(size, (vbio->vbio_max_bytes - BIO_BI_SIZE(bio)) &
	vbio->vbio_lbs_mask);
	if (ssize > 0 &&
	bio_add_page(bio, page, ssize, offset) == ssize) {
	/* Accepted, adjust and load any remaining. */
	size -= ssize;
	offset += ssize;
	continue;
	}

	/* No room, set up for a new BIO and loop */
	vbio->vbio_offset += BIO_BI_SIZE(bio);

	/* Signal new BIO allocation wanted */
	bio = NULL;
	}

	return (0);
	}

	/* Iterator callback to submit ABD pages to the vbio. */
	static int
	vbio_fill_cb(struct page page, size_t off, size_t len, void priv)
	{
	vbio_t *vbio = priv;
	return (vbio_add_page(vbio, page, len, off));
	}

	/* Create some BIOs, fill them with data and submit them */
	static void
	vbio_submit(vbio_t vbio, abd_t abd, uint64_t size)
	{
	/*
	* We plug so we can submit the BIOs as we go and only unplug them when
	* they are fully created and submitted. This is important; if we don't
	* plug, then the kernel may start executing earlier BIOs while we're
	* still creating and executing later ones, and if the device goes
	* away while that's happening, older kernels can get confused and
	* trample memory.
	*/
	struct blk_plug plug;
	blk_start_plug(&plug);

	(void) abd_iterate_page_func(abd, 0, size, vbio_fill_cb, vbio);
	ASSERT(vbio->vbio_bio);

	vbio->vbio_bio->bi_end_io = vbio_completion;
	vbio->vbio_bio->bi_private = vbio;

	/*
	* Once submitted, vbio_bio now owns vbio (through bi_private) and we
	* can't touch it again. The bio may complete and vbio_completion() be
	* called and free the vbio before this task is run again, so we must
	* consider it invalid from this point.
	*/
	vdev_submit_bio(vbio->vbio_bio);

	blk_finish_plug(&plug);
	}

	/* IO completion callback */
	BIO_END_IO_PROTO(vbio_completion, bio, error)
	{
	vbio_t *vbio = bio->bi_private;
	zio_t *zio = vbio->vbio_zio;

	ASSERT(zio);

	/* Capture and log any errors */
	#ifdef HAVE_1ARG_BIO_END_IO_T
	zio->io_error = BIO_END_IO_ERROR(bio);
	#else
	zio->io_error = 0;
	if (error)
	zio->io_error = -(error);
	else if (!test_bit(BIO_UPTODATE, &bio->bi_flags))
	zio->io_error = EIO;
	#endif
	ASSERT3U(zio->io_error, >=, 0);

	if (zio->io_error)
	vdev_disk_error(zio);

	/* Return the BIO to the kernel */
	bio_put(bio);

	/*
	* If we copied the ABD before issuing it, clean up and return the copy
	* to the ADB, with changes if appropriate.
	*/
	if (vbio->vbio_abd != NULL) {
	void *buf = abd_to_buf(vbio->vbio_abd);
	abd_free(vbio->vbio_abd);
	vbio->vbio_abd = NULL;

	if (zio->io_type == ZIO_TYPE_READ)
	abd_return_buf_copy(zio->io_abd, buf, zio->io_size);
	else
	abd_return_buf(zio->io_abd, buf, zio->io_size);
	}

	/* Final cleanup */
	kmem_free(vbio, sizeof (vbio_t));

	/* All done, submit for processing */
	zio_delay_interrupt(zio);
	}

	/*
	* Iterator callback to count ABD pages and check their size & alignment.
	*
	* On Linux, each BIO segment can take a page pointer, and an offset+length of
	* the data within that page. A page can be arbitrarily large ("compound"
	* pages) but we still have to ensure the data portion is correctly sized and
	* aligned to the logical block size, to ensure that if the kernel wants to
	* split the BIO, the two halves will still be properly aligned.
	*
	* NOTE: if you change this function, change the copy in
	* tests/zfs-tests/tests/functional/vdev_disk/page_alignment.c, and add test
	* data there to validate the change you're making.
	*
	*/
	typedef struct {
	uint_t bmask;
	uint_t npages;
	uint_t end;
	} vdev_disk_check_pages_t;

	static int
	vdev_disk_check_pages_cb(struct page page, size_t off, size_t len, void priv)
	{
	(void) page;
	vdev_disk_check_pages_t *s = priv;

	/*
	* If we didn't finish on a block size boundary last time, then there
	* would be a gap if we tried to use this ABD as-is, so abort.
	*/
	if (s->end != 0)
	return (1);

	/*
	* Note if we're taking less than a full block, so we can check it
	* above on the next call.
	*/
	s->end = (off+len) & s->bmask;

	/* All blocks after the first must start on a block size boundary. */
	if (s->npages != 0 && (off & s->bmask) != 0)
	return (1);

	s->npages++;
	return (0);
	}

	/*
	* Check if we can submit the pages in this ABD to the kernel as-is. Returns
	* the number of pages, or 0 if it can't be submitted like this.
	*/
	static boolean_t
	vdev_disk_check_pages(abd_t abd, uint64_t size, struct block_device bdev)
	{
	vdev_disk_check_pages_t s = {
	.bmask = bdev_logical_block_size(bdev)-1,
	.npages = 0,
	.end = 0,
	};

	if (abd_iterate_page_func(abd, 0, size, vdev_disk_check_pages_cb, &s))
	return (B_FALSE);

	return (B_TRUE);
	}

	static int
	vdev_disk_io_rw(zio_t *zio)
	{
	vdev_t *v = zio->io_vd;
	vdev_disk_t *vd = v->vdev_tsd;
	struct block_device *bdev = BDH_BDEV(vd->vd_bdh);
	int flags = 0;

	/*
	* Accessing outside the block device is never allowed.
	*/
	if (zio->io_offset + zio->io_size > bdev->bd_inode->i_size) {
	vdev_dbgmsg(zio->io_vd,
	"Illegal access %llu size %llu, device size %llu",
	(u_longlong_t)zio->io_offset,
	(u_longlong_t)zio->io_size,
	(u_longlong_t)i_size_read(bdev->bd_inode));
	return (SET_ERROR(EIO));
	}

	if (!(zio->io_flags & (ZIO_FLAG_IO_RETRY \| ZIO_FLAG_TRYHARD)) &&
	v->vdev_failfast == B_TRUE) {
	bio_set_flags_failfast(bdev, &flags, zfs_vdev_failfast_mask & 1,
	zfs_vdev_failfast_mask & 2, zfs_vdev_failfast_mask & 4);
	}

	/*
	* Check alignment of the incoming ABD. If any part of it would require
	* submitting a page that is not aligned to the logical block size,
	* then we take a copy into a linear buffer and submit that instead.
	* This should be impossible on a 512b LBS, and fairly rare on 4K,
	* usually requiring abnormally-small data blocks (eg gang blocks)
	* mixed into the same ABD as larger ones (eg aggregated).
	*/
	abd_t *abd = zio->io_abd;
	if (!vdev_disk_check_pages(abd, zio->io_size, bdev)) {
	void *buf;
	if (zio->io_type == ZIO_TYPE_READ)
	buf = abd_borrow_buf(zio->io_abd, zio->io_size);
	else
	buf = abd_borrow_buf_copy(zio->io_abd, zio->io_size);

	/*
	* Wrap the copy in an abd_t, so we can use the same iterators
	* to count and fill the vbio later.
	*/
	abd = abd_get_from_buf(buf, zio->io_size);

	/*
	* False here would mean the borrowed copy has an invalid
	* alignment too, which would mean we've somehow been passed a
	* linear ABD with an interior page that has a non-zero offset
	* or a size not a multiple of PAGE_SIZE. This is not possible.
	* It would mean either zio_buf_alloc() or its underlying
	* allocators have done something extremely strange, or our
	* math in vdev_disk_check_pages() is wrong. In either case,
	* something in seriously wrong and its not safe to continue.
	*/
	VERIFY(vdev_disk_check_pages(abd, zio->io_size, bdev));
	}

	/* Allocate vbio, with a pointer to the borrowed ABD if necessary */
	vbio_t *vbio = vbio_alloc(zio, bdev, flags);
	if (abd != zio->io_abd)
	vbio->vbio_abd = abd;

	/* Fill it with data pages and submit it to the kernel */
	vbio_submit(vbio, abd, zio->io_size);
	return (0);
	}

	/* ========== */

	/*
	* This is the classic, battle-tested BIO submission code. Until we're totally
	* sure that the new code is safe and correct in all cases, this will remain
	* available and can be enabled by setting zfs_vdev_disk_classic=1 at module
	* load time.
	*
	* These functions have been renamed to vdev_classic_* to make it clear what
	* they belong to, but their implementations are unchanged.
	*/

	/*
	* Virtual device vector for disks.
	*/
	typedef struct dio_request {
	zio_t dr_zio; / Parent ZIO */
	atomic_t dr_ref; /* References */
	int dr_error; /* Bio error */
	int dr_bio_count; /* Count of bio's */
	struct bio dr_bio[]; / Attached bio's */
	} dio_request_t;

	static dio_request_t *
	vdev_classic_dio_alloc(int bio_count)
	{
	dio_request_t *dr = kmem_zalloc(sizeof (dio_request_t) +
	sizeof (struct bio ) bio_count, KM_SLEEP);
	atomic_set(&dr->dr_ref, 0);
	dr->dr_bio_count = bio_count;
	dr->dr_error = 0;

	for (int i = 0; i < dr->dr_bio_count; i++)
	dr->dr_bio[i] = NULL;

	return (dr);
	}

	static void
	vdev_classic_dio_free(dio_request_t *dr)
	{
	int i;

	for (i = 0; i < dr->dr_bio_count; i++)
	if (dr->dr_bio[i])
	bio_put(dr->dr_bio[i]);

	kmem_free(dr, sizeof (dio_request_t) +
	sizeof (struct bio ) dr->dr_bio_count);
	}

	static void
	vdev_classic_dio_get(dio_request_t *dr)
	{
	atomic_inc(&dr->dr_ref);
	}

	static void
	vdev_classic_dio_put(dio_request_t *dr)
	{
	int rc = atomic_dec_return(&dr->dr_ref);

	/*
	* Free the dio_request when the last reference is dropped and
	* ensure zio_interpret is called only once with the correct zio
	*/
	if (rc == 0) {
	zio_t *zio = dr->dr_zio;
	int error = dr->dr_error;

	vdev_classic_dio_free(dr);

	if (zio) {
	zio->io_error = error;
	ASSERT3S(zio->io_error, >=, 0);
	if (zio->io_error)
	vdev_disk_error(zio);

	zio_delay_interrupt(zio);
	}
	}
	}

	BIO_END_IO_PROTO(vdev_classic_physio_completion, bio, error)
	{
	dio_request_t *dr = bio->bi_private;

	if (dr->dr_error == 0) {
	#ifdef HAVE_1ARG_BIO_END_IO_T
	dr->dr_error = BIO_END_IO_ERROR(bio);
	#else
	if (error)
	dr->dr_error = -(error);
	else if (!test_bit(BIO_UPTODATE, &bio->bi_flags))
	dr->dr_error = EIO;
	#endif
	}

	/* Drop reference acquired by vdev_classic_physio */
	vdev_classic_dio_put(dr);
	}

	static inline unsigned int
	vdev_classic_bio_max_segs(zio_t *zio, int bio_size, uint64_t abd_offset)
	{
	unsigned long nr_segs = abd_nr_pages_off(zio->io_abd,
	bio_size, abd_offset);

	#ifdef HAVE_BIO_MAX_SEGS
	return (bio_max_segs(nr_segs));
	#else
	return (MIN(nr_segs, BIO_MAX_PAGES));
	#endif
	}

	static int
	vdev_classic_physio(zio_t *zio)
	{
	vdev_t *v = zio->io_vd;
	vdev_disk_t *vd = v->vdev_tsd;
	struct block_device *bdev = BDH_BDEV(vd->vd_bdh);
	size_t io_size = zio->io_size;
	uint64_t io_offset = zio->io_offset;
	int rw = zio->io_type == ZIO_TYPE_READ ? READ : WRITE;
	int flags = 0;

	dio_request_t *dr;
	uint64_t abd_offset;
	uint64_t bio_offset;
	int bio_size;
	int bio_count = 16;
	int error = 0;
	struct blk_plug plug;
	unsigned short nr_vecs;

	/*
	* Accessing outside the block device is never allowed.
	*/
	if (io_offset + io_size > bdev->bd_inode->i_size) {
	vdev_dbgmsg(zio->io_vd,
	"Illegal access %llu size %llu, device size %llu",
	(u_longlong_t)io_offset,
	(u_longlong_t)io_size,
	(u_longlong_t)i_size_read(bdev->bd_inode));
	return (SET_ERROR(EIO));
	}

	retry:
	dr = vdev_classic_dio_alloc(bio_count);

	if (!(zio->io_flags & (ZIO_FLAG_IO_RETRY \| ZIO_FLAG_TRYHARD)) &&
	zio->io_vd->vdev_failfast == B_TRUE) {
	bio_set_flags_failfast(bdev, &flags, zfs_vdev_failfast_mask & 1,
	zfs_vdev_failfast_mask & 2, zfs_vdev_failfast_mask & 4);
	}

	dr->dr_zio = zio;

	/*
	* Since bio's can have up to BIO_MAX_PAGES=256 iovec's, each of which
	* is at least 512 bytes and at most PAGESIZE (typically 4K), one bio
	* can cover at least 128KB and at most 1MB. When the required number
	* of iovec's exceeds this, we are forced to break the IO in multiple
	* bio's and wait for them all to complete. This is likely if the
	* recordsize property is increased beyond 1MB. The default
	* bio_count=16 should typically accommodate the maximum-size zio of
	* 16MB.
	*/

	abd_offset = 0;
	bio_offset = io_offset;
	bio_size = io_size;
	for (int i = 0; i <= dr->dr_bio_count; i++) {

	/* Finished constructing bio's for given buffer */
	if (bio_size <= 0)
	break;

	/*
	* If additional bio's are required, we have to retry, but
	* this should be rare - see the comment above.
	*/
	if (dr->dr_bio_count == i) {
	vdev_classic_dio_free(dr);
	bio_count *= 2;
	goto retry;
	}

	nr_vecs = vdev_classic_bio_max_segs(zio, bio_size, abd_offset);
	dr->dr_bio[i] = vdev_bio_alloc(bdev, GFP_NOIO, nr_vecs);
	if (unlikely(dr->dr_bio[i] == NULL)) {
	vdev_classic_dio_free(dr);
	return (SET_ERROR(ENOMEM));
	}

	/* Matching put called by vdev_classic_physio_completion */
	vdev_classic_dio_get(dr);

	BIO_BI_SECTOR(dr->dr_bio[i]) = bio_offset >> 9;
	dr->dr_bio[i]->bi_end_io = vdev_classic_physio_completion;
	dr->dr_bio[i]->bi_private = dr;
	bio_set_op_attrs(dr->dr_bio[i], rw, flags);

	/* Remaining size is returned to become the new size */
	bio_size = abd_bio_map_off(dr->dr_bio[i], zio->io_abd,
	bio_size, abd_offset);

	/* Advance in buffer and construct another bio if needed */
	abd_offset += BIO_BI_SIZE(dr->dr_bio[i]);
	bio_offset += BIO_BI_SIZE(dr->dr_bio[i]);
	}

	/* Extra reference to protect dio_request during vdev_submit_bio */
	vdev_classic_dio_get(dr);

	if (dr->dr_bio_count > 1)
	blk_start_plug(&plug);

	/* Submit all bio's associated with this dio */
	for (int i = 0; i < dr->dr_bio_count; i++) {
	if (dr->dr_bio[i])
	vdev_submit_bio(dr->dr_bio[i]);
	}

	if (dr->dr_bio_count > 1)
	blk_finish_plug(&plug);

	vdev_classic_dio_put(dr);

	return (error);
	}

	/* ========== */

	BIO_END_IO_PROTO(vdev_disk_io_flush_completion, bio, error)
	{
	zio_t *zio = bio->bi_private;
	#ifdef HAVE_1ARG_BIO_END_IO_T
	zio->io_error = BIO_END_IO_ERROR(bio);
	#else
	zio->io_error = -error;
	#endif

	if (zio->io_error && (zio->io_error == EOPNOTSUPP))
	zio->io_vd->vdev_nowritecache = B_TRUE;

	bio_put(bio);
	ASSERT3S(zio->io_error, >=, 0);
	if (zio->io_error)
	vdev_disk_error(zio);
	zio_interrupt(zio);
	}

	static int
	vdev_disk_io_flush(struct block_device bdev, zio_t zio)
	{
	struct request_queue *q;
	struct bio *bio;

	q = bdev_get_queue(bdev);
	if (!q)
	return (SET_ERROR(ENXIO));

	bio = vdev_bio_alloc(bdev, GFP_NOIO, 0);
	if (unlikely(bio == NULL))
	return (SET_ERROR(ENOMEM));

	bio->bi_end_io = vdev_disk_io_flush_completion;
	bio->bi_private = zio;
	bio_set_flush(bio);
	vdev_submit_bio(bio);
	invalidate_bdev(bdev);

	return (0);
	}

	BIO_END_IO_PROTO(vdev_disk_discard_end_io, bio, error)
	{
	zio_t *zio = bio->bi_private;
	#ifdef HAVE_1ARG_BIO_END_IO_T
	zio->io_error = BIO_END_IO_ERROR(bio);
	#else
	zio->io_error = -error;
	#endif
	bio_put(bio);
	if (zio->io_error)
	vdev_disk_error(zio);
	zio_interrupt(zio);
	}

	/*
	* Wrappers for the different secure erase and discard APIs. We use async
	* when available; in this case, *biop is set to the last bio in the chain.
	*/
	static int
	vdev_bdev_issue_secure_erase(zfs_bdev_handle_t *bdh, sector_t sector,
	sector_t nsect, struct bio **biop)
	{
	*biop = NULL;
	int error;

	#if defined(HAVE_BLKDEV_ISSUE_SECURE_ERASE)
	error = blkdev_issue_secure_erase(BDH_BDEV(bdh),
	sector, nsect, GFP_NOFS);
	#elif defined(HAVE_BLKDEV_ISSUE_DISCARD_ASYNC_FLAGS)
	error = __blkdev_issue_discard(BDH_BDEV(bdh),
	sector, nsect, GFP_NOFS, BLKDEV_DISCARD_SECURE, biop);
	#elif defined(HAVE_BLKDEV_ISSUE_DISCARD_FLAGS)
	error = blkdev_issue_discard(BDH_BDEV(bdh),
	sector, nsect, GFP_NOFS, BLKDEV_DISCARD_SECURE);
	#else
	#error "unsupported kernel"
	#endif

	return (error);
	}

	static int
	vdev_bdev_issue_discard(zfs_bdev_handle_t *bdh, sector_t sector,
	sector_t nsect, struct bio **biop)
	{
	*biop = NULL;
	int error;

	#if defined(HAVE_BLKDEV_ISSUE_DISCARD_ASYNC_FLAGS)
	error = __blkdev_issue_discard(BDH_BDEV(bdh),
	sector, nsect, GFP_NOFS, 0, biop);
	#elif defined(HAVE_BLKDEV_ISSUE_DISCARD_ASYNC_NOFLAGS)
	error = __blkdev_issue_discard(BDH_BDEV(bdh),
	sector, nsect, GFP_NOFS, biop);
	#elif defined(HAVE_BLKDEV_ISSUE_DISCARD_FLAGS)
	error = blkdev_issue_discard(BDH_BDEV(bdh),
	sector, nsect, GFP_NOFS, 0);
	#elif defined(HAVE_BLKDEV_ISSUE_DISCARD_NOFLAGS)
	error = blkdev_issue_discard(BDH_BDEV(bdh),
	sector, nsect, GFP_NOFS);
	#else
	#error "unsupported kernel"
	#endif

	return (error);
	}

	/*
	* Entry point for TRIM ops. This calls the right wrapper for secure erase or
	* discard, and then does the appropriate finishing work for error vs success
	* and async vs sync.
	*/
	static int
	vdev_disk_io_trim(zio_t *zio)
	{
	int error;
	struct bio *bio;

	zfs_bdev_handle_t bdh = ((vdev_disk_t )zio->io_vd->vdev_tsd)->vd_bdh;
	sector_t sector = zio->io_offset >> 9;
	sector_t nsects = zio->io_size >> 9;

	if (zio->io_trim_flags & ZIO_TRIM_SECURE)
	error = vdev_bdev_issue_secure_erase(bdh, sector, nsects, &bio);
	else
	error = vdev_bdev_issue_discard(bdh, sector, nsects, &bio);

	if (error != 0)
	return (SET_ERROR(-error));

	if (bio == NULL) {
	/*
	* This was a synchronous op that completed successfully, so
	* return it to ZFS immediately.
	*/
	zio_interrupt(zio);
	} else {
	/*
	* This was an asynchronous op; set up completion callback and
	* issue it.
	*/
	bio->bi_private = zio;
	bio->bi_end_io = vdev_disk_discard_end_io;
	vdev_submit_bio(bio);
	}

	return (0);
	}

	int (vdev_disk_io_rw_fn)(zio_t zio) = NULL;

	static void
	vdev_disk_io_start(zio_t *zio)
	{
	vdev_t *v = zio->io_vd;
	vdev_disk_t *vd = v->vdev_tsd;
	int error;

	/*
	* If the vdev is closed, it's likely in the REMOVED or FAULTED state.
	* Nothing to be done here but return failure.
	*/
	if (vd == NULL) {
	zio->io_error = ENXIO;
	zio_interrupt(zio);
	return;
	}

	rw_enter(&vd->vd_lock, RW_READER);

	/*
	* If the vdev is closed, it's likely due to a failed reopen and is
	* in the UNAVAIL state. Nothing to be done here but return failure.
	*/
	if (vd->vd_bdh == NULL) {
	rw_exit(&vd->vd_lock);
	zio->io_error = ENXIO;
	zio_interrupt(zio);
	return;
	}

	switch (zio->io_type) {
	case ZIO_TYPE_FLUSH:

	if (!vdev_readable(v)) {
	/* Drive not there, can't flush */
	error = SET_ERROR(ENXIO);
	} else if (zfs_nocacheflush) {
	/* Flushing disabled by operator, declare success */
	error = 0;
	} else if (v->vdev_nowritecache) {
	/* This vdev not capable of flushing */
	error = SET_ERROR(ENOTSUP);
	} else {
	/*
	* Issue the flush. If successful, the response will
	* be handled in the completion callback, so we're done.
	*/
	error = vdev_disk_io_flush(BDH_BDEV(vd->vd_bdh), zio);
	if (error == 0) {
	rw_exit(&vd->vd_lock);
	return;
	}
	}

	/* Couldn't issue the flush, so set the error and return it */
	rw_exit(&vd->vd_lock);
	zio->io_error = error;
	zio_execute(zio);
	return;

	case ZIO_TYPE_TRIM:
	error = vdev_disk_io_trim(zio);
	rw_exit(&vd->vd_lock);
	if (error) {
	zio->io_error = error;
	zio_execute(zio);
	}
	return;

	case ZIO_TYPE_READ:
	case ZIO_TYPE_WRITE:
	zio->io_target_timestamp = zio_handle_io_delay(zio);
	error = vdev_disk_io_rw_fn(zio);
	rw_exit(&vd->vd_lock);
	if (error) {
	zio->io_error = error;
	zio_interrupt(zio);
	}
	return;

	default:
	/*
	* Getting here means our parent vdev has made a very strange
	* request of us, and shouldn't happen. Assert here to force a
	* crash in dev builds, but in production return the IO
	* unhandled. The pool will likely suspend anyway but that's
	* nicer than crashing the kernel.
	*/
	ASSERT3S(zio->io_type, ==, -1);

	rw_exit(&vd->vd_lock);
	zio->io_error = SET_ERROR(ENOTSUP);
	zio_interrupt(zio);
	return;
	}

	__builtin_unreachable();
	}

	static void
	vdev_disk_io_done(zio_t *zio)
	{
	/*
	* If the device returned EIO, we revalidate the media. If it is
	* determined the media has changed this triggers the asynchronous
	* removal of the device from the configuration.
	*/
	if (zio->io_error == EIO) {
	vdev_t *v = zio->io_vd;
	vdev_disk_t *vd = v->vdev_tsd;

	if (!zfs_check_disk_status(BDH_BDEV(vd->vd_bdh))) {
	invalidate_bdev(BDH_BDEV(vd->vd_bdh));
	v->vdev_remove_wanted = B_TRUE;
	spa_async_request(zio->io_spa, SPA_ASYNC_REMOVE);
	}
	}
	}

	static void
	vdev_disk_hold(vdev_t *vd)
	{
	ASSERT(spa_config_held(vd->vdev_spa, SCL_STATE, RW_WRITER));

	/* We must have a pathname, and it must be absolute. */
	if (vd->vdev_path == NULL \|\| vd->vdev_path[0] != '/')
	return;

	/*
	* Only prefetch path and devid info if the device has
	* never been opened.
	*/
	if (vd->vdev_tsd != NULL)
	return;

	}

	static void
	vdev_disk_rele(vdev_t *vd)
	{
	ASSERT(spa_config_held(vd->vdev_spa, SCL_STATE, RW_WRITER));

	/* XXX: Implement me as a vnode rele for the device */
	}

	/*
	* BIO submission method. See comment above about vdev_classic.
	* Set zfs_vdev_disk_classic=0 for new, =1 for classic
	*/
	static uint_t zfs_vdev_disk_classic = 0; /* default new */

	/* Set submission function from module parameter */
	static int
	vdev_disk_param_set_classic(const char buf, zfs_kernel_param_t kp)
	{
	int err = param_set_uint(buf, kp);
	if (err < 0)
	return (SET_ERROR(err));

	vdev_disk_io_rw_fn =
	zfs_vdev_disk_classic ? vdev_classic_physio : vdev_disk_io_rw;

	printk(KERN_INFO "ZFS: forcing %s BIO submission\n",
	zfs_vdev_disk_classic ? "classic" : "new");

	return (0);
	}

	/*
	* At first use vdev use, set the submission function from the default value if
	* it hasn't been set already.
	*/
	static int
	vdev_disk_init(spa_t spa, nvlist_t nv, void **tsd)
	{
	(void) spa;
	(void) nv;
	(void) tsd;

	if (vdev_disk_io_rw_fn == NULL)
	vdev_disk_io_rw_fn = zfs_vdev_disk_classic ?
	vdev_classic_physio : vdev_disk_io_rw;

	return (0);
	}

	vdev_ops_t vdev_disk_ops = {
	.vdev_op_init = vdev_disk_init,
	.vdev_op_fini = NULL,
	.vdev_op_open = vdev_disk_open,
	.vdev_op_close = vdev_disk_close,
	.vdev_op_asize = vdev_default_asize,
	.vdev_op_min_asize = vdev_default_min_asize,
	.vdev_op_min_alloc = NULL,
	.vdev_op_io_start = vdev_disk_io_start,
	.vdev_op_io_done = vdev_disk_io_done,
	.vdev_op_state_change = NULL,
	.vdev_op_need_resilver = NULL,
	.vdev_op_hold = vdev_disk_hold,
	.vdev_op_rele = vdev_disk_rele,
	.vdev_op_remap = NULL,
	.vdev_op_xlate = vdev_default_xlate,
	.vdev_op_rebuild_asize = NULL,
	.vdev_op_metaslab_init = NULL,
	.vdev_op_config_generate = NULL,
	.vdev_op_nparity = NULL,
	.vdev_op_ndisks = NULL,
	.vdev_op_type = VDEV_TYPE_DISK, /* name of this vdev type */
	.vdev_op_leaf = B_TRUE, /* leaf vdev */
	.vdev_op_kobj_evt_post = vdev_disk_kobj_evt_post
	};

	/*
	* The zfs_vdev_scheduler module option has been deprecated. Setting this
	* value no longer has any effect. It has not yet been entirely removed
	* to allow the module to be loaded if this option is specified in the
	* /etc/modprobe.d/zfs.conf file. The following warning will be logged.
	*/
	static int
	param_set_vdev_scheduler(const char val, zfs_kernel_param_t kp)
	{
	int error = param_set_charp(val, kp);
	if (error == 0) {
	printk(KERN_INFO "The 'zfs_vdev_scheduler' module option "
	"is not supported.\n");
	}

	return (error);
	}

	static const char *zfs_vdev_scheduler = "unused";
	module_param_call(zfs_vdev_scheduler, param_set_vdev_scheduler,
	param_get_charp, &zfs_vdev_scheduler, 0644);
	MODULE_PARM_DESC(zfs_vdev_scheduler, "I/O scheduler");

	int
	param_set_min_auto_ashift(const char buf, zfs_kernel_param_t kp)
	{
	uint_t val;
	int error;

	error = kstrtouint(buf, 0, &val);
	if (error < 0)
	return (SET_ERROR(error));

	if (val < ASHIFT_MIN \|\| val > zfs_vdev_max_auto_ashift)
	return (SET_ERROR(-EINVAL));

	error = param_set_uint(buf, kp);
	if (error < 0)
	return (SET_ERROR(error));

	return (0);
	}

	int
	param_set_max_auto_ashift(const char buf, zfs_kernel_param_t kp)
	{
	uint_t val;
	int error;

	error = kstrtouint(buf, 0, &val);
	if (error < 0)
	return (SET_ERROR(error));

	if (val > ASHIFT_MAX \|\| val < zfs_vdev_min_auto_ashift)
	return (SET_ERROR(-EINVAL));

	error = param_set_uint(buf, kp);
	if (error < 0)
	return (SET_ERROR(error));

	return (0);
	}

	ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, open_timeout_ms, UINT, ZMOD_RW,
	"Timeout before determining that a device is missing");

	ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, failfast_mask, UINT, ZMOD_RW,
	"Defines failfast mask: 1 - device, 2 - transport, 4 - driver");

	ZFS_MODULE_PARAM(zfs_vdev_disk, zfs_vdev_disk_, max_segs, UINT, ZMOD_RW,
	"Maximum number of data segments to add to an IO request (min 4)");

	ZFS_MODULE_PARAM_CALL(zfs_vdev_disk, zfs_vdev_disk_, classic,
	vdev_disk_param_set_classic, param_get_uint, ZMOD_RD,
	"Use classic BIO submission method");
	diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zfs_debug.c b/sys/contrib/openzfs/module/os/linux/zfs/zfs_debug.c
	index f707959c9445..9ee40771fc19 100644
	--- a/sys/contrib/openzfs/module/os/linux/zfs/zfs_debug.c
	+++ b/sys/contrib/openzfs/module/os/linux/zfs/zfs_debug.c
	@@ -1,260 +1,260 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2014 by Delphix. All rights reserved.
	*/

	#include <sys/zfs_context.h>
	#include <sys/trace_zfs.h>

	typedef struct zfs_dbgmsg {
	procfs_list_node_t zdm_node;
	uint64_t zdm_timestamp;
	uint_t zdm_size;
	char zdm_msg[]; /* variable length allocation */
	} zfs_dbgmsg_t;

	static procfs_list_t zfs_dbgmsgs;
	static uint_t zfs_dbgmsg_size = 0;
	static uint_t zfs_dbgmsg_maxsize = 4<<20; /* 4MB */

	/*
	* Internal ZFS debug messages are enabled by default.
	*
	* # Print debug messages
	* cat /proc/spl/kstat/zfs/dbgmsg
	*
	* # Disable the kernel debug message log.
	* echo 0 > /sys/module/zfs/parameters/zfs_dbgmsg_enable
	*
	* # Clear the kernel debug message log.
	* echo 0 >/proc/spl/kstat/zfs/dbgmsg
	*/
	int zfs_dbgmsg_enable = B_TRUE;

	static int
	zfs_dbgmsg_show_header(struct seq_file *f)
	{
	seq_printf(f, "%-12s %-8s\n", "timestamp", "message");
	return (0);
	}

	static int
	zfs_dbgmsg_show(struct seq_file f, void p)
	{
	zfs_dbgmsg_t zdm = (zfs_dbgmsg_t )p;
	seq_printf(f, "%-12llu %-s\n",
	(u_longlong_t)zdm->zdm_timestamp, zdm->zdm_msg);
	return (0);
	}

	static void
	zfs_dbgmsg_purge(uint_t max_size)
	{
	while (zfs_dbgmsg_size > max_size) {
	zfs_dbgmsg_t *zdm = list_remove_head(&zfs_dbgmsgs.pl_list);
	if (zdm == NULL)
	return;

	uint_t size = zdm->zdm_size;
	kmem_free(zdm, size);
	zfs_dbgmsg_size -= size;
	}
	}

	static int
	zfs_dbgmsg_clear(procfs_list_t *procfs_list)
	{
	(void) procfs_list;
	mutex_enter(&zfs_dbgmsgs.pl_lock);
	zfs_dbgmsg_purge(0);
	mutex_exit(&zfs_dbgmsgs.pl_lock);
	return (0);
	}

	void
	zfs_dbgmsg_init(void)
	{
	procfs_list_install("zfs",
	NULL,
	"dbgmsg",
	0600,
	&zfs_dbgmsgs,
	zfs_dbgmsg_show,
	zfs_dbgmsg_show_header,
	zfs_dbgmsg_clear,
	offsetof(zfs_dbgmsg_t, zdm_node));
	}

	void
	zfs_dbgmsg_fini(void)
	{
	procfs_list_uninstall(&zfs_dbgmsgs);
	zfs_dbgmsg_purge(0);

	/*
	* TODO - decide how to make this permanent
	*/
	#ifdef _KERNEL
	procfs_list_destroy(&zfs_dbgmsgs);
	#endif
	}

	void
	__set_error(const char file, const char func, int line, int err)
	{
	/*
	* To enable this:
	*
	* $ echo 512 >/sys/module/zfs/parameters/zfs_flags
	*/
	if (zfs_flags & ZFS_DEBUG_SET_ERROR)
	__dprintf(B_FALSE, file, func, line, "error %lu",
	(ulong_t)err);
	}

	void
	__zfs_dbgmsg(char *buf)
	{
	uint_t size = sizeof (zfs_dbgmsg_t) + strlen(buf) + 1;
	zfs_dbgmsg_t *zdm = kmem_zalloc(size, KM_SLEEP);
	zdm->zdm_size = size;
	zdm->zdm_timestamp = gethrestime_sec();
	strcpy(zdm->zdm_msg, buf);

	mutex_enter(&zfs_dbgmsgs.pl_lock);
	procfs_list_add(&zfs_dbgmsgs, zdm);
	zfs_dbgmsg_size += size;
	zfs_dbgmsg_purge(zfs_dbgmsg_maxsize);
	mutex_exit(&zfs_dbgmsgs.pl_lock);
	}

	#ifdef _KERNEL

	void
	__dprintf(boolean_t dprint, const char file, const char func,
	int line, const char *fmt, ...)
	{
	const char *newfile;
	va_list adx;
	size_t size;
	char *buf;
	char *nl;
	int i;
	char *prefix = (dprint) ? "dprintf: " : "";

	size = 1024;
	buf = kmem_alloc(size, KM_SLEEP);

	/*
	* Get rid of annoying prefix to filename.
	*/
	newfile = strrchr(file, '/');
	if (newfile != NULL) {
	newfile = newfile + 1; /* Get rid of leading / */
	} else {
	newfile = file;
	}

	i = snprintf(buf, size, "%px %s%s:%d:%s(): ",
	curthread, prefix, newfile, line, func);

	if (i < size) {
	va_start(adx, fmt);
	(void) vsnprintf(buf + i, size - i, fmt, adx);
	va_end(adx);
	}

	/*
	* Get rid of trailing newline for dprintf logs.
	*/
	if (dprint && buf[0] != '\0') {
	nl = &buf[strlen(buf) - 1];
	if (*nl == '\n')
	*nl = '\0';
	}

	/*
	* To get this data enable the zfs__dprintf trace point as shown:
	*
	* # Enable zfs__dprintf tracepoint, clear the tracepoint ring buffer
	* $ echo 1 > /sys/kernel/debug/tracing/events/zfs/enable
	* $ echo 0 > /sys/kernel/debug/tracing/trace
	*
	* # Dump the ring buffer.
	* $ cat /sys/kernel/debug/tracing/trace
	*/
	DTRACE_PROBE1(zfs__dprintf, char *, buf);

	/*
	* To get this data:
	*
	* $ cat /proc/spl/kstat/zfs/dbgmsg
	*
	* To clear the buffer:
	* $ echo 0 > /proc/spl/kstat/zfs/dbgmsg
	*/
	__zfs_dbgmsg(buf);

	kmem_free(buf, size);
	}

	#else

	void
	-zfs_dbgmsg_print(const char *tag)
	+zfs_dbgmsg_print(int fd, const char *tag)
	{
	ssize_t ret __attribute__((unused));

	+ mutex_enter(&zfs_dbgmsgs.pl_lock);
	+
	/*
	* We use write() in this function instead of printf()
	* so it is safe to call from a signal handler.
	*/
	- ret = write(STDOUT_FILENO, "ZFS_DBGMSG(", 11);
	- ret = write(STDOUT_FILENO, tag, strlen(tag));
	- ret = write(STDOUT_FILENO, ") START:\n", 9);
	+ ret = write(fd, "ZFS_DBGMSG(", 11);
	+ ret = write(fd, tag, strlen(tag));
	+ ret = write(fd, ") START:\n", 9);

	- mutex_enter(&zfs_dbgmsgs.pl_lock);
	for (zfs_dbgmsg_t *zdm = list_head(&zfs_dbgmsgs.pl_list); zdm != NULL;
	zdm = list_next(&zfs_dbgmsgs.pl_list, zdm)) {
	- ret = write(STDOUT_FILENO, zdm->zdm_msg,
	- strlen(zdm->zdm_msg));
	- ret = write(STDOUT_FILENO, "\n", 1);
	+ ret = write(fd, zdm->zdm_msg, strlen(zdm->zdm_msg));
	+ ret = write(fd, "\n", 1);
	}

	- ret = write(STDOUT_FILENO, "ZFS_DBGMSG(", 11);
	- ret = write(STDOUT_FILENO, tag, strlen(tag));
	- ret = write(STDOUT_FILENO, ") END\n", 6);
	+ ret = write(fd, "ZFS_DBGMSG(", 11);
	+ ret = write(fd, tag, strlen(tag));
	+ ret = write(fd, ") END\n", 6);

	mutex_exit(&zfs_dbgmsgs.pl_lock);
	}
	#endif /* _KERNEL */

	#ifdef _KERNEL
	module_param(zfs_dbgmsg_enable, int, 0644);
	MODULE_PARM_DESC(zfs_dbgmsg_enable, "Enable ZFS debug message log");

	/* BEGIN CSTYLED */
	module_param(zfs_dbgmsg_maxsize, uint, 0644);
	/* END CSTYLED */
	MODULE_PARM_DESC(zfs_dbgmsg_maxsize, "Maximum ZFS debug log size");
	#endif
	diff --git a/sys/contrib/openzfs/module/os/linux/zfs/zvol_os.c b/sys/contrib/openzfs/module/os/linux/zfs/zvol_os.c
	index 2a036dc5136b..3e020e532263 100644
	--- a/sys/contrib/openzfs/module/os/linux/zfs/zvol_os.c
	+++ b/sys/contrib/openzfs/module/os/linux/zfs/zvol_os.c
	@@ -1,1771 +1,1772 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2012, 2020 by Delphix. All rights reserved.
	*/

	#include <sys/dataset_kstats.h>
	#include <sys/dbuf.h>
	#include <sys/dmu_traverse.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_prop.h>
	#include <sys/dsl_dir.h>
	#include <sys/zap.h>
	#include <sys/zfeature.h>
	#include <sys/zil_impl.h>
	#include <sys/dmu_tx.h>
	#include <sys/zio.h>
	#include <sys/zfs_rlock.h>
	#include <sys/spa_impl.h>
	#include <sys/zvol.h>
	#include <sys/zvol_impl.h>
	#include <cityhash.h>

	#include <linux/blkdev_compat.h>
	#include <linux/task_io_accounting_ops.h>

	#ifdef HAVE_BLK_MQ
	#include <linux/blk-mq.h>
	#endif

	static void zvol_request_impl(zvol_state_t zv, struct bio bio,
	struct request *rq, boolean_t force_sync);

	static unsigned int zvol_major = ZVOL_MAJOR;
	static unsigned int zvol_request_sync = 0;
	static unsigned int zvol_prefetch_bytes = (128 * 1024);
	static unsigned long zvol_max_discard_blocks = 16384;

	/*
	* Switch taskq at multiple of 512 MB offset. This can be set to a lower value
	* to utilize more threads for small files but may affect prefetch hits.
	*/
	#define ZVOL_TASKQ_OFFSET_SHIFT 29

	#ifndef HAVE_BLKDEV_GET_ERESTARTSYS
	static unsigned int zvol_open_timeout_ms = 1000;
	#endif

	static unsigned int zvol_threads = 0;
	#ifdef HAVE_BLK_MQ
	static unsigned int zvol_blk_mq_threads = 0;
	static unsigned int zvol_blk_mq_actual_threads;
	static boolean_t zvol_use_blk_mq = B_FALSE;

	/*
	* The maximum number of volblocksize blocks to process per thread. Typically,
	* write heavy workloads preform better with higher values here, and read
	* heavy workloads preform better with lower values, but that's not a hard
	* and fast rule. It's basically a knob to tune between "less overhead with
	* less parallelism" and "more overhead, but more parallelism".
	*
	* '8' was chosen as a reasonable, balanced, default based off of sequential
	* read and write tests to a zvol in an NVMe pool (with 16 CPUs).
	*/
	static unsigned int zvol_blk_mq_blocks_per_thread = 8;
	#endif

	static unsigned int zvol_num_taskqs = 0;

	#ifndef BLKDEV_DEFAULT_RQ
	/* BLKDEV_MAX_RQ was renamed to BLKDEV_DEFAULT_RQ in the 5.16 kernel */
	#define BLKDEV_DEFAULT_RQ BLKDEV_MAX_RQ
	#endif

	/*
	* Finalize our BIO or request.
	*/
	#ifdef HAVE_BLK_MQ
	#define END_IO(zv, bio, rq, error) do { \
	if (bio) { \
	BIO_END_IO(bio, error); \
	} else { \
	blk_mq_end_request(rq, errno_to_bi_status(error)); \
	} \
	} while (0)
	#else
	#define END_IO(zv, bio, rq, error) BIO_END_IO(bio, error)
	#endif

	#ifdef HAVE_BLK_MQ
	static unsigned int zvol_blk_mq_queue_depth = BLKDEV_DEFAULT_RQ;
	static unsigned int zvol_actual_blk_mq_queue_depth;
	#endif

	struct zvol_state_os {
	struct gendisk zvo_disk; / generic disk */
	struct request_queue zvo_queue; / request queue */
	dev_t zvo_dev; /* device id */

	#ifdef HAVE_BLK_MQ
	struct blk_mq_tag_set tag_set;
	#endif

	/* Set from the global 'zvol_use_blk_mq' at zvol load */
	boolean_t use_blk_mq;
	};

	typedef struct zv_taskq {
	uint_t tqs_cnt;
	taskq_t **tqs_taskq;
	} zv_taskq_t;
	static zv_taskq_t zvol_taskqs;
	static struct ida zvol_ida;

	typedef struct zv_request_stack {
	zvol_state_t *zv;
	struct bio *bio;
	struct request *rq;
	} zv_request_t;

	typedef struct zv_work {
	struct request *rq;
	struct work_struct work;
	} zv_work_t;

	typedef struct zv_request_task {
	zv_request_t zvr;
	taskq_ent_t ent;
	} zv_request_task_t;

	static zv_request_task_t *
	zv_request_task_create(zv_request_t zvr)
	{
	zv_request_task_t *task;
	task = kmem_alloc(sizeof (zv_request_task_t), KM_SLEEP);
	taskq_init_ent(&task->ent);
	task->zvr = zvr;
	return (task);
	}

	static void
	zv_request_task_free(zv_request_task_t *task)
	{
	kmem_free(task, sizeof (*task));
	}

	#ifdef HAVE_BLK_MQ

	/*
	* This is called when a new block multiqueue request comes in. A request
	* contains one or more BIOs.
	*/
	static blk_status_t zvol_mq_queue_rq(struct blk_mq_hw_ctx *hctx,
	const struct blk_mq_queue_data *bd)
	{
	struct request *rq = bd->rq;
	zvol_state_t *zv = rq->q->queuedata;

	/* Tell the kernel that we are starting to process this request */
	blk_mq_start_request(rq);

	if (blk_rq_is_passthrough(rq)) {
	/* Skip non filesystem request */
	blk_mq_end_request(rq, BLK_STS_IOERR);
	return (BLK_STS_IOERR);
	}

	zvol_request_impl(zv, NULL, rq, 0);

	/* Acknowledge to the kernel that we got this request */
	return (BLK_STS_OK);
	}

	static struct blk_mq_ops zvol_blk_mq_queue_ops = {
	.queue_rq = zvol_mq_queue_rq,
	};

	/* Initialize our blk-mq struct */
	static int zvol_blk_mq_alloc_tag_set(zvol_state_t *zv)
	{
	struct zvol_state_os *zso = zv->zv_zso;

	memset(&zso->tag_set, 0, sizeof (zso->tag_set));

	/* Initialize tag set. */
	zso->tag_set.ops = &zvol_blk_mq_queue_ops;
	zso->tag_set.nr_hw_queues = zvol_blk_mq_actual_threads;
	zso->tag_set.queue_depth = zvol_actual_blk_mq_queue_depth;
	zso->tag_set.numa_node = NUMA_NO_NODE;
	zso->tag_set.cmd_size = 0;

	/*
	* We need BLK_MQ_F_BLOCKING here since we do blocking calls in
	* zvol_request_impl()
	*/
	zso->tag_set.flags = BLK_MQ_F_SHOULD_MERGE \| BLK_MQ_F_BLOCKING;
	zso->tag_set.driver_data = zv;

	return (blk_mq_alloc_tag_set(&zso->tag_set));
	}
	#endif /* HAVE_BLK_MQ */

	/*
	* Given a path, return TRUE if path is a ZVOL.
	*/
	boolean_t
	zvol_os_is_zvol(const char *path)
	{
	dev_t dev = 0;

	if (vdev_lookup_bdev(path, &dev) != 0)
	return (B_FALSE);

	if (MAJOR(dev) == zvol_major)
	return (B_TRUE);

	return (B_FALSE);
	}

	static void
	zvol_write(zv_request_t *zvr)
	{
	struct bio *bio = zvr->bio;
	struct request *rq = zvr->rq;
	int error = 0;
	zfs_uio_t uio;
	zvol_state_t *zv = zvr->zv;
	struct request_queue *q;
	struct gendisk *disk;
	unsigned long start_time = 0;
	boolean_t acct = B_FALSE;

	ASSERT3P(zv, !=, NULL);
	ASSERT3U(zv->zv_open_count, >, 0);
	ASSERT3P(zv->zv_zilog, !=, NULL);

	q = zv->zv_zso->zvo_queue;
	disk = zv->zv_zso->zvo_disk;

	/* bio marked as FLUSH need to flush before write */
	if (io_is_flush(bio, rq))
	zil_commit(zv->zv_zilog, ZVOL_OBJ);

	/* Some requests are just for flush and nothing else. */
	if (io_size(bio, rq) == 0) {
	rw_exit(&zv->zv_suspend_lock);
	END_IO(zv, bio, rq, 0);
	return;
	}

	zfs_uio_bvec_init(&uio, bio, rq);

	ssize_t start_resid = uio.uio_resid;

	/*
	* With use_blk_mq, accounting is done by blk_mq_start_request()
	* and blk_mq_end_request(), so we can skip it here.
	*/
	if (bio) {
	acct = blk_queue_io_stat(q);
	if (acct) {
	start_time = blk_generic_start_io_acct(q, disk, WRITE,
	bio);
	}
	}

	boolean_t sync =
	io_is_fua(bio, rq) \|\| zv->zv_objset->os_sync == ZFS_SYNC_ALWAYS;

	zfs_locked_range_t *lr = zfs_rangelock_enter(&zv->zv_rangelock,
	uio.uio_loffset, uio.uio_resid, RL_WRITER);

	uint64_t volsize = zv->zv_volsize;
	while (uio.uio_resid > 0 && uio.uio_loffset < volsize) {
	uint64_t bytes = MIN(uio.uio_resid, DMU_MAX_ACCESS >> 1);
	uint64_t off = uio.uio_loffset;
	dmu_tx_t *tx = dmu_tx_create(zv->zv_objset);

	if (bytes > volsize - off) /* don't write past the end */
	bytes = volsize - off;

	dmu_tx_hold_write_by_dnode(tx, zv->zv_dn, off, bytes);

	/* This will only fail for ENOSPC */
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error) {
	dmu_tx_abort(tx);
	break;
	}
	error = dmu_write_uio_dnode(zv->zv_dn, &uio, bytes, tx);
	if (error == 0) {
	zvol_log_write(zv, tx, off, bytes, sync);
	}
	dmu_tx_commit(tx);

	if (error)
	break;
	}
	zfs_rangelock_exit(lr);

	int64_t nwritten = start_resid - uio.uio_resid;
	dataset_kstats_update_write_kstats(&zv->zv_kstat, nwritten);
	task_io_account_write(nwritten);

	if (sync)
	zil_commit(zv->zv_zilog, ZVOL_OBJ);

	rw_exit(&zv->zv_suspend_lock);

	if (bio && acct) {
	blk_generic_end_io_acct(q, disk, WRITE, bio, start_time);
	}

	END_IO(zv, bio, rq, -error);
	}

	static void
	zvol_write_task(void *arg)
	{
	zv_request_task_t *task = arg;
	zvol_write(&task->zvr);
	zv_request_task_free(task);
	}

	static void
	zvol_discard(zv_request_t *zvr)
	{
	struct bio *bio = zvr->bio;
	struct request *rq = zvr->rq;
	zvol_state_t *zv = zvr->zv;
	uint64_t start = io_offset(bio, rq);
	uint64_t size = io_size(bio, rq);
	uint64_t end = start + size;
	boolean_t sync;
	int error = 0;
	dmu_tx_t *tx;
	struct request_queue *q = zv->zv_zso->zvo_queue;
	struct gendisk *disk = zv->zv_zso->zvo_disk;
	unsigned long start_time = 0;
	boolean_t acct = B_FALSE;

	ASSERT3P(zv, !=, NULL);
	ASSERT3U(zv->zv_open_count, >, 0);
	ASSERT3P(zv->zv_zilog, !=, NULL);

	if (bio) {
	acct = blk_queue_io_stat(q);
	if (acct) {
	start_time = blk_generic_start_io_acct(q, disk, WRITE,
	bio);
	}
	}

	sync = io_is_fua(bio, rq) \|\| zv->zv_objset->os_sync == ZFS_SYNC_ALWAYS;

	if (end > zv->zv_volsize) {
	error = SET_ERROR(EIO);
	goto unlock;
	}

	/*
	* Align the request to volume block boundaries when a secure erase is
	* not required. This will prevent dnode_free_range() from zeroing out
	* the unaligned parts which is slow (read-modify-write) and useless
	* since we are not freeing any space by doing so.
	*/
	if (!io_is_secure_erase(bio, rq)) {
	start = P2ROUNDUP(start, zv->zv_volblocksize);
	- end = P2ALIGN(end, zv->zv_volblocksize);
	+ end = P2ALIGN_TYPED(end, zv->zv_volblocksize, uint64_t);
	size = end - start;
	}

	if (start >= end)
	goto unlock;

	zfs_locked_range_t *lr = zfs_rangelock_enter(&zv->zv_rangelock,
	start, size, RL_WRITER);

	tx = dmu_tx_create(zv->zv_objset);
	dmu_tx_mark_netfree(tx);
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error != 0) {
	dmu_tx_abort(tx);
	} else {
	zvol_log_truncate(zv, tx, start, size);
	dmu_tx_commit(tx);
	error = dmu_free_long_range(zv->zv_objset,
	ZVOL_OBJ, start, size);
	}
	zfs_rangelock_exit(lr);

	if (error == 0 && sync)
	zil_commit(zv->zv_zilog, ZVOL_OBJ);

	unlock:
	rw_exit(&zv->zv_suspend_lock);

	if (bio && acct) {
	blk_generic_end_io_acct(q, disk, WRITE, bio,
	start_time);
	}

	END_IO(zv, bio, rq, -error);
	}

	static void
	zvol_discard_task(void *arg)
	{
	zv_request_task_t *task = arg;
	zvol_discard(&task->zvr);
	zv_request_task_free(task);
	}

	static void
	zvol_read(zv_request_t *zvr)
	{
	struct bio *bio = zvr->bio;
	struct request *rq = zvr->rq;
	int error = 0;
	zfs_uio_t uio;
	boolean_t acct = B_FALSE;
	zvol_state_t *zv = zvr->zv;
	struct request_queue *q;
	struct gendisk *disk;
	unsigned long start_time = 0;

	ASSERT3P(zv, !=, NULL);
	ASSERT3U(zv->zv_open_count, >, 0);

	zfs_uio_bvec_init(&uio, bio, rq);

	q = zv->zv_zso->zvo_queue;
	disk = zv->zv_zso->zvo_disk;

	ssize_t start_resid = uio.uio_resid;

	/*
	* When blk-mq is being used, accounting is done by
	* blk_mq_start_request() and blk_mq_end_request().
	*/
	if (bio) {
	acct = blk_queue_io_stat(q);
	if (acct)
	start_time = blk_generic_start_io_acct(q, disk, READ,
	bio);
	}

	zfs_locked_range_t *lr = zfs_rangelock_enter(&zv->zv_rangelock,
	uio.uio_loffset, uio.uio_resid, RL_READER);

	uint64_t volsize = zv->zv_volsize;

	while (uio.uio_resid > 0 && uio.uio_loffset < volsize) {
	uint64_t bytes = MIN(uio.uio_resid, DMU_MAX_ACCESS >> 1);

	/* don't read past the end */
	if (bytes > volsize - uio.uio_loffset)
	bytes = volsize - uio.uio_loffset;

	error = dmu_read_uio_dnode(zv->zv_dn, &uio, bytes);
	if (error) {
	/* convert checksum errors into IO errors */
	if (error == ECKSUM)
	error = SET_ERROR(EIO);
	break;
	}
	}
	zfs_rangelock_exit(lr);

	int64_t nread = start_resid - uio.uio_resid;
	dataset_kstats_update_read_kstats(&zv->zv_kstat, nread);
	task_io_account_read(nread);

	rw_exit(&zv->zv_suspend_lock);

	if (bio && acct) {
	blk_generic_end_io_acct(q, disk, READ, bio, start_time);
	}

	END_IO(zv, bio, rq, -error);
	}

	static void
	zvol_read_task(void *arg)
	{
	zv_request_task_t *task = arg;
	zvol_read(&task->zvr);
	zv_request_task_free(task);
	}


	/*
	* Process a BIO or request
	*
	* Either 'bio' or 'rq' should be set depending on if we are processing a
	* bio or a request (both should not be set).
	*
	* force_sync: Set to 0 to defer processing to a background taskq
	* Set to 1 to process data synchronously
	*/
	static void
	zvol_request_impl(zvol_state_t zv, struct bio bio, struct request *rq,
	boolean_t force_sync)
	{
	fstrans_cookie_t cookie = spl_fstrans_mark();
	uint64_t offset = io_offset(bio, rq);
	uint64_t size = io_size(bio, rq);
	int rw = io_data_dir(bio, rq);

	if (zvol_request_sync \|\| zv->zv_threading == B_FALSE)
	force_sync = 1;

	zv_request_t zvr = {
	.zv = zv,
	.bio = bio,
	.rq = rq,
	};

	if (io_has_data(bio, rq) && offset + size > zv->zv_volsize) {
	printk(KERN_INFO "%s: bad access: offset=%llu, size=%lu\n",
	zv->zv_zso->zvo_disk->disk_name,
	(long long unsigned)offset,
	(long unsigned)size);

	END_IO(zv, bio, rq, -SET_ERROR(EIO));
	goto out;
	}

	zv_request_task_t *task;
	zv_taskq_t *ztqs = &zvol_taskqs;
	uint_t blk_mq_hw_queue = 0;
	uint_t tq_idx;
	uint_t taskq_hash;
	#ifdef HAVE_BLK_MQ
	if (rq)
	#ifdef HAVE_BLK_MQ_RQ_HCTX
	blk_mq_hw_queue = rq->mq_hctx->queue_num;
	#else
	blk_mq_hw_queue =
	rq->q->queue_hw_ctx[rq->q->mq_map[rq->cpu]]->queue_num;
	#endif
	#endif
	taskq_hash = cityhash4((uintptr_t)zv, offset >> ZVOL_TASKQ_OFFSET_SHIFT,
	blk_mq_hw_queue, 0);
	tq_idx = taskq_hash % ztqs->tqs_cnt;

	if (rw == WRITE) {
	if (unlikely(zv->zv_flags & ZVOL_RDONLY)) {
	END_IO(zv, bio, rq, -SET_ERROR(EROFS));
	goto out;
	}

	/*
	* Prevents the zvol from being suspended, or the ZIL being
	* concurrently opened. Will be released after the i/o
	* completes.
	*/
	rw_enter(&zv->zv_suspend_lock, RW_READER);

	/*
	* Open a ZIL if this is the first time we have written to this
	* zvol. We protect zv->zv_zilog with zv_suspend_lock rather
	* than zv_state_lock so that we don't need to acquire an
	* additional lock in this path.
	*/
	if (zv->zv_zilog == NULL) {
	rw_exit(&zv->zv_suspend_lock);
	rw_enter(&zv->zv_suspend_lock, RW_WRITER);
	if (zv->zv_zilog == NULL) {
	zv->zv_zilog = zil_open(zv->zv_objset,
	zvol_get_data, &zv->zv_kstat.dk_zil_sums);
	zv->zv_flags \|= ZVOL_WRITTEN_TO;
	/* replay / destroy done in zvol_create_minor */
	VERIFY0((zv->zv_zilog->zl_header->zh_flags &
	ZIL_REPLAY_NEEDED));
	}
	rw_downgrade(&zv->zv_suspend_lock);
	}

	/*
	* We don't want this thread to be blocked waiting for i/o to
	* complete, so we instead wait from a taskq callback. The
	* i/o may be a ZIL write (via zil_commit()), or a read of an
	* indirect block, or a read of a data block (if this is a
	* partial-block write). We will indicate that the i/o is
	* complete by calling END_IO() from the taskq callback.
	*
	* This design allows the calling thread to continue and
	* initiate more concurrent operations by calling
	* zvol_request() again. There are typically only a small
	* number of threads available to call zvol_request() (e.g.
	* one per iSCSI target), so keeping the latency of
	* zvol_request() low is important for performance.
	*
	* The zvol_request_sync module parameter allows this
	* behavior to be altered, for performance evaluation
	* purposes. If the callback blocks, setting
	* zvol_request_sync=1 will result in much worse performance.
	*
	* We can have up to zvol_threads concurrent i/o's being
	* processed for all zvols on the system. This is typically
	* a vast improvement over the zvol_request_sync=1 behavior
	* of one i/o at a time per zvol. However, an even better
	* design would be for zvol_request() to initiate the zio
	* directly, and then be notified by the zio_done callback,
	* which would call END_IO(). Unfortunately, the DMU/ZIL
	* interfaces lack this functionality (they block waiting for
	* the i/o to complete).
	*/
	if (io_is_discard(bio, rq) \|\| io_is_secure_erase(bio, rq)) {
	if (force_sync) {
	zvol_discard(&zvr);
	} else {
	task = zv_request_task_create(zvr);
	taskq_dispatch_ent(ztqs->tqs_taskq[tq_idx],
	zvol_discard_task, task, 0, &task->ent);
	}
	} else {
	if (force_sync) {
	zvol_write(&zvr);
	} else {
	task = zv_request_task_create(zvr);
	taskq_dispatch_ent(ztqs->tqs_taskq[tq_idx],
	zvol_write_task, task, 0, &task->ent);
	}
	}
	} else {
	/*
	* The SCST driver, and possibly others, may issue READ I/Os
	* with a length of zero bytes. These empty I/Os contain no
	* data and require no additional handling.
	*/
	if (size == 0) {
	END_IO(zv, bio, rq, 0);
	goto out;
	}

	rw_enter(&zv->zv_suspend_lock, RW_READER);

	/* See comment in WRITE case above. */
	if (force_sync) {
	zvol_read(&zvr);
	} else {
	task = zv_request_task_create(zvr);
	taskq_dispatch_ent(ztqs->tqs_taskq[tq_idx],
	zvol_read_task, task, 0, &task->ent);
	}
	}

	out:
	spl_fstrans_unmark(cookie);
	}

	#ifdef HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS
	#ifdef HAVE_BDEV_SUBMIT_BIO_RETURNS_VOID
	static void
	zvol_submit_bio(struct bio *bio)
	#else
	static blk_qc_t
	zvol_submit_bio(struct bio *bio)
	#endif
	#else
	static MAKE_REQUEST_FN_RET
	zvol_request(struct request_queue q, struct bio bio)
	#endif
	{
	#ifdef HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS
	#if defined(HAVE_BIO_BDEV_DISK)
	struct request_queue *q = bio->bi_bdev->bd_disk->queue;
	#else
	struct request_queue *q = bio->bi_disk->queue;
	#endif
	#endif
	zvol_state_t *zv = q->queuedata;

	zvol_request_impl(zv, bio, NULL, 0);
	#if defined(HAVE_MAKE_REQUEST_FN_RET_QC) \|\| \
	defined(HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS) && \
	!defined(HAVE_BDEV_SUBMIT_BIO_RETURNS_VOID)
	return (BLK_QC_T_NONE);
	#endif
	}

	static int
	#ifdef HAVE_BLK_MODE_T
	zvol_open(struct gendisk *disk, blk_mode_t flag)
	#else
	zvol_open(struct block_device *bdev, fmode_t flag)
	#endif
	{
	zvol_state_t *zv;
	int error = 0;
	boolean_t drop_suspend = B_FALSE;
	#ifndef HAVE_BLKDEV_GET_ERESTARTSYS
	hrtime_t timeout = MSEC2NSEC(zvol_open_timeout_ms);
	hrtime_t start = gethrtime();

	retry:
	#endif
	rw_enter(&zvol_state_lock, RW_READER);
	/*
	* Obtain a copy of private_data under the zvol_state_lock to make
	* sure that either the result of zvol free code path setting
	* disk->private_data to NULL is observed, or zvol_os_free()
	* is not called on this zv because of the positive zv_open_count.
	*/
	#ifdef HAVE_BLK_MODE_T
	zv = disk->private_data;
	#else
	zv = bdev->bd_disk->private_data;
	#endif
	if (zv == NULL) {
	rw_exit(&zvol_state_lock);
	return (SET_ERROR(-ENXIO));
	}

	mutex_enter(&zv->zv_state_lock);
	/*
	* Make sure zvol is not suspended during first open
	* (hold zv_suspend_lock) and respect proper lock acquisition
	* ordering - zv_suspend_lock before zv_state_lock
	*/
	if (zv->zv_open_count == 0) {
	if (!rw_tryenter(&zv->zv_suspend_lock, RW_READER)) {
	mutex_exit(&zv->zv_state_lock);
	rw_enter(&zv->zv_suspend_lock, RW_READER);
	mutex_enter(&zv->zv_state_lock);
	/* check to see if zv_suspend_lock is needed */
	if (zv->zv_open_count != 0) {
	rw_exit(&zv->zv_suspend_lock);
	} else {
	drop_suspend = B_TRUE;
	}
	} else {
	drop_suspend = B_TRUE;
	}
	}
	rw_exit(&zvol_state_lock);

	ASSERT(MUTEX_HELD(&zv->zv_state_lock));

	if (zv->zv_open_count == 0) {
	boolean_t drop_namespace = B_FALSE;

	ASSERT(RW_READ_HELD(&zv->zv_suspend_lock));

	/*
	* In all other call paths the spa_namespace_lock is taken
	* before the bdev->bd_mutex lock. However, on open(2)
	* the __blkdev_get() function calls fops->open() with the
	* bdev->bd_mutex lock held. This can result in a deadlock
	* when zvols from one pool are used as vdevs in another.
	*
	* To prevent a lock inversion deadlock we preemptively
	* take the spa_namespace_lock. Normally the lock will not
	* be contended and this is safe because spa_open_common()
	* handles the case where the caller already holds the
	* spa_namespace_lock.
	*
	* When the lock cannot be aquired after multiple retries
	* this must be the vdev on zvol deadlock case and we have
	* no choice but to return an error. For 5.12 and older
	* kernels returning -ERESTARTSYS will result in the
	* bdev->bd_mutex being dropped, then reacquired, and
	* fops->open() being called again. This process can be
	* repeated safely until both locks are acquired. For 5.13
	* and newer the -ERESTARTSYS retry logic was removed from
	* the kernel so the only option is to return the error for
	* the caller to handle it.
	*/
	if (!mutex_owned(&spa_namespace_lock)) {
	if (!mutex_tryenter(&spa_namespace_lock)) {
	mutex_exit(&zv->zv_state_lock);
	rw_exit(&zv->zv_suspend_lock);

	#ifdef HAVE_BLKDEV_GET_ERESTARTSYS
	schedule();
	return (SET_ERROR(-ERESTARTSYS));
	#else
	if ((gethrtime() - start) > timeout)
	return (SET_ERROR(-ERESTARTSYS));

	- schedule_timeout(MSEC_TO_TICK(10));
	+ schedule_timeout_interruptible(
	+ MSEC_TO_TICK(10));
	goto retry;
	#endif
	} else {
	drop_namespace = B_TRUE;
	}
	}

	error = -zvol_first_open(zv, !(blk_mode_is_open_write(flag)));

	if (drop_namespace)
	mutex_exit(&spa_namespace_lock);
	}

	if (error == 0) {
	if ((blk_mode_is_open_write(flag)) &&
	(zv->zv_flags & ZVOL_RDONLY)) {
	if (zv->zv_open_count == 0)
	zvol_last_close(zv);

	error = SET_ERROR(-EROFS);
	} else {
	zv->zv_open_count++;
	}
	}

	mutex_exit(&zv->zv_state_lock);
	if (drop_suspend)
	rw_exit(&zv->zv_suspend_lock);

	if (error == 0)
	#ifdef HAVE_BLK_MODE_T
	disk_check_media_change(disk);
	#else
	zfs_check_media_change(bdev);
	#endif

	return (error);
	}

	static void
	#ifdef HAVE_BLOCK_DEVICE_OPERATIONS_RELEASE_1ARG
	zvol_release(struct gendisk *disk)
	#else
	zvol_release(struct gendisk *disk, fmode_t unused)
	#endif
	{
	#if !defined(HAVE_BLOCK_DEVICE_OPERATIONS_RELEASE_1ARG)
	(void) unused;
	#endif
	zvol_state_t *zv;
	boolean_t drop_suspend = B_TRUE;

	rw_enter(&zvol_state_lock, RW_READER);
	zv = disk->private_data;

	mutex_enter(&zv->zv_state_lock);
	ASSERT3U(zv->zv_open_count, >, 0);
	/*
	* make sure zvol is not suspended during last close
	* (hold zv_suspend_lock) and respect proper lock acquisition
	* ordering - zv_suspend_lock before zv_state_lock
	*/
	if (zv->zv_open_count == 1) {
	if (!rw_tryenter(&zv->zv_suspend_lock, RW_READER)) {
	mutex_exit(&zv->zv_state_lock);
	rw_enter(&zv->zv_suspend_lock, RW_READER);
	mutex_enter(&zv->zv_state_lock);
	/* check to see if zv_suspend_lock is needed */
	if (zv->zv_open_count != 1) {
	rw_exit(&zv->zv_suspend_lock);
	drop_suspend = B_FALSE;
	}
	}
	} else {
	drop_suspend = B_FALSE;
	}
	rw_exit(&zvol_state_lock);

	ASSERT(MUTEX_HELD(&zv->zv_state_lock));

	zv->zv_open_count--;
	if (zv->zv_open_count == 0) {
	ASSERT(RW_READ_HELD(&zv->zv_suspend_lock));
	zvol_last_close(zv);
	}

	mutex_exit(&zv->zv_state_lock);

	if (drop_suspend)
	rw_exit(&zv->zv_suspend_lock);
	}

	static int
	zvol_ioctl(struct block_device *bdev, fmode_t mode,
	unsigned int cmd, unsigned long arg)
	{
	zvol_state_t *zv = bdev->bd_disk->private_data;
	int error = 0;

	ASSERT3U(zv->zv_open_count, >, 0);

	switch (cmd) {
	case BLKFLSBUF:
	#ifdef HAVE_FSYNC_BDEV
	fsync_bdev(bdev);
	#elif defined(HAVE_SYNC_BLOCKDEV)
	sync_blockdev(bdev);
	#else
	#error "Neither fsync_bdev() nor sync_blockdev() found"
	#endif
	invalidate_bdev(bdev);
	rw_enter(&zv->zv_suspend_lock, RW_READER);

	if (!(zv->zv_flags & ZVOL_RDONLY))
	txg_wait_synced(dmu_objset_pool(zv->zv_objset), 0);

	rw_exit(&zv->zv_suspend_lock);
	break;

	case BLKZNAME:
	mutex_enter(&zv->zv_state_lock);
	error = copy_to_user((void *)arg, zv->zv_name, MAXNAMELEN);
	mutex_exit(&zv->zv_state_lock);
	break;

	default:
	error = -ENOTTY;
	break;
	}

	return (SET_ERROR(error));
	}

	#ifdef CONFIG_COMPAT
	static int
	zvol_compat_ioctl(struct block_device *bdev, fmode_t mode,
	unsigned cmd, unsigned long arg)
	{
	return (zvol_ioctl(bdev, mode, cmd, arg));
	}
	#else
	#define zvol_compat_ioctl NULL
	#endif

	static unsigned int
	zvol_check_events(struct gendisk *disk, unsigned int clearing)
	{
	unsigned int mask = 0;

	rw_enter(&zvol_state_lock, RW_READER);

	zvol_state_t *zv = disk->private_data;
	if (zv != NULL) {
	mutex_enter(&zv->zv_state_lock);
	mask = zv->zv_changed ? DISK_EVENT_MEDIA_CHANGE : 0;
	zv->zv_changed = 0;
	mutex_exit(&zv->zv_state_lock);
	}

	rw_exit(&zvol_state_lock);

	return (mask);
	}

	static int
	zvol_revalidate_disk(struct gendisk *disk)
	{
	rw_enter(&zvol_state_lock, RW_READER);

	zvol_state_t *zv = disk->private_data;
	if (zv != NULL) {
	mutex_enter(&zv->zv_state_lock);
	set_capacity(zv->zv_zso->zvo_disk,
	zv->zv_volsize >> SECTOR_BITS);
	mutex_exit(&zv->zv_state_lock);
	}

	rw_exit(&zvol_state_lock);

	return (0);
	}

	int
	zvol_os_update_volsize(zvol_state_t *zv, uint64_t volsize)
	{
	struct gendisk *disk = zv->zv_zso->zvo_disk;

	#if defined(HAVE_REVALIDATE_DISK_SIZE)
	revalidate_disk_size(disk, zvol_revalidate_disk(disk) == 0);
	#elif defined(HAVE_REVALIDATE_DISK)
	revalidate_disk(disk);
	#else
	zvol_revalidate_disk(disk);
	#endif
	return (0);
	}

	void
	zvol_os_clear_private(zvol_state_t *zv)
	{
	/*
	* Cleared while holding zvol_state_lock as a writer
	* which will prevent zvol_open() from opening it.
	*/
	zv->zv_zso->zvo_disk->private_data = NULL;
	}

	/*
	* Provide a simple virtual geometry for legacy compatibility. For devices
	* smaller than 1 MiB a small head and sector count is used to allow very
	* tiny devices. For devices over 1 Mib a standard head and sector count
	* is used to keep the cylinders count reasonable.
	*/
	static int
	zvol_getgeo(struct block_device bdev, struct hd_geometry geo)
	{
	zvol_state_t *zv = bdev->bd_disk->private_data;
	sector_t sectors;

	ASSERT3U(zv->zv_open_count, >, 0);

	sectors = get_capacity(zv->zv_zso->zvo_disk);

	if (sectors > 2048) {
	geo->heads = 16;
	geo->sectors = 63;
	} else {
	geo->heads = 2;
	geo->sectors = 4;
	}

	geo->start = 0;
	geo->cylinders = sectors / (geo->heads * geo->sectors);

	return (0);
	}

	/*
	* Why have two separate block_device_operations structs?
	*
	* Normally we'd just have one, and assign 'submit_bio' as needed. However,
	* it's possible the user's kernel is built with CONSTIFY_PLUGIN, meaning we
	* can't just change submit_bio dynamically at runtime. So just create two
	* separate structs to get around this.
	*/
	static const struct block_device_operations zvol_ops_blk_mq = {
	.open = zvol_open,
	.release = zvol_release,
	.ioctl = zvol_ioctl,
	.compat_ioctl = zvol_compat_ioctl,
	.check_events = zvol_check_events,
	#ifdef HAVE_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK
	.revalidate_disk = zvol_revalidate_disk,
	#endif
	.getgeo = zvol_getgeo,
	.owner = THIS_MODULE,
	};

	static const struct block_device_operations zvol_ops = {
	.open = zvol_open,
	.release = zvol_release,
	.ioctl = zvol_ioctl,
	.compat_ioctl = zvol_compat_ioctl,
	.check_events = zvol_check_events,
	#ifdef HAVE_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK
	.revalidate_disk = zvol_revalidate_disk,
	#endif
	.getgeo = zvol_getgeo,
	.owner = THIS_MODULE,
	#ifdef HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS
	.submit_bio = zvol_submit_bio,
	#endif
	};

	static int
	zvol_alloc_non_blk_mq(struct zvol_state_os *zso)
	{
	#if defined(HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS)
	#if defined(HAVE_BLK_ALLOC_DISK)
	zso->zvo_disk = blk_alloc_disk(NUMA_NO_NODE);
	if (zso->zvo_disk == NULL)
	return (1);

	zso->zvo_disk->minors = ZVOL_MINORS;
	zso->zvo_queue = zso->zvo_disk->queue;
	#elif defined(HAVE_BLK_ALLOC_DISK_2ARG)
	struct gendisk *disk = blk_alloc_disk(NULL, NUMA_NO_NODE);
	if (IS_ERR(disk)) {
	zso->zvo_disk = NULL;
	return (1);
	}

	zso->zvo_disk = disk;
	zso->zvo_disk->minors = ZVOL_MINORS;
	zso->zvo_queue = zso->zvo_disk->queue;
	#else
	zso->zvo_queue = blk_alloc_queue(NUMA_NO_NODE);
	if (zso->zvo_queue == NULL)
	return (1);

	zso->zvo_disk = alloc_disk(ZVOL_MINORS);
	if (zso->zvo_disk == NULL) {
	blk_cleanup_queue(zso->zvo_queue);
	return (1);
	}

	zso->zvo_disk->queue = zso->zvo_queue;
	#endif /* HAVE_BLK_ALLOC_DISK */
	#else
	zso->zvo_queue = blk_generic_alloc_queue(zvol_request, NUMA_NO_NODE);
	if (zso->zvo_queue == NULL)
	return (1);

	zso->zvo_disk = alloc_disk(ZVOL_MINORS);
	if (zso->zvo_disk == NULL) {
	blk_cleanup_queue(zso->zvo_queue);
	return (1);
	}

	zso->zvo_disk->queue = zso->zvo_queue;
	#endif /* HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS */
	return (0);

	}

	static int
	zvol_alloc_blk_mq(zvol_state_t *zv)
	{
	#ifdef HAVE_BLK_MQ
	struct zvol_state_os *zso = zv->zv_zso;

	/* Allocate our blk-mq tag_set */
	if (zvol_blk_mq_alloc_tag_set(zv) != 0)
	return (1);

	#if defined(HAVE_BLK_ALLOC_DISK)
	zso->zvo_disk = blk_mq_alloc_disk(&zso->tag_set, zv);
	if (zso->zvo_disk == NULL) {
	blk_mq_free_tag_set(&zso->tag_set);
	return (1);
	}
	zso->zvo_queue = zso->zvo_disk->queue;
	zso->zvo_disk->minors = ZVOL_MINORS;
	#elif defined(HAVE_BLK_ALLOC_DISK_2ARG)
	struct gendisk *disk = blk_mq_alloc_disk(&zso->tag_set, NULL, zv);
	if (IS_ERR(disk)) {
	zso->zvo_disk = NULL;
	blk_mq_free_tag_set(&zso->tag_set);
	return (1);
	}

	zso->zvo_disk = disk;
	zso->zvo_queue = zso->zvo_disk->queue;
	zso->zvo_disk->minors = ZVOL_MINORS;
	#else
	zso->zvo_disk = alloc_disk(ZVOL_MINORS);
	if (zso->zvo_disk == NULL) {
	blk_cleanup_queue(zso->zvo_queue);
	blk_mq_free_tag_set(&zso->tag_set);
	return (1);
	}
	/* Allocate queue */
	zso->zvo_queue = blk_mq_init_queue(&zso->tag_set);
	if (IS_ERR(zso->zvo_queue)) {
	blk_mq_free_tag_set(&zso->tag_set);
	return (1);
	}

	/* Our queue is now created, assign it to our disk */
	zso->zvo_disk->queue = zso->zvo_queue;

	#endif
	#endif
	return (0);
	}

	/*
	* Allocate memory for a new zvol_state_t and setup the required
	* request queue and generic disk structures for the block device.
	*/
	static zvol_state_t *
	zvol_alloc(dev_t dev, const char *name)
	{
	zvol_state_t *zv;
	struct zvol_state_os *zso;
	uint64_t volmode;
	int ret;

	if (dsl_prop_get_integer(name, "volmode", &volmode, NULL) != 0)
	return (NULL);

	if (volmode == ZFS_VOLMODE_DEFAULT)
	volmode = zvol_volmode;

	if (volmode == ZFS_VOLMODE_NONE)
	return (NULL);

	zv = kmem_zalloc(sizeof (zvol_state_t), KM_SLEEP);
	zso = kmem_zalloc(sizeof (struct zvol_state_os), KM_SLEEP);
	zv->zv_zso = zso;
	zv->zv_volmode = volmode;

	list_link_init(&zv->zv_next);
	mutex_init(&zv->zv_state_lock, NULL, MUTEX_DEFAULT, NULL);

	#ifdef HAVE_BLK_MQ
	zv->zv_zso->use_blk_mq = zvol_use_blk_mq;
	#endif

	/*
	* The block layer has 3 interfaces for getting BIOs:
	*
	* 1. blk-mq request queues (new)
	* 2. submit_bio() (oldest)
	* 3. regular request queues (old).
	*
	* Each of those interfaces has two permutations:
	*
	* a) We have blk_alloc_disk()/blk_mq_alloc_disk(), which allocates
	* both the disk and its queue (5.14 kernel or newer)
	*
	* b) We don't have blk_*alloc_disk(), and have to allocate the
	* disk and the queue separately. (5.13 kernel or older)
	*/
	if (zv->zv_zso->use_blk_mq) {
	ret = zvol_alloc_blk_mq(zv);
	zso->zvo_disk->fops = &zvol_ops_blk_mq;
	} else {
	ret = zvol_alloc_non_blk_mq(zso);
	zso->zvo_disk->fops = &zvol_ops;
	}
	if (ret != 0)
	goto out_kmem;

	blk_queue_set_write_cache(zso->zvo_queue, B_TRUE, B_TRUE);

	/* Limit read-ahead to a single page to prevent over-prefetching. */
	blk_queue_set_read_ahead(zso->zvo_queue, 1);

	if (!zv->zv_zso->use_blk_mq) {
	/* Disable write merging in favor of the ZIO pipeline. */
	blk_queue_flag_set(QUEUE_FLAG_NOMERGES, zso->zvo_queue);
	}

	/* Enable /proc/diskstats */
	blk_queue_flag_set(QUEUE_FLAG_IO_STAT, zso->zvo_queue);

	zso->zvo_queue->queuedata = zv;
	zso->zvo_dev = dev;
	zv->zv_open_count = 0;
	strlcpy(zv->zv_name, name, sizeof (zv->zv_name));

	zfs_rangelock_init(&zv->zv_rangelock, NULL, NULL);
	rw_init(&zv->zv_suspend_lock, NULL, RW_DEFAULT, NULL);

	zso->zvo_disk->major = zvol_major;
	zso->zvo_disk->events = DISK_EVENT_MEDIA_CHANGE;

	/*
	* Setting ZFS_VOLMODE_DEV disables partitioning on ZVOL devices.
	* This is accomplished by limiting the number of minors for the
	* device to one and explicitly disabling partition scanning.
	*/
	if (volmode == ZFS_VOLMODE_DEV) {
	zso->zvo_disk->minors = 1;
	zso->zvo_disk->flags &= ~ZFS_GENHD_FL_EXT_DEVT;
	zso->zvo_disk->flags \|= ZFS_GENHD_FL_NO_PART;
	}

	zso->zvo_disk->first_minor = (dev & MINORMASK);
	zso->zvo_disk->private_data = zv;
	snprintf(zso->zvo_disk->disk_name, DISK_NAME_LEN, "%s%d",
	ZVOL_DEV_NAME, (dev & MINORMASK));

	return (zv);

	out_kmem:
	kmem_free(zso, sizeof (struct zvol_state_os));
	kmem_free(zv, sizeof (zvol_state_t));
	return (NULL);
	}

	/*
	* Cleanup then free a zvol_state_t which was created by zvol_alloc().
	* At this time, the structure is not opened by anyone, is taken off
	* the zvol_state_list, and has its private data set to NULL.
	* The zvol_state_lock is dropped.
	*
	* This function may take many milliseconds to complete (e.g. we've seen
	* it take over 256ms), due to the calls to "blk_cleanup_queue" and
	* "del_gendisk". Thus, consumers need to be careful to account for this
	* latency when calling this function.
	*/
	void
	zvol_os_free(zvol_state_t *zv)
	{

	ASSERT(!RW_LOCK_HELD(&zv->zv_suspend_lock));
	ASSERT(!MUTEX_HELD(&zv->zv_state_lock));
	ASSERT0(zv->zv_open_count);
	ASSERT3P(zv->zv_zso->zvo_disk->private_data, ==, NULL);

	rw_destroy(&zv->zv_suspend_lock);
	zfs_rangelock_fini(&zv->zv_rangelock);

	del_gendisk(zv->zv_zso->zvo_disk);
	#if defined(HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS) && \
	(defined(HAVE_BLK_ALLOC_DISK) \|\| defined(HAVE_BLK_ALLOC_DISK_2ARG))
	#if defined(HAVE_BLK_CLEANUP_DISK)
	blk_cleanup_disk(zv->zv_zso->zvo_disk);
	#else
	put_disk(zv->zv_zso->zvo_disk);
	#endif
	#else
	blk_cleanup_queue(zv->zv_zso->zvo_queue);
	put_disk(zv->zv_zso->zvo_disk);
	#endif

	#ifdef HAVE_BLK_MQ
	if (zv->zv_zso->use_blk_mq)
	blk_mq_free_tag_set(&zv->zv_zso->tag_set);
	#endif

	ida_simple_remove(&zvol_ida,
	MINOR(zv->zv_zso->zvo_dev) >> ZVOL_MINOR_BITS);

	mutex_destroy(&zv->zv_state_lock);
	dataset_kstats_destroy(&zv->zv_kstat);

	kmem_free(zv->zv_zso, sizeof (struct zvol_state_os));
	kmem_free(zv, sizeof (zvol_state_t));
	}

	void
	zvol_wait_close(zvol_state_t *zv)
	{
	}

	/*
	* Create a block device minor node and setup the linkage between it
	* and the specified volume. Once this function returns the block
	* device is live and ready for use.
	*/
	int
	zvol_os_create_minor(const char *name)
	{
	zvol_state_t *zv;
	objset_t *os;
	dmu_object_info_t *doi;
	uint64_t volsize;
	uint64_t len;
	unsigned minor = 0;
	int error = 0;
	int idx;
	uint64_t hash = zvol_name_hash(name);
	uint64_t volthreading;
	bool replayed_zil = B_FALSE;

	if (zvol_inhibit_dev)
	return (0);

	idx = ida_simple_get(&zvol_ida, 0, 0, kmem_flags_convert(KM_SLEEP));
	if (idx < 0)
	return (SET_ERROR(-idx));
	minor = idx << ZVOL_MINOR_BITS;
	if (MINOR(minor) != minor) {
	/* too many partitions can cause an overflow */
	zfs_dbgmsg("zvol: create minor overflow: %s, minor %u/%u",
	name, minor, MINOR(minor));
	ida_simple_remove(&zvol_ida, idx);
	return (SET_ERROR(EINVAL));
	}

	zv = zvol_find_by_name_hash(name, hash, RW_NONE);
	if (zv) {
	ASSERT(MUTEX_HELD(&zv->zv_state_lock));
	mutex_exit(&zv->zv_state_lock);
	ida_simple_remove(&zvol_ida, idx);
	return (SET_ERROR(EEXIST));
	}

	doi = kmem_alloc(sizeof (dmu_object_info_t), KM_SLEEP);

	error = dmu_objset_own(name, DMU_OST_ZVOL, B_TRUE, B_TRUE, FTAG, &os);
	if (error)
	goto out_doi;

	error = dmu_object_info(os, ZVOL_OBJ, doi);
	if (error)
	goto out_dmu_objset_disown;

	error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
	if (error)
	goto out_dmu_objset_disown;

	zv = zvol_alloc(MKDEV(zvol_major, minor), name);
	if (zv == NULL) {
	error = SET_ERROR(EAGAIN);
	goto out_dmu_objset_disown;
	}
	zv->zv_hash = hash;

	if (dmu_objset_is_snapshot(os))
	zv->zv_flags \|= ZVOL_RDONLY;

	zv->zv_volblocksize = doi->doi_data_block_size;
	zv->zv_volsize = volsize;
	zv->zv_objset = os;

	/* Default */
	zv->zv_threading = B_TRUE;
	if (dsl_prop_get_integer(name, "volthreading", &volthreading, NULL)
	== 0)
	zv->zv_threading = volthreading;

	set_capacity(zv->zv_zso->zvo_disk, zv->zv_volsize >> 9);

	blk_queue_max_hw_sectors(zv->zv_zso->zvo_queue,
	(DMU_MAX_ACCESS / 4) >> 9);

	if (zv->zv_zso->use_blk_mq) {
	/*
	* IO requests can be really big (1MB). When an IO request
	* comes in, it is passed off to zvol_read() or zvol_write()
	* in a new thread, where it is chunked up into 'volblocksize'
	* sized pieces and processed. So for example, if the request
	* is a 1MB write and your volblocksize is 128k, one zvol_write
	* thread will take that request and sequentially do ten 128k
	* IOs. This is due to the fact that the thread needs to lock
	* each volblocksize sized block. So you might be wondering:
	* "instead of passing the whole 1MB request to one thread,
	* why not pass ten individual 128k chunks to ten threads and
	* process the whole write in parallel?" The short answer is
	* that there's a sweet spot number of chunks that balances
	* the greater parallelism with the added overhead of more
	* threads. The sweet spot can be different depending on if you
	* have a read or write heavy workload. Writes typically want
	* high chunk counts while reads typically want lower ones. On
	* a test pool with 6 NVMe drives in a 3x 2-disk mirror
	* configuration, with volblocksize=8k, the sweet spot for good
	* sequential reads and writes was at 8 chunks.
	*/

	/*
	* Below we tell the kernel how big we want our requests
	* to be. You would think that blk_queue_io_opt() would be
	* used to do this since it is used to "set optimal request
	* size for the queue", but that doesn't seem to do
	* anything - the kernel still gives you huge requests
	* with tons of little PAGE_SIZE segments contained within it.
	*
	* Knowing that the kernel will just give you PAGE_SIZE segments
	* no matter what, you can say "ok, I want PAGE_SIZE byte
	* segments, and I want 'N' of them per request", where N is
	* the correct number of segments for the volblocksize and
	* number of chunks you want.
	*/
	#ifdef HAVE_BLK_MQ
	if (zvol_blk_mq_blocks_per_thread != 0) {
	unsigned int chunks;
	chunks = MIN(zvol_blk_mq_blocks_per_thread, UINT16_MAX);

	blk_queue_max_segment_size(zv->zv_zso->zvo_queue,
	PAGE_SIZE);
	blk_queue_max_segments(zv->zv_zso->zvo_queue,
	(zv->zv_volblocksize * chunks) / PAGE_SIZE);
	} else {
	/*
	* Special case: zvol_blk_mq_blocks_per_thread = 0
	* Max everything out.
	*/
	blk_queue_max_segments(zv->zv_zso->zvo_queue,
	UINT16_MAX);
	blk_queue_max_segment_size(zv->zv_zso->zvo_queue,
	UINT_MAX);
	}
	#endif
	} else {
	blk_queue_max_segments(zv->zv_zso->zvo_queue, UINT16_MAX);
	blk_queue_max_segment_size(zv->zv_zso->zvo_queue, UINT_MAX);
	}

	blk_queue_physical_block_size(zv->zv_zso->zvo_queue,
	zv->zv_volblocksize);
	blk_queue_io_opt(zv->zv_zso->zvo_queue, zv->zv_volblocksize);
	blk_queue_max_discard_sectors(zv->zv_zso->zvo_queue,
	(zvol_max_discard_blocks * zv->zv_volblocksize) >> 9);
	blk_queue_discard_granularity(zv->zv_zso->zvo_queue,
	zv->zv_volblocksize);
	#ifdef QUEUE_FLAG_DISCARD
	blk_queue_flag_set(QUEUE_FLAG_DISCARD, zv->zv_zso->zvo_queue);
	#endif
	#ifdef QUEUE_FLAG_NONROT
	blk_queue_flag_set(QUEUE_FLAG_NONROT, zv->zv_zso->zvo_queue);
	#endif
	#ifdef QUEUE_FLAG_ADD_RANDOM
	blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, zv->zv_zso->zvo_queue);
	#endif
	/* This flag was introduced in kernel version 4.12. */
	#ifdef QUEUE_FLAG_SCSI_PASSTHROUGH
	blk_queue_flag_set(QUEUE_FLAG_SCSI_PASSTHROUGH, zv->zv_zso->zvo_queue);
	#endif

	ASSERT3P(zv->zv_kstat.dk_kstats, ==, NULL);
	error = dataset_kstats_create(&zv->zv_kstat, zv->zv_objset);
	if (error)
	goto out_dmu_objset_disown;
	ASSERT3P(zv->zv_zilog, ==, NULL);
	zv->zv_zilog = zil_open(os, zvol_get_data, &zv->zv_kstat.dk_zil_sums);
	if (spa_writeable(dmu_objset_spa(os))) {
	if (zil_replay_disable)
	replayed_zil = zil_destroy(zv->zv_zilog, B_FALSE);
	else
	replayed_zil = zil_replay(os, zv, zvol_replay_vector);
	}
	if (replayed_zil)
	zil_close(zv->zv_zilog);
	zv->zv_zilog = NULL;

	/*
	* When udev detects the addition of the device it will immediately
	* invoke blkid(8) to determine the type of content on the device.
	* Prefetching the blocks commonly scanned by blkid(8) will speed
	* up this process.
	*/
	len = MIN(zvol_prefetch_bytes, SPA_MAXBLOCKSIZE);
	if (len > 0) {
	dmu_prefetch(os, ZVOL_OBJ, 0, 0, len, ZIO_PRIORITY_SYNC_READ);
	dmu_prefetch(os, ZVOL_OBJ, 0, volsize - len, len,
	ZIO_PRIORITY_SYNC_READ);
	}

	zv->zv_objset = NULL;
	out_dmu_objset_disown:
	dmu_objset_disown(os, B_TRUE, FTAG);
	out_doi:
	kmem_free(doi, sizeof (dmu_object_info_t));

	/*
	* Keep in mind that once add_disk() is called, the zvol is
	* announced to the world, and zvol_open()/zvol_release() can
	* be called at any time. Incidentally, add_disk() itself calls
	* zvol_open()->zvol_first_open() and zvol_release()->zvol_last_close()
	* directly as well.
	*/
	if (error == 0) {
	rw_enter(&zvol_state_lock, RW_WRITER);
	zvol_insert(zv);
	rw_exit(&zvol_state_lock);
	#ifdef HAVE_ADD_DISK_RET
	error = add_disk(zv->zv_zso->zvo_disk);
	#else
	add_disk(zv->zv_zso->zvo_disk);
	#endif
	} else {
	ida_simple_remove(&zvol_ida, idx);
	}

	return (error);
	}

	void
	zvol_os_rename_minor(zvol_state_t zv, const char newname)
	{
	int readonly = get_disk_ro(zv->zv_zso->zvo_disk);

	ASSERT(RW_LOCK_HELD(&zvol_state_lock));
	ASSERT(MUTEX_HELD(&zv->zv_state_lock));

	strlcpy(zv->zv_name, newname, sizeof (zv->zv_name));

	/* move to new hashtable entry */
	zv->zv_hash = zvol_name_hash(newname);
	hlist_del(&zv->zv_hlink);
	hlist_add_head(&zv->zv_hlink, ZVOL_HT_HEAD(zv->zv_hash));

	/*
	* The block device's read-only state is briefly changed causing
	* a KOBJ_CHANGE uevent to be issued. This ensures udev detects
	* the name change and fixes the symlinks. This does not change
	* ZVOL_RDONLY in zv->zv_flags so the actual read-only state never
	* changes. This would normally be done using kobject_uevent() but
	* that is a GPL-only symbol which is why we need this workaround.
	*/
	set_disk_ro(zv->zv_zso->zvo_disk, !readonly);
	set_disk_ro(zv->zv_zso->zvo_disk, readonly);

	dataset_kstats_rename(&zv->zv_kstat, newname);
	}

	void
	zvol_os_set_disk_ro(zvol_state_t *zv, int flags)
	{

	set_disk_ro(zv->zv_zso->zvo_disk, flags);
	}

	void
	zvol_os_set_capacity(zvol_state_t *zv, uint64_t capacity)
	{

	set_capacity(zv->zv_zso->zvo_disk, capacity);
	}

	int
	zvol_init(void)
	{
	int error;

	/*
	* zvol_threads is the module param the user passes in.
	*
	* zvol_actual_threads is what we use internally, since the user can
	* pass zvol_thread = 0 to mean "use all the CPUs" (the default).
	*/
	static unsigned int zvol_actual_threads;

	if (zvol_threads == 0) {
	/*
	* See dde9380a1 for why 32 was chosen here. This should
	* probably be refined to be some multiple of the number
	* of CPUs.
	*/
	zvol_actual_threads = MAX(num_online_cpus(), 32);
	} else {
	zvol_actual_threads = MIN(MAX(zvol_threads, 1), 1024);
	}

	/*
	* Use atleast 32 zvol_threads but for many core system,
	* prefer 6 threads per taskq, but no more taskqs
	* than threads in them on large systems.
	*
	* taskq total
	* cpus taskqs threads threads
	* ------- ------- ------- -------
	* 1 1 32 32
	* 2 1 32 32
	* 4 1 32 32
	* 8 2 16 32
	* 16 3 11 33
	* 32 5 7 35
	* 64 8 8 64
	* 128 11 12 132
	* 256 16 16 256
	*/
	zv_taskq_t *ztqs = &zvol_taskqs;
	uint_t num_tqs = MIN(num_online_cpus(), zvol_num_taskqs);
	if (num_tqs == 0) {
	num_tqs = 1 + num_online_cpus() / 6;
	while (num_tqs * num_tqs > zvol_actual_threads)
	num_tqs--;
	}
	uint_t per_tq_thread = zvol_actual_threads / num_tqs;
	if (per_tq_thread * num_tqs < zvol_actual_threads)
	per_tq_thread++;
	ztqs->tqs_cnt = num_tqs;
	ztqs->tqs_taskq = kmem_alloc(num_tqs * sizeof (taskq_t *), KM_SLEEP);
	error = register_blkdev(zvol_major, ZVOL_DRIVER);
	if (error) {
	kmem_free(ztqs->tqs_taskq, ztqs->tqs_cnt * sizeof (taskq_t *));
	ztqs->tqs_taskq = NULL;
	printk(KERN_INFO "ZFS: register_blkdev() failed %d\n", error);
	return (error);
	}

	#ifdef HAVE_BLK_MQ
	if (zvol_blk_mq_queue_depth == 0) {
	zvol_actual_blk_mq_queue_depth = BLKDEV_DEFAULT_RQ;
	} else {
	zvol_actual_blk_mq_queue_depth =
	MAX(zvol_blk_mq_queue_depth, BLKDEV_MIN_RQ);
	}

	if (zvol_blk_mq_threads == 0) {
	zvol_blk_mq_actual_threads = num_online_cpus();
	} else {
	zvol_blk_mq_actual_threads = MIN(MAX(zvol_blk_mq_threads, 1),
	1024);
	}
	#endif
	for (uint_t i = 0; i < num_tqs; i++) {
	char name[32];
	(void) snprintf(name, sizeof (name), "%s_tq-%u",
	ZVOL_DRIVER, i);
	ztqs->tqs_taskq[i] = taskq_create(name, per_tq_thread,
	maxclsyspri, per_tq_thread, INT_MAX,
	TASKQ_PREPOPULATE \| TASKQ_DYNAMIC);
	if (ztqs->tqs_taskq[i] == NULL) {
	for (int j = i - 1; j >= 0; j--)
	taskq_destroy(ztqs->tqs_taskq[j]);
	unregister_blkdev(zvol_major, ZVOL_DRIVER);
	kmem_free(ztqs->tqs_taskq, ztqs->tqs_cnt *
	sizeof (taskq_t *));
	ztqs->tqs_taskq = NULL;
	return (-ENOMEM);
	}
	}

	zvol_init_impl();
	ida_init(&zvol_ida);
	return (0);
	}

	void
	zvol_fini(void)
	{
	zv_taskq_t *ztqs = &zvol_taskqs;
	zvol_fini_impl();
	unregister_blkdev(zvol_major, ZVOL_DRIVER);

	if (ztqs->tqs_taskq == NULL) {
	ASSERT3U(ztqs->tqs_cnt, ==, 0);
	} else {
	for (uint_t i = 0; i < ztqs->tqs_cnt; i++) {
	ASSERT3P(ztqs->tqs_taskq[i], !=, NULL);
	taskq_destroy(ztqs->tqs_taskq[i]);
	}
	kmem_free(ztqs->tqs_taskq, ztqs->tqs_cnt *
	sizeof (taskq_t *));
	ztqs->tqs_taskq = NULL;
	}

	ida_destroy(&zvol_ida);
	}

	/* BEGIN CSTYLED */
	module_param(zvol_inhibit_dev, uint, 0644);
	MODULE_PARM_DESC(zvol_inhibit_dev, "Do not create zvol device nodes");

	module_param(zvol_major, uint, 0444);
	MODULE_PARM_DESC(zvol_major, "Major number for zvol device");

	module_param(zvol_threads, uint, 0444);
	MODULE_PARM_DESC(zvol_threads, "Number of threads to handle I/O requests. Set"
	"to 0 to use all active CPUs");

	module_param(zvol_request_sync, uint, 0644);
	MODULE_PARM_DESC(zvol_request_sync, "Synchronously handle bio requests");

	module_param(zvol_max_discard_blocks, ulong, 0444);
	MODULE_PARM_DESC(zvol_max_discard_blocks, "Max number of blocks to discard");

	module_param(zvol_num_taskqs, uint, 0444);
	MODULE_PARM_DESC(zvol_num_taskqs, "Number of zvol taskqs");

	module_param(zvol_prefetch_bytes, uint, 0644);
	MODULE_PARM_DESC(zvol_prefetch_bytes, "Prefetch N bytes at zvol start+end");

	module_param(zvol_volmode, uint, 0644);
	MODULE_PARM_DESC(zvol_volmode, "Default volmode property value");

	#ifdef HAVE_BLK_MQ
	module_param(zvol_blk_mq_queue_depth, uint, 0644);
	MODULE_PARM_DESC(zvol_blk_mq_queue_depth, "Default blk-mq queue depth");

	module_param(zvol_use_blk_mq, uint, 0644);
	MODULE_PARM_DESC(zvol_use_blk_mq, "Use the blk-mq API for zvols");

	module_param(zvol_blk_mq_blocks_per_thread, uint, 0644);
	MODULE_PARM_DESC(zvol_blk_mq_blocks_per_thread,
	"Process volblocksize blocks per thread");
	#endif

	#ifndef HAVE_BLKDEV_GET_ERESTARTSYS
	module_param(zvol_open_timeout_ms, uint, 0644);
	MODULE_PARM_DESC(zvol_open_timeout_ms, "Timeout for ZVOL open retries");
	#endif

	/* END CSTYLED */
	diff --git a/sys/contrib/openzfs/module/zcommon/zfs_fletcher.c b/sys/contrib/openzfs/module/zcommon/zfs_fletcher.c
	index 619ddef0243a..74b8c2a475a1 100644
	--- a/sys/contrib/openzfs/module/zcommon/zfs_fletcher.c
	+++ b/sys/contrib/openzfs/module/zcommon/zfs_fletcher.c
	@@ -1,1011 +1,1013 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright 2009 Sun Microsystems, Inc. All rights reserved.
	* Use is subject to license terms.
	* Copyright (C) 2016 Gvozden Nešković. All rights reserved.
	*/
	/*
	* Copyright 2013 Saso Kiselkov. All rights reserved.
	*/

	/*
	* Copyright (c) 2016 by Delphix. All rights reserved.
	*/

	/*
	* Fletcher Checksums
	* ------------------
	*
	* ZFS's 2nd and 4th order Fletcher checksums are defined by the following
	* recurrence relations:
	*
	* a = a + f
	* i i-1 i-1
	*
	* b = b + a
	* i i-1 i
	*
	* c = c + b (fletcher-4 only)
	* i i-1 i
	*
	* d = d + c (fletcher-4 only)
	* i i-1 i
	*
	* Where
	* a_0 = b_0 = c_0 = d_0 = 0
	* and
	* f_0 .. f_(n-1) are the input data.
	*
	* Using standard techniques, these translate into the following series:
	*
	* __n_ __n_
	* \ \| \ \|
	* a = > f b = > i * f
	* n /___\| n - i n /___\| n - i
	* i = 1 i = 1
	*
	*
	* __n_ __n_
	* \ \| i(i+1) \ \| i(i+1)*(i+2)
	* c = > ------- f d = > ------------- f
	* n /___\| 2 n - i n /___\| 6 n - i
	* i = 1 i = 1
	*
	* For fletcher-2, the f_is are 64-bit, and [ab]_i are 64-bit accumulators.
	* Since the additions are done mod (2^64), errors in the high bits may not
	* be noticed. For this reason, fletcher-2 is deprecated.
	*
	* For fletcher-4, the f_is are 32-bit, and [abcd]_i are 64-bit accumulators.
	* A conservative estimate of how big the buffer can get before we overflow
	* can be estimated using f_i = 0xffffffff for all i:
	*
	* % bc
	* f=2^32-1;d=0; for (i = 1; d<2^64; i++) { d += fi(i+1)(i+2)/6 }; (i-1)4
	* 2264
	* quit
	* %
	*
	* So blocks of up to 2k will not overflow. Our largest block size is
	* 128k, which has 32k 4-byte words, so we can compute the largest possible
	* accumulators, then divide by 2^64 to figure the max amount of overflow:
	*
	* % bc
	* a=b=c=d=0; f=2^32-1; for (i=1; i<=32*1024; i++) { a+=f; b+=a; c+=b; d+=c }
	* a/2^64;b/2^64;c/2^64;d/2^64
	* 0
	* 0
	* 1365
	* 11186858
	* quit
	* %
	*
	* So a and b cannot overflow. To make sure each bit of input has some
	* effect on the contents of c and d, we can look at what the factors of
	* the coefficients in the equations for c_n and d_n are. The number of 2s
	* in the factors determines the lowest set bit in the multiplier. Running
	* through the cases for n*(n+1)/2 reveals that the highest power of 2 is
	* 2^14, and for n(n+1)(n+2)/6 it is 2^15. So while some data may overflow
	* the 64-bit accumulators, every bit of every f_i effects every accumulator,
	* even for 128k blocks.
	*
	* If we wanted to make a stronger version of fletcher4 (fletcher4c?),
	* we could do our calculations mod (2^32 - 1) by adding in the carries
	* periodically, and store the number of carries in the top 32-bits.
	*
	* --------------------
	* Checksum Performance
	* --------------------
	*
	* There are two interesting components to checksum performance: cached and
	* uncached performance. With cached data, fletcher-2 is about four times
	* faster than fletcher-4. With uncached data, the performance difference is
	* negligible, since the cost of a cache fill dominates the processing time.
	* Even though fletcher-4 is slower than fletcher-2, it is still a pretty
	* efficient pass over the data.
	*
	* In normal operation, the data which is being checksummed is in a buffer
	* which has been filled either by:
	*
	* 1. a compression step, which will be mostly cached, or
	* 2. a memcpy() or copyin(), which will be uncached
	* (because the copy is cache-bypassing).
	*
	* For both cached and uncached data, both fletcher checksums are much faster
	* than sha-256, and slower than 'off', which doesn't touch the data at all.
	*/

	#include <sys/types.h>
	#include <sys/sysmacros.h>
	#include <sys/byteorder.h>
	#include <sys/simd.h>
	#include <sys/spa.h>
	#include <sys/zio_checksum.h>
	#include <sys/zfs_context.h>
	#include <zfs_fletcher.h>

	#define FLETCHER_MIN_SIMD_SIZE 64

	static void fletcher_4_scalar_init(fletcher_4_ctx_t *ctx);
	static void fletcher_4_scalar_fini(fletcher_4_ctx_t ctx, zio_cksum_t zcp);
	static void fletcher_4_scalar_native(fletcher_4_ctx_t *ctx,
	const void *buf, uint64_t size);
	static void fletcher_4_scalar_byteswap(fletcher_4_ctx_t *ctx,
	const void *buf, uint64_t size);
	static boolean_t fletcher_4_scalar_valid(void);

	static const fletcher_4_ops_t fletcher_4_scalar_ops = {
	.init_native = fletcher_4_scalar_init,
	.fini_native = fletcher_4_scalar_fini,
	.compute_native = fletcher_4_scalar_native,
	.init_byteswap = fletcher_4_scalar_init,
	.fini_byteswap = fletcher_4_scalar_fini,
	.compute_byteswap = fletcher_4_scalar_byteswap,
	.valid = fletcher_4_scalar_valid,
	.uses_fpu = B_FALSE,
	.name = "scalar"
	};

	static fletcher_4_ops_t fletcher_4_fastest_impl = {
	.name = "fastest",
	.valid = fletcher_4_scalar_valid
	};

	static const fletcher_4_ops_t *fletcher_4_impls[] = {
	&fletcher_4_scalar_ops,
	&fletcher_4_superscalar_ops,
	&fletcher_4_superscalar4_ops,
	#if defined(HAVE_SSE2)
	&fletcher_4_sse2_ops,
	#endif
	#if defined(HAVE_SSE2) && defined(HAVE_SSSE3)
	&fletcher_4_ssse3_ops,
	#endif
	#if defined(HAVE_AVX) && defined(HAVE_AVX2)
	&fletcher_4_avx2_ops,
	#endif
	#if defined(__x86_64) && defined(HAVE_AVX512F)
	&fletcher_4_avx512f_ops,
	#endif
	#if defined(__x86_64) && defined(HAVE_AVX512BW)
	&fletcher_4_avx512bw_ops,
	#endif
	#if defined(__aarch64__) && !defined(__FreeBSD__)
	&fletcher_4_aarch64_neon_ops,
	#endif
	};

	/* Hold all supported implementations */
	static uint32_t fletcher_4_supp_impls_cnt = 0;
	static fletcher_4_ops_t *fletcher_4_supp_impls[ARRAY_SIZE(fletcher_4_impls)];

	/* Select fletcher4 implementation */
	#define IMPL_FASTEST (UINT32_MAX)
	#define IMPL_CYCLE (UINT32_MAX - 1)
	#define IMPL_SCALAR (0)

	static uint32_t fletcher_4_impl_chosen = IMPL_FASTEST;

	#define IMPL_READ(i) ((volatile uint32_t ) &(i))

	static struct fletcher_4_impl_selector {
	const char *fis_name;
	uint32_t fis_sel;
	} fletcher_4_impl_selectors[] = {
	{ "cycle", IMPL_CYCLE },
	{ "fastest", IMPL_FASTEST },
	{ "scalar", IMPL_SCALAR }
	};

	#if defined(_KERNEL)
	static kstat_t *fletcher_4_kstat;

	static struct fletcher_4_kstat {
	uint64_t native;
	uint64_t byteswap;
	} fletcher_4_stat_data[ARRAY_SIZE(fletcher_4_impls) + 1];
	#endif

	/* Indicate that benchmark has been completed */
	static boolean_t fletcher_4_initialized = B_FALSE;

	void
	fletcher_init(zio_cksum_t *zcp)
	{
	ZIO_SET_CHECKSUM(zcp, 0, 0, 0, 0);
	}

	int
	fletcher_2_incremental_native(void buf, size_t size, void data)
	{
	zio_cksum_t *zcp = data;

	const uint64_t *ip = buf;
	const uint64_t *ipend = ip + (size / sizeof (uint64_t));
	uint64_t a0, b0, a1, b1;

	a0 = zcp->zc_word[0];
	a1 = zcp->zc_word[1];
	b0 = zcp->zc_word[2];
	b1 = zcp->zc_word[3];

	for (; ip < ipend; ip += 2) {
	a0 += ip[0];
	a1 += ip[1];
	b0 += a0;
	b1 += a1;
	}

	ZIO_SET_CHECKSUM(zcp, a0, a1, b0, b1);
	return (0);
	}

	void
	fletcher_2_native(const void *buf, uint64_t size,
	const void ctx_template, zio_cksum_t zcp)
	{
	(void) ctx_template;
	fletcher_init(zcp);
	(void) fletcher_2_incremental_native((void *) buf, size, zcp);
	}

	int
	fletcher_2_incremental_byteswap(void buf, size_t size, void data)
	{
	zio_cksum_t *zcp = data;

	const uint64_t *ip = buf;
	const uint64_t *ipend = ip + (size / sizeof (uint64_t));
	uint64_t a0, b0, a1, b1;

	a0 = zcp->zc_word[0];
	a1 = zcp->zc_word[1];
	b0 = zcp->zc_word[2];
	b1 = zcp->zc_word[3];

	for (; ip < ipend; ip += 2) {
	a0 += BSWAP_64(ip[0]);
	a1 += BSWAP_64(ip[1]);
	b0 += a0;
	b1 += a1;
	}

	ZIO_SET_CHECKSUM(zcp, a0, a1, b0, b1);
	return (0);
	}

	void
	fletcher_2_byteswap(const void *buf, uint64_t size,
	const void ctx_template, zio_cksum_t zcp)
	{
	(void) ctx_template;
	fletcher_init(zcp);
	(void) fletcher_2_incremental_byteswap((void *) buf, size, zcp);
	}

	static void
	fletcher_4_scalar_init(fletcher_4_ctx_t *ctx)
	{
	ZIO_SET_CHECKSUM(&ctx->scalar, 0, 0, 0, 0);
	}

	static void
	fletcher_4_scalar_fini(fletcher_4_ctx_t ctx, zio_cksum_t zcp)
	{
	memcpy(zcp, &ctx->scalar, sizeof (zio_cksum_t));
	}

	static void
	fletcher_4_scalar_native(fletcher_4_ctx_t ctx, const void buf,
	uint64_t size)
	{
	const uint32_t *ip = buf;
	const uint32_t *ipend = ip + (size / sizeof (uint32_t));
	uint64_t a, b, c, d;

	a = ctx->scalar.zc_word[0];
	b = ctx->scalar.zc_word[1];
	c = ctx->scalar.zc_word[2];
	d = ctx->scalar.zc_word[3];

	for (; ip < ipend; ip++) {
	a += ip[0];
	b += a;
	c += b;
	d += c;
	}

	ZIO_SET_CHECKSUM(&ctx->scalar, a, b, c, d);
	}

	static void
	fletcher_4_scalar_byteswap(fletcher_4_ctx_t ctx, const void buf,
	uint64_t size)
	{
	const uint32_t *ip = buf;
	const uint32_t *ipend = ip + (size / sizeof (uint32_t));
	uint64_t a, b, c, d;

	a = ctx->scalar.zc_word[0];
	b = ctx->scalar.zc_word[1];
	c = ctx->scalar.zc_word[2];
	d = ctx->scalar.zc_word[3];

	for (; ip < ipend; ip++) {
	a += BSWAP_32(ip[0]);
	b += a;
	c += b;
	d += c;
	}

	ZIO_SET_CHECKSUM(&ctx->scalar, a, b, c, d);
	}

	static boolean_t
	fletcher_4_scalar_valid(void)
	{
	return (B_TRUE);
	}

	int
	fletcher_4_impl_set(const char *val)
	{
	int err = -EINVAL;
	uint32_t impl = IMPL_READ(fletcher_4_impl_chosen);
	size_t i, val_len;

	val_len = strlen(val);
	while ((val_len > 0) && !!isspace(val[val_len-1])) /* trim '\n' */
	val_len--;

	/* check mandatory implementations */
	for (i = 0; i < ARRAY_SIZE(fletcher_4_impl_selectors); i++) {
	const char *name = fletcher_4_impl_selectors[i].fis_name;

	if (val_len == strlen(name) &&
	strncmp(val, name, val_len) == 0) {
	impl = fletcher_4_impl_selectors[i].fis_sel;
	err = 0;
	break;
	}
	}

	if (err != 0 && fletcher_4_initialized) {
	/* check all supported implementations */
	for (i = 0; i < fletcher_4_supp_impls_cnt; i++) {
	const char *name = fletcher_4_supp_impls[i]->name;

	if (val_len == strlen(name) &&
	strncmp(val, name, val_len) == 0) {
	impl = i;
	err = 0;
	break;
	}
	}
	}

	if (err == 0) {
	atomic_swap_32(&fletcher_4_impl_chosen, impl);
	membar_producer();
	}

	return (err);
	}

	/*
	* Returns the Fletcher 4 operations for checksums. When a SIMD
	* implementation is not allowed in the current context, then fallback
	* to the fastest generic implementation.
	*/
	static inline const fletcher_4_ops_t *
	fletcher_4_impl_get(void)
	{
	if (!kfpu_allowed())
	return (&fletcher_4_superscalar4_ops);

	const fletcher_4_ops_t *ops = NULL;
	uint32_t impl = IMPL_READ(fletcher_4_impl_chosen);

	switch (impl) {
	case IMPL_FASTEST:
	ASSERT(fletcher_4_initialized);
	ops = &fletcher_4_fastest_impl;
	break;
	case IMPL_CYCLE:
	/* Cycle through supported implementations */
	ASSERT(fletcher_4_initialized);
	ASSERT3U(fletcher_4_supp_impls_cnt, >, 0);
	static uint32_t cycle_count = 0;
	uint32_t idx = (++cycle_count) % fletcher_4_supp_impls_cnt;
	ops = fletcher_4_supp_impls[idx];
	break;
	default:
	ASSERT3U(fletcher_4_supp_impls_cnt, >, 0);
	ASSERT3U(impl, <, fletcher_4_supp_impls_cnt);
	ops = fletcher_4_supp_impls[impl];
	break;
	}

	ASSERT3P(ops, !=, NULL);

	return (ops);
	}

	static inline void
	fletcher_4_native_impl(const void buf, uint64_t size, zio_cksum_t zcp)
	{
	fletcher_4_ctx_t ctx;
	const fletcher_4_ops_t *ops = fletcher_4_impl_get();

	if (ops->uses_fpu == B_TRUE) {
	kfpu_begin();
	}
	ops->init_native(&ctx);
	ops->compute_native(&ctx, buf, size);
	ops->fini_native(&ctx, zcp);
	if (ops->uses_fpu == B_TRUE) {
	kfpu_end();
	}
	}

	void
	fletcher_4_native(const void *buf, uint64_t size,
	const void ctx_template, zio_cksum_t zcp)
	{
	(void) ctx_template;
	- const uint64_t p2size = P2ALIGN(size, FLETCHER_MIN_SIMD_SIZE);
	+ const uint64_t p2size = P2ALIGN_TYPED(size, FLETCHER_MIN_SIMD_SIZE,
	+ uint64_t);

	ASSERT(IS_P2ALIGNED(size, sizeof (uint32_t)));

	if (size == 0 \|\| p2size == 0) {
	ZIO_SET_CHECKSUM(zcp, 0, 0, 0, 0);

	if (size > 0)
	fletcher_4_scalar_native((fletcher_4_ctx_t *)zcp,
	buf, size);
	} else {
	fletcher_4_native_impl(buf, p2size, zcp);

	if (p2size < size)
	fletcher_4_scalar_native((fletcher_4_ctx_t *)zcp,
	(char *)buf + p2size, size - p2size);
	}
	}

	void
	fletcher_4_native_varsize(const void buf, uint64_t size, zio_cksum_t zcp)
	{
	ZIO_SET_CHECKSUM(zcp, 0, 0, 0, 0);
	fletcher_4_scalar_native((fletcher_4_ctx_t *)zcp, buf, size);
	}

	static inline void
	fletcher_4_byteswap_impl(const void buf, uint64_t size, zio_cksum_t zcp)
	{
	fletcher_4_ctx_t ctx;
	const fletcher_4_ops_t *ops = fletcher_4_impl_get();

	if (ops->uses_fpu == B_TRUE) {
	kfpu_begin();
	}
	ops->init_byteswap(&ctx);
	ops->compute_byteswap(&ctx, buf, size);
	ops->fini_byteswap(&ctx, zcp);
	if (ops->uses_fpu == B_TRUE) {
	kfpu_end();
	}
	}

	void
	fletcher_4_byteswap(const void *buf, uint64_t size,
	const void ctx_template, zio_cksum_t zcp)
	{
	(void) ctx_template;
	- const uint64_t p2size = P2ALIGN(size, FLETCHER_MIN_SIMD_SIZE);
	+ const uint64_t p2size = P2ALIGN_TYPED(size, FLETCHER_MIN_SIMD_SIZE,
	+ uint64_t);

	ASSERT(IS_P2ALIGNED(size, sizeof (uint32_t)));

	if (size == 0 \|\| p2size == 0) {
	ZIO_SET_CHECKSUM(zcp, 0, 0, 0, 0);

	if (size > 0)
	fletcher_4_scalar_byteswap((fletcher_4_ctx_t *)zcp,
	buf, size);
	} else {
	fletcher_4_byteswap_impl(buf, p2size, zcp);

	if (p2size < size)
	fletcher_4_scalar_byteswap((fletcher_4_ctx_t *)zcp,
	(char *)buf + p2size, size - p2size);
	}
	}

	/* Incremental Fletcher 4 */

	#define ZFS_FLETCHER_4_INC_MAX_SIZE (8ULL << 20)

	static inline void
	fletcher_4_incremental_combine(zio_cksum_t *zcp, const uint64_t size,
	const zio_cksum_t *nzcp)
	{
	const uint64_t c1 = size / sizeof (uint32_t);
	const uint64_t c2 = c1 * (c1 + 1) / 2;
	const uint64_t c3 = c2 * (c1 + 2) / 3;

	/*
	* Value of 'c3' overflows on buffer sizes close to 16MiB. For that
	* reason we split incremental fletcher4 computation of large buffers
	* to steps of (ZFS_FLETCHER_4_INC_MAX_SIZE) size.
	*/
	ASSERT3U(size, <=, ZFS_FLETCHER_4_INC_MAX_SIZE);

	zcp->zc_word[3] += nzcp->zc_word[3] + c1 * zcp->zc_word[2] +
	c2 * zcp->zc_word[1] + c3 * zcp->zc_word[0];
	zcp->zc_word[2] += nzcp->zc_word[2] + c1 * zcp->zc_word[1] +
	c2 * zcp->zc_word[0];
	zcp->zc_word[1] += nzcp->zc_word[1] + c1 * zcp->zc_word[0];
	zcp->zc_word[0] += nzcp->zc_word[0];
	}

	static inline void
	fletcher_4_incremental_impl(boolean_t native, const void *buf, uint64_t size,
	zio_cksum_t *zcp)
	{
	while (size > 0) {
	zio_cksum_t nzc;
	uint64_t len = MIN(size, ZFS_FLETCHER_4_INC_MAX_SIZE);

	if (native)
	fletcher_4_native(buf, len, NULL, &nzc);
	else
	fletcher_4_byteswap(buf, len, NULL, &nzc);

	fletcher_4_incremental_combine(zcp, len, &nzc);

	size -= len;
	buf += len;
	}
	}

	int
	fletcher_4_incremental_native(void buf, size_t size, void data)
	{
	zio_cksum_t *zcp = data;
	/* Use scalar impl to directly update cksum of small blocks */
	if (size < SPA_MINBLOCKSIZE)
	fletcher_4_scalar_native((fletcher_4_ctx_t *)zcp, buf, size);
	else
	fletcher_4_incremental_impl(B_TRUE, buf, size, zcp);
	return (0);
	}

	int
	fletcher_4_incremental_byteswap(void buf, size_t size, void data)
	{
	zio_cksum_t *zcp = data;
	/* Use scalar impl to directly update cksum of small blocks */
	if (size < SPA_MINBLOCKSIZE)
	fletcher_4_scalar_byteswap((fletcher_4_ctx_t *)zcp, buf, size);
	else
	fletcher_4_incremental_impl(B_FALSE, buf, size, zcp);
	return (0);
	}

	#if defined(_KERNEL)
	/*
	* Fletcher 4 kstats
	*/
	static int
	fletcher_4_kstat_headers(char *buf, size_t size)
	{
	ssize_t off = 0;

	off += snprintf(buf + off, size, "%-17s", "implementation");
	off += snprintf(buf + off, size - off, "%-15s", "native");
	(void) snprintf(buf + off, size - off, "%-15s\n", "byteswap");

	return (0);
	}

	static int
	fletcher_4_kstat_data(char buf, size_t size, void data)
	{
	struct fletcher_4_kstat *fastest_stat =
	&fletcher_4_stat_data[fletcher_4_supp_impls_cnt];
	struct fletcher_4_kstat curr_stat = (struct fletcher_4_kstat )data;
	ssize_t off = 0;

	if (curr_stat == fastest_stat) {
	off += snprintf(buf + off, size - off, "%-17s", "fastest");
	off += snprintf(buf + off, size - off, "%-15s",
	fletcher_4_supp_impls[fastest_stat->native]->name);
	(void) snprintf(buf + off, size - off, "%-15s\n",
	fletcher_4_supp_impls[fastest_stat->byteswap]->name);
	} else {
	ptrdiff_t id = curr_stat - fletcher_4_stat_data;

	off += snprintf(buf + off, size - off, "%-17s",
	fletcher_4_supp_impls[id]->name);
	off += snprintf(buf + off, size - off, "%-15llu",
	(u_longlong_t)curr_stat->native);
	(void) snprintf(buf + off, size - off, "%-15llu\n",
	(u_longlong_t)curr_stat->byteswap);
	}

	return (0);
	}

	static void *
	fletcher_4_kstat_addr(kstat_t *ksp, loff_t n)
	{
	if (n <= fletcher_4_supp_impls_cnt)
	ksp->ks_private = (void *) (fletcher_4_stat_data + n);
	else
	ksp->ks_private = NULL;

	return (ksp->ks_private);
	}
	#endif

	#define FLETCHER_4_FASTEST_FN_COPY(type, src) \
	{ \
	fletcher_4_fastest_impl.init_ ## type = src->init_ ## type; \
	fletcher_4_fastest_impl.fini_ ## type = src->fini_ ## type; \
	fletcher_4_fastest_impl.compute_ ## type = src->compute_ ## type; \
	fletcher_4_fastest_impl.uses_fpu = src->uses_fpu; \
	}

	#define FLETCHER_4_BENCH_NS (MSEC2NSEC(1)) /* 1ms */

	typedef void fletcher_checksum_func_t(const void , uint64_t, const void ,
	zio_cksum_t *);

	#if defined(_KERNEL)
	static void
	fletcher_4_benchmark_impl(boolean_t native, char *data, uint64_t data_size)
	{

	struct fletcher_4_kstat *fastest_stat =
	&fletcher_4_stat_data[fletcher_4_supp_impls_cnt];
	hrtime_t start;
	uint64_t run_bw, run_time_ns, best_run = 0;
	zio_cksum_t zc;
	uint32_t i, l, sel_save = IMPL_READ(fletcher_4_impl_chosen);

	fletcher_checksum_func_t *fletcher_4_test = native ?
	fletcher_4_native : fletcher_4_byteswap;

	for (i = 0; i < fletcher_4_supp_impls_cnt; i++) {
	struct fletcher_4_kstat *stat = &fletcher_4_stat_data[i];
	uint64_t run_count = 0;

	/* temporary set an implementation */
	fletcher_4_impl_chosen = i;

	kpreempt_disable();
	start = gethrtime();
	do {
	for (l = 0; l < 32; l++, run_count++)
	fletcher_4_test(data, data_size, NULL, &zc);

	run_time_ns = gethrtime() - start;
	} while (run_time_ns < FLETCHER_4_BENCH_NS);
	kpreempt_enable();

	run_bw = data_size * run_count * NANOSEC;
	run_bw /= run_time_ns; /* B/s */

	if (native)
	stat->native = run_bw;
	else
	stat->byteswap = run_bw;

	if (run_bw > best_run) {
	best_run = run_bw;

	if (native) {
	fastest_stat->native = i;
	FLETCHER_4_FASTEST_FN_COPY(native,
	fletcher_4_supp_impls[i]);
	} else {
	fastest_stat->byteswap = i;
	FLETCHER_4_FASTEST_FN_COPY(byteswap,
	fletcher_4_supp_impls[i]);
	}
	}
	}

	/* restore original selection */
	atomic_swap_32(&fletcher_4_impl_chosen, sel_save);
	}
	#endif /* _KERNEL */

	/*
	* Initialize and benchmark all supported implementations.
	*/
	static void
	fletcher_4_benchmark(void)
	{
	fletcher_4_ops_t *curr_impl;
	int i, c;

	/* Move supported implementations into fletcher_4_supp_impls */
	for (i = 0, c = 0; i < ARRAY_SIZE(fletcher_4_impls); i++) {
	curr_impl = (fletcher_4_ops_t *)fletcher_4_impls[i];

	if (curr_impl->valid && curr_impl->valid())
	fletcher_4_supp_impls[c++] = curr_impl;
	}
	membar_producer(); /* complete fletcher_4_supp_impls[] init */
	fletcher_4_supp_impls_cnt = c; /* number of supported impl */

	#if defined(_KERNEL)
	static const size_t data_size = 1 << SPA_OLD_MAXBLOCKSHIFT; /* 128kiB */
	char *databuf = vmem_alloc(data_size, KM_SLEEP);

	for (i = 0; i < data_size / sizeof (uint64_t); i++)
	((uint64_t )databuf)[i] = (uintptr_t)(databuf+i); / warm-up */

	fletcher_4_benchmark_impl(B_FALSE, databuf, data_size);
	fletcher_4_benchmark_impl(B_TRUE, databuf, data_size);

	vmem_free(databuf, data_size);
	#else
	/*
	* Skip the benchmark in user space to avoid impacting libzpool
	* consumers (zdb, zhack, zinject, ztest). The last implementation
	* is assumed to be the fastest and used by default.
	*/
	memcpy(&fletcher_4_fastest_impl,
	fletcher_4_supp_impls[fletcher_4_supp_impls_cnt - 1],
	sizeof (fletcher_4_fastest_impl));
	fletcher_4_fastest_impl.name = "fastest";
	membar_producer();
	#endif /* _KERNEL */
	}

	void
	fletcher_4_init(void)
	{
	/* Determine the fastest available implementation. */
	fletcher_4_benchmark();

	#if defined(_KERNEL)
	/* Install kstats for all implementations */
	fletcher_4_kstat = kstat_create("zfs", 0, "fletcher_4_bench", "misc",
	KSTAT_TYPE_RAW, 0, KSTAT_FLAG_VIRTUAL);
	if (fletcher_4_kstat != NULL) {
	fletcher_4_kstat->ks_data = NULL;
	fletcher_4_kstat->ks_ndata = UINT32_MAX;
	kstat_set_raw_ops(fletcher_4_kstat,
	fletcher_4_kstat_headers,
	fletcher_4_kstat_data,
	fletcher_4_kstat_addr);
	kstat_install(fletcher_4_kstat);
	}
	#endif

	/* Finish initialization */
	fletcher_4_initialized = B_TRUE;
	}

	void
	fletcher_4_fini(void)
	{
	#if defined(_KERNEL)
	if (fletcher_4_kstat != NULL) {
	kstat_delete(fletcher_4_kstat);
	fletcher_4_kstat = NULL;
	}
	#endif
	}

	/* ABD adapters */

	static void
	abd_fletcher_4_init(zio_abd_checksum_data_t *cdp)
	{
	const fletcher_4_ops_t *ops = fletcher_4_impl_get();
	cdp->acd_private = (void *) ops;

	if (ops->uses_fpu == B_TRUE) {
	kfpu_begin();
	}
	if (cdp->acd_byteorder == ZIO_CHECKSUM_NATIVE)
	ops->init_native(cdp->acd_ctx);
	else
	ops->init_byteswap(cdp->acd_ctx);

	}

	static void
	abd_fletcher_4_fini(zio_abd_checksum_data_t *cdp)
	{
	fletcher_4_ops_t ops = (fletcher_4_ops_t )cdp->acd_private;

	ASSERT(ops);

	if (cdp->acd_byteorder == ZIO_CHECKSUM_NATIVE)
	ops->fini_native(cdp->acd_ctx, cdp->acd_zcp);
	else
	ops->fini_byteswap(cdp->acd_ctx, cdp->acd_zcp);

	if (ops->uses_fpu == B_TRUE) {
	kfpu_end();
	}
	}


	static void
	abd_fletcher_4_simd2scalar(boolean_t native, void *data, size_t size,
	zio_abd_checksum_data_t *cdp)
	{
	zio_cksum_t *zcp = cdp->acd_zcp;

	ASSERT3U(size, <, FLETCHER_MIN_SIMD_SIZE);

	abd_fletcher_4_fini(cdp);
	cdp->acd_private = (void *)&fletcher_4_scalar_ops;

	if (native)
	fletcher_4_incremental_native(data, size, zcp);
	else
	fletcher_4_incremental_byteswap(data, size, zcp);
	}

	static int
	abd_fletcher_4_iter(void data, size_t size, void private)
	{
	zio_abd_checksum_data_t cdp = (zio_abd_checksum_data_t )private;
	fletcher_4_ctx_t *ctx = cdp->acd_ctx;
	fletcher_4_ops_t ops = (fletcher_4_ops_t )cdp->acd_private;
	boolean_t native = cdp->acd_byteorder == ZIO_CHECKSUM_NATIVE;
	- uint64_t asize = P2ALIGN(size, FLETCHER_MIN_SIMD_SIZE);
	+ uint64_t asize = P2ALIGN_TYPED(size, FLETCHER_MIN_SIMD_SIZE, uint64_t);

	ASSERT(IS_P2ALIGNED(size, sizeof (uint32_t)));

	if (asize > 0) {
	if (native)
	ops->compute_native(ctx, data, asize);
	else
	ops->compute_byteswap(ctx, data, asize);

	size -= asize;
	data = (char *)data + asize;
	}

	if (size > 0) {
	ASSERT3U(size, <, FLETCHER_MIN_SIMD_SIZE);
	/* At this point we have to switch to scalar impl */
	abd_fletcher_4_simd2scalar(native, data, size, cdp);
	}

	return (0);
	}

	zio_abd_checksum_func_t fletcher_4_abd_ops = {
	.acf_init = abd_fletcher_4_init,
	.acf_fini = abd_fletcher_4_fini,
	.acf_iter = abd_fletcher_4_iter
	};

	#if defined(_KERNEL)

	#define IMPL_FMT(impl, i) (((impl) == (i)) ? "[%s] " : "%s ")

	#if defined(__linux__)

	static int
	fletcher_4_param_get(char buffer, zfs_kernel_param_t unused)
	{
	const uint32_t impl = IMPL_READ(fletcher_4_impl_chosen);
	char *fmt;
	int cnt = 0;

	/* list fastest */
	fmt = IMPL_FMT(impl, IMPL_FASTEST);
	cnt += kmem_scnprintf(buffer + cnt, PAGE_SIZE - cnt, fmt, "fastest");

	/* list all supported implementations */
	for (uint32_t i = 0; i < fletcher_4_supp_impls_cnt; ++i) {
	fmt = IMPL_FMT(impl, i);
	cnt += kmem_scnprintf(buffer + cnt, PAGE_SIZE - cnt, fmt,
	fletcher_4_supp_impls[i]->name);
	}

	return (cnt);
	}

	static int
	fletcher_4_param_set(const char val, zfs_kernel_param_t unused)
	{
	return (fletcher_4_impl_set(val));
	}

	#else

	#include <sys/sbuf.h>

	static int
	fletcher_4_param(ZFS_MODULE_PARAM_ARGS)
	{
	int err;

	if (req->newptr == NULL) {
	const uint32_t impl = IMPL_READ(fletcher_4_impl_chosen);
	const int init_buflen = 64;
	const char *fmt;
	struct sbuf *s;

	s = sbuf_new_for_sysctl(NULL, NULL, init_buflen, req);

	/* list fastest */
	fmt = IMPL_FMT(impl, IMPL_FASTEST);
	(void) sbuf_printf(s, fmt, "fastest");

	/* list all supported implementations */
	for (uint32_t i = 0; i < fletcher_4_supp_impls_cnt; ++i) {
	fmt = IMPL_FMT(impl, i);
	(void) sbuf_printf(s, fmt,
	fletcher_4_supp_impls[i]->name);
	}

	err = sbuf_finish(s);
	sbuf_delete(s);

	return (err);
	}

	char buf[16];

	err = sysctl_handle_string(oidp, buf, sizeof (buf), req);
	if (err)
	return (err);
	return (-fletcher_4_impl_set(buf));
	}

	#endif

	#undef IMPL_FMT

	/*
	* Choose a fletcher 4 implementation in ZFS.
	* Users can choose "cycle" to exercise all implementations, but this is
	* for testing purpose therefore it can only be set in user space.
	*/
	ZFS_MODULE_VIRTUAL_PARAM_CALL(zfs, zfs_, fletcher_4_impl,
	fletcher_4_param_set, fletcher_4_param_get, ZMOD_RW,
	"Select fletcher 4 implementation.");

	EXPORT_SYMBOL(fletcher_init);
	EXPORT_SYMBOL(fletcher_2_incremental_native);
	EXPORT_SYMBOL(fletcher_2_incremental_byteswap);
	EXPORT_SYMBOL(fletcher_4_init);
	EXPORT_SYMBOL(fletcher_4_fini);
	EXPORT_SYMBOL(fletcher_2_native);
	EXPORT_SYMBOL(fletcher_2_byteswap);
	EXPORT_SYMBOL(fletcher_4_native);
	EXPORT_SYMBOL(fletcher_4_native_varsize);
	EXPORT_SYMBOL(fletcher_4_byteswap);
	EXPORT_SYMBOL(fletcher_4_incremental_native);
	EXPORT_SYMBOL(fletcher_4_incremental_byteswap);
	EXPORT_SYMBOL(fletcher_4_abd_ops);
	#endif
	diff --git a/sys/contrib/openzfs/module/zfs/arc.c b/sys/contrib/openzfs/module/zfs/arc.c
	index 51039af9bcc0..30d30b98a6c6 100644
	--- a/sys/contrib/openzfs/module/zfs/arc.c
	+++ b/sys/contrib/openzfs/module/zfs/arc.c
	@@ -1,10751 +1,10751 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2018, Joyent, Inc.
	* Copyright (c) 2011, 2020, Delphix. All rights reserved.
	* Copyright (c) 2014, Saso Kiselkov. All rights reserved.
	* Copyright (c) 2017, Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2019, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
	* Copyright (c) 2020, George Amanakis. All rights reserved.
	* Copyright (c) 2019, Klara Inc.
	* Copyright (c) 2019, Allan Jude
	* Copyright (c) 2020, The FreeBSD Foundation [1]
	*
	* [1] Portions of this software were developed by Allan Jude
	* under sponsorship from the FreeBSD Foundation.
	*/

	/*
	* DVA-based Adjustable Replacement Cache
	*
	* While much of the theory of operation used here is
	* based on the self-tuning, low overhead replacement cache
	* presented by Megiddo and Modha at FAST 2003, there are some
	* significant differences:
	*
	* 1. The Megiddo and Modha model assumes any page is evictable.
	* Pages in its cache cannot be "locked" into memory. This makes
	* the eviction algorithm simple: evict the last page in the list.
	* This also make the performance characteristics easy to reason
	* about. Our cache is not so simple. At any given moment, some
	* subset of the blocks in the cache are un-evictable because we
	* have handed out a reference to them. Blocks are only evictable
	* when there are no external references active. This makes
	* eviction far more problematic: we choose to evict the evictable
	* blocks that are the "lowest" in the list.
	*
	* There are times when it is not possible to evict the requested
	* space. In these circumstances we are unable to adjust the cache
	* size. To prevent the cache growing unbounded at these times we
	* implement a "cache throttle" that slows the flow of new data
	* into the cache until we can make space available.
	*
	* 2. The Megiddo and Modha model assumes a fixed cache size.
	* Pages are evicted when the cache is full and there is a cache
	* miss. Our model has a variable sized cache. It grows with
	* high use, but also tries to react to memory pressure from the
	* operating system: decreasing its size when system memory is
	* tight.
	*
	* 3. The Megiddo and Modha model assumes a fixed page size. All
	* elements of the cache are therefore exactly the same size. So
	* when adjusting the cache size following a cache miss, its simply
	* a matter of choosing a single page to evict. In our model, we
	* have variable sized cache blocks (ranging from 512 bytes to
	* 128K bytes). We therefore choose a set of blocks to evict to make
	* space for a cache miss that approximates as closely as possible
	* the space used by the new block.
	*
	* See also: "ARC: A Self-Tuning, Low Overhead Replacement Cache"
	* by N. Megiddo & D. Modha, FAST 2003
	*/

	/*
	* The locking model:
	*
	* A new reference to a cache buffer can be obtained in two
	* ways: 1) via a hash table lookup using the DVA as a key,
	* or 2) via one of the ARC lists. The arc_read() interface
	* uses method 1, while the internal ARC algorithms for
	* adjusting the cache use method 2. We therefore provide two
	* types of locks: 1) the hash table lock array, and 2) the
	* ARC list locks.
	*
	* Buffers do not have their own mutexes, rather they rely on the
	* hash table mutexes for the bulk of their protection (i.e. most
	* fields in the arc_buf_hdr_t are protected by these mutexes).
	*
	* buf_hash_find() returns the appropriate mutex (held) when it
	* locates the requested buffer in the hash table. It returns
	* NULL for the mutex if the buffer was not in the table.
	*
	* buf_hash_remove() expects the appropriate hash mutex to be
	* already held before it is invoked.
	*
	* Each ARC state also has a mutex which is used to protect the
	* buffer list associated with the state. When attempting to
	* obtain a hash table lock while holding an ARC list lock you
	* must use: mutex_tryenter() to avoid deadlock. Also note that
	* the active state mutex must be held before the ghost state mutex.
	*
	* It as also possible to register a callback which is run when the
	* metadata limit is reached and no buffers can be safely evicted. In
	* this case the arc user should drop a reference on some arc buffers so
	* they can be reclaimed. For example, when using the ZPL each dentry
	* holds a references on a znode. These dentries must be pruned before
	* the arc buffer holding the znode can be safely evicted.
	*
	* Note that the majority of the performance stats are manipulated
	* with atomic operations.
	*
	* The L2ARC uses the l2ad_mtx on each vdev for the following:
	*
	* - L2ARC buflist creation
	* - L2ARC buflist eviction
	* - L2ARC write completion, which walks L2ARC buflists
	* - ARC header destruction, as it removes from L2ARC buflists
	* - ARC header release, as it removes from L2ARC buflists
	*/

	/*
	* ARC operation:
	*
	* Every block that is in the ARC is tracked by an arc_buf_hdr_t structure.
	* This structure can point either to a block that is still in the cache or to
	* one that is only accessible in an L2 ARC device, or it can provide
	* information about a block that was recently evicted. If a block is
	* only accessible in the L2ARC, then the arc_buf_hdr_t only has enough
	* information to retrieve it from the L2ARC device. This information is
	* stored in the l2arc_buf_hdr_t sub-structure of the arc_buf_hdr_t. A block
	* that is in this state cannot access the data directly.
	*
	* Blocks that are actively being referenced or have not been evicted
	* are cached in the L1ARC. The L1ARC (l1arc_buf_hdr_t) is a structure within
	* the arc_buf_hdr_t that will point to the data block in memory. A block can
	* only be read by a consumer if it has an l1arc_buf_hdr_t. The L1ARC
	* caches data in two ways -- in a list of ARC buffers (arc_buf_t) and
	* also in the arc_buf_hdr_t's private physical data block pointer (b_pabd).
	*
	* The L1ARC's data pointer may or may not be uncompressed. The ARC has the
	* ability to store the physical data (b_pabd) associated with the DVA of the
	* arc_buf_hdr_t. Since the b_pabd is a copy of the on-disk physical block,
	* it will match its on-disk compression characteristics. This behavior can be
	* disabled by setting 'zfs_compressed_arc_enabled' to B_FALSE. When the
	* compressed ARC functionality is disabled, the b_pabd will point to an
	* uncompressed version of the on-disk data.
	*
	* Data in the L1ARC is not accessed by consumers of the ARC directly. Each
	* arc_buf_hdr_t can have multiple ARC buffers (arc_buf_t) which reference it.
	* Each ARC buffer (arc_buf_t) is being actively accessed by a specific ARC
	* consumer. The ARC will provide references to this data and will keep it
	* cached until it is no longer in use. The ARC caches only the L1ARC's physical
	* data block and will evict any arc_buf_t that is no longer referenced. The
	* amount of memory consumed by the arc_buf_ts' data buffers can be seen via the
	* "overhead_size" kstat.
	*
	* Depending on the consumer, an arc_buf_t can be requested in uncompressed or
	* compressed form. The typical case is that consumers will want uncompressed
	* data, and when that happens a new data buffer is allocated where the data is
	* decompressed for them to use. Currently the only consumer who wants
	* compressed arc_buf_t's is "zfs send", when it streams data exactly as it
	* exists on disk. When this happens, the arc_buf_t's data buffer is shared
	* with the arc_buf_hdr_t.
	*
	* Here is a diagram showing an arc_buf_hdr_t referenced by two arc_buf_t's. The
	* first one is owned by a compressed send consumer (and therefore references
	* the same compressed data buffer as the arc_buf_hdr_t) and the second could be
	* used by any other consumer (and has its own uncompressed copy of the data
	* buffer).
	*
	* arc_buf_hdr_t
	* +-----------+
	* \| fields \|
	* \| common to \|
	* \| L1- and \|
	* \| L2ARC \|
	* +-----------+
	* \| l2arc_buf_hdr_t
	* \| \|
	* +-----------+
	* \| l1arc_buf_hdr_t
	* \| \| arc_buf_t
	* \| b_buf +------------>+-----------+ arc_buf_t
	* \| b_pabd +-+ \|b_next +---->+-----------+
	* +-----------+ \| \|-----------\| \|b_next +-->NULL
	* \| \|b_comp = T \| +-----------+
	* \| \|b_data +-+ \|b_comp = F \|
	* \| +-----------+ \| \|b_data +-+
	* +->+------+ \| +-----------+ \|
	* compressed \| \| \| \|
	* data \| \|<--------------+ \| uncompressed
	* +------+ compressed, \| data
	* shared +-->+------+
	* data \| \|
	* \| \|
	* +------+
	*
	* When a consumer reads a block, the ARC must first look to see if the
	* arc_buf_hdr_t is cached. If the hdr is cached then the ARC allocates a new
	* arc_buf_t and either copies uncompressed data into a new data buffer from an
	* existing uncompressed arc_buf_t, decompresses the hdr's b_pabd buffer into a
	* new data buffer, or shares the hdr's b_pabd buffer, depending on whether the
	* hdr is compressed and the desired compression characteristics of the
	* arc_buf_t consumer. If the arc_buf_t ends up sharing data with the
	* arc_buf_hdr_t and both of them are uncompressed then the arc_buf_t must be
	* the last buffer in the hdr's b_buf list, however a shared compressed buf can
	* be anywhere in the hdr's list.
	*
	* The diagram below shows an example of an uncompressed ARC hdr that is
	* sharing its data with an arc_buf_t (note that the shared uncompressed buf is
	* the last element in the buf list):
	*
	* arc_buf_hdr_t
	* +-----------+
	* \| \|
	* \| \|
	* \| \|
	* +-----------+
	* l2arc_buf_hdr_t\| \|
	* \| \|
	* +-----------+
	* l1arc_buf_hdr_t\| \|
	* \| \| arc_buf_t (shared)
	* \| b_buf +------------>+---------+ arc_buf_t
	* \| \| \|b_next +---->+---------+
	* \| b_pabd +-+ \|---------\| \|b_next +-->NULL
	* +-----------+ \| \| \| +---------+
	* \| \|b_data +-+ \| \|
	* \| +---------+ \| \|b_data +-+
	* +->+------+ \| +---------+ \|
	* \| \| \| \|
	* uncompressed \| \| \| \|
	* data +------+ \| \|
	* ^ +->+------+ \|
	* \| uncompressed \| \| \|
	* \| data \| \| \|
	* \| +------+ \|
	* +---------------------------------+
	*
	* Writing to the ARC requires that the ARC first discard the hdr's b_pabd
	* since the physical block is about to be rewritten. The new data contents
	* will be contained in the arc_buf_t. As the I/O pipeline performs the write,
	* it may compress the data before writing it to disk. The ARC will be called
	* with the transformed data and will memcpy the transformed on-disk block into
	* a newly allocated b_pabd. Writes are always done into buffers which have
	* either been loaned (and hence are new and don't have other readers) or
	* buffers which have been released (and hence have their own hdr, if there
	* were originally other readers of the buf's original hdr). This ensures that
	* the ARC only needs to update a single buf and its hdr after a write occurs.
	*
	* When the L2ARC is in use, it will also take advantage of the b_pabd. The
	* L2ARC will always write the contents of b_pabd to the L2ARC. This means
	* that when compressed ARC is enabled that the L2ARC blocks are identical
	* to the on-disk block in the main data pool. This provides a significant
	* advantage since the ARC can leverage the bp's checksum when reading from the
	* L2ARC to determine if the contents are valid. However, if the compressed
	* ARC is disabled, then the L2ARC's block must be transformed to look
	* like the physical block in the main data pool before comparing the
	* checksum and determining its validity.
	*
	* The L1ARC has a slightly different system for storing encrypted data.
	* Raw (encrypted + possibly compressed) data has a few subtle differences from
	* data that is just compressed. The biggest difference is that it is not
	* possible to decrypt encrypted data (or vice-versa) if the keys aren't loaded.
	* The other difference is that encryption cannot be treated as a suggestion.
	* If a caller would prefer compressed data, but they actually wind up with
	* uncompressed data the worst thing that could happen is there might be a
	* performance hit. If the caller requests encrypted data, however, we must be
	* sure they actually get it or else secret information could be leaked. Raw
	* data is stored in hdr->b_crypt_hdr.b_rabd. An encrypted header, therefore,
	* may have both an encrypted version and a decrypted version of its data at
	* once. When a caller needs a raw arc_buf_t, it is allocated and the data is
	* copied out of this header. To avoid complications with b_pabd, raw buffers
	* cannot be shared.
	*/

	#include <sys/spa.h>
	#include <sys/zio.h>
	#include <sys/spa_impl.h>
	#include <sys/zio_compress.h>
	#include <sys/zio_checksum.h>
	#include <sys/zfs_context.h>
	#include <sys/arc.h>
	#include <sys/zfs_refcount.h>
	#include <sys/vdev.h>
	#include <sys/vdev_impl.h>
	#include <sys/dsl_pool.h>
	#include <sys/multilist.h>
	#include <sys/abd.h>
	#include <sys/zil.h>
	#include <sys/fm/fs/zfs.h>
	#include <sys/callb.h>
	#include <sys/kstat.h>
	#include <sys/zthr.h>
	#include <zfs_fletcher.h>
	#include <sys/arc_impl.h>
	#include <sys/trace_zfs.h>
	#include <sys/aggsum.h>
	#include <sys/wmsum.h>
	#include <cityhash.h>
	#include <sys/vdev_trim.h>
	#include <sys/zfs_racct.h>
	#include <sys/zstd/zstd.h>

	#ifndef _KERNEL
	/* set with ZFS_DEBUG=watch, to enable watchpoints on frozen buffers */
	boolean_t arc_watch = B_FALSE;
	#endif

	/*
	* This thread's job is to keep enough free memory in the system, by
	* calling arc_kmem_reap_soon() plus arc_reduce_target_size(), which improves
	* arc_available_memory().
	*/
	static zthr_t *arc_reap_zthr;

	/*
	* This thread's job is to keep arc_size under arc_c, by calling
	* arc_evict(), which improves arc_is_overflowing().
	*/
	static zthr_t *arc_evict_zthr;
	static arc_buf_hdr_t **arc_state_evict_markers;
	static int arc_state_evict_marker_count;

	static kmutex_t arc_evict_lock;
	static boolean_t arc_evict_needed = B_FALSE;
	static clock_t arc_last_uncached_flush;

	/*
	* Count of bytes evicted since boot.
	*/
	static uint64_t arc_evict_count;

	/*
	* List of arc_evict_waiter_t's, representing threads waiting for the
	* arc_evict_count to reach specific values.
	*/
	static list_t arc_evict_waiters;

	/*
	* When arc_is_overflowing(), arc_get_data_impl() waits for this percent of
	* the requested amount of data to be evicted. For example, by default for
	* every 2KB that's evicted, 1KB of it may be "reused" by a new allocation.
	* Since this is above 100%, it ensures that progress is made towards getting
	* arc_size under arc_c. Since this is finite, it ensures that allocations
	* can still happen, even during the potentially long time that arc_size is
	* more than arc_c.
	*/
	static uint_t zfs_arc_eviction_pct = 200;

	/*
	* The number of headers to evict in arc_evict_state_impl() before
	* dropping the sublist lock and evicting from another sublist. A lower
	* value means we're more likely to evict the "correct" header (i.e. the
	* oldest header in the arc state), but comes with higher overhead
	* (i.e. more invocations of arc_evict_state_impl()).
	*/
	static uint_t zfs_arc_evict_batch_limit = 10;

	/* number of seconds before growing cache again */
	uint_t arc_grow_retry = 5;

	/*
	* Minimum time between calls to arc_kmem_reap_soon().
	*/
	static const int arc_kmem_cache_reap_retry_ms = 1000;

	/* shift of arc_c for calculating overflow limit in arc_get_data_impl */
	static int zfs_arc_overflow_shift = 8;

	/* log2(fraction of arc to reclaim) */
	uint_t arc_shrink_shift = 7;

	/* percent of pagecache to reclaim arc to */
	#ifdef _KERNEL
	uint_t zfs_arc_pc_percent = 0;
	#endif

	/*
	* log2(fraction of ARC which must be free to allow growing).
	* I.e. If there is less than arc_c >> arc_no_grow_shift free memory,
	* when reading a new block into the ARC, we will evict an equal-sized block
	* from the ARC.
	*
	* This must be less than arc_shrink_shift, so that when we shrink the ARC,
	* we will still not allow it to grow.
	*/
	uint_t arc_no_grow_shift = 5;


	/*
	* minimum lifespan of a prefetch block in clock ticks
	* (initialized in arc_init())
	*/
	static uint_t arc_min_prefetch_ms;
	static uint_t arc_min_prescient_prefetch_ms;

	/*
	* If this percent of memory is free, don't throttle.
	*/
	uint_t arc_lotsfree_percent = 10;

	/*
	* The arc has filled available memory and has now warmed up.
	*/
	boolean_t arc_warm;

	/*
	* These tunables are for performance analysis.
	*/
	uint64_t zfs_arc_max = 0;
	uint64_t zfs_arc_min = 0;
	static uint64_t zfs_arc_dnode_limit = 0;
	static uint_t zfs_arc_dnode_reduce_percent = 10;
	static uint_t zfs_arc_grow_retry = 0;
	static uint_t zfs_arc_shrink_shift = 0;
	uint_t zfs_arc_average_blocksize = 8 * 1024; /* 8KB */

	/*
	* ARC dirty data constraints for arc_tempreserve_space() throttle:
	* * total dirty data limit
	* * anon block dirty limit
	* * each pool's anon allowance
	*/
	static const unsigned long zfs_arc_dirty_limit_percent = 50;
	static const unsigned long zfs_arc_anon_limit_percent = 25;
	static const unsigned long zfs_arc_pool_dirty_percent = 20;

	/*
	* Enable or disable compressed arc buffers.
	*/
	int zfs_compressed_arc_enabled = B_TRUE;

	/*
	* Balance between metadata and data on ghost hits. Values above 100
	* increase metadata caching by proportionally reducing effect of ghost
	* data hits on target data/metadata rate.
	*/
	static uint_t zfs_arc_meta_balance = 500;

	/*
	* Percentage that can be consumed by dnodes of ARC meta buffers.
	*/
	static uint_t zfs_arc_dnode_limit_percent = 10;

	/*
	* These tunables are Linux-specific
	*/
	static uint64_t zfs_arc_sys_free = 0;
	static uint_t zfs_arc_min_prefetch_ms = 0;
	static uint_t zfs_arc_min_prescient_prefetch_ms = 0;
	static uint_t zfs_arc_lotsfree_percent = 10;

	/*
	* Number of arc_prune threads
	*/
	static int zfs_arc_prune_task_threads = 1;

	/* The 7 states: */
	arc_state_t ARC_anon;
	arc_state_t ARC_mru;
	arc_state_t ARC_mru_ghost;
	arc_state_t ARC_mfu;
	arc_state_t ARC_mfu_ghost;
	arc_state_t ARC_l2c_only;
	arc_state_t ARC_uncached;

	arc_stats_t arc_stats = {
	{ "hits", KSTAT_DATA_UINT64 },
	{ "iohits", KSTAT_DATA_UINT64 },
	{ "misses", KSTAT_DATA_UINT64 },
	{ "demand_data_hits", KSTAT_DATA_UINT64 },
	{ "demand_data_iohits", KSTAT_DATA_UINT64 },
	{ "demand_data_misses", KSTAT_DATA_UINT64 },
	{ "demand_metadata_hits", KSTAT_DATA_UINT64 },
	{ "demand_metadata_iohits", KSTAT_DATA_UINT64 },
	{ "demand_metadata_misses", KSTAT_DATA_UINT64 },
	{ "prefetch_data_hits", KSTAT_DATA_UINT64 },
	{ "prefetch_data_iohits", KSTAT_DATA_UINT64 },
	{ "prefetch_data_misses", KSTAT_DATA_UINT64 },
	{ "prefetch_metadata_hits", KSTAT_DATA_UINT64 },
	{ "prefetch_metadata_iohits", KSTAT_DATA_UINT64 },
	{ "prefetch_metadata_misses", KSTAT_DATA_UINT64 },
	{ "mru_hits", KSTAT_DATA_UINT64 },
	{ "mru_ghost_hits", KSTAT_DATA_UINT64 },
	{ "mfu_hits", KSTAT_DATA_UINT64 },
	{ "mfu_ghost_hits", KSTAT_DATA_UINT64 },
	{ "uncached_hits", KSTAT_DATA_UINT64 },
	{ "deleted", KSTAT_DATA_UINT64 },
	{ "mutex_miss", KSTAT_DATA_UINT64 },
	{ "access_skip", KSTAT_DATA_UINT64 },
	{ "evict_skip", KSTAT_DATA_UINT64 },
	{ "evict_not_enough", KSTAT_DATA_UINT64 },
	{ "evict_l2_cached", KSTAT_DATA_UINT64 },
	{ "evict_l2_eligible", KSTAT_DATA_UINT64 },
	{ "evict_l2_eligible_mfu", KSTAT_DATA_UINT64 },
	{ "evict_l2_eligible_mru", KSTAT_DATA_UINT64 },
	{ "evict_l2_ineligible", KSTAT_DATA_UINT64 },
	{ "evict_l2_skip", KSTAT_DATA_UINT64 },
	{ "hash_elements", KSTAT_DATA_UINT64 },
	{ "hash_elements_max", KSTAT_DATA_UINT64 },
	{ "hash_collisions", KSTAT_DATA_UINT64 },
	{ "hash_chains", KSTAT_DATA_UINT64 },
	{ "hash_chain_max", KSTAT_DATA_UINT64 },
	{ "meta", KSTAT_DATA_UINT64 },
	{ "pd", KSTAT_DATA_UINT64 },
	{ "pm", KSTAT_DATA_UINT64 },
	{ "c", KSTAT_DATA_UINT64 },
	{ "c_min", KSTAT_DATA_UINT64 },
	{ "c_max", KSTAT_DATA_UINT64 },
	{ "size", KSTAT_DATA_UINT64 },
	{ "compressed_size", KSTAT_DATA_UINT64 },
	{ "uncompressed_size", KSTAT_DATA_UINT64 },
	{ "overhead_size", KSTAT_DATA_UINT64 },
	{ "hdr_size", KSTAT_DATA_UINT64 },
	{ "data_size", KSTAT_DATA_UINT64 },
	{ "metadata_size", KSTAT_DATA_UINT64 },
	{ "dbuf_size", KSTAT_DATA_UINT64 },
	{ "dnode_size", KSTAT_DATA_UINT64 },
	{ "bonus_size", KSTAT_DATA_UINT64 },
	#if defined(COMPAT_FREEBSD11)
	{ "other_size", KSTAT_DATA_UINT64 },
	#endif
	{ "anon_size", KSTAT_DATA_UINT64 },
	{ "anon_data", KSTAT_DATA_UINT64 },
	{ "anon_metadata", KSTAT_DATA_UINT64 },
	{ "anon_evictable_data", KSTAT_DATA_UINT64 },
	{ "anon_evictable_metadata", KSTAT_DATA_UINT64 },
	{ "mru_size", KSTAT_DATA_UINT64 },
	{ "mru_data", KSTAT_DATA_UINT64 },
	{ "mru_metadata", KSTAT_DATA_UINT64 },
	{ "mru_evictable_data", KSTAT_DATA_UINT64 },
	{ "mru_evictable_metadata", KSTAT_DATA_UINT64 },
	{ "mru_ghost_size", KSTAT_DATA_UINT64 },
	{ "mru_ghost_data", KSTAT_DATA_UINT64 },
	{ "mru_ghost_metadata", KSTAT_DATA_UINT64 },
	{ "mru_ghost_evictable_data", KSTAT_DATA_UINT64 },
	{ "mru_ghost_evictable_metadata", KSTAT_DATA_UINT64 },
	{ "mfu_size", KSTAT_DATA_UINT64 },
	{ "mfu_data", KSTAT_DATA_UINT64 },
	{ "mfu_metadata", KSTAT_DATA_UINT64 },
	{ "mfu_evictable_data", KSTAT_DATA_UINT64 },
	{ "mfu_evictable_metadata", KSTAT_DATA_UINT64 },
	{ "mfu_ghost_size", KSTAT_DATA_UINT64 },
	{ "mfu_ghost_data", KSTAT_DATA_UINT64 },
	{ "mfu_ghost_metadata", KSTAT_DATA_UINT64 },
	{ "mfu_ghost_evictable_data", KSTAT_DATA_UINT64 },
	{ "mfu_ghost_evictable_metadata", KSTAT_DATA_UINT64 },
	{ "uncached_size", KSTAT_DATA_UINT64 },
	{ "uncached_data", KSTAT_DATA_UINT64 },
	{ "uncached_metadata", KSTAT_DATA_UINT64 },
	{ "uncached_evictable_data", KSTAT_DATA_UINT64 },
	{ "uncached_evictable_metadata", KSTAT_DATA_UINT64 },
	{ "l2_hits", KSTAT_DATA_UINT64 },
	{ "l2_misses", KSTAT_DATA_UINT64 },
	{ "l2_prefetch_asize", KSTAT_DATA_UINT64 },
	{ "l2_mru_asize", KSTAT_DATA_UINT64 },
	{ "l2_mfu_asize", KSTAT_DATA_UINT64 },
	{ "l2_bufc_data_asize", KSTAT_DATA_UINT64 },
	{ "l2_bufc_metadata_asize", KSTAT_DATA_UINT64 },
	{ "l2_feeds", KSTAT_DATA_UINT64 },
	{ "l2_rw_clash", KSTAT_DATA_UINT64 },
	{ "l2_read_bytes", KSTAT_DATA_UINT64 },
	{ "l2_write_bytes", KSTAT_DATA_UINT64 },
	{ "l2_writes_sent", KSTAT_DATA_UINT64 },
	{ "l2_writes_done", KSTAT_DATA_UINT64 },
	{ "l2_writes_error", KSTAT_DATA_UINT64 },
	{ "l2_writes_lock_retry", KSTAT_DATA_UINT64 },
	{ "l2_evict_lock_retry", KSTAT_DATA_UINT64 },
	{ "l2_evict_reading", KSTAT_DATA_UINT64 },
	{ "l2_evict_l1cached", KSTAT_DATA_UINT64 },
	{ "l2_free_on_write", KSTAT_DATA_UINT64 },
	{ "l2_abort_lowmem", KSTAT_DATA_UINT64 },
	{ "l2_cksum_bad", KSTAT_DATA_UINT64 },
	{ "l2_io_error", KSTAT_DATA_UINT64 },
	{ "l2_size", KSTAT_DATA_UINT64 },
	{ "l2_asize", KSTAT_DATA_UINT64 },
	{ "l2_hdr_size", KSTAT_DATA_UINT64 },
	{ "l2_log_blk_writes", KSTAT_DATA_UINT64 },
	{ "l2_log_blk_avg_asize", KSTAT_DATA_UINT64 },
	{ "l2_log_blk_asize", KSTAT_DATA_UINT64 },
	{ "l2_log_blk_count", KSTAT_DATA_UINT64 },
	{ "l2_data_to_meta_ratio", KSTAT_DATA_UINT64 },
	{ "l2_rebuild_success", KSTAT_DATA_UINT64 },
	{ "l2_rebuild_unsupported", KSTAT_DATA_UINT64 },
	{ "l2_rebuild_io_errors", KSTAT_DATA_UINT64 },
	{ "l2_rebuild_dh_errors", KSTAT_DATA_UINT64 },
	{ "l2_rebuild_cksum_lb_errors", KSTAT_DATA_UINT64 },
	{ "l2_rebuild_lowmem", KSTAT_DATA_UINT64 },
	{ "l2_rebuild_size", KSTAT_DATA_UINT64 },
	{ "l2_rebuild_asize", KSTAT_DATA_UINT64 },
	{ "l2_rebuild_bufs", KSTAT_DATA_UINT64 },
	{ "l2_rebuild_bufs_precached", KSTAT_DATA_UINT64 },
	{ "l2_rebuild_log_blks", KSTAT_DATA_UINT64 },
	{ "memory_throttle_count", KSTAT_DATA_UINT64 },
	{ "memory_direct_count", KSTAT_DATA_UINT64 },
	{ "memory_indirect_count", KSTAT_DATA_UINT64 },
	{ "memory_all_bytes", KSTAT_DATA_UINT64 },
	{ "memory_free_bytes", KSTAT_DATA_UINT64 },
	{ "memory_available_bytes", KSTAT_DATA_INT64 },
	{ "arc_no_grow", KSTAT_DATA_UINT64 },
	{ "arc_tempreserve", KSTAT_DATA_UINT64 },
	{ "arc_loaned_bytes", KSTAT_DATA_UINT64 },
	{ "arc_prune", KSTAT_DATA_UINT64 },
	{ "arc_meta_used", KSTAT_DATA_UINT64 },
	{ "arc_dnode_limit", KSTAT_DATA_UINT64 },
	{ "async_upgrade_sync", KSTAT_DATA_UINT64 },
	{ "predictive_prefetch", KSTAT_DATA_UINT64 },
	{ "demand_hit_predictive_prefetch", KSTAT_DATA_UINT64 },
	{ "demand_iohit_predictive_prefetch", KSTAT_DATA_UINT64 },
	{ "prescient_prefetch", KSTAT_DATA_UINT64 },
	{ "demand_hit_prescient_prefetch", KSTAT_DATA_UINT64 },
	{ "demand_iohit_prescient_prefetch", KSTAT_DATA_UINT64 },
	{ "arc_need_free", KSTAT_DATA_UINT64 },
	{ "arc_sys_free", KSTAT_DATA_UINT64 },
	{ "arc_raw_size", KSTAT_DATA_UINT64 },
	{ "cached_only_in_progress", KSTAT_DATA_UINT64 },
	{ "abd_chunk_waste_size", KSTAT_DATA_UINT64 },
	};

	arc_sums_t arc_sums;

	#define ARCSTAT_MAX(stat, val) { \
	uint64_t m; \
	while ((val) > (m = arc_stats.stat.value.ui64) && \
	(m != atomic_cas_64(&arc_stats.stat.value.ui64, m, (val)))) \
	continue; \
	}

	/*
	* We define a macro to allow ARC hits/misses to be easily broken down by
	* two separate conditions, giving a total of four different subtypes for
	* each of hits and misses (so eight statistics total).
	*/
	#define ARCSTAT_CONDSTAT(cond1, stat1, notstat1, cond2, stat2, notstat2, stat) \
	if (cond1) { \
	if (cond2) { \
	ARCSTAT_BUMP(arcstat_##stat1##_##stat2##_##stat); \
	} else { \
	ARCSTAT_BUMP(arcstat_##stat1##_##notstat2##_##stat); \
	} \
	} else { \
	if (cond2) { \
	ARCSTAT_BUMP(arcstat_##notstat1##_##stat2##_##stat); \
	} else { \
	ARCSTAT_BUMP(arcstat_##notstat1##_##notstat2##_##stat);\
	} \
	}

	/*
	* This macro allows us to use kstats as floating averages. Each time we
	* update this kstat, we first factor it and the update value by
	* ARCSTAT_AVG_FACTOR to shrink the new value's contribution to the overall
	* average. This macro assumes that integer loads and stores are atomic, but
	* is not safe for multiple writers updating the kstat in parallel (only the
	* last writer's update will remain).
	*/
	#define ARCSTAT_F_AVG_FACTOR 3
	#define ARCSTAT_F_AVG(stat, value) \
	do { \
	uint64_t x = ARCSTAT(stat); \
	x = x - x / ARCSTAT_F_AVG_FACTOR + \
	(value) / ARCSTAT_F_AVG_FACTOR; \
	ARCSTAT(stat) = x; \
	} while (0)

	static kstat_t *arc_ksp;

	/*
	* There are several ARC variables that are critical to export as kstats --
	* but we don't want to have to grovel around in the kstat whenever we wish to
	* manipulate them. For these variables, we therefore define them to be in
	* terms of the statistic variable. This assures that we are not introducing
	* the possibility of inconsistency by having shadow copies of the variables,
	* while still allowing the code to be readable.
	*/
	#define arc_tempreserve ARCSTAT(arcstat_tempreserve)
	#define arc_loaned_bytes ARCSTAT(arcstat_loaned_bytes)
	#define arc_dnode_limit ARCSTAT(arcstat_dnode_limit) /* max size for dnodes */
	#define arc_need_free ARCSTAT(arcstat_need_free) /* waiting to be evicted */

	hrtime_t arc_growtime;
	list_t arc_prune_list;
	kmutex_t arc_prune_mtx;
	taskq_t *arc_prune_taskq;

	#define GHOST_STATE(state) \
	((state) == arc_mru_ghost \|\| (state) == arc_mfu_ghost \|\| \
	(state) == arc_l2c_only)

	#define HDR_IN_HASH_TABLE(hdr) ((hdr)->b_flags & ARC_FLAG_IN_HASH_TABLE)
	#define HDR_IO_IN_PROGRESS(hdr) ((hdr)->b_flags & ARC_FLAG_IO_IN_PROGRESS)
	#define HDR_IO_ERROR(hdr) ((hdr)->b_flags & ARC_FLAG_IO_ERROR)
	#define HDR_PREFETCH(hdr) ((hdr)->b_flags & ARC_FLAG_PREFETCH)
	#define HDR_PRESCIENT_PREFETCH(hdr) \
	((hdr)->b_flags & ARC_FLAG_PRESCIENT_PREFETCH)
	#define HDR_COMPRESSION_ENABLED(hdr) \
	((hdr)->b_flags & ARC_FLAG_COMPRESSED_ARC)

	#define HDR_L2CACHE(hdr) ((hdr)->b_flags & ARC_FLAG_L2CACHE)
	#define HDR_UNCACHED(hdr) ((hdr)->b_flags & ARC_FLAG_UNCACHED)
	#define HDR_L2_READING(hdr) \
	(((hdr)->b_flags & ARC_FLAG_IO_IN_PROGRESS) && \
	((hdr)->b_flags & ARC_FLAG_HAS_L2HDR))
	#define HDR_L2_WRITING(hdr) ((hdr)->b_flags & ARC_FLAG_L2_WRITING)
	#define HDR_L2_EVICTED(hdr) ((hdr)->b_flags & ARC_FLAG_L2_EVICTED)
	#define HDR_L2_WRITE_HEAD(hdr) ((hdr)->b_flags & ARC_FLAG_L2_WRITE_HEAD)
	#define HDR_PROTECTED(hdr) ((hdr)->b_flags & ARC_FLAG_PROTECTED)
	#define HDR_NOAUTH(hdr) ((hdr)->b_flags & ARC_FLAG_NOAUTH)
	#define HDR_SHARED_DATA(hdr) ((hdr)->b_flags & ARC_FLAG_SHARED_DATA)

	#define HDR_ISTYPE_METADATA(hdr) \
	((hdr)->b_flags & ARC_FLAG_BUFC_METADATA)
	#define HDR_ISTYPE_DATA(hdr) (!HDR_ISTYPE_METADATA(hdr))

	#define HDR_HAS_L1HDR(hdr) ((hdr)->b_flags & ARC_FLAG_HAS_L1HDR)
	#define HDR_HAS_L2HDR(hdr) ((hdr)->b_flags & ARC_FLAG_HAS_L2HDR)
	#define HDR_HAS_RABD(hdr) \
	(HDR_HAS_L1HDR(hdr) && HDR_PROTECTED(hdr) && \
	(hdr)->b_crypt_hdr.b_rabd != NULL)
	#define HDR_ENCRYPTED(hdr) \
	(HDR_PROTECTED(hdr) && DMU_OT_IS_ENCRYPTED((hdr)->b_crypt_hdr.b_ot))
	#define HDR_AUTHENTICATED(hdr) \
	(HDR_PROTECTED(hdr) && !DMU_OT_IS_ENCRYPTED((hdr)->b_crypt_hdr.b_ot))

	/* For storing compression mode in b_flags */
	#define HDR_COMPRESS_OFFSET (highbit64(ARC_FLAG_COMPRESS_0) - 1)

	#define HDR_GET_COMPRESS(hdr) ((enum zio_compress)BF32_GET((hdr)->b_flags, \
	HDR_COMPRESS_OFFSET, SPA_COMPRESSBITS))
	#define HDR_SET_COMPRESS(hdr, cmp) BF32_SET((hdr)->b_flags, \
	HDR_COMPRESS_OFFSET, SPA_COMPRESSBITS, (cmp));

	#define ARC_BUF_LAST(buf) ((buf)->b_next == NULL)
	#define ARC_BUF_SHARED(buf) ((buf)->b_flags & ARC_BUF_FLAG_SHARED)
	#define ARC_BUF_COMPRESSED(buf) ((buf)->b_flags & ARC_BUF_FLAG_COMPRESSED)
	#define ARC_BUF_ENCRYPTED(buf) ((buf)->b_flags & ARC_BUF_FLAG_ENCRYPTED)

	/*
	* Other sizes
	*/

	#define HDR_FULL_SIZE ((int64_t)sizeof (arc_buf_hdr_t))
	#define HDR_L2ONLY_SIZE ((int64_t)offsetof(arc_buf_hdr_t, b_l1hdr))

	/*
	* Hash table routines
	*/

	#define BUF_LOCKS 2048
	typedef struct buf_hash_table {
	uint64_t ht_mask;
	arc_buf_hdr_t **ht_table;
	kmutex_t ht_locks[BUF_LOCKS] ____cacheline_aligned;
	} buf_hash_table_t;

	static buf_hash_table_t buf_hash_table;

	#define BUF_HASH_INDEX(spa, dva, birth) \
	(buf_hash(spa, dva, birth) & buf_hash_table.ht_mask)
	#define BUF_HASH_LOCK(idx) (&buf_hash_table.ht_locks[idx & (BUF_LOCKS-1)])
	#define HDR_LOCK(hdr) \
	(BUF_HASH_LOCK(BUF_HASH_INDEX(hdr->b_spa, &hdr->b_dva, hdr->b_birth)))

	uint64_t zfs_crc64_table[256];

	/*
	* Level 2 ARC
	*/

	#define L2ARC_WRITE_SIZE (32 * 1024 * 1024) /* initial write max */
	#define L2ARC_HEADROOM 8 /* num of writes */

	/*
	* If we discover during ARC scan any buffers to be compressed, we boost
	* our headroom for the next scanning cycle by this percentage multiple.
	*/
	#define L2ARC_HEADROOM_BOOST 200
	#define L2ARC_FEED_SECS 1 /* caching interval secs */
	#define L2ARC_FEED_MIN_MS 200 /* min caching interval ms */

	/*
	* We can feed L2ARC from two states of ARC buffers, mru and mfu,
	* and each of the state has two types: data and metadata.
	*/
	#define L2ARC_FEED_TYPES 4

	/* L2ARC Performance Tunables */
	uint64_t l2arc_write_max = L2ARC_WRITE_SIZE; /* def max write size */
	uint64_t l2arc_write_boost = L2ARC_WRITE_SIZE; /* extra warmup write */
	uint64_t l2arc_headroom = L2ARC_HEADROOM; /* # of dev writes */
	uint64_t l2arc_headroom_boost = L2ARC_HEADROOM_BOOST;
	uint64_t l2arc_feed_secs = L2ARC_FEED_SECS; /* interval seconds */
	uint64_t l2arc_feed_min_ms = L2ARC_FEED_MIN_MS; /* min interval msecs */
	int l2arc_noprefetch = B_TRUE; /* don't cache prefetch bufs */
	int l2arc_feed_again = B_TRUE; /* turbo warmup */
	int l2arc_norw = B_FALSE; /* no reads during writes */
	static uint_t l2arc_meta_percent = 33; /* limit on headers size */

	/*
	* L2ARC Internals
	*/
	static list_t L2ARC_dev_list; /* device list */
	static list_t l2arc_dev_list; / device list pointer */
	static kmutex_t l2arc_dev_mtx; /* device list mutex */
	static l2arc_dev_t l2arc_dev_last; / last device used */
	static list_t L2ARC_free_on_write; /* free after write buf list */
	static list_t l2arc_free_on_write; / free after write list ptr */
	static kmutex_t l2arc_free_on_write_mtx; /* mutex for list */
	static uint64_t l2arc_ndev; /* number of devices */

	typedef struct l2arc_read_callback {
	arc_buf_hdr_t l2rcb_hdr; / read header */
	blkptr_t l2rcb_bp; /* original blkptr */
	zbookmark_phys_t l2rcb_zb; /* original bookmark */
	int l2rcb_flags; /* original flags */
	abd_t l2rcb_abd; / temporary buffer */
	} l2arc_read_callback_t;

	typedef struct l2arc_data_free {
	/* protected by l2arc_free_on_write_mtx */
	abd_t *l2df_abd;
	size_t l2df_size;
	arc_buf_contents_t l2df_type;
	list_node_t l2df_list_node;
	} l2arc_data_free_t;

	typedef enum arc_fill_flags {
	ARC_FILL_LOCKED = 1 << 0, /* hdr lock is held */
	ARC_FILL_COMPRESSED = 1 << 1, /* fill with compressed data */
	ARC_FILL_ENCRYPTED = 1 << 2, /* fill with encrypted data */
	ARC_FILL_NOAUTH = 1 << 3, /* don't attempt to authenticate */
	ARC_FILL_IN_PLACE = 1 << 4 /* fill in place (special case) */
	} arc_fill_flags_t;

	typedef enum arc_ovf_level {
	ARC_OVF_NONE, /* ARC within target size. */
	ARC_OVF_SOME, /* ARC is slightly overflowed. */
	ARC_OVF_SEVERE /* ARC is severely overflowed. */
	} arc_ovf_level_t;

	static kmutex_t l2arc_feed_thr_lock;
	static kcondvar_t l2arc_feed_thr_cv;
	static uint8_t l2arc_thread_exit;

	static kmutex_t l2arc_rebuild_thr_lock;
	static kcondvar_t l2arc_rebuild_thr_cv;

	enum arc_hdr_alloc_flags {
	ARC_HDR_ALLOC_RDATA = 0x1,
	ARC_HDR_USE_RESERVE = 0x4,
	ARC_HDR_ALLOC_LINEAR = 0x8,
	};


	static abd_t arc_get_data_abd(arc_buf_hdr_t , uint64_t, const void *, int);
	static void arc_get_data_buf(arc_buf_hdr_t , uint64_t, const void *);
	static void arc_get_data_impl(arc_buf_hdr_t , uint64_t, const void , int);
	static void arc_free_data_abd(arc_buf_hdr_t , abd_t , uint64_t, const void *);
	static void arc_free_data_buf(arc_buf_hdr_t , void , uint64_t, const void *);
	static void arc_free_data_impl(arc_buf_hdr_t *hdr, uint64_t size,
	const void *tag);
	static void arc_hdr_free_abd(arc_buf_hdr_t *, boolean_t);
	static void arc_hdr_alloc_abd(arc_buf_hdr_t *, int);
	static void arc_hdr_destroy(arc_buf_hdr_t *);
	static void arc_access(arc_buf_hdr_t *, arc_flags_t, boolean_t);
	static void arc_buf_watch(arc_buf_t *);
	static void arc_change_state(arc_state_t , arc_buf_hdr_t );

	static arc_buf_contents_t arc_buf_type(arc_buf_hdr_t *);
	static uint32_t arc_bufc_to_flags(arc_buf_contents_t);
	static inline void arc_hdr_set_flags(arc_buf_hdr_t *hdr, arc_flags_t flags);
	static inline void arc_hdr_clear_flags(arc_buf_hdr_t *hdr, arc_flags_t flags);

	static boolean_t l2arc_write_eligible(uint64_t, arc_buf_hdr_t *);
	static void l2arc_read_done(zio_t *);
	static void l2arc_do_free_on_write(void);
	static void l2arc_hdr_arcstats_update(arc_buf_hdr_t *hdr, boolean_t incr,
	boolean_t state_only);

	static void arc_prune_async(uint64_t adjust);

	#define l2arc_hdr_arcstats_increment(hdr) \
	l2arc_hdr_arcstats_update((hdr), B_TRUE, B_FALSE)
	#define l2arc_hdr_arcstats_decrement(hdr) \
	l2arc_hdr_arcstats_update((hdr), B_FALSE, B_FALSE)
	#define l2arc_hdr_arcstats_increment_state(hdr) \
	l2arc_hdr_arcstats_update((hdr), B_TRUE, B_TRUE)
	#define l2arc_hdr_arcstats_decrement_state(hdr) \
	l2arc_hdr_arcstats_update((hdr), B_FALSE, B_TRUE)

	/*
	* l2arc_exclude_special : A zfs module parameter that controls whether buffers
	* present on special vdevs are eligibile for caching in L2ARC. If
	* set to 1, exclude dbufs on special vdevs from being cached to
	* L2ARC.
	*/
	int l2arc_exclude_special = 0;

	/*
	* l2arc_mfuonly : A ZFS module parameter that controls whether only MFU
	* metadata and data are cached from ARC into L2ARC.
	*/
	static int l2arc_mfuonly = 0;

	/*
	* L2ARC TRIM
	* l2arc_trim_ahead : A ZFS module parameter that controls how much ahead of
	* the current write size (l2arc_write_max) we should TRIM if we
	* have filled the device. It is defined as a percentage of the
	* write size. If set to 100 we trim twice the space required to
	* accommodate upcoming writes. A minimum of 64MB will be trimmed.
	* It also enables TRIM of the whole L2ARC device upon creation or
	* addition to an existing pool or if the header of the device is
	* invalid upon importing a pool or onlining a cache device. The
	* default is 0, which disables TRIM on L2ARC altogether as it can
	* put significant stress on the underlying storage devices. This
	* will vary depending of how well the specific device handles
	* these commands.
	*/
	static uint64_t l2arc_trim_ahead = 0;

	/*
	* Performance tuning of L2ARC persistence:
	*
	* l2arc_rebuild_enabled : A ZFS module parameter that controls whether adding
	* an L2ARC device (either at pool import or later) will attempt
	* to rebuild L2ARC buffer contents.
	* l2arc_rebuild_blocks_min_l2size : A ZFS module parameter that controls
	* whether log blocks are written to the L2ARC device. If the L2ARC
	* device is less than 1GB, the amount of data l2arc_evict()
	* evicts is significant compared to the amount of restored L2ARC
	* data. In this case do not write log blocks in L2ARC in order
	* not to waste space.
	*/
	static int l2arc_rebuild_enabled = B_TRUE;
	static uint64_t l2arc_rebuild_blocks_min_l2size = 1024 * 1024 * 1024;

	/* L2ARC persistence rebuild control routines. */
	void l2arc_rebuild_vdev(vdev_t *vd, boolean_t reopen);
	static __attribute__((noreturn)) void l2arc_dev_rebuild_thread(void *arg);
	static int l2arc_rebuild(l2arc_dev_t *dev);

	/* L2ARC persistence read I/O routines. */
	static int l2arc_dev_hdr_read(l2arc_dev_t *dev);
	static int l2arc_log_blk_read(l2arc_dev_t *dev,
	const l2arc_log_blkptr_t this_lp, const l2arc_log_blkptr_t next_lp,
	l2arc_log_blk_phys_t this_lb, l2arc_log_blk_phys_t next_lb,
	zio_t this_io, zio_t *next_io);
	static zio_t l2arc_log_blk_fetch(vdev_t vd,
	const l2arc_log_blkptr_t lp, l2arc_log_blk_phys_t lb);
	static void l2arc_log_blk_fetch_abort(zio_t *zio);

	/* L2ARC persistence block restoration routines. */
	static void l2arc_log_blk_restore(l2arc_dev_t *dev,
	const l2arc_log_blk_phys_t *lb, uint64_t lb_asize);
	static void l2arc_hdr_restore(const l2arc_log_ent_phys_t *le,
	l2arc_dev_t *dev);

	/* L2ARC persistence write I/O routines. */
	static uint64_t l2arc_log_blk_commit(l2arc_dev_t dev, zio_t pio,
	l2arc_write_callback_t *cb);

	/* L2ARC persistence auxiliary routines. */
	boolean_t l2arc_log_blkptr_valid(l2arc_dev_t *dev,
	const l2arc_log_blkptr_t *lbp);
	static boolean_t l2arc_log_blk_insert(l2arc_dev_t *dev,
	const arc_buf_hdr_t *ab);
	boolean_t l2arc_range_check_overlap(uint64_t bottom,
	uint64_t top, uint64_t check);
	static void l2arc_blk_fetch_done(zio_t *zio);
	static inline uint64_t
	l2arc_log_blk_overhead(uint64_t write_sz, l2arc_dev_t *dev);

	/*
	* We use Cityhash for this. It's fast, and has good hash properties without
	* requiring any large static buffers.
	*/
	static uint64_t
	buf_hash(uint64_t spa, const dva_t *dva, uint64_t birth)
	{
	return (cityhash4(spa, dva->dva_word[0], dva->dva_word[1], birth));
	}

	#define HDR_EMPTY(hdr) \
	((hdr)->b_dva.dva_word[0] == 0 && \
	(hdr)->b_dva.dva_word[1] == 0)

	#define HDR_EMPTY_OR_LOCKED(hdr) \
	(HDR_EMPTY(hdr) \|\| MUTEX_HELD(HDR_LOCK(hdr)))

	#define HDR_EQUAL(spa, dva, birth, hdr) \
	((hdr)->b_dva.dva_word[0] == (dva)->dva_word[0]) && \
	((hdr)->b_dva.dva_word[1] == (dva)->dva_word[1]) && \
	((hdr)->b_birth == birth) && ((hdr)->b_spa == spa)

	static void
	buf_discard_identity(arc_buf_hdr_t *hdr)
	{
	hdr->b_dva.dva_word[0] = 0;
	hdr->b_dva.dva_word[1] = 0;
	hdr->b_birth = 0;
	}

	static arc_buf_hdr_t *
	buf_hash_find(uint64_t spa, const blkptr_t bp, kmutex_t *lockp)
	{
	const dva_t *dva = BP_IDENTITY(bp);
	uint64_t birth = BP_GET_BIRTH(bp);
	uint64_t idx = BUF_HASH_INDEX(spa, dva, birth);
	kmutex_t *hash_lock = BUF_HASH_LOCK(idx);
	arc_buf_hdr_t *hdr;

	mutex_enter(hash_lock);
	for (hdr = buf_hash_table.ht_table[idx]; hdr != NULL;
	hdr = hdr->b_hash_next) {
	if (HDR_EQUAL(spa, dva, birth, hdr)) {
	*lockp = hash_lock;
	return (hdr);
	}
	}
	mutex_exit(hash_lock);
	*lockp = NULL;
	return (NULL);
	}

	/*
	* Insert an entry into the hash table. If there is already an element
	* equal to elem in the hash table, then the already existing element
	* will be returned and the new element will not be inserted.
	* Otherwise returns NULL.
	* If lockp == NULL, the caller is assumed to already hold the hash lock.
	*/
	static arc_buf_hdr_t *
	buf_hash_insert(arc_buf_hdr_t hdr, kmutex_t *lockp)
	{
	uint64_t idx = BUF_HASH_INDEX(hdr->b_spa, &hdr->b_dva, hdr->b_birth);
	kmutex_t *hash_lock = BUF_HASH_LOCK(idx);
	arc_buf_hdr_t *fhdr;
	uint32_t i;

	ASSERT(!DVA_IS_EMPTY(&hdr->b_dva));
	ASSERT(hdr->b_birth != 0);
	ASSERT(!HDR_IN_HASH_TABLE(hdr));

	if (lockp != NULL) {
	*lockp = hash_lock;
	mutex_enter(hash_lock);
	} else {
	ASSERT(MUTEX_HELD(hash_lock));
	}

	for (fhdr = buf_hash_table.ht_table[idx], i = 0; fhdr != NULL;
	fhdr = fhdr->b_hash_next, i++) {
	if (HDR_EQUAL(hdr->b_spa, &hdr->b_dva, hdr->b_birth, fhdr))
	return (fhdr);
	}

	hdr->b_hash_next = buf_hash_table.ht_table[idx];
	buf_hash_table.ht_table[idx] = hdr;
	arc_hdr_set_flags(hdr, ARC_FLAG_IN_HASH_TABLE);

	/* collect some hash table performance data */
	if (i > 0) {
	ARCSTAT_BUMP(arcstat_hash_collisions);
	if (i == 1)
	ARCSTAT_BUMP(arcstat_hash_chains);

	ARCSTAT_MAX(arcstat_hash_chain_max, i);
	}
	uint64_t he = atomic_inc_64_nv(
	&arc_stats.arcstat_hash_elements.value.ui64);
	ARCSTAT_MAX(arcstat_hash_elements_max, he);

	return (NULL);
	}

	static void
	buf_hash_remove(arc_buf_hdr_t *hdr)
	{
	arc_buf_hdr_t fhdr, *hdrp;
	uint64_t idx = BUF_HASH_INDEX(hdr->b_spa, &hdr->b_dva, hdr->b_birth);

	ASSERT(MUTEX_HELD(BUF_HASH_LOCK(idx)));
	ASSERT(HDR_IN_HASH_TABLE(hdr));

	hdrp = &buf_hash_table.ht_table[idx];
	while ((fhdr = *hdrp) != hdr) {
	ASSERT3P(fhdr, !=, NULL);
	hdrp = &fhdr->b_hash_next;
	}
	*hdrp = hdr->b_hash_next;
	hdr->b_hash_next = NULL;
	arc_hdr_clear_flags(hdr, ARC_FLAG_IN_HASH_TABLE);

	/* collect some hash table performance data */
	atomic_dec_64(&arc_stats.arcstat_hash_elements.value.ui64);

	if (buf_hash_table.ht_table[idx] &&
	buf_hash_table.ht_table[idx]->b_hash_next == NULL)
	ARCSTAT_BUMPDOWN(arcstat_hash_chains);
	}

	/*
	* Global data structures and functions for the buf kmem cache.
	*/

	static kmem_cache_t *hdr_full_cache;
	static kmem_cache_t *hdr_l2only_cache;
	static kmem_cache_t *buf_cache;

	static void
	buf_fini(void)
	{
	#if defined(_KERNEL)
	/*
	* Large allocations which do not require contiguous pages
	* should be using vmem_free() in the linux kernel\
	*/
	vmem_free(buf_hash_table.ht_table,
	(buf_hash_table.ht_mask + 1) * sizeof (void *));
	#else
	kmem_free(buf_hash_table.ht_table,
	(buf_hash_table.ht_mask + 1) * sizeof (void *));
	#endif
	for (int i = 0; i < BUF_LOCKS; i++)
	mutex_destroy(BUF_HASH_LOCK(i));
	kmem_cache_destroy(hdr_full_cache);
	kmem_cache_destroy(hdr_l2only_cache);
	kmem_cache_destroy(buf_cache);
	}

	/*
	* Constructor callback - called when the cache is empty
	* and a new buf is requested.
	*/
	static int
	hdr_full_cons(void vbuf, void unused, int kmflag)
	{
	(void) unused, (void) kmflag;
	arc_buf_hdr_t *hdr = vbuf;

	memset(hdr, 0, HDR_FULL_SIZE);
	hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS;
	zfs_refcount_create(&hdr->b_l1hdr.b_refcnt);
	#ifdef ZFS_DEBUG
	mutex_init(&hdr->b_l1hdr.b_freeze_lock, NULL, MUTEX_DEFAULT, NULL);
	#endif
	multilist_link_init(&hdr->b_l1hdr.b_arc_node);
	list_link_init(&hdr->b_l2hdr.b_l2node);
	arc_space_consume(HDR_FULL_SIZE, ARC_SPACE_HDRS);

	return (0);
	}

	static int
	hdr_l2only_cons(void vbuf, void unused, int kmflag)
	{
	(void) unused, (void) kmflag;
	arc_buf_hdr_t *hdr = vbuf;

	memset(hdr, 0, HDR_L2ONLY_SIZE);
	arc_space_consume(HDR_L2ONLY_SIZE, ARC_SPACE_L2HDRS);

	return (0);
	}

	static int
	buf_cons(void vbuf, void unused, int kmflag)
	{
	(void) unused, (void) kmflag;
	arc_buf_t *buf = vbuf;

	memset(buf, 0, sizeof (arc_buf_t));
	arc_space_consume(sizeof (arc_buf_t), ARC_SPACE_HDRS);

	return (0);
	}

	/*
	* Destructor callback - called when a cached buf is
	* no longer required.
	*/
	static void
	hdr_full_dest(void vbuf, void unused)
	{
	(void) unused;
	arc_buf_hdr_t *hdr = vbuf;

	ASSERT(HDR_EMPTY(hdr));
	zfs_refcount_destroy(&hdr->b_l1hdr.b_refcnt);
	#ifdef ZFS_DEBUG
	mutex_destroy(&hdr->b_l1hdr.b_freeze_lock);
	#endif
	ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node));
	arc_space_return(HDR_FULL_SIZE, ARC_SPACE_HDRS);
	}

	static void
	hdr_l2only_dest(void vbuf, void unused)
	{
	(void) unused;
	arc_buf_hdr_t *hdr = vbuf;

	ASSERT(HDR_EMPTY(hdr));
	arc_space_return(HDR_L2ONLY_SIZE, ARC_SPACE_L2HDRS);
	}

	static void
	buf_dest(void vbuf, void unused)
	{
	(void) unused;
	(void) vbuf;

	arc_space_return(sizeof (arc_buf_t), ARC_SPACE_HDRS);
	}

	static void
	buf_init(void)
	{
	uint64_t *ct = NULL;
	uint64_t hsize = 1ULL << 12;
	int i, j;

	/*
	* The hash table is big enough to fill all of physical memory
	* with an average block size of zfs_arc_average_blocksize (default 8K).
	* By default, the table will take up
	* totalmem * sizeof(void*) / 8K (1MB per GB with 8-byte pointers).
	*/
	while (hsize * zfs_arc_average_blocksize < arc_all_memory())
	hsize <<= 1;
	retry:
	buf_hash_table.ht_mask = hsize - 1;
	#if defined(_KERNEL)
	/*
	* Large allocations which do not require contiguous pages
	* should be using vmem_alloc() in the linux kernel
	*/
	buf_hash_table.ht_table =
	vmem_zalloc(hsize * sizeof (void*), KM_SLEEP);
	#else
	buf_hash_table.ht_table =
	kmem_zalloc(hsize * sizeof (void*), KM_NOSLEEP);
	#endif
	if (buf_hash_table.ht_table == NULL) {
	ASSERT(hsize > (1ULL << 8));
	hsize >>= 1;
	goto retry;
	}

	hdr_full_cache = kmem_cache_create("arc_buf_hdr_t_full", HDR_FULL_SIZE,
	0, hdr_full_cons, hdr_full_dest, NULL, NULL, NULL, 0);
	hdr_l2only_cache = kmem_cache_create("arc_buf_hdr_t_l2only",
	HDR_L2ONLY_SIZE, 0, hdr_l2only_cons, hdr_l2only_dest, NULL,
	NULL, NULL, 0);
	buf_cache = kmem_cache_create("arc_buf_t", sizeof (arc_buf_t),
	0, buf_cons, buf_dest, NULL, NULL, NULL, 0);

	for (i = 0; i < 256; i++)
	for (ct = zfs_crc64_table + i, *ct = i, j = 8; j > 0; j--)
	ct = (ct >> 1) ^ (-(*ct & 1) & ZFS_CRC64_POLY);

	for (i = 0; i < BUF_LOCKS; i++)
	mutex_init(BUF_HASH_LOCK(i), NULL, MUTEX_DEFAULT, NULL);
	}

	#define ARC_MINTIME (hz>>4) /* 62 ms */

	/*
	* This is the size that the buf occupies in memory. If the buf is compressed,
	* it will correspond to the compressed size. You should use this method of
	* getting the buf size unless you explicitly need the logical size.
	*/
	uint64_t
	arc_buf_size(arc_buf_t *buf)
	{
	return (ARC_BUF_COMPRESSED(buf) ?
	HDR_GET_PSIZE(buf->b_hdr) : HDR_GET_LSIZE(buf->b_hdr));
	}

	uint64_t
	arc_buf_lsize(arc_buf_t *buf)
	{
	return (HDR_GET_LSIZE(buf->b_hdr));
	}

	/*
	* This function will return B_TRUE if the buffer is encrypted in memory.
	* This buffer can be decrypted by calling arc_untransform().
	*/
	boolean_t
	arc_is_encrypted(arc_buf_t *buf)
	{
	return (ARC_BUF_ENCRYPTED(buf) != 0);
	}

	/*
	* Returns B_TRUE if the buffer represents data that has not had its MAC
	* verified yet.
	*/
	boolean_t
	arc_is_unauthenticated(arc_buf_t *buf)
	{
	return (HDR_NOAUTH(buf->b_hdr) != 0);
	}

	void
	arc_get_raw_params(arc_buf_t buf, boolean_t byteorder, uint8_t *salt,
	uint8_t iv, uint8_t mac)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;

	ASSERT(HDR_PROTECTED(hdr));

	memcpy(salt, hdr->b_crypt_hdr.b_salt, ZIO_DATA_SALT_LEN);
	memcpy(iv, hdr->b_crypt_hdr.b_iv, ZIO_DATA_IV_LEN);
	memcpy(mac, hdr->b_crypt_hdr.b_mac, ZIO_DATA_MAC_LEN);
	*byteorder = (hdr->b_l1hdr.b_byteswap == DMU_BSWAP_NUMFUNCS) ?
	ZFS_HOST_BYTEORDER : !ZFS_HOST_BYTEORDER;
	}

	/*
	* Indicates how this buffer is compressed in memory. If it is not compressed
	* the value will be ZIO_COMPRESS_OFF. It can be made normally readable with
	* arc_untransform() as long as it is also unencrypted.
	*/
	enum zio_compress
	arc_get_compression(arc_buf_t *buf)
	{
	return (ARC_BUF_COMPRESSED(buf) ?
	HDR_GET_COMPRESS(buf->b_hdr) : ZIO_COMPRESS_OFF);
	}

	/*
	* Return the compression algorithm used to store this data in the ARC. If ARC
	* compression is enabled or this is an encrypted block, this will be the same
	* as what's used to store it on-disk. Otherwise, this will be ZIO_COMPRESS_OFF.
	*/
	static inline enum zio_compress
	arc_hdr_get_compress(arc_buf_hdr_t *hdr)
	{
	return (HDR_COMPRESSION_ENABLED(hdr) ?
	HDR_GET_COMPRESS(hdr) : ZIO_COMPRESS_OFF);
	}

	uint8_t
	arc_get_complevel(arc_buf_t *buf)
	{
	return (buf->b_hdr->b_complevel);
	}

	static inline boolean_t
	arc_buf_is_shared(arc_buf_t *buf)
	{
	boolean_t shared = (buf->b_data != NULL &&
	buf->b_hdr->b_l1hdr.b_pabd != NULL &&
	abd_is_linear(buf->b_hdr->b_l1hdr.b_pabd) &&
	buf->b_data == abd_to_buf(buf->b_hdr->b_l1hdr.b_pabd));
	IMPLY(shared, HDR_SHARED_DATA(buf->b_hdr));
	EQUIV(shared, ARC_BUF_SHARED(buf));
	IMPLY(shared, ARC_BUF_COMPRESSED(buf) \|\| ARC_BUF_LAST(buf));

	/*
	* It would be nice to assert arc_can_share() too, but the "hdr isn't
	* already being shared" requirement prevents us from doing that.
	*/

	return (shared);
	}

	/*
	* Free the checksum associated with this header. If there is no checksum, this
	* is a no-op.
	*/
	static inline void
	arc_cksum_free(arc_buf_hdr_t *hdr)
	{
	#ifdef ZFS_DEBUG
	ASSERT(HDR_HAS_L1HDR(hdr));

	mutex_enter(&hdr->b_l1hdr.b_freeze_lock);
	if (hdr->b_l1hdr.b_freeze_cksum != NULL) {
	kmem_free(hdr->b_l1hdr.b_freeze_cksum, sizeof (zio_cksum_t));
	hdr->b_l1hdr.b_freeze_cksum = NULL;
	}
	mutex_exit(&hdr->b_l1hdr.b_freeze_lock);
	#endif
	}

	/*
	* Return true iff at least one of the bufs on hdr is not compressed.
	* Encrypted buffers count as compressed.
	*/
	static boolean_t
	arc_hdr_has_uncompressed_buf(arc_buf_hdr_t *hdr)
	{
	ASSERT(hdr->b_l1hdr.b_state == arc_anon \|\| HDR_EMPTY_OR_LOCKED(hdr));

	for (arc_buf_t *b = hdr->b_l1hdr.b_buf; b != NULL; b = b->b_next) {
	if (!ARC_BUF_COMPRESSED(b)) {
	return (B_TRUE);
	}
	}
	return (B_FALSE);
	}


	/*
	* If we've turned on the ZFS_DEBUG_MODIFY flag, verify that the buf's data
	* matches the checksum that is stored in the hdr. If there is no checksum,
	* or if the buf is compressed, this is a no-op.
	*/
	static void
	arc_cksum_verify(arc_buf_t *buf)
	{
	#ifdef ZFS_DEBUG
	arc_buf_hdr_t *hdr = buf->b_hdr;
	zio_cksum_t zc;

	if (!(zfs_flags & ZFS_DEBUG_MODIFY))
	return;

	if (ARC_BUF_COMPRESSED(buf))
	return;

	ASSERT(HDR_HAS_L1HDR(hdr));

	mutex_enter(&hdr->b_l1hdr.b_freeze_lock);

	if (hdr->b_l1hdr.b_freeze_cksum == NULL \|\| HDR_IO_ERROR(hdr)) {
	mutex_exit(&hdr->b_l1hdr.b_freeze_lock);
	return;
	}

	fletcher_2_native(buf->b_data, arc_buf_size(buf), NULL, &zc);
	if (!ZIO_CHECKSUM_EQUAL(*hdr->b_l1hdr.b_freeze_cksum, zc))
	panic("buffer modified while frozen!");
	mutex_exit(&hdr->b_l1hdr.b_freeze_lock);
	#endif
	}

	/*
	* This function makes the assumption that data stored in the L2ARC
	* will be transformed exactly as it is in the main pool. Because of
	* this we can verify the checksum against the reading process's bp.
	*/
	static boolean_t
	arc_cksum_is_equal(arc_buf_hdr_t hdr, zio_t zio)
	{
	ASSERT(!BP_IS_EMBEDDED(zio->io_bp));
	VERIFY3U(BP_GET_PSIZE(zio->io_bp), ==, HDR_GET_PSIZE(hdr));

	/*
	* Block pointers always store the checksum for the logical data.
	* If the block pointer has the gang bit set, then the checksum
	* it represents is for the reconstituted data and not for an
	* individual gang member. The zio pipeline, however, must be able to
	* determine the checksum of each of the gang constituents so it
	* treats the checksum comparison differently than what we need
	* for l2arc blocks. This prevents us from using the
	* zio_checksum_error() interface directly. Instead we must call the
	* zio_checksum_error_impl() so that we can ensure the checksum is
	* generated using the correct checksum algorithm and accounts for the
	* logical I/O size and not just a gang fragment.
	*/
	return (zio_checksum_error_impl(zio->io_spa, zio->io_bp,
	BP_GET_CHECKSUM(zio->io_bp), zio->io_abd, zio->io_size,
	zio->io_offset, NULL) == 0);
	}

	/*
	* Given a buf full of data, if ZFS_DEBUG_MODIFY is enabled this computes a
	* checksum and attaches it to the buf's hdr so that we can ensure that the buf
	* isn't modified later on. If buf is compressed or there is already a checksum
	* on the hdr, this is a no-op (we only checksum uncompressed bufs).
	*/
	static void
	arc_cksum_compute(arc_buf_t *buf)
	{
	if (!(zfs_flags & ZFS_DEBUG_MODIFY))
	return;

	#ifdef ZFS_DEBUG
	arc_buf_hdr_t *hdr = buf->b_hdr;
	ASSERT(HDR_HAS_L1HDR(hdr));
	mutex_enter(&hdr->b_l1hdr.b_freeze_lock);
	if (hdr->b_l1hdr.b_freeze_cksum != NULL \|\| ARC_BUF_COMPRESSED(buf)) {
	mutex_exit(&hdr->b_l1hdr.b_freeze_lock);
	return;
	}

	ASSERT(!ARC_BUF_ENCRYPTED(buf));
	ASSERT(!ARC_BUF_COMPRESSED(buf));
	hdr->b_l1hdr.b_freeze_cksum = kmem_alloc(sizeof (zio_cksum_t),
	KM_SLEEP);
	fletcher_2_native(buf->b_data, arc_buf_size(buf), NULL,
	hdr->b_l1hdr.b_freeze_cksum);
	mutex_exit(&hdr->b_l1hdr.b_freeze_lock);
	#endif
	arc_buf_watch(buf);
	}

	#ifndef _KERNEL
	void
	arc_buf_sigsegv(int sig, siginfo_t si, void unused)
	{
	(void) sig, (void) unused;
	panic("Got SIGSEGV at address: 0x%lx\n", (long)si->si_addr);
	}
	#endif

	static void
	arc_buf_unwatch(arc_buf_t *buf)
	{
	#ifndef _KERNEL
	if (arc_watch) {
	ASSERT0(mprotect(buf->b_data, arc_buf_size(buf),
	PROT_READ \| PROT_WRITE));
	}
	#else
	(void) buf;
	#endif
	}

	static void
	arc_buf_watch(arc_buf_t *buf)
	{
	#ifndef _KERNEL
	if (arc_watch)
	ASSERT0(mprotect(buf->b_data, arc_buf_size(buf),
	PROT_READ));
	#else
	(void) buf;
	#endif
	}

	static arc_buf_contents_t
	arc_buf_type(arc_buf_hdr_t *hdr)
	{
	arc_buf_contents_t type;
	if (HDR_ISTYPE_METADATA(hdr)) {
	type = ARC_BUFC_METADATA;
	} else {
	type = ARC_BUFC_DATA;
	}
	VERIFY3U(hdr->b_type, ==, type);
	return (type);
	}

	boolean_t
	arc_is_metadata(arc_buf_t *buf)
	{
	return (HDR_ISTYPE_METADATA(buf->b_hdr) != 0);
	}

	static uint32_t
	arc_bufc_to_flags(arc_buf_contents_t type)
	{
	switch (type) {
	case ARC_BUFC_DATA:
	/* metadata field is 0 if buffer contains normal data */
	return (0);
	case ARC_BUFC_METADATA:
	return (ARC_FLAG_BUFC_METADATA);
	default:
	break;
	}
	panic("undefined ARC buffer type!");
	return ((uint32_t)-1);
	}

	void
	arc_buf_thaw(arc_buf_t *buf)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;

	ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon);
	ASSERT(!HDR_IO_IN_PROGRESS(hdr));

	arc_cksum_verify(buf);

	/*
	* Compressed buffers do not manipulate the b_freeze_cksum.
	*/
	if (ARC_BUF_COMPRESSED(buf))
	return;

	ASSERT(HDR_HAS_L1HDR(hdr));
	arc_cksum_free(hdr);
	arc_buf_unwatch(buf);
	}

	void
	arc_buf_freeze(arc_buf_t *buf)
	{
	if (!(zfs_flags & ZFS_DEBUG_MODIFY))
	return;

	if (ARC_BUF_COMPRESSED(buf))
	return;

	ASSERT(HDR_HAS_L1HDR(buf->b_hdr));
	arc_cksum_compute(buf);
	}

	/*
	* The arc_buf_hdr_t's b_flags should never be modified directly. Instead,
	* the following functions should be used to ensure that the flags are
	* updated in a thread-safe way. When manipulating the flags either
	* the hash_lock must be held or the hdr must be undiscoverable. This
	* ensures that we're not racing with any other threads when updating
	* the flags.
	*/
	static inline void
	arc_hdr_set_flags(arc_buf_hdr_t *hdr, arc_flags_t flags)
	{
	ASSERT(HDR_EMPTY_OR_LOCKED(hdr));
	hdr->b_flags \|= flags;
	}

	static inline void
	arc_hdr_clear_flags(arc_buf_hdr_t *hdr, arc_flags_t flags)
	{
	ASSERT(HDR_EMPTY_OR_LOCKED(hdr));
	hdr->b_flags &= ~flags;
	}

	/*
	* Setting the compression bits in the arc_buf_hdr_t's b_flags is
	* done in a special way since we have to clear and set bits
	* at the same time. Consumers that wish to set the compression bits
	* must use this function to ensure that the flags are updated in
	* thread-safe manner.
	*/
	static void
	arc_hdr_set_compress(arc_buf_hdr_t *hdr, enum zio_compress cmp)
	{
	ASSERT(HDR_EMPTY_OR_LOCKED(hdr));

	/*
	* Holes and embedded blocks will always have a psize = 0 so
	* we ignore the compression of the blkptr and set the
	* want to uncompress them. Mark them as uncompressed.
	*/
	if (!zfs_compressed_arc_enabled \|\| HDR_GET_PSIZE(hdr) == 0) {
	arc_hdr_clear_flags(hdr, ARC_FLAG_COMPRESSED_ARC);
	ASSERT(!HDR_COMPRESSION_ENABLED(hdr));
	} else {
	arc_hdr_set_flags(hdr, ARC_FLAG_COMPRESSED_ARC);
	ASSERT(HDR_COMPRESSION_ENABLED(hdr));
	}

	HDR_SET_COMPRESS(hdr, cmp);
	ASSERT3U(HDR_GET_COMPRESS(hdr), ==, cmp);
	}

	/*
	* Looks for another buf on the same hdr which has the data decompressed, copies
	* from it, and returns true. If no such buf exists, returns false.
	*/
	static boolean_t
	arc_buf_try_copy_decompressed_data(arc_buf_t *buf)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;
	boolean_t copied = B_FALSE;

	ASSERT(HDR_HAS_L1HDR(hdr));
	ASSERT3P(buf->b_data, !=, NULL);
	ASSERT(!ARC_BUF_COMPRESSED(buf));

	for (arc_buf_t *from = hdr->b_l1hdr.b_buf; from != NULL;
	from = from->b_next) {
	/* can't use our own data buffer */
	if (from == buf) {
	continue;
	}

	if (!ARC_BUF_COMPRESSED(from)) {
	memcpy(buf->b_data, from->b_data, arc_buf_size(buf));
	copied = B_TRUE;
	break;
	}
	}

	#ifdef ZFS_DEBUG
	/*
	* There were no decompressed bufs, so there should not be a
	* checksum on the hdr either.
	*/
	if (zfs_flags & ZFS_DEBUG_MODIFY)
	EQUIV(!copied, hdr->b_l1hdr.b_freeze_cksum == NULL);
	#endif

	return (copied);
	}

	/*
	* Allocates an ARC buf header that's in an evicted & L2-cached state.
	* This is used during l2arc reconstruction to make empty ARC buffers
	* which circumvent the regular disk->arc->l2arc path and instead come
	* into being in the reverse order, i.e. l2arc->arc.
	*/
	static arc_buf_hdr_t *
	arc_buf_alloc_l2only(size_t size, arc_buf_contents_t type, l2arc_dev_t *dev,
	dva_t dva, uint64_t daddr, int32_t psize, uint64_t birth,
	enum zio_compress compress, uint8_t complevel, boolean_t protected,
	boolean_t prefetch, arc_state_type_t arcs_state)
	{
	arc_buf_hdr_t *hdr;

	ASSERT(size != 0);
	hdr = kmem_cache_alloc(hdr_l2only_cache, KM_SLEEP);
	hdr->b_birth = birth;
	hdr->b_type = type;
	hdr->b_flags = 0;
	arc_hdr_set_flags(hdr, arc_bufc_to_flags(type) \| ARC_FLAG_HAS_L2HDR);
	HDR_SET_LSIZE(hdr, size);
	HDR_SET_PSIZE(hdr, psize);
	arc_hdr_set_compress(hdr, compress);
	hdr->b_complevel = complevel;
	if (protected)
	arc_hdr_set_flags(hdr, ARC_FLAG_PROTECTED);
	if (prefetch)
	arc_hdr_set_flags(hdr, ARC_FLAG_PREFETCH);
	hdr->b_spa = spa_load_guid(dev->l2ad_vdev->vdev_spa);

	hdr->b_dva = dva;

	hdr->b_l2hdr.b_dev = dev;
	hdr->b_l2hdr.b_daddr = daddr;
	hdr->b_l2hdr.b_arcs_state = arcs_state;

	return (hdr);
	}

	/*
	* Return the size of the block, b_pabd, that is stored in the arc_buf_hdr_t.
	*/
	static uint64_t
	arc_hdr_size(arc_buf_hdr_t *hdr)
	{
	uint64_t size;

	if (arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF &&
	HDR_GET_PSIZE(hdr) > 0) {
	size = HDR_GET_PSIZE(hdr);
	} else {
	ASSERT3U(HDR_GET_LSIZE(hdr), !=, 0);
	size = HDR_GET_LSIZE(hdr);
	}
	return (size);
	}

	static int
	arc_hdr_authenticate(arc_buf_hdr_t hdr, spa_t spa, uint64_t dsobj)
	{
	int ret;
	uint64_t csize;
	uint64_t lsize = HDR_GET_LSIZE(hdr);
	uint64_t psize = HDR_GET_PSIZE(hdr);
	void *tmpbuf = NULL;
	abd_t *abd = hdr->b_l1hdr.b_pabd;

	ASSERT(HDR_EMPTY_OR_LOCKED(hdr));
	ASSERT(HDR_AUTHENTICATED(hdr));
	ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);

	/*
	* The MAC is calculated on the compressed data that is stored on disk.
	* However, if compressed arc is disabled we will only have the
	* decompressed data available to us now. Compress it into a temporary
	* abd so we can verify the MAC. The performance overhead of this will
	* be relatively low, since most objects in an encrypted objset will
	* be encrypted (instead of authenticated) anyway.
	*/
	if (HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF &&
	!HDR_COMPRESSION_ENABLED(hdr)) {

	csize = zio_compress_data(HDR_GET_COMPRESS(hdr),
	hdr->b_l1hdr.b_pabd, &tmpbuf, lsize, hdr->b_complevel);
	ASSERT3P(tmpbuf, !=, NULL);
	ASSERT3U(csize, <=, psize);
	abd = abd_get_from_buf(tmpbuf, lsize);
	abd_take_ownership_of_buf(abd, B_TRUE);
	abd_zero_off(abd, csize, psize - csize);
	}

	/*
	* Authentication is best effort. We authenticate whenever the key is
	* available. If we succeed we clear ARC_FLAG_NOAUTH.
	*/
	if (hdr->b_crypt_hdr.b_ot == DMU_OT_OBJSET) {
	ASSERT3U(HDR_GET_COMPRESS(hdr), ==, ZIO_COMPRESS_OFF);
	ASSERT3U(lsize, ==, psize);
	ret = spa_do_crypt_objset_mac_abd(B_FALSE, spa, dsobj, abd,
	psize, hdr->b_l1hdr.b_byteswap != DMU_BSWAP_NUMFUNCS);
	} else {
	ret = spa_do_crypt_mac_abd(B_FALSE, spa, dsobj, abd, psize,
	hdr->b_crypt_hdr.b_mac);
	}

	if (ret == 0)
	arc_hdr_clear_flags(hdr, ARC_FLAG_NOAUTH);
	else if (ret != ENOENT)
	goto error;

	if (tmpbuf != NULL)
	abd_free(abd);

	return (0);

	error:
	if (tmpbuf != NULL)
	abd_free(abd);

	return (ret);
	}

	/*
	* This function will take a header that only has raw encrypted data in
	* b_crypt_hdr.b_rabd and decrypt it into a new buffer which is stored in
	* b_l1hdr.b_pabd. If designated in the header flags, this function will
	* also decompress the data.
	*/
	static int
	arc_hdr_decrypt(arc_buf_hdr_t hdr, spa_t spa, const zbookmark_phys_t *zb)
	{
	int ret;
	abd_t *cabd = NULL;
	void *tmp = NULL;
	boolean_t no_crypt = B_FALSE;
	boolean_t bswap = (hdr->b_l1hdr.b_byteswap != DMU_BSWAP_NUMFUNCS);

	ASSERT(HDR_EMPTY_OR_LOCKED(hdr));
	ASSERT(HDR_ENCRYPTED(hdr));

	arc_hdr_alloc_abd(hdr, 0);

	ret = spa_do_crypt_abd(B_FALSE, spa, zb, hdr->b_crypt_hdr.b_ot,
	B_FALSE, bswap, hdr->b_crypt_hdr.b_salt, hdr->b_crypt_hdr.b_iv,
	hdr->b_crypt_hdr.b_mac, HDR_GET_PSIZE(hdr), hdr->b_l1hdr.b_pabd,
	hdr->b_crypt_hdr.b_rabd, &no_crypt);
	if (ret != 0)
	goto error;

	if (no_crypt) {
	abd_copy(hdr->b_l1hdr.b_pabd, hdr->b_crypt_hdr.b_rabd,
	HDR_GET_PSIZE(hdr));
	}

	/*
	* If this header has disabled arc compression but the b_pabd is
	* compressed after decrypting it, we need to decompress the newly
	* decrypted data.
	*/
	if (HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF &&
	!HDR_COMPRESSION_ENABLED(hdr)) {
	/*
	* We want to make sure that we are correctly honoring the
	* zfs_abd_scatter_enabled setting, so we allocate an abd here
	* and then loan a buffer from it, rather than allocating a
	* linear buffer and wrapping it in an abd later.
	*/
	cabd = arc_get_data_abd(hdr, arc_hdr_size(hdr), hdr, 0);
	tmp = abd_borrow_buf(cabd, arc_hdr_size(hdr));

	ret = zio_decompress_data(HDR_GET_COMPRESS(hdr),
	hdr->b_l1hdr.b_pabd, tmp, HDR_GET_PSIZE(hdr),
	HDR_GET_LSIZE(hdr), &hdr->b_complevel);
	if (ret != 0) {
	abd_return_buf(cabd, tmp, arc_hdr_size(hdr));
	goto error;
	}

	abd_return_buf_copy(cabd, tmp, arc_hdr_size(hdr));
	arc_free_data_abd(hdr, hdr->b_l1hdr.b_pabd,
	arc_hdr_size(hdr), hdr);
	hdr->b_l1hdr.b_pabd = cabd;
	}

	return (0);

	error:
	arc_hdr_free_abd(hdr, B_FALSE);
	if (cabd != NULL)
	arc_free_data_buf(hdr, cabd, arc_hdr_size(hdr), hdr);

	return (ret);
	}

	/*
	* This function is called during arc_buf_fill() to prepare the header's
	* abd plaintext pointer for use. This involves authenticated protected
	* data and decrypting encrypted data into the plaintext abd.
	*/
	static int
	arc_fill_hdr_crypt(arc_buf_hdr_t hdr, kmutex_t hash_lock, spa_t *spa,
	const zbookmark_phys_t *zb, boolean_t noauth)
	{
	int ret;

	ASSERT(HDR_PROTECTED(hdr));

	if (hash_lock != NULL)
	mutex_enter(hash_lock);

	if (HDR_NOAUTH(hdr) && !noauth) {
	/*
	* The caller requested authenticated data but our data has
	* not been authenticated yet. Verify the MAC now if we can.
	*/
	ret = arc_hdr_authenticate(hdr, spa, zb->zb_objset);
	if (ret != 0)
	goto error;
	} else if (HDR_HAS_RABD(hdr) && hdr->b_l1hdr.b_pabd == NULL) {
	/*
	* If we only have the encrypted version of the data, but the
	* unencrypted version was requested we take this opportunity
	* to store the decrypted version in the header for future use.
	*/
	ret = arc_hdr_decrypt(hdr, spa, zb);
	if (ret != 0)
	goto error;
	}

	ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);

	if (hash_lock != NULL)
	mutex_exit(hash_lock);

	return (0);

	error:
	if (hash_lock != NULL)
	mutex_exit(hash_lock);

	return (ret);
	}

	/*
	* This function is used by the dbuf code to decrypt bonus buffers in place.
	* The dbuf code itself doesn't have any locking for decrypting a shared dnode
	* block, so we use the hash lock here to protect against concurrent calls to
	* arc_buf_fill().
	*/
	static void
	arc_buf_untransform_in_place(arc_buf_t *buf)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;

	ASSERT(HDR_ENCRYPTED(hdr));
	ASSERT3U(hdr->b_crypt_hdr.b_ot, ==, DMU_OT_DNODE);
	ASSERT(HDR_EMPTY_OR_LOCKED(hdr));
	ASSERT3PF(hdr->b_l1hdr.b_pabd, !=, NULL, "hdr %px buf %px", hdr, buf);

	zio_crypt_copy_dnode_bonus(hdr->b_l1hdr.b_pabd, buf->b_data,
	arc_buf_size(buf));
	buf->b_flags &= ~ARC_BUF_FLAG_ENCRYPTED;
	buf->b_flags &= ~ARC_BUF_FLAG_COMPRESSED;
	}

	/*
	* Given a buf that has a data buffer attached to it, this function will
	* efficiently fill the buf with data of the specified compression setting from
	* the hdr and update the hdr's b_freeze_cksum if necessary. If the buf and hdr
	* are already sharing a data buf, no copy is performed.
	*
	* If the buf is marked as compressed but uncompressed data was requested, this
	* will allocate a new data buffer for the buf, remove that flag, and fill the
	* buf with uncompressed data. You can't request a compressed buf on a hdr with
	* uncompressed data, and (since we haven't added support for it yet) if you
	* want compressed data your buf must already be marked as compressed and have
	* the correct-sized data buffer.
	*/
	static int
	arc_buf_fill(arc_buf_t buf, spa_t spa, const zbookmark_phys_t *zb,
	arc_fill_flags_t flags)
	{
	int error = 0;
	arc_buf_hdr_t *hdr = buf->b_hdr;
	boolean_t hdr_compressed =
	(arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF);
	boolean_t compressed = (flags & ARC_FILL_COMPRESSED) != 0;
	boolean_t encrypted = (flags & ARC_FILL_ENCRYPTED) != 0;
	dmu_object_byteswap_t bswap = hdr->b_l1hdr.b_byteswap;
	kmutex_t *hash_lock = (flags & ARC_FILL_LOCKED) ? NULL : HDR_LOCK(hdr);

	ASSERT3P(buf->b_data, !=, NULL);
	IMPLY(compressed, hdr_compressed \|\| ARC_BUF_ENCRYPTED(buf));
	IMPLY(compressed, ARC_BUF_COMPRESSED(buf));
	IMPLY(encrypted, HDR_ENCRYPTED(hdr));
	IMPLY(encrypted, ARC_BUF_ENCRYPTED(buf));
	IMPLY(encrypted, ARC_BUF_COMPRESSED(buf));
	IMPLY(encrypted, !arc_buf_is_shared(buf));

	/*
	* If the caller wanted encrypted data we just need to copy it from
	* b_rabd and potentially byteswap it. We won't be able to do any
	* further transforms on it.
	*/
	if (encrypted) {
	ASSERT(HDR_HAS_RABD(hdr));
	abd_copy_to_buf(buf->b_data, hdr->b_crypt_hdr.b_rabd,
	HDR_GET_PSIZE(hdr));
	goto byteswap;
	}

	/*
	* Adjust encrypted and authenticated headers to accommodate
	* the request if needed. Dnode blocks (ARC_FILL_IN_PLACE) are
	* allowed to fail decryption due to keys not being loaded
	* without being marked as an IO error.
	*/
	if (HDR_PROTECTED(hdr)) {
	error = arc_fill_hdr_crypt(hdr, hash_lock, spa,
	zb, !!(flags & ARC_FILL_NOAUTH));
	if (error == EACCES && (flags & ARC_FILL_IN_PLACE) != 0) {
	return (error);
	} else if (error != 0) {
	if (hash_lock != NULL)
	mutex_enter(hash_lock);
	arc_hdr_set_flags(hdr, ARC_FLAG_IO_ERROR);
	if (hash_lock != NULL)
	mutex_exit(hash_lock);
	return (error);
	}
	}

	/*
	* There is a special case here for dnode blocks which are
	* decrypting their bonus buffers. These blocks may request to
	* be decrypted in-place. This is necessary because there may
	* be many dnodes pointing into this buffer and there is
	* currently no method to synchronize replacing the backing
	* b_data buffer and updating all of the pointers. Here we use
	* the hash lock to ensure there are no races. If the need
	* arises for other types to be decrypted in-place, they must
	* add handling here as well.
	*/
	if ((flags & ARC_FILL_IN_PLACE) != 0) {
	ASSERT(!hdr_compressed);
	ASSERT(!compressed);
	ASSERT(!encrypted);

	if (HDR_ENCRYPTED(hdr) && ARC_BUF_ENCRYPTED(buf)) {
	ASSERT3U(hdr->b_crypt_hdr.b_ot, ==, DMU_OT_DNODE);

	if (hash_lock != NULL)
	mutex_enter(hash_lock);
	arc_buf_untransform_in_place(buf);
	if (hash_lock != NULL)
	mutex_exit(hash_lock);

	/* Compute the hdr's checksum if necessary */
	arc_cksum_compute(buf);
	}

	return (0);
	}

	if (hdr_compressed == compressed) {
	if (ARC_BUF_SHARED(buf)) {
	ASSERT(arc_buf_is_shared(buf));
	} else {
	abd_copy_to_buf(buf->b_data, hdr->b_l1hdr.b_pabd,
	arc_buf_size(buf));
	}
	} else {
	ASSERT(hdr_compressed);
	ASSERT(!compressed);

	/*
	* If the buf is sharing its data with the hdr, unlink it and
	* allocate a new data buffer for the buf.
	*/
	if (ARC_BUF_SHARED(buf)) {
	ASSERTF(ARC_BUF_COMPRESSED(buf),
	"buf %p was uncompressed", buf);

	/* We need to give the buf its own b_data */
	buf->b_flags &= ~ARC_BUF_FLAG_SHARED;
	buf->b_data =
	arc_get_data_buf(hdr, HDR_GET_LSIZE(hdr), buf);
	arc_hdr_clear_flags(hdr, ARC_FLAG_SHARED_DATA);

	/* Previously overhead was 0; just add new overhead */
	ARCSTAT_INCR(arcstat_overhead_size, HDR_GET_LSIZE(hdr));
	} else if (ARC_BUF_COMPRESSED(buf)) {
	ASSERT(!arc_buf_is_shared(buf));

	/* We need to reallocate the buf's b_data */
	arc_free_data_buf(hdr, buf->b_data, HDR_GET_PSIZE(hdr),
	buf);
	buf->b_data =
	arc_get_data_buf(hdr, HDR_GET_LSIZE(hdr), buf);

	/* We increased the size of b_data; update overhead */
	ARCSTAT_INCR(arcstat_overhead_size,
	HDR_GET_LSIZE(hdr) - HDR_GET_PSIZE(hdr));
	}

	/*
	* Regardless of the buf's previous compression settings, it
	* should not be compressed at the end of this function.
	*/
	buf->b_flags &= ~ARC_BUF_FLAG_COMPRESSED;

	/*
	* Try copying the data from another buf which already has a
	* decompressed version. If that's not possible, it's time to
	* bite the bullet and decompress the data from the hdr.
	*/
	if (arc_buf_try_copy_decompressed_data(buf)) {
	/* Skip byteswapping and checksumming (already done) */
	return (0);
	} else {
	error = zio_decompress_data(HDR_GET_COMPRESS(hdr),
	hdr->b_l1hdr.b_pabd, buf->b_data,
	HDR_GET_PSIZE(hdr), HDR_GET_LSIZE(hdr),
	&hdr->b_complevel);

	/*
	* Absent hardware errors or software bugs, this should
	* be impossible, but log it anyway so we can debug it.
	*/
	if (error != 0) {
	zfs_dbgmsg(
	"hdr %px, compress %d, psize %d, lsize %d",
	hdr, arc_hdr_get_compress(hdr),
	HDR_GET_PSIZE(hdr), HDR_GET_LSIZE(hdr));
	if (hash_lock != NULL)
	mutex_enter(hash_lock);
	arc_hdr_set_flags(hdr, ARC_FLAG_IO_ERROR);
	if (hash_lock != NULL)
	mutex_exit(hash_lock);
	return (SET_ERROR(EIO));
	}
	}
	}

	byteswap:
	/* Byteswap the buf's data if necessary */
	if (bswap != DMU_BSWAP_NUMFUNCS) {
	ASSERT(!HDR_SHARED_DATA(hdr));
	ASSERT3U(bswap, <, DMU_BSWAP_NUMFUNCS);
	dmu_ot_byteswap[bswap].ob_func(buf->b_data, HDR_GET_LSIZE(hdr));
	}

	/* Compute the hdr's checksum if necessary */
	arc_cksum_compute(buf);

	return (0);
	}

	/*
	* If this function is being called to decrypt an encrypted buffer or verify an
	* authenticated one, the key must be loaded and a mapping must be made
	* available in the keystore via spa_keystore_create_mapping() or one of its
	* callers.
	*/
	int
	arc_untransform(arc_buf_t buf, spa_t spa, const zbookmark_phys_t *zb,
	boolean_t in_place)
	{
	int ret;
	arc_fill_flags_t flags = 0;

	if (in_place)
	flags \|= ARC_FILL_IN_PLACE;

	ret = arc_buf_fill(buf, spa, zb, flags);
	if (ret == ECKSUM) {
	/*
	* Convert authentication and decryption errors to EIO
	* (and generate an ereport) before leaving the ARC.
	*/
	ret = SET_ERROR(EIO);
	spa_log_error(spa, zb, buf->b_hdr->b_birth);
	(void) zfs_ereport_post(FM_EREPORT_ZFS_AUTHENTICATION,
	spa, NULL, zb, NULL, 0);
	}

	return (ret);
	}

	/*
	* Increment the amount of evictable space in the arc_state_t's refcount.
	* We account for the space used by the hdr and the arc buf individually
	* so that we can add and remove them from the refcount individually.
	*/
	static void
	arc_evictable_space_increment(arc_buf_hdr_t hdr, arc_state_t state)
	{
	arc_buf_contents_t type = arc_buf_type(hdr);

	ASSERT(HDR_HAS_L1HDR(hdr));

	if (GHOST_STATE(state)) {
	ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
	ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
	ASSERT(!HDR_HAS_RABD(hdr));
	(void) zfs_refcount_add_many(&state->arcs_esize[type],
	HDR_GET_LSIZE(hdr), hdr);
	return;
	}

	if (hdr->b_l1hdr.b_pabd != NULL) {
	(void) zfs_refcount_add_many(&state->arcs_esize[type],
	arc_hdr_size(hdr), hdr);
	}
	if (HDR_HAS_RABD(hdr)) {
	(void) zfs_refcount_add_many(&state->arcs_esize[type],
	HDR_GET_PSIZE(hdr), hdr);
	}

	for (arc_buf_t *buf = hdr->b_l1hdr.b_buf; buf != NULL;
	buf = buf->b_next) {
	if (ARC_BUF_SHARED(buf))
	continue;
	(void) zfs_refcount_add_many(&state->arcs_esize[type],
	arc_buf_size(buf), buf);
	}
	}

	/*
	* Decrement the amount of evictable space in the arc_state_t's refcount.
	* We account for the space used by the hdr and the arc buf individually
	* so that we can add and remove them from the refcount individually.
	*/
	static void
	arc_evictable_space_decrement(arc_buf_hdr_t hdr, arc_state_t state)
	{
	arc_buf_contents_t type = arc_buf_type(hdr);

	ASSERT(HDR_HAS_L1HDR(hdr));

	if (GHOST_STATE(state)) {
	ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
	ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
	ASSERT(!HDR_HAS_RABD(hdr));
	(void) zfs_refcount_remove_many(&state->arcs_esize[type],
	HDR_GET_LSIZE(hdr), hdr);
	return;
	}

	if (hdr->b_l1hdr.b_pabd != NULL) {
	(void) zfs_refcount_remove_many(&state->arcs_esize[type],
	arc_hdr_size(hdr), hdr);
	}
	if (HDR_HAS_RABD(hdr)) {
	(void) zfs_refcount_remove_many(&state->arcs_esize[type],
	HDR_GET_PSIZE(hdr), hdr);
	}

	for (arc_buf_t *buf = hdr->b_l1hdr.b_buf; buf != NULL;
	buf = buf->b_next) {
	if (ARC_BUF_SHARED(buf))
	continue;
	(void) zfs_refcount_remove_many(&state->arcs_esize[type],
	arc_buf_size(buf), buf);
	}
	}

	/*
	* Add a reference to this hdr indicating that someone is actively
	* referencing that memory. When the refcount transitions from 0 to 1,
	* we remove it from the respective arc_state_t list to indicate that
	* it is not evictable.
	*/
	static void
	add_reference(arc_buf_hdr_t hdr, const void tag)
	{
	arc_state_t *state = hdr->b_l1hdr.b_state;

	ASSERT(HDR_HAS_L1HDR(hdr));
	if (!HDR_EMPTY(hdr) && !MUTEX_HELD(HDR_LOCK(hdr))) {
	ASSERT(state == arc_anon);
	ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
	ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
	}

	if ((zfs_refcount_add(&hdr->b_l1hdr.b_refcnt, tag) == 1) &&
	state != arc_anon && state != arc_l2c_only) {
	/* We don't use the L2-only state list. */
	multilist_remove(&state->arcs_list[arc_buf_type(hdr)], hdr);
	arc_evictable_space_decrement(hdr, state);
	}
	}

	/*
	* Remove a reference from this hdr. When the reference transitions from
	* 1 to 0 and we're not anonymous, then we add this hdr to the arc_state_t's
	* list making it eligible for eviction.
	*/
	static int
	remove_reference(arc_buf_hdr_t hdr, const void tag)
	{
	int cnt;
	arc_state_t *state = hdr->b_l1hdr.b_state;

	ASSERT(HDR_HAS_L1HDR(hdr));
	ASSERT(state == arc_anon \|\| MUTEX_HELD(HDR_LOCK(hdr)));
	ASSERT(!GHOST_STATE(state)); /* arc_l2c_only counts as a ghost. */

	if ((cnt = zfs_refcount_remove(&hdr->b_l1hdr.b_refcnt, tag)) != 0)
	return (cnt);

	if (state == arc_anon) {
	arc_hdr_destroy(hdr);
	return (0);
	}
	if (state == arc_uncached && !HDR_PREFETCH(hdr)) {
	arc_change_state(arc_anon, hdr);
	arc_hdr_destroy(hdr);
	return (0);
	}
	multilist_insert(&state->arcs_list[arc_buf_type(hdr)], hdr);
	arc_evictable_space_increment(hdr, state);
	return (0);
	}

	/*
	* Returns detailed information about a specific arc buffer. When the
	* state_index argument is set the function will calculate the arc header
	* list position for its arc state. Since this requires a linear traversal
	* callers are strongly encourage not to do this. However, it can be helpful
	* for targeted analysis so the functionality is provided.
	*/
	void
	arc_buf_info(arc_buf_t ab, arc_buf_info_t abi, int state_index)
	{
	(void) state_index;
	arc_buf_hdr_t *hdr = ab->b_hdr;
	l1arc_buf_hdr_t *l1hdr = NULL;
	l2arc_buf_hdr_t *l2hdr = NULL;
	arc_state_t *state = NULL;

	memset(abi, 0, sizeof (arc_buf_info_t));

	if (hdr == NULL)
	return;

	abi->abi_flags = hdr->b_flags;

	if (HDR_HAS_L1HDR(hdr)) {
	l1hdr = &hdr->b_l1hdr;
	state = l1hdr->b_state;
	}
	if (HDR_HAS_L2HDR(hdr))
	l2hdr = &hdr->b_l2hdr;

	if (l1hdr) {
	abi->abi_bufcnt = 0;
	for (arc_buf_t *buf = l1hdr->b_buf; buf; buf = buf->b_next)
	abi->abi_bufcnt++;
	abi->abi_access = l1hdr->b_arc_access;
	abi->abi_mru_hits = l1hdr->b_mru_hits;
	abi->abi_mru_ghost_hits = l1hdr->b_mru_ghost_hits;
	abi->abi_mfu_hits = l1hdr->b_mfu_hits;
	abi->abi_mfu_ghost_hits = l1hdr->b_mfu_ghost_hits;
	abi->abi_holds = zfs_refcount_count(&l1hdr->b_refcnt);
	}

	if (l2hdr) {
	abi->abi_l2arc_dattr = l2hdr->b_daddr;
	abi->abi_l2arc_hits = l2hdr->b_hits;
	}

	abi->abi_state_type = state ? state->arcs_state : ARC_STATE_ANON;
	abi->abi_state_contents = arc_buf_type(hdr);
	abi->abi_size = arc_hdr_size(hdr);
	}

	/*
	* Move the supplied buffer to the indicated state. The hash lock
	* for the buffer must be held by the caller.
	*/
	static void
	arc_change_state(arc_state_t new_state, arc_buf_hdr_t hdr)
	{
	arc_state_t *old_state;
	int64_t refcnt;
	boolean_t update_old, update_new;
	arc_buf_contents_t type = arc_buf_type(hdr);

	/*
	* We almost always have an L1 hdr here, since we call arc_hdr_realloc()
	* in arc_read() when bringing a buffer out of the L2ARC. However, the
	* L1 hdr doesn't always exist when we change state to arc_anon before
	* destroying a header, in which case reallocating to add the L1 hdr is
	* pointless.
	*/
	if (HDR_HAS_L1HDR(hdr)) {
	old_state = hdr->b_l1hdr.b_state;
	refcnt = zfs_refcount_count(&hdr->b_l1hdr.b_refcnt);
	update_old = (hdr->b_l1hdr.b_buf != NULL \|\|
	hdr->b_l1hdr.b_pabd != NULL \|\| HDR_HAS_RABD(hdr));

	IMPLY(GHOST_STATE(old_state), hdr->b_l1hdr.b_buf == NULL);
	IMPLY(GHOST_STATE(new_state), hdr->b_l1hdr.b_buf == NULL);
	IMPLY(old_state == arc_anon, hdr->b_l1hdr.b_buf == NULL \|\|
	ARC_BUF_LAST(hdr->b_l1hdr.b_buf));
	} else {
	old_state = arc_l2c_only;
	refcnt = 0;
	update_old = B_FALSE;
	}
	update_new = update_old;
	if (GHOST_STATE(old_state))
	update_old = B_TRUE;
	if (GHOST_STATE(new_state))
	update_new = B_TRUE;

	ASSERT(MUTEX_HELD(HDR_LOCK(hdr)));
	ASSERT3P(new_state, !=, old_state);

	/*
	* If this buffer is evictable, transfer it from the
	* old state list to the new state list.
	*/
	if (refcnt == 0) {
	if (old_state != arc_anon && old_state != arc_l2c_only) {
	ASSERT(HDR_HAS_L1HDR(hdr));
	/* remove_reference() saves on insert. */
	if (multilist_link_active(&hdr->b_l1hdr.b_arc_node)) {
	multilist_remove(&old_state->arcs_list[type],
	hdr);
	arc_evictable_space_decrement(hdr, old_state);
	}
	}
	if (new_state != arc_anon && new_state != arc_l2c_only) {
	/*
	* An L1 header always exists here, since if we're
	* moving to some L1-cached state (i.e. not l2c_only or
	* anonymous), we realloc the header to add an L1hdr
	* beforehand.
	*/
	ASSERT(HDR_HAS_L1HDR(hdr));
	multilist_insert(&new_state->arcs_list[type], hdr);
	arc_evictable_space_increment(hdr, new_state);
	}
	}

	ASSERT(!HDR_EMPTY(hdr));
	if (new_state == arc_anon && HDR_IN_HASH_TABLE(hdr))
	buf_hash_remove(hdr);

	/* adjust state sizes (ignore arc_l2c_only) */

	if (update_new && new_state != arc_l2c_only) {
	ASSERT(HDR_HAS_L1HDR(hdr));
	if (GHOST_STATE(new_state)) {

	/*
	* When moving a header to a ghost state, we first
	* remove all arc buffers. Thus, we'll have no arc
	* buffer to use for the reference. As a result, we
	* use the arc header pointer for the reference.
	*/
	(void) zfs_refcount_add_many(
	&new_state->arcs_size[type],
	HDR_GET_LSIZE(hdr), hdr);
	ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
	ASSERT(!HDR_HAS_RABD(hdr));
	} else {

	/*
	* Each individual buffer holds a unique reference,
	* thus we must remove each of these references one
	* at a time.
	*/
	for (arc_buf_t *buf = hdr->b_l1hdr.b_buf; buf != NULL;
	buf = buf->b_next) {

	/*
	* When the arc_buf_t is sharing the data
	* block with the hdr, the owner of the
	* reference belongs to the hdr. Only
	* add to the refcount if the arc_buf_t is
	* not shared.
	*/
	if (ARC_BUF_SHARED(buf))
	continue;

	(void) zfs_refcount_add_many(
	&new_state->arcs_size[type],
	arc_buf_size(buf), buf);
	}

	if (hdr->b_l1hdr.b_pabd != NULL) {
	(void) zfs_refcount_add_many(
	&new_state->arcs_size[type],
	arc_hdr_size(hdr), hdr);
	}

	if (HDR_HAS_RABD(hdr)) {
	(void) zfs_refcount_add_many(
	&new_state->arcs_size[type],
	HDR_GET_PSIZE(hdr), hdr);
	}
	}
	}

	if (update_old && old_state != arc_l2c_only) {
	ASSERT(HDR_HAS_L1HDR(hdr));
	if (GHOST_STATE(old_state)) {
	ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
	ASSERT(!HDR_HAS_RABD(hdr));

	/*
	* When moving a header off of a ghost state,
	* the header will not contain any arc buffers.
	* We use the arc header pointer for the reference
	* which is exactly what we did when we put the
	* header on the ghost state.
	*/

	(void) zfs_refcount_remove_many(
	&old_state->arcs_size[type],
	HDR_GET_LSIZE(hdr), hdr);
	} else {

	/*
	* Each individual buffer holds a unique reference,
	* thus we must remove each of these references one
	* at a time.
	*/
	for (arc_buf_t *buf = hdr->b_l1hdr.b_buf; buf != NULL;
	buf = buf->b_next) {

	/*
	* When the arc_buf_t is sharing the data
	* block with the hdr, the owner of the
	* reference belongs to the hdr. Only
	* add to the refcount if the arc_buf_t is
	* not shared.
	*/
	if (ARC_BUF_SHARED(buf))
	continue;

	(void) zfs_refcount_remove_many(
	&old_state->arcs_size[type],
	arc_buf_size(buf), buf);
	}
	ASSERT(hdr->b_l1hdr.b_pabd != NULL \|\|
	HDR_HAS_RABD(hdr));

	if (hdr->b_l1hdr.b_pabd != NULL) {
	(void) zfs_refcount_remove_many(
	&old_state->arcs_size[type],
	arc_hdr_size(hdr), hdr);
	}

	if (HDR_HAS_RABD(hdr)) {
	(void) zfs_refcount_remove_many(
	&old_state->arcs_size[type],
	HDR_GET_PSIZE(hdr), hdr);
	}
	}
	}

	if (HDR_HAS_L1HDR(hdr)) {
	hdr->b_l1hdr.b_state = new_state;

	if (HDR_HAS_L2HDR(hdr) && new_state != arc_l2c_only) {
	l2arc_hdr_arcstats_decrement_state(hdr);
	hdr->b_l2hdr.b_arcs_state = new_state->arcs_state;
	l2arc_hdr_arcstats_increment_state(hdr);
	}
	}
	}

	void
	arc_space_consume(uint64_t space, arc_space_type_t type)
	{
	ASSERT(type >= 0 && type < ARC_SPACE_NUMTYPES);

	switch (type) {
	default:
	break;
	case ARC_SPACE_DATA:
	ARCSTAT_INCR(arcstat_data_size, space);
	break;
	case ARC_SPACE_META:
	ARCSTAT_INCR(arcstat_metadata_size, space);
	break;
	case ARC_SPACE_BONUS:
	ARCSTAT_INCR(arcstat_bonus_size, space);
	break;
	case ARC_SPACE_DNODE:
	ARCSTAT_INCR(arcstat_dnode_size, space);
	break;
	case ARC_SPACE_DBUF:
	ARCSTAT_INCR(arcstat_dbuf_size, space);
	break;
	case ARC_SPACE_HDRS:
	ARCSTAT_INCR(arcstat_hdr_size, space);
	break;
	case ARC_SPACE_L2HDRS:
	aggsum_add(&arc_sums.arcstat_l2_hdr_size, space);
	break;
	case ARC_SPACE_ABD_CHUNK_WASTE:
	/*
	* Note: this includes space wasted by all scatter ABD's, not
	* just those allocated by the ARC. But the vast majority of
	* scatter ABD's come from the ARC, because other users are
	* very short-lived.
	*/
	ARCSTAT_INCR(arcstat_abd_chunk_waste_size, space);
	break;
	}

	if (type != ARC_SPACE_DATA && type != ARC_SPACE_ABD_CHUNK_WASTE)
	ARCSTAT_INCR(arcstat_meta_used, space);

	aggsum_add(&arc_sums.arcstat_size, space);
	}

	void
	arc_space_return(uint64_t space, arc_space_type_t type)
	{
	ASSERT(type >= 0 && type < ARC_SPACE_NUMTYPES);

	switch (type) {
	default:
	break;
	case ARC_SPACE_DATA:
	ARCSTAT_INCR(arcstat_data_size, -space);
	break;
	case ARC_SPACE_META:
	ARCSTAT_INCR(arcstat_metadata_size, -space);
	break;
	case ARC_SPACE_BONUS:
	ARCSTAT_INCR(arcstat_bonus_size, -space);
	break;
	case ARC_SPACE_DNODE:
	ARCSTAT_INCR(arcstat_dnode_size, -space);
	break;
	case ARC_SPACE_DBUF:
	ARCSTAT_INCR(arcstat_dbuf_size, -space);
	break;
	case ARC_SPACE_HDRS:
	ARCSTAT_INCR(arcstat_hdr_size, -space);
	break;
	case ARC_SPACE_L2HDRS:
	aggsum_add(&arc_sums.arcstat_l2_hdr_size, -space);
	break;
	case ARC_SPACE_ABD_CHUNK_WASTE:
	ARCSTAT_INCR(arcstat_abd_chunk_waste_size, -space);
	break;
	}

	if (type != ARC_SPACE_DATA && type != ARC_SPACE_ABD_CHUNK_WASTE)
	ARCSTAT_INCR(arcstat_meta_used, -space);

	ASSERT(aggsum_compare(&arc_sums.arcstat_size, space) >= 0);
	aggsum_add(&arc_sums.arcstat_size, -space);
	}

	/*
	* Given a hdr and a buf, returns whether that buf can share its b_data buffer
	* with the hdr's b_pabd.
	*/
	static boolean_t
	arc_can_share(arc_buf_hdr_t hdr, arc_buf_t buf)
	{
	/*
	* The criteria for sharing a hdr's data are:
	* 1. the buffer is not encrypted
	* 2. the hdr's compression matches the buf's compression
	* 3. the hdr doesn't need to be byteswapped
	* 4. the hdr isn't already being shared
	* 5. the buf is either compressed or it is the last buf in the hdr list
	*
	* Criterion #5 maintains the invariant that shared uncompressed
	* bufs must be the final buf in the hdr's b_buf list. Reading this, you
	* might ask, "if a compressed buf is allocated first, won't that be the
	* last thing in the list?", but in that case it's impossible to create
	* a shared uncompressed buf anyway (because the hdr must be compressed
	* to have the compressed buf). You might also think that #3 is
	* sufficient to make this guarantee, however it's possible
	* (specifically in the rare L2ARC write race mentioned in
	* arc_buf_alloc_impl()) there will be an existing uncompressed buf that
	* is shareable, but wasn't at the time of its allocation. Rather than
	* allow a new shared uncompressed buf to be created and then shuffle
	* the list around to make it the last element, this simply disallows
	* sharing if the new buf isn't the first to be added.
	*/
	ASSERT3P(buf->b_hdr, ==, hdr);
	boolean_t hdr_compressed =
	arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF;
	boolean_t buf_compressed = ARC_BUF_COMPRESSED(buf) != 0;
	return (!ARC_BUF_ENCRYPTED(buf) &&
	buf_compressed == hdr_compressed &&
	hdr->b_l1hdr.b_byteswap == DMU_BSWAP_NUMFUNCS &&
	!HDR_SHARED_DATA(hdr) &&
	(ARC_BUF_LAST(buf) \|\| ARC_BUF_COMPRESSED(buf)));
	}

	/*
	* Allocate a buf for this hdr. If you care about the data that's in the hdr,
	* or if you want a compressed buffer, pass those flags in. Returns 0 if the
	* copy was made successfully, or an error code otherwise.
	*/
	static int
	arc_buf_alloc_impl(arc_buf_hdr_t hdr, spa_t spa, const zbookmark_phys_t *zb,
	const void *tag, boolean_t encrypted, boolean_t compressed,
	boolean_t noauth, boolean_t fill, arc_buf_t **ret)
	{
	arc_buf_t *buf;
	arc_fill_flags_t flags = ARC_FILL_LOCKED;

	ASSERT(HDR_HAS_L1HDR(hdr));
	ASSERT3U(HDR_GET_LSIZE(hdr), >, 0);
	VERIFY(hdr->b_type == ARC_BUFC_DATA \|\|
	hdr->b_type == ARC_BUFC_METADATA);
	ASSERT3P(ret, !=, NULL);
	ASSERT3P(*ret, ==, NULL);
	IMPLY(encrypted, compressed);

	buf = *ret = kmem_cache_alloc(buf_cache, KM_PUSHPAGE);
	buf->b_hdr = hdr;
	buf->b_data = NULL;
	buf->b_next = hdr->b_l1hdr.b_buf;
	buf->b_flags = 0;

	add_reference(hdr, tag);

	/*
	* We're about to change the hdr's b_flags. We must either
	* hold the hash_lock or be undiscoverable.
	*/
	ASSERT(HDR_EMPTY_OR_LOCKED(hdr));

	/*
	* Only honor requests for compressed bufs if the hdr is actually
	* compressed. This must be overridden if the buffer is encrypted since
	* encrypted buffers cannot be decompressed.
	*/
	if (encrypted) {
	buf->b_flags \|= ARC_BUF_FLAG_COMPRESSED;
	buf->b_flags \|= ARC_BUF_FLAG_ENCRYPTED;
	flags \|= ARC_FILL_COMPRESSED \| ARC_FILL_ENCRYPTED;
	} else if (compressed &&
	arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF) {
	buf->b_flags \|= ARC_BUF_FLAG_COMPRESSED;
	flags \|= ARC_FILL_COMPRESSED;
	}

	if (noauth) {
	ASSERT0(encrypted);
	flags \|= ARC_FILL_NOAUTH;
	}

	/*
	* If the hdr's data can be shared then we share the data buffer and
	* set the appropriate bit in the hdr's b_flags to indicate the hdr is
	* sharing it's b_pabd with the arc_buf_t. Otherwise, we allocate a new
	* buffer to store the buf's data.
	*
	* There are two additional restrictions here because we're sharing
	* hdr -> buf instead of the usual buf -> hdr. First, the hdr can't be
	* actively involved in an L2ARC write, because if this buf is used by
	* an arc_write() then the hdr's data buffer will be released when the
	* write completes, even though the L2ARC write might still be using it.
	* Second, the hdr's ABD must be linear so that the buf's user doesn't
	* need to be ABD-aware. It must be allocated via
	* zio_[data_]buf_alloc(), not as a page, because we need to be able
	* to abd_release_ownership_of_buf(), which isn't allowed on "linear
	* page" buffers because the ABD code needs to handle freeing them
	* specially.
	*/
	boolean_t can_share = arc_can_share(hdr, buf) &&
	!HDR_L2_WRITING(hdr) &&
	hdr->b_l1hdr.b_pabd != NULL &&
	abd_is_linear(hdr->b_l1hdr.b_pabd) &&
	!abd_is_linear_page(hdr->b_l1hdr.b_pabd);

	/* Set up b_data and sharing */
	if (can_share) {
	buf->b_data = abd_to_buf(hdr->b_l1hdr.b_pabd);
	buf->b_flags \|= ARC_BUF_FLAG_SHARED;
	arc_hdr_set_flags(hdr, ARC_FLAG_SHARED_DATA);
	} else {
	buf->b_data =
	arc_get_data_buf(hdr, arc_buf_size(buf), buf);
	ARCSTAT_INCR(arcstat_overhead_size, arc_buf_size(buf));
	}
	VERIFY3P(buf->b_data, !=, NULL);

	hdr->b_l1hdr.b_buf = buf;

	/*
	* If the user wants the data from the hdr, we need to either copy or
	* decompress the data.
	*/
	if (fill) {
	ASSERT3P(zb, !=, NULL);
	return (arc_buf_fill(buf, spa, zb, flags));
	}

	return (0);
	}

	static const char *arc_onloan_tag = "onloan";

	static inline void
	arc_loaned_bytes_update(int64_t delta)
	{
	atomic_add_64(&arc_loaned_bytes, delta);

	/* assert that it did not wrap around */
	ASSERT3S(atomic_add_64_nv(&arc_loaned_bytes, 0), >=, 0);
	}

	/*
	* Loan out an anonymous arc buffer. Loaned buffers are not counted as in
	* flight data by arc_tempreserve_space() until they are "returned". Loaned
	* buffers must be returned to the arc before they can be used by the DMU or
	* freed.
	*/
	arc_buf_t *
	arc_loan_buf(spa_t *spa, boolean_t is_metadata, int size)
	{
	arc_buf_t *buf = arc_alloc_buf(spa, arc_onloan_tag,
	is_metadata ? ARC_BUFC_METADATA : ARC_BUFC_DATA, size);

	arc_loaned_bytes_update(arc_buf_size(buf));

	return (buf);
	}

	arc_buf_t *
	arc_loan_compressed_buf(spa_t *spa, uint64_t psize, uint64_t lsize,
	enum zio_compress compression_type, uint8_t complevel)
	{
	arc_buf_t *buf = arc_alloc_compressed_buf(spa, arc_onloan_tag,
	psize, lsize, compression_type, complevel);

	arc_loaned_bytes_update(arc_buf_size(buf));

	return (buf);
	}

	arc_buf_t *
	arc_loan_raw_buf(spa_t *spa, uint64_t dsobj, boolean_t byteorder,
	const uint8_t salt, const uint8_t iv, const uint8_t *mac,
	dmu_object_type_t ot, uint64_t psize, uint64_t lsize,
	enum zio_compress compression_type, uint8_t complevel)
	{
	arc_buf_t *buf = arc_alloc_raw_buf(spa, arc_onloan_tag, dsobj,
	byteorder, salt, iv, mac, ot, psize, lsize, compression_type,
	complevel);

	atomic_add_64(&arc_loaned_bytes, psize);
	return (buf);
	}


	/*
	* Return a loaned arc buffer to the arc.
	*/
	void
	arc_return_buf(arc_buf_t buf, const void tag)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;

	ASSERT3P(buf->b_data, !=, NULL);
	ASSERT(HDR_HAS_L1HDR(hdr));
	(void) zfs_refcount_add(&hdr->b_l1hdr.b_refcnt, tag);
	(void) zfs_refcount_remove(&hdr->b_l1hdr.b_refcnt, arc_onloan_tag);

	arc_loaned_bytes_update(-arc_buf_size(buf));
	}

	/* Detach an arc_buf from a dbuf (tag) */
	void
	arc_loan_inuse_buf(arc_buf_t buf, const void tag)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;

	ASSERT3P(buf->b_data, !=, NULL);
	ASSERT(HDR_HAS_L1HDR(hdr));
	(void) zfs_refcount_add(&hdr->b_l1hdr.b_refcnt, arc_onloan_tag);
	(void) zfs_refcount_remove(&hdr->b_l1hdr.b_refcnt, tag);

	arc_loaned_bytes_update(arc_buf_size(buf));
	}

	static void
	l2arc_free_abd_on_write(abd_t *abd, size_t size, arc_buf_contents_t type)
	{
	l2arc_data_free_t df = kmem_alloc(sizeof (df), KM_SLEEP);

	df->l2df_abd = abd;
	df->l2df_size = size;
	df->l2df_type = type;
	mutex_enter(&l2arc_free_on_write_mtx);
	list_insert_head(l2arc_free_on_write, df);
	mutex_exit(&l2arc_free_on_write_mtx);
	}

	static void
	arc_hdr_free_on_write(arc_buf_hdr_t *hdr, boolean_t free_rdata)
	{
	arc_state_t *state = hdr->b_l1hdr.b_state;
	arc_buf_contents_t type = arc_buf_type(hdr);
	uint64_t size = (free_rdata) ? HDR_GET_PSIZE(hdr) : arc_hdr_size(hdr);

	/* protected by hash lock, if in the hash table */
	if (multilist_link_active(&hdr->b_l1hdr.b_arc_node)) {
	ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
	ASSERT(state != arc_anon && state != arc_l2c_only);

	(void) zfs_refcount_remove_many(&state->arcs_esize[type],
	size, hdr);
	}
	(void) zfs_refcount_remove_many(&state->arcs_size[type], size, hdr);
	if (type == ARC_BUFC_METADATA) {
	arc_space_return(size, ARC_SPACE_META);
	} else {
	ASSERT(type == ARC_BUFC_DATA);
	arc_space_return(size, ARC_SPACE_DATA);
	}

	if (free_rdata) {
	l2arc_free_abd_on_write(hdr->b_crypt_hdr.b_rabd, size, type);
	} else {
	l2arc_free_abd_on_write(hdr->b_l1hdr.b_pabd, size, type);
	}
	}

	/*
	* Share the arc_buf_t's data with the hdr. Whenever we are sharing the
	* data buffer, we transfer the refcount ownership to the hdr and update
	* the appropriate kstats.
	*/
	static void
	arc_share_buf(arc_buf_hdr_t hdr, arc_buf_t buf)
	{
	ASSERT(arc_can_share(hdr, buf));
	ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
	ASSERT(!ARC_BUF_ENCRYPTED(buf));
	ASSERT(HDR_EMPTY_OR_LOCKED(hdr));

	/*
	* Start sharing the data buffer. We transfer the
	* refcount ownership to the hdr since it always owns
	* the refcount whenever an arc_buf_t is shared.
	*/
	zfs_refcount_transfer_ownership_many(
	&hdr->b_l1hdr.b_state->arcs_size[arc_buf_type(hdr)],
	arc_hdr_size(hdr), buf, hdr);
	hdr->b_l1hdr.b_pabd = abd_get_from_buf(buf->b_data, arc_buf_size(buf));
	abd_take_ownership_of_buf(hdr->b_l1hdr.b_pabd,
	HDR_ISTYPE_METADATA(hdr));
	arc_hdr_set_flags(hdr, ARC_FLAG_SHARED_DATA);
	buf->b_flags \|= ARC_BUF_FLAG_SHARED;

	/*
	* Since we've transferred ownership to the hdr we need
	* to increment its compressed and uncompressed kstats and
	* decrement the overhead size.
	*/
	ARCSTAT_INCR(arcstat_compressed_size, arc_hdr_size(hdr));
	ARCSTAT_INCR(arcstat_uncompressed_size, HDR_GET_LSIZE(hdr));
	ARCSTAT_INCR(arcstat_overhead_size, -arc_buf_size(buf));
	}

	static void
	arc_unshare_buf(arc_buf_hdr_t hdr, arc_buf_t buf)
	{
	ASSERT(arc_buf_is_shared(buf));
	ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);
	ASSERT(HDR_EMPTY_OR_LOCKED(hdr));

	/*
	* We are no longer sharing this buffer so we need
	* to transfer its ownership to the rightful owner.
	*/
	zfs_refcount_transfer_ownership_many(
	&hdr->b_l1hdr.b_state->arcs_size[arc_buf_type(hdr)],
	arc_hdr_size(hdr), hdr, buf);
	arc_hdr_clear_flags(hdr, ARC_FLAG_SHARED_DATA);
	abd_release_ownership_of_buf(hdr->b_l1hdr.b_pabd);
	abd_free(hdr->b_l1hdr.b_pabd);
	hdr->b_l1hdr.b_pabd = NULL;
	buf->b_flags &= ~ARC_BUF_FLAG_SHARED;

	/*
	* Since the buffer is no longer shared between
	* the arc buf and the hdr, count it as overhead.
	*/
	ARCSTAT_INCR(arcstat_compressed_size, -arc_hdr_size(hdr));
	ARCSTAT_INCR(arcstat_uncompressed_size, -HDR_GET_LSIZE(hdr));
	ARCSTAT_INCR(arcstat_overhead_size, arc_buf_size(buf));
	}

	/*
	* Remove an arc_buf_t from the hdr's buf list and return the last
	* arc_buf_t on the list. If no buffers remain on the list then return
	* NULL.
	*/
	static arc_buf_t *
	arc_buf_remove(arc_buf_hdr_t hdr, arc_buf_t buf)
	{
	ASSERT(HDR_HAS_L1HDR(hdr));
	ASSERT(HDR_EMPTY_OR_LOCKED(hdr));

	arc_buf_t **bufp = &hdr->b_l1hdr.b_buf;
	arc_buf_t *lastbuf = NULL;

	/*
	* Remove the buf from the hdr list and locate the last
	* remaining buffer on the list.
	*/
	while (*bufp != NULL) {
	if (*bufp == buf)
	*bufp = buf->b_next;

	/*
	* If we've removed a buffer in the middle of
	* the list then update the lastbuf and update
	* bufp.
	*/
	if (*bufp != NULL) {
	lastbuf = *bufp;
	bufp = &(*bufp)->b_next;
	}
	}
	buf->b_next = NULL;
	ASSERT3P(lastbuf, !=, buf);
	IMPLY(lastbuf != NULL, ARC_BUF_LAST(lastbuf));

	return (lastbuf);
	}

	/*
	* Free up buf->b_data and pull the arc_buf_t off of the arc_buf_hdr_t's
	* list and free it.
	*/
	static void
	arc_buf_destroy_impl(arc_buf_t *buf)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;

	/*
	* Free up the data associated with the buf but only if we're not
	* sharing this with the hdr. If we are sharing it with the hdr, the
	* hdr is responsible for doing the free.
	*/
	if (buf->b_data != NULL) {
	/*
	* We're about to change the hdr's b_flags. We must either
	* hold the hash_lock or be undiscoverable.
	*/
	ASSERT(HDR_EMPTY_OR_LOCKED(hdr));

	arc_cksum_verify(buf);
	arc_buf_unwatch(buf);

	if (ARC_BUF_SHARED(buf)) {
	arc_hdr_clear_flags(hdr, ARC_FLAG_SHARED_DATA);
	} else {
	ASSERT(!arc_buf_is_shared(buf));
	uint64_t size = arc_buf_size(buf);
	arc_free_data_buf(hdr, buf->b_data, size, buf);
	ARCSTAT_INCR(arcstat_overhead_size, -size);
	}
	buf->b_data = NULL;

	/*
	* If we have no more encrypted buffers and we've already
	* gotten a copy of the decrypted data we can free b_rabd
	* to save some space.
	*/
	if (ARC_BUF_ENCRYPTED(buf) && HDR_HAS_RABD(hdr) &&
	hdr->b_l1hdr.b_pabd != NULL && !HDR_IO_IN_PROGRESS(hdr)) {
	arc_buf_t *b;
	for (b = hdr->b_l1hdr.b_buf; b; b = b->b_next) {
	if (b != buf && ARC_BUF_ENCRYPTED(b))
	break;
	}
	if (b == NULL)
	arc_hdr_free_abd(hdr, B_TRUE);
	}
	}

	arc_buf_t *lastbuf = arc_buf_remove(hdr, buf);

	if (ARC_BUF_SHARED(buf) && !ARC_BUF_COMPRESSED(buf)) {
	/*
	* If the current arc_buf_t is sharing its data buffer with the
	* hdr, then reassign the hdr's b_pabd to share it with the new
	* buffer at the end of the list. The shared buffer is always
	* the last one on the hdr's buffer list.
	*
	* There is an equivalent case for compressed bufs, but since
	* they aren't guaranteed to be the last buf in the list and
	* that is an exceedingly rare case, we just allow that space be
	* wasted temporarily. We must also be careful not to share
	* encrypted buffers, since they cannot be shared.
	*/
	if (lastbuf != NULL && !ARC_BUF_ENCRYPTED(lastbuf)) {
	/* Only one buf can be shared at once */
	ASSERT(!arc_buf_is_shared(lastbuf));
	/* hdr is uncompressed so can't have compressed buf */
	ASSERT(!ARC_BUF_COMPRESSED(lastbuf));

	ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);
	arc_hdr_free_abd(hdr, B_FALSE);

	/*
	* We must setup a new shared block between the
	* last buffer and the hdr. The data would have
	* been allocated by the arc buf so we need to transfer
	* ownership to the hdr since it's now being shared.
	*/
	arc_share_buf(hdr, lastbuf);
	}
	} else if (HDR_SHARED_DATA(hdr)) {
	/*
	* Uncompressed shared buffers are always at the end
	* of the list. Compressed buffers don't have the
	* same requirements. This makes it hard to
	* simply assert that the lastbuf is shared so
	* we rely on the hdr's compression flags to determine
	* if we have a compressed, shared buffer.
	*/
	ASSERT3P(lastbuf, !=, NULL);
	ASSERT(arc_buf_is_shared(lastbuf) \|\|
	arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF);
	}

	/*
	* Free the checksum if we're removing the last uncompressed buf from
	* this hdr.
	*/
	if (!arc_hdr_has_uncompressed_buf(hdr)) {
	arc_cksum_free(hdr);
	}

	/* clean up the buf */
	buf->b_hdr = NULL;
	kmem_cache_free(buf_cache, buf);
	}

	static void
	arc_hdr_alloc_abd(arc_buf_hdr_t *hdr, int alloc_flags)
	{
	uint64_t size;
	boolean_t alloc_rdata = ((alloc_flags & ARC_HDR_ALLOC_RDATA) != 0);

	ASSERT3U(HDR_GET_LSIZE(hdr), >, 0);
	ASSERT(HDR_HAS_L1HDR(hdr));
	ASSERT(!HDR_SHARED_DATA(hdr) \|\| alloc_rdata);
	IMPLY(alloc_rdata, HDR_PROTECTED(hdr));

	if (alloc_rdata) {
	size = HDR_GET_PSIZE(hdr);
	ASSERT3P(hdr->b_crypt_hdr.b_rabd, ==, NULL);
	hdr->b_crypt_hdr.b_rabd = arc_get_data_abd(hdr, size, hdr,
	alloc_flags);
	ASSERT3P(hdr->b_crypt_hdr.b_rabd, !=, NULL);
	ARCSTAT_INCR(arcstat_raw_size, size);
	} else {
	size = arc_hdr_size(hdr);
	ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
	hdr->b_l1hdr.b_pabd = arc_get_data_abd(hdr, size, hdr,
	alloc_flags);
	ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);
	}

	ARCSTAT_INCR(arcstat_compressed_size, size);
	ARCSTAT_INCR(arcstat_uncompressed_size, HDR_GET_LSIZE(hdr));
	}

	static void
	arc_hdr_free_abd(arc_buf_hdr_t *hdr, boolean_t free_rdata)
	{
	uint64_t size = (free_rdata) ? HDR_GET_PSIZE(hdr) : arc_hdr_size(hdr);

	ASSERT(HDR_HAS_L1HDR(hdr));
	ASSERT(hdr->b_l1hdr.b_pabd != NULL \|\| HDR_HAS_RABD(hdr));
	IMPLY(free_rdata, HDR_HAS_RABD(hdr));

	/*
	* If the hdr is currently being written to the l2arc then
	* we defer freeing the data by adding it to the l2arc_free_on_write
	* list. The l2arc will free the data once it's finished
	* writing it to the l2arc device.
	*/
	if (HDR_L2_WRITING(hdr)) {
	arc_hdr_free_on_write(hdr, free_rdata);
	ARCSTAT_BUMP(arcstat_l2_free_on_write);
	} else if (free_rdata) {
	arc_free_data_abd(hdr, hdr->b_crypt_hdr.b_rabd, size, hdr);
	} else {
	arc_free_data_abd(hdr, hdr->b_l1hdr.b_pabd, size, hdr);
	}

	if (free_rdata) {
	hdr->b_crypt_hdr.b_rabd = NULL;
	ARCSTAT_INCR(arcstat_raw_size, -size);
	} else {
	hdr->b_l1hdr.b_pabd = NULL;
	}

	if (hdr->b_l1hdr.b_pabd == NULL && !HDR_HAS_RABD(hdr))
	hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS;

	ARCSTAT_INCR(arcstat_compressed_size, -size);
	ARCSTAT_INCR(arcstat_uncompressed_size, -HDR_GET_LSIZE(hdr));
	}

	/*
	* Allocate empty anonymous ARC header. The header will get its identity
	* assigned and buffers attached later as part of read or write operations.
	*
	* In case of read arc_read() assigns header its identify (b_dva + b_birth),
	* inserts it into ARC hash to become globally visible and allocates physical
	* (b_pabd) or raw (b_rabd) ABD buffer to read into from disk. On disk read
	* completion arc_read_done() allocates ARC buffer(s) as needed, potentially
	* sharing one of them with the physical ABD buffer.
	*
	* In case of write arc_alloc_buf() allocates ARC buffer to be filled with
	* data. Then after compression and/or encryption arc_write_ready() allocates
	* and fills (or potentially shares) physical (b_pabd) or raw (b_rabd) ABD
	* buffer. On disk write completion arc_write_done() assigns the header its
	* new identity (b_dva + b_birth) and inserts into ARC hash.
	*
	* In case of partial overwrite the old data is read first as described. Then
	* arc_release() either allocates new anonymous ARC header and moves the ARC
	* buffer to it, or reuses the old ARC header by discarding its identity and
	* removing it from ARC hash. After buffer modification normal write process
	* follows as described.
	*/
	static arc_buf_hdr_t *
	arc_hdr_alloc(uint64_t spa, int32_t psize, int32_t lsize,
	boolean_t protected, enum zio_compress compression_type, uint8_t complevel,
	arc_buf_contents_t type)
	{
	arc_buf_hdr_t *hdr;

	VERIFY(type == ARC_BUFC_DATA \|\| type == ARC_BUFC_METADATA);
	hdr = kmem_cache_alloc(hdr_full_cache, KM_PUSHPAGE);

	ASSERT(HDR_EMPTY(hdr));
	#ifdef ZFS_DEBUG
	ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL);
	#endif
	HDR_SET_PSIZE(hdr, psize);
	HDR_SET_LSIZE(hdr, lsize);
	hdr->b_spa = spa;
	hdr->b_type = type;
	hdr->b_flags = 0;
	arc_hdr_set_flags(hdr, arc_bufc_to_flags(type) \| ARC_FLAG_HAS_L1HDR);
	arc_hdr_set_compress(hdr, compression_type);
	hdr->b_complevel = complevel;
	if (protected)
	arc_hdr_set_flags(hdr, ARC_FLAG_PROTECTED);

	hdr->b_l1hdr.b_state = arc_anon;
	hdr->b_l1hdr.b_arc_access = 0;
	hdr->b_l1hdr.b_mru_hits = 0;
	hdr->b_l1hdr.b_mru_ghost_hits = 0;
	hdr->b_l1hdr.b_mfu_hits = 0;
	hdr->b_l1hdr.b_mfu_ghost_hits = 0;
	hdr->b_l1hdr.b_buf = NULL;

	ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt));

	return (hdr);
	}

	/*
	* Transition between the two allocation states for the arc_buf_hdr struct.
	* The arc_buf_hdr struct can be allocated with (hdr_full_cache) or without
	* (hdr_l2only_cache) the fields necessary for the L1 cache - the smaller
	* version is used when a cache buffer is only in the L2ARC in order to reduce
	* memory usage.
	*/
	static arc_buf_hdr_t *
	arc_hdr_realloc(arc_buf_hdr_t hdr, kmem_cache_t old, kmem_cache_t *new)
	{
	ASSERT(HDR_HAS_L2HDR(hdr));

	arc_buf_hdr_t *nhdr;
	l2arc_dev_t *dev = hdr->b_l2hdr.b_dev;

	ASSERT((old == hdr_full_cache && new == hdr_l2only_cache) \|\|
	(old == hdr_l2only_cache && new == hdr_full_cache));

	nhdr = kmem_cache_alloc(new, KM_PUSHPAGE);

	ASSERT(MUTEX_HELD(HDR_LOCK(hdr)));
	buf_hash_remove(hdr);

	memcpy(nhdr, hdr, HDR_L2ONLY_SIZE);

	if (new == hdr_full_cache) {
	arc_hdr_set_flags(nhdr, ARC_FLAG_HAS_L1HDR);
	/*
	* arc_access and arc_change_state need to be aware that a
	* header has just come out of L2ARC, so we set its state to
	* l2c_only even though it's about to change.
	*/
	nhdr->b_l1hdr.b_state = arc_l2c_only;

	/* Verify previous threads set to NULL before freeing */
	ASSERT3P(nhdr->b_l1hdr.b_pabd, ==, NULL);
	ASSERT(!HDR_HAS_RABD(hdr));
	} else {
	ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
	#ifdef ZFS_DEBUG
	ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL);
	#endif

	/*
	* If we've reached here, We must have been called from
	* arc_evict_hdr(), as such we should have already been
	* removed from any ghost list we were previously on
	* (which protects us from racing with arc_evict_state),
	* thus no locking is needed during this check.
	*/
	ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node));

	/*
	* A buffer must not be moved into the arc_l2c_only
	* state if it's not finished being written out to the
	* l2arc device. Otherwise, the b_l1hdr.b_pabd field
	* might try to be accessed, even though it was removed.
	*/
	VERIFY(!HDR_L2_WRITING(hdr));
	VERIFY3P(hdr->b_l1hdr.b_pabd, ==, NULL);
	ASSERT(!HDR_HAS_RABD(hdr));

	arc_hdr_clear_flags(nhdr, ARC_FLAG_HAS_L1HDR);
	}
	/*
	* The header has been reallocated so we need to re-insert it into any
	* lists it was on.
	*/
	(void) buf_hash_insert(nhdr, NULL);

	ASSERT(list_link_active(&hdr->b_l2hdr.b_l2node));

	mutex_enter(&dev->l2ad_mtx);

	/*
	* We must place the realloc'ed header back into the list at
	* the same spot. Otherwise, if it's placed earlier in the list,
	* l2arc_write_buffers() could find it during the function's
	* write phase, and try to write it out to the l2arc.
	*/
	list_insert_after(&dev->l2ad_buflist, hdr, nhdr);
	list_remove(&dev->l2ad_buflist, hdr);

	mutex_exit(&dev->l2ad_mtx);

	/*
	* Since we're using the pointer address as the tag when
	* incrementing and decrementing the l2ad_alloc refcount, we
	* must remove the old pointer (that we're about to destroy) and
	* add the new pointer to the refcount. Otherwise we'd remove
	* the wrong pointer address when calling arc_hdr_destroy() later.
	*/

	(void) zfs_refcount_remove_many(&dev->l2ad_alloc,
	arc_hdr_size(hdr), hdr);
	(void) zfs_refcount_add_many(&dev->l2ad_alloc,
	arc_hdr_size(nhdr), nhdr);

	buf_discard_identity(hdr);
	kmem_cache_free(old, hdr);

	return (nhdr);
	}

	/*
	* This function is used by the send / receive code to convert a newly
	* allocated arc_buf_t to one that is suitable for a raw encrypted write. It
	* is also used to allow the root objset block to be updated without altering
	* its embedded MACs. Both block types will always be uncompressed so we do not
	* have to worry about compression type or psize.
	*/
	void
	arc_convert_to_raw(arc_buf_t *buf, uint64_t dsobj, boolean_t byteorder,
	dmu_object_type_t ot, const uint8_t salt, const uint8_t iv,
	const uint8_t *mac)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;

	ASSERT(ot == DMU_OT_DNODE \|\| ot == DMU_OT_OBJSET);
	ASSERT(HDR_HAS_L1HDR(hdr));
	ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon);

	buf->b_flags \|= (ARC_BUF_FLAG_COMPRESSED \| ARC_BUF_FLAG_ENCRYPTED);
	arc_hdr_set_flags(hdr, ARC_FLAG_PROTECTED);
	hdr->b_crypt_hdr.b_dsobj = dsobj;
	hdr->b_crypt_hdr.b_ot = ot;
	hdr->b_l1hdr.b_byteswap = (byteorder == ZFS_HOST_BYTEORDER) ?
	DMU_BSWAP_NUMFUNCS : DMU_OT_BYTESWAP(ot);
	if (!arc_hdr_has_uncompressed_buf(hdr))
	arc_cksum_free(hdr);

	if (salt != NULL)
	memcpy(hdr->b_crypt_hdr.b_salt, salt, ZIO_DATA_SALT_LEN);
	if (iv != NULL)
	memcpy(hdr->b_crypt_hdr.b_iv, iv, ZIO_DATA_IV_LEN);
	if (mac != NULL)
	memcpy(hdr->b_crypt_hdr.b_mac, mac, ZIO_DATA_MAC_LEN);
	}

	/*
	* Allocate a new arc_buf_hdr_t and arc_buf_t and return the buf to the caller.
	* The buf is returned thawed since we expect the consumer to modify it.
	*/
	arc_buf_t *
	arc_alloc_buf(spa_t spa, const void tag, arc_buf_contents_t type,
	int32_t size)
	{
	arc_buf_hdr_t *hdr = arc_hdr_alloc(spa_load_guid(spa), size, size,
	B_FALSE, ZIO_COMPRESS_OFF, 0, type);

	arc_buf_t *buf = NULL;
	VERIFY0(arc_buf_alloc_impl(hdr, spa, NULL, tag, B_FALSE, B_FALSE,
	B_FALSE, B_FALSE, &buf));
	arc_buf_thaw(buf);

	return (buf);
	}

	/*
	* Allocate a compressed buf in the same manner as arc_alloc_buf. Don't use this
	* for bufs containing metadata.
	*/
	arc_buf_t *
	arc_alloc_compressed_buf(spa_t spa, const void tag, uint64_t psize,
	uint64_t lsize, enum zio_compress compression_type, uint8_t complevel)
	{
	ASSERT3U(lsize, >, 0);
	ASSERT3U(lsize, >=, psize);
	ASSERT3U(compression_type, >, ZIO_COMPRESS_OFF);
	ASSERT3U(compression_type, <, ZIO_COMPRESS_FUNCTIONS);

	arc_buf_hdr_t *hdr = arc_hdr_alloc(spa_load_guid(spa), psize, lsize,
	B_FALSE, compression_type, complevel, ARC_BUFC_DATA);

	arc_buf_t *buf = NULL;
	VERIFY0(arc_buf_alloc_impl(hdr, spa, NULL, tag, B_FALSE,
	B_TRUE, B_FALSE, B_FALSE, &buf));
	arc_buf_thaw(buf);

	/*
	* To ensure that the hdr has the correct data in it if we call
	* arc_untransform() on this buf before it's been written to disk,
	* it's easiest if we just set up sharing between the buf and the hdr.
	*/
	arc_share_buf(hdr, buf);

	return (buf);
	}

	arc_buf_t *
	arc_alloc_raw_buf(spa_t spa, const void tag, uint64_t dsobj,
	boolean_t byteorder, const uint8_t salt, const uint8_t iv,
	const uint8_t *mac, dmu_object_type_t ot, uint64_t psize, uint64_t lsize,
	enum zio_compress compression_type, uint8_t complevel)
	{
	arc_buf_hdr_t *hdr;
	arc_buf_t *buf;
	arc_buf_contents_t type = DMU_OT_IS_METADATA(ot) ?
	ARC_BUFC_METADATA : ARC_BUFC_DATA;

	ASSERT3U(lsize, >, 0);
	ASSERT3U(lsize, >=, psize);
	ASSERT3U(compression_type, >=, ZIO_COMPRESS_OFF);
	ASSERT3U(compression_type, <, ZIO_COMPRESS_FUNCTIONS);

	hdr = arc_hdr_alloc(spa_load_guid(spa), psize, lsize, B_TRUE,
	compression_type, complevel, type);

	hdr->b_crypt_hdr.b_dsobj = dsobj;
	hdr->b_crypt_hdr.b_ot = ot;
	hdr->b_l1hdr.b_byteswap = (byteorder == ZFS_HOST_BYTEORDER) ?
	DMU_BSWAP_NUMFUNCS : DMU_OT_BYTESWAP(ot);
	memcpy(hdr->b_crypt_hdr.b_salt, salt, ZIO_DATA_SALT_LEN);
	memcpy(hdr->b_crypt_hdr.b_iv, iv, ZIO_DATA_IV_LEN);
	memcpy(hdr->b_crypt_hdr.b_mac, mac, ZIO_DATA_MAC_LEN);

	/*
	* This buffer will be considered encrypted even if the ot is not an
	* encrypted type. It will become authenticated instead in
	* arc_write_ready().
	*/
	buf = NULL;
	VERIFY0(arc_buf_alloc_impl(hdr, spa, NULL, tag, B_TRUE, B_TRUE,
	B_FALSE, B_FALSE, &buf));
	arc_buf_thaw(buf);

	return (buf);
	}

	static void
	l2arc_hdr_arcstats_update(arc_buf_hdr_t *hdr, boolean_t incr,
	boolean_t state_only)
	{
	l2arc_buf_hdr_t *l2hdr = &hdr->b_l2hdr;
	l2arc_dev_t *dev = l2hdr->b_dev;
	uint64_t lsize = HDR_GET_LSIZE(hdr);
	uint64_t psize = HDR_GET_PSIZE(hdr);
	uint64_t asize = vdev_psize_to_asize(dev->l2ad_vdev, psize);
	arc_buf_contents_t type = hdr->b_type;
	int64_t lsize_s;
	int64_t psize_s;
	int64_t asize_s;

	if (incr) {
	lsize_s = lsize;
	psize_s = psize;
	asize_s = asize;
	} else {
	lsize_s = -lsize;
	psize_s = -psize;
	asize_s = -asize;
	}

	/* If the buffer is a prefetch, count it as such. */
	if (HDR_PREFETCH(hdr)) {
	ARCSTAT_INCR(arcstat_l2_prefetch_asize, asize_s);
	} else {
	/*
	* We use the value stored in the L2 header upon initial
	* caching in L2ARC. This value will be updated in case
	* an MRU/MRU_ghost buffer transitions to MFU but the L2ARC
	* metadata (log entry) cannot currently be updated. Having
	* the ARC state in the L2 header solves the problem of a
	* possibly absent L1 header (apparent in buffers restored
	* from persistent L2ARC).
	*/
	switch (hdr->b_l2hdr.b_arcs_state) {
	case ARC_STATE_MRU_GHOST:
	case ARC_STATE_MRU:
	ARCSTAT_INCR(arcstat_l2_mru_asize, asize_s);
	break;
	case ARC_STATE_MFU_GHOST:
	case ARC_STATE_MFU:
	ARCSTAT_INCR(arcstat_l2_mfu_asize, asize_s);
	break;
	default:
	break;
	}
	}

	if (state_only)
	return;

	ARCSTAT_INCR(arcstat_l2_psize, psize_s);
	ARCSTAT_INCR(arcstat_l2_lsize, lsize_s);

	switch (type) {
	case ARC_BUFC_DATA:
	ARCSTAT_INCR(arcstat_l2_bufc_data_asize, asize_s);
	break;
	case ARC_BUFC_METADATA:
	ARCSTAT_INCR(arcstat_l2_bufc_metadata_asize, asize_s);
	break;
	default:
	break;
	}
	}


	static void
	arc_hdr_l2hdr_destroy(arc_buf_hdr_t *hdr)
	{
	l2arc_buf_hdr_t *l2hdr = &hdr->b_l2hdr;
	l2arc_dev_t *dev = l2hdr->b_dev;
	uint64_t psize = HDR_GET_PSIZE(hdr);
	uint64_t asize = vdev_psize_to_asize(dev->l2ad_vdev, psize);

	ASSERT(MUTEX_HELD(&dev->l2ad_mtx));
	ASSERT(HDR_HAS_L2HDR(hdr));

	list_remove(&dev->l2ad_buflist, hdr);

	l2arc_hdr_arcstats_decrement(hdr);
	vdev_space_update(dev->l2ad_vdev, -asize, 0, 0);

	(void) zfs_refcount_remove_many(&dev->l2ad_alloc, arc_hdr_size(hdr),
	hdr);
	arc_hdr_clear_flags(hdr, ARC_FLAG_HAS_L2HDR);
	}

	static void
	arc_hdr_destroy(arc_buf_hdr_t *hdr)
	{
	if (HDR_HAS_L1HDR(hdr)) {
	ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
	ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon);
	}
	ASSERT(!HDR_IO_IN_PROGRESS(hdr));
	ASSERT(!HDR_IN_HASH_TABLE(hdr));

	if (HDR_HAS_L2HDR(hdr)) {
	l2arc_dev_t *dev = hdr->b_l2hdr.b_dev;
	boolean_t buflist_held = MUTEX_HELD(&dev->l2ad_mtx);

	if (!buflist_held)
	mutex_enter(&dev->l2ad_mtx);

	/*
	* Even though we checked this conditional above, we
	* need to check this again now that we have the
	* l2ad_mtx. This is because we could be racing with
	* another thread calling l2arc_evict() which might have
	* destroyed this header's L2 portion as we were waiting
	* to acquire the l2ad_mtx. If that happens, we don't
	* want to re-destroy the header's L2 portion.
	*/
	if (HDR_HAS_L2HDR(hdr)) {

	if (!HDR_EMPTY(hdr))
	buf_discard_identity(hdr);

	arc_hdr_l2hdr_destroy(hdr);
	}

	if (!buflist_held)
	mutex_exit(&dev->l2ad_mtx);
	}

	/*
	* The header's identify can only be safely discarded once it is no
	* longer discoverable. This requires removing it from the hash table
	* and the l2arc header list. After this point the hash lock can not
	* be used to protect the header.
	*/
	if (!HDR_EMPTY(hdr))
	buf_discard_identity(hdr);

	if (HDR_HAS_L1HDR(hdr)) {
	arc_cksum_free(hdr);

	while (hdr->b_l1hdr.b_buf != NULL)
	arc_buf_destroy_impl(hdr->b_l1hdr.b_buf);

	if (hdr->b_l1hdr.b_pabd != NULL)
	arc_hdr_free_abd(hdr, B_FALSE);

	if (HDR_HAS_RABD(hdr))
	arc_hdr_free_abd(hdr, B_TRUE);
	}

	ASSERT3P(hdr->b_hash_next, ==, NULL);
	if (HDR_HAS_L1HDR(hdr)) {
	ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node));
	ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL);
	#ifdef ZFS_DEBUG
	ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL);
	#endif
	kmem_cache_free(hdr_full_cache, hdr);
	} else {
	kmem_cache_free(hdr_l2only_cache, hdr);
	}
	}

	void
	arc_buf_destroy(arc_buf_t buf, const void tag)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;

	if (hdr->b_l1hdr.b_state == arc_anon) {
	ASSERT3P(hdr->b_l1hdr.b_buf, ==, buf);
	ASSERT(ARC_BUF_LAST(buf));
	ASSERT(!HDR_IO_IN_PROGRESS(hdr));
	VERIFY0(remove_reference(hdr, tag));
	return;
	}

	kmutex_t *hash_lock = HDR_LOCK(hdr);
	mutex_enter(hash_lock);

	ASSERT3P(hdr, ==, buf->b_hdr);
	ASSERT3P(hdr->b_l1hdr.b_buf, !=, NULL);
	ASSERT3P(hash_lock, ==, HDR_LOCK(hdr));
	ASSERT3P(hdr->b_l1hdr.b_state, !=, arc_anon);
	ASSERT3P(buf->b_data, !=, NULL);

	arc_buf_destroy_impl(buf);
	(void) remove_reference(hdr, tag);
	mutex_exit(hash_lock);
	}

	/*
	* Evict the arc_buf_hdr that is provided as a parameter. The resultant
	* state of the header is dependent on its state prior to entering this
	* function. The following transitions are possible:
	*
	* - arc_mru -> arc_mru_ghost
	* - arc_mfu -> arc_mfu_ghost
	* - arc_mru_ghost -> arc_l2c_only
	* - arc_mru_ghost -> deleted
	* - arc_mfu_ghost -> arc_l2c_only
	* - arc_mfu_ghost -> deleted
	* - arc_uncached -> deleted
	*
	* Return total size of evicted data buffers for eviction progress tracking.
	* When evicting from ghost states return logical buffer size to make eviction
	* progress at the same (or at least comparable) rate as from non-ghost states.
	*
	* Return *real_evicted for actual ARC size reduction to wake up threads
	* waiting for it. For non-ghost states it includes size of evicted data
	* buffers (the headers are not freed there). For ghost states it includes
	* only the evicted headers size.
	*/
	static int64_t
	arc_evict_hdr(arc_buf_hdr_t hdr, uint64_t real_evicted)
	{
	arc_state_t evicted_state, state;
	int64_t bytes_evicted = 0;
	uint_t min_lifetime = HDR_PRESCIENT_PREFETCH(hdr) ?
	arc_min_prescient_prefetch_ms : arc_min_prefetch_ms;

	ASSERT(MUTEX_HELD(HDR_LOCK(hdr)));
	ASSERT(HDR_HAS_L1HDR(hdr));
	ASSERT(!HDR_IO_IN_PROGRESS(hdr));
	ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
	ASSERT0(zfs_refcount_count(&hdr->b_l1hdr.b_refcnt));

	*real_evicted = 0;
	state = hdr->b_l1hdr.b_state;
	if (GHOST_STATE(state)) {

	/*
	* l2arc_write_buffers() relies on a header's L1 portion
	* (i.e. its b_pabd field) during it's write phase.
	* Thus, we cannot push a header onto the arc_l2c_only
	* state (removing its L1 piece) until the header is
	* done being written to the l2arc.
	*/
	if (HDR_HAS_L2HDR(hdr) && HDR_L2_WRITING(hdr)) {
	ARCSTAT_BUMP(arcstat_evict_l2_skip);
	return (bytes_evicted);
	}

	ARCSTAT_BUMP(arcstat_deleted);
	bytes_evicted += HDR_GET_LSIZE(hdr);

	DTRACE_PROBE1(arc__delete, arc_buf_hdr_t *, hdr);

	if (HDR_HAS_L2HDR(hdr)) {
	ASSERT(hdr->b_l1hdr.b_pabd == NULL);
	ASSERT(!HDR_HAS_RABD(hdr));
	/*
	* This buffer is cached on the 2nd Level ARC;
	* don't destroy the header.
	*/
	arc_change_state(arc_l2c_only, hdr);
	/*
	* dropping from L1+L2 cached to L2-only,
	* realloc to remove the L1 header.
	*/
	(void) arc_hdr_realloc(hdr, hdr_full_cache,
	hdr_l2only_cache);
	*real_evicted += HDR_FULL_SIZE - HDR_L2ONLY_SIZE;
	} else {
	arc_change_state(arc_anon, hdr);
	arc_hdr_destroy(hdr);
	*real_evicted += HDR_FULL_SIZE;
	}
	return (bytes_evicted);
	}

	ASSERT(state == arc_mru \|\| state == arc_mfu \|\| state == arc_uncached);
	evicted_state = (state == arc_uncached) ? arc_anon :
	((state == arc_mru) ? arc_mru_ghost : arc_mfu_ghost);

	/* prefetch buffers have a minimum lifespan */
	if ((hdr->b_flags & (ARC_FLAG_PREFETCH \| ARC_FLAG_INDIRECT)) &&
	ddi_get_lbolt() - hdr->b_l1hdr.b_arc_access <
	MSEC_TO_TICK(min_lifetime)) {
	ARCSTAT_BUMP(arcstat_evict_skip);
	return (bytes_evicted);
	}

	if (HDR_HAS_L2HDR(hdr)) {
	ARCSTAT_INCR(arcstat_evict_l2_cached, HDR_GET_LSIZE(hdr));
	} else {
	if (l2arc_write_eligible(hdr->b_spa, hdr)) {
	ARCSTAT_INCR(arcstat_evict_l2_eligible,
	HDR_GET_LSIZE(hdr));

	switch (state->arcs_state) {
	case ARC_STATE_MRU:
	ARCSTAT_INCR(
	arcstat_evict_l2_eligible_mru,
	HDR_GET_LSIZE(hdr));
	break;
	case ARC_STATE_MFU:
	ARCSTAT_INCR(
	arcstat_evict_l2_eligible_mfu,
	HDR_GET_LSIZE(hdr));
	break;
	default:
	break;
	}
	} else {
	ARCSTAT_INCR(arcstat_evict_l2_ineligible,
	HDR_GET_LSIZE(hdr));
	}
	}

	bytes_evicted += arc_hdr_size(hdr);
	*real_evicted += arc_hdr_size(hdr);

	/*
	* If this hdr is being evicted and has a compressed buffer then we
	* discard it here before we change states. This ensures that the
	* accounting is updated correctly in arc_free_data_impl().
	*/
	if (hdr->b_l1hdr.b_pabd != NULL)
	arc_hdr_free_abd(hdr, B_FALSE);

	if (HDR_HAS_RABD(hdr))
	arc_hdr_free_abd(hdr, B_TRUE);

	arc_change_state(evicted_state, hdr);
	DTRACE_PROBE1(arc__evict, arc_buf_hdr_t *, hdr);
	if (evicted_state == arc_anon) {
	arc_hdr_destroy(hdr);
	*real_evicted += HDR_FULL_SIZE;
	} else {
	ASSERT(HDR_IN_HASH_TABLE(hdr));
	}

	return (bytes_evicted);
	}

	static void
	arc_set_need_free(void)
	{
	ASSERT(MUTEX_HELD(&arc_evict_lock));
	int64_t remaining = arc_free_memory() - arc_sys_free / 2;
	arc_evict_waiter_t *aw = list_tail(&arc_evict_waiters);
	if (aw == NULL) {
	arc_need_free = MAX(-remaining, 0);
	} else {
	arc_need_free =
	MAX(-remaining, (int64_t)(aw->aew_count - arc_evict_count));
	}
	}

	static uint64_t
	arc_evict_state_impl(multilist_t ml, int idx, arc_buf_hdr_t marker,
	uint64_t spa, uint64_t bytes)
	{
	multilist_sublist_t *mls;
	uint64_t bytes_evicted = 0, real_evicted = 0;
	arc_buf_hdr_t *hdr;
	kmutex_t *hash_lock;
	uint_t evict_count = zfs_arc_evict_batch_limit;

	ASSERT3P(marker, !=, NULL);

	mls = multilist_sublist_lock_idx(ml, idx);

	for (hdr = multilist_sublist_prev(mls, marker); likely(hdr != NULL);
	hdr = multilist_sublist_prev(mls, marker)) {
	if ((evict_count == 0) \|\| (bytes_evicted >= bytes))
	break;

	/*
	* To keep our iteration location, move the marker
	* forward. Since we're not holding hdr's hash lock, we
	* must be very careful and not remove 'hdr' from the
	* sublist. Otherwise, other consumers might mistake the
	* 'hdr' as not being on a sublist when they call the
	* multilist_link_active() function (they all rely on
	* the hash lock protecting concurrent insertions and
	* removals). multilist_sublist_move_forward() was
	* specifically implemented to ensure this is the case
	* (only 'marker' will be removed and re-inserted).
	*/
	multilist_sublist_move_forward(mls, marker);

	/*
	* The only case where the b_spa field should ever be
	* zero, is the marker headers inserted by
	* arc_evict_state(). It's possible for multiple threads
	* to be calling arc_evict_state() concurrently (e.g.
	* dsl_pool_close() and zio_inject_fault()), so we must
	* skip any markers we see from these other threads.
	*/
	if (hdr->b_spa == 0)
	continue;

	/* we're only interested in evicting buffers of a certain spa */
	if (spa != 0 && hdr->b_spa != spa) {
	ARCSTAT_BUMP(arcstat_evict_skip);
	continue;
	}

	hash_lock = HDR_LOCK(hdr);

	/*
	* We aren't calling this function from any code path
	* that would already be holding a hash lock, so we're
	* asserting on this assumption to be defensive in case
	* this ever changes. Without this check, it would be
	* possible to incorrectly increment arcstat_mutex_miss
	* below (e.g. if the code changed such that we called
	* this function with a hash lock held).
	*/
	ASSERT(!MUTEX_HELD(hash_lock));

	if (mutex_tryenter(hash_lock)) {
	uint64_t revicted;
	uint64_t evicted = arc_evict_hdr(hdr, &revicted);
	mutex_exit(hash_lock);

	bytes_evicted += evicted;
	real_evicted += revicted;

	/*
	* If evicted is zero, arc_evict_hdr() must have
	* decided to skip this header, don't increment
	* evict_count in this case.
	*/
	if (evicted != 0)
	evict_count--;

	} else {
	ARCSTAT_BUMP(arcstat_mutex_miss);
	}
	}

	multilist_sublist_unlock(mls);

	/*
	* Increment the count of evicted bytes, and wake up any threads that
	* are waiting for the count to reach this value. Since the list is
	* ordered by ascending aew_count, we pop off the beginning of the
	* list until we reach the end, or a waiter that's past the current
	* "count". Doing this outside the loop reduces the number of times
	* we need to acquire the global arc_evict_lock.
	*
	* Only wake when there's sufficient free memory in the system
	* (specifically, arc_sys_free/2, which by default is a bit more than
	* 1/64th of RAM). See the comments in arc_wait_for_eviction().
	*/
	mutex_enter(&arc_evict_lock);
	arc_evict_count += real_evicted;

	if (arc_free_memory() > arc_sys_free / 2) {
	arc_evict_waiter_t *aw;
	while ((aw = list_head(&arc_evict_waiters)) != NULL &&
	aw->aew_count <= arc_evict_count) {
	list_remove(&arc_evict_waiters, aw);
	cv_broadcast(&aw->aew_cv);
	}
	}
	arc_set_need_free();
	mutex_exit(&arc_evict_lock);

	/*
	* If the ARC size is reduced from arc_c_max to arc_c_min (especially
	* if the average cached block is small), eviction can be on-CPU for
	* many seconds. To ensure that other threads that may be bound to
	* this CPU are able to make progress, make a voluntary preemption
	* call here.
	*/
	kpreempt(KPREEMPT_SYNC);

	return (bytes_evicted);
	}

	static arc_buf_hdr_t *
	arc_state_alloc_marker(void)
	{
	arc_buf_hdr_t *marker = kmem_cache_alloc(hdr_full_cache, KM_SLEEP);

	/*
	* A b_spa of 0 is used to indicate that this header is
	* a marker. This fact is used in arc_evict_state_impl().
	*/
	marker->b_spa = 0;

	return (marker);
	}

	static void
	arc_state_free_marker(arc_buf_hdr_t *marker)
	{
	kmem_cache_free(hdr_full_cache, marker);
	}

	/*
	* Allocate an array of buffer headers used as placeholders during arc state
	* eviction.
	*/
	static arc_buf_hdr_t **
	arc_state_alloc_markers(int count)
	{
	arc_buf_hdr_t **markers;

	markers = kmem_zalloc(sizeof (markers) count, KM_SLEEP);
	for (int i = 0; i < count; i++)
	markers[i] = arc_state_alloc_marker();
	return (markers);
	}

	static void
	arc_state_free_markers(arc_buf_hdr_t **markers, int count)
	{
	for (int i = 0; i < count; i++)
	arc_state_free_marker(markers[i]);
	kmem_free(markers, sizeof (markers) count);
	}

	/*
	* Evict buffers from the given arc state, until we've removed the
	* specified number of bytes. Move the removed buffers to the
	* appropriate evict state.
	*
	* This function makes a "best effort". It skips over any buffers
	* it can't get a hash_lock on, and so, may not catch all candidates.
	* It may also return without evicting as much space as requested.
	*
	* If bytes is specified using the special value ARC_EVICT_ALL, this
	* will evict all available (i.e. unlocked and evictable) buffers from
	* the given arc state; which is used by arc_flush().
	*/
	static uint64_t
	arc_evict_state(arc_state_t *state, arc_buf_contents_t type, uint64_t spa,
	uint64_t bytes)
	{
	uint64_t total_evicted = 0;
	multilist_t *ml = &state->arcs_list[type];
	int num_sublists;
	arc_buf_hdr_t **markers;

	num_sublists = multilist_get_num_sublists(ml);

	/*
	* If we've tried to evict from each sublist, made some
	* progress, but still have not hit the target number of bytes
	* to evict, we want to keep trying. The markers allow us to
	* pick up where we left off for each individual sublist, rather
	* than starting from the tail each time.
	*/
	if (zthr_iscurthread(arc_evict_zthr)) {
	markers = arc_state_evict_markers;
	ASSERT3S(num_sublists, <=, arc_state_evict_marker_count);
	} else {
	markers = arc_state_alloc_markers(num_sublists);
	}
	for (int i = 0; i < num_sublists; i++) {
	multilist_sublist_t *mls;

	mls = multilist_sublist_lock_idx(ml, i);
	multilist_sublist_insert_tail(mls, markers[i]);
	multilist_sublist_unlock(mls);
	}

	/*
	* While we haven't hit our target number of bytes to evict, or
	* we're evicting all available buffers.
	*/
	while (total_evicted < bytes) {
	int sublist_idx = multilist_get_random_index(ml);
	uint64_t scan_evicted = 0;

	/*
	* Start eviction using a randomly selected sublist,
	* this is to try and evenly balance eviction across all
	* sublists. Always starting at the same sublist
	* (e.g. index 0) would cause evictions to favor certain
	* sublists over others.
	*/
	for (int i = 0; i < num_sublists; i++) {
	uint64_t bytes_remaining;
	uint64_t bytes_evicted;

	if (total_evicted < bytes)
	bytes_remaining = bytes - total_evicted;
	else
	break;

	bytes_evicted = arc_evict_state_impl(ml, sublist_idx,
	markers[sublist_idx], spa, bytes_remaining);

	scan_evicted += bytes_evicted;
	total_evicted += bytes_evicted;

	/* we've reached the end, wrap to the beginning */
	if (++sublist_idx >= num_sublists)
	sublist_idx = 0;
	}

	/*
	* If we didn't evict anything during this scan, we have
	* no reason to believe we'll evict more during another
	* scan, so break the loop.
	*/
	if (scan_evicted == 0) {
	/* This isn't possible, let's make that obvious */
	ASSERT3S(bytes, !=, 0);

	/*
	* When bytes is ARC_EVICT_ALL, the only way to
	* break the loop is when scan_evicted is zero.
	* In that case, we actually have evicted enough,
	* so we don't want to increment the kstat.
	*/
	if (bytes != ARC_EVICT_ALL) {
	ASSERT3S(total_evicted, <, bytes);
	ARCSTAT_BUMP(arcstat_evict_not_enough);
	}

	break;
	}
	}

	for (int i = 0; i < num_sublists; i++) {
	multilist_sublist_t *mls = multilist_sublist_lock_idx(ml, i);
	multilist_sublist_remove(mls, markers[i]);
	multilist_sublist_unlock(mls);
	}
	if (markers != arc_state_evict_markers)
	arc_state_free_markers(markers, num_sublists);

	return (total_evicted);
	}

	/*
	* Flush all "evictable" data of the given type from the arc state
	* specified. This will not evict any "active" buffers (i.e. referenced).
	*
	* When 'retry' is set to B_FALSE, the function will make a single pass
	* over the state and evict any buffers that it can. Since it doesn't
	* continually retry the eviction, it might end up leaving some buffers
	* in the ARC due to lock misses.
	*
	* When 'retry' is set to B_TRUE, the function will continually retry the
	* eviction until all evictable buffers have been removed from the
	* state. As a result, if concurrent insertions into the state are
	* allowed (e.g. if the ARC isn't shutting down), this function might
	* wind up in an infinite loop, continually trying to evict buffers.
	*/
	static uint64_t
	arc_flush_state(arc_state_t *state, uint64_t spa, arc_buf_contents_t type,
	boolean_t retry)
	{
	uint64_t evicted = 0;

	while (zfs_refcount_count(&state->arcs_esize[type]) != 0) {
	evicted += arc_evict_state(state, type, spa, ARC_EVICT_ALL);

	if (!retry)
	break;
	}

	return (evicted);
	}

	/*
	* Evict the specified number of bytes from the state specified. This
	* function prevents us from trying to evict more from a state's list
	* than is "evictable", and to skip evicting altogether when passed a
	* negative value for "bytes". In contrast, arc_evict_state() will
	* evict everything it can, when passed a negative value for "bytes".
	*/
	static uint64_t
	arc_evict_impl(arc_state_t *state, arc_buf_contents_t type, int64_t bytes)
	{
	uint64_t delta;

	if (bytes > 0 && zfs_refcount_count(&state->arcs_esize[type]) > 0) {
	delta = MIN(zfs_refcount_count(&state->arcs_esize[type]),
	bytes);
	return (arc_evict_state(state, type, 0, delta));
	}

	return (0);
	}

	/*
	* Adjust specified fraction, taking into account initial ghost state(s) size,
	* ghost hit bytes towards increasing the fraction, ghost hit bytes towards
	* decreasing it, plus a balance factor, controlling the decrease rate, used
	* to balance metadata vs data.
	*/
	static uint64_t
	arc_evict_adj(uint64_t frac, uint64_t total, uint64_t up, uint64_t down,
	uint_t balance)
	{
	if (total < 8 \|\| up + down == 0)
	return (frac);

	/*
	* We should not have more ghost hits than ghost size, but they
	* may get close. Restrict maximum adjustment in that case.
	*/
	if (up + down >= total / 4) {
	uint64_t scale = (up + down) / (total / 8);
	up /= scale;
	down /= scale;
	}

	/* Get maximal dynamic range by choosing optimal shifts. */
	int s = highbit64(total);
	s = MIN(64 - s, 32);

	uint64_t ofrac = (1ULL << 32) - frac;

	if (frac >= 4 * ofrac)
	up /= frac / (2 * ofrac + 1);
	up = (up << s) / (total >> (32 - s));
	if (ofrac >= 4 * frac)
	down /= ofrac / (2 * frac + 1);
	down = (down << s) / (total >> (32 - s));
	down = down * 100 / balance;

	return (frac + up - down);
	}

	/*
	* Evict buffers from the cache, such that arcstat_size is capped by arc_c.
	*/
	static uint64_t
	arc_evict(void)
	{
	uint64_t asize, bytes, total_evicted = 0;
	int64_t e, mrud, mrum, mfud, mfum, w;
	static uint64_t ogrd, ogrm, ogfd, ogfm;
	static uint64_t gsrd, gsrm, gsfd, gsfm;
	uint64_t ngrd, ngrm, ngfd, ngfm;

	/* Get current size of ARC states we can evict from. */
	mrud = zfs_refcount_count(&arc_mru->arcs_size[ARC_BUFC_DATA]) +
	zfs_refcount_count(&arc_anon->arcs_size[ARC_BUFC_DATA]);
	mrum = zfs_refcount_count(&arc_mru->arcs_size[ARC_BUFC_METADATA]) +
	zfs_refcount_count(&arc_anon->arcs_size[ARC_BUFC_METADATA]);
	mfud = zfs_refcount_count(&arc_mfu->arcs_size[ARC_BUFC_DATA]);
	mfum = zfs_refcount_count(&arc_mfu->arcs_size[ARC_BUFC_METADATA]);
	uint64_t d = mrud + mfud;
	uint64_t m = mrum + mfum;
	uint64_t t = d + m;

	/* Get ARC ghost hits since last eviction. */
	ngrd = wmsum_value(&arc_mru_ghost->arcs_hits[ARC_BUFC_DATA]);
	uint64_t grd = ngrd - ogrd;
	ogrd = ngrd;
	ngrm = wmsum_value(&arc_mru_ghost->arcs_hits[ARC_BUFC_METADATA]);
	uint64_t grm = ngrm - ogrm;
	ogrm = ngrm;
	ngfd = wmsum_value(&arc_mfu_ghost->arcs_hits[ARC_BUFC_DATA]);
	uint64_t gfd = ngfd - ogfd;
	ogfd = ngfd;
	ngfm = wmsum_value(&arc_mfu_ghost->arcs_hits[ARC_BUFC_METADATA]);
	uint64_t gfm = ngfm - ogfm;
	ogfm = ngfm;

	/* Adjust ARC states balance based on ghost hits. */
	arc_meta = arc_evict_adj(arc_meta, gsrd + gsrm + gsfd + gsfm,
	grm + gfm, grd + gfd, zfs_arc_meta_balance);
	arc_pd = arc_evict_adj(arc_pd, gsrd + gsfd, grd, gfd, 100);
	arc_pm = arc_evict_adj(arc_pm, gsrm + gsfm, grm, gfm, 100);

	asize = aggsum_value(&arc_sums.arcstat_size);
	int64_t wt = t - (asize - arc_c);

	/*
	* Try to reduce pinned dnodes if more than 3/4 of wanted metadata
	* target is not evictable or if they go over arc_dnode_limit.
	*/
	int64_t prune = 0;
	int64_t dn = wmsum_value(&arc_sums.arcstat_dnode_size);
	w = wt * (int64_t)(arc_meta >> 16) >> 16;
	if (zfs_refcount_count(&arc_mru->arcs_size[ARC_BUFC_METADATA]) +
	zfs_refcount_count(&arc_mfu->arcs_size[ARC_BUFC_METADATA]) -
	zfs_refcount_count(&arc_mru->arcs_esize[ARC_BUFC_METADATA]) -
	zfs_refcount_count(&arc_mfu->arcs_esize[ARC_BUFC_METADATA]) >
	w * 3 / 4) {
	prune = dn / sizeof (dnode_t) *
	zfs_arc_dnode_reduce_percent / 100;
	} else if (dn > arc_dnode_limit) {
	prune = (dn - arc_dnode_limit) / sizeof (dnode_t) *
	zfs_arc_dnode_reduce_percent / 100;
	}
	if (prune > 0)
	arc_prune_async(prune);

	/* Evict MRU metadata. */
	w = wt * (int64_t)(arc_meta * arc_pm >> 48) >> 16;
	e = MIN((int64_t)(asize - arc_c), (int64_t)(mrum - w));
	bytes = arc_evict_impl(arc_mru, ARC_BUFC_METADATA, e);
	total_evicted += bytes;
	mrum -= bytes;
	asize -= bytes;

	/* Evict MFU metadata. */
	w = wt * (int64_t)(arc_meta >> 16) >> 16;
	e = MIN((int64_t)(asize - arc_c), (int64_t)(m - w));
	bytes = arc_evict_impl(arc_mfu, ARC_BUFC_METADATA, e);
	total_evicted += bytes;
	mfum -= bytes;
	asize -= bytes;

	/* Evict MRU data. */
	wt -= m - total_evicted;
	w = wt * (int64_t)(arc_pd >> 16) >> 16;
	e = MIN((int64_t)(asize - arc_c), (int64_t)(mrud - w));
	bytes = arc_evict_impl(arc_mru, ARC_BUFC_DATA, e);
	total_evicted += bytes;
	mrud -= bytes;
	asize -= bytes;

	/* Evict MFU data. */
	e = asize - arc_c;
	bytes = arc_evict_impl(arc_mfu, ARC_BUFC_DATA, e);
	mfud -= bytes;
	total_evicted += bytes;

	/*
	* Evict ghost lists
	*
	* Size of each state's ghost list represents how much that state
	* may grow by shrinking the other states. Would it need to shrink
	* other states to zero (that is unlikely), its ghost size would be
	* equal to sum of other three state sizes. But excessive ghost
	* size may result in false ghost hits (too far back), that may
	* never result in real cache hits if several states are competing.
	* So choose some arbitraty point of 1/2 of other state sizes.
	*/
	gsrd = (mrum + mfud + mfum) / 2;
	e = zfs_refcount_count(&arc_mru_ghost->arcs_size[ARC_BUFC_DATA]) -
	gsrd;
	(void) arc_evict_impl(arc_mru_ghost, ARC_BUFC_DATA, e);

	gsrm = (mrud + mfud + mfum) / 2;
	e = zfs_refcount_count(&arc_mru_ghost->arcs_size[ARC_BUFC_METADATA]) -
	gsrm;
	(void) arc_evict_impl(arc_mru_ghost, ARC_BUFC_METADATA, e);

	gsfd = (mrud + mrum + mfum) / 2;
	e = zfs_refcount_count(&arc_mfu_ghost->arcs_size[ARC_BUFC_DATA]) -
	gsfd;
	(void) arc_evict_impl(arc_mfu_ghost, ARC_BUFC_DATA, e);

	gsfm = (mrud + mrum + mfud) / 2;
	e = zfs_refcount_count(&arc_mfu_ghost->arcs_size[ARC_BUFC_METADATA]) -
	gsfm;
	(void) arc_evict_impl(arc_mfu_ghost, ARC_BUFC_METADATA, e);

	return (total_evicted);
	}

	void
	arc_flush(spa_t *spa, boolean_t retry)
	{
	uint64_t guid = 0;

	/*
	* If retry is B_TRUE, a spa must not be specified since we have
	* no good way to determine if all of a spa's buffers have been
	* evicted from an arc state.
	*/
	ASSERT(!retry \|\| spa == NULL);

	if (spa != NULL)
	guid = spa_load_guid(spa);

	(void) arc_flush_state(arc_mru, guid, ARC_BUFC_DATA, retry);
	(void) arc_flush_state(arc_mru, guid, ARC_BUFC_METADATA, retry);

	(void) arc_flush_state(arc_mfu, guid, ARC_BUFC_DATA, retry);
	(void) arc_flush_state(arc_mfu, guid, ARC_BUFC_METADATA, retry);

	(void) arc_flush_state(arc_mru_ghost, guid, ARC_BUFC_DATA, retry);
	(void) arc_flush_state(arc_mru_ghost, guid, ARC_BUFC_METADATA, retry);

	(void) arc_flush_state(arc_mfu_ghost, guid, ARC_BUFC_DATA, retry);
	(void) arc_flush_state(arc_mfu_ghost, guid, ARC_BUFC_METADATA, retry);

	(void) arc_flush_state(arc_uncached, guid, ARC_BUFC_DATA, retry);
	(void) arc_flush_state(arc_uncached, guid, ARC_BUFC_METADATA, retry);
	}

	void
	arc_reduce_target_size(int64_t to_free)
	{
	uint64_t c = arc_c;

	if (c <= arc_c_min)
	return;

	/*
	* All callers want the ARC to actually evict (at least) this much
	* memory. Therefore we reduce from the lower of the current size and
	* the target size. This way, even if arc_c is much higher than
	* arc_size (as can be the case after many calls to arc_freed(), we will
	* immediately have arc_c < arc_size and therefore the arc_evict_zthr
	* will evict.
	*/
	uint64_t asize = aggsum_value(&arc_sums.arcstat_size);
	if (asize < c)
	to_free += c - asize;
	arc_c = MAX((int64_t)c - to_free, (int64_t)arc_c_min);

	/* See comment in arc_evict_cb_check() on why lock+flag */
	mutex_enter(&arc_evict_lock);
	arc_evict_needed = B_TRUE;
	mutex_exit(&arc_evict_lock);
	zthr_wakeup(arc_evict_zthr);
	}

	/*
	* Determine if the system is under memory pressure and is asking
	* to reclaim memory. A return value of B_TRUE indicates that the system
	* is under memory pressure and that the arc should adjust accordingly.
	*/
	boolean_t
	arc_reclaim_needed(void)
	{
	return (arc_available_memory() < 0);
	}

	void
	arc_kmem_reap_soon(void)
	{
	size_t i;
	kmem_cache_t *prev_cache = NULL;
	kmem_cache_t *prev_data_cache = NULL;

	#ifdef _KERNEL
	#if defined(_ILP32)
	/*
	* Reclaim unused memory from all kmem caches.
	*/
	kmem_reap();
	#endif
	#endif

	for (i = 0; i < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; i++) {
	#if defined(_ILP32)
	/* reach upper limit of cache size on 32-bit */
	if (zio_buf_cache[i] == NULL)
	break;
	#endif
	if (zio_buf_cache[i] != prev_cache) {
	prev_cache = zio_buf_cache[i];
	kmem_cache_reap_now(zio_buf_cache[i]);
	}
	if (zio_data_buf_cache[i] != prev_data_cache) {
	prev_data_cache = zio_data_buf_cache[i];
	kmem_cache_reap_now(zio_data_buf_cache[i]);
	}
	}
	kmem_cache_reap_now(buf_cache);
	kmem_cache_reap_now(hdr_full_cache);
	kmem_cache_reap_now(hdr_l2only_cache);
	kmem_cache_reap_now(zfs_btree_leaf_cache);
	abd_cache_reap_now();
	}

	static boolean_t
	arc_evict_cb_check(void arg, zthr_t zthr)
	{
	(void) arg, (void) zthr;

	#ifdef ZFS_DEBUG
	/*
	* This is necessary in order to keep the kstat information
	* up to date for tools that display kstat data such as the
	* mdb ::arc dcmd and the Linux crash utility. These tools
	* typically do not call kstat's update function, but simply
	* dump out stats from the most recent update. Without
	* this call, these commands may show stale stats for the
	* anon, mru, mru_ghost, mfu, and mfu_ghost lists. Even
	* with this call, the data might be out of date if the
	* evict thread hasn't been woken recently; but that should
	* suffice. The arc_state_t structures can be queried
	* directly if more accurate information is needed.
	*/
	if (arc_ksp != NULL)
	arc_ksp->ks_update(arc_ksp, KSTAT_READ);
	#endif

	/*
	* We have to rely on arc_wait_for_eviction() to tell us when to
	* evict, rather than checking if we are overflowing here, so that we
	* are sure to not leave arc_wait_for_eviction() waiting on aew_cv.
	* If we have become "not overflowing" since arc_wait_for_eviction()
	* checked, we need to wake it up. We could broadcast the CV here,
	* but arc_wait_for_eviction() may have not yet gone to sleep. We
	* would need to use a mutex to ensure that this function doesn't
	* broadcast until arc_wait_for_eviction() has gone to sleep (e.g.
	* the arc_evict_lock). However, the lock ordering of such a lock
	* would necessarily be incorrect with respect to the zthr_lock,
	* which is held before this function is called, and is held by
	* arc_wait_for_eviction() when it calls zthr_wakeup().
	*/
	if (arc_evict_needed)
	return (B_TRUE);

	/*
	* If we have buffers in uncached state, evict them periodically.
	*/
	return ((zfs_refcount_count(&arc_uncached->arcs_esize[ARC_BUFC_DATA]) +
	zfs_refcount_count(&arc_uncached->arcs_esize[ARC_BUFC_METADATA]) &&
	ddi_get_lbolt() - arc_last_uncached_flush >
	MSEC_TO_TICK(arc_min_prefetch_ms / 2)));
	}

	/*
	* Keep arc_size under arc_c by running arc_evict which evicts data
	* from the ARC.
	*/
	static void
	arc_evict_cb(void arg, zthr_t zthr)
	{
	(void) arg;

	uint64_t evicted = 0;
	fstrans_cookie_t cookie = spl_fstrans_mark();

	/* Always try to evict from uncached state. */
	arc_last_uncached_flush = ddi_get_lbolt();
	evicted += arc_flush_state(arc_uncached, 0, ARC_BUFC_DATA, B_FALSE);
	evicted += arc_flush_state(arc_uncached, 0, ARC_BUFC_METADATA, B_FALSE);

	/* Evict from other states only if told to. */
	if (arc_evict_needed)
	evicted += arc_evict();

	/*
	* If evicted is zero, we couldn't evict anything
	* via arc_evict(). This could be due to hash lock
	* collisions, but more likely due to the majority of
	* arc buffers being unevictable. Therefore, even if
	* arc_size is above arc_c, another pass is unlikely to
	* be helpful and could potentially cause us to enter an
	* infinite loop. Additionally, zthr_iscancelled() is
	* checked here so that if the arc is shutting down, the
	* broadcast will wake any remaining arc evict waiters.
	*
	* Note we cancel using zthr instead of arc_evict_zthr
	* because the latter may not yet be initializd when the
	* callback is first invoked.
	*/
	mutex_enter(&arc_evict_lock);
	arc_evict_needed = !zthr_iscancelled(zthr) &&
	evicted > 0 && aggsum_compare(&arc_sums.arcstat_size, arc_c) > 0;
	if (!arc_evict_needed) {
	/*
	* We're either no longer overflowing, or we
	* can't evict anything more, so we should wake
	* arc_get_data_impl() sooner.
	*/
	arc_evict_waiter_t *aw;
	while ((aw = list_remove_head(&arc_evict_waiters)) != NULL) {
	cv_broadcast(&aw->aew_cv);
	}
	arc_set_need_free();
	}
	mutex_exit(&arc_evict_lock);
	spl_fstrans_unmark(cookie);
	}

	static boolean_t
	arc_reap_cb_check(void arg, zthr_t zthr)
	{
	(void) arg, (void) zthr;

	int64_t free_memory = arc_available_memory();
	static int reap_cb_check_counter = 0;

	/*
	* If a kmem reap is already active, don't schedule more. We must
	* check for this because kmem_cache_reap_soon() won't actually
	* block on the cache being reaped (this is to prevent callers from
	* becoming implicitly blocked by a system-wide kmem reap -- which,
	* on a system with many, many full magazines, can take minutes).
	*/
	if (!kmem_cache_reap_active() && free_memory < 0) {

	arc_no_grow = B_TRUE;
	arc_warm = B_TRUE;
	/*
	* Wait at least zfs_grow_retry (default 5) seconds
	* before considering growing.
	*/
	arc_growtime = gethrtime() + SEC2NSEC(arc_grow_retry);
	return (B_TRUE);
	} else if (free_memory < arc_c >> arc_no_grow_shift) {
	arc_no_grow = B_TRUE;
	} else if (gethrtime() >= arc_growtime) {
	arc_no_grow = B_FALSE;
	}

	/*
	* Called unconditionally every 60 seconds to reclaim unused
	* zstd compression and decompression context. This is done
	* here to avoid the need for an independent thread.
	*/
	if (!((reap_cb_check_counter++) % 60))
	zfs_zstd_cache_reap_now();

	return (B_FALSE);
	}

	/*
	* Keep enough free memory in the system by reaping the ARC's kmem
	* caches. To cause more slabs to be reapable, we may reduce the
	* target size of the cache (arc_c), causing the arc_evict_cb()
	* to free more buffers.
	*/
	static void
	arc_reap_cb(void arg, zthr_t zthr)
	{
	(void) arg, (void) zthr;

	int64_t free_memory;
	fstrans_cookie_t cookie = spl_fstrans_mark();

	/*
	* Kick off asynchronous kmem_reap()'s of all our caches.
	*/
	arc_kmem_reap_soon();

	/*
	* Wait at least arc_kmem_cache_reap_retry_ms between
	* arc_kmem_reap_soon() calls. Without this check it is possible to
	* end up in a situation where we spend lots of time reaping
	* caches, while we're near arc_c_min. Waiting here also gives the
	* subsequent free memory check a chance of finding that the
	* asynchronous reap has already freed enough memory, and we don't
	* need to call arc_reduce_target_size().
	*/
	delay((hz * arc_kmem_cache_reap_retry_ms + 999) / 1000);

	/*
	* Reduce the target size as needed to maintain the amount of free
	* memory in the system at a fraction of the arc_size (1/128th by
	* default). If oversubscribed (free_memory < 0) then reduce the
	* target arc_size by the deficit amount plus the fractional
	* amount. If free memory is positive but less than the fractional
	* amount, reduce by what is needed to hit the fractional amount.
	*/
	free_memory = arc_available_memory();

	int64_t can_free = arc_c - arc_c_min;
	if (can_free > 0) {
	int64_t to_free = (can_free >> arc_shrink_shift) - free_memory;
	if (to_free > 0)
	arc_reduce_target_size(to_free);
	}
	spl_fstrans_unmark(cookie);
	}

	#ifdef _KERNEL
	/*
	* Determine the amount of memory eligible for eviction contained in the
	* ARC. All clean data reported by the ghost lists can always be safely
	* evicted. Due to arc_c_min, the same does not hold for all clean data
	* contained by the regular mru and mfu lists.
	*
	* In the case of the regular mru and mfu lists, we need to report as
	* much clean data as possible, such that evicting that same reported
	* data will not bring arc_size below arc_c_min. Thus, in certain
	* circumstances, the total amount of clean data in the mru and mfu
	* lists might not actually be evictable.
	*
	* The following two distinct cases are accounted for:
	*
	* 1. The sum of the amount of dirty data contained by both the mru and
	* mfu lists, plus the ARC's other accounting (e.g. the anon list),
	* is greater than or equal to arc_c_min.
	* (i.e. amount of dirty data >= arc_c_min)
	*
	* This is the easy case; all clean data contained by the mru and mfu
	* lists is evictable. Evicting all clean data can only drop arc_size
	* to the amount of dirty data, which is greater than arc_c_min.
	*
	* 2. The sum of the amount of dirty data contained by both the mru and
	* mfu lists, plus the ARC's other accounting (e.g. the anon list),
	* is less than arc_c_min.
	* (i.e. arc_c_min > amount of dirty data)
	*
	* 2.1. arc_size is greater than or equal arc_c_min.
	* (i.e. arc_size >= arc_c_min > amount of dirty data)
	*
	* In this case, not all clean data from the regular mru and mfu
	* lists is actually evictable; we must leave enough clean data
	* to keep arc_size above arc_c_min. Thus, the maximum amount of
	* evictable data from the two lists combined, is exactly the
	* difference between arc_size and arc_c_min.
	*
	* 2.2. arc_size is less than arc_c_min
	* (i.e. arc_c_min > arc_size > amount of dirty data)
	*
	* In this case, none of the data contained in the mru and mfu
	* lists is evictable, even if it's clean. Since arc_size is
	* already below arc_c_min, evicting any more would only
	* increase this negative difference.
	*/

	#endif /* _KERNEL */

	/*
	* Adapt arc info given the number of bytes we are trying to add and
	* the state that we are coming from. This function is only called
	* when we are adding new content to the cache.
	*/
	static void
	arc_adapt(uint64_t bytes)
	{
	/*
	* Wake reap thread if we do not have any available memory
	*/
	if (arc_reclaim_needed()) {
	zthr_wakeup(arc_reap_zthr);
	return;
	}

	if (arc_no_grow)
	return;

	if (arc_c >= arc_c_max)
	return;

	/*
	* If we're within (2 * maxblocksize) bytes of the target
	* cache size, increment the target cache size
	*/
	if (aggsum_upper_bound(&arc_sums.arcstat_size) +
	2 * SPA_MAXBLOCKSIZE >= arc_c) {
	uint64_t dc = MAX(bytes, SPA_OLD_MAXBLOCKSIZE);
	if (atomic_add_64_nv(&arc_c, dc) > arc_c_max)
	arc_c = arc_c_max;
	}
	}

	/*
	* Check if arc_size has grown past our upper threshold, determined by
	* zfs_arc_overflow_shift.
	*/
	static arc_ovf_level_t
	arc_is_overflowing(boolean_t use_reserve)
	{
	/* Always allow at least one block of overflow */
	int64_t overflow = MAX(SPA_MAXBLOCKSIZE,
	arc_c >> zfs_arc_overflow_shift);

	/*
	* We just compare the lower bound here for performance reasons. Our
	* primary goals are to make sure that the arc never grows without
	* bound, and that it can reach its maximum size. This check
	* accomplishes both goals. The maximum amount we could run over by is
	* 2 * aggsum_borrow_multiplier * NUM_CPUS * the average size of a block
	* in the ARC. In practice, that's in the tens of MB, which is low
	* enough to be safe.
	*/
	int64_t over = aggsum_lower_bound(&arc_sums.arcstat_size) -
	arc_c - overflow / 2;
	if (!use_reserve)
	overflow /= 2;
	return (over < 0 ? ARC_OVF_NONE :
	over < overflow ? ARC_OVF_SOME : ARC_OVF_SEVERE);
	}

	static abd_t *
	arc_get_data_abd(arc_buf_hdr_t hdr, uint64_t size, const void tag,
	int alloc_flags)
	{
	arc_buf_contents_t type = arc_buf_type(hdr);

	arc_get_data_impl(hdr, size, tag, alloc_flags);
	if (alloc_flags & ARC_HDR_ALLOC_LINEAR)
	return (abd_alloc_linear(size, type == ARC_BUFC_METADATA));
	else
	return (abd_alloc(size, type == ARC_BUFC_METADATA));
	}

	static void *
	arc_get_data_buf(arc_buf_hdr_t hdr, uint64_t size, const void tag)
	{
	arc_buf_contents_t type = arc_buf_type(hdr);

	arc_get_data_impl(hdr, size, tag, 0);
	if (type == ARC_BUFC_METADATA) {
	return (zio_buf_alloc(size));
	} else {
	ASSERT(type == ARC_BUFC_DATA);
	return (zio_data_buf_alloc(size));
	}
	}

	/*
	* Wait for the specified amount of data (in bytes) to be evicted from the
	* ARC, and for there to be sufficient free memory in the system. Waiting for
	* eviction ensures that the memory used by the ARC decreases. Waiting for
	* free memory ensures that the system won't run out of free pages, regardless
	* of ARC behavior and settings. See arc_lowmem_init().
	*/
	void
	arc_wait_for_eviction(uint64_t amount, boolean_t use_reserve)
	{
	switch (arc_is_overflowing(use_reserve)) {
	case ARC_OVF_NONE:
	return;
	case ARC_OVF_SOME:
	/*
	* This is a bit racy without taking arc_evict_lock, but the
	* worst that can happen is we either call zthr_wakeup() extra
	* time due to race with other thread here, or the set flag
	* get cleared by arc_evict_cb(), which is unlikely due to
	* big hysteresis, but also not important since at this level
	* of overflow the eviction is purely advisory. Same time
	* taking the global lock here every time without waiting for
	* the actual eviction creates a significant lock contention.
	*/
	if (!arc_evict_needed) {
	arc_evict_needed = B_TRUE;
	zthr_wakeup(arc_evict_zthr);
	}
	return;
	case ARC_OVF_SEVERE:
	default:
	{
	arc_evict_waiter_t aw;
	list_link_init(&aw.aew_node);
	cv_init(&aw.aew_cv, NULL, CV_DEFAULT, NULL);

	uint64_t last_count = 0;
	mutex_enter(&arc_evict_lock);
	if (!list_is_empty(&arc_evict_waiters)) {
	arc_evict_waiter_t *last =
	list_tail(&arc_evict_waiters);
	last_count = last->aew_count;
	} else if (!arc_evict_needed) {
	arc_evict_needed = B_TRUE;
	zthr_wakeup(arc_evict_zthr);
	}
	/*
	* Note, the last waiter's count may be less than
	* arc_evict_count if we are low on memory in which
	* case arc_evict_state_impl() may have deferred
	* wakeups (but still incremented arc_evict_count).
	*/
	aw.aew_count = MAX(last_count, arc_evict_count) + amount;

	list_insert_tail(&arc_evict_waiters, &aw);

	arc_set_need_free();

	DTRACE_PROBE3(arc__wait__for__eviction,
	uint64_t, amount,
	uint64_t, arc_evict_count,
	uint64_t, aw.aew_count);

	/*
	* We will be woken up either when arc_evict_count reaches
	* aew_count, or when the ARC is no longer overflowing and
	* eviction completes.
	* In case of "false" wakeup, we will still be on the list.
	*/
	do {
	cv_wait(&aw.aew_cv, &arc_evict_lock);
	} while (list_link_active(&aw.aew_node));
	mutex_exit(&arc_evict_lock);

	cv_destroy(&aw.aew_cv);
	}
	}
	}

	/*
	* Allocate a block and return it to the caller. If we are hitting the
	* hard limit for the cache size, we must sleep, waiting for the eviction
	* thread to catch up. If we're past the target size but below the hard
	* limit, we'll only signal the reclaim thread and continue on.
	*/
	static void
	arc_get_data_impl(arc_buf_hdr_t hdr, uint64_t size, const void tag,
	int alloc_flags)
	{
	arc_adapt(size);

	/*
	* If arc_size is currently overflowing, we must be adding data
	* faster than we are evicting. To ensure we don't compound the
	* problem by adding more data and forcing arc_size to grow even
	* further past it's target size, we wait for the eviction thread to
	* make some progress. We also wait for there to be sufficient free
	* memory in the system, as measured by arc_free_memory().
	*
	* Specifically, we wait for zfs_arc_eviction_pct percent of the
	* requested size to be evicted. This should be more than 100%, to
	* ensure that that progress is also made towards getting arc_size
	* under arc_c. See the comment above zfs_arc_eviction_pct.
	*/
	arc_wait_for_eviction(size * zfs_arc_eviction_pct / 100,
	alloc_flags & ARC_HDR_USE_RESERVE);

	arc_buf_contents_t type = arc_buf_type(hdr);
	if (type == ARC_BUFC_METADATA) {
	arc_space_consume(size, ARC_SPACE_META);
	} else {
	arc_space_consume(size, ARC_SPACE_DATA);
	}

	/*
	* Update the state size. Note that ghost states have a
	* "ghost size" and so don't need to be updated.
	*/
	arc_state_t *state = hdr->b_l1hdr.b_state;
	if (!GHOST_STATE(state)) {

	(void) zfs_refcount_add_many(&state->arcs_size[type], size,
	tag);

	/*
	* If this is reached via arc_read, the link is
	* protected by the hash lock. If reached via
	* arc_buf_alloc, the header should not be accessed by
	* any other thread. And, if reached via arc_read_done,
	* the hash lock will protect it if it's found in the
	* hash table; otherwise no other thread should be
	* trying to [add\|remove]_reference it.
	*/
	if (multilist_link_active(&hdr->b_l1hdr.b_arc_node)) {
	ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
	(void) zfs_refcount_add_many(&state->arcs_esize[type],
	size, tag);
	}
	}
	}

	static void
	arc_free_data_abd(arc_buf_hdr_t hdr, abd_t abd, uint64_t size,
	const void *tag)
	{
	arc_free_data_impl(hdr, size, tag);
	abd_free(abd);
	}

	static void
	arc_free_data_buf(arc_buf_hdr_t hdr, void buf, uint64_t size, const void *tag)
	{
	arc_buf_contents_t type = arc_buf_type(hdr);

	arc_free_data_impl(hdr, size, tag);
	if (type == ARC_BUFC_METADATA) {
	zio_buf_free(buf, size);
	} else {
	ASSERT(type == ARC_BUFC_DATA);
	zio_data_buf_free(buf, size);
	}
	}

	/*
	* Free the arc data buffer.
	*/
	static void
	arc_free_data_impl(arc_buf_hdr_t hdr, uint64_t size, const void tag)
	{
	arc_state_t *state = hdr->b_l1hdr.b_state;
	arc_buf_contents_t type = arc_buf_type(hdr);

	/* protected by hash lock, if in the hash table */
	if (multilist_link_active(&hdr->b_l1hdr.b_arc_node)) {
	ASSERT(zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt));
	ASSERT(state != arc_anon && state != arc_l2c_only);

	(void) zfs_refcount_remove_many(&state->arcs_esize[type],
	size, tag);
	}
	(void) zfs_refcount_remove_many(&state->arcs_size[type], size, tag);

	VERIFY3U(hdr->b_type, ==, type);
	if (type == ARC_BUFC_METADATA) {
	arc_space_return(size, ARC_SPACE_META);
	} else {
	ASSERT(type == ARC_BUFC_DATA);
	arc_space_return(size, ARC_SPACE_DATA);
	}
	}

	/*
	* This routine is called whenever a buffer is accessed.
	*/
	static void
	arc_access(arc_buf_hdr_t *hdr, arc_flags_t arc_flags, boolean_t hit)
	{
	ASSERT(MUTEX_HELD(HDR_LOCK(hdr)));
	ASSERT(HDR_HAS_L1HDR(hdr));

	/*
	* Update buffer prefetch status.
	*/
	boolean_t was_prefetch = HDR_PREFETCH(hdr);
	boolean_t now_prefetch = arc_flags & ARC_FLAG_PREFETCH;
	if (was_prefetch != now_prefetch) {
	if (was_prefetch) {
	ARCSTAT_CONDSTAT(hit, demand_hit, demand_iohit,
	HDR_PRESCIENT_PREFETCH(hdr), prescient, predictive,
	prefetch);
	}
	if (HDR_HAS_L2HDR(hdr))
	l2arc_hdr_arcstats_decrement_state(hdr);
	if (was_prefetch) {
	arc_hdr_clear_flags(hdr,
	ARC_FLAG_PREFETCH \| ARC_FLAG_PRESCIENT_PREFETCH);
	} else {
	arc_hdr_set_flags(hdr, ARC_FLAG_PREFETCH);
	}
	if (HDR_HAS_L2HDR(hdr))
	l2arc_hdr_arcstats_increment_state(hdr);
	}
	if (now_prefetch) {
	if (arc_flags & ARC_FLAG_PRESCIENT_PREFETCH) {
	arc_hdr_set_flags(hdr, ARC_FLAG_PRESCIENT_PREFETCH);
	ARCSTAT_BUMP(arcstat_prescient_prefetch);
	} else {
	ARCSTAT_BUMP(arcstat_predictive_prefetch);
	}
	}
	if (arc_flags & ARC_FLAG_L2CACHE)
	arc_hdr_set_flags(hdr, ARC_FLAG_L2CACHE);

	clock_t now = ddi_get_lbolt();
	if (hdr->b_l1hdr.b_state == arc_anon) {
	arc_state_t *new_state;
	/*
	* This buffer is not in the cache, and does not appear in
	* our "ghost" lists. Add it to the MRU or uncached state.
	*/
	ASSERT0(hdr->b_l1hdr.b_arc_access);
	hdr->b_l1hdr.b_arc_access = now;
	if (HDR_UNCACHED(hdr)) {
	new_state = arc_uncached;
	DTRACE_PROBE1(new_state__uncached, arc_buf_hdr_t *,
	hdr);
	} else {
	new_state = arc_mru;
	DTRACE_PROBE1(new_state__mru, arc_buf_hdr_t *, hdr);
	}
	arc_change_state(new_state, hdr);
	} else if (hdr->b_l1hdr.b_state == arc_mru) {
	/*
	* This buffer has been accessed once recently and either
	* its read is still in progress or it is in the cache.
	*/
	if (HDR_IO_IN_PROGRESS(hdr)) {
	hdr->b_l1hdr.b_arc_access = now;
	return;
	}
	hdr->b_l1hdr.b_mru_hits++;
	ARCSTAT_BUMP(arcstat_mru_hits);

	/*
	* If the previous access was a prefetch, then it already
	* handled possible promotion, so nothing more to do for now.
	*/
	if (was_prefetch) {
	hdr->b_l1hdr.b_arc_access = now;
	return;
	}

	/*
	* If more than ARC_MINTIME have passed from the previous
	* hit, promote the buffer to the MFU state.
	*/
	if (ddi_time_after(now, hdr->b_l1hdr.b_arc_access +
	ARC_MINTIME)) {
	hdr->b_l1hdr.b_arc_access = now;
	DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, hdr);
	arc_change_state(arc_mfu, hdr);
	}
	} else if (hdr->b_l1hdr.b_state == arc_mru_ghost) {
	arc_state_t *new_state;
	/*
	* This buffer has been accessed once recently, but was
	* evicted from the cache. Would we have bigger MRU, it
	* would be an MRU hit, so handle it the same way, except
	* we don't need to check the previous access time.
	*/
	hdr->b_l1hdr.b_mru_ghost_hits++;
	ARCSTAT_BUMP(arcstat_mru_ghost_hits);
	hdr->b_l1hdr.b_arc_access = now;
	wmsum_add(&arc_mru_ghost->arcs_hits[arc_buf_type(hdr)],
	arc_hdr_size(hdr));
	if (was_prefetch) {
	new_state = arc_mru;
	DTRACE_PROBE1(new_state__mru, arc_buf_hdr_t *, hdr);
	} else {
	new_state = arc_mfu;
	DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, hdr);
	}
	arc_change_state(new_state, hdr);
	} else if (hdr->b_l1hdr.b_state == arc_mfu) {
	/*
	* This buffer has been accessed more than once and either
	* still in the cache or being restored from one of ghosts.
	*/
	if (!HDR_IO_IN_PROGRESS(hdr)) {
	hdr->b_l1hdr.b_mfu_hits++;
	ARCSTAT_BUMP(arcstat_mfu_hits);
	}
	hdr->b_l1hdr.b_arc_access = now;
	} else if (hdr->b_l1hdr.b_state == arc_mfu_ghost) {
	/*
	* This buffer has been accessed more than once recently, but
	* has been evicted from the cache. Would we have bigger MFU
	* it would stay in cache, so move it back to MFU state.
	*/
	hdr->b_l1hdr.b_mfu_ghost_hits++;
	ARCSTAT_BUMP(arcstat_mfu_ghost_hits);
	hdr->b_l1hdr.b_arc_access = now;
	wmsum_add(&arc_mfu_ghost->arcs_hits[arc_buf_type(hdr)],
	arc_hdr_size(hdr));
	DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, hdr);
	arc_change_state(arc_mfu, hdr);
	} else if (hdr->b_l1hdr.b_state == arc_uncached) {
	/*
	* This buffer is uncacheable, but we got a hit. Probably
	* a demand read after prefetch. Nothing more to do here.
	*/
	if (!HDR_IO_IN_PROGRESS(hdr))
	ARCSTAT_BUMP(arcstat_uncached_hits);
	hdr->b_l1hdr.b_arc_access = now;
	} else if (hdr->b_l1hdr.b_state == arc_l2c_only) {
	/*
	* This buffer is on the 2nd Level ARC and was not accessed
	* for a long time, so treat it as new and put into MRU.
	*/
	hdr->b_l1hdr.b_arc_access = now;
	DTRACE_PROBE1(new_state__mru, arc_buf_hdr_t *, hdr);
	arc_change_state(arc_mru, hdr);
	} else {
	cmn_err(CE_PANIC, "invalid arc state 0x%p",
	hdr->b_l1hdr.b_state);
	}
	}

	/*
	* This routine is called by dbuf_hold() to update the arc_access() state
	* which otherwise would be skipped for entries in the dbuf cache.
	*/
	void
	arc_buf_access(arc_buf_t *buf)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;

	/*
	* Avoid taking the hash_lock when possible as an optimization.
	* The header must be checked again under the hash_lock in order
	* to handle the case where it is concurrently being released.
	*/
	if (hdr->b_l1hdr.b_state == arc_anon \|\| HDR_EMPTY(hdr))
	return;

	kmutex_t *hash_lock = HDR_LOCK(hdr);
	mutex_enter(hash_lock);

	if (hdr->b_l1hdr.b_state == arc_anon \|\| HDR_EMPTY(hdr)) {
	mutex_exit(hash_lock);
	ARCSTAT_BUMP(arcstat_access_skip);
	return;
	}

	ASSERT(hdr->b_l1hdr.b_state == arc_mru \|\|
	hdr->b_l1hdr.b_state == arc_mfu \|\|
	hdr->b_l1hdr.b_state == arc_uncached);

	DTRACE_PROBE1(arc__hit, arc_buf_hdr_t *, hdr);
	arc_access(hdr, 0, B_TRUE);
	mutex_exit(hash_lock);

	ARCSTAT_BUMP(arcstat_hits);
	ARCSTAT_CONDSTAT(B_TRUE /* demand */, demand, prefetch,
	!HDR_ISTYPE_METADATA(hdr), data, metadata, hits);
	}

	/* a generic arc_read_done_func_t which you can use */
	void
	arc_bcopy_func(zio_t zio, const zbookmark_phys_t zb, const blkptr_t *bp,
	arc_buf_t buf, void arg)
	{
	(void) zio, (void) zb, (void) bp;

	if (buf == NULL)
	return;

	memcpy(arg, buf->b_data, arc_buf_size(buf));
	arc_buf_destroy(buf, arg);
	}

	/* a generic arc_read_done_func_t */
	void
	arc_getbuf_func(zio_t zio, const zbookmark_phys_t zb, const blkptr_t *bp,
	arc_buf_t buf, void arg)
	{
	(void) zb, (void) bp;
	arc_buf_t **bufp = arg;

	if (buf == NULL) {
	ASSERT(zio == NULL \|\| zio->io_error != 0);
	*bufp = NULL;
	} else {
	ASSERT(zio == NULL \|\| zio->io_error == 0);
	*bufp = buf;
	ASSERT(buf->b_data != NULL);
	}
	}

	static void
	arc_hdr_verify(arc_buf_hdr_t hdr, blkptr_t bp)
	{
	if (BP_IS_HOLE(bp) \|\| BP_IS_EMBEDDED(bp)) {
	ASSERT3U(HDR_GET_PSIZE(hdr), ==, 0);
	ASSERT3U(arc_hdr_get_compress(hdr), ==, ZIO_COMPRESS_OFF);
	} else {
	if (HDR_COMPRESSION_ENABLED(hdr)) {
	ASSERT3U(arc_hdr_get_compress(hdr), ==,
	BP_GET_COMPRESS(bp));
	}
	ASSERT3U(HDR_GET_LSIZE(hdr), ==, BP_GET_LSIZE(bp));
	ASSERT3U(HDR_GET_PSIZE(hdr), ==, BP_GET_PSIZE(bp));
	ASSERT3U(!!HDR_PROTECTED(hdr), ==, BP_IS_PROTECTED(bp));
	}
	}

	static void
	arc_read_done(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;
	arc_buf_hdr_t *hdr = zio->io_private;
	kmutex_t *hash_lock = NULL;
	arc_callback_t *callback_list;
	arc_callback_t *acb;

	/*
	* The hdr was inserted into hash-table and removed from lists
	* prior to starting I/O. We should find this header, since
	* it's in the hash table, and it should be legit since it's
	* not possible to evict it during the I/O. The only possible
	* reason for it not to be found is if we were freed during the
	* read.
	*/
	if (HDR_IN_HASH_TABLE(hdr)) {
	arc_buf_hdr_t *found;

	ASSERT3U(hdr->b_birth, ==, BP_GET_BIRTH(zio->io_bp));
	ASSERT3U(hdr->b_dva.dva_word[0], ==,
	BP_IDENTITY(zio->io_bp)->dva_word[0]);
	ASSERT3U(hdr->b_dva.dva_word[1], ==,
	BP_IDENTITY(zio->io_bp)->dva_word[1]);

	found = buf_hash_find(hdr->b_spa, zio->io_bp, &hash_lock);

	ASSERT((found == hdr &&
	DVA_EQUAL(&hdr->b_dva, BP_IDENTITY(zio->io_bp))) \|\|
	(found == hdr && HDR_L2_READING(hdr)));
	ASSERT3P(hash_lock, !=, NULL);
	}

	if (BP_IS_PROTECTED(bp)) {
	hdr->b_crypt_hdr.b_ot = BP_GET_TYPE(bp);
	hdr->b_crypt_hdr.b_dsobj = zio->io_bookmark.zb_objset;
	zio_crypt_decode_params_bp(bp, hdr->b_crypt_hdr.b_salt,
	hdr->b_crypt_hdr.b_iv);

	if (zio->io_error == 0) {
	if (BP_GET_TYPE(bp) == DMU_OT_INTENT_LOG) {
	void *tmpbuf;

	tmpbuf = abd_borrow_buf_copy(zio->io_abd,
	sizeof (zil_chain_t));
	zio_crypt_decode_mac_zil(tmpbuf,
	hdr->b_crypt_hdr.b_mac);
	abd_return_buf(zio->io_abd, tmpbuf,
	sizeof (zil_chain_t));
	} else {
	zio_crypt_decode_mac_bp(bp,
	hdr->b_crypt_hdr.b_mac);
	}
	}
	}

	if (zio->io_error == 0) {
	/* byteswap if necessary */
	if (BP_SHOULD_BYTESWAP(zio->io_bp)) {
	if (BP_GET_LEVEL(zio->io_bp) > 0) {
	hdr->b_l1hdr.b_byteswap = DMU_BSWAP_UINT64;
	} else {
	hdr->b_l1hdr.b_byteswap =
	DMU_OT_BYTESWAP(BP_GET_TYPE(zio->io_bp));
	}
	} else {
	hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS;
	}
	if (!HDR_L2_READING(hdr)) {
	hdr->b_complevel = zio->io_prop.zp_complevel;
	}
	}

	arc_hdr_clear_flags(hdr, ARC_FLAG_L2_EVICTED);
	if (l2arc_noprefetch && HDR_PREFETCH(hdr))
	arc_hdr_clear_flags(hdr, ARC_FLAG_L2CACHE);

	callback_list = hdr->b_l1hdr.b_acb;
	ASSERT3P(callback_list, !=, NULL);
	hdr->b_l1hdr.b_acb = NULL;

	/*
	* If a read request has a callback (i.e. acb_done is not NULL), then we
	* make a buf containing the data according to the parameters which were
	* passed in. The implementation of arc_buf_alloc_impl() ensures that we
	* aren't needlessly decompressing the data multiple times.
	*/
	int callback_cnt = 0;
	for (acb = callback_list; acb != NULL; acb = acb->acb_next) {

	/* We need the last one to call below in original order. */
	callback_list = acb;

	if (!acb->acb_done \|\| acb->acb_nobuf)
	continue;

	callback_cnt++;

	if (zio->io_error != 0)
	continue;

	int error = arc_buf_alloc_impl(hdr, zio->io_spa,
	&acb->acb_zb, acb->acb_private, acb->acb_encrypted,
	acb->acb_compressed, acb->acb_noauth, B_TRUE,
	&acb->acb_buf);

	/*
	* Assert non-speculative zios didn't fail because an
	* encryption key wasn't loaded
	*/
	ASSERT((zio->io_flags & ZIO_FLAG_SPECULATIVE) \|\|
	error != EACCES);

	/*
	* If we failed to decrypt, report an error now (as the zio
	* layer would have done if it had done the transforms).
	*/
	if (error == ECKSUM) {
	ASSERT(BP_IS_PROTECTED(bp));
	error = SET_ERROR(EIO);
	if ((zio->io_flags & ZIO_FLAG_SPECULATIVE) == 0) {
	spa_log_error(zio->io_spa, &acb->acb_zb,
	BP_GET_LOGICAL_BIRTH(zio->io_bp));
	(void) zfs_ereport_post(
	FM_EREPORT_ZFS_AUTHENTICATION,
	zio->io_spa, NULL, &acb->acb_zb, zio, 0);
	}
	}

	if (error != 0) {
	/*
	* Decompression or decryption failed. Set
	* io_error so that when we call acb_done
	* (below), we will indicate that the read
	* failed. Note that in the unusual case
	* where one callback is compressed and another
	* uncompressed, we will mark all of them
	* as failed, even though the uncompressed
	* one can't actually fail. In this case,
	* the hdr will not be anonymous, because
	* if there are multiple callbacks, it's
	* because multiple threads found the same
	* arc buf in the hash table.
	*/
	zio->io_error = error;
	}
	}

	/*
	* If there are multiple callbacks, we must have the hash lock,
	* because the only way for multiple threads to find this hdr is
	* in the hash table. This ensures that if there are multiple
	* callbacks, the hdr is not anonymous. If it were anonymous,
	* we couldn't use arc_buf_destroy() in the error case below.
	*/
	ASSERT(callback_cnt < 2 \|\| hash_lock != NULL);

	if (zio->io_error == 0) {
	arc_hdr_verify(hdr, zio->io_bp);
	} else {
	arc_hdr_set_flags(hdr, ARC_FLAG_IO_ERROR);
	if (hdr->b_l1hdr.b_state != arc_anon)
	arc_change_state(arc_anon, hdr);
	if (HDR_IN_HASH_TABLE(hdr))
	buf_hash_remove(hdr);
	}

	arc_hdr_clear_flags(hdr, ARC_FLAG_IO_IN_PROGRESS);
	(void) remove_reference(hdr, hdr);

	if (hash_lock != NULL)
	mutex_exit(hash_lock);

	/* execute each callback and free its structure */
	while ((acb = callback_list) != NULL) {
	if (acb->acb_done != NULL) {
	if (zio->io_error != 0 && acb->acb_buf != NULL) {
	/*
	* If arc_buf_alloc_impl() fails during
	* decompression, the buf will still be
	* allocated, and needs to be freed here.
	*/
	arc_buf_destroy(acb->acb_buf,
	acb->acb_private);
	acb->acb_buf = NULL;
	}
	acb->acb_done(zio, &zio->io_bookmark, zio->io_bp,
	acb->acb_buf, acb->acb_private);
	}

	if (acb->acb_zio_dummy != NULL) {
	acb->acb_zio_dummy->io_error = zio->io_error;
	zio_nowait(acb->acb_zio_dummy);
	}

	callback_list = acb->acb_prev;
	if (acb->acb_wait) {
	mutex_enter(&acb->acb_wait_lock);
	acb->acb_wait_error = zio->io_error;
	acb->acb_wait = B_FALSE;
	cv_signal(&acb->acb_wait_cv);
	mutex_exit(&acb->acb_wait_lock);
	/* acb will be freed by the waiting thread. */
	} else {
	kmem_free(acb, sizeof (arc_callback_t));
	}
	}
	}

	/*
	* "Read" the block at the specified DVA (in bp) via the
	* cache. If the block is found in the cache, invoke the provided
	* callback immediately and return. Note that the `zio' parameter
	* in the callback will be NULL in this case, since no IO was
	* required. If the block is not in the cache pass the read request
	* on to the spa with a substitute callback function, so that the
	* requested block will be added to the cache.
	*
	* If a read request arrives for a block that has a read in-progress,
	* either wait for the in-progress read to complete (and return the
	* results); or, if this is a read with a "done" func, add a record
	* to the read to invoke the "done" func when the read completes,
	* and return; or just return.
	*
	* arc_read_done() will invoke all the requested "done" functions
	* for readers of this block.
	*/
	int
	arc_read(zio_t pio, spa_t spa, const blkptr_t *bp,
	arc_read_done_func_t done, void private, zio_priority_t priority,
	int zio_flags, arc_flags_t arc_flags, const zbookmark_phys_t zb)
	{
	arc_buf_hdr_t *hdr = NULL;
	kmutex_t *hash_lock = NULL;
	zio_t *rzio;
	uint64_t guid = spa_load_guid(spa);
	boolean_t compressed_read = (zio_flags & ZIO_FLAG_RAW_COMPRESS) != 0;
	boolean_t encrypted_read = BP_IS_ENCRYPTED(bp) &&
	(zio_flags & ZIO_FLAG_RAW_ENCRYPT) != 0;
	boolean_t noauth_read = BP_IS_AUTHENTICATED(bp) &&
	(zio_flags & ZIO_FLAG_RAW_ENCRYPT) != 0;
	boolean_t embedded_bp = !!BP_IS_EMBEDDED(bp);
	boolean_t no_buf = *arc_flags & ARC_FLAG_NO_BUF;
	arc_buf_t *buf = NULL;
	int rc = 0;

	ASSERT(!embedded_bp \|\|
	BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA);
	ASSERT(!BP_IS_HOLE(bp));
	ASSERT(!BP_IS_REDACTED(bp));

	/*
	* Normally SPL_FSTRANS will already be set since kernel threads which
	* expect to call the DMU interfaces will set it when created. System
	* calls are similarly handled by setting/cleaning the bit in the
	* registered callback (module/os/.../zfs/zpl_*).
	*
	* External consumers such as Lustre which call the exported DMU
	* interfaces may not have set SPL_FSTRANS. To avoid a deadlock
	* on the hash_lock always set and clear the bit.
	*/
	fstrans_cookie_t cookie = spl_fstrans_mark();
	top:
	/*
	* Verify the block pointer contents are reasonable. This should
	* always be the case since the blkptr is protected by a checksum.
	* However, if there is damage it's desirable to detect this early
	* and treat it as a checksum error. This allows an alternate blkptr
	* to be tried when one is available (e.g. ditto blocks).
	*/
	if (!zfs_blkptr_verify(spa, bp, (zio_flags & ZIO_FLAG_CONFIG_WRITER) ?
	BLK_CONFIG_HELD : BLK_CONFIG_NEEDED, BLK_VERIFY_LOG)) {
	rc = SET_ERROR(ECKSUM);
	goto done;
	}

	if (!embedded_bp) {
	/*
	* Embedded BP's have no DVA and require no I/O to "read".
	* Create an anonymous arc buf to back it.
	*/
	hdr = buf_hash_find(guid, bp, &hash_lock);
	}

	/*
	* Determine if we have an L1 cache hit or a cache miss. For simplicity
	* we maintain encrypted data separately from compressed / uncompressed
	* data. If the user is requesting raw encrypted data and we don't have
	* that in the header we will read from disk to guarantee that we can
	* get it even if the encryption keys aren't loaded.
	*/
	if (hdr != NULL && HDR_HAS_L1HDR(hdr) && (HDR_HAS_RABD(hdr) \|\|
	(hdr->b_l1hdr.b_pabd != NULL && !encrypted_read))) {
	boolean_t is_data = !HDR_ISTYPE_METADATA(hdr);

	if (HDR_IO_IN_PROGRESS(hdr)) {
	if (*arc_flags & ARC_FLAG_CACHED_ONLY) {
	mutex_exit(hash_lock);
	ARCSTAT_BUMP(arcstat_cached_only_in_progress);
	rc = SET_ERROR(ENOENT);
	goto done;
	}

	zio_t *head_zio = hdr->b_l1hdr.b_acb->acb_zio_head;
	ASSERT3P(head_zio, !=, NULL);
	if ((hdr->b_flags & ARC_FLAG_PRIO_ASYNC_READ) &&
	priority == ZIO_PRIORITY_SYNC_READ) {
	/*
	* This is a sync read that needs to wait for
	* an in-flight async read. Request that the
	* zio have its priority upgraded.
	*/
	zio_change_priority(head_zio, priority);
	DTRACE_PROBE1(arc__async__upgrade__sync,
	arc_buf_hdr_t *, hdr);
	ARCSTAT_BUMP(arcstat_async_upgrade_sync);
	}

	DTRACE_PROBE1(arc__iohit, arc_buf_hdr_t *, hdr);
	arc_access(hdr, *arc_flags, B_FALSE);

	/*
	* If there are multiple threads reading the same block
	* and that block is not yet in the ARC, then only one
	* thread will do the physical I/O and all other
	* threads will wait until that I/O completes.
	* Synchronous reads use the acb_wait_cv whereas nowait
	* reads register a callback. Both are signalled/called
	* in arc_read_done.
	*
	* Errors of the physical I/O may need to be propagated.
	* Synchronous read errors are returned here from
	* arc_read_done via acb_wait_error. Nowait reads
	* attach the acb_zio_dummy zio to pio and
	* arc_read_done propagates the physical I/O's io_error
	* to acb_zio_dummy, and thereby to pio.
	*/
	arc_callback_t *acb = NULL;
	if (done \|\| pio \|\| *arc_flags & ARC_FLAG_WAIT) {
	acb = kmem_zalloc(sizeof (arc_callback_t),
	KM_SLEEP);
	acb->acb_done = done;
	acb->acb_private = private;
	acb->acb_compressed = compressed_read;
	acb->acb_encrypted = encrypted_read;
	acb->acb_noauth = noauth_read;
	acb->acb_nobuf = no_buf;
	if (*arc_flags & ARC_FLAG_WAIT) {
	acb->acb_wait = B_TRUE;
	mutex_init(&acb->acb_wait_lock, NULL,
	MUTEX_DEFAULT, NULL);
	cv_init(&acb->acb_wait_cv, NULL,
	CV_DEFAULT, NULL);
	}
	acb->acb_zb = *zb;
	if (pio != NULL) {
	acb->acb_zio_dummy = zio_null(pio,
	spa, NULL, NULL, NULL, zio_flags);
	}
	acb->acb_zio_head = head_zio;
	acb->acb_next = hdr->b_l1hdr.b_acb;
	hdr->b_l1hdr.b_acb->acb_prev = acb;
	hdr->b_l1hdr.b_acb = acb;
	}
	mutex_exit(hash_lock);

	ARCSTAT_BUMP(arcstat_iohits);
	ARCSTAT_CONDSTAT(!(*arc_flags & ARC_FLAG_PREFETCH),
	demand, prefetch, is_data, data, metadata, iohits);

	if (*arc_flags & ARC_FLAG_WAIT) {
	mutex_enter(&acb->acb_wait_lock);
	while (acb->acb_wait) {
	cv_wait(&acb->acb_wait_cv,
	&acb->acb_wait_lock);
	}
	rc = acb->acb_wait_error;
	mutex_exit(&acb->acb_wait_lock);
	mutex_destroy(&acb->acb_wait_lock);
	cv_destroy(&acb->acb_wait_cv);
	kmem_free(acb, sizeof (arc_callback_t));
	}
	goto out;
	}

	ASSERT(hdr->b_l1hdr.b_state == arc_mru \|\|
	hdr->b_l1hdr.b_state == arc_mfu \|\|
	hdr->b_l1hdr.b_state == arc_uncached);

	DTRACE_PROBE1(arc__hit, arc_buf_hdr_t *, hdr);
	arc_access(hdr, *arc_flags, B_TRUE);

	if (done && !no_buf) {
	ASSERT(!embedded_bp \|\| !BP_IS_HOLE(bp));

	/* Get a buf with the desired data in it. */
	rc = arc_buf_alloc_impl(hdr, spa, zb, private,
	encrypted_read, compressed_read, noauth_read,
	B_TRUE, &buf);
	if (rc == ECKSUM) {
	/*
	* Convert authentication and decryption errors
	* to EIO (and generate an ereport if needed)
	* before leaving the ARC.
	*/
	rc = SET_ERROR(EIO);
	if ((zio_flags & ZIO_FLAG_SPECULATIVE) == 0) {
	spa_log_error(spa, zb, hdr->b_birth);
	(void) zfs_ereport_post(
	FM_EREPORT_ZFS_AUTHENTICATION,
	spa, NULL, zb, NULL, 0);
	}
	}
	if (rc != 0) {
	arc_buf_destroy_impl(buf);
	buf = NULL;
	(void) remove_reference(hdr, private);
	}

	/* assert any errors weren't due to unloaded keys */
	ASSERT((zio_flags & ZIO_FLAG_SPECULATIVE) \|\|
	rc != EACCES);
	}
	mutex_exit(hash_lock);
	ARCSTAT_BUMP(arcstat_hits);
	ARCSTAT_CONDSTAT(!(*arc_flags & ARC_FLAG_PREFETCH),
	demand, prefetch, is_data, data, metadata, hits);
	*arc_flags \|= ARC_FLAG_CACHED;
	goto done;
	} else {
	uint64_t lsize = BP_GET_LSIZE(bp);
	uint64_t psize = BP_GET_PSIZE(bp);
	arc_callback_t *acb;
	vdev_t *vd = NULL;
	uint64_t addr = 0;
	boolean_t devw = B_FALSE;
	uint64_t size;
	abd_t *hdr_abd;
	int alloc_flags = encrypted_read ? ARC_HDR_ALLOC_RDATA : 0;
	arc_buf_contents_t type = BP_GET_BUFC_TYPE(bp);

	if (*arc_flags & ARC_FLAG_CACHED_ONLY) {
	if (hash_lock != NULL)
	mutex_exit(hash_lock);
	rc = SET_ERROR(ENOENT);
	goto done;
	}

	if (hdr == NULL) {
	/*
	* This block is not in the cache or it has
	* embedded data.
	*/
	arc_buf_hdr_t *exists = NULL;
	hdr = arc_hdr_alloc(guid, psize, lsize,
	BP_IS_PROTECTED(bp), BP_GET_COMPRESS(bp), 0, type);

	if (!embedded_bp) {
	hdr->b_dva = *BP_IDENTITY(bp);
	hdr->b_birth = BP_GET_BIRTH(bp);
	exists = buf_hash_insert(hdr, &hash_lock);
	}
	if (exists != NULL) {
	/* somebody beat us to the hash insert */
	mutex_exit(hash_lock);
	buf_discard_identity(hdr);
	arc_hdr_destroy(hdr);
	goto top; /* restart the IO request */
	}
	} else {
	/*
	* This block is in the ghost cache or encrypted data
	* was requested and we didn't have it. If it was
	* L2-only (and thus didn't have an L1 hdr),
	* we realloc the header to add an L1 hdr.
	*/
	if (!HDR_HAS_L1HDR(hdr)) {
	hdr = arc_hdr_realloc(hdr, hdr_l2only_cache,
	hdr_full_cache);
	}

	if (GHOST_STATE(hdr->b_l1hdr.b_state)) {
	ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
	ASSERT(!HDR_HAS_RABD(hdr));
	ASSERT(!HDR_IO_IN_PROGRESS(hdr));
	ASSERT0(zfs_refcount_count(
	&hdr->b_l1hdr.b_refcnt));
	ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL);
	#ifdef ZFS_DEBUG
	ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL);
	#endif
	} else if (HDR_IO_IN_PROGRESS(hdr)) {
	/*
	* If this header already had an IO in progress
	* and we are performing another IO to fetch
	* encrypted data we must wait until the first
	* IO completes so as not to confuse
	* arc_read_done(). This should be very rare
	* and so the performance impact shouldn't
	* matter.
	*/
	arc_callback_t *acb = kmem_zalloc(
	sizeof (arc_callback_t), KM_SLEEP);
	acb->acb_wait = B_TRUE;
	mutex_init(&acb->acb_wait_lock, NULL,
	MUTEX_DEFAULT, NULL);
	cv_init(&acb->acb_wait_cv, NULL, CV_DEFAULT,
	NULL);
	acb->acb_zio_head =
	hdr->b_l1hdr.b_acb->acb_zio_head;
	acb->acb_next = hdr->b_l1hdr.b_acb;
	hdr->b_l1hdr.b_acb->acb_prev = acb;
	hdr->b_l1hdr.b_acb = acb;
	mutex_exit(hash_lock);
	mutex_enter(&acb->acb_wait_lock);
	while (acb->acb_wait) {
	cv_wait(&acb->acb_wait_cv,
	&acb->acb_wait_lock);
	}
	mutex_exit(&acb->acb_wait_lock);
	mutex_destroy(&acb->acb_wait_lock);
	cv_destroy(&acb->acb_wait_cv);
	kmem_free(acb, sizeof (arc_callback_t));
	goto top;
	}
	}
	if (*arc_flags & ARC_FLAG_UNCACHED) {
	arc_hdr_set_flags(hdr, ARC_FLAG_UNCACHED);
	if (!encrypted_read)
	alloc_flags \|= ARC_HDR_ALLOC_LINEAR;
	}

	/*
	* Take additional reference for IO_IN_PROGRESS. It stops
	* arc_access() from putting this header without any buffers
	* and so other references but obviously nonevictable onto
	* the evictable list of MRU or MFU state.
	*/
	add_reference(hdr, hdr);
	if (!embedded_bp)
	arc_access(hdr, *arc_flags, B_FALSE);
	arc_hdr_set_flags(hdr, ARC_FLAG_IO_IN_PROGRESS);
	arc_hdr_alloc_abd(hdr, alloc_flags);
	if (encrypted_read) {
	ASSERT(HDR_HAS_RABD(hdr));
	size = HDR_GET_PSIZE(hdr);
	hdr_abd = hdr->b_crypt_hdr.b_rabd;
	zio_flags \|= ZIO_FLAG_RAW;
	} else {
	ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);
	size = arc_hdr_size(hdr);
	hdr_abd = hdr->b_l1hdr.b_pabd;

	if (arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF) {
	zio_flags \|= ZIO_FLAG_RAW_COMPRESS;
	}

	/*
	* For authenticated bp's, we do not ask the ZIO layer
	* to authenticate them since this will cause the entire
	* IO to fail if the key isn't loaded. Instead, we
	* defer authentication until arc_buf_fill(), which will
	* verify the data when the key is available.
	*/
	if (BP_IS_AUTHENTICATED(bp))
	zio_flags \|= ZIO_FLAG_RAW_ENCRYPT;
	}

	if (BP_IS_AUTHENTICATED(bp))
	arc_hdr_set_flags(hdr, ARC_FLAG_NOAUTH);
	if (BP_GET_LEVEL(bp) > 0)
	arc_hdr_set_flags(hdr, ARC_FLAG_INDIRECT);
	ASSERT(!GHOST_STATE(hdr->b_l1hdr.b_state));

	acb = kmem_zalloc(sizeof (arc_callback_t), KM_SLEEP);
	acb->acb_done = done;
	acb->acb_private = private;
	acb->acb_compressed = compressed_read;
	acb->acb_encrypted = encrypted_read;
	acb->acb_noauth = noauth_read;
	acb->acb_zb = *zb;

	ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL);
	hdr->b_l1hdr.b_acb = acb;

	if (HDR_HAS_L2HDR(hdr) &&
	(vd = hdr->b_l2hdr.b_dev->l2ad_vdev) != NULL) {
	devw = hdr->b_l2hdr.b_dev->l2ad_writing;
	addr = hdr->b_l2hdr.b_daddr;
	/*
	* Lock out L2ARC device removal.
	*/
	if (vdev_is_dead(vd) \|\|
	!spa_config_tryenter(spa, SCL_L2ARC, vd, RW_READER))
	vd = NULL;
	}

	/*
	* We count both async reads and scrub IOs as asynchronous so
	* that both can be upgraded in the event of a cache hit while
	* the read IO is still in-flight.
	*/
	if (priority == ZIO_PRIORITY_ASYNC_READ \|\|
	priority == ZIO_PRIORITY_SCRUB)
	arc_hdr_set_flags(hdr, ARC_FLAG_PRIO_ASYNC_READ);
	else
	arc_hdr_clear_flags(hdr, ARC_FLAG_PRIO_ASYNC_READ);

	/*
	* At this point, we have a level 1 cache miss or a blkptr
	* with embedded data. Try again in L2ARC if possible.
	*/
	ASSERT3U(HDR_GET_LSIZE(hdr), ==, lsize);

	/*
	* Skip ARC stat bump for block pointers with embedded
	* data. The data are read from the blkptr itself via
	* decode_embedded_bp_compressed().
	*/
	if (!embedded_bp) {
	DTRACE_PROBE4(arc__miss, arc_buf_hdr_t *, hdr,
	blkptr_t *, bp, uint64_t, lsize,
	zbookmark_phys_t *, zb);
	ARCSTAT_BUMP(arcstat_misses);
	ARCSTAT_CONDSTAT(!(*arc_flags & ARC_FLAG_PREFETCH),
	demand, prefetch, !HDR_ISTYPE_METADATA(hdr), data,
	metadata, misses);
	zfs_racct_read(size, 1);
	}

	/* Check if the spa even has l2 configured */
	const boolean_t spa_has_l2 = l2arc_ndev != 0 &&
	spa->spa_l2cache.sav_count > 0;

	if (vd != NULL && spa_has_l2 && !(l2arc_norw && devw)) {
	/*
	* Read from the L2ARC if the following are true:
	* 1. The L2ARC vdev was previously cached.
	* 2. This buffer still has L2ARC metadata.
	* 3. This buffer isn't currently writing to the L2ARC.
	* 4. The L2ARC entry wasn't evicted, which may
	* also have invalidated the vdev.
	*/
	if (HDR_HAS_L2HDR(hdr) &&
	!HDR_L2_WRITING(hdr) && !HDR_L2_EVICTED(hdr)) {
	l2arc_read_callback_t *cb;
	abd_t *abd;
	uint64_t asize;

	DTRACE_PROBE1(l2arc__hit, arc_buf_hdr_t *, hdr);
	ARCSTAT_BUMP(arcstat_l2_hits);
	hdr->b_l2hdr.b_hits++;

	cb = kmem_zalloc(sizeof (l2arc_read_callback_t),
	KM_SLEEP);
	cb->l2rcb_hdr = hdr;
	cb->l2rcb_bp = *bp;
	cb->l2rcb_zb = *zb;
	cb->l2rcb_flags = zio_flags;

	/*
	* When Compressed ARC is disabled, but the
	* L2ARC block is compressed, arc_hdr_size()
	* will have returned LSIZE rather than PSIZE.
	*/
	if (HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF &&
	!HDR_COMPRESSION_ENABLED(hdr) &&
	HDR_GET_PSIZE(hdr) != 0) {
	size = HDR_GET_PSIZE(hdr);
	}

	asize = vdev_psize_to_asize(vd, size);
	if (asize != size) {
	abd = abd_alloc_for_io(asize,
	HDR_ISTYPE_METADATA(hdr));
	cb->l2rcb_abd = abd;
	} else {
	abd = hdr_abd;
	}

	ASSERT(addr >= VDEV_LABEL_START_SIZE &&
	addr + asize <= vd->vdev_psize -
	VDEV_LABEL_END_SIZE);

	/*
	* l2arc read. The SCL_L2ARC lock will be
	* released by l2arc_read_done().
	* Issue a null zio if the underlying buffer
	* was squashed to zero size by compression.
	*/
	ASSERT3U(arc_hdr_get_compress(hdr), !=,
	ZIO_COMPRESS_EMPTY);
	rzio = zio_read_phys(pio, vd, addr,
	asize, abd,
	ZIO_CHECKSUM_OFF,
	l2arc_read_done, cb, priority,
	zio_flags \| ZIO_FLAG_CANFAIL \|
	ZIO_FLAG_DONT_PROPAGATE \|
	ZIO_FLAG_DONT_RETRY, B_FALSE);
	acb->acb_zio_head = rzio;

	if (hash_lock != NULL)
	mutex_exit(hash_lock);

	DTRACE_PROBE2(l2arc__read, vdev_t *, vd,
	zio_t *, rzio);
	ARCSTAT_INCR(arcstat_l2_read_bytes,
	HDR_GET_PSIZE(hdr));

	if (*arc_flags & ARC_FLAG_NOWAIT) {
	zio_nowait(rzio);
	goto out;
	}

	ASSERT(*arc_flags & ARC_FLAG_WAIT);
	if (zio_wait(rzio) == 0)
	goto out;

	/* l2arc read error; goto zio_read() */
	if (hash_lock != NULL)
	mutex_enter(hash_lock);
	} else {
	DTRACE_PROBE1(l2arc__miss,
	arc_buf_hdr_t *, hdr);
	ARCSTAT_BUMP(arcstat_l2_misses);
	if (HDR_L2_WRITING(hdr))
	ARCSTAT_BUMP(arcstat_l2_rw_clash);
	spa_config_exit(spa, SCL_L2ARC, vd);
	}
	} else {
	if (vd != NULL)
	spa_config_exit(spa, SCL_L2ARC, vd);

	/*
	* Only a spa with l2 should contribute to l2
	* miss stats. (Including the case of having a
	* faulted cache device - that's also a miss.)
	*/
	if (spa_has_l2) {
	/*
	* Skip ARC stat bump for block pointers with
	* embedded data. The data are read from the
	* blkptr itself via
	* decode_embedded_bp_compressed().
	*/
	if (!embedded_bp) {
	DTRACE_PROBE1(l2arc__miss,
	arc_buf_hdr_t *, hdr);
	ARCSTAT_BUMP(arcstat_l2_misses);
	}
	}
	}

	rzio = zio_read(pio, spa, bp, hdr_abd, size,
	arc_read_done, hdr, priority, zio_flags, zb);
	acb->acb_zio_head = rzio;

	if (hash_lock != NULL)
	mutex_exit(hash_lock);

	if (*arc_flags & ARC_FLAG_WAIT) {
	rc = zio_wait(rzio);
	goto out;
	}

	ASSERT(*arc_flags & ARC_FLAG_NOWAIT);
	zio_nowait(rzio);
	}

	out:
	/* embedded bps don't actually go to disk */
	if (!embedded_bp)
	spa_read_history_add(spa, zb, *arc_flags);
	spl_fstrans_unmark(cookie);
	return (rc);

	done:
	if (done)
	done(NULL, zb, bp, buf, private);
	if (pio && rc != 0) {
	zio_t *zio = zio_null(pio, spa, NULL, NULL, NULL, zio_flags);
	zio->io_error = rc;
	zio_nowait(zio);
	}
	goto out;
	}

	arc_prune_t *
	arc_add_prune_callback(arc_prune_func_t func, void private)
	{
	arc_prune_t *p;

	p = kmem_alloc(sizeof (*p), KM_SLEEP);
	p->p_pfunc = func;
	p->p_private = private;
	list_link_init(&p->p_node);
	zfs_refcount_create(&p->p_refcnt);

	mutex_enter(&arc_prune_mtx);
	zfs_refcount_add(&p->p_refcnt, &arc_prune_list);
	list_insert_head(&arc_prune_list, p);
	mutex_exit(&arc_prune_mtx);

	return (p);
	}

	void
	arc_remove_prune_callback(arc_prune_t *p)
	{
	boolean_t wait = B_FALSE;
	mutex_enter(&arc_prune_mtx);
	list_remove(&arc_prune_list, p);
	if (zfs_refcount_remove(&p->p_refcnt, &arc_prune_list) > 0)
	wait = B_TRUE;
	mutex_exit(&arc_prune_mtx);

	/* wait for arc_prune_task to finish */
	if (wait)
	taskq_wait_outstanding(arc_prune_taskq, 0);
	ASSERT0(zfs_refcount_count(&p->p_refcnt));
	zfs_refcount_destroy(&p->p_refcnt);
	kmem_free(p, sizeof (*p));
	}

	/*
	* Helper function for arc_prune_async() it is responsible for safely
	* handling the execution of a registered arc_prune_func_t.
	*/
	static void
	arc_prune_task(void *ptr)
	{
	arc_prune_t ap = (arc_prune_t )ptr;
	arc_prune_func_t *func = ap->p_pfunc;

	if (func != NULL)
	func(ap->p_adjust, ap->p_private);

	(void) zfs_refcount_remove(&ap->p_refcnt, func);
	}

	/*
	* Notify registered consumers they must drop holds on a portion of the ARC
	* buffers they reference. This provides a mechanism to ensure the ARC can
	* honor the metadata limit and reclaim otherwise pinned ARC buffers.
	*
	* This operation is performed asynchronously so it may be safely called
	* in the context of the arc_reclaim_thread(). A reference is taken here
	* for each registered arc_prune_t and the arc_prune_task() is responsible
	* for releasing it once the registered arc_prune_func_t has completed.
	*/
	static void
	arc_prune_async(uint64_t adjust)
	{
	arc_prune_t *ap;

	mutex_enter(&arc_prune_mtx);
	for (ap = list_head(&arc_prune_list); ap != NULL;
	ap = list_next(&arc_prune_list, ap)) {

	if (zfs_refcount_count(&ap->p_refcnt) >= 2)
	continue;

	zfs_refcount_add(&ap->p_refcnt, ap->p_pfunc);
	ap->p_adjust = adjust;
	if (taskq_dispatch(arc_prune_taskq, arc_prune_task,
	ap, TQ_SLEEP) == TASKQID_INVALID) {
	(void) zfs_refcount_remove(&ap->p_refcnt, ap->p_pfunc);
	continue;
	}
	ARCSTAT_BUMP(arcstat_prune);
	}
	mutex_exit(&arc_prune_mtx);
	}

	/*
	* Notify the arc that a block was freed, and thus will never be used again.
	*/
	void
	arc_freed(spa_t spa, const blkptr_t bp)
	{
	arc_buf_hdr_t *hdr;
	kmutex_t *hash_lock;
	uint64_t guid = spa_load_guid(spa);

	ASSERT(!BP_IS_EMBEDDED(bp));

	hdr = buf_hash_find(guid, bp, &hash_lock);
	if (hdr == NULL)
	return;

	/*
	* We might be trying to free a block that is still doing I/O
	* (i.e. prefetch) or has some other reference (i.e. a dedup-ed,
	* dmu_sync-ed block). A block may also have a reference if it is
	* part of a dedup-ed, dmu_synced write. The dmu_sync() function would
	* have written the new block to its final resting place on disk but
	* without the dedup flag set. This would have left the hdr in the MRU
	* state and discoverable. When the txg finally syncs it detects that
	* the block was overridden in open context and issues an override I/O.
	* Since this is a dedup block, the override I/O will determine if the
	* block is already in the DDT. If so, then it will replace the io_bp
	* with the bp from the DDT and allow the I/O to finish. When the I/O
	* reaches the done callback, dbuf_write_override_done, it will
	* check to see if the io_bp and io_bp_override are identical.
	* If they are not, then it indicates that the bp was replaced with
	* the bp in the DDT and the override bp is freed. This allows
	* us to arrive here with a reference on a block that is being
	* freed. So if we have an I/O in progress, or a reference to
	* this hdr, then we don't destroy the hdr.
	*/
	if (!HDR_HAS_L1HDR(hdr) \|\|
	zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt)) {
	arc_change_state(arc_anon, hdr);
	arc_hdr_destroy(hdr);
	mutex_exit(hash_lock);
	} else {
	mutex_exit(hash_lock);
	}

	}

	/*
	* Release this buffer from the cache, making it an anonymous buffer. This
	* must be done after a read and prior to modifying the buffer contents.
	* If the buffer has more than one reference, we must make
	* a new hdr for the buffer.
	*/
	void
	arc_release(arc_buf_t buf, const void tag)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;

	/*
	* It would be nice to assert that if its DMU metadata (level >
	* 0 \|\| it's the dnode file), then it must be syncing context.
	* But we don't know that information at this level.
	*/

	ASSERT(HDR_HAS_L1HDR(hdr));

	/*
	* We don't grab the hash lock prior to this check, because if
	* the buffer's header is in the arc_anon state, it won't be
	* linked into the hash table.
	*/
	if (hdr->b_l1hdr.b_state == arc_anon) {
	ASSERT(!HDR_IO_IN_PROGRESS(hdr));
	ASSERT(!HDR_IN_HASH_TABLE(hdr));
	ASSERT(!HDR_HAS_L2HDR(hdr));

	ASSERT3P(hdr->b_l1hdr.b_buf, ==, buf);
	ASSERT(ARC_BUF_LAST(buf));
	ASSERT3S(zfs_refcount_count(&hdr->b_l1hdr.b_refcnt), ==, 1);
	ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node));

	hdr->b_l1hdr.b_arc_access = 0;

	/*
	* If the buf is being overridden then it may already
	* have a hdr that is not empty.
	*/
	buf_discard_identity(hdr);
	arc_buf_thaw(buf);

	return;
	}

	kmutex_t *hash_lock = HDR_LOCK(hdr);
	mutex_enter(hash_lock);

	/*
	* This assignment is only valid as long as the hash_lock is
	* held, we must be careful not to reference state or the
	* b_state field after dropping the lock.
	*/
	arc_state_t *state = hdr->b_l1hdr.b_state;
	ASSERT3P(hash_lock, ==, HDR_LOCK(hdr));
	ASSERT3P(state, !=, arc_anon);

	/* this buffer is not on any list */
	ASSERT3S(zfs_refcount_count(&hdr->b_l1hdr.b_refcnt), >, 0);

	if (HDR_HAS_L2HDR(hdr)) {
	mutex_enter(&hdr->b_l2hdr.b_dev->l2ad_mtx);

	/*
	* We have to recheck this conditional again now that
	* we're holding the l2ad_mtx to prevent a race with
	* another thread which might be concurrently calling
	* l2arc_evict(). In that case, l2arc_evict() might have
	* destroyed the header's L2 portion as we were waiting
	* to acquire the l2ad_mtx.
	*/
	if (HDR_HAS_L2HDR(hdr))
	arc_hdr_l2hdr_destroy(hdr);

	mutex_exit(&hdr->b_l2hdr.b_dev->l2ad_mtx);
	}

	/*
	* Do we have more than one buf?
	*/
	if (hdr->b_l1hdr.b_buf != buf \|\| !ARC_BUF_LAST(buf)) {
	arc_buf_hdr_t *nhdr;
	uint64_t spa = hdr->b_spa;
	uint64_t psize = HDR_GET_PSIZE(hdr);
	uint64_t lsize = HDR_GET_LSIZE(hdr);
	boolean_t protected = HDR_PROTECTED(hdr);
	enum zio_compress compress = arc_hdr_get_compress(hdr);
	arc_buf_contents_t type = arc_buf_type(hdr);
	VERIFY3U(hdr->b_type, ==, type);

	ASSERT(hdr->b_l1hdr.b_buf != buf \|\| buf->b_next != NULL);
	VERIFY3S(remove_reference(hdr, tag), >, 0);

	if (ARC_BUF_SHARED(buf) && !ARC_BUF_COMPRESSED(buf)) {
	ASSERT3P(hdr->b_l1hdr.b_buf, !=, buf);
	ASSERT(ARC_BUF_LAST(buf));
	}

	/*
	* Pull the data off of this hdr and attach it to
	* a new anonymous hdr. Also find the last buffer
	* in the hdr's buffer list.
	*/
	arc_buf_t *lastbuf = arc_buf_remove(hdr, buf);
	ASSERT3P(lastbuf, !=, NULL);

	/*
	* If the current arc_buf_t and the hdr are sharing their data
	* buffer, then we must stop sharing that block.
	*/
	if (ARC_BUF_SHARED(buf)) {
	ASSERT3P(hdr->b_l1hdr.b_buf, !=, buf);
	ASSERT(!arc_buf_is_shared(lastbuf));

	/*
	* First, sever the block sharing relationship between
	* buf and the arc_buf_hdr_t.
	*/
	arc_unshare_buf(hdr, buf);

	/*
	* Now we need to recreate the hdr's b_pabd. Since we
	* have lastbuf handy, we try to share with it, but if
	* we can't then we allocate a new b_pabd and copy the
	* data from buf into it.
	*/
	if (arc_can_share(hdr, lastbuf)) {
	arc_share_buf(hdr, lastbuf);
	} else {
	arc_hdr_alloc_abd(hdr, 0);
	abd_copy_from_buf(hdr->b_l1hdr.b_pabd,
	buf->b_data, psize);
	}
	VERIFY3P(lastbuf->b_data, !=, NULL);
	} else if (HDR_SHARED_DATA(hdr)) {
	/*
	* Uncompressed shared buffers are always at the end
	* of the list. Compressed buffers don't have the
	* same requirements. This makes it hard to
	* simply assert that the lastbuf is shared so
	* we rely on the hdr's compression flags to determine
	* if we have a compressed, shared buffer.
	*/
	ASSERT(arc_buf_is_shared(lastbuf) \|\|
	arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF);
	ASSERT(!arc_buf_is_shared(buf));
	}

	ASSERT(hdr->b_l1hdr.b_pabd != NULL \|\| HDR_HAS_RABD(hdr));
	ASSERT3P(state, !=, arc_l2c_only);

	(void) zfs_refcount_remove_many(&state->arcs_size[type],
	arc_buf_size(buf), buf);

	if (zfs_refcount_is_zero(&hdr->b_l1hdr.b_refcnt)) {
	ASSERT3P(state, !=, arc_l2c_only);
	(void) zfs_refcount_remove_many(
	&state->arcs_esize[type],
	arc_buf_size(buf), buf);
	}

	arc_cksum_verify(buf);
	arc_buf_unwatch(buf);

	/* if this is the last uncompressed buf free the checksum */
	if (!arc_hdr_has_uncompressed_buf(hdr))
	arc_cksum_free(hdr);

	mutex_exit(hash_lock);

	nhdr = arc_hdr_alloc(spa, psize, lsize, protected,
	compress, hdr->b_complevel, type);
	ASSERT3P(nhdr->b_l1hdr.b_buf, ==, NULL);
	ASSERT0(zfs_refcount_count(&nhdr->b_l1hdr.b_refcnt));
	VERIFY3U(nhdr->b_type, ==, type);
	ASSERT(!HDR_SHARED_DATA(nhdr));

	nhdr->b_l1hdr.b_buf = buf;
	(void) zfs_refcount_add(&nhdr->b_l1hdr.b_refcnt, tag);
	buf->b_hdr = nhdr;

	(void) zfs_refcount_add_many(&arc_anon->arcs_size[type],
	arc_buf_size(buf), buf);
	} else {
	ASSERT(zfs_refcount_count(&hdr->b_l1hdr.b_refcnt) == 1);
	/* protected by hash lock, or hdr is on arc_anon */
	ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node));
	ASSERT(!HDR_IO_IN_PROGRESS(hdr));
	hdr->b_l1hdr.b_mru_hits = 0;
	hdr->b_l1hdr.b_mru_ghost_hits = 0;
	hdr->b_l1hdr.b_mfu_hits = 0;
	hdr->b_l1hdr.b_mfu_ghost_hits = 0;
	arc_change_state(arc_anon, hdr);
	hdr->b_l1hdr.b_arc_access = 0;

	mutex_exit(hash_lock);
	buf_discard_identity(hdr);
	arc_buf_thaw(buf);
	}
	}

	int
	arc_released(arc_buf_t *buf)
	{
	return (buf->b_data != NULL &&
	buf->b_hdr->b_l1hdr.b_state == arc_anon);
	}

	#ifdef ZFS_DEBUG
	int
	arc_referenced(arc_buf_t *buf)
	{
	return (zfs_refcount_count(&buf->b_hdr->b_l1hdr.b_refcnt));
	}
	#endif

	static void
	arc_write_ready(zio_t *zio)
	{
	arc_write_callback_t *callback = zio->io_private;
	arc_buf_t *buf = callback->awcb_buf;
	arc_buf_hdr_t *hdr = buf->b_hdr;
	blkptr_t *bp = zio->io_bp;
	uint64_t psize = BP_IS_HOLE(bp) ? 0 : BP_GET_PSIZE(bp);
	fstrans_cookie_t cookie = spl_fstrans_mark();

	ASSERT(HDR_HAS_L1HDR(hdr));
	ASSERT(!zfs_refcount_is_zero(&buf->b_hdr->b_l1hdr.b_refcnt));
	ASSERT3P(hdr->b_l1hdr.b_buf, !=, NULL);

	/*
	* If we're reexecuting this zio because the pool suspended, then
	* cleanup any state that was previously set the first time the
	* callback was invoked.
	*/
	if (zio->io_flags & ZIO_FLAG_REEXECUTED) {
	arc_cksum_free(hdr);
	arc_buf_unwatch(buf);
	if (hdr->b_l1hdr.b_pabd != NULL) {
	if (ARC_BUF_SHARED(buf)) {
	arc_unshare_buf(hdr, buf);
	} else {
	ASSERT(!arc_buf_is_shared(buf));
	arc_hdr_free_abd(hdr, B_FALSE);
	}
	}

	if (HDR_HAS_RABD(hdr))
	arc_hdr_free_abd(hdr, B_TRUE);
	}
	ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);
	ASSERT(!HDR_HAS_RABD(hdr));
	ASSERT(!HDR_SHARED_DATA(hdr));
	ASSERT(!arc_buf_is_shared(buf));

	callback->awcb_ready(zio, buf, callback->awcb_private);

	if (HDR_IO_IN_PROGRESS(hdr)) {
	ASSERT(zio->io_flags & ZIO_FLAG_REEXECUTED);
	} else {
	arc_hdr_set_flags(hdr, ARC_FLAG_IO_IN_PROGRESS);
	add_reference(hdr, hdr); /* For IO_IN_PROGRESS. */
	}

	if (BP_IS_PROTECTED(bp)) {
	/* ZIL blocks are written through zio_rewrite */
	ASSERT3U(BP_GET_TYPE(bp), !=, DMU_OT_INTENT_LOG);

	if (BP_SHOULD_BYTESWAP(bp)) {
	if (BP_GET_LEVEL(bp) > 0) {
	hdr->b_l1hdr.b_byteswap = DMU_BSWAP_UINT64;
	} else {
	hdr->b_l1hdr.b_byteswap =
	DMU_OT_BYTESWAP(BP_GET_TYPE(bp));
	}
	} else {
	hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS;
	}

	arc_hdr_set_flags(hdr, ARC_FLAG_PROTECTED);
	hdr->b_crypt_hdr.b_ot = BP_GET_TYPE(bp);
	hdr->b_crypt_hdr.b_dsobj = zio->io_bookmark.zb_objset;
	zio_crypt_decode_params_bp(bp, hdr->b_crypt_hdr.b_salt,
	hdr->b_crypt_hdr.b_iv);
	zio_crypt_decode_mac_bp(bp, hdr->b_crypt_hdr.b_mac);
	} else {
	arc_hdr_clear_flags(hdr, ARC_FLAG_PROTECTED);
	}

	/*
	* If this block was written for raw encryption but the zio layer
	* ended up only authenticating it, adjust the buffer flags now.
	*/
	if (BP_IS_AUTHENTICATED(bp) && ARC_BUF_ENCRYPTED(buf)) {
	arc_hdr_set_flags(hdr, ARC_FLAG_NOAUTH);
	buf->b_flags &= ~ARC_BUF_FLAG_ENCRYPTED;
	if (BP_GET_COMPRESS(bp) == ZIO_COMPRESS_OFF)
	buf->b_flags &= ~ARC_BUF_FLAG_COMPRESSED;
	} else if (BP_IS_HOLE(bp) && ARC_BUF_ENCRYPTED(buf)) {
	buf->b_flags &= ~ARC_BUF_FLAG_ENCRYPTED;
	buf->b_flags &= ~ARC_BUF_FLAG_COMPRESSED;
	}

	/* this must be done after the buffer flags are adjusted */
	arc_cksum_compute(buf);

	enum zio_compress compress;
	if (BP_IS_HOLE(bp) \|\| BP_IS_EMBEDDED(bp)) {
	compress = ZIO_COMPRESS_OFF;
	} else {
	ASSERT3U(HDR_GET_LSIZE(hdr), ==, BP_GET_LSIZE(bp));
	compress = BP_GET_COMPRESS(bp);
	}
	HDR_SET_PSIZE(hdr, psize);
	arc_hdr_set_compress(hdr, compress);
	hdr->b_complevel = zio->io_prop.zp_complevel;

	if (zio->io_error != 0 \|\| psize == 0)
	goto out;

	/*
	* Fill the hdr with data. If the buffer is encrypted we have no choice
	* but to copy the data into b_radb. If the hdr is compressed, the data
	* we want is available from the zio, otherwise we can take it from
	* the buf.
	*
	* We might be able to share the buf's data with the hdr here. However,
	* doing so would cause the ARC to be full of linear ABDs if we write a
	* lot of shareable data. As a compromise, we check whether scattered
	* ABDs are allowed, and assume that if they are then the user wants
	* the ARC to be primarily filled with them regardless of the data being
	* written. Therefore, if they're allowed then we allocate one and copy
	* the data into it; otherwise, we share the data directly if we can.
	*/
	if (ARC_BUF_ENCRYPTED(buf)) {
	ASSERT3U(psize, >, 0);
	ASSERT(ARC_BUF_COMPRESSED(buf));
	arc_hdr_alloc_abd(hdr, ARC_HDR_ALLOC_RDATA \|
	ARC_HDR_USE_RESERVE);
	abd_copy(hdr->b_crypt_hdr.b_rabd, zio->io_abd, psize);
	} else if (!(HDR_UNCACHED(hdr) \|\|
	abd_size_alloc_linear(arc_buf_size(buf))) \|\|
	!arc_can_share(hdr, buf)) {
	/*
	* Ideally, we would always copy the io_abd into b_pabd, but the
	* user may have disabled compressed ARC, thus we must check the
	* hdr's compression setting rather than the io_bp's.
	*/
	if (BP_IS_ENCRYPTED(bp)) {
	ASSERT3U(psize, >, 0);
	arc_hdr_alloc_abd(hdr, ARC_HDR_ALLOC_RDATA \|
	ARC_HDR_USE_RESERVE);
	abd_copy(hdr->b_crypt_hdr.b_rabd, zio->io_abd, psize);
	} else if (arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF &&
	!ARC_BUF_COMPRESSED(buf)) {
	ASSERT3U(psize, >, 0);
	arc_hdr_alloc_abd(hdr, ARC_HDR_USE_RESERVE);
	abd_copy(hdr->b_l1hdr.b_pabd, zio->io_abd, psize);
	} else {
	ASSERT3U(zio->io_orig_size, ==, arc_hdr_size(hdr));
	arc_hdr_alloc_abd(hdr, ARC_HDR_USE_RESERVE);
	abd_copy_from_buf(hdr->b_l1hdr.b_pabd, buf->b_data,
	arc_buf_size(buf));
	}
	} else {
	ASSERT3P(buf->b_data, ==, abd_to_buf(zio->io_orig_abd));
	ASSERT3U(zio->io_orig_size, ==, arc_buf_size(buf));
	ASSERT3P(hdr->b_l1hdr.b_buf, ==, buf);
	ASSERT(ARC_BUF_LAST(buf));

	arc_share_buf(hdr, buf);
	}

	out:
	arc_hdr_verify(hdr, bp);
	spl_fstrans_unmark(cookie);
	}

	static void
	arc_write_children_ready(zio_t *zio)
	{
	arc_write_callback_t *callback = zio->io_private;
	arc_buf_t *buf = callback->awcb_buf;

	callback->awcb_children_ready(zio, buf, callback->awcb_private);
	}

	static void
	arc_write_done(zio_t *zio)
	{
	arc_write_callback_t *callback = zio->io_private;
	arc_buf_t *buf = callback->awcb_buf;
	arc_buf_hdr_t *hdr = buf->b_hdr;

	ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL);

	if (zio->io_error == 0) {
	arc_hdr_verify(hdr, zio->io_bp);

	if (BP_IS_HOLE(zio->io_bp) \|\| BP_IS_EMBEDDED(zio->io_bp)) {
	buf_discard_identity(hdr);
	} else {
	hdr->b_dva = *BP_IDENTITY(zio->io_bp);
	hdr->b_birth = BP_GET_BIRTH(zio->io_bp);
	}
	} else {
	ASSERT(HDR_EMPTY(hdr));
	}

	/*
	* If the block to be written was all-zero or compressed enough to be
	* embedded in the BP, no write was performed so there will be no
	* dva/birth/checksum. The buffer must therefore remain anonymous
	* (and uncached).
	*/
	if (!HDR_EMPTY(hdr)) {
	arc_buf_hdr_t *exists;
	kmutex_t *hash_lock;

	ASSERT3U(zio->io_error, ==, 0);

	arc_cksum_verify(buf);

	exists = buf_hash_insert(hdr, &hash_lock);
	if (exists != NULL) {
	/*
	* This can only happen if we overwrite for
	* sync-to-convergence, because we remove
	* buffers from the hash table when we arc_free().
	*/
	if (zio->io_flags & ZIO_FLAG_IO_REWRITE) {
	if (!BP_EQUAL(&zio->io_bp_orig, zio->io_bp))
	panic("bad overwrite, hdr=%p exists=%p",
	(void )hdr, (void )exists);
	ASSERT(zfs_refcount_is_zero(
	&exists->b_l1hdr.b_refcnt));
	arc_change_state(arc_anon, exists);
	arc_hdr_destroy(exists);
	mutex_exit(hash_lock);
	exists = buf_hash_insert(hdr, &hash_lock);
	ASSERT3P(exists, ==, NULL);
	} else if (zio->io_flags & ZIO_FLAG_NOPWRITE) {
	/* nopwrite */
	ASSERT(zio->io_prop.zp_nopwrite);
	if (!BP_EQUAL(&zio->io_bp_orig, zio->io_bp))
	panic("bad nopwrite, hdr=%p exists=%p",
	(void )hdr, (void )exists);
	} else {
	/* Dedup */
	ASSERT3P(hdr->b_l1hdr.b_buf, !=, NULL);
	ASSERT(ARC_BUF_LAST(hdr->b_l1hdr.b_buf));
	ASSERT(hdr->b_l1hdr.b_state == arc_anon);
	ASSERT(BP_GET_DEDUP(zio->io_bp));
	ASSERT(BP_GET_LEVEL(zio->io_bp) == 0);
	}
	}
	arc_hdr_clear_flags(hdr, ARC_FLAG_IO_IN_PROGRESS);
	VERIFY3S(remove_reference(hdr, hdr), >, 0);
	/* if it's not anon, we are doing a scrub */
	if (exists == NULL && hdr->b_l1hdr.b_state == arc_anon)
	arc_access(hdr, 0, B_FALSE);
	mutex_exit(hash_lock);
	} else {
	arc_hdr_clear_flags(hdr, ARC_FLAG_IO_IN_PROGRESS);
	VERIFY3S(remove_reference(hdr, hdr), >, 0);
	}

	callback->awcb_done(zio, buf, callback->awcb_private);

	abd_free(zio->io_abd);
	kmem_free(callback, sizeof (arc_write_callback_t));
	}

	zio_t *
	arc_write(zio_t pio, spa_t spa, uint64_t txg,
	blkptr_t bp, arc_buf_t buf, boolean_t uncached, boolean_t l2arc,
	const zio_prop_t zp, arc_write_done_func_t ready,
	arc_write_done_func_t children_ready, arc_write_done_func_t done,
	void *private, zio_priority_t priority, int zio_flags,
	const zbookmark_phys_t *zb)
	{
	arc_buf_hdr_t *hdr = buf->b_hdr;
	arc_write_callback_t *callback;
	zio_t *zio;
	zio_prop_t localprop = *zp;

	ASSERT3P(ready, !=, NULL);
	ASSERT3P(done, !=, NULL);
	ASSERT(!HDR_IO_ERROR(hdr));
	ASSERT(!HDR_IO_IN_PROGRESS(hdr));
	ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL);
	ASSERT3P(hdr->b_l1hdr.b_buf, !=, NULL);
	if (uncached)
	arc_hdr_set_flags(hdr, ARC_FLAG_UNCACHED);
	else if (l2arc)
	arc_hdr_set_flags(hdr, ARC_FLAG_L2CACHE);

	if (ARC_BUF_ENCRYPTED(buf)) {
	ASSERT(ARC_BUF_COMPRESSED(buf));
	localprop.zp_encrypt = B_TRUE;
	localprop.zp_compress = HDR_GET_COMPRESS(hdr);
	localprop.zp_complevel = hdr->b_complevel;
	localprop.zp_byteorder =
	(hdr->b_l1hdr.b_byteswap == DMU_BSWAP_NUMFUNCS) ?
	ZFS_HOST_BYTEORDER : !ZFS_HOST_BYTEORDER;
	memcpy(localprop.zp_salt, hdr->b_crypt_hdr.b_salt,
	ZIO_DATA_SALT_LEN);
	memcpy(localprop.zp_iv, hdr->b_crypt_hdr.b_iv,
	ZIO_DATA_IV_LEN);
	memcpy(localprop.zp_mac, hdr->b_crypt_hdr.b_mac,
	ZIO_DATA_MAC_LEN);
	if (DMU_OT_IS_ENCRYPTED(localprop.zp_type)) {
	localprop.zp_nopwrite = B_FALSE;
	localprop.zp_copies =
	MIN(localprop.zp_copies, SPA_DVAS_PER_BP - 1);
	}
	zio_flags \|= ZIO_FLAG_RAW;
	} else if (ARC_BUF_COMPRESSED(buf)) {
	ASSERT3U(HDR_GET_LSIZE(hdr), !=, arc_buf_size(buf));
	localprop.zp_compress = HDR_GET_COMPRESS(hdr);
	localprop.zp_complevel = hdr->b_complevel;
	zio_flags \|= ZIO_FLAG_RAW_COMPRESS;
	}
	callback = kmem_zalloc(sizeof (arc_write_callback_t), KM_SLEEP);
	callback->awcb_ready = ready;
	callback->awcb_children_ready = children_ready;
	callback->awcb_done = done;
	callback->awcb_private = private;
	callback->awcb_buf = buf;

	/*
	* The hdr's b_pabd is now stale, free it now. A new data block
	* will be allocated when the zio pipeline calls arc_write_ready().
	*/
	if (hdr->b_l1hdr.b_pabd != NULL) {
	/*
	* If the buf is currently sharing the data block with
	* the hdr then we need to break that relationship here.
	* The hdr will remain with a NULL data pointer and the
	* buf will take sole ownership of the block.
	*/
	if (ARC_BUF_SHARED(buf)) {
	arc_unshare_buf(hdr, buf);
	} else {
	ASSERT(!arc_buf_is_shared(buf));
	arc_hdr_free_abd(hdr, B_FALSE);
	}
	VERIFY3P(buf->b_data, !=, NULL);
	}

	if (HDR_HAS_RABD(hdr))
	arc_hdr_free_abd(hdr, B_TRUE);

	if (!(zio_flags & ZIO_FLAG_RAW))
	arc_hdr_set_compress(hdr, ZIO_COMPRESS_OFF);

	ASSERT(!arc_buf_is_shared(buf));
	ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL);

	zio = zio_write(pio, spa, txg, bp,
	abd_get_from_buf(buf->b_data, HDR_GET_LSIZE(hdr)),
	HDR_GET_LSIZE(hdr), arc_buf_size(buf), &localprop, arc_write_ready,
	(children_ready != NULL) ? arc_write_children_ready : NULL,
	arc_write_done, callback, priority, zio_flags, zb);

	return (zio);
	}

	void
	arc_tempreserve_clear(uint64_t reserve)
	{
	atomic_add_64(&arc_tempreserve, -reserve);
	ASSERT((int64_t)arc_tempreserve >= 0);
	}

	int
	arc_tempreserve_space(spa_t *spa, uint64_t reserve, uint64_t txg)
	{
	int error;
	uint64_t anon_size;

	if (!arc_no_grow &&
	reserve > arc_c/4 &&
	reserve * 4 > (2ULL << SPA_MAXBLOCKSHIFT))
	arc_c = MIN(arc_c_max, reserve * 4);

	/*
	* Throttle when the calculated memory footprint for the TXG
	* exceeds the target ARC size.
	*/
	if (reserve > arc_c) {
	DMU_TX_STAT_BUMP(dmu_tx_memory_reserve);
	return (SET_ERROR(ERESTART));
	}

	/*
	* Don't count loaned bufs as in flight dirty data to prevent long
	* network delays from blocking transactions that are ready to be
	* assigned to a txg.
	*/

	/* assert that it has not wrapped around */
	ASSERT3S(atomic_add_64_nv(&arc_loaned_bytes, 0), >=, 0);

	anon_size = MAX((int64_t)
	(zfs_refcount_count(&arc_anon->arcs_size[ARC_BUFC_DATA]) +
	zfs_refcount_count(&arc_anon->arcs_size[ARC_BUFC_METADATA]) -
	arc_loaned_bytes), 0);

	/*
	* Writes will, almost always, require additional memory allocations
	* in order to compress/encrypt/etc the data. We therefore need to
	* make sure that there is sufficient available memory for this.
	*/
	error = arc_memory_throttle(spa, reserve, txg);
	if (error != 0)
	return (error);

	/*
	* Throttle writes when the amount of dirty data in the cache
	* gets too large. We try to keep the cache less than half full
	* of dirty blocks so that our sync times don't grow too large.
	*
	* In the case of one pool being built on another pool, we want
	* to make sure we don't end up throttling the lower (backing)
	* pool when the upper pool is the majority contributor to dirty
	* data. To insure we make forward progress during throttling, we
	* also check the current pool's net dirty data and only throttle
	* if it exceeds zfs_arc_pool_dirty_percent of the anonymous dirty
	* data in the cache.
	*
	* Note: if two requests come in concurrently, we might let them
	* both succeed, when one of them should fail. Not a huge deal.
	*/
	uint64_t total_dirty = reserve + arc_tempreserve + anon_size;
	uint64_t spa_dirty_anon = spa_dirty_data(spa);
	uint64_t rarc_c = arc_warm ? arc_c : arc_c_max;
	if (total_dirty > rarc_c * zfs_arc_dirty_limit_percent / 100 &&
	anon_size > rarc_c * zfs_arc_anon_limit_percent / 100 &&
	spa_dirty_anon > anon_size * zfs_arc_pool_dirty_percent / 100) {
	#ifdef ZFS_DEBUG
	uint64_t meta_esize = zfs_refcount_count(
	&arc_anon->arcs_esize[ARC_BUFC_METADATA]);
	uint64_t data_esize =
	zfs_refcount_count(&arc_anon->arcs_esize[ARC_BUFC_DATA]);
	dprintf("failing, arc_tempreserve=%lluK anon_meta=%lluK "
	"anon_data=%lluK tempreserve=%lluK rarc_c=%lluK\n",
	(u_longlong_t)arc_tempreserve >> 10,
	(u_longlong_t)meta_esize >> 10,
	(u_longlong_t)data_esize >> 10,
	(u_longlong_t)reserve >> 10,
	(u_longlong_t)rarc_c >> 10);
	#endif
	DMU_TX_STAT_BUMP(dmu_tx_dirty_throttle);
	return (SET_ERROR(ERESTART));
	}
	atomic_add_64(&arc_tempreserve, reserve);
	return (0);
	}

	static void
	arc_kstat_update_state(arc_state_t state, kstat_named_t size,
	kstat_named_t data, kstat_named_t metadata,
	kstat_named_t evict_data, kstat_named_t evict_metadata)
	{
	data->value.ui64 =
	zfs_refcount_count(&state->arcs_size[ARC_BUFC_DATA]);
	metadata->value.ui64 =
	zfs_refcount_count(&state->arcs_size[ARC_BUFC_METADATA]);
	size->value.ui64 = data->value.ui64 + metadata->value.ui64;
	evict_data->value.ui64 =
	zfs_refcount_count(&state->arcs_esize[ARC_BUFC_DATA]);
	evict_metadata->value.ui64 =
	zfs_refcount_count(&state->arcs_esize[ARC_BUFC_METADATA]);
	}

	static int
	arc_kstat_update(kstat_t *ksp, int rw)
	{
	arc_stats_t *as = ksp->ks_data;

	if (rw == KSTAT_WRITE)
	return (SET_ERROR(EACCES));

	as->arcstat_hits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_hits);
	as->arcstat_iohits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_iohits);
	as->arcstat_misses.value.ui64 =
	wmsum_value(&arc_sums.arcstat_misses);
	as->arcstat_demand_data_hits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_demand_data_hits);
	as->arcstat_demand_data_iohits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_demand_data_iohits);
	as->arcstat_demand_data_misses.value.ui64 =
	wmsum_value(&arc_sums.arcstat_demand_data_misses);
	as->arcstat_demand_metadata_hits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_demand_metadata_hits);
	as->arcstat_demand_metadata_iohits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_demand_metadata_iohits);
	as->arcstat_demand_metadata_misses.value.ui64 =
	wmsum_value(&arc_sums.arcstat_demand_metadata_misses);
	as->arcstat_prefetch_data_hits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_prefetch_data_hits);
	as->arcstat_prefetch_data_iohits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_prefetch_data_iohits);
	as->arcstat_prefetch_data_misses.value.ui64 =
	wmsum_value(&arc_sums.arcstat_prefetch_data_misses);
	as->arcstat_prefetch_metadata_hits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_prefetch_metadata_hits);
	as->arcstat_prefetch_metadata_iohits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_prefetch_metadata_iohits);
	as->arcstat_prefetch_metadata_misses.value.ui64 =
	wmsum_value(&arc_sums.arcstat_prefetch_metadata_misses);
	as->arcstat_mru_hits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_mru_hits);
	as->arcstat_mru_ghost_hits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_mru_ghost_hits);
	as->arcstat_mfu_hits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_mfu_hits);
	as->arcstat_mfu_ghost_hits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_mfu_ghost_hits);
	as->arcstat_uncached_hits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_uncached_hits);
	as->arcstat_deleted.value.ui64 =
	wmsum_value(&arc_sums.arcstat_deleted);
	as->arcstat_mutex_miss.value.ui64 =
	wmsum_value(&arc_sums.arcstat_mutex_miss);
	as->arcstat_access_skip.value.ui64 =
	wmsum_value(&arc_sums.arcstat_access_skip);
	as->arcstat_evict_skip.value.ui64 =
	wmsum_value(&arc_sums.arcstat_evict_skip);
	as->arcstat_evict_not_enough.value.ui64 =
	wmsum_value(&arc_sums.arcstat_evict_not_enough);
	as->arcstat_evict_l2_cached.value.ui64 =
	wmsum_value(&arc_sums.arcstat_evict_l2_cached);
	as->arcstat_evict_l2_eligible.value.ui64 =
	wmsum_value(&arc_sums.arcstat_evict_l2_eligible);
	as->arcstat_evict_l2_eligible_mfu.value.ui64 =
	wmsum_value(&arc_sums.arcstat_evict_l2_eligible_mfu);
	as->arcstat_evict_l2_eligible_mru.value.ui64 =
	wmsum_value(&arc_sums.arcstat_evict_l2_eligible_mru);
	as->arcstat_evict_l2_ineligible.value.ui64 =
	wmsum_value(&arc_sums.arcstat_evict_l2_ineligible);
	as->arcstat_evict_l2_skip.value.ui64 =
	wmsum_value(&arc_sums.arcstat_evict_l2_skip);
	as->arcstat_hash_collisions.value.ui64 =
	wmsum_value(&arc_sums.arcstat_hash_collisions);
	as->arcstat_hash_chains.value.ui64 =
	wmsum_value(&arc_sums.arcstat_hash_chains);
	as->arcstat_size.value.ui64 =
	aggsum_value(&arc_sums.arcstat_size);
	as->arcstat_compressed_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_compressed_size);
	as->arcstat_uncompressed_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_uncompressed_size);
	as->arcstat_overhead_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_overhead_size);
	as->arcstat_hdr_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_hdr_size);
	as->arcstat_data_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_data_size);
	as->arcstat_metadata_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_metadata_size);
	as->arcstat_dbuf_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_dbuf_size);
	#if defined(COMPAT_FREEBSD11)
	as->arcstat_other_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_bonus_size) +
	wmsum_value(&arc_sums.arcstat_dnode_size) +
	wmsum_value(&arc_sums.arcstat_dbuf_size);
	#endif

	arc_kstat_update_state(arc_anon,
	&as->arcstat_anon_size,
	&as->arcstat_anon_data,
	&as->arcstat_anon_metadata,
	&as->arcstat_anon_evictable_data,
	&as->arcstat_anon_evictable_metadata);
	arc_kstat_update_state(arc_mru,
	&as->arcstat_mru_size,
	&as->arcstat_mru_data,
	&as->arcstat_mru_metadata,
	&as->arcstat_mru_evictable_data,
	&as->arcstat_mru_evictable_metadata);
	arc_kstat_update_state(arc_mru_ghost,
	&as->arcstat_mru_ghost_size,
	&as->arcstat_mru_ghost_data,
	&as->arcstat_mru_ghost_metadata,
	&as->arcstat_mru_ghost_evictable_data,
	&as->arcstat_mru_ghost_evictable_metadata);
	arc_kstat_update_state(arc_mfu,
	&as->arcstat_mfu_size,
	&as->arcstat_mfu_data,
	&as->arcstat_mfu_metadata,
	&as->arcstat_mfu_evictable_data,
	&as->arcstat_mfu_evictable_metadata);
	arc_kstat_update_state(arc_mfu_ghost,
	&as->arcstat_mfu_ghost_size,
	&as->arcstat_mfu_ghost_data,
	&as->arcstat_mfu_ghost_metadata,
	&as->arcstat_mfu_ghost_evictable_data,
	&as->arcstat_mfu_ghost_evictable_metadata);
	arc_kstat_update_state(arc_uncached,
	&as->arcstat_uncached_size,
	&as->arcstat_uncached_data,
	&as->arcstat_uncached_metadata,
	&as->arcstat_uncached_evictable_data,
	&as->arcstat_uncached_evictable_metadata);

	as->arcstat_dnode_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_dnode_size);
	as->arcstat_bonus_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_bonus_size);
	as->arcstat_l2_hits.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_hits);
	as->arcstat_l2_misses.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_misses);
	as->arcstat_l2_prefetch_asize.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_prefetch_asize);
	as->arcstat_l2_mru_asize.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_mru_asize);
	as->arcstat_l2_mfu_asize.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_mfu_asize);
	as->arcstat_l2_bufc_data_asize.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_bufc_data_asize);
	as->arcstat_l2_bufc_metadata_asize.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_bufc_metadata_asize);
	as->arcstat_l2_feeds.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_feeds);
	as->arcstat_l2_rw_clash.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rw_clash);
	as->arcstat_l2_read_bytes.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_read_bytes);
	as->arcstat_l2_write_bytes.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_write_bytes);
	as->arcstat_l2_writes_sent.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_writes_sent);
	as->arcstat_l2_writes_done.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_writes_done);
	as->arcstat_l2_writes_error.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_writes_error);
	as->arcstat_l2_writes_lock_retry.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_writes_lock_retry);
	as->arcstat_l2_evict_lock_retry.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_evict_lock_retry);
	as->arcstat_l2_evict_reading.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_evict_reading);
	as->arcstat_l2_evict_l1cached.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_evict_l1cached);
	as->arcstat_l2_free_on_write.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_free_on_write);
	as->arcstat_l2_abort_lowmem.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_abort_lowmem);
	as->arcstat_l2_cksum_bad.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_cksum_bad);
	as->arcstat_l2_io_error.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_io_error);
	as->arcstat_l2_lsize.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_lsize);
	as->arcstat_l2_psize.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_psize);
	as->arcstat_l2_hdr_size.value.ui64 =
	aggsum_value(&arc_sums.arcstat_l2_hdr_size);
	as->arcstat_l2_log_blk_writes.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_log_blk_writes);
	as->arcstat_l2_log_blk_asize.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_log_blk_asize);
	as->arcstat_l2_log_blk_count.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_log_blk_count);
	as->arcstat_l2_rebuild_success.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rebuild_success);
	as->arcstat_l2_rebuild_abort_unsupported.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rebuild_abort_unsupported);
	as->arcstat_l2_rebuild_abort_io_errors.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rebuild_abort_io_errors);
	as->arcstat_l2_rebuild_abort_dh_errors.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rebuild_abort_dh_errors);
	as->arcstat_l2_rebuild_abort_cksum_lb_errors.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rebuild_abort_cksum_lb_errors);
	as->arcstat_l2_rebuild_abort_lowmem.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rebuild_abort_lowmem);
	as->arcstat_l2_rebuild_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rebuild_size);
	as->arcstat_l2_rebuild_asize.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rebuild_asize);
	as->arcstat_l2_rebuild_bufs.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rebuild_bufs);
	as->arcstat_l2_rebuild_bufs_precached.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rebuild_bufs_precached);
	as->arcstat_l2_rebuild_log_blks.value.ui64 =
	wmsum_value(&arc_sums.arcstat_l2_rebuild_log_blks);
	as->arcstat_memory_throttle_count.value.ui64 =
	wmsum_value(&arc_sums.arcstat_memory_throttle_count);
	as->arcstat_memory_direct_count.value.ui64 =
	wmsum_value(&arc_sums.arcstat_memory_direct_count);
	as->arcstat_memory_indirect_count.value.ui64 =
	wmsum_value(&arc_sums.arcstat_memory_indirect_count);

	as->arcstat_memory_all_bytes.value.ui64 =
	arc_all_memory();
	as->arcstat_memory_free_bytes.value.ui64 =
	arc_free_memory();
	as->arcstat_memory_available_bytes.value.i64 =
	arc_available_memory();

	as->arcstat_prune.value.ui64 =
	wmsum_value(&arc_sums.arcstat_prune);
	as->arcstat_meta_used.value.ui64 =
	wmsum_value(&arc_sums.arcstat_meta_used);
	as->arcstat_async_upgrade_sync.value.ui64 =
	wmsum_value(&arc_sums.arcstat_async_upgrade_sync);
	as->arcstat_predictive_prefetch.value.ui64 =
	wmsum_value(&arc_sums.arcstat_predictive_prefetch);
	as->arcstat_demand_hit_predictive_prefetch.value.ui64 =
	wmsum_value(&arc_sums.arcstat_demand_hit_predictive_prefetch);
	as->arcstat_demand_iohit_predictive_prefetch.value.ui64 =
	wmsum_value(&arc_sums.arcstat_demand_iohit_predictive_prefetch);
	as->arcstat_prescient_prefetch.value.ui64 =
	wmsum_value(&arc_sums.arcstat_prescient_prefetch);
	as->arcstat_demand_hit_prescient_prefetch.value.ui64 =
	wmsum_value(&arc_sums.arcstat_demand_hit_prescient_prefetch);
	as->arcstat_demand_iohit_prescient_prefetch.value.ui64 =
	wmsum_value(&arc_sums.arcstat_demand_iohit_prescient_prefetch);
	as->arcstat_raw_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_raw_size);
	as->arcstat_cached_only_in_progress.value.ui64 =
	wmsum_value(&arc_sums.arcstat_cached_only_in_progress);
	as->arcstat_abd_chunk_waste_size.value.ui64 =
	wmsum_value(&arc_sums.arcstat_abd_chunk_waste_size);

	return (0);
	}

	/*
	* This function must return indices evenly distributed between all
	* sublists of the multilist. This is needed due to how the ARC eviction
	* code is laid out; arc_evict_state() assumes ARC buffers are evenly
	* distributed between all sublists and uses this assumption when
	* deciding which sublist to evict from and how much to evict from it.
	*/
	static unsigned int
	arc_state_multilist_index_func(multilist_t ml, void obj)
	{
	arc_buf_hdr_t *hdr = obj;

	/*
	* We rely on b_dva to generate evenly distributed index
	* numbers using buf_hash below. So, as an added precaution,
	* let's make sure we never add empty buffers to the arc lists.
	*/
	ASSERT(!HDR_EMPTY(hdr));

	/*
	* The assumption here, is the hash value for a given
	* arc_buf_hdr_t will remain constant throughout its lifetime
	* (i.e. its b_spa, b_dva, and b_birth fields don't change).
	* Thus, we don't need to store the header's sublist index
	* on insertion, as this index can be recalculated on removal.
	*
	* Also, the low order bits of the hash value are thought to be
	* distributed evenly. Otherwise, in the case that the multilist
	* has a power of two number of sublists, each sublists' usage
	* would not be evenly distributed. In this context full 64bit
	* division would be a waste of time, so limit it to 32 bits.
	*/
	return ((unsigned int)buf_hash(hdr->b_spa, &hdr->b_dva, hdr->b_birth) %
	multilist_get_num_sublists(ml));
	}

	static unsigned int
	arc_state_l2c_multilist_index_func(multilist_t ml, void obj)
	{
	panic("Header %p insert into arc_l2c_only %p", obj, ml);
	}

	#define WARN_IF_TUNING_IGNORED(tuning, value, do_warn) do { \
	if ((do_warn) && (tuning) && ((tuning) != (value))) { \
	cmn_err(CE_WARN, \
	"ignoring tunable %s (using %llu instead)", \
	(#tuning), (u_longlong_t)(value)); \
	} \
	} while (0)

	/*
	* Called during module initialization and periodically thereafter to
	* apply reasonable changes to the exposed performance tunings. Can also be
	* called explicitly by param_set_arc_*() functions when ARC tunables are
	* updated manually. Non-zero zfs_* values which differ from the currently set
	* values will be applied.
	*/
	void
	arc_tuning_update(boolean_t verbose)
	{
	uint64_t allmem = arc_all_memory();

	/* Valid range: 32M - <arc_c_max> */
	if ((zfs_arc_min) && (zfs_arc_min != arc_c_min) &&
	(zfs_arc_min >= 2ULL << SPA_MAXBLOCKSHIFT) &&
	(zfs_arc_min <= arc_c_max)) {
	arc_c_min = zfs_arc_min;
	arc_c = MAX(arc_c, arc_c_min);
	}
	WARN_IF_TUNING_IGNORED(zfs_arc_min, arc_c_min, verbose);

	/* Valid range: 64M - <all physical memory> */
	if ((zfs_arc_max) && (zfs_arc_max != arc_c_max) &&
	(zfs_arc_max >= MIN_ARC_MAX) && (zfs_arc_max < allmem) &&
	(zfs_arc_max > arc_c_min)) {
	arc_c_max = zfs_arc_max;
	arc_c = MIN(arc_c, arc_c_max);
	if (arc_dnode_limit > arc_c_max)
	arc_dnode_limit = arc_c_max;
	}
	WARN_IF_TUNING_IGNORED(zfs_arc_max, arc_c_max, verbose);

	/* Valid range: 0 - <all physical memory> */
	arc_dnode_limit = zfs_arc_dnode_limit ? zfs_arc_dnode_limit :
	MIN(zfs_arc_dnode_limit_percent, 100) * arc_c_max / 100;
	WARN_IF_TUNING_IGNORED(zfs_arc_dnode_limit, arc_dnode_limit, verbose);

	/* Valid range: 1 - N */
	if (zfs_arc_grow_retry)
	arc_grow_retry = zfs_arc_grow_retry;

	/* Valid range: 1 - N */
	if (zfs_arc_shrink_shift) {
	arc_shrink_shift = zfs_arc_shrink_shift;
	arc_no_grow_shift = MIN(arc_no_grow_shift, arc_shrink_shift -1);
	}

	/* Valid range: 1 - N ms */
	if (zfs_arc_min_prefetch_ms)
	arc_min_prefetch_ms = zfs_arc_min_prefetch_ms;

	/* Valid range: 1 - N ms */
	if (zfs_arc_min_prescient_prefetch_ms) {
	arc_min_prescient_prefetch_ms =
	zfs_arc_min_prescient_prefetch_ms;
	}

	/* Valid range: 0 - 100 */
	if (zfs_arc_lotsfree_percent <= 100)
	arc_lotsfree_percent = zfs_arc_lotsfree_percent;
	WARN_IF_TUNING_IGNORED(zfs_arc_lotsfree_percent, arc_lotsfree_percent,
	verbose);

	/* Valid range: 0 - <all physical memory> */
	if ((zfs_arc_sys_free) && (zfs_arc_sys_free != arc_sys_free))
	arc_sys_free = MIN(zfs_arc_sys_free, allmem);
	WARN_IF_TUNING_IGNORED(zfs_arc_sys_free, arc_sys_free, verbose);
	}

	static void
	arc_state_multilist_init(multilist_t *ml,
	multilist_sublist_index_func_t index_func, int maxcountp)
	{
	multilist_create(ml, sizeof (arc_buf_hdr_t),
	offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node), index_func);
	maxcountp = MAX(maxcountp, multilist_get_num_sublists(ml));
	}

	static void
	arc_state_init(void)
	{
	int num_sublists = 0;

	arc_state_multilist_init(&arc_mru->arcs_list[ARC_BUFC_METADATA],
	arc_state_multilist_index_func, &num_sublists);
	arc_state_multilist_init(&arc_mru->arcs_list[ARC_BUFC_DATA],
	arc_state_multilist_index_func, &num_sublists);
	arc_state_multilist_init(&arc_mru_ghost->arcs_list[ARC_BUFC_METADATA],
	arc_state_multilist_index_func, &num_sublists);
	arc_state_multilist_init(&arc_mru_ghost->arcs_list[ARC_BUFC_DATA],
	arc_state_multilist_index_func, &num_sublists);
	arc_state_multilist_init(&arc_mfu->arcs_list[ARC_BUFC_METADATA],
	arc_state_multilist_index_func, &num_sublists);
	arc_state_multilist_init(&arc_mfu->arcs_list[ARC_BUFC_DATA],
	arc_state_multilist_index_func, &num_sublists);
	arc_state_multilist_init(&arc_mfu_ghost->arcs_list[ARC_BUFC_METADATA],
	arc_state_multilist_index_func, &num_sublists);
	arc_state_multilist_init(&arc_mfu_ghost->arcs_list[ARC_BUFC_DATA],
	arc_state_multilist_index_func, &num_sublists);
	arc_state_multilist_init(&arc_uncached->arcs_list[ARC_BUFC_METADATA],
	arc_state_multilist_index_func, &num_sublists);
	arc_state_multilist_init(&arc_uncached->arcs_list[ARC_BUFC_DATA],
	arc_state_multilist_index_func, &num_sublists);

	/*
	* L2 headers should never be on the L2 state list since they don't
	* have L1 headers allocated. Special index function asserts that.
	*/
	arc_state_multilist_init(&arc_l2c_only->arcs_list[ARC_BUFC_METADATA],
	arc_state_l2c_multilist_index_func, &num_sublists);
	arc_state_multilist_init(&arc_l2c_only->arcs_list[ARC_BUFC_DATA],
	arc_state_l2c_multilist_index_func, &num_sublists);

	/*
	* Keep track of the number of markers needed to reclaim buffers from
	* any ARC state. The markers will be pre-allocated so as to minimize
	* the number of memory allocations performed by the eviction thread.
	*/
	arc_state_evict_marker_count = num_sublists;

	zfs_refcount_create(&arc_anon->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_create(&arc_anon->arcs_esize[ARC_BUFC_DATA]);
	zfs_refcount_create(&arc_mru->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_create(&arc_mru->arcs_esize[ARC_BUFC_DATA]);
	zfs_refcount_create(&arc_mru_ghost->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_create(&arc_mru_ghost->arcs_esize[ARC_BUFC_DATA]);
	zfs_refcount_create(&arc_mfu->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_create(&arc_mfu->arcs_esize[ARC_BUFC_DATA]);
	zfs_refcount_create(&arc_mfu_ghost->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_create(&arc_mfu_ghost->arcs_esize[ARC_BUFC_DATA]);
	zfs_refcount_create(&arc_l2c_only->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_create(&arc_l2c_only->arcs_esize[ARC_BUFC_DATA]);
	zfs_refcount_create(&arc_uncached->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_create(&arc_uncached->arcs_esize[ARC_BUFC_DATA]);

	zfs_refcount_create(&arc_anon->arcs_size[ARC_BUFC_DATA]);
	zfs_refcount_create(&arc_anon->arcs_size[ARC_BUFC_METADATA]);
	zfs_refcount_create(&arc_mru->arcs_size[ARC_BUFC_DATA]);
	zfs_refcount_create(&arc_mru->arcs_size[ARC_BUFC_METADATA]);
	zfs_refcount_create(&arc_mru_ghost->arcs_size[ARC_BUFC_DATA]);
	zfs_refcount_create(&arc_mru_ghost->arcs_size[ARC_BUFC_METADATA]);
	zfs_refcount_create(&arc_mfu->arcs_size[ARC_BUFC_DATA]);
	zfs_refcount_create(&arc_mfu->arcs_size[ARC_BUFC_METADATA]);
	zfs_refcount_create(&arc_mfu_ghost->arcs_size[ARC_BUFC_DATA]);
	zfs_refcount_create(&arc_mfu_ghost->arcs_size[ARC_BUFC_METADATA]);
	zfs_refcount_create(&arc_l2c_only->arcs_size[ARC_BUFC_DATA]);
	zfs_refcount_create(&arc_l2c_only->arcs_size[ARC_BUFC_METADATA]);
	zfs_refcount_create(&arc_uncached->arcs_size[ARC_BUFC_DATA]);
	zfs_refcount_create(&arc_uncached->arcs_size[ARC_BUFC_METADATA]);

	wmsum_init(&arc_mru_ghost->arcs_hits[ARC_BUFC_DATA], 0);
	wmsum_init(&arc_mru_ghost->arcs_hits[ARC_BUFC_METADATA], 0);
	wmsum_init(&arc_mfu_ghost->arcs_hits[ARC_BUFC_DATA], 0);
	wmsum_init(&arc_mfu_ghost->arcs_hits[ARC_BUFC_METADATA], 0);

	wmsum_init(&arc_sums.arcstat_hits, 0);
	wmsum_init(&arc_sums.arcstat_iohits, 0);
	wmsum_init(&arc_sums.arcstat_misses, 0);
	wmsum_init(&arc_sums.arcstat_demand_data_hits, 0);
	wmsum_init(&arc_sums.arcstat_demand_data_iohits, 0);
	wmsum_init(&arc_sums.arcstat_demand_data_misses, 0);
	wmsum_init(&arc_sums.arcstat_demand_metadata_hits, 0);
	wmsum_init(&arc_sums.arcstat_demand_metadata_iohits, 0);
	wmsum_init(&arc_sums.arcstat_demand_metadata_misses, 0);
	wmsum_init(&arc_sums.arcstat_prefetch_data_hits, 0);
	wmsum_init(&arc_sums.arcstat_prefetch_data_iohits, 0);
	wmsum_init(&arc_sums.arcstat_prefetch_data_misses, 0);
	wmsum_init(&arc_sums.arcstat_prefetch_metadata_hits, 0);
	wmsum_init(&arc_sums.arcstat_prefetch_metadata_iohits, 0);
	wmsum_init(&arc_sums.arcstat_prefetch_metadata_misses, 0);
	wmsum_init(&arc_sums.arcstat_mru_hits, 0);
	wmsum_init(&arc_sums.arcstat_mru_ghost_hits, 0);
	wmsum_init(&arc_sums.arcstat_mfu_hits, 0);
	wmsum_init(&arc_sums.arcstat_mfu_ghost_hits, 0);
	wmsum_init(&arc_sums.arcstat_uncached_hits, 0);
	wmsum_init(&arc_sums.arcstat_deleted, 0);
	wmsum_init(&arc_sums.arcstat_mutex_miss, 0);
	wmsum_init(&arc_sums.arcstat_access_skip, 0);
	wmsum_init(&arc_sums.arcstat_evict_skip, 0);
	wmsum_init(&arc_sums.arcstat_evict_not_enough, 0);
	wmsum_init(&arc_sums.arcstat_evict_l2_cached, 0);
	wmsum_init(&arc_sums.arcstat_evict_l2_eligible, 0);
	wmsum_init(&arc_sums.arcstat_evict_l2_eligible_mfu, 0);
	wmsum_init(&arc_sums.arcstat_evict_l2_eligible_mru, 0);
	wmsum_init(&arc_sums.arcstat_evict_l2_ineligible, 0);
	wmsum_init(&arc_sums.arcstat_evict_l2_skip, 0);
	wmsum_init(&arc_sums.arcstat_hash_collisions, 0);
	wmsum_init(&arc_sums.arcstat_hash_chains, 0);
	aggsum_init(&arc_sums.arcstat_size, 0);
	wmsum_init(&arc_sums.arcstat_compressed_size, 0);
	wmsum_init(&arc_sums.arcstat_uncompressed_size, 0);
	wmsum_init(&arc_sums.arcstat_overhead_size, 0);
	wmsum_init(&arc_sums.arcstat_hdr_size, 0);
	wmsum_init(&arc_sums.arcstat_data_size, 0);
	wmsum_init(&arc_sums.arcstat_metadata_size, 0);
	wmsum_init(&arc_sums.arcstat_dbuf_size, 0);
	wmsum_init(&arc_sums.arcstat_dnode_size, 0);
	wmsum_init(&arc_sums.arcstat_bonus_size, 0);
	wmsum_init(&arc_sums.arcstat_l2_hits, 0);
	wmsum_init(&arc_sums.arcstat_l2_misses, 0);
	wmsum_init(&arc_sums.arcstat_l2_prefetch_asize, 0);
	wmsum_init(&arc_sums.arcstat_l2_mru_asize, 0);
	wmsum_init(&arc_sums.arcstat_l2_mfu_asize, 0);
	wmsum_init(&arc_sums.arcstat_l2_bufc_data_asize, 0);
	wmsum_init(&arc_sums.arcstat_l2_bufc_metadata_asize, 0);
	wmsum_init(&arc_sums.arcstat_l2_feeds, 0);
	wmsum_init(&arc_sums.arcstat_l2_rw_clash, 0);
	wmsum_init(&arc_sums.arcstat_l2_read_bytes, 0);
	wmsum_init(&arc_sums.arcstat_l2_write_bytes, 0);
	wmsum_init(&arc_sums.arcstat_l2_writes_sent, 0);
	wmsum_init(&arc_sums.arcstat_l2_writes_done, 0);
	wmsum_init(&arc_sums.arcstat_l2_writes_error, 0);
	wmsum_init(&arc_sums.arcstat_l2_writes_lock_retry, 0);
	wmsum_init(&arc_sums.arcstat_l2_evict_lock_retry, 0);
	wmsum_init(&arc_sums.arcstat_l2_evict_reading, 0);
	wmsum_init(&arc_sums.arcstat_l2_evict_l1cached, 0);
	wmsum_init(&arc_sums.arcstat_l2_free_on_write, 0);
	wmsum_init(&arc_sums.arcstat_l2_abort_lowmem, 0);
	wmsum_init(&arc_sums.arcstat_l2_cksum_bad, 0);
	wmsum_init(&arc_sums.arcstat_l2_io_error, 0);
	wmsum_init(&arc_sums.arcstat_l2_lsize, 0);
	wmsum_init(&arc_sums.arcstat_l2_psize, 0);
	aggsum_init(&arc_sums.arcstat_l2_hdr_size, 0);
	wmsum_init(&arc_sums.arcstat_l2_log_blk_writes, 0);
	wmsum_init(&arc_sums.arcstat_l2_log_blk_asize, 0);
	wmsum_init(&arc_sums.arcstat_l2_log_blk_count, 0);
	wmsum_init(&arc_sums.arcstat_l2_rebuild_success, 0);
	wmsum_init(&arc_sums.arcstat_l2_rebuild_abort_unsupported, 0);
	wmsum_init(&arc_sums.arcstat_l2_rebuild_abort_io_errors, 0);
	wmsum_init(&arc_sums.arcstat_l2_rebuild_abort_dh_errors, 0);
	wmsum_init(&arc_sums.arcstat_l2_rebuild_abort_cksum_lb_errors, 0);
	wmsum_init(&arc_sums.arcstat_l2_rebuild_abort_lowmem, 0);
	wmsum_init(&arc_sums.arcstat_l2_rebuild_size, 0);
	wmsum_init(&arc_sums.arcstat_l2_rebuild_asize, 0);
	wmsum_init(&arc_sums.arcstat_l2_rebuild_bufs, 0);
	wmsum_init(&arc_sums.arcstat_l2_rebuild_bufs_precached, 0);
	wmsum_init(&arc_sums.arcstat_l2_rebuild_log_blks, 0);
	wmsum_init(&arc_sums.arcstat_memory_throttle_count, 0);
	wmsum_init(&arc_sums.arcstat_memory_direct_count, 0);
	wmsum_init(&arc_sums.arcstat_memory_indirect_count, 0);
	wmsum_init(&arc_sums.arcstat_prune, 0);
	wmsum_init(&arc_sums.arcstat_meta_used, 0);
	wmsum_init(&arc_sums.arcstat_async_upgrade_sync, 0);
	wmsum_init(&arc_sums.arcstat_predictive_prefetch, 0);
	wmsum_init(&arc_sums.arcstat_demand_hit_predictive_prefetch, 0);
	wmsum_init(&arc_sums.arcstat_demand_iohit_predictive_prefetch, 0);
	wmsum_init(&arc_sums.arcstat_prescient_prefetch, 0);
	wmsum_init(&arc_sums.arcstat_demand_hit_prescient_prefetch, 0);
	wmsum_init(&arc_sums.arcstat_demand_iohit_prescient_prefetch, 0);
	wmsum_init(&arc_sums.arcstat_raw_size, 0);
	wmsum_init(&arc_sums.arcstat_cached_only_in_progress, 0);
	wmsum_init(&arc_sums.arcstat_abd_chunk_waste_size, 0);

	arc_anon->arcs_state = ARC_STATE_ANON;
	arc_mru->arcs_state = ARC_STATE_MRU;
	arc_mru_ghost->arcs_state = ARC_STATE_MRU_GHOST;
	arc_mfu->arcs_state = ARC_STATE_MFU;
	arc_mfu_ghost->arcs_state = ARC_STATE_MFU_GHOST;
	arc_l2c_only->arcs_state = ARC_STATE_L2C_ONLY;
	arc_uncached->arcs_state = ARC_STATE_UNCACHED;
	}

	static void
	arc_state_fini(void)
	{
	zfs_refcount_destroy(&arc_anon->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_destroy(&arc_anon->arcs_esize[ARC_BUFC_DATA]);
	zfs_refcount_destroy(&arc_mru->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_destroy(&arc_mru->arcs_esize[ARC_BUFC_DATA]);
	zfs_refcount_destroy(&arc_mru_ghost->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_destroy(&arc_mru_ghost->arcs_esize[ARC_BUFC_DATA]);
	zfs_refcount_destroy(&arc_mfu->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_destroy(&arc_mfu->arcs_esize[ARC_BUFC_DATA]);
	zfs_refcount_destroy(&arc_mfu_ghost->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_destroy(&arc_mfu_ghost->arcs_esize[ARC_BUFC_DATA]);
	zfs_refcount_destroy(&arc_l2c_only->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_destroy(&arc_l2c_only->arcs_esize[ARC_BUFC_DATA]);
	zfs_refcount_destroy(&arc_uncached->arcs_esize[ARC_BUFC_METADATA]);
	zfs_refcount_destroy(&arc_uncached->arcs_esize[ARC_BUFC_DATA]);

	zfs_refcount_destroy(&arc_anon->arcs_size[ARC_BUFC_DATA]);
	zfs_refcount_destroy(&arc_anon->arcs_size[ARC_BUFC_METADATA]);
	zfs_refcount_destroy(&arc_mru->arcs_size[ARC_BUFC_DATA]);
	zfs_refcount_destroy(&arc_mru->arcs_size[ARC_BUFC_METADATA]);
	zfs_refcount_destroy(&arc_mru_ghost->arcs_size[ARC_BUFC_DATA]);
	zfs_refcount_destroy(&arc_mru_ghost->arcs_size[ARC_BUFC_METADATA]);
	zfs_refcount_destroy(&arc_mfu->arcs_size[ARC_BUFC_DATA]);
	zfs_refcount_destroy(&arc_mfu->arcs_size[ARC_BUFC_METADATA]);
	zfs_refcount_destroy(&arc_mfu_ghost->arcs_size[ARC_BUFC_DATA]);
	zfs_refcount_destroy(&arc_mfu_ghost->arcs_size[ARC_BUFC_METADATA]);
	zfs_refcount_destroy(&arc_l2c_only->arcs_size[ARC_BUFC_DATA]);
	zfs_refcount_destroy(&arc_l2c_only->arcs_size[ARC_BUFC_METADATA]);
	zfs_refcount_destroy(&arc_uncached->arcs_size[ARC_BUFC_DATA]);
	zfs_refcount_destroy(&arc_uncached->arcs_size[ARC_BUFC_METADATA]);

	multilist_destroy(&arc_mru->arcs_list[ARC_BUFC_METADATA]);
	multilist_destroy(&arc_mru_ghost->arcs_list[ARC_BUFC_METADATA]);
	multilist_destroy(&arc_mfu->arcs_list[ARC_BUFC_METADATA]);
	multilist_destroy(&arc_mfu_ghost->arcs_list[ARC_BUFC_METADATA]);
	multilist_destroy(&arc_mru->arcs_list[ARC_BUFC_DATA]);
	multilist_destroy(&arc_mru_ghost->arcs_list[ARC_BUFC_DATA]);
	multilist_destroy(&arc_mfu->arcs_list[ARC_BUFC_DATA]);
	multilist_destroy(&arc_mfu_ghost->arcs_list[ARC_BUFC_DATA]);
	multilist_destroy(&arc_l2c_only->arcs_list[ARC_BUFC_METADATA]);
	multilist_destroy(&arc_l2c_only->arcs_list[ARC_BUFC_DATA]);
	multilist_destroy(&arc_uncached->arcs_list[ARC_BUFC_METADATA]);
	multilist_destroy(&arc_uncached->arcs_list[ARC_BUFC_DATA]);

	wmsum_fini(&arc_mru_ghost->arcs_hits[ARC_BUFC_DATA]);
	wmsum_fini(&arc_mru_ghost->arcs_hits[ARC_BUFC_METADATA]);
	wmsum_fini(&arc_mfu_ghost->arcs_hits[ARC_BUFC_DATA]);
	wmsum_fini(&arc_mfu_ghost->arcs_hits[ARC_BUFC_METADATA]);

	wmsum_fini(&arc_sums.arcstat_hits);
	wmsum_fini(&arc_sums.arcstat_iohits);
	wmsum_fini(&arc_sums.arcstat_misses);
	wmsum_fini(&arc_sums.arcstat_demand_data_hits);
	wmsum_fini(&arc_sums.arcstat_demand_data_iohits);
	wmsum_fini(&arc_sums.arcstat_demand_data_misses);
	wmsum_fini(&arc_sums.arcstat_demand_metadata_hits);
	wmsum_fini(&arc_sums.arcstat_demand_metadata_iohits);
	wmsum_fini(&arc_sums.arcstat_demand_metadata_misses);
	wmsum_fini(&arc_sums.arcstat_prefetch_data_hits);
	wmsum_fini(&arc_sums.arcstat_prefetch_data_iohits);
	wmsum_fini(&arc_sums.arcstat_prefetch_data_misses);
	wmsum_fini(&arc_sums.arcstat_prefetch_metadata_hits);
	wmsum_fini(&arc_sums.arcstat_prefetch_metadata_iohits);
	wmsum_fini(&arc_sums.arcstat_prefetch_metadata_misses);
	wmsum_fini(&arc_sums.arcstat_mru_hits);
	wmsum_fini(&arc_sums.arcstat_mru_ghost_hits);
	wmsum_fini(&arc_sums.arcstat_mfu_hits);
	wmsum_fini(&arc_sums.arcstat_mfu_ghost_hits);
	wmsum_fini(&arc_sums.arcstat_uncached_hits);
	wmsum_fini(&arc_sums.arcstat_deleted);
	wmsum_fini(&arc_sums.arcstat_mutex_miss);
	wmsum_fini(&arc_sums.arcstat_access_skip);
	wmsum_fini(&arc_sums.arcstat_evict_skip);
	wmsum_fini(&arc_sums.arcstat_evict_not_enough);
	wmsum_fini(&arc_sums.arcstat_evict_l2_cached);
	wmsum_fini(&arc_sums.arcstat_evict_l2_eligible);
	wmsum_fini(&arc_sums.arcstat_evict_l2_eligible_mfu);
	wmsum_fini(&arc_sums.arcstat_evict_l2_eligible_mru);
	wmsum_fini(&arc_sums.arcstat_evict_l2_ineligible);
	wmsum_fini(&arc_sums.arcstat_evict_l2_skip);
	wmsum_fini(&arc_sums.arcstat_hash_collisions);
	wmsum_fini(&arc_sums.arcstat_hash_chains);
	aggsum_fini(&arc_sums.arcstat_size);
	wmsum_fini(&arc_sums.arcstat_compressed_size);
	wmsum_fini(&arc_sums.arcstat_uncompressed_size);
	wmsum_fini(&arc_sums.arcstat_overhead_size);
	wmsum_fini(&arc_sums.arcstat_hdr_size);
	wmsum_fini(&arc_sums.arcstat_data_size);
	wmsum_fini(&arc_sums.arcstat_metadata_size);
	wmsum_fini(&arc_sums.arcstat_dbuf_size);
	wmsum_fini(&arc_sums.arcstat_dnode_size);
	wmsum_fini(&arc_sums.arcstat_bonus_size);
	wmsum_fini(&arc_sums.arcstat_l2_hits);
	wmsum_fini(&arc_sums.arcstat_l2_misses);
	wmsum_fini(&arc_sums.arcstat_l2_prefetch_asize);
	wmsum_fini(&arc_sums.arcstat_l2_mru_asize);
	wmsum_fini(&arc_sums.arcstat_l2_mfu_asize);
	wmsum_fini(&arc_sums.arcstat_l2_bufc_data_asize);
	wmsum_fini(&arc_sums.arcstat_l2_bufc_metadata_asize);
	wmsum_fini(&arc_sums.arcstat_l2_feeds);
	wmsum_fini(&arc_sums.arcstat_l2_rw_clash);
	wmsum_fini(&arc_sums.arcstat_l2_read_bytes);
	wmsum_fini(&arc_sums.arcstat_l2_write_bytes);
	wmsum_fini(&arc_sums.arcstat_l2_writes_sent);
	wmsum_fini(&arc_sums.arcstat_l2_writes_done);
	wmsum_fini(&arc_sums.arcstat_l2_writes_error);
	wmsum_fini(&arc_sums.arcstat_l2_writes_lock_retry);
	wmsum_fini(&arc_sums.arcstat_l2_evict_lock_retry);
	wmsum_fini(&arc_sums.arcstat_l2_evict_reading);
	wmsum_fini(&arc_sums.arcstat_l2_evict_l1cached);
	wmsum_fini(&arc_sums.arcstat_l2_free_on_write);
	wmsum_fini(&arc_sums.arcstat_l2_abort_lowmem);
	wmsum_fini(&arc_sums.arcstat_l2_cksum_bad);
	wmsum_fini(&arc_sums.arcstat_l2_io_error);
	wmsum_fini(&arc_sums.arcstat_l2_lsize);
	wmsum_fini(&arc_sums.arcstat_l2_psize);
	aggsum_fini(&arc_sums.arcstat_l2_hdr_size);
	wmsum_fini(&arc_sums.arcstat_l2_log_blk_writes);
	wmsum_fini(&arc_sums.arcstat_l2_log_blk_asize);
	wmsum_fini(&arc_sums.arcstat_l2_log_blk_count);
	wmsum_fini(&arc_sums.arcstat_l2_rebuild_success);
	wmsum_fini(&arc_sums.arcstat_l2_rebuild_abort_unsupported);
	wmsum_fini(&arc_sums.arcstat_l2_rebuild_abort_io_errors);
	wmsum_fini(&arc_sums.arcstat_l2_rebuild_abort_dh_errors);
	wmsum_fini(&arc_sums.arcstat_l2_rebuild_abort_cksum_lb_errors);
	wmsum_fini(&arc_sums.arcstat_l2_rebuild_abort_lowmem);
	wmsum_fini(&arc_sums.arcstat_l2_rebuild_size);
	wmsum_fini(&arc_sums.arcstat_l2_rebuild_asize);
	wmsum_fini(&arc_sums.arcstat_l2_rebuild_bufs);
	wmsum_fini(&arc_sums.arcstat_l2_rebuild_bufs_precached);
	wmsum_fini(&arc_sums.arcstat_l2_rebuild_log_blks);
	wmsum_fini(&arc_sums.arcstat_memory_throttle_count);
	wmsum_fini(&arc_sums.arcstat_memory_direct_count);
	wmsum_fini(&arc_sums.arcstat_memory_indirect_count);
	wmsum_fini(&arc_sums.arcstat_prune);
	wmsum_fini(&arc_sums.arcstat_meta_used);
	wmsum_fini(&arc_sums.arcstat_async_upgrade_sync);
	wmsum_fini(&arc_sums.arcstat_predictive_prefetch);
	wmsum_fini(&arc_sums.arcstat_demand_hit_predictive_prefetch);
	wmsum_fini(&arc_sums.arcstat_demand_iohit_predictive_prefetch);
	wmsum_fini(&arc_sums.arcstat_prescient_prefetch);
	wmsum_fini(&arc_sums.arcstat_demand_hit_prescient_prefetch);
	wmsum_fini(&arc_sums.arcstat_demand_iohit_prescient_prefetch);
	wmsum_fini(&arc_sums.arcstat_raw_size);
	wmsum_fini(&arc_sums.arcstat_cached_only_in_progress);
	wmsum_fini(&arc_sums.arcstat_abd_chunk_waste_size);
	}

	uint64_t
	arc_target_bytes(void)
	{
	return (arc_c);
	}

	void
	arc_set_limits(uint64_t allmem)
	{
	/* Set min cache to 1/32 of all memory, or 32MB, whichever is more. */
	arc_c_min = MAX(allmem / 32, 2ULL << SPA_MAXBLOCKSHIFT);

	/* How to set default max varies by platform. */
	arc_c_max = arc_default_max(arc_c_min, allmem);
	}
	void
	arc_init(void)
	{
	uint64_t percent, allmem = arc_all_memory();
	mutex_init(&arc_evict_lock, NULL, MUTEX_DEFAULT, NULL);
	list_create(&arc_evict_waiters, sizeof (arc_evict_waiter_t),
	offsetof(arc_evict_waiter_t, aew_node));

	arc_min_prefetch_ms = 1000;
	arc_min_prescient_prefetch_ms = 6000;

	#if defined(_KERNEL)
	arc_lowmem_init();
	#endif

	arc_set_limits(allmem);

	#ifdef _KERNEL
	/*
	* If zfs_arc_max is non-zero at init, meaning it was set in the kernel
	* environment before the module was loaded, don't block setting the
	* maximum because it is less than arc_c_min, instead, reset arc_c_min
	* to a lower value.
	* zfs_arc_min will be handled by arc_tuning_update().
	*/
	if (zfs_arc_max != 0 && zfs_arc_max >= MIN_ARC_MAX &&
	zfs_arc_max < allmem) {
	arc_c_max = zfs_arc_max;
	if (arc_c_min >= arc_c_max) {
	arc_c_min = MAX(zfs_arc_max / 2,
	2ULL << SPA_MAXBLOCKSHIFT);
	}
	}
	#else
	/*
	* In userland, there's only the memory pressure that we artificially
	* create (see arc_available_memory()). Don't let arc_c get too
	* small, because it can cause transactions to be larger than
	* arc_c, causing arc_tempreserve_space() to fail.
	*/
	arc_c_min = MAX(arc_c_max / 2, 2ULL << SPA_MAXBLOCKSHIFT);
	#endif

	arc_c = arc_c_min;
	/*
	* 32-bit fixed point fractions of metadata from total ARC size,
	* MRU data from all data and MRU metadata from all metadata.
	*/
	arc_meta = (1ULL << 32) / 4; /* Metadata is 25% of arc_c. */
	arc_pd = (1ULL << 32) / 2; /* Data MRU is 50% of data. */
	arc_pm = (1ULL << 32) / 2; /* Metadata MRU is 50% of metadata. */

	percent = MIN(zfs_arc_dnode_limit_percent, 100);
	arc_dnode_limit = arc_c_max * percent / 100;

	/* Apply user specified tunings */
	arc_tuning_update(B_TRUE);

	/* if kmem_flags are set, lets try to use less memory */
	if (kmem_debugging())
	arc_c = arc_c / 2;
	if (arc_c < arc_c_min)
	arc_c = arc_c_min;

	arc_register_hotplug();

	arc_state_init();

	buf_init();

	list_create(&arc_prune_list, sizeof (arc_prune_t),
	offsetof(arc_prune_t, p_node));
	mutex_init(&arc_prune_mtx, NULL, MUTEX_DEFAULT, NULL);

	arc_prune_taskq = taskq_create("arc_prune", zfs_arc_prune_task_threads,
	defclsyspri, 100, INT_MAX, TASKQ_PREPOPULATE \| TASKQ_DYNAMIC);

	arc_ksp = kstat_create("zfs", 0, "arcstats", "misc", KSTAT_TYPE_NAMED,
	sizeof (arc_stats) / sizeof (kstat_named_t), KSTAT_FLAG_VIRTUAL);

	if (arc_ksp != NULL) {
	arc_ksp->ks_data = &arc_stats;
	arc_ksp->ks_update = arc_kstat_update;
	kstat_install(arc_ksp);
	}

	arc_state_evict_markers =
	arc_state_alloc_markers(arc_state_evict_marker_count);
	arc_evict_zthr = zthr_create_timer("arc_evict",
	arc_evict_cb_check, arc_evict_cb, NULL, SEC2NSEC(1), defclsyspri);
	arc_reap_zthr = zthr_create_timer("arc_reap",
	arc_reap_cb_check, arc_reap_cb, NULL, SEC2NSEC(1), minclsyspri);

	arc_warm = B_FALSE;

	/*
	* Calculate maximum amount of dirty data per pool.
	*
	* If it has been set by a module parameter, take that.
	* Otherwise, use a percentage of physical memory defined by
	* zfs_dirty_data_max_percent (default 10%) with a cap at
	* zfs_dirty_data_max_max (default 4G or 25% of physical memory).
	*/
	#ifdef __LP64__
	if (zfs_dirty_data_max_max == 0)
	zfs_dirty_data_max_max = MIN(4ULL * 1024 * 1024 * 1024,
	allmem * zfs_dirty_data_max_max_percent / 100);
	#else
	if (zfs_dirty_data_max_max == 0)
	zfs_dirty_data_max_max = MIN(1ULL * 1024 * 1024 * 1024,
	allmem * zfs_dirty_data_max_max_percent / 100);
	#endif

	if (zfs_dirty_data_max == 0) {
	zfs_dirty_data_max = allmem *
	zfs_dirty_data_max_percent / 100;
	zfs_dirty_data_max = MIN(zfs_dirty_data_max,
	zfs_dirty_data_max_max);
	}

	if (zfs_wrlog_data_max == 0) {

	/*
	* dp_wrlog_total is reduced for each txg at the end of
	* spa_sync(). However, dp_dirty_total is reduced every time
	* a block is written out. Thus under normal operation,
	* dp_wrlog_total could grow 2 times as big as
	* zfs_dirty_data_max.
	*/
	zfs_wrlog_data_max = zfs_dirty_data_max * 2;
	}
	}

	void
	arc_fini(void)
	{
	arc_prune_t *p;

	#ifdef _KERNEL
	arc_lowmem_fini();
	#endif /* _KERNEL */

	/* Use B_TRUE to ensure all buffers are evicted */
	arc_flush(NULL, B_TRUE);

	if (arc_ksp != NULL) {
	kstat_delete(arc_ksp);
	arc_ksp = NULL;
	}

	taskq_wait(arc_prune_taskq);
	taskq_destroy(arc_prune_taskq);

	mutex_enter(&arc_prune_mtx);
	while ((p = list_remove_head(&arc_prune_list)) != NULL) {
	(void) zfs_refcount_remove(&p->p_refcnt, &arc_prune_list);
	zfs_refcount_destroy(&p->p_refcnt);
	kmem_free(p, sizeof (*p));
	}
	mutex_exit(&arc_prune_mtx);

	list_destroy(&arc_prune_list);
	mutex_destroy(&arc_prune_mtx);

	(void) zthr_cancel(arc_evict_zthr);
	(void) zthr_cancel(arc_reap_zthr);
	arc_state_free_markers(arc_state_evict_markers,
	arc_state_evict_marker_count);

	mutex_destroy(&arc_evict_lock);
	list_destroy(&arc_evict_waiters);

	/*
	* Free any buffers that were tagged for destruction. This needs
	* to occur before arc_state_fini() runs and destroys the aggsum
	* values which are updated when freeing scatter ABDs.
	*/
	l2arc_do_free_on_write();

	/*
	* buf_fini() must proceed arc_state_fini() because buf_fin() may
	* trigger the release of kmem magazines, which can callback to
	* arc_space_return() which accesses aggsums freed in act_state_fini().
	*/
	buf_fini();
	arc_state_fini();

	arc_unregister_hotplug();

	/*
	* We destroy the zthrs after all the ARC state has been
	* torn down to avoid the case of them receiving any
	* wakeup() signals after they are destroyed.
	*/
	zthr_destroy(arc_evict_zthr);
	zthr_destroy(arc_reap_zthr);

	ASSERT0(arc_loaned_bytes);
	}

	/*
	* Level 2 ARC
	*
	* The level 2 ARC (L2ARC) is a cache layer in-between main memory and disk.
	* It uses dedicated storage devices to hold cached data, which are populated
	* using large infrequent writes. The main role of this cache is to boost
	* the performance of random read workloads. The intended L2ARC devices
	* include short-stroked disks, solid state disks, and other media with
	* substantially faster read latency than disk.
	*
	* +-----------------------+
	* \| ARC \|
	* +-----------------------+
	* \| ^ ^
	* \| \| \|
	* l2arc_feed_thread() arc_read()
	* \| \| \|
	* \| l2arc read \|
	* V \| \|
	* +---------------+ \|
	* \| L2ARC \| \|
	* +---------------+ \|
	* \| ^ \|
	* l2arc_write() \| \|
	* \| \| \|
	* V \| \|
	* +-------+ +-------+
	* \| vdev \| \| vdev \|
	* \| cache \| \| cache \|
	* +-------+ +-------+
	* +=========+ .-----.
	* : L2ARC : \|-_____-\|
	* : devices : \| Disks \|
	* +=========+ `-_____-'
	*
	* Read requests are satisfied from the following sources, in order:
	*
	* 1) ARC
	* 2) vdev cache of L2ARC devices
	* 3) L2ARC devices
	* 4) vdev cache of disks
	* 5) disks
	*
	* Some L2ARC device types exhibit extremely slow write performance.
	* To accommodate for this there are some significant differences between
	* the L2ARC and traditional cache design:
	*
	* 1. There is no eviction path from the ARC to the L2ARC. Evictions from
	* the ARC behave as usual, freeing buffers and placing headers on ghost
	* lists. The ARC does not send buffers to the L2ARC during eviction as
	* this would add inflated write latencies for all ARC memory pressure.
	*
	* 2. The L2ARC attempts to cache data from the ARC before it is evicted.
	* It does this by periodically scanning buffers from the eviction-end of
	* the MFU and MRU ARC lists, copying them to the L2ARC devices if they are
	* not already there. It scans until a headroom of buffers is satisfied,
	* which itself is a buffer for ARC eviction. If a compressible buffer is
	* found during scanning and selected for writing to an L2ARC device, we
	* temporarily boost scanning headroom during the next scan cycle to make
	* sure we adapt to compression effects (which might significantly reduce
	* the data volume we write to L2ARC). The thread that does this is
	* l2arc_feed_thread(), illustrated below; example sizes are included to
	* provide a better sense of ratio than this diagram:
	*
	* head --> tail
	* +---------------------+----------+
	* ARC_mfu \|:::::#:::::::::::::::\|o#o###o###\|-->. # already on L2ARC
	* +---------------------+----------+ \| o L2ARC eligible
	* ARC_mru \|:#:::::::::::::::::::\|#o#ooo####\|-->\| : ARC buffer
	* +---------------------+----------+ \|
	* 15.9 Gbytes ^ 32 Mbytes \|
	* headroom \|
	* l2arc_feed_thread()
	* \|
	* l2arc write hand <--[oooo]--'
	* \| 8 Mbyte
	* \| write max
	* V
	* +==============================+
	* L2ARC dev \|####\|#\|###\|###\| \|####\| ... \|
	* +==============================+
	* 32 Gbytes
	*
	* 3. If an ARC buffer is copied to the L2ARC but then hit instead of
	* evicted, then the L2ARC has cached a buffer much sooner than it probably
	* needed to, potentially wasting L2ARC device bandwidth and storage. It is
	* safe to say that this is an uncommon case, since buffers at the end of
	* the ARC lists have moved there due to inactivity.
	*
	* 4. If the ARC evicts faster than the L2ARC can maintain a headroom,
	* then the L2ARC simply misses copying some buffers. This serves as a
	* pressure valve to prevent heavy read workloads from both stalling the ARC
	* with waits and clogging the L2ARC with writes. This also helps prevent
	* the potential for the L2ARC to churn if it attempts to cache content too
	* quickly, such as during backups of the entire pool.
	*
	* 5. After system boot and before the ARC has filled main memory, there are
	* no evictions from the ARC and so the tails of the ARC_mfu and ARC_mru
	* lists can remain mostly static. Instead of searching from tail of these
	* lists as pictured, the l2arc_feed_thread() will search from the list heads
	* for eligible buffers, greatly increasing its chance of finding them.
	*
	* The L2ARC device write speed is also boosted during this time so that
	* the L2ARC warms up faster. Since there have been no ARC evictions yet,
	* there are no L2ARC reads, and no fear of degrading read performance
	* through increased writes.
	*
	* 6. Writes to the L2ARC devices are grouped and sent in-sequence, so that
	* the vdev queue can aggregate them into larger and fewer writes. Each
	* device is written to in a rotor fashion, sweeping writes through
	* available space then repeating.
	*
	* 7. The L2ARC does not store dirty content. It never needs to flush
	* write buffers back to disk based storage.
	*
	* 8. If an ARC buffer is written (and dirtied) which also exists in the
	* L2ARC, the now stale L2ARC buffer is immediately dropped.
	*
	* The performance of the L2ARC can be tweaked by a number of tunables, which
	* may be necessary for different workloads:
	*
	* l2arc_write_max max write bytes per interval
	* l2arc_write_boost extra write bytes during device warmup
	* l2arc_noprefetch skip caching prefetched buffers
	* l2arc_headroom number of max device writes to precache
	* l2arc_headroom_boost when we find compressed buffers during ARC
	* scanning, we multiply headroom by this
	* percentage factor for the next scan cycle,
	* since more compressed buffers are likely to
	* be present
	* l2arc_feed_secs seconds between L2ARC writing
	*
	* Tunables may be removed or added as future performance improvements are
	* integrated, and also may become zpool properties.
	*
	* There are three key functions that control how the L2ARC warms up:
	*
	* l2arc_write_eligible() check if a buffer is eligible to cache
	* l2arc_write_size() calculate how much to write
	* l2arc_write_interval() calculate sleep delay between writes
	*
	* These three functions determine what to write, how much, and how quickly
	* to send writes.
	*
	* L2ARC persistence:
	*
	* When writing buffers to L2ARC, we periodically add some metadata to
	* make sure we can pick them up after reboot, thus dramatically reducing
	* the impact that any downtime has on the performance of storage systems
	* with large caches.
	*
	* The implementation works fairly simply by integrating the following two
	* modifications:
	*
	* *) When writing to the L2ARC, we occasionally write a "l2arc log block",
	* which is an additional piece of metadata which describes what's been
	* written. This allows us to rebuild the arc_buf_hdr_t structures of the
	* main ARC buffers. There are 2 linked-lists of log blocks headed by
	* dh_start_lbps[2]. We alternate which chain we append to, so they are
	* time-wise and offset-wise interleaved, but that is an optimization rather
	* than for correctness. The log block also includes a pointer to the
	* previous block in its chain.
	*
	* *) We reserve SPA_MINBLOCKSIZE of space at the start of each L2ARC device
	* for our header bookkeeping purposes. This contains a device header,
	* which contains our top-level reference structures. We update it each
	* time we write a new log block, so that we're able to locate it in the
	* L2ARC device. If this write results in an inconsistent device header
	* (e.g. due to power failure), we detect this by verifying the header's
	* checksum and simply fail to reconstruct the L2ARC after reboot.
	*
	* Implementation diagram:
	*
	* +=== L2ARC device (not to scale) ======================================+
	* \| ___two newest log block pointers__.__________ \|
	* \| / \dh_start_lbps[1] \|
	* \| / \ \dh_start_lbps[0]\|
	* \|.___/__. V V \|
	* \|\|L2 dev\|....\|lb \|bufs \|lb \|bufs \|lb \|bufs \|lb \|bufs \|lb \|---(empty)---\|
	* \|\| hdr\| ^ /^ /^ / / \|
	* \|+------+ ...--\-------/ \-----/--\------/ / \|
	* \| \--------------/ \--------------/ \|
	* +======================================================================+
	*
	* As can be seen on the diagram, rather than using a simple linked list,
	* we use a pair of linked lists with alternating elements. This is a
	* performance enhancement due to the fact that we only find out the
	* address of the next log block access once the current block has been
	* completely read in. Obviously, this hurts performance, because we'd be
	* keeping the device's I/O queue at only a 1 operation deep, thus
	* incurring a large amount of I/O round-trip latency. Having two lists
	* allows us to fetch two log blocks ahead of where we are currently
	* rebuilding L2ARC buffers.
	*
	* On-device data structures:
	*
	* L2ARC device header: l2arc_dev_hdr_phys_t
	* L2ARC log block: l2arc_log_blk_phys_t
	*
	* L2ARC reconstruction:
	*
	* When writing data, we simply write in the standard rotary fashion,
	* evicting buffers as we go and simply writing new data over them (writing
	* a new log block every now and then). This obviously means that once we
	* loop around the end of the device, we will start cutting into an already
	* committed log block (and its referenced data buffers), like so:
	*
	* current write head__ __old tail
	* \ /
	* V V
	* <--\|bufs \|lb \|bufs \|lb \| \|bufs \|lb \|bufs \|lb \|-->
	* ^ ^^^^^^^^^___________________________________
	* \| \
	* <<nextwrite>> may overwrite this blk and/or its bufs --'
	*
	* When importing the pool, we detect this situation and use it to stop
	* our scanning process (see l2arc_rebuild).
	*
	* There is one significant caveat to consider when rebuilding ARC contents
	* from an L2ARC device: what about invalidated buffers? Given the above
	* construction, we cannot update blocks which we've already written to amend
	* them to remove buffers which were invalidated. Thus, during reconstruction,
	* we might be populating the cache with buffers for data that's not on the
	* main pool anymore, or may have been overwritten!
	*
	* As it turns out, this isn't a problem. Every arc_read request includes
	* both the DVA and, crucially, the birth TXG of the BP the caller is
	* looking for. So even if the cache were populated by completely rotten
	* blocks for data that had been long deleted and/or overwritten, we'll
	* never actually return bad data from the cache, since the DVA with the
	* birth TXG uniquely identify a block in space and time - once created,
	* a block is immutable on disk. The worst thing we have done is wasted
	* some time and memory at l2arc rebuild to reconstruct outdated ARC
	* entries that will get dropped from the l2arc as it is being updated
	* with new blocks.
	*
	* L2ARC buffers that have been evicted by l2arc_evict() ahead of the write
	* hand are not restored. This is done by saving the offset (in bytes)
	* l2arc_evict() has evicted to in the L2ARC device header and taking it
	* into account when restoring buffers.
	*/

	static boolean_t
	l2arc_write_eligible(uint64_t spa_guid, arc_buf_hdr_t *hdr)
	{
	/*
	* A buffer is not eligible for the L2ARC if it:
	* 1. belongs to a different spa.
	* 2. is already cached on the L2ARC.
	* 3. has an I/O in progress (it may be an incomplete read).
	* 4. is flagged not eligible (zfs property).
	*/
	if (hdr->b_spa != spa_guid \|\| HDR_HAS_L2HDR(hdr) \|\|
	HDR_IO_IN_PROGRESS(hdr) \|\| !HDR_L2CACHE(hdr))
	return (B_FALSE);

	return (B_TRUE);
	}

	static uint64_t
	l2arc_write_size(l2arc_dev_t *dev)
	{
	uint64_t size;

	/*
	* Make sure our globals have meaningful values in case the user
	* altered them.
	*/
	size = l2arc_write_max;
	if (size == 0) {
	cmn_err(CE_NOTE, "l2arc_write_max must be greater than zero, "
	"resetting it to the default (%d)", L2ARC_WRITE_SIZE);
	size = l2arc_write_max = L2ARC_WRITE_SIZE;
	}

	if (arc_warm == B_FALSE)
	size += l2arc_write_boost;

	/* We need to add in the worst case scenario of log block overhead. */
	size += l2arc_log_blk_overhead(size, dev);
	if (dev->l2ad_vdev->vdev_has_trim && l2arc_trim_ahead > 0) {
	/*
	* Trim ahead of the write size 64MB or (l2arc_trim_ahead/100)
	* times the writesize, whichever is greater.
	*/
	size += MAX(64 * 1024 * 1024,
	(size * l2arc_trim_ahead) / 100);
	}

	/*
	* Make sure the write size does not exceed the size of the cache
	* device. This is important in l2arc_evict(), otherwise infinite
	* iteration can occur.
	*/
	size = MIN(size, (dev->l2ad_end - dev->l2ad_start) / 4);

	size = P2ROUNDUP(size, 1ULL << dev->l2ad_vdev->vdev_ashift);

	return (size);

	}

	static clock_t
	l2arc_write_interval(clock_t began, uint64_t wanted, uint64_t wrote)
	{
	clock_t interval, next, now;

	/*
	* If the ARC lists are busy, increase our write rate; if the
	* lists are stale, idle back. This is achieved by checking
	* how much we previously wrote - if it was more than half of
	* what we wanted, schedule the next write much sooner.
	*/
	if (l2arc_feed_again && wrote > (wanted / 2))
	interval = (hz * l2arc_feed_min_ms) / 1000;
	else
	interval = hz * l2arc_feed_secs;

	now = ddi_get_lbolt();
	next = MAX(now, MIN(now + interval, began + interval));

	return (next);
	}

	/*
	* Cycle through L2ARC devices. This is how L2ARC load balances.
	* If a device is returned, this also returns holding the spa config lock.
	*/
	static l2arc_dev_t *
	l2arc_dev_get_next(void)
	{
	l2arc_dev_t first, next = NULL;

	/*
	* Lock out the removal of spas (spa_namespace_lock), then removal
	* of cache devices (l2arc_dev_mtx). Once a device has been selected,
	* both locks will be dropped and a spa config lock held instead.
	*/
	mutex_enter(&spa_namespace_lock);
	mutex_enter(&l2arc_dev_mtx);

	/* if there are no vdevs, there is nothing to do */
	if (l2arc_ndev == 0)
	goto out;

	first = NULL;
	next = l2arc_dev_last;
	do {
	/* loop around the list looking for a non-faulted vdev */
	if (next == NULL) {
	next = list_head(l2arc_dev_list);
	} else {
	next = list_next(l2arc_dev_list, next);
	if (next == NULL)
	next = list_head(l2arc_dev_list);
	}

	/* if we have come back to the start, bail out */
	if (first == NULL)
	first = next;
	else if (next == first)
	break;

	ASSERT3P(next, !=, NULL);
	} while (vdev_is_dead(next->l2ad_vdev) \|\| next->l2ad_rebuild \|\|
	- next->l2ad_trim_all);
	+ next->l2ad_trim_all \|\| next->l2ad_spa->spa_is_exporting);

	/* if we were unable to find any usable vdevs, return NULL */
	if (vdev_is_dead(next->l2ad_vdev) \|\| next->l2ad_rebuild \|\|
	- next->l2ad_trim_all)
	+ next->l2ad_trim_all \|\| next->l2ad_spa->spa_is_exporting)
	next = NULL;

	l2arc_dev_last = next;

	out:
	mutex_exit(&l2arc_dev_mtx);

	/*
	* Grab the config lock to prevent the 'next' device from being
	* removed while we are writing to it.
	*/
	if (next != NULL)
	spa_config_enter(next->l2ad_spa, SCL_L2ARC, next, RW_READER);
	mutex_exit(&spa_namespace_lock);

	return (next);
	}

	/*
	* Free buffers that were tagged for destruction.
	*/
	static void
	l2arc_do_free_on_write(void)
	{
	l2arc_data_free_t *df;

	mutex_enter(&l2arc_free_on_write_mtx);
	while ((df = list_remove_head(l2arc_free_on_write)) != NULL) {
	ASSERT3P(df->l2df_abd, !=, NULL);
	abd_free(df->l2df_abd);
	kmem_free(df, sizeof (l2arc_data_free_t));
	}
	mutex_exit(&l2arc_free_on_write_mtx);
	}

	/*
	* A write to a cache device has completed. Update all headers to allow
	* reads from these buffers to begin.
	*/
	static void
	l2arc_write_done(zio_t *zio)
	{
	l2arc_write_callback_t *cb;
	l2arc_lb_abd_buf_t *abd_buf;
	l2arc_lb_ptr_buf_t *lb_ptr_buf;
	l2arc_dev_t *dev;
	l2arc_dev_hdr_phys_t *l2dhdr;
	list_t *buflist;
	arc_buf_hdr_t head, hdr, *hdr_prev;
	kmutex_t *hash_lock;
	int64_t bytes_dropped = 0;

	cb = zio->io_private;
	ASSERT3P(cb, !=, NULL);
	dev = cb->l2wcb_dev;
	l2dhdr = dev->l2ad_dev_hdr;
	ASSERT3P(dev, !=, NULL);
	head = cb->l2wcb_head;
	ASSERT3P(head, !=, NULL);
	buflist = &dev->l2ad_buflist;
	ASSERT3P(buflist, !=, NULL);
	DTRACE_PROBE2(l2arc__iodone, zio_t *, zio,
	l2arc_write_callback_t *, cb);

	/*
	* All writes completed, or an error was hit.
	*/
	top:
	mutex_enter(&dev->l2ad_mtx);
	for (hdr = list_prev(buflist, head); hdr; hdr = hdr_prev) {
	hdr_prev = list_prev(buflist, hdr);

	hash_lock = HDR_LOCK(hdr);

	/*
	* We cannot use mutex_enter or else we can deadlock
	* with l2arc_write_buffers (due to swapping the order
	* the hash lock and l2ad_mtx are taken).
	*/
	if (!mutex_tryenter(hash_lock)) {
	/*
	* Missed the hash lock. We must retry so we
	* don't leave the ARC_FLAG_L2_WRITING bit set.
	*/
	ARCSTAT_BUMP(arcstat_l2_writes_lock_retry);

	/*
	* We don't want to rescan the headers we've
	* already marked as having been written out, so
	* we reinsert the head node so we can pick up
	* where we left off.
	*/
	list_remove(buflist, head);
	list_insert_after(buflist, hdr, head);

	mutex_exit(&dev->l2ad_mtx);

	/*
	* We wait for the hash lock to become available
	* to try and prevent busy waiting, and increase
	* the chance we'll be able to acquire the lock
	* the next time around.
	*/
	mutex_enter(hash_lock);
	mutex_exit(hash_lock);
	goto top;
	}

	/*
	* We could not have been moved into the arc_l2c_only
	* state while in-flight due to our ARC_FLAG_L2_WRITING
	* bit being set. Let's just ensure that's being enforced.
	*/
	ASSERT(HDR_HAS_L1HDR(hdr));

	/*
	* Skipped - drop L2ARC entry and mark the header as no
	* longer L2 eligibile.
	*/
	if (zio->io_error != 0) {
	/*
	* Error - drop L2ARC entry.
	*/
	list_remove(buflist, hdr);
	arc_hdr_clear_flags(hdr, ARC_FLAG_HAS_L2HDR);

	uint64_t psize = HDR_GET_PSIZE(hdr);
	l2arc_hdr_arcstats_decrement(hdr);

	bytes_dropped +=
	vdev_psize_to_asize(dev->l2ad_vdev, psize);
	(void) zfs_refcount_remove_many(&dev->l2ad_alloc,
	arc_hdr_size(hdr), hdr);
	}

	/*
	* Allow ARC to begin reads and ghost list evictions to
	* this L2ARC entry.
	*/
	arc_hdr_clear_flags(hdr, ARC_FLAG_L2_WRITING);

	mutex_exit(hash_lock);
	}

	/*
	* Free the allocated abd buffers for writing the log blocks.
	* If the zio failed reclaim the allocated space and remove the
	* pointers to these log blocks from the log block pointer list
	* of the L2ARC device.
	*/
	while ((abd_buf = list_remove_tail(&cb->l2wcb_abd_list)) != NULL) {
	abd_free(abd_buf->abd);
	zio_buf_free(abd_buf, sizeof (*abd_buf));
	if (zio->io_error != 0) {
	lb_ptr_buf = list_remove_head(&dev->l2ad_lbptr_list);
	/*
	* L2BLK_GET_PSIZE returns aligned size for log
	* blocks.
	*/
	uint64_t asize =
	L2BLK_GET_PSIZE((lb_ptr_buf->lb_ptr)->lbp_prop);
	bytes_dropped += asize;
	ARCSTAT_INCR(arcstat_l2_log_blk_asize, -asize);
	ARCSTAT_BUMPDOWN(arcstat_l2_log_blk_count);
	zfs_refcount_remove_many(&dev->l2ad_lb_asize, asize,
	lb_ptr_buf);
	(void) zfs_refcount_remove(&dev->l2ad_lb_count,
	lb_ptr_buf);
	kmem_free(lb_ptr_buf->lb_ptr,
	sizeof (l2arc_log_blkptr_t));
	kmem_free(lb_ptr_buf, sizeof (l2arc_lb_ptr_buf_t));
	}
	}
	list_destroy(&cb->l2wcb_abd_list);

	if (zio->io_error != 0) {
	ARCSTAT_BUMP(arcstat_l2_writes_error);

	/*
	* Restore the lbps array in the header to its previous state.
	* If the list of log block pointers is empty, zero out the
	* log block pointers in the device header.
	*/
	lb_ptr_buf = list_head(&dev->l2ad_lbptr_list);
	for (int i = 0; i < 2; i++) {
	if (lb_ptr_buf == NULL) {
	/*
	* If the list is empty zero out the device
	* header. Otherwise zero out the second log
	* block pointer in the header.
	*/
	if (i == 0) {
	memset(l2dhdr, 0,
	dev->l2ad_dev_hdr_asize);
	} else {
	memset(&l2dhdr->dh_start_lbps[i], 0,
	sizeof (l2arc_log_blkptr_t));
	}
	break;
	}
	memcpy(&l2dhdr->dh_start_lbps[i], lb_ptr_buf->lb_ptr,
	sizeof (l2arc_log_blkptr_t));
	lb_ptr_buf = list_next(&dev->l2ad_lbptr_list,
	lb_ptr_buf);
	}
	}

	ARCSTAT_BUMP(arcstat_l2_writes_done);
	list_remove(buflist, head);
	ASSERT(!HDR_HAS_L1HDR(head));
	kmem_cache_free(hdr_l2only_cache, head);
	mutex_exit(&dev->l2ad_mtx);

	ASSERT(dev->l2ad_vdev != NULL);
	vdev_space_update(dev->l2ad_vdev, -bytes_dropped, 0, 0);

	l2arc_do_free_on_write();

	kmem_free(cb, sizeof (l2arc_write_callback_t));
	}

	static int
	l2arc_untransform(zio_t zio, l2arc_read_callback_t cb)
	{
	int ret;
	spa_t *spa = zio->io_spa;
	arc_buf_hdr_t *hdr = cb->l2rcb_hdr;
	blkptr_t *bp = zio->io_bp;
	uint8_t salt[ZIO_DATA_SALT_LEN];
	uint8_t iv[ZIO_DATA_IV_LEN];
	uint8_t mac[ZIO_DATA_MAC_LEN];
	boolean_t no_crypt = B_FALSE;

	/*
	* ZIL data is never be written to the L2ARC, so we don't need
	* special handling for its unique MAC storage.
	*/
	ASSERT3U(BP_GET_TYPE(bp), !=, DMU_OT_INTENT_LOG);
	ASSERT(MUTEX_HELD(HDR_LOCK(hdr)));
	ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);

	/*
	* If the data was encrypted, decrypt it now. Note that
	* we must check the bp here and not the hdr, since the
	* hdr does not have its encryption parameters updated
	* until arc_read_done().
	*/
	if (BP_IS_ENCRYPTED(bp)) {
	abd_t *eabd = arc_get_data_abd(hdr, arc_hdr_size(hdr), hdr,
	ARC_HDR_USE_RESERVE);

	zio_crypt_decode_params_bp(bp, salt, iv);
	zio_crypt_decode_mac_bp(bp, mac);

	ret = spa_do_crypt_abd(B_FALSE, spa, &cb->l2rcb_zb,
	BP_GET_TYPE(bp), BP_GET_DEDUP(bp), BP_SHOULD_BYTESWAP(bp),
	salt, iv, mac, HDR_GET_PSIZE(hdr), eabd,
	hdr->b_l1hdr.b_pabd, &no_crypt);
	if (ret != 0) {
	arc_free_data_abd(hdr, eabd, arc_hdr_size(hdr), hdr);
	goto error;
	}

	/*
	* If we actually performed decryption, replace b_pabd
	* with the decrypted data. Otherwise we can just throw
	* our decryption buffer away.
	*/
	if (!no_crypt) {
	arc_free_data_abd(hdr, hdr->b_l1hdr.b_pabd,
	arc_hdr_size(hdr), hdr);
	hdr->b_l1hdr.b_pabd = eabd;
	zio->io_abd = eabd;
	} else {
	arc_free_data_abd(hdr, eabd, arc_hdr_size(hdr), hdr);
	}
	}

	/*
	* If the L2ARC block was compressed, but ARC compression
	* is disabled we decompress the data into a new buffer and
	* replace the existing data.
	*/
	if (HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF &&
	!HDR_COMPRESSION_ENABLED(hdr)) {
	abd_t *cabd = arc_get_data_abd(hdr, arc_hdr_size(hdr), hdr,
	ARC_HDR_USE_RESERVE);
	void *tmp = abd_borrow_buf(cabd, arc_hdr_size(hdr));

	ret = zio_decompress_data(HDR_GET_COMPRESS(hdr),
	hdr->b_l1hdr.b_pabd, tmp, HDR_GET_PSIZE(hdr),
	HDR_GET_LSIZE(hdr), &hdr->b_complevel);
	if (ret != 0) {
	abd_return_buf_copy(cabd, tmp, arc_hdr_size(hdr));
	arc_free_data_abd(hdr, cabd, arc_hdr_size(hdr), hdr);
	goto error;
	}

	abd_return_buf_copy(cabd, tmp, arc_hdr_size(hdr));
	arc_free_data_abd(hdr, hdr->b_l1hdr.b_pabd,
	arc_hdr_size(hdr), hdr);
	hdr->b_l1hdr.b_pabd = cabd;
	zio->io_abd = cabd;
	zio->io_size = HDR_GET_LSIZE(hdr);
	}

	return (0);

	error:
	return (ret);
	}


	/*
	* A read to a cache device completed. Validate buffer contents before
	* handing over to the regular ARC routines.
	*/
	static void
	l2arc_read_done(zio_t *zio)
	{
	int tfm_error = 0;
	l2arc_read_callback_t *cb = zio->io_private;
	arc_buf_hdr_t *hdr;
	kmutex_t *hash_lock;
	boolean_t valid_cksum;
	boolean_t using_rdata = (BP_IS_ENCRYPTED(&cb->l2rcb_bp) &&
	(cb->l2rcb_flags & ZIO_FLAG_RAW_ENCRYPT));

	ASSERT3P(zio->io_vd, !=, NULL);
	ASSERT(zio->io_flags & ZIO_FLAG_DONT_PROPAGATE);

	spa_config_exit(zio->io_spa, SCL_L2ARC, zio->io_vd);

	ASSERT3P(cb, !=, NULL);
	hdr = cb->l2rcb_hdr;
	ASSERT3P(hdr, !=, NULL);

	hash_lock = HDR_LOCK(hdr);
	mutex_enter(hash_lock);
	ASSERT3P(hash_lock, ==, HDR_LOCK(hdr));

	/*
	* If the data was read into a temporary buffer,
	* move it and free the buffer.
	*/
	if (cb->l2rcb_abd != NULL) {
	ASSERT3U(arc_hdr_size(hdr), <, zio->io_size);
	if (zio->io_error == 0) {
	if (using_rdata) {
	abd_copy(hdr->b_crypt_hdr.b_rabd,
	cb->l2rcb_abd, arc_hdr_size(hdr));
	} else {
	abd_copy(hdr->b_l1hdr.b_pabd,
	cb->l2rcb_abd, arc_hdr_size(hdr));
	}
	}

	/*
	* The following must be done regardless of whether
	* there was an error:
	* - free the temporary buffer
	* - point zio to the real ARC buffer
	* - set zio size accordingly
	* These are required because zio is either re-used for
	* an I/O of the block in the case of the error
	* or the zio is passed to arc_read_done() and it
	* needs real data.
	*/
	abd_free(cb->l2rcb_abd);
	zio->io_size = zio->io_orig_size = arc_hdr_size(hdr);

	if (using_rdata) {
	ASSERT(HDR_HAS_RABD(hdr));
	zio->io_abd = zio->io_orig_abd =
	hdr->b_crypt_hdr.b_rabd;
	} else {
	ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL);
	zio->io_abd = zio->io_orig_abd = hdr->b_l1hdr.b_pabd;
	}
	}

	ASSERT3P(zio->io_abd, !=, NULL);

	/*
	* Check this survived the L2ARC journey.
	*/
	ASSERT(zio->io_abd == hdr->b_l1hdr.b_pabd \|\|
	(HDR_HAS_RABD(hdr) && zio->io_abd == hdr->b_crypt_hdr.b_rabd));
	zio->io_bp_copy = cb->l2rcb_bp; /* XXX fix in L2ARC 2.0 */
	zio->io_bp = &zio->io_bp_copy; /* XXX fix in L2ARC 2.0 */
	zio->io_prop.zp_complevel = hdr->b_complevel;

	valid_cksum = arc_cksum_is_equal(hdr, zio);

	/*
	* b_rabd will always match the data as it exists on disk if it is
	* being used. Therefore if we are reading into b_rabd we do not
	* attempt to untransform the data.
	*/
	if (valid_cksum && !using_rdata)
	tfm_error = l2arc_untransform(zio, cb);

	if (valid_cksum && tfm_error == 0 && zio->io_error == 0 &&
	!HDR_L2_EVICTED(hdr)) {
	mutex_exit(hash_lock);
	zio->io_private = hdr;
	arc_read_done(zio);
	} else {
	/*
	* Buffer didn't survive caching. Increment stats and
	* reissue to the original storage device.
	*/
	if (zio->io_error != 0) {
	ARCSTAT_BUMP(arcstat_l2_io_error);
	} else {
	zio->io_error = SET_ERROR(EIO);
	}
	if (!valid_cksum \|\| tfm_error != 0)
	ARCSTAT_BUMP(arcstat_l2_cksum_bad);

	/*
	* If there's no waiter, issue an async i/o to the primary
	* storage now. If there is a waiter, the caller must
	* issue the i/o in a context where it's OK to block.
	*/
	if (zio->io_waiter == NULL) {
	zio_t *pio = zio_unique_parent(zio);
	void *abd = (using_rdata) ?
	hdr->b_crypt_hdr.b_rabd : hdr->b_l1hdr.b_pabd;

	ASSERT(!pio \|\| pio->io_child_type == ZIO_CHILD_LOGICAL);

	zio = zio_read(pio, zio->io_spa, zio->io_bp,
	abd, zio->io_size, arc_read_done,
	hdr, zio->io_priority, cb->l2rcb_flags,
	&cb->l2rcb_zb);

	/*
	* Original ZIO will be freed, so we need to update
	* ARC header with the new ZIO pointer to be used
	* by zio_change_priority() in arc_read().
	*/
	for (struct arc_callback *acb = hdr->b_l1hdr.b_acb;
	acb != NULL; acb = acb->acb_next)
	acb->acb_zio_head = zio;

	mutex_exit(hash_lock);
	zio_nowait(zio);
	} else {
	mutex_exit(hash_lock);
	}
	}

	kmem_free(cb, sizeof (l2arc_read_callback_t));
	}

	/*
	* This is the list priority from which the L2ARC will search for pages to
	* cache. This is used within loops (0..3) to cycle through lists in the
	* desired order. This order can have a significant effect on cache
	* performance.
	*
	* Currently the metadata lists are hit first, MFU then MRU, followed by
	* the data lists. This function returns a locked list, and also returns
	* the lock pointer.
	*/
	static multilist_sublist_t *
	l2arc_sublist_lock(int list_num)
	{
	multilist_t *ml = NULL;
	unsigned int idx;

	ASSERT(list_num >= 0 && list_num < L2ARC_FEED_TYPES);

	switch (list_num) {
	case 0:
	ml = &arc_mfu->arcs_list[ARC_BUFC_METADATA];
	break;
	case 1:
	ml = &arc_mru->arcs_list[ARC_BUFC_METADATA];
	break;
	case 2:
	ml = &arc_mfu->arcs_list[ARC_BUFC_DATA];
	break;
	case 3:
	ml = &arc_mru->arcs_list[ARC_BUFC_DATA];
	break;
	default:
	return (NULL);
	}

	/*
	* Return a randomly-selected sublist. This is acceptable
	* because the caller feeds only a little bit of data for each
	* call (8MB). Subsequent calls will result in different
	* sublists being selected.
	*/
	idx = multilist_get_random_index(ml);
	return (multilist_sublist_lock_idx(ml, idx));
	}

	/*
	* Calculates the maximum overhead of L2ARC metadata log blocks for a given
	* L2ARC write size. l2arc_evict and l2arc_write_size need to include this
	* overhead in processing to make sure there is enough headroom available
	* when writing buffers.
	*/
	static inline uint64_t
	l2arc_log_blk_overhead(uint64_t write_sz, l2arc_dev_t *dev)
	{
	if (dev->l2ad_log_entries == 0) {
	return (0);
	} else {
	uint64_t log_entries = write_sz >> SPA_MINBLOCKSHIFT;

	uint64_t log_blocks = (log_entries +
	dev->l2ad_log_entries - 1) /
	dev->l2ad_log_entries;

	return (vdev_psize_to_asize(dev->l2ad_vdev,
	sizeof (l2arc_log_blk_phys_t)) * log_blocks);
	}
	}

	/*
	* Evict buffers from the device write hand to the distance specified in
	* bytes. This distance may span populated buffers, it may span nothing.
	* This is clearing a region on the L2ARC device ready for writing.
	* If the 'all' boolean is set, every buffer is evicted.
	*/
	static void
	l2arc_evict(l2arc_dev_t *dev, uint64_t distance, boolean_t all)
	{
	list_t *buflist;
	arc_buf_hdr_t hdr, hdr_prev;
	kmutex_t *hash_lock;
	uint64_t taddr;
	l2arc_lb_ptr_buf_t lb_ptr_buf, lb_ptr_buf_prev;
	vdev_t *vd = dev->l2ad_vdev;
	boolean_t rerun;

	buflist = &dev->l2ad_buflist;

	top:
	rerun = B_FALSE;
	if (dev->l2ad_hand + distance > dev->l2ad_end) {
	/*
	* When there is no space to accommodate upcoming writes,
	* evict to the end. Then bump the write and evict hands
	* to the start and iterate. This iteration does not
	* happen indefinitely as we make sure in
	* l2arc_write_size() that when the write hand is reset,
	* the write size does not exceed the end of the device.
	*/
	rerun = B_TRUE;
	taddr = dev->l2ad_end;
	} else {
	taddr = dev->l2ad_hand + distance;
	}
	DTRACE_PROBE4(l2arc__evict, l2arc_dev_t , dev, list_t , buflist,
	uint64_t, taddr, boolean_t, all);

	if (!all) {
	/*
	* This check has to be placed after deciding whether to
	* iterate (rerun).
	*/
	if (dev->l2ad_first) {
	/*
	* This is the first sweep through the device. There is
	* nothing to evict. We have already trimmmed the
	* whole device.
	*/
	goto out;
	} else {
	/*
	* Trim the space to be evicted.
	*/
	if (vd->vdev_has_trim && dev->l2ad_evict < taddr &&
	l2arc_trim_ahead > 0) {
	/*
	* We have to drop the spa_config lock because
	* vdev_trim_range() will acquire it.
	* l2ad_evict already accounts for the label
	* size. To prevent vdev_trim_ranges() from
	* adding it again, we subtract it from
	* l2ad_evict.
	*/
	spa_config_exit(dev->l2ad_spa, SCL_L2ARC, dev);
	vdev_trim_simple(vd,
	dev->l2ad_evict - VDEV_LABEL_START_SIZE,
	taddr - dev->l2ad_evict);
	spa_config_enter(dev->l2ad_spa, SCL_L2ARC, dev,
	RW_READER);
	}

	/*
	* When rebuilding L2ARC we retrieve the evict hand
	* from the header of the device. Of note, l2arc_evict()
	* does not actually delete buffers from the cache
	* device, but trimming may do so depending on the
	* hardware implementation. Thus keeping track of the
	* evict hand is useful.
	*/
	dev->l2ad_evict = MAX(dev->l2ad_evict, taddr);
	}
	}

	retry:
	mutex_enter(&dev->l2ad_mtx);
	/*
	* We have to account for evicted log blocks. Run vdev_space_update()
	* on log blocks whose offset (in bytes) is before the evicted offset
	* (in bytes) by searching in the list of pointers to log blocks
	* present in the L2ARC device.
	*/
	for (lb_ptr_buf = list_tail(&dev->l2ad_lbptr_list); lb_ptr_buf;
	lb_ptr_buf = lb_ptr_buf_prev) {

	lb_ptr_buf_prev = list_prev(&dev->l2ad_lbptr_list, lb_ptr_buf);

	/* L2BLK_GET_PSIZE returns aligned size for log blocks */
	uint64_t asize = L2BLK_GET_PSIZE(
	(lb_ptr_buf->lb_ptr)->lbp_prop);

	/*
	* We don't worry about log blocks left behind (ie
	* lbp_payload_start < l2ad_hand) because l2arc_write_buffers()
	* will never write more than l2arc_evict() evicts.
	*/
	if (!all && l2arc_log_blkptr_valid(dev, lb_ptr_buf->lb_ptr)) {
	break;
	} else {
	vdev_space_update(vd, -asize, 0, 0);
	ARCSTAT_INCR(arcstat_l2_log_blk_asize, -asize);
	ARCSTAT_BUMPDOWN(arcstat_l2_log_blk_count);
	zfs_refcount_remove_many(&dev->l2ad_lb_asize, asize,
	lb_ptr_buf);
	(void) zfs_refcount_remove(&dev->l2ad_lb_count,
	lb_ptr_buf);
	list_remove(&dev->l2ad_lbptr_list, lb_ptr_buf);
	kmem_free(lb_ptr_buf->lb_ptr,
	sizeof (l2arc_log_blkptr_t));
	kmem_free(lb_ptr_buf, sizeof (l2arc_lb_ptr_buf_t));
	}
	}

	for (hdr = list_tail(buflist); hdr; hdr = hdr_prev) {
	hdr_prev = list_prev(buflist, hdr);

	ASSERT(!HDR_EMPTY(hdr));
	hash_lock = HDR_LOCK(hdr);

	/*
	* We cannot use mutex_enter or else we can deadlock
	* with l2arc_write_buffers (due to swapping the order
	* the hash lock and l2ad_mtx are taken).
	*/
	if (!mutex_tryenter(hash_lock)) {
	/*
	* Missed the hash lock. Retry.
	*/
	ARCSTAT_BUMP(arcstat_l2_evict_lock_retry);
	mutex_exit(&dev->l2ad_mtx);
	mutex_enter(hash_lock);
	mutex_exit(hash_lock);
	goto retry;
	}

	/*
	* A header can't be on this list if it doesn't have L2 header.
	*/
	ASSERT(HDR_HAS_L2HDR(hdr));

	/* Ensure this header has finished being written. */
	ASSERT(!HDR_L2_WRITING(hdr));
	ASSERT(!HDR_L2_WRITE_HEAD(hdr));

	if (!all && (hdr->b_l2hdr.b_daddr >= dev->l2ad_evict \|\|
	hdr->b_l2hdr.b_daddr < dev->l2ad_hand)) {
	/*
	* We've evicted to the target address,
	* or the end of the device.
	*/
	mutex_exit(hash_lock);
	break;
	}

	if (!HDR_HAS_L1HDR(hdr)) {
	ASSERT(!HDR_L2_READING(hdr));
	/*
	* This doesn't exist in the ARC. Destroy.
	* arc_hdr_destroy() will call list_remove()
	* and decrement arcstat_l2_lsize.
	*/
	arc_change_state(arc_anon, hdr);
	arc_hdr_destroy(hdr);
	} else {
	ASSERT(hdr->b_l1hdr.b_state != arc_l2c_only);
	ARCSTAT_BUMP(arcstat_l2_evict_l1cached);
	/*
	* Invalidate issued or about to be issued
	* reads, since we may be about to write
	* over this location.
	*/
	if (HDR_L2_READING(hdr)) {
	ARCSTAT_BUMP(arcstat_l2_evict_reading);
	arc_hdr_set_flags(hdr, ARC_FLAG_L2_EVICTED);
	}

	arc_hdr_l2hdr_destroy(hdr);
	}
	mutex_exit(hash_lock);
	}
	mutex_exit(&dev->l2ad_mtx);

	out:
	/*
	* We need to check if we evict all buffers, otherwise we may iterate
	* unnecessarily.
	*/
	if (!all && rerun) {
	/*
	* Bump device hand to the device start if it is approaching the
	* end. l2arc_evict() has already evicted ahead for this case.
	*/
	dev->l2ad_hand = dev->l2ad_start;
	dev->l2ad_evict = dev->l2ad_start;
	dev->l2ad_first = B_FALSE;
	goto top;
	}

	if (!all) {
	/*
	* In case of cache device removal (all) the following
	* assertions may be violated without functional consequences
	* as the device is about to be removed.
	*/
	- ASSERT3U(dev->l2ad_hand + distance, <, dev->l2ad_end);
	+ ASSERT3U(dev->l2ad_hand + distance, <=, dev->l2ad_end);
	if (!dev->l2ad_first)
	ASSERT3U(dev->l2ad_hand, <=, dev->l2ad_evict);
	}
	}

	/*
	* Handle any abd transforms that might be required for writing to the L2ARC.
	* If successful, this function will always return an abd with the data
	* transformed as it is on disk in a new abd of asize bytes.
	*/
	static int
	l2arc_apply_transforms(spa_t spa, arc_buf_hdr_t hdr, uint64_t asize,
	abd_t **abd_out)
	{
	int ret;
	abd_t cabd = NULL, eabd = NULL, *to_write = hdr->b_l1hdr.b_pabd;
	enum zio_compress compress = HDR_GET_COMPRESS(hdr);
	uint64_t psize = HDR_GET_PSIZE(hdr);
	uint64_t size = arc_hdr_size(hdr);
	boolean_t ismd = HDR_ISTYPE_METADATA(hdr);
	boolean_t bswap = (hdr->b_l1hdr.b_byteswap != DMU_BSWAP_NUMFUNCS);
	dsl_crypto_key_t *dck = NULL;
	uint8_t mac[ZIO_DATA_MAC_LEN] = { 0 };
	boolean_t no_crypt = B_FALSE;

	ASSERT((HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF &&
	!HDR_COMPRESSION_ENABLED(hdr)) \|\|
	HDR_ENCRYPTED(hdr) \|\| HDR_SHARED_DATA(hdr) \|\| psize != asize);
	ASSERT3U(psize, <=, asize);

	/*
	* If this data simply needs its own buffer, we simply allocate it
	* and copy the data. This may be done to eliminate a dependency on a
	* shared buffer or to reallocate the buffer to match asize.
	*/
	if (HDR_HAS_RABD(hdr)) {
	ASSERT3U(asize, >, psize);
	to_write = abd_alloc_for_io(asize, ismd);
	abd_copy(to_write, hdr->b_crypt_hdr.b_rabd, psize);
	abd_zero_off(to_write, psize, asize - psize);
	goto out;
	}

	if ((compress == ZIO_COMPRESS_OFF \|\| HDR_COMPRESSION_ENABLED(hdr)) &&
	!HDR_ENCRYPTED(hdr)) {
	ASSERT3U(size, ==, psize);
	to_write = abd_alloc_for_io(asize, ismd);
	abd_copy(to_write, hdr->b_l1hdr.b_pabd, size);
	if (asize > size)
	abd_zero_off(to_write, size, asize - size);
	goto out;
	}

	if (compress != ZIO_COMPRESS_OFF && !HDR_COMPRESSION_ENABLED(hdr)) {
	size_t bufsize = MAX(size, asize);
	void *buf = zio_buf_alloc(bufsize);
	uint64_t csize = zio_compress_data(compress, to_write, &buf,
	size, hdr->b_complevel);
	if (csize > psize) {
	/*
	* We can't re-compress the block into the original
	* psize. Even if it fits into asize, it does not
	* matter, since checksum will never match on read.
	*/
	zio_buf_free(buf, bufsize);
	return (SET_ERROR(EIO));
	}
	if (asize > csize)
	memset((char *)buf + csize, 0, asize - csize);
	to_write = cabd = abd_get_from_buf(buf, bufsize);
	abd_take_ownership_of_buf(cabd, B_TRUE);
	}

	if (HDR_ENCRYPTED(hdr)) {
	eabd = abd_alloc_for_io(asize, ismd);

	/*
	* If the dataset was disowned before the buffer
	* made it to this point, the key to re-encrypt
	* it won't be available. In this case we simply
	* won't write the buffer to the L2ARC.
	*/
	ret = spa_keystore_lookup_key(spa, hdr->b_crypt_hdr.b_dsobj,
	FTAG, &dck);
	if (ret != 0)
	goto error;

	ret = zio_do_crypt_abd(B_TRUE, &dck->dck_key,
	hdr->b_crypt_hdr.b_ot, bswap, hdr->b_crypt_hdr.b_salt,
	hdr->b_crypt_hdr.b_iv, mac, psize, to_write, eabd,
	&no_crypt);
	if (ret != 0)
	goto error;

	if (no_crypt)
	abd_copy(eabd, to_write, psize);

	if (psize != asize)
	abd_zero_off(eabd, psize, asize - psize);

	/* assert that the MAC we got here matches the one we saved */
	ASSERT0(memcmp(mac, hdr->b_crypt_hdr.b_mac, ZIO_DATA_MAC_LEN));
	spa_keystore_dsl_key_rele(spa, dck, FTAG);

	if (to_write == cabd)
	abd_free(cabd);

	to_write = eabd;
	}

	out:
	ASSERT3P(to_write, !=, hdr->b_l1hdr.b_pabd);
	*abd_out = to_write;
	return (0);

	error:
	if (dck != NULL)
	spa_keystore_dsl_key_rele(spa, dck, FTAG);
	if (cabd != NULL)
	abd_free(cabd);
	if (eabd != NULL)
	abd_free(eabd);

	*abd_out = NULL;
	return (ret);
	}

	static void
	l2arc_blk_fetch_done(zio_t *zio)
	{
	l2arc_read_callback_t *cb;

	cb = zio->io_private;
	if (cb->l2rcb_abd != NULL)
	abd_free(cb->l2rcb_abd);
	kmem_free(cb, sizeof (l2arc_read_callback_t));
	}

	/*
	* Find and write ARC buffers to the L2ARC device.
	*
	* An ARC_FLAG_L2_WRITING flag is set so that the L2ARC buffers are not valid
	* for reading until they have completed writing.
	* The headroom_boost is an in-out parameter used to maintain headroom boost
	* state between calls to this function.
	*
	* Returns the number of bytes actually written (which may be smaller than
	* the delta by which the device hand has changed due to alignment and the
	* writing of log blocks).
	*/
	static uint64_t
	l2arc_write_buffers(spa_t spa, l2arc_dev_t dev, uint64_t target_sz)
	{
	arc_buf_hdr_t hdr, head, *marker;
	uint64_t write_asize, write_psize, headroom;
	boolean_t full, from_head = !arc_warm;
	l2arc_write_callback_t *cb = NULL;
	zio_t pio, wzio;
	uint64_t guid = spa_load_guid(spa);
	l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr;

	ASSERT3P(dev->l2ad_vdev, !=, NULL);

	pio = NULL;
	write_asize = write_psize = 0;
	full = B_FALSE;
	head = kmem_cache_alloc(hdr_l2only_cache, KM_PUSHPAGE);
	arc_hdr_set_flags(head, ARC_FLAG_L2_WRITE_HEAD \| ARC_FLAG_HAS_L2HDR);
	marker = arc_state_alloc_marker();

	/*
	* Copy buffers for L2ARC writing.
	*/
	for (int pass = 0; pass < L2ARC_FEED_TYPES; pass++) {
	/*
	* If pass == 1 or 3, we cache MRU metadata and data
	* respectively.
	*/
	if (l2arc_mfuonly) {
	if (pass == 1 \|\| pass == 3)
	continue;
	}

	uint64_t passed_sz = 0;
	headroom = target_sz * l2arc_headroom;
	if (zfs_compressed_arc_enabled)
	headroom = (headroom * l2arc_headroom_boost) / 100;

	/*
	* Until the ARC is warm and starts to evict, read from the
	* head of the ARC lists rather than the tail.
	*/
	multilist_sublist_t *mls = l2arc_sublist_lock(pass);
	ASSERT3P(mls, !=, NULL);
	if (from_head)
	hdr = multilist_sublist_head(mls);
	else
	hdr = multilist_sublist_tail(mls);

	while (hdr != NULL) {
	kmutex_t *hash_lock;
	abd_t *to_write = NULL;

	hash_lock = HDR_LOCK(hdr);
	if (!mutex_tryenter(hash_lock)) {
	skip:
	/* Skip this buffer rather than waiting. */
	if (from_head)
	hdr = multilist_sublist_next(mls, hdr);
	else
	hdr = multilist_sublist_prev(mls, hdr);
	continue;
	}

	passed_sz += HDR_GET_LSIZE(hdr);
	if (l2arc_headroom != 0 && passed_sz > headroom) {
	/*
	* Searched too far.
	*/
	mutex_exit(hash_lock);
	break;
	}

	if (!l2arc_write_eligible(guid, hdr)) {
	mutex_exit(hash_lock);
	goto skip;
	}

	ASSERT(HDR_HAS_L1HDR(hdr));
	ASSERT3U(HDR_GET_PSIZE(hdr), >, 0);
	ASSERT3U(arc_hdr_size(hdr), >, 0);
	ASSERT(hdr->b_l1hdr.b_pabd != NULL \|\|
	HDR_HAS_RABD(hdr));
	uint64_t psize = HDR_GET_PSIZE(hdr);
	uint64_t asize = vdev_psize_to_asize(dev->l2ad_vdev,
	psize);

	/*
	* If the allocated size of this buffer plus the max
	* size for the pending log block exceeds the evicted
	* target size, terminate writing buffers for this run.
	*/
	if (write_asize + asize +
	sizeof (l2arc_log_blk_phys_t) > target_sz) {
	full = B_TRUE;
	mutex_exit(hash_lock);
	break;
	}

	/*
	* We should not sleep with sublist lock held or it
	* may block ARC eviction. Insert a marker to save
	* the position and drop the lock.
	*/
	if (from_head) {
	multilist_sublist_insert_after(mls, hdr,
	marker);
	} else {
	multilist_sublist_insert_before(mls, hdr,
	marker);
	}
	multilist_sublist_unlock(mls);

	/*
	* If this header has b_rabd, we can use this since it
	* must always match the data exactly as it exists on
	* disk. Otherwise, the L2ARC can normally use the
	* hdr's data, but if we're sharing data between the
	* hdr and one of its bufs, L2ARC needs its own copy of
	* the data so that the ZIO below can't race with the
	* buf consumer. To ensure that this copy will be
	* available for the lifetime of the ZIO and be cleaned
	* up afterwards, we add it to the l2arc_free_on_write
	* queue. If we need to apply any transforms to the
	* data (compression, encryption) we will also need the
	* extra buffer.
	*/
	if (HDR_HAS_RABD(hdr) && psize == asize) {
	to_write = hdr->b_crypt_hdr.b_rabd;
	} else if ((HDR_COMPRESSION_ENABLED(hdr) \|\|
	HDR_GET_COMPRESS(hdr) == ZIO_COMPRESS_OFF) &&
	!HDR_ENCRYPTED(hdr) && !HDR_SHARED_DATA(hdr) &&
	psize == asize) {
	to_write = hdr->b_l1hdr.b_pabd;
	} else {
	int ret;
	arc_buf_contents_t type = arc_buf_type(hdr);

	ret = l2arc_apply_transforms(spa, hdr, asize,
	&to_write);
	if (ret != 0) {
	arc_hdr_clear_flags(hdr,
	ARC_FLAG_L2CACHE);
	mutex_exit(hash_lock);
	goto next;
	}

	l2arc_free_abd_on_write(to_write, asize, type);
	}

	hdr->b_l2hdr.b_dev = dev;
	hdr->b_l2hdr.b_daddr = dev->l2ad_hand;
	hdr->b_l2hdr.b_hits = 0;
	hdr->b_l2hdr.b_arcs_state =
	hdr->b_l1hdr.b_state->arcs_state;
	mutex_enter(&dev->l2ad_mtx);
	if (pio == NULL) {
	/*
	* Insert a dummy header on the buflist so
	* l2arc_write_done() can find where the
	* write buffers begin without searching.
	*/
	list_insert_head(&dev->l2ad_buflist, head);
	}
	list_insert_head(&dev->l2ad_buflist, hdr);
	mutex_exit(&dev->l2ad_mtx);
	arc_hdr_set_flags(hdr, ARC_FLAG_HAS_L2HDR \|
	ARC_FLAG_L2_WRITING);

	(void) zfs_refcount_add_many(&dev->l2ad_alloc,
	arc_hdr_size(hdr), hdr);
	l2arc_hdr_arcstats_increment(hdr);

	boolean_t commit = l2arc_log_blk_insert(dev, hdr);
	mutex_exit(hash_lock);

	if (pio == NULL) {
	cb = kmem_alloc(
	sizeof (l2arc_write_callback_t), KM_SLEEP);
	cb->l2wcb_dev = dev;
	cb->l2wcb_head = head;
	list_create(&cb->l2wcb_abd_list,
	sizeof (l2arc_lb_abd_buf_t),
	offsetof(l2arc_lb_abd_buf_t, node));
	pio = zio_root(spa, l2arc_write_done, cb,
	ZIO_FLAG_CANFAIL);
	}

	wzio = zio_write_phys(pio, dev->l2ad_vdev,
	dev->l2ad_hand, asize, to_write,
	ZIO_CHECKSUM_OFF, NULL, hdr,
	ZIO_PRIORITY_ASYNC_WRITE,
	ZIO_FLAG_CANFAIL, B_FALSE);

	DTRACE_PROBE2(l2arc__write, vdev_t *, dev->l2ad_vdev,
	zio_t *, wzio);
	zio_nowait(wzio);

	write_psize += psize;
	write_asize += asize;
	dev->l2ad_hand += asize;
	vdev_space_update(dev->l2ad_vdev, asize, 0, 0);

	if (commit) {
	/* l2ad_hand will be adjusted inside. */
	write_asize +=
	l2arc_log_blk_commit(dev, pio, cb);
	}

	next:
	multilist_sublist_lock(mls);
	if (from_head)
	hdr = multilist_sublist_next(mls, marker);
	else
	hdr = multilist_sublist_prev(mls, marker);
	multilist_sublist_remove(mls, marker);
	}

	multilist_sublist_unlock(mls);

	if (full == B_TRUE)
	break;
	}

	arc_state_free_marker(marker);

	/* No buffers selected for writing? */
	if (pio == NULL) {
	ASSERT0(write_psize);
	ASSERT(!HDR_HAS_L1HDR(head));
	kmem_cache_free(hdr_l2only_cache, head);

	/*
	* Although we did not write any buffers l2ad_evict may
	* have advanced.
	*/
	if (dev->l2ad_evict != l2dhdr->dh_evict)
	l2arc_dev_hdr_update(dev);

	return (0);
	}

	if (!dev->l2ad_first)
	ASSERT3U(dev->l2ad_hand, <=, dev->l2ad_evict);

	ASSERT3U(write_asize, <=, target_sz);
	ARCSTAT_BUMP(arcstat_l2_writes_sent);
	ARCSTAT_INCR(arcstat_l2_write_bytes, write_psize);

	dev->l2ad_writing = B_TRUE;
	(void) zio_wait(pio);
	dev->l2ad_writing = B_FALSE;

	/*
	* Update the device header after the zio completes as
	* l2arc_write_done() may have updated the memory holding the log block
	* pointers in the device header.
	*/
	l2arc_dev_hdr_update(dev);

	return (write_asize);
	}

	static boolean_t
	l2arc_hdr_limit_reached(void)
	{
	int64_t s = aggsum_upper_bound(&arc_sums.arcstat_l2_hdr_size);

	return (arc_reclaim_needed() \|\|
	(s > (arc_warm ? arc_c : arc_c_max) * l2arc_meta_percent / 100));
	}

	/*
	* This thread feeds the L2ARC at regular intervals. This is the beating
	* heart of the L2ARC.
	*/
	static __attribute__((noreturn)) void
	l2arc_feed_thread(void *unused)
	{
	(void) unused;
	callb_cpr_t cpr;
	l2arc_dev_t *dev;
	spa_t *spa;
	uint64_t size, wrote;
	clock_t begin, next = ddi_get_lbolt();
	fstrans_cookie_t cookie;

	CALLB_CPR_INIT(&cpr, &l2arc_feed_thr_lock, callb_generic_cpr, FTAG);

	mutex_enter(&l2arc_feed_thr_lock);

	cookie = spl_fstrans_mark();
	while (l2arc_thread_exit == 0) {
	CALLB_CPR_SAFE_BEGIN(&cpr);
	(void) cv_timedwait_idle(&l2arc_feed_thr_cv,
	&l2arc_feed_thr_lock, next);
	CALLB_CPR_SAFE_END(&cpr, &l2arc_feed_thr_lock);
	next = ddi_get_lbolt() + hz;

	/*
	* Quick check for L2ARC devices.
	*/
	mutex_enter(&l2arc_dev_mtx);
	if (l2arc_ndev == 0) {
	mutex_exit(&l2arc_dev_mtx);
	continue;
	}
	mutex_exit(&l2arc_dev_mtx);
	begin = ddi_get_lbolt();

	/*
	* This selects the next l2arc device to write to, and in
	* doing so the next spa to feed from: dev->l2ad_spa. This
	* will return NULL if there are now no l2arc devices or if
	* they are all faulted.
	*
	* If a device is returned, its spa's config lock is also
	* held to prevent device removal. l2arc_dev_get_next()
	* will grab and release l2arc_dev_mtx.
	*/
	if ((dev = l2arc_dev_get_next()) == NULL)
	continue;

	spa = dev->l2ad_spa;
	ASSERT3P(spa, !=, NULL);

	/*
	* If the pool is read-only then force the feed thread to
	* sleep a little longer.
	*/
	if (!spa_writeable(spa)) {
	next = ddi_get_lbolt() + 5 * l2arc_feed_secs * hz;
	spa_config_exit(spa, SCL_L2ARC, dev);
	continue;
	}

	/*
	* Avoid contributing to memory pressure.
	*/
	if (l2arc_hdr_limit_reached()) {
	ARCSTAT_BUMP(arcstat_l2_abort_lowmem);
	spa_config_exit(spa, SCL_L2ARC, dev);
	continue;
	}

	ARCSTAT_BUMP(arcstat_l2_feeds);

	size = l2arc_write_size(dev);

	/*
	* Evict L2ARC buffers that will be overwritten.
	*/
	l2arc_evict(dev, size, B_FALSE);

	/*
	* Write ARC buffers.
	*/
	wrote = l2arc_write_buffers(spa, dev, size);

	/*
	* Calculate interval between writes.
	*/
	next = l2arc_write_interval(begin, size, wrote);
	spa_config_exit(spa, SCL_L2ARC, dev);
	}
	spl_fstrans_unmark(cookie);

	l2arc_thread_exit = 0;
	cv_broadcast(&l2arc_feed_thr_cv);
	CALLB_CPR_EXIT(&cpr); /* drops l2arc_feed_thr_lock */
	thread_exit();
	}

	boolean_t
	l2arc_vdev_present(vdev_t *vd)
	{
	return (l2arc_vdev_get(vd) != NULL);
	}

	/*
	* Returns the l2arc_dev_t associated with a particular vdev_t or NULL if
	* the vdev_t isn't an L2ARC device.
	*/
	l2arc_dev_t *
	l2arc_vdev_get(vdev_t *vd)
	{
	l2arc_dev_t *dev;

	mutex_enter(&l2arc_dev_mtx);
	for (dev = list_head(l2arc_dev_list); dev != NULL;
	dev = list_next(l2arc_dev_list, dev)) {
	if (dev->l2ad_vdev == vd)
	break;
	}
	mutex_exit(&l2arc_dev_mtx);

	return (dev);
	}

	static void
	l2arc_rebuild_dev(l2arc_dev_t *dev, boolean_t reopen)
	{
	l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr;
	uint64_t l2dhdr_asize = dev->l2ad_dev_hdr_asize;
	spa_t *spa = dev->l2ad_spa;

	/*
	* The L2ARC has to hold at least the payload of one log block for
	* them to be restored (persistent L2ARC). The payload of a log block
	* depends on the amount of its log entries. We always write log blocks
	* with 1022 entries. How many of them are committed or restored depends
	* on the size of the L2ARC device. Thus the maximum payload of
	* one log block is 1022 * SPA_MAXBLOCKSIZE = 16GB. If the L2ARC device
	* is less than that, we reduce the amount of committed and restored
	* log entries per block so as to enable persistence.
	*/
	if (dev->l2ad_end < l2arc_rebuild_blocks_min_l2size) {
	dev->l2ad_log_entries = 0;
	} else {
	dev->l2ad_log_entries = MIN((dev->l2ad_end -
	dev->l2ad_start) >> SPA_MAXBLOCKSHIFT,
	L2ARC_LOG_BLK_MAX_ENTRIES);
	}

	/*
	* Read the device header, if an error is returned do not rebuild L2ARC.
	*/
	if (l2arc_dev_hdr_read(dev) == 0 && dev->l2ad_log_entries > 0) {
	/*
	* If we are onlining a cache device (vdev_reopen) that was
	* still present (l2arc_vdev_present()) and rebuild is enabled,
	* we should evict all ARC buffers and pointers to log blocks
	* and reclaim their space before restoring its contents to
	* L2ARC.
	*/
	if (reopen) {
	if (!l2arc_rebuild_enabled) {
	return;
	} else {
	l2arc_evict(dev, 0, B_TRUE);
	/* start a new log block */
	dev->l2ad_log_ent_idx = 0;
	dev->l2ad_log_blk_payload_asize = 0;
	dev->l2ad_log_blk_payload_start = 0;
	}
	}
	/*
	* Just mark the device as pending for a rebuild. We won't
	* be starting a rebuild in line here as it would block pool
	* import. Instead spa_load_impl will hand that off to an
	* async task which will call l2arc_spa_rebuild_start.
	*/
	dev->l2ad_rebuild = B_TRUE;
	} else if (spa_writeable(spa)) {
	/*
	* In this case TRIM the whole device if l2arc_trim_ahead > 0,
	* otherwise create a new header. We zero out the memory holding
	* the header to reset dh_start_lbps. If we TRIM the whole
	* device the new header will be written by
	* vdev_trim_l2arc_thread() at the end of the TRIM to update the
	* trim_state in the header too. When reading the header, if
	* trim_state is not VDEV_TRIM_COMPLETE and l2arc_trim_ahead > 0
	* we opt to TRIM the whole device again.
	*/
	if (l2arc_trim_ahead > 0) {
	dev->l2ad_trim_all = B_TRUE;
	} else {
	memset(l2dhdr, 0, l2dhdr_asize);
	l2arc_dev_hdr_update(dev);
	}
	}
	}

	/*
	* Add a vdev for use by the L2ARC. By this point the spa has already
	* validated the vdev and opened it.
	*/
	void
	l2arc_add_vdev(spa_t spa, vdev_t vd)
	{
	l2arc_dev_t *adddev;
	uint64_t l2dhdr_asize;

	ASSERT(!l2arc_vdev_present(vd));

	/*
	* Create a new l2arc device entry.
	*/
	adddev = vmem_zalloc(sizeof (l2arc_dev_t), KM_SLEEP);
	adddev->l2ad_spa = spa;
	adddev->l2ad_vdev = vd;
	/* leave extra size for an l2arc device header */
	l2dhdr_asize = adddev->l2ad_dev_hdr_asize =
	MAX(sizeof (*adddev->l2ad_dev_hdr), 1 << vd->vdev_ashift);
	adddev->l2ad_start = VDEV_LABEL_START_SIZE + l2dhdr_asize;
	adddev->l2ad_end = VDEV_LABEL_START_SIZE + vdev_get_min_asize(vd);
	ASSERT3U(adddev->l2ad_start, <, adddev->l2ad_end);
	adddev->l2ad_hand = adddev->l2ad_start;
	adddev->l2ad_evict = adddev->l2ad_start;
	adddev->l2ad_first = B_TRUE;
	adddev->l2ad_writing = B_FALSE;
	adddev->l2ad_trim_all = B_FALSE;
	list_link_init(&adddev->l2ad_node);
	adddev->l2ad_dev_hdr = kmem_zalloc(l2dhdr_asize, KM_SLEEP);

	mutex_init(&adddev->l2ad_mtx, NULL, MUTEX_DEFAULT, NULL);
	/*
	* This is a list of all ARC buffers that are still valid on the
	* device.
	*/
	list_create(&adddev->l2ad_buflist, sizeof (arc_buf_hdr_t),
	offsetof(arc_buf_hdr_t, b_l2hdr.b_l2node));

	/*
	* This is a list of pointers to log blocks that are still present
	* on the device.
	*/
	list_create(&adddev->l2ad_lbptr_list, sizeof (l2arc_lb_ptr_buf_t),
	offsetof(l2arc_lb_ptr_buf_t, node));

	vdev_space_update(vd, 0, 0, adddev->l2ad_end - adddev->l2ad_hand);
	zfs_refcount_create(&adddev->l2ad_alloc);
	zfs_refcount_create(&adddev->l2ad_lb_asize);
	zfs_refcount_create(&adddev->l2ad_lb_count);

	/*
	* Decide if dev is eligible for L2ARC rebuild or whole device
	* trimming. This has to happen before the device is added in the
	* cache device list and l2arc_dev_mtx is released. Otherwise
	* l2arc_feed_thread() might already start writing on the
	* device.
	*/
	l2arc_rebuild_dev(adddev, B_FALSE);

	/*
	* Add device to global list
	*/
	mutex_enter(&l2arc_dev_mtx);
	list_insert_head(l2arc_dev_list, adddev);
	atomic_inc_64(&l2arc_ndev);
	mutex_exit(&l2arc_dev_mtx);
	}

	/*
	* Decide if a vdev is eligible for L2ARC rebuild, called from vdev_reopen()
	* in case of onlining a cache device.
	*/
	void
	l2arc_rebuild_vdev(vdev_t *vd, boolean_t reopen)
	{
	l2arc_dev_t *dev = NULL;

	dev = l2arc_vdev_get(vd);
	ASSERT3P(dev, !=, NULL);

	/*
	* In contrast to l2arc_add_vdev() we do not have to worry about
	* l2arc_feed_thread() invalidating previous content when onlining a
	* cache device. The device parameters (l2ad*) are not cleared when
	* offlining the device and writing new buffers will not invalidate
	* all previous content. In worst case only buffers that have not had
	* their log block written to the device will be lost.
	* When onlining the cache device (ie offline->online without exporting
	* the pool in between) this happens:
	* vdev_reopen() -> vdev_open() -> l2arc_rebuild_vdev()
	* \| \|
	* vdev_is_dead() = B_FALSE l2ad_rebuild = B_TRUE
	* During the time where vdev_is_dead = B_FALSE and until l2ad_rebuild
	* is set to B_TRUE we might write additional buffers to the device.
	*/
	l2arc_rebuild_dev(dev, reopen);
	}

	/*
	* Remove a vdev from the L2ARC.
	*/
	void
	l2arc_remove_vdev(vdev_t *vd)
	{
	l2arc_dev_t *remdev = NULL;

	/*
	* Find the device by vdev
	*/
	remdev = l2arc_vdev_get(vd);
	ASSERT3P(remdev, !=, NULL);

	/*
	* Cancel any ongoing or scheduled rebuild.
	*/
	mutex_enter(&l2arc_rebuild_thr_lock);
	if (remdev->l2ad_rebuild_began == B_TRUE) {
	remdev->l2ad_rebuild_cancel = B_TRUE;
	while (remdev->l2ad_rebuild == B_TRUE)
	cv_wait(&l2arc_rebuild_thr_cv, &l2arc_rebuild_thr_lock);
	}
	mutex_exit(&l2arc_rebuild_thr_lock);

	/*
	* Remove device from global list
	*/
	mutex_enter(&l2arc_dev_mtx);
	list_remove(l2arc_dev_list, remdev);
	l2arc_dev_last = NULL; /* may have been invalidated */
	atomic_dec_64(&l2arc_ndev);
	mutex_exit(&l2arc_dev_mtx);

	/*
	* Clear all buflists and ARC references. L2ARC device flush.
	*/
	l2arc_evict(remdev, 0, B_TRUE);
	list_destroy(&remdev->l2ad_buflist);
	ASSERT(list_is_empty(&remdev->l2ad_lbptr_list));
	list_destroy(&remdev->l2ad_lbptr_list);
	mutex_destroy(&remdev->l2ad_mtx);
	zfs_refcount_destroy(&remdev->l2ad_alloc);
	zfs_refcount_destroy(&remdev->l2ad_lb_asize);
	zfs_refcount_destroy(&remdev->l2ad_lb_count);
	kmem_free(remdev->l2ad_dev_hdr, remdev->l2ad_dev_hdr_asize);
	vmem_free(remdev, sizeof (l2arc_dev_t));
	}

	void
	l2arc_init(void)
	{
	l2arc_thread_exit = 0;
	l2arc_ndev = 0;

	mutex_init(&l2arc_feed_thr_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&l2arc_feed_thr_cv, NULL, CV_DEFAULT, NULL);
	mutex_init(&l2arc_rebuild_thr_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&l2arc_rebuild_thr_cv, NULL, CV_DEFAULT, NULL);
	mutex_init(&l2arc_dev_mtx, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&l2arc_free_on_write_mtx, NULL, MUTEX_DEFAULT, NULL);

	l2arc_dev_list = &L2ARC_dev_list;
	l2arc_free_on_write = &L2ARC_free_on_write;
	list_create(l2arc_dev_list, sizeof (l2arc_dev_t),
	offsetof(l2arc_dev_t, l2ad_node));
	list_create(l2arc_free_on_write, sizeof (l2arc_data_free_t),
	offsetof(l2arc_data_free_t, l2df_list_node));
	}

	void
	l2arc_fini(void)
	{
	mutex_destroy(&l2arc_feed_thr_lock);
	cv_destroy(&l2arc_feed_thr_cv);
	mutex_destroy(&l2arc_rebuild_thr_lock);
	cv_destroy(&l2arc_rebuild_thr_cv);
	mutex_destroy(&l2arc_dev_mtx);
	mutex_destroy(&l2arc_free_on_write_mtx);

	list_destroy(l2arc_dev_list);
	list_destroy(l2arc_free_on_write);
	}

	void
	l2arc_start(void)
	{
	if (!(spa_mode_global & SPA_MODE_WRITE))
	return;

	(void) thread_create(NULL, 0, l2arc_feed_thread, NULL, 0, &p0,
	TS_RUN, defclsyspri);
	}

	void
	l2arc_stop(void)
	{
	if (!(spa_mode_global & SPA_MODE_WRITE))
	return;

	mutex_enter(&l2arc_feed_thr_lock);
	cv_signal(&l2arc_feed_thr_cv); /* kick thread out of startup */
	l2arc_thread_exit = 1;
	while (l2arc_thread_exit != 0)
	cv_wait(&l2arc_feed_thr_cv, &l2arc_feed_thr_lock);
	mutex_exit(&l2arc_feed_thr_lock);
	}

	/*
	* Punches out rebuild threads for the L2ARC devices in a spa. This should
	* be called after pool import from the spa async thread, since starting
	* these threads directly from spa_import() will make them part of the
	* "zpool import" context and delay process exit (and thus pool import).
	*/
	void
	l2arc_spa_rebuild_start(spa_t *spa)
	{
	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	/*
	* Locate the spa's l2arc devices and kick off rebuild threads.
	*/
	for (int i = 0; i < spa->spa_l2cache.sav_count; i++) {
	l2arc_dev_t *dev =
	l2arc_vdev_get(spa->spa_l2cache.sav_vdevs[i]);
	if (dev == NULL) {
	/* Don't attempt a rebuild if the vdev is UNAVAIL */
	continue;
	}
	mutex_enter(&l2arc_rebuild_thr_lock);
	if (dev->l2ad_rebuild && !dev->l2ad_rebuild_cancel) {
	dev->l2ad_rebuild_began = B_TRUE;
	(void) thread_create(NULL, 0, l2arc_dev_rebuild_thread,
	dev, 0, &p0, TS_RUN, minclsyspri);
	}
	mutex_exit(&l2arc_rebuild_thr_lock);
	}
	}

	/*
	* Main entry point for L2ARC rebuilding.
	*/
	static __attribute__((noreturn)) void
	l2arc_dev_rebuild_thread(void *arg)
	{
	l2arc_dev_t *dev = arg;

	VERIFY(!dev->l2ad_rebuild_cancel);
	VERIFY(dev->l2ad_rebuild);
	(void) l2arc_rebuild(dev);
	mutex_enter(&l2arc_rebuild_thr_lock);
	dev->l2ad_rebuild_began = B_FALSE;
	dev->l2ad_rebuild = B_FALSE;
	mutex_exit(&l2arc_rebuild_thr_lock);

	thread_exit();
	}

	/*
	* This function implements the actual L2ARC metadata rebuild. It:
	* starts reading the log block chain and restores each block's contents
	* to memory (reconstructing arc_buf_hdr_t's).
	*
	* Operation stops under any of the following conditions:
	*
	* 1) We reach the end of the log block chain.
	* 2) We encounter any error condition (cksum errors, io errors)
	*/
	static int
	l2arc_rebuild(l2arc_dev_t *dev)
	{
	vdev_t *vd = dev->l2ad_vdev;
	spa_t *spa = vd->vdev_spa;
	int err = 0;
	l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr;
	l2arc_log_blk_phys_t this_lb, next_lb;
	zio_t this_io = NULL, next_io = NULL;
	l2arc_log_blkptr_t lbps[2];
	l2arc_lb_ptr_buf_t *lb_ptr_buf;
	boolean_t lock_held;

	this_lb = vmem_zalloc(sizeof (*this_lb), KM_SLEEP);
	next_lb = vmem_zalloc(sizeof (*next_lb), KM_SLEEP);

	/*
	* We prevent device removal while issuing reads to the device,
	* then during the rebuilding phases we drop this lock again so
	* that a spa_unload or device remove can be initiated - this is
	* safe, because the spa will signal us to stop before removing
	* our device and wait for us to stop.
	*/
	spa_config_enter(spa, SCL_L2ARC, vd, RW_READER);
	lock_held = B_TRUE;

	/*
	* Retrieve the persistent L2ARC device state.
	* L2BLK_GET_PSIZE returns aligned size for log blocks.
	*/
	dev->l2ad_evict = MAX(l2dhdr->dh_evict, dev->l2ad_start);
	dev->l2ad_hand = MAX(l2dhdr->dh_start_lbps[0].lbp_daddr +
	L2BLK_GET_PSIZE((&l2dhdr->dh_start_lbps[0])->lbp_prop),
	dev->l2ad_start);
	dev->l2ad_first = !!(l2dhdr->dh_flags & L2ARC_DEV_HDR_EVICT_FIRST);

	vd->vdev_trim_action_time = l2dhdr->dh_trim_action_time;
	vd->vdev_trim_state = l2dhdr->dh_trim_state;

	/*
	* In case the zfs module parameter l2arc_rebuild_enabled is false
	* we do not start the rebuild process.
	*/
	if (!l2arc_rebuild_enabled)
	goto out;

	/* Prepare the rebuild process */
	memcpy(lbps, l2dhdr->dh_start_lbps, sizeof (lbps));

	/* Start the rebuild process */
	for (;;) {
	if (!l2arc_log_blkptr_valid(dev, &lbps[0]))
	break;

	if ((err = l2arc_log_blk_read(dev, &lbps[0], &lbps[1],
	this_lb, next_lb, this_io, &next_io)) != 0)
	goto out;

	/*
	* Our memory pressure valve. If the system is running low
	* on memory, rather than swamping memory with new ARC buf
	* hdrs, we opt not to rebuild the L2ARC. At this point,
	* however, we have already set up our L2ARC dev to chain in
	* new metadata log blocks, so the user may choose to offline/
	* online the L2ARC dev at a later time (or re-import the pool)
	* to reconstruct it (when there's less memory pressure).
	*/
	if (l2arc_hdr_limit_reached()) {
	ARCSTAT_BUMP(arcstat_l2_rebuild_abort_lowmem);
	cmn_err(CE_NOTE, "System running low on memory, "
	"aborting L2ARC rebuild.");
	err = SET_ERROR(ENOMEM);
	goto out;
	}

	spa_config_exit(spa, SCL_L2ARC, vd);
	lock_held = B_FALSE;

	/*
	* Now that we know that the next_lb checks out alright, we
	* can start reconstruction from this log block.
	* L2BLK_GET_PSIZE returns aligned size for log blocks.
	*/
	uint64_t asize = L2BLK_GET_PSIZE((&lbps[0])->lbp_prop);
	l2arc_log_blk_restore(dev, this_lb, asize);

	/*
	* log block restored, include its pointer in the list of
	* pointers to log blocks present in the L2ARC device.
	*/
	lb_ptr_buf = kmem_zalloc(sizeof (l2arc_lb_ptr_buf_t), KM_SLEEP);
	lb_ptr_buf->lb_ptr = kmem_zalloc(sizeof (l2arc_log_blkptr_t),
	KM_SLEEP);
	memcpy(lb_ptr_buf->lb_ptr, &lbps[0],
	sizeof (l2arc_log_blkptr_t));
	mutex_enter(&dev->l2ad_mtx);
	list_insert_tail(&dev->l2ad_lbptr_list, lb_ptr_buf);
	ARCSTAT_INCR(arcstat_l2_log_blk_asize, asize);
	ARCSTAT_BUMP(arcstat_l2_log_blk_count);
	zfs_refcount_add_many(&dev->l2ad_lb_asize, asize, lb_ptr_buf);
	zfs_refcount_add(&dev->l2ad_lb_count, lb_ptr_buf);
	mutex_exit(&dev->l2ad_mtx);
	vdev_space_update(vd, asize, 0, 0);

	/*
	* Protection against loops of log blocks:
	*
	* l2ad_hand l2ad_evict
	* V V
	* l2ad_start \|=======================================\| l2ad_end
	* -----\|\|\|----\|\|\|---\|\|\|----\|\|\|
	* (3) (2) (1) (0)
	* ---\|\|\|---\|\|\|----\|\|\|---\|\|\|
	* (7) (6) (5) (4)
	*
	* In this situation the pointer of log block (4) passes
	* l2arc_log_blkptr_valid() but the log block should not be
	* restored as it is overwritten by the payload of log block
	* (0). Only log blocks (0)-(3) should be restored. We check
	* whether l2ad_evict lies in between the payload starting
	* offset of the next log block (lbps[1].lbp_payload_start)
	* and the payload starting offset of the present log block
	* (lbps[0].lbp_payload_start). If true and this isn't the
	* first pass, we are looping from the beginning and we should
	* stop.
	*/
	if (l2arc_range_check_overlap(lbps[1].lbp_payload_start,
	lbps[0].lbp_payload_start, dev->l2ad_evict) &&
	!dev->l2ad_first)
	goto out;

	kpreempt(KPREEMPT_SYNC);
	for (;;) {
	mutex_enter(&l2arc_rebuild_thr_lock);
	if (dev->l2ad_rebuild_cancel) {
	dev->l2ad_rebuild = B_FALSE;
	cv_signal(&l2arc_rebuild_thr_cv);
	mutex_exit(&l2arc_rebuild_thr_lock);
	err = SET_ERROR(ECANCELED);
	goto out;
	}
	mutex_exit(&l2arc_rebuild_thr_lock);
	if (spa_config_tryenter(spa, SCL_L2ARC, vd,
	RW_READER)) {
	lock_held = B_TRUE;
	break;
	}
	/*
	* L2ARC config lock held by somebody in writer,
	* possibly due to them trying to remove us. They'll
	* likely to want us to shut down, so after a little
	* delay, we check l2ad_rebuild_cancel and retry
	* the lock again.
	*/
	delay(1);
	}

	/*
	* Continue with the next log block.
	*/
	lbps[0] = lbps[1];
	lbps[1] = this_lb->lb_prev_lbp;
	PTR_SWAP(this_lb, next_lb);
	this_io = next_io;
	next_io = NULL;
	}

	if (this_io != NULL)
	l2arc_log_blk_fetch_abort(this_io);
	out:
	if (next_io != NULL)
	l2arc_log_blk_fetch_abort(next_io);
	vmem_free(this_lb, sizeof (*this_lb));
	vmem_free(next_lb, sizeof (*next_lb));

	if (!l2arc_rebuild_enabled) {
	spa_history_log_internal(spa, "L2ARC rebuild", NULL,
	"disabled");
	} else if (err == 0 && zfs_refcount_count(&dev->l2ad_lb_count) > 0) {
	ARCSTAT_BUMP(arcstat_l2_rebuild_success);
	spa_history_log_internal(spa, "L2ARC rebuild", NULL,
	"successful, restored %llu blocks",
	(u_longlong_t)zfs_refcount_count(&dev->l2ad_lb_count));
	} else if (err == 0 && zfs_refcount_count(&dev->l2ad_lb_count) == 0) {
	/*
	* No error but also nothing restored, meaning the lbps array
	* in the device header points to invalid/non-present log
	* blocks. Reset the header.
	*/
	spa_history_log_internal(spa, "L2ARC rebuild", NULL,
	"no valid log blocks");
	memset(l2dhdr, 0, dev->l2ad_dev_hdr_asize);
	l2arc_dev_hdr_update(dev);
	} else if (err == ECANCELED) {
	/*
	* In case the rebuild was canceled do not log to spa history
	* log as the pool may be in the process of being removed.
	*/
	zfs_dbgmsg("L2ARC rebuild aborted, restored %llu blocks",
	(u_longlong_t)zfs_refcount_count(&dev->l2ad_lb_count));
	} else if (err != 0) {
	spa_history_log_internal(spa, "L2ARC rebuild", NULL,
	"aborted, restored %llu blocks",
	(u_longlong_t)zfs_refcount_count(&dev->l2ad_lb_count));
	}

	if (lock_held)
	spa_config_exit(spa, SCL_L2ARC, vd);

	return (err);
	}

	/*
	* Attempts to read the device header on the provided L2ARC device and writes
	* it to `hdr'. On success, this function returns 0, otherwise the appropriate
	* error code is returned.
	*/
	static int
	l2arc_dev_hdr_read(l2arc_dev_t *dev)
	{
	int err;
	uint64_t guid;
	l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr;
	const uint64_t l2dhdr_asize = dev->l2ad_dev_hdr_asize;
	abd_t *abd;

	guid = spa_guid(dev->l2ad_vdev->vdev_spa);

	abd = abd_get_from_buf(l2dhdr, l2dhdr_asize);

	err = zio_wait(zio_read_phys(NULL, dev->l2ad_vdev,
	VDEV_LABEL_START_SIZE, l2dhdr_asize, abd,
	ZIO_CHECKSUM_LABEL, NULL, NULL, ZIO_PRIORITY_SYNC_READ,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_DONT_PROPAGATE \| ZIO_FLAG_DONT_RETRY \|
	ZIO_FLAG_SPECULATIVE, B_FALSE));

	abd_free(abd);

	if (err != 0) {
	ARCSTAT_BUMP(arcstat_l2_rebuild_abort_dh_errors);
	zfs_dbgmsg("L2ARC IO error (%d) while reading device header, "
	"vdev guid: %llu", err,
	(u_longlong_t)dev->l2ad_vdev->vdev_guid);
	return (err);
	}

	if (l2dhdr->dh_magic == BSWAP_64(L2ARC_DEV_HDR_MAGIC))
	byteswap_uint64_array(l2dhdr, sizeof (*l2dhdr));

	if (l2dhdr->dh_magic != L2ARC_DEV_HDR_MAGIC \|\|
	l2dhdr->dh_spa_guid != guid \|\|
	l2dhdr->dh_vdev_guid != dev->l2ad_vdev->vdev_guid \|\|
	l2dhdr->dh_version != L2ARC_PERSISTENT_VERSION \|\|
	l2dhdr->dh_log_entries != dev->l2ad_log_entries \|\|
	l2dhdr->dh_end != dev->l2ad_end \|\|
	!l2arc_range_check_overlap(dev->l2ad_start, dev->l2ad_end,
	l2dhdr->dh_evict) \|\|
	(l2dhdr->dh_trim_state != VDEV_TRIM_COMPLETE &&
	l2arc_trim_ahead > 0)) {
	/*
	* Attempt to rebuild a device containing no actual dev hdr
	* or containing a header from some other pool or from another
	* version of persistent L2ARC.
	*/
	ARCSTAT_BUMP(arcstat_l2_rebuild_abort_unsupported);
	return (SET_ERROR(ENOTSUP));
	}

	return (0);
	}

	/*
	* Reads L2ARC log blocks from storage and validates their contents.
	*
	* This function implements a simple fetcher to make sure that while
	* we're processing one buffer the L2ARC is already fetching the next
	* one in the chain.
	*
	* The arguments this_lp and next_lp point to the current and next log block
	* address in the block chain. Similarly, this_lb and next_lb hold the
	* l2arc_log_blk_phys_t's of the current and next L2ARC blk.
	*
	* The `this_io' and `next_io' arguments are used for block fetching.
	* When issuing the first blk IO during rebuild, you should pass NULL for
	* `this_io'. This function will then issue a sync IO to read the block and
	* also issue an async IO to fetch the next block in the block chain. The
	* fetched IO is returned in `next_io'. On subsequent calls to this
	* function, pass the value returned in `next_io' from the previous call
	* as `this_io' and a fresh `next_io' pointer to hold the next fetch IO.
	* Prior to the call, you should initialize your `next_io' pointer to be
	* NULL. If no fetch IO was issued, the pointer is left set at NULL.
	*
	* On success, this function returns 0, otherwise it returns an appropriate
	* error code. On error the fetching IO is aborted and cleared before
	* returning from this function. Therefore, if we return `success', the
	* caller can assume that we have taken care of cleanup of fetch IOs.
	*/
	static int
	l2arc_log_blk_read(l2arc_dev_t *dev,
	const l2arc_log_blkptr_t this_lbp, const l2arc_log_blkptr_t next_lbp,
	l2arc_log_blk_phys_t this_lb, l2arc_log_blk_phys_t next_lb,
	zio_t this_io, zio_t *next_io)
	{
	int err = 0;
	zio_cksum_t cksum;
	abd_t *abd = NULL;
	uint64_t asize;

	ASSERT(this_lbp != NULL && next_lbp != NULL);
	ASSERT(this_lb != NULL && next_lb != NULL);
	ASSERT(next_io != NULL && *next_io == NULL);
	ASSERT(l2arc_log_blkptr_valid(dev, this_lbp));

	/*
	* Check to see if we have issued the IO for this log block in a
	* previous run. If not, this is the first call, so issue it now.
	*/
	if (this_io == NULL) {
	this_io = l2arc_log_blk_fetch(dev->l2ad_vdev, this_lbp,
	this_lb);
	}

	/*
	* Peek to see if we can start issuing the next IO immediately.
	*/
	if (l2arc_log_blkptr_valid(dev, next_lbp)) {
	/*
	* Start issuing IO for the next log block early - this
	* should help keep the L2ARC device busy while we
	* decompress and restore this log block.
	*/
	*next_io = l2arc_log_blk_fetch(dev->l2ad_vdev, next_lbp,
	next_lb);
	}

	/* Wait for the IO to read this log block to complete */
	if ((err = zio_wait(this_io)) != 0) {
	ARCSTAT_BUMP(arcstat_l2_rebuild_abort_io_errors);
	zfs_dbgmsg("L2ARC IO error (%d) while reading log block, "
	"offset: %llu, vdev guid: %llu", err,
	(u_longlong_t)this_lbp->lbp_daddr,
	(u_longlong_t)dev->l2ad_vdev->vdev_guid);
	goto cleanup;
	}

	/*
	* Make sure the buffer checks out.
	* L2BLK_GET_PSIZE returns aligned size for log blocks.
	*/
	asize = L2BLK_GET_PSIZE((this_lbp)->lbp_prop);
	fletcher_4_native(this_lb, asize, NULL, &cksum);
	if (!ZIO_CHECKSUM_EQUAL(cksum, this_lbp->lbp_cksum)) {
	ARCSTAT_BUMP(arcstat_l2_rebuild_abort_cksum_lb_errors);
	zfs_dbgmsg("L2ARC log block cksum failed, offset: %llu, "
	"vdev guid: %llu, l2ad_hand: %llu, l2ad_evict: %llu",
	(u_longlong_t)this_lbp->lbp_daddr,
	(u_longlong_t)dev->l2ad_vdev->vdev_guid,
	(u_longlong_t)dev->l2ad_hand,
	(u_longlong_t)dev->l2ad_evict);
	err = SET_ERROR(ECKSUM);
	goto cleanup;
	}

	/* Now we can take our time decoding this buffer */
	switch (L2BLK_GET_COMPRESS((this_lbp)->lbp_prop)) {
	case ZIO_COMPRESS_OFF:
	break;
	case ZIO_COMPRESS_LZ4:
	abd = abd_alloc_for_io(asize, B_TRUE);
	abd_copy_from_buf_off(abd, this_lb, 0, asize);
	if ((err = zio_decompress_data(
	L2BLK_GET_COMPRESS((this_lbp)->lbp_prop),
	abd, this_lb, asize, sizeof (*this_lb), NULL)) != 0) {
	err = SET_ERROR(EINVAL);
	goto cleanup;
	}
	break;
	default:
	err = SET_ERROR(EINVAL);
	goto cleanup;
	}
	if (this_lb->lb_magic == BSWAP_64(L2ARC_LOG_BLK_MAGIC))
	byteswap_uint64_array(this_lb, sizeof (*this_lb));
	if (this_lb->lb_magic != L2ARC_LOG_BLK_MAGIC) {
	err = SET_ERROR(EINVAL);
	goto cleanup;
	}
	cleanup:
	/* Abort an in-flight fetch I/O in case of error */
	if (err != 0 && *next_io != NULL) {
	l2arc_log_blk_fetch_abort(*next_io);
	*next_io = NULL;
	}
	if (abd != NULL)
	abd_free(abd);
	return (err);
	}

	/*
	* Restores the payload of a log block to ARC. This creates empty ARC hdr
	* entries which only contain an l2arc hdr, essentially restoring the
	* buffers to their L2ARC evicted state. This function also updates space
	* usage on the L2ARC vdev to make sure it tracks restored buffers.
	*/
	static void
	l2arc_log_blk_restore(l2arc_dev_t dev, const l2arc_log_blk_phys_t lb,
	uint64_t lb_asize)
	{
	uint64_t size = 0, asize = 0;
	uint64_t log_entries = dev->l2ad_log_entries;

	/*
	* Usually arc_adapt() is called only for data, not headers, but
	* since we may allocate significant amount of memory here, let ARC
	* grow its arc_c.
	*/
	arc_adapt(log_entries * HDR_L2ONLY_SIZE);

	for (int i = log_entries - 1; i >= 0; i--) {
	/*
	* Restore goes in the reverse temporal direction to preserve
	* correct temporal ordering of buffers in the l2ad_buflist.
	* l2arc_hdr_restore also does a list_insert_tail instead of
	* list_insert_head on the l2ad_buflist:
	*
	* LIST l2ad_buflist LIST
	* HEAD <------ (time) ------ TAIL
	* direction +-----+-----+-----+-----+-----+ direction
	* of l2arc <== \| buf \| buf \| buf \| buf \| buf \| ===> of rebuild
	* fill +-----+-----+-----+-----+-----+
	* ^ ^
	* \| \|
	* \| \|
	* l2arc_feed_thread l2arc_rebuild
	* will place new bufs here restores bufs here
	*
	* During l2arc_rebuild() the device is not used by
	* l2arc_feed_thread() as dev->l2ad_rebuild is set to true.
	*/
	size += L2BLK_GET_LSIZE((&lb->lb_entries[i])->le_prop);
	asize += vdev_psize_to_asize(dev->l2ad_vdev,
	L2BLK_GET_PSIZE((&lb->lb_entries[i])->le_prop));
	l2arc_hdr_restore(&lb->lb_entries[i], dev);
	}

	/*
	* Record rebuild stats:
	* size Logical size of restored buffers in the L2ARC
	* asize Aligned size of restored buffers in the L2ARC
	*/
	ARCSTAT_INCR(arcstat_l2_rebuild_size, size);
	ARCSTAT_INCR(arcstat_l2_rebuild_asize, asize);
	ARCSTAT_INCR(arcstat_l2_rebuild_bufs, log_entries);
	ARCSTAT_F_AVG(arcstat_l2_log_blk_avg_asize, lb_asize);
	ARCSTAT_F_AVG(arcstat_l2_data_to_meta_ratio, asize / lb_asize);
	ARCSTAT_BUMP(arcstat_l2_rebuild_log_blks);
	}

	/*
	* Restores a single ARC buf hdr from a log entry. The ARC buffer is put
	* into a state indicating that it has been evicted to L2ARC.
	*/
	static void
	l2arc_hdr_restore(const l2arc_log_ent_phys_t le, l2arc_dev_t dev)
	{
	arc_buf_hdr_t hdr, exists;
	kmutex_t *hash_lock;
	arc_buf_contents_t type = L2BLK_GET_TYPE((le)->le_prop);
	uint64_t asize;

	/*
	* Do all the allocation before grabbing any locks, this lets us
	* sleep if memory is full and we don't have to deal with failed
	* allocations.
	*/
	hdr = arc_buf_alloc_l2only(L2BLK_GET_LSIZE((le)->le_prop), type,
	dev, le->le_dva, le->le_daddr,
	L2BLK_GET_PSIZE((le)->le_prop), le->le_birth,
	L2BLK_GET_COMPRESS((le)->le_prop), le->le_complevel,
	L2BLK_GET_PROTECTED((le)->le_prop),
	L2BLK_GET_PREFETCH((le)->le_prop),
	L2BLK_GET_STATE((le)->le_prop));
	asize = vdev_psize_to_asize(dev->l2ad_vdev,
	L2BLK_GET_PSIZE((le)->le_prop));

	/*
	* vdev_space_update() has to be called before arc_hdr_destroy() to
	* avoid underflow since the latter also calls vdev_space_update().
	*/
	l2arc_hdr_arcstats_increment(hdr);
	vdev_space_update(dev->l2ad_vdev, asize, 0, 0);

	mutex_enter(&dev->l2ad_mtx);
	list_insert_tail(&dev->l2ad_buflist, hdr);
	(void) zfs_refcount_add_many(&dev->l2ad_alloc, arc_hdr_size(hdr), hdr);
	mutex_exit(&dev->l2ad_mtx);

	exists = buf_hash_insert(hdr, &hash_lock);
	if (exists) {
	/* Buffer was already cached, no need to restore it. */
	arc_hdr_destroy(hdr);
	/*
	* If the buffer is already cached, check whether it has
	* L2ARC metadata. If not, enter them and update the flag.
	* This is important is case of onlining a cache device, since
	* we previously evicted all L2ARC metadata from ARC.
	*/
	if (!HDR_HAS_L2HDR(exists)) {
	arc_hdr_set_flags(exists, ARC_FLAG_HAS_L2HDR);
	exists->b_l2hdr.b_dev = dev;
	exists->b_l2hdr.b_daddr = le->le_daddr;
	exists->b_l2hdr.b_arcs_state =
	L2BLK_GET_STATE((le)->le_prop);
	mutex_enter(&dev->l2ad_mtx);
	list_insert_tail(&dev->l2ad_buflist, exists);
	(void) zfs_refcount_add_many(&dev->l2ad_alloc,
	arc_hdr_size(exists), exists);
	mutex_exit(&dev->l2ad_mtx);
	l2arc_hdr_arcstats_increment(exists);
	vdev_space_update(dev->l2ad_vdev, asize, 0, 0);
	}
	ARCSTAT_BUMP(arcstat_l2_rebuild_bufs_precached);
	}

	mutex_exit(hash_lock);
	}

	/*
	* Starts an asynchronous read IO to read a log block. This is used in log
	* block reconstruction to start reading the next block before we are done
	* decoding and reconstructing the current block, to keep the l2arc device
	* nice and hot with read IO to process.
	* The returned zio will contain a newly allocated memory buffers for the IO
	* data which should then be freed by the caller once the zio is no longer
	* needed (i.e. due to it having completed). If you wish to abort this
	* zio, you should do so using l2arc_log_blk_fetch_abort, which takes
	* care of disposing of the allocated buffers correctly.
	*/
	static zio_t *
	l2arc_log_blk_fetch(vdev_t vd, const l2arc_log_blkptr_t lbp,
	l2arc_log_blk_phys_t *lb)
	{
	uint32_t asize;
	zio_t *pio;
	l2arc_read_callback_t *cb;

	/* L2BLK_GET_PSIZE returns aligned size for log blocks */
	asize = L2BLK_GET_PSIZE((lbp)->lbp_prop);
	ASSERT(asize <= sizeof (l2arc_log_blk_phys_t));

	cb = kmem_zalloc(sizeof (l2arc_read_callback_t), KM_SLEEP);
	cb->l2rcb_abd = abd_get_from_buf(lb, asize);
	pio = zio_root(vd->vdev_spa, l2arc_blk_fetch_done, cb,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_DONT_PROPAGATE \| ZIO_FLAG_DONT_RETRY);
	(void) zio_nowait(zio_read_phys(pio, vd, lbp->lbp_daddr, asize,
	cb->l2rcb_abd, ZIO_CHECKSUM_OFF, NULL, NULL,
	ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL \|
	ZIO_FLAG_DONT_PROPAGATE \| ZIO_FLAG_DONT_RETRY, B_FALSE));

	return (pio);
	}

	/*
	* Aborts a zio returned from l2arc_log_blk_fetch and frees the data
	* buffers allocated for it.
	*/
	static void
	l2arc_log_blk_fetch_abort(zio_t *zio)
	{
	(void) zio_wait(zio);
	}

	/*
	* Creates a zio to update the device header on an l2arc device.
	*/
	void
	l2arc_dev_hdr_update(l2arc_dev_t *dev)
	{
	l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr;
	const uint64_t l2dhdr_asize = dev->l2ad_dev_hdr_asize;
	abd_t *abd;
	int err;

	VERIFY(spa_config_held(dev->l2ad_spa, SCL_STATE_ALL, RW_READER));

	l2dhdr->dh_magic = L2ARC_DEV_HDR_MAGIC;
	l2dhdr->dh_version = L2ARC_PERSISTENT_VERSION;
	l2dhdr->dh_spa_guid = spa_guid(dev->l2ad_vdev->vdev_spa);
	l2dhdr->dh_vdev_guid = dev->l2ad_vdev->vdev_guid;
	l2dhdr->dh_log_entries = dev->l2ad_log_entries;
	l2dhdr->dh_evict = dev->l2ad_evict;
	l2dhdr->dh_start = dev->l2ad_start;
	l2dhdr->dh_end = dev->l2ad_end;
	l2dhdr->dh_lb_asize = zfs_refcount_count(&dev->l2ad_lb_asize);
	l2dhdr->dh_lb_count = zfs_refcount_count(&dev->l2ad_lb_count);
	l2dhdr->dh_flags = 0;
	l2dhdr->dh_trim_action_time = dev->l2ad_vdev->vdev_trim_action_time;
	l2dhdr->dh_trim_state = dev->l2ad_vdev->vdev_trim_state;
	if (dev->l2ad_first)
	l2dhdr->dh_flags \|= L2ARC_DEV_HDR_EVICT_FIRST;

	abd = abd_get_from_buf(l2dhdr, l2dhdr_asize);

	err = zio_wait(zio_write_phys(NULL, dev->l2ad_vdev,
	VDEV_LABEL_START_SIZE, l2dhdr_asize, abd, ZIO_CHECKSUM_LABEL, NULL,
	NULL, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_CANFAIL, B_FALSE));

	abd_free(abd);

	if (err != 0) {
	zfs_dbgmsg("L2ARC IO error (%d) while writing device header, "
	"vdev guid: %llu", err,
	(u_longlong_t)dev->l2ad_vdev->vdev_guid);
	}
	}

	/*
	* Commits a log block to the L2ARC device. This routine is invoked from
	* l2arc_write_buffers when the log block fills up.
	* This function allocates some memory to temporarily hold the serialized
	* buffer to be written. This is then released in l2arc_write_done.
	*/
	static uint64_t
	l2arc_log_blk_commit(l2arc_dev_t dev, zio_t pio, l2arc_write_callback_t *cb)
	{
	l2arc_log_blk_phys_t *lb = &dev->l2ad_log_blk;
	l2arc_dev_hdr_phys_t *l2dhdr = dev->l2ad_dev_hdr;
	uint64_t psize, asize;
	zio_t *wzio;
	l2arc_lb_abd_buf_t *abd_buf;
	uint8_t *tmpbuf = NULL;
	l2arc_lb_ptr_buf_t *lb_ptr_buf;

	VERIFY3S(dev->l2ad_log_ent_idx, ==, dev->l2ad_log_entries);

	abd_buf = zio_buf_alloc(sizeof (*abd_buf));
	abd_buf->abd = abd_get_from_buf(lb, sizeof (*lb));
	lb_ptr_buf = kmem_zalloc(sizeof (l2arc_lb_ptr_buf_t), KM_SLEEP);
	lb_ptr_buf->lb_ptr = kmem_zalloc(sizeof (l2arc_log_blkptr_t), KM_SLEEP);

	/* link the buffer into the block chain */
	lb->lb_prev_lbp = l2dhdr->dh_start_lbps[1];
	lb->lb_magic = L2ARC_LOG_BLK_MAGIC;

	/*
	* l2arc_log_blk_commit() may be called multiple times during a single
	* l2arc_write_buffers() call. Save the allocated abd buffers in a list
	* so we can free them in l2arc_write_done() later on.
	*/
	list_insert_tail(&cb->l2wcb_abd_list, abd_buf);

	/* try to compress the buffer */
	psize = zio_compress_data(ZIO_COMPRESS_LZ4,
	abd_buf->abd, (void *) &tmpbuf, sizeof (lb), 0);

	/* a log block is never entirely zero */
	ASSERT(psize != 0);
	asize = vdev_psize_to_asize(dev->l2ad_vdev, psize);
	ASSERT(asize <= sizeof (*lb));

	/*
	* Update the start log block pointer in the device header to point
	* to the log block we're about to write.
	*/
	l2dhdr->dh_start_lbps[1] = l2dhdr->dh_start_lbps[0];
	l2dhdr->dh_start_lbps[0].lbp_daddr = dev->l2ad_hand;
	l2dhdr->dh_start_lbps[0].lbp_payload_asize =
	dev->l2ad_log_blk_payload_asize;
	l2dhdr->dh_start_lbps[0].lbp_payload_start =
	dev->l2ad_log_blk_payload_start;
	L2BLK_SET_LSIZE(
	(&l2dhdr->dh_start_lbps[0])->lbp_prop, sizeof (*lb));
	L2BLK_SET_PSIZE(
	(&l2dhdr->dh_start_lbps[0])->lbp_prop, asize);
	L2BLK_SET_CHECKSUM(
	(&l2dhdr->dh_start_lbps[0])->lbp_prop,
	ZIO_CHECKSUM_FLETCHER_4);
	if (asize < sizeof (*lb)) {
	/* compression succeeded */
	memset(tmpbuf + psize, 0, asize - psize);
	L2BLK_SET_COMPRESS(
	(&l2dhdr->dh_start_lbps[0])->lbp_prop,
	ZIO_COMPRESS_LZ4);
	} else {
	/* compression failed */
	memcpy(tmpbuf, lb, sizeof (*lb));
	L2BLK_SET_COMPRESS(
	(&l2dhdr->dh_start_lbps[0])->lbp_prop,
	ZIO_COMPRESS_OFF);
	}

	/* checksum what we're about to write */
	fletcher_4_native(tmpbuf, asize, NULL,
	&l2dhdr->dh_start_lbps[0].lbp_cksum);

	abd_free(abd_buf->abd);

	/* perform the write itself */
	abd_buf->abd = abd_get_from_buf(tmpbuf, sizeof (*lb));
	abd_take_ownership_of_buf(abd_buf->abd, B_TRUE);
	wzio = zio_write_phys(pio, dev->l2ad_vdev, dev->l2ad_hand,
	asize, abd_buf->abd, ZIO_CHECKSUM_OFF, NULL, NULL,
	ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_CANFAIL, B_FALSE);
	DTRACE_PROBE2(l2arc__write, vdev_t , dev->l2ad_vdev, zio_t , wzio);
	(void) zio_nowait(wzio);

	dev->l2ad_hand += asize;
	/*
	* Include the committed log block's pointer in the list of pointers
	* to log blocks present in the L2ARC device.
	*/
	memcpy(lb_ptr_buf->lb_ptr, &l2dhdr->dh_start_lbps[0],
	sizeof (l2arc_log_blkptr_t));
	mutex_enter(&dev->l2ad_mtx);
	list_insert_head(&dev->l2ad_lbptr_list, lb_ptr_buf);
	ARCSTAT_INCR(arcstat_l2_log_blk_asize, asize);
	ARCSTAT_BUMP(arcstat_l2_log_blk_count);
	zfs_refcount_add_many(&dev->l2ad_lb_asize, asize, lb_ptr_buf);
	zfs_refcount_add(&dev->l2ad_lb_count, lb_ptr_buf);
	mutex_exit(&dev->l2ad_mtx);
	vdev_space_update(dev->l2ad_vdev, asize, 0, 0);

	/* bump the kstats */
	ARCSTAT_INCR(arcstat_l2_write_bytes, asize);
	ARCSTAT_BUMP(arcstat_l2_log_blk_writes);
	ARCSTAT_F_AVG(arcstat_l2_log_blk_avg_asize, asize);
	ARCSTAT_F_AVG(arcstat_l2_data_to_meta_ratio,
	dev->l2ad_log_blk_payload_asize / asize);

	/* start a new log block */
	dev->l2ad_log_ent_idx = 0;
	dev->l2ad_log_blk_payload_asize = 0;
	dev->l2ad_log_blk_payload_start = 0;

	return (asize);
	}

	/*
	* Validates an L2ARC log block address to make sure that it can be read
	* from the provided L2ARC device.
	*/
	boolean_t
	l2arc_log_blkptr_valid(l2arc_dev_t dev, const l2arc_log_blkptr_t lbp)
	{
	/* L2BLK_GET_PSIZE returns aligned size for log blocks */
	uint64_t asize = L2BLK_GET_PSIZE((lbp)->lbp_prop);
	uint64_t end = lbp->lbp_daddr + asize - 1;
	uint64_t start = lbp->lbp_payload_start;
	boolean_t evicted = B_FALSE;

	/*
	* A log block is valid if all of the following conditions are true:
	* - it fits entirely (including its payload) between l2ad_start and
	* l2ad_end
	* - it has a valid size
	* - neither the log block itself nor part of its payload was evicted
	* by l2arc_evict():
	*
	* l2ad_hand l2ad_evict
	* \| \| lbp_daddr
	* \| start \| \| end
	* \| \| \| \| \|
	* V V V V V
	* l2ad_start ============================================ l2ad_end
	* --------------------------\|\|\|\|
	* ^ ^
	* \| log block
	* payload
	*/

	evicted =
	l2arc_range_check_overlap(start, end, dev->l2ad_hand) \|\|
	l2arc_range_check_overlap(start, end, dev->l2ad_evict) \|\|
	l2arc_range_check_overlap(dev->l2ad_hand, dev->l2ad_evict, start) \|\|
	l2arc_range_check_overlap(dev->l2ad_hand, dev->l2ad_evict, end);

	return (start >= dev->l2ad_start && end <= dev->l2ad_end &&
	asize > 0 && asize <= sizeof (l2arc_log_blk_phys_t) &&
	(!evicted \|\| dev->l2ad_first));
	}

	/*
	* Inserts ARC buffer header `hdr' into the current L2ARC log block on
	* the device. The buffer being inserted must be present in L2ARC.
	* Returns B_TRUE if the L2ARC log block is full and needs to be committed
	* to L2ARC, or B_FALSE if it still has room for more ARC buffers.
	*/
	static boolean_t
	l2arc_log_blk_insert(l2arc_dev_t dev, const arc_buf_hdr_t hdr)
	{
	l2arc_log_blk_phys_t *lb = &dev->l2ad_log_blk;
	l2arc_log_ent_phys_t *le;

	if (dev->l2ad_log_entries == 0)
	return (B_FALSE);

	int index = dev->l2ad_log_ent_idx++;

	ASSERT3S(index, <, dev->l2ad_log_entries);
	ASSERT(HDR_HAS_L2HDR(hdr));

	le = &lb->lb_entries[index];
	memset(le, 0, sizeof (*le));
	le->le_dva = hdr->b_dva;
	le->le_birth = hdr->b_birth;
	le->le_daddr = hdr->b_l2hdr.b_daddr;
	if (index == 0)
	dev->l2ad_log_blk_payload_start = le->le_daddr;
	L2BLK_SET_LSIZE((le)->le_prop, HDR_GET_LSIZE(hdr));
	L2BLK_SET_PSIZE((le)->le_prop, HDR_GET_PSIZE(hdr));
	L2BLK_SET_COMPRESS((le)->le_prop, HDR_GET_COMPRESS(hdr));
	le->le_complevel = hdr->b_complevel;
	L2BLK_SET_TYPE((le)->le_prop, hdr->b_type);
	L2BLK_SET_PROTECTED((le)->le_prop, !!(HDR_PROTECTED(hdr)));
	L2BLK_SET_PREFETCH((le)->le_prop, !!(HDR_PREFETCH(hdr)));
	L2BLK_SET_STATE((le)->le_prop, hdr->b_l2hdr.b_arcs_state);

	dev->l2ad_log_blk_payload_asize += vdev_psize_to_asize(dev->l2ad_vdev,
	HDR_GET_PSIZE(hdr));

	return (dev->l2ad_log_ent_idx == dev->l2ad_log_entries);
	}

	/*
	* Checks whether a given L2ARC device address sits in a time-sequential
	* range. The trick here is that the L2ARC is a rotary buffer, so we can't
	* just do a range comparison, we need to handle the situation in which the
	* range wraps around the end of the L2ARC device. Arguments:
	* bottom -- Lower end of the range to check (written to earlier).
	* top -- Upper end of the range to check (written to later).
	* check -- The address for which we want to determine if it sits in
	* between the top and bottom.
	*
	* The 3-way conditional below represents the following cases:
	*
	* bottom < top : Sequentially ordered case:
	* <check>--------+-------------------+
	* \| (overlap here?) \|
	* L2ARC dev V V
	* \|---------------<bottom>============<top>--------------\|
	*
	* bottom > top: Looped-around case:
	* <check>--------+------------------+
	* \| (overlap here?) \|
	* L2ARC dev V V
	* \|===============<top>---------------<bottom>===========\|
	* ^ ^
	* \| (or here?) \|
	* +---------------+---------<check>
	*
	* top == bottom : Just a single address comparison.
	*/
	boolean_t
	l2arc_range_check_overlap(uint64_t bottom, uint64_t top, uint64_t check)
	{
	if (bottom < top)
	return (bottom <= check && check <= top);
	else if (bottom > top)
	return (check <= top \|\| bottom <= check);
	else
	return (check == top);
	}

	EXPORT_SYMBOL(arc_buf_size);
	EXPORT_SYMBOL(arc_write);
	EXPORT_SYMBOL(arc_read);
	EXPORT_SYMBOL(arc_buf_info);
	EXPORT_SYMBOL(arc_getbuf_func);
	EXPORT_SYMBOL(arc_add_prune_callback);
	EXPORT_SYMBOL(arc_remove_prune_callback);

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, min, param_set_arc_min,
	spl_param_get_u64, ZMOD_RW, "Minimum ARC size in bytes");

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, max, param_set_arc_max,
	spl_param_get_u64, ZMOD_RW, "Maximum ARC size in bytes");

	ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, meta_balance, UINT, ZMOD_RW,
	"Balance between metadata and data on ghost hits.");

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, grow_retry, param_set_arc_int,
	param_get_uint, ZMOD_RW, "Seconds before growing ARC size");

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, shrink_shift, param_set_arc_int,
	param_get_uint, ZMOD_RW, "log2(fraction of ARC to reclaim)");

	ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, pc_percent, UINT, ZMOD_RW,
	"Percent of pagecache to reclaim ARC to");

	ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, average_blocksize, UINT, ZMOD_RD,
	"Target average block size");

	ZFS_MODULE_PARAM(zfs, zfs_, compressed_arc_enabled, INT, ZMOD_RW,
	"Disable compressed ARC buffers");

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, min_prefetch_ms, param_set_arc_int,
	param_get_uint, ZMOD_RW, "Min life of prefetch block in ms");

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, min_prescient_prefetch_ms,
	param_set_arc_int, param_get_uint, ZMOD_RW,
	"Min life of prescient prefetched block in ms");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, write_max, U64, ZMOD_RW,
	"Max write bytes per interval");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, write_boost, U64, ZMOD_RW,
	"Extra write bytes during device warmup");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, headroom, U64, ZMOD_RW,
	"Number of max device writes to precache");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, headroom_boost, U64, ZMOD_RW,
	"Compressed l2arc_headroom multiplier");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, trim_ahead, U64, ZMOD_RW,
	"TRIM ahead L2ARC write size multiplier");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, feed_secs, U64, ZMOD_RW,
	"Seconds between L2ARC writing");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, feed_min_ms, U64, ZMOD_RW,
	"Min feed interval in milliseconds");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, noprefetch, INT, ZMOD_RW,
	"Skip caching prefetched buffers");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, feed_again, INT, ZMOD_RW,
	"Turbo L2ARC warmup");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, norw, INT, ZMOD_RW,
	"No reads during writes");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, meta_percent, UINT, ZMOD_RW,
	"Percent of ARC size allowed for L2ARC-only headers");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, rebuild_enabled, INT, ZMOD_RW,
	"Rebuild the L2ARC when importing a pool");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, rebuild_blocks_min_l2size, U64, ZMOD_RW,
	"Min size in bytes to write rebuild log blocks in L2ARC");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, mfuonly, INT, ZMOD_RW,
	"Cache only MFU data from ARC into L2ARC");

	ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, exclude_special, INT, ZMOD_RW,
	"Exclude dbufs on special vdevs from being cached to L2ARC if set.");

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, lotsfree_percent, param_set_arc_int,
	param_get_uint, ZMOD_RW, "System free memory I/O throttle in bytes");

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, sys_free, param_set_arc_u64,
	spl_param_get_u64, ZMOD_RW, "System free memory target size in bytes");

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, dnode_limit, param_set_arc_u64,
	spl_param_get_u64, ZMOD_RW, "Minimum bytes of dnodes in ARC");

	ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, dnode_limit_percent,
	param_set_arc_int, param_get_uint, ZMOD_RW,
	"Percent of ARC meta buffers for dnodes");

	ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, dnode_reduce_percent, UINT, ZMOD_RW,
	"Percentage of excess dnodes to try to unpin");

	ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, eviction_pct, UINT, ZMOD_RW,
	"When full, ARC allocation waits for eviction of this % of alloc size");

	ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, evict_batch_limit, UINT, ZMOD_RW,
	"The number of headers to evict per sublist before moving to the next");

	ZFS_MODULE_PARAM(zfs_arc, zfs_arc_, prune_task_threads, INT, ZMOD_RW,
	"Number of arc_prune threads");
	diff --git a/sys/contrib/openzfs/module/zfs/btree.c b/sys/contrib/openzfs/module/zfs/btree.c
	index af2b94a850be..9c52083603f1 100644
	--- a/sys/contrib/openzfs/module/zfs/btree.c
	+++ b/sys/contrib/openzfs/module/zfs/btree.c
	@@ -1,2215 +1,2215 @@
	/*
	* CDDL HEADER START
	*
	* This file and its contents are supplied under the terms of the
	* Common Development and Distribution License ("CDDL"), version 1.0.
	* You may only use this file in accordance with the terms of version
	* 1.0 of the CDDL.
	*
	* A full copy of the text of the CDDL should have accompanied this
	* source. A copy of the CDDL is also available via the Internet at
	* http://www.illumos.org/license/CDDL.
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2019 by Delphix. All rights reserved.
	*/

	#include <sys/btree.h>
	#include <sys/bitops.h>
	#include <sys/zfs_context.h>

	kmem_cache_t *zfs_btree_leaf_cache;

	/*
	* Control the extent of the verification that occurs when zfs_btree_verify is
	* called. Primarily used for debugging when extending the btree logic and
	* functionality. As the intensity is increased, new verification steps are
	* added. These steps are cumulative; intensity = 3 includes the intensity = 1
	* and intensity = 2 steps as well.
	*
	* Intensity 1: Verify that the tree's height is consistent throughout.
	* Intensity 2: Verify that a core node's children's parent pointers point
	* to the core node.
	* Intensity 3: Verify that the total number of elements in the tree matches the
	* sum of the number of elements in each node. Also verifies that each node's
	* count obeys the invariants (less than or equal to maximum value, greater than
	* or equal to half the maximum minus one).
	* Intensity 4: Verify that each element compares less than the element
	* immediately after it and greater than the one immediately before it using the
	* comparator function. For core nodes, also checks that each element is greater
	* than the last element in the first of the two nodes it separates, and less
	* than the first element in the second of the two nodes.
	* Intensity 5: Verifies, if ZFS_DEBUG is defined, that all unused memory inside
	* of each node is poisoned appropriately. Note that poisoning always occurs if
	* ZFS_DEBUG is set, so it is safe to set the intensity to 5 during normal
	* operation.
	*
	* Intensity 4 and 5 are particularly expensive to perform; the previous levels
	* are a few memory operations per node, while these levels require multiple
	* operations per element. In addition, when creating large btrees, these
	* operations are called at every step, resulting in extremely slow operation
	* (while the asymptotic complexity of the other steps is the same, the
	* importance of the constant factors cannot be denied).
	*/
	uint_t zfs_btree_verify_intensity = 0;

	/*
	* Convenience functions to silence warnings from memcpy/memmove's
	* return values and change argument order to src, dest.
	*/
	static void
	bcpy(const void src, void dest, size_t size)
	{
	(void) memcpy(dest, src, size);
	}

	static void
	bmov(const void src, void dest, size_t size)
	{
	(void) memmove(dest, src, size);
	}

	static boolean_t
	zfs_btree_is_core(struct zfs_btree_hdr *hdr)
	{
	return (hdr->bth_first == -1);
	}

	#ifdef _ILP32
	#define BTREE_POISON 0xabadb10c
	#else
	#define BTREE_POISON 0xabadb10cdeadbeef
	#endif

	static void
	zfs_btree_poison_node(zfs_btree_t tree, zfs_btree_hdr_t hdr)
	{
	#ifdef ZFS_DEBUG
	size_t size = tree->bt_elem_size;
	if (zfs_btree_is_core(hdr)) {
	zfs_btree_core_t node = (zfs_btree_core_t )hdr;
	for (uint32_t i = hdr->bth_count + 1; i <= BTREE_CORE_ELEMS;
	i++) {
	node->btc_children[i] =
	(zfs_btree_hdr_t *)BTREE_POISON;
	}
	(void) memset(node->btc_elems + hdr->bth_count * size, 0x0f,
	(BTREE_CORE_ELEMS - hdr->bth_count) * size);
	} else {
	zfs_btree_leaf_t leaf = (zfs_btree_leaf_t )hdr;
	(void) memset(leaf->btl_elems, 0x0f, hdr->bth_first * size);
	(void) memset(leaf->btl_elems +
	(hdr->bth_first + hdr->bth_count) * size, 0x0f,
	tree->bt_leaf_size - offsetof(zfs_btree_leaf_t, btl_elems) -
	(hdr->bth_first + hdr->bth_count) * size);
	}
	#endif
	}

	static inline void
	zfs_btree_poison_node_at(zfs_btree_t tree, zfs_btree_hdr_t hdr,
	uint32_t idx, uint32_t count)
	{
	#ifdef ZFS_DEBUG
	size_t size = tree->bt_elem_size;
	if (zfs_btree_is_core(hdr)) {
	ASSERT3U(idx, >=, hdr->bth_count);
	ASSERT3U(idx, <=, BTREE_CORE_ELEMS);
	ASSERT3U(idx + count, <=, BTREE_CORE_ELEMS);
	zfs_btree_core_t node = (zfs_btree_core_t )hdr;
	for (uint32_t i = 1; i <= count; i++) {
	node->btc_children[idx + i] =
	(zfs_btree_hdr_t *)BTREE_POISON;
	}
	(void) memset(node->btc_elems + idx * size, 0x0f, count * size);
	} else {
	ASSERT3U(idx, <=, tree->bt_leaf_cap);
	ASSERT3U(idx + count, <=, tree->bt_leaf_cap);
	zfs_btree_leaf_t leaf = (zfs_btree_leaf_t )hdr;
	(void) memset(leaf->btl_elems +
	(hdr->bth_first + idx) * size, 0x0f, count * size);
	}
	#endif
	}

	static inline void
	zfs_btree_verify_poison_at(zfs_btree_t tree, zfs_btree_hdr_t hdr,
	uint32_t idx)
	{
	#ifdef ZFS_DEBUG
	size_t size = tree->bt_elem_size;
	if (zfs_btree_is_core(hdr)) {
	ASSERT3U(idx, <, BTREE_CORE_ELEMS);
	zfs_btree_core_t node = (zfs_btree_core_t )hdr;
	zfs_btree_hdr_t cval = (zfs_btree_hdr_t )BTREE_POISON;
	VERIFY3P(node->btc_children[idx + 1], ==, cval);
	for (size_t i = 0; i < size; i++)
	VERIFY3U(node->btc_elems[idx * size + i], ==, 0x0f);
	} else {
	ASSERT3U(idx, <, tree->bt_leaf_cap);
	zfs_btree_leaf_t leaf = (zfs_btree_leaf_t )hdr;
	if (idx >= tree->bt_leaf_cap - hdr->bth_first)
	return;
	for (size_t i = 0; i < size; i++) {
	VERIFY3U(leaf->btl_elems[(hdr->bth_first + idx)
	* size + i], ==, 0x0f);
	}
	}
	#endif
	}

	void
	zfs_btree_init(void)
	{
	zfs_btree_leaf_cache = kmem_cache_create("zfs_btree_leaf_cache",
	BTREE_LEAF_SIZE, 0, NULL, NULL, NULL, NULL, NULL, 0);
	}

	void
	zfs_btree_fini(void)
	{
	kmem_cache_destroy(zfs_btree_leaf_cache);
	}

	static void *
	zfs_btree_leaf_alloc(zfs_btree_t *tree)
	{
	if (tree->bt_leaf_size == BTREE_LEAF_SIZE)
	return (kmem_cache_alloc(zfs_btree_leaf_cache, KM_SLEEP));
	else
	return (kmem_alloc(tree->bt_leaf_size, KM_SLEEP));
	}

	static void
	zfs_btree_leaf_free(zfs_btree_t tree, void ptr)
	{
	if (tree->bt_leaf_size == BTREE_LEAF_SIZE)
	return (kmem_cache_free(zfs_btree_leaf_cache, ptr));
	else
	return (kmem_free(ptr, tree->bt_leaf_size));
	}

	void
	zfs_btree_create(zfs_btree_t tree, int (compar) (const void , const void ),
	bt_find_in_buf_f bt_find_in_buf, size_t size)
	{
	zfs_btree_create_custom(tree, compar, bt_find_in_buf, size,
	BTREE_LEAF_SIZE);
	}

	static void *
	zfs_btree_find_in_buf(zfs_btree_t tree, uint8_t buf, uint32_t nelems,
	const void value, zfs_btree_index_t where);

	void
	zfs_btree_create_custom(zfs_btree_t *tree,
	int (compar) (const void , const void *),
	bt_find_in_buf_f bt_find_in_buf,
	size_t size, size_t lsize)
	{
	size_t esize = lsize - offsetof(zfs_btree_leaf_t, btl_elems);

	ASSERT3U(size, <=, esize / 2);
	memset(tree, 0, sizeof (*tree));
	tree->bt_compar = compar;
	tree->bt_find_in_buf = (bt_find_in_buf == NULL) ?
	zfs_btree_find_in_buf : bt_find_in_buf;
	tree->bt_elem_size = size;
	tree->bt_leaf_size = lsize;
	- tree->bt_leaf_cap = P2ALIGN(esize / size, 2);
	+ tree->bt_leaf_cap = P2ALIGN_TYPED(esize / size, 2, size_t);
	tree->bt_height = -1;
	tree->bt_bulk = NULL;
	}

	/*
	* Find value in the array of elements provided. Uses a simple binary search.
	*/
	static void *
	zfs_btree_find_in_buf(zfs_btree_t tree, uint8_t buf, uint32_t nelems,
	const void value, zfs_btree_index_t where)
	{
	uint32_t max = nelems;
	uint32_t min = 0;
	while (max > min) {
	uint32_t idx = (min + max) / 2;
	uint8_t cur = buf + idx tree->bt_elem_size;
	int comp = tree->bt_compar(cur, value);
	if (comp < 0) {
	min = idx + 1;
	} else if (comp > 0) {
	max = idx;
	} else {
	where->bti_offset = idx;
	where->bti_before = B_FALSE;
	return (cur);
	}
	}

	where->bti_offset = max;
	where->bti_before = B_TRUE;
	return (NULL);
	}

	/*
	* Find the given value in the tree. where may be passed as null to use as a
	* membership test or if the btree is being used as a map.
	*/
	void *
	zfs_btree_find(zfs_btree_t tree, const void value, zfs_btree_index_t *where)
	{
	if (tree->bt_height == -1) {
	if (where != NULL) {
	where->bti_node = NULL;
	where->bti_offset = 0;
	}
	ASSERT0(tree->bt_num_elems);
	return (NULL);
	}

	/*
	* If we're in bulk-insert mode, we check the last spot in the tree
	* and the last leaf in the tree before doing the normal search,
	* because for most workloads the vast majority of finds in
	* bulk-insert mode are to insert new elements.
	*/
	zfs_btree_index_t idx;
	size_t size = tree->bt_elem_size;
	if (tree->bt_bulk != NULL) {
	zfs_btree_leaf_t *last_leaf = tree->bt_bulk;
	int comp = tree->bt_compar(last_leaf->btl_elems +
	(last_leaf->btl_hdr.bth_first +
	last_leaf->btl_hdr.bth_count - 1) * size, value);
	if (comp < 0) {
	/*
	* If what they're looking for is after the last
	* element, it's not in the tree.
	*/
	if (where != NULL) {
	where->bti_node = (zfs_btree_hdr_t *)last_leaf;
	where->bti_offset =
	last_leaf->btl_hdr.bth_count;
	where->bti_before = B_TRUE;
	}
	return (NULL);
	} else if (comp == 0) {
	if (where != NULL) {
	where->bti_node = (zfs_btree_hdr_t *)last_leaf;
	where->bti_offset =
	last_leaf->btl_hdr.bth_count - 1;
	where->bti_before = B_FALSE;
	}
	return (last_leaf->btl_elems +
	(last_leaf->btl_hdr.bth_first +
	last_leaf->btl_hdr.bth_count - 1) * size);
	}
	if (tree->bt_compar(last_leaf->btl_elems +
	last_leaf->btl_hdr.bth_first * size, value) <= 0) {
	/*
	* If what they're looking for is after the first
	* element in the last leaf, it's in the last leaf or
	* it's not in the tree.
	*/
	void *d = tree->bt_find_in_buf(tree,
	last_leaf->btl_elems +
	last_leaf->btl_hdr.bth_first * size,
	last_leaf->btl_hdr.bth_count, value, &idx);

	if (where != NULL) {
	idx.bti_node = (zfs_btree_hdr_t *)last_leaf;
	*where = idx;
	}
	return (d);
	}
	}

	zfs_btree_core_t *node = NULL;
	uint32_t child = 0;
	uint32_t depth = 0;

	/*
	* Iterate down the tree, finding which child the value should be in
	* by comparing with the separators.
	*/
	for (node = (zfs_btree_core_t *)tree->bt_root; depth < tree->bt_height;
	node = (zfs_btree_core_t *)node->btc_children[child], depth++) {
	ASSERT3P(node, !=, NULL);
	void *d = tree->bt_find_in_buf(tree, node->btc_elems,
	node->btc_hdr.bth_count, value, &idx);
	EQUIV(d != NULL, !idx.bti_before);
	if (d != NULL) {
	if (where != NULL) {
	idx.bti_node = (zfs_btree_hdr_t *)node;
	*where = idx;
	}
	return (d);
	}
	ASSERT(idx.bti_before);
	child = idx.bti_offset;
	}

	/*
	* The value is in this leaf, or it would be if it were in the
	* tree. Find its proper location and return it.
	*/
	zfs_btree_leaf_t *leaf = (depth == 0 ?
	(zfs_btree_leaf_t )tree->bt_root : (zfs_btree_leaf_t )node);
	void *d = tree->bt_find_in_buf(tree, leaf->btl_elems +
	leaf->btl_hdr.bth_first * size,
	leaf->btl_hdr.bth_count, value, &idx);

	if (where != NULL) {
	idx.bti_node = (zfs_btree_hdr_t *)leaf;
	*where = idx;
	}

	return (d);
	}

	/*
	* To explain the following functions, it is useful to understand the four
	* kinds of shifts used in btree operation. First, a shift is a movement of
	* elements within a node. It is used to create gaps for inserting new
	* elements and children, or cover gaps created when things are removed. A
	* shift has two fundamental properties, each of which can be one of two
	* values, making four types of shifts. There is the direction of the shift
	* (left or right) and the shape of the shift (parallelogram or isoceles
	* trapezoid (shortened to trapezoid hereafter)). The shape distinction only
	* applies to shifts of core nodes.
	*
	* The names derive from the following imagining of the layout of a node:
	*
	* Elements: * * * * * * * ... * * *
	* Children: * * * * * * * * ... * * *
	*
	* This layout follows from the fact that the elements act as separators
	* between pairs of children, and that children root subtrees "below" the
	* current node. A left and right shift are fairly self-explanatory; a left
	* shift moves things to the left, while a right shift moves things to the
	* right. A parallelogram shift is a shift with the same number of elements
	* and children being moved, while a trapezoid shift is a shift that moves one
	* more children than elements. An example follows:
	*
	* A parallelogram shift could contain the following:
	* _______________
	* \* * * * \ * * * ... * * *
	* * \ * * * \ * * ... * * *
	* ---------------
	* A trapezoid shift could contain the following:
	* ___________
	* * / * * * \ * * * ... * * *
	* * / * * * \ * * ... * * *
	* ---------------
	*
	* Note that a parallelogram shift is always shaped like a "left-leaning"
	* parallelogram, where the starting index of the children being moved is
	* always one higher than the starting index of the elements being moved. No
	* "right-leaning" parallelogram shifts are needed (shifts where the starting
	* element index and starting child index being moved are the same) to achieve
	* any btree operations, so we ignore them.
	*/

	enum bt_shift_shape {
	BSS_TRAPEZOID,
	BSS_PARALLELOGRAM
	};

	enum bt_shift_direction {
	BSD_LEFT,
	BSD_RIGHT
	};

	/*
	* Shift elements and children in the provided core node by off spots. The
	* first element moved is idx, and count elements are moved. The shape of the
	* shift is determined by shape. The direction is determined by dir.
	*/
	static inline void
	bt_shift_core(zfs_btree_t tree, zfs_btree_core_t node, uint32_t idx,
	uint32_t count, uint32_t off, enum bt_shift_shape shape,
	enum bt_shift_direction dir)
	{
	size_t size = tree->bt_elem_size;
	ASSERT(zfs_btree_is_core(&node->btc_hdr));

	uint8_t e_start = node->btc_elems + idx size;
	uint8_t e_out = (dir == BSD_LEFT ? e_start - off size :
	e_start + off * size);
	bmov(e_start, e_out, count * size);

	zfs_btree_hdr_t **c_start = node->btc_children + idx +
	(shape == BSS_TRAPEZOID ? 0 : 1);
	zfs_btree_hdr_t **c_out = (dir == BSD_LEFT ? c_start - off :
	c_start + off);
	uint32_t c_count = count + (shape == BSS_TRAPEZOID ? 1 : 0);
	bmov(c_start, c_out, c_count * sizeof (*c_start));
	}

	/*
	* Shift elements and children in the provided core node left by one spot.
	* The first element moved is idx, and count elements are moved. The
	* shape of the shift is determined by trap; true if the shift is a trapezoid,
	* false if it is a parallelogram.
	*/
	static inline void
	bt_shift_core_left(zfs_btree_t tree, zfs_btree_core_t node, uint32_t idx,
	uint32_t count, enum bt_shift_shape shape)
	{
	bt_shift_core(tree, node, idx, count, 1, shape, BSD_LEFT);
	}

	/*
	* Shift elements and children in the provided core node right by one spot.
	* Starts with elements[idx] and children[idx] and one more child than element.
	*/
	static inline void
	bt_shift_core_right(zfs_btree_t tree, zfs_btree_core_t node, uint32_t idx,
	uint32_t count, enum bt_shift_shape shape)
	{
	bt_shift_core(tree, node, idx, count, 1, shape, BSD_RIGHT);
	}

	/*
	* Shift elements and children in the provided leaf node by off spots.
	* The first element moved is idx, and count elements are moved. The direction
	* is determined by left.
	*/
	static inline void
	bt_shift_leaf(zfs_btree_t tree, zfs_btree_leaf_t node, uint32_t idx,
	uint32_t count, uint32_t off, enum bt_shift_direction dir)
	{
	size_t size = tree->bt_elem_size;
	zfs_btree_hdr_t *hdr = &node->btl_hdr;
	ASSERT(!zfs_btree_is_core(hdr));

	if (count == 0)
	return;
	uint8_t start = node->btl_elems + (hdr->bth_first + idx) size;
	uint8_t out = (dir == BSD_LEFT ? start - off size :
	start + off * size);
	bmov(start, out, count * size);
	}

	/*
	* Grow leaf for n new elements before idx.
	*/
	static void
	bt_grow_leaf(zfs_btree_t tree, zfs_btree_leaf_t leaf, uint32_t idx,
	uint32_t n)
	{
	zfs_btree_hdr_t *hdr = &leaf->btl_hdr;
	ASSERT(!zfs_btree_is_core(hdr));
	ASSERT3U(idx, <=, hdr->bth_count);
	uint32_t capacity = tree->bt_leaf_cap;
	ASSERT3U(hdr->bth_count + n, <=, capacity);
	boolean_t cl = (hdr->bth_first >= n);
	boolean_t cr = (hdr->bth_first + hdr->bth_count + n <= capacity);

	if (cl && (!cr \|\| idx <= hdr->bth_count / 2)) {
	/* Grow left. */
	hdr->bth_first -= n;
	bt_shift_leaf(tree, leaf, n, idx, n, BSD_LEFT);
	} else if (cr) {
	/* Grow right. */
	bt_shift_leaf(tree, leaf, idx, hdr->bth_count - idx, n,
	BSD_RIGHT);
	} else {
	/* Grow both ways. */
	uint32_t fn = hdr->bth_first -
	(capacity - (hdr->bth_count + n)) / 2;
	hdr->bth_first -= fn;
	bt_shift_leaf(tree, leaf, fn, idx, fn, BSD_LEFT);
	bt_shift_leaf(tree, leaf, fn + idx, hdr->bth_count - idx,
	n - fn, BSD_RIGHT);
	}
	hdr->bth_count += n;
	}

	/*
	* Shrink leaf for count elements starting from idx.
	*/
	static void
	bt_shrink_leaf(zfs_btree_t tree, zfs_btree_leaf_t leaf, uint32_t idx,
	uint32_t n)
	{
	zfs_btree_hdr_t *hdr = &leaf->btl_hdr;
	ASSERT(!zfs_btree_is_core(hdr));
	ASSERT3U(idx, <=, hdr->bth_count);
	ASSERT3U(idx + n, <=, hdr->bth_count);

	if (idx <= (hdr->bth_count - n) / 2) {
	bt_shift_leaf(tree, leaf, 0, idx, n, BSD_RIGHT);
	zfs_btree_poison_node_at(tree, hdr, 0, n);
	hdr->bth_first += n;
	} else {
	bt_shift_leaf(tree, leaf, idx + n, hdr->bth_count - idx - n, n,
	BSD_LEFT);
	zfs_btree_poison_node_at(tree, hdr, hdr->bth_count - n, n);
	}
	hdr->bth_count -= n;
	}

	/*
	* Move children and elements from one core node to another. The shape
	* parameter behaves the same as it does in the shift logic.
	*/
	static inline void
	bt_transfer_core(zfs_btree_t tree, zfs_btree_core_t source, uint32_t sidx,
	uint32_t count, zfs_btree_core_t *dest, uint32_t didx,
	enum bt_shift_shape shape)
	{
	size_t size = tree->bt_elem_size;
	ASSERT(zfs_btree_is_core(&source->btc_hdr));
	ASSERT(zfs_btree_is_core(&dest->btc_hdr));

	bcpy(source->btc_elems + sidx * size, dest->btc_elems + didx * size,
	count * size);

	uint32_t c_count = count + (shape == BSS_TRAPEZOID ? 1 : 0);
	bcpy(source->btc_children + sidx + (shape == BSS_TRAPEZOID ? 0 : 1),
	dest->btc_children + didx + (shape == BSS_TRAPEZOID ? 0 : 1),
	c_count * sizeof (*source->btc_children));
	}

	static inline void
	bt_transfer_leaf(zfs_btree_t tree, zfs_btree_leaf_t source, uint32_t sidx,
	uint32_t count, zfs_btree_leaf_t *dest, uint32_t didx)
	{
	size_t size = tree->bt_elem_size;
	ASSERT(!zfs_btree_is_core(&source->btl_hdr));
	ASSERT(!zfs_btree_is_core(&dest->btl_hdr));

	bcpy(source->btl_elems + (source->btl_hdr.bth_first + sidx) * size,
	dest->btl_elems + (dest->btl_hdr.bth_first + didx) * size,
	count * size);
	}

	/*
	* Find the first element in the subtree rooted at hdr, return its value and
	* put its location in where if non-null.
	*/
	static void *
	zfs_btree_first_helper(zfs_btree_t tree, zfs_btree_hdr_t hdr,
	zfs_btree_index_t *where)
	{
	zfs_btree_hdr_t *node;

	for (node = hdr; zfs_btree_is_core(node);
	node = ((zfs_btree_core_t *)node)->btc_children[0])
	;

	ASSERT(!zfs_btree_is_core(node));
	zfs_btree_leaf_t leaf = (zfs_btree_leaf_t )node;
	if (where != NULL) {
	where->bti_node = node;
	where->bti_offset = 0;
	where->bti_before = B_FALSE;
	}
	return (&leaf->btl_elems[node->bth_first * tree->bt_elem_size]);
	}

	/* Insert an element and a child into a core node at the given offset. */
	static void
	zfs_btree_insert_core_impl(zfs_btree_t tree, zfs_btree_core_t parent,
	uint32_t offset, zfs_btree_hdr_t new_node, void buf)
	{
	size_t size = tree->bt_elem_size;
	zfs_btree_hdr_t *par_hdr = &parent->btc_hdr;
	ASSERT3P(par_hdr, ==, new_node->bth_parent);
	ASSERT3U(par_hdr->bth_count, <, BTREE_CORE_ELEMS);

	if (zfs_btree_verify_intensity >= 5) {
	zfs_btree_verify_poison_at(tree, par_hdr,
	par_hdr->bth_count);
	}
	/* Shift existing elements and children */
	uint32_t count = par_hdr->bth_count - offset;
	bt_shift_core_right(tree, parent, offset, count,
	BSS_PARALLELOGRAM);

	/* Insert new values */
	parent->btc_children[offset + 1] = new_node;
	bcpy(buf, parent->btc_elems + offset * size, size);
	par_hdr->bth_count++;
	}

	/*
	* Insert new_node into the parent of old_node directly after old_node, with
	* buf as the dividing element between the two.
	*/
	static void
	zfs_btree_insert_into_parent(zfs_btree_t tree, zfs_btree_hdr_t old_node,
	zfs_btree_hdr_t new_node, void buf)
	{
	ASSERT3P(old_node->bth_parent, ==, new_node->bth_parent);
	size_t size = tree->bt_elem_size;
	zfs_btree_core_t *parent = old_node->bth_parent;

	/*
	* If this is the root node we were splitting, we create a new root
	* and increase the height of the tree.
	*/
	if (parent == NULL) {
	ASSERT3P(old_node, ==, tree->bt_root);
	tree->bt_num_nodes++;
	zfs_btree_core_t *new_root =
	kmem_alloc(sizeof (zfs_btree_core_t) + BTREE_CORE_ELEMS *
	size, KM_SLEEP);
	zfs_btree_hdr_t *new_root_hdr = &new_root->btc_hdr;
	new_root_hdr->bth_parent = NULL;
	new_root_hdr->bth_first = -1;
	new_root_hdr->bth_count = 1;

	old_node->bth_parent = new_node->bth_parent = new_root;
	new_root->btc_children[0] = old_node;
	new_root->btc_children[1] = new_node;
	bcpy(buf, new_root->btc_elems, size);

	tree->bt_height++;
	tree->bt_root = new_root_hdr;
	zfs_btree_poison_node(tree, new_root_hdr);
	return;
	}

	/*
	* Since we have the new separator, binary search for where to put
	* new_node.
	*/
	zfs_btree_hdr_t *par_hdr = &parent->btc_hdr;
	zfs_btree_index_t idx;
	ASSERT(zfs_btree_is_core(par_hdr));
	VERIFY3P(tree->bt_find_in_buf(tree, parent->btc_elems,
	par_hdr->bth_count, buf, &idx), ==, NULL);
	ASSERT(idx.bti_before);
	uint32_t offset = idx.bti_offset;
	ASSERT3U(offset, <=, par_hdr->bth_count);
	ASSERT3P(parent->btc_children[offset], ==, old_node);

	/*
	* If the parent isn't full, shift things to accommodate our insertions
	* and return.
	*/
	if (par_hdr->bth_count != BTREE_CORE_ELEMS) {
	zfs_btree_insert_core_impl(tree, parent, offset, new_node, buf);
	return;
	}

	/*
	* We need to split this core node into two. Currently there are
	* BTREE_CORE_ELEMS + 1 child nodes, and we are adding one for
	* BTREE_CORE_ELEMS + 2. Some of the children will be part of the
	* current node, and the others will be moved to the new core node.
	* There are BTREE_CORE_ELEMS + 1 elements including the new one. One
	* will be used as the new separator in our parent, and the others
	* will be split among the two core nodes.
	*
	* Usually we will split the node in half evenly, with
	* BTREE_CORE_ELEMS/2 elements in each node. If we're bulk loading, we
	* instead move only about a quarter of the elements (and children) to
	* the new node. Since the average state after a long time is a 3/4
	* full node, shortcutting directly to that state improves efficiency.
	*
	* We do this in two stages: first we split into two nodes, and then we
	* reuse our existing logic to insert the new element and child.
	*/
	uint32_t move_count = MAX((BTREE_CORE_ELEMS / (tree->bt_bulk == NULL ?
	2 : 4)) - 1, 2);
	uint32_t keep_count = BTREE_CORE_ELEMS - move_count - 1;
	ASSERT3U(BTREE_CORE_ELEMS - move_count, >=, 2);
	tree->bt_num_nodes++;
	zfs_btree_core_t *new_parent = kmem_alloc(sizeof (zfs_btree_core_t) +
	BTREE_CORE_ELEMS * size, KM_SLEEP);
	zfs_btree_hdr_t *new_par_hdr = &new_parent->btc_hdr;
	new_par_hdr->bth_parent = par_hdr->bth_parent;
	new_par_hdr->bth_first = -1;
	new_par_hdr->bth_count = move_count;
	zfs_btree_poison_node(tree, new_par_hdr);

	par_hdr->bth_count = keep_count;

	bt_transfer_core(tree, parent, keep_count + 1, move_count, new_parent,
	0, BSS_TRAPEZOID);

	/* Store the new separator in a buffer. */
	uint8_t *tmp_buf = kmem_alloc(size, KM_SLEEP);
	bcpy(parent->btc_elems + keep_count * size, tmp_buf,
	size);
	zfs_btree_poison_node(tree, par_hdr);

	if (offset < keep_count) {
	/* Insert the new node into the left half */
	zfs_btree_insert_core_impl(tree, parent, offset, new_node,
	buf);

	/*
	* Move the new separator to the existing buffer.
	*/
	bcpy(tmp_buf, buf, size);
	} else if (offset > keep_count) {
	/* Insert the new node into the right half */
	new_node->bth_parent = new_parent;
	zfs_btree_insert_core_impl(tree, new_parent,
	offset - keep_count - 1, new_node, buf);

	/*
	* Move the new separator to the existing buffer.
	*/
	bcpy(tmp_buf, buf, size);
	} else {
	/*
	* Move the new separator into the right half, and replace it
	* with buf. We also need to shift back the elements in the
	* right half to accommodate new_node.
	*/
	bt_shift_core_right(tree, new_parent, 0, move_count,
	BSS_TRAPEZOID);
	new_parent->btc_children[0] = new_node;
	bcpy(tmp_buf, new_parent->btc_elems, size);
	new_par_hdr->bth_count++;
	}
	kmem_free(tmp_buf, size);
	zfs_btree_poison_node(tree, par_hdr);

	for (uint32_t i = 0; i <= new_parent->btc_hdr.bth_count; i++)
	new_parent->btc_children[i]->bth_parent = new_parent;

	for (uint32_t i = 0; i <= parent->btc_hdr.bth_count; i++)
	ASSERT3P(parent->btc_children[i]->bth_parent, ==, parent);

	/*
	* Now that the node is split, we need to insert the new node into its
	* parent. This may cause further splitting.
	*/
	zfs_btree_insert_into_parent(tree, &parent->btc_hdr,
	&new_parent->btc_hdr, buf);
	}

	/* Insert an element into a leaf node at the given offset. */
	static void
	zfs_btree_insert_leaf_impl(zfs_btree_t tree, zfs_btree_leaf_t leaf,
	uint32_t idx, const void *value)
	{
	size_t size = tree->bt_elem_size;
	zfs_btree_hdr_t *hdr = &leaf->btl_hdr;
	ASSERT3U(leaf->btl_hdr.bth_count, <, tree->bt_leaf_cap);

	if (zfs_btree_verify_intensity >= 5) {
	zfs_btree_verify_poison_at(tree, &leaf->btl_hdr,
	leaf->btl_hdr.bth_count);
	}

	bt_grow_leaf(tree, leaf, idx, 1);
	uint8_t start = leaf->btl_elems + (hdr->bth_first + idx) size;
	bcpy(value, start, size);
	}

	static void
	zfs_btree_verify_order_helper(zfs_btree_t tree, zfs_btree_hdr_t hdr);

	/* Helper function for inserting a new value into leaf at the given index. */
	static void
	zfs_btree_insert_into_leaf(zfs_btree_t tree, zfs_btree_leaf_t leaf,
	const void *value, uint32_t idx)
	{
	size_t size = tree->bt_elem_size;
	uint32_t capacity = tree->bt_leaf_cap;

	/*
	* If the leaf isn't full, shift the elements after idx and insert
	* value.
	*/
	if (leaf->btl_hdr.bth_count != capacity) {
	zfs_btree_insert_leaf_impl(tree, leaf, idx, value);
	return;
	}

	/*
	* Otherwise, we split the leaf node into two nodes. If we're not bulk
	* inserting, each is of size (capacity / 2). If we are bulk
	* inserting, we move a quarter of the elements to the new node so
	* inserts into the old node don't cause immediate splitting but the
	* tree stays relatively dense. Since the average state after a long
	* time is a 3/4 full node, shortcutting directly to that state
	* improves efficiency. At the end of the bulk insertion process
	* we'll need to go through and fix up any nodes (the last leaf and
	* its ancestors, potentially) that are below the minimum.
	*
	* In either case, we're left with one extra element. The leftover
	* element will become the new dividing element between the two nodes.
	*/
	uint32_t move_count = MAX(capacity / (tree->bt_bulk ? 4 : 2), 1) - 1;
	uint32_t keep_count = capacity - move_count - 1;
	ASSERT3U(keep_count, >=, 1);
	/* If we insert on left. move one more to keep leaves balanced. */
	if (idx < keep_count) {
	keep_count--;
	move_count++;
	}
	tree->bt_num_nodes++;
	zfs_btree_leaf_t *new_leaf = zfs_btree_leaf_alloc(tree);
	zfs_btree_hdr_t *new_hdr = &new_leaf->btl_hdr;
	new_hdr->bth_parent = leaf->btl_hdr.bth_parent;
	new_hdr->bth_first = (tree->bt_bulk ? 0 : capacity / 4) +
	(idx >= keep_count && idx <= keep_count + move_count / 2);
	new_hdr->bth_count = move_count;
	zfs_btree_poison_node(tree, new_hdr);

	if (tree->bt_bulk != NULL && leaf == tree->bt_bulk)
	tree->bt_bulk = new_leaf;

	/* Copy the back part to the new leaf. */
	bt_transfer_leaf(tree, leaf, keep_count + 1, move_count, new_leaf, 0);

	/* We store the new separator in a buffer we control for simplicity. */
	uint8_t *buf = kmem_alloc(size, KM_SLEEP);
	bcpy(leaf->btl_elems + (leaf->btl_hdr.bth_first + keep_count) * size,
	buf, size);

	bt_shrink_leaf(tree, leaf, keep_count, 1 + move_count);

	if (idx < keep_count) {
	/* Insert into the existing leaf. */
	zfs_btree_insert_leaf_impl(tree, leaf, idx, value);
	} else if (idx > keep_count) {
	/* Insert into the new leaf. */
	zfs_btree_insert_leaf_impl(tree, new_leaf, idx - keep_count -
	1, value);
	} else {
	/*
	* Insert planned separator into the new leaf, and use
	* the new value as the new separator.
	*/
	zfs_btree_insert_leaf_impl(tree, new_leaf, 0, buf);
	bcpy(value, buf, size);
	}

	/*
	* Now that the node is split, we need to insert the new node into its
	* parent. This may cause further splitting, bur only of core nodes.
	*/
	zfs_btree_insert_into_parent(tree, &leaf->btl_hdr, &new_leaf->btl_hdr,
	buf);
	kmem_free(buf, size);
	}

	static uint32_t
	zfs_btree_find_parent_idx(zfs_btree_t tree, zfs_btree_hdr_t hdr)
	{
	void *buf;
	if (zfs_btree_is_core(hdr)) {
	buf = ((zfs_btree_core_t *)hdr)->btc_elems;
	} else {
	buf = ((zfs_btree_leaf_t *)hdr)->btl_elems +
	hdr->bth_first * tree->bt_elem_size;
	}
	zfs_btree_index_t idx;
	zfs_btree_core_t *parent = hdr->bth_parent;
	VERIFY3P(tree->bt_find_in_buf(tree, parent->btc_elems,
	parent->btc_hdr.bth_count, buf, &idx), ==, NULL);
	ASSERT(idx.bti_before);
	ASSERT3U(idx.bti_offset, <=, parent->btc_hdr.bth_count);
	ASSERT3P(parent->btc_children[idx.bti_offset], ==, hdr);
	return (idx.bti_offset);
	}

	/*
	* Take the b-tree out of bulk insert mode. During bulk-insert mode, some
	* nodes may violate the invariant that non-root nodes must be at least half
	* full. All nodes violating this invariant should be the last node in their
	* particular level. To correct the invariant, we take values from their left
	* neighbor until they are half full. They must have a left neighbor at their
	* level because the last node at a level is not the first node unless it's
	* the root.
	*/
	static void
	zfs_btree_bulk_finish(zfs_btree_t *tree)
	{
	ASSERT3P(tree->bt_bulk, !=, NULL);
	ASSERT3P(tree->bt_root, !=, NULL);
	zfs_btree_leaf_t *leaf = tree->bt_bulk;
	zfs_btree_hdr_t *hdr = &leaf->btl_hdr;
	zfs_btree_core_t *parent = hdr->bth_parent;
	size_t size = tree->bt_elem_size;
	uint32_t capacity = tree->bt_leaf_cap;

	/*
	* The invariant doesn't apply to the root node, if that's the only
	* node in the tree we're done.
	*/
	if (parent == NULL) {
	tree->bt_bulk = NULL;
	return;
	}

	/* First, take elements to rebalance the leaf node. */
	if (hdr->bth_count < capacity / 2) {
	/*
	* First, find the left neighbor. The simplest way to do this
	* is to call zfs_btree_prev twice; the first time finds some
	* ancestor of this node, and the second time finds the left
	* neighbor. The ancestor found is the lowest common ancestor
	* of leaf and the neighbor.
	*/
	zfs_btree_index_t idx = {
	.bti_node = hdr,
	.bti_offset = 0
	};
	VERIFY3P(zfs_btree_prev(tree, &idx, &idx), !=, NULL);
	ASSERT(zfs_btree_is_core(idx.bti_node));
	zfs_btree_core_t common = (zfs_btree_core_t )idx.bti_node;
	uint32_t common_idx = idx.bti_offset;

	VERIFY3P(zfs_btree_prev(tree, &idx, &idx), !=, NULL);
	ASSERT(!zfs_btree_is_core(idx.bti_node));
	zfs_btree_leaf_t l_neighbor = (zfs_btree_leaf_t )idx.bti_node;
	zfs_btree_hdr_t *l_hdr = idx.bti_node;
	uint32_t move_count = (capacity / 2) - hdr->bth_count;
	ASSERT3U(l_neighbor->btl_hdr.bth_count - move_count, >=,
	capacity / 2);

	if (zfs_btree_verify_intensity >= 5) {
	for (uint32_t i = 0; i < move_count; i++) {
	zfs_btree_verify_poison_at(tree, hdr,
	leaf->btl_hdr.bth_count + i);
	}
	}

	/* First, shift elements in leaf back. */
	bt_grow_leaf(tree, leaf, 0, move_count);

	/* Next, move the separator from the common ancestor to leaf. */
	uint8_t separator = common->btc_elems + common_idx size;
	uint8_t *out = leaf->btl_elems +
	(hdr->bth_first + move_count - 1) * size;
	bcpy(separator, out, size);

	/*
	* Now we move elements from the tail of the left neighbor to
	* fill the remaining spots in leaf.
	*/
	bt_transfer_leaf(tree, l_neighbor, l_hdr->bth_count -
	(move_count - 1), move_count - 1, leaf, 0);

	/*
	* Finally, move the new last element in the left neighbor to
	* the separator.
	*/
	bcpy(l_neighbor->btl_elems + (l_hdr->bth_first +
	l_hdr->bth_count - move_count) * size, separator, size);

	/* Adjust the node's counts, and we're done. */
	bt_shrink_leaf(tree, l_neighbor, l_hdr->bth_count - move_count,
	move_count);

	ASSERT3U(l_hdr->bth_count, >=, capacity / 2);
	ASSERT3U(hdr->bth_count, >=, capacity / 2);
	}

	/*
	* Now we have to rebalance any ancestors of leaf that may also
	* violate the invariant.
	*/
	capacity = BTREE_CORE_ELEMS;
	while (parent->btc_hdr.bth_parent != NULL) {
	zfs_btree_core_t *cur = parent;
	zfs_btree_hdr_t *hdr = &cur->btc_hdr;
	parent = hdr->bth_parent;
	/*
	* If the invariant isn't violated, move on to the next
	* ancestor.
	*/
	if (hdr->bth_count >= capacity / 2)
	continue;

	/*
	* Because the smallest number of nodes we can move when
	* splitting is 2, we never need to worry about not having a
	* left sibling (a sibling is a neighbor with the same parent).
	*/
	uint32_t parent_idx = zfs_btree_find_parent_idx(tree, hdr);
	ASSERT3U(parent_idx, >, 0);
	zfs_btree_core_t *l_neighbor =
	(zfs_btree_core_t *)parent->btc_children[parent_idx - 1];
	uint32_t move_count = (capacity / 2) - hdr->bth_count;
	ASSERT3U(l_neighbor->btc_hdr.bth_count - move_count, >=,
	capacity / 2);

	if (zfs_btree_verify_intensity >= 5) {
	for (uint32_t i = 0; i < move_count; i++) {
	zfs_btree_verify_poison_at(tree, hdr,
	hdr->bth_count + i);
	}
	}
	/* First, shift things in the right node back. */
	bt_shift_core(tree, cur, 0, hdr->bth_count, move_count,
	BSS_TRAPEZOID, BSD_RIGHT);

	/* Next, move the separator to the right node. */
	uint8_t separator = parent->btc_elems + ((parent_idx - 1)
	size);
	uint8_t e_out = cur->btc_elems + ((move_count - 1) size);
	bcpy(separator, e_out, size);

	/*
	* Now, move elements and children from the left node to the
	* right. We move one more child than elements.
	*/
	move_count--;
	uint32_t move_idx = l_neighbor->btc_hdr.bth_count - move_count;
	bt_transfer_core(tree, l_neighbor, move_idx, move_count, cur, 0,
	BSS_TRAPEZOID);

	/*
	* Finally, move the last element in the left node to the
	* separator's position.
	*/
	move_idx--;
	bcpy(l_neighbor->btc_elems + move_idx * size, separator, size);

	l_neighbor->btc_hdr.bth_count -= move_count + 1;
	hdr->bth_count += move_count + 1;

	ASSERT3U(l_neighbor->btc_hdr.bth_count, >=, capacity / 2);
	ASSERT3U(hdr->bth_count, >=, capacity / 2);

	zfs_btree_poison_node(tree, &l_neighbor->btc_hdr);

	for (uint32_t i = 0; i <= hdr->bth_count; i++)
	cur->btc_children[i]->bth_parent = cur;
	}

	tree->bt_bulk = NULL;
	zfs_btree_verify(tree);
	}

	/*
	* Insert value into tree at the location specified by where.
	*/
	void
	zfs_btree_add_idx(zfs_btree_t tree, const void value,
	const zfs_btree_index_t *where)
	{
	zfs_btree_index_t idx = {0};

	/* If we're not inserting in the last leaf, end bulk insert mode. */
	if (tree->bt_bulk != NULL) {
	if (where->bti_node != &tree->bt_bulk->btl_hdr) {
	zfs_btree_bulk_finish(tree);
	VERIFY3P(zfs_btree_find(tree, value, &idx), ==, NULL);
	where = &idx;
	}
	}

	tree->bt_num_elems++;
	/*
	* If this is the first element in the tree, create a leaf root node
	* and add the value to it.
	*/
	if (where->bti_node == NULL) {
	ASSERT3U(tree->bt_num_elems, ==, 1);
	ASSERT3S(tree->bt_height, ==, -1);
	ASSERT3P(tree->bt_root, ==, NULL);
	ASSERT0(where->bti_offset);

	tree->bt_num_nodes++;
	zfs_btree_leaf_t *leaf = zfs_btree_leaf_alloc(tree);
	tree->bt_root = &leaf->btl_hdr;
	tree->bt_height++;

	zfs_btree_hdr_t *hdr = &leaf->btl_hdr;
	hdr->bth_parent = NULL;
	hdr->bth_first = 0;
	hdr->bth_count = 0;
	zfs_btree_poison_node(tree, hdr);

	zfs_btree_insert_into_leaf(tree, leaf, value, 0);
	tree->bt_bulk = leaf;
	} else if (!zfs_btree_is_core(where->bti_node)) {
	/*
	* If we're inserting into a leaf, go directly to the helper
	* function.
	*/
	zfs_btree_insert_into_leaf(tree,
	(zfs_btree_leaf_t *)where->bti_node, value,
	where->bti_offset);
	} else {
	/*
	* If we're inserting into a core node, we can't just shift
	* the existing element in that slot in the same node without
	* breaking our ordering invariants. Instead we place the new
	* value in the node at that spot and then insert the old
	* separator into the first slot in the subtree to the right.
	*/
	zfs_btree_core_t node = (zfs_btree_core_t )where->bti_node;

	/*
	* We can ignore bti_before, because either way the value
	* should end up in bti_offset.
	*/
	uint32_t off = where->bti_offset;
	zfs_btree_hdr_t *subtree = node->btc_children[off + 1];
	size_t size = tree->bt_elem_size;
	uint8_t *buf = kmem_alloc(size, KM_SLEEP);
	bcpy(node->btc_elems + off * size, buf, size);
	bcpy(value, node->btc_elems + off * size, size);

	/*
	* Find the first slot in the subtree to the right, insert
	* there.
	*/
	zfs_btree_index_t new_idx;
	VERIFY3P(zfs_btree_first_helper(tree, subtree, &new_idx), !=,
	NULL);
	ASSERT0(new_idx.bti_offset);
	ASSERT(!zfs_btree_is_core(new_idx.bti_node));
	zfs_btree_insert_into_leaf(tree,
	(zfs_btree_leaf_t *)new_idx.bti_node, buf, 0);
	kmem_free(buf, size);
	}
	zfs_btree_verify(tree);
	}

	/*
	* Return the first element in the tree, and put its location in where if
	* non-null.
	*/
	void *
	zfs_btree_first(zfs_btree_t tree, zfs_btree_index_t where)
	{
	if (tree->bt_height == -1) {
	ASSERT0(tree->bt_num_elems);
	return (NULL);
	}
	return (zfs_btree_first_helper(tree, tree->bt_root, where));
	}

	/*
	* Find the last element in the subtree rooted at hdr, return its value and
	* put its location in where if non-null.
	*/
	static void *
	zfs_btree_last_helper(zfs_btree_t btree, zfs_btree_hdr_t hdr,
	zfs_btree_index_t *where)
	{
	zfs_btree_hdr_t *node;

	for (node = hdr; zfs_btree_is_core(node); node =
	((zfs_btree_core_t *)node)->btc_children[node->bth_count])
	;

	zfs_btree_leaf_t leaf = (zfs_btree_leaf_t )node;
	if (where != NULL) {
	where->bti_node = node;
	where->bti_offset = node->bth_count - 1;
	where->bti_before = B_FALSE;
	}
	return (leaf->btl_elems + (node->bth_first + node->bth_count - 1) *
	btree->bt_elem_size);
	}

	/*
	* Return the last element in the tree, and put its location in where if
	* non-null.
	*/
	void *
	zfs_btree_last(zfs_btree_t tree, zfs_btree_index_t where)
	{
	if (tree->bt_height == -1) {
	ASSERT0(tree->bt_num_elems);
	return (NULL);
	}
	return (zfs_btree_last_helper(tree, tree->bt_root, where));
	}

	/*
	* This function contains the logic to find the next node in the tree. A
	* helper function is used because there are multiple internal consumemrs of
	* this logic. The done_func is used by zfs_btree_destroy_nodes to clean up each
	* node after we've finished with it.
	*/
	static void *
	zfs_btree_next_helper(zfs_btree_t tree, const zfs_btree_index_t idx,
	zfs_btree_index_t *out_idx,
	void (done_func)(zfs_btree_t , zfs_btree_hdr_t *))
	{
	if (idx->bti_node == NULL) {
	ASSERT3S(tree->bt_height, ==, -1);
	return (NULL);
	}

	uint32_t offset = idx->bti_offset;
	if (!zfs_btree_is_core(idx->bti_node)) {
	/*
	* When finding the next element of an element in a leaf,
	* there are two cases. If the element isn't the last one in
	* the leaf, in which case we just return the next element in
	* the leaf. Otherwise, we need to traverse up our parents
	* until we find one where our ancestor isn't the last child
	* of its parent. Once we do, the next element is the
	* separator after our ancestor in its parent.
	*/
	zfs_btree_leaf_t leaf = (zfs_btree_leaf_t )idx->bti_node;
	uint32_t new_off = offset + (idx->bti_before ? 0 : 1);
	if (leaf->btl_hdr.bth_count > new_off) {
	out_idx->bti_node = &leaf->btl_hdr;
	out_idx->bti_offset = new_off;
	out_idx->bti_before = B_FALSE;
	return (leaf->btl_elems + (leaf->btl_hdr.bth_first +
	new_off) * tree->bt_elem_size);
	}

	zfs_btree_hdr_t *prev = &leaf->btl_hdr;
	for (zfs_btree_core_t *node = leaf->btl_hdr.bth_parent;
	node != NULL; node = node->btc_hdr.bth_parent) {
	zfs_btree_hdr_t *hdr = &node->btc_hdr;
	ASSERT(zfs_btree_is_core(hdr));
	uint32_t i = zfs_btree_find_parent_idx(tree, prev);
	if (done_func != NULL)
	done_func(tree, prev);
	if (i == hdr->bth_count) {
	prev = hdr;
	continue;
	}
	out_idx->bti_node = hdr;
	out_idx->bti_offset = i;
	out_idx->bti_before = B_FALSE;
	return (node->btc_elems + i * tree->bt_elem_size);
	}
	if (done_func != NULL)
	done_func(tree, prev);
	/*
	* We've traversed all the way up and been at the end of the
	* node every time, so this was the last element in the tree.
	*/
	return (NULL);
	}

	/* If we were before an element in a core node, return that element. */
	ASSERT(zfs_btree_is_core(idx->bti_node));
	zfs_btree_core_t node = (zfs_btree_core_t )idx->bti_node;
	if (idx->bti_before) {
	out_idx->bti_before = B_FALSE;
	return (node->btc_elems + offset * tree->bt_elem_size);
	}

	/*
	* The next element from one in a core node is the first element in
	* the subtree just to the right of the separator.
	*/
	zfs_btree_hdr_t *child = node->btc_children[offset + 1];
	return (zfs_btree_first_helper(tree, child, out_idx));
	}

	/*
	* Return the next valued node in the tree. The same address can be safely
	* passed for idx and out_idx.
	*/
	void *
	zfs_btree_next(zfs_btree_t tree, const zfs_btree_index_t idx,
	zfs_btree_index_t *out_idx)
	{
	return (zfs_btree_next_helper(tree, idx, out_idx, NULL));
	}

	/*
	* Return the previous valued node in the tree. The same value can be safely
	* passed for idx and out_idx.
	*/
	void *
	zfs_btree_prev(zfs_btree_t tree, const zfs_btree_index_t idx,
	zfs_btree_index_t *out_idx)
	{
	if (idx->bti_node == NULL) {
	ASSERT3S(tree->bt_height, ==, -1);
	return (NULL);
	}

	uint32_t offset = idx->bti_offset;
	if (!zfs_btree_is_core(idx->bti_node)) {
	/*
	* When finding the previous element of an element in a leaf,
	* there are two cases. If the element isn't the first one in
	* the leaf, in which case we just return the previous element
	* in the leaf. Otherwise, we need to traverse up our parents
	* until we find one where our previous ancestor isn't the
	* first child. Once we do, the previous element is the
	* separator after our previous ancestor.
	*/
	zfs_btree_leaf_t leaf = (zfs_btree_leaf_t )idx->bti_node;
	if (offset != 0) {
	out_idx->bti_node = &leaf->btl_hdr;
	out_idx->bti_offset = offset - 1;
	out_idx->bti_before = B_FALSE;
	return (leaf->btl_elems + (leaf->btl_hdr.bth_first +
	offset - 1) * tree->bt_elem_size);
	}
	zfs_btree_hdr_t *prev = &leaf->btl_hdr;
	for (zfs_btree_core_t *node = leaf->btl_hdr.bth_parent;
	node != NULL; node = node->btc_hdr.bth_parent) {
	zfs_btree_hdr_t *hdr = &node->btc_hdr;
	ASSERT(zfs_btree_is_core(hdr));
	uint32_t i = zfs_btree_find_parent_idx(tree, prev);
	if (i == 0) {
	prev = hdr;
	continue;
	}
	out_idx->bti_node = hdr;
	out_idx->bti_offset = i - 1;
	out_idx->bti_before = B_FALSE;
	return (node->btc_elems + (i - 1) * tree->bt_elem_size);
	}
	/*
	* We've traversed all the way up and been at the start of the
	* node every time, so this was the first node in the tree.
	*/
	return (NULL);
	}

	/*
	* The previous element from one in a core node is the last element in
	* the subtree just to the left of the separator.
	*/
	ASSERT(zfs_btree_is_core(idx->bti_node));
	zfs_btree_core_t node = (zfs_btree_core_t )idx->bti_node;
	zfs_btree_hdr_t *child = node->btc_children[offset];
	return (zfs_btree_last_helper(tree, child, out_idx));
	}

	/*
	* Get the value at the provided index in the tree.
	*
	* Note that the value returned from this function can be mutated, but only
	* if it will not change the ordering of the element with respect to any other
	* elements that could be in the tree.
	*/
	void *
	zfs_btree_get(zfs_btree_t tree, zfs_btree_index_t idx)
	{
	ASSERT(!idx->bti_before);
	size_t size = tree->bt_elem_size;
	if (!zfs_btree_is_core(idx->bti_node)) {
	zfs_btree_leaf_t leaf = (zfs_btree_leaf_t )idx->bti_node;
	return (leaf->btl_elems + (leaf->btl_hdr.bth_first +
	idx->bti_offset) * size);
	}
	zfs_btree_core_t node = (zfs_btree_core_t )idx->bti_node;
	return (node->btc_elems + idx->bti_offset * size);
	}

	/* Add the given value to the tree. Must not already be in the tree. */
	void
	zfs_btree_add(zfs_btree_t tree, const void node)
	{
	zfs_btree_index_t where = {0};
	VERIFY3P(zfs_btree_find(tree, node, &where), ==, NULL);
	zfs_btree_add_idx(tree, node, &where);
	}

	/* Helper function to free a tree node. */
	static void
	zfs_btree_node_destroy(zfs_btree_t tree, zfs_btree_hdr_t node)
	{
	tree->bt_num_nodes--;
	if (!zfs_btree_is_core(node)) {
	zfs_btree_leaf_free(tree, node);
	} else {
	kmem_free(node, sizeof (zfs_btree_core_t) +
	BTREE_CORE_ELEMS * tree->bt_elem_size);
	}
	}

	/*
	* Remove the rm_hdr and the separator to its left from the parent node. The
	* buffer that rm_hdr was stored in may already be freed, so its contents
	* cannot be accessed.
	*/
	static void
	zfs_btree_remove_from_node(zfs_btree_t tree, zfs_btree_core_t node,
	zfs_btree_hdr_t *rm_hdr)
	{
	size_t size = tree->bt_elem_size;
	uint32_t min_count = (BTREE_CORE_ELEMS / 2) - 1;
	zfs_btree_hdr_t *hdr = &node->btc_hdr;
	/*
	* If the node is the root node and rm_hdr is one of two children,
	* promote the other child to the root.
	*/
	if (hdr->bth_parent == NULL && hdr->bth_count <= 1) {
	ASSERT3U(hdr->bth_count, ==, 1);
	ASSERT3P(tree->bt_root, ==, node);
	ASSERT3P(node->btc_children[1], ==, rm_hdr);
	tree->bt_root = node->btc_children[0];
	node->btc_children[0]->bth_parent = NULL;
	zfs_btree_node_destroy(tree, hdr);
	tree->bt_height--;
	return;
	}

	uint32_t idx;
	for (idx = 0; idx <= hdr->bth_count; idx++) {
	if (node->btc_children[idx] == rm_hdr)
	break;
	}
	ASSERT3U(idx, <=, hdr->bth_count);

	/*
	* If the node is the root or it has more than the minimum number of
	* children, just remove the child and separator, and return.
	*/
	if (hdr->bth_parent == NULL \|\|
	hdr->bth_count > min_count) {
	/*
	* Shift the element and children to the right of rm_hdr to
	* the left by one spot.
	*/
	bt_shift_core_left(tree, node, idx, hdr->bth_count - idx,
	BSS_PARALLELOGRAM);
	hdr->bth_count--;
	zfs_btree_poison_node_at(tree, hdr, hdr->bth_count, 1);
	return;
	}

	ASSERT3U(hdr->bth_count, ==, min_count);

	/*
	* Now we try to take a node from a neighbor. We check left, then
	* right. If the neighbor exists and has more than the minimum number
	* of elements, we move the separator between us and them to our
	* node, move their closest element (last for left, first for right)
	* to the separator, and move their closest child to our node. Along
	* the way we need to collapse the gap made by idx, and (for our right
	* neighbor) the gap made by removing their first element and child.
	*
	* Note: this logic currently doesn't support taking from a neighbor
	* that isn't a sibling (i.e. a neighbor with a different
	* parent). This isn't critical functionality, but may be worth
	* implementing in the future for completeness' sake.
	*/
	zfs_btree_core_t *parent = hdr->bth_parent;
	uint32_t parent_idx = zfs_btree_find_parent_idx(tree, hdr);

	zfs_btree_hdr_t *l_hdr = (parent_idx == 0 ? NULL :
	parent->btc_children[parent_idx - 1]);
	if (l_hdr != NULL && l_hdr->bth_count > min_count) {
	/* We can take a node from the left neighbor. */
	ASSERT(zfs_btree_is_core(l_hdr));
	zfs_btree_core_t neighbor = (zfs_btree_core_t )l_hdr;

	/*
	* Start by shifting the elements and children in the current
	* node to the right by one spot.
	*/
	bt_shift_core_right(tree, node, 0, idx - 1, BSS_TRAPEZOID);

	/*
	* Move the separator between node and neighbor to the first
	* element slot in the current node.
	*/
	uint8_t separator = parent->btc_elems + (parent_idx - 1)
	size;
	bcpy(separator, node->btc_elems, size);

	/* Move the last child of neighbor to our first child slot. */
	node->btc_children[0] =
	neighbor->btc_children[l_hdr->bth_count];
	node->btc_children[0]->bth_parent = node;

	/* Move the last element of neighbor to the separator spot. */
	uint8_t *take_elem = neighbor->btc_elems +
	(l_hdr->bth_count - 1) * size;
	bcpy(take_elem, separator, size);
	l_hdr->bth_count--;
	zfs_btree_poison_node_at(tree, l_hdr, l_hdr->bth_count, 1);
	return;
	}

	zfs_btree_hdr_t *r_hdr = (parent_idx == parent->btc_hdr.bth_count ?
	NULL : parent->btc_children[parent_idx + 1]);
	if (r_hdr != NULL && r_hdr->bth_count > min_count) {
	/* We can take a node from the right neighbor. */
	ASSERT(zfs_btree_is_core(r_hdr));
	zfs_btree_core_t neighbor = (zfs_btree_core_t )r_hdr;

	/*
	* Shift elements in node left by one spot to overwrite rm_hdr
	* and the separator before it.
	*/
	bt_shift_core_left(tree, node, idx, hdr->bth_count - idx,
	BSS_PARALLELOGRAM);

	/*
	* Move the separator between node and neighbor to the last
	* element spot in node.
	*/
	uint8_t separator = parent->btc_elems + parent_idx size;
	bcpy(separator, node->btc_elems + (hdr->bth_count - 1) * size,
	size);

	/*
	* Move the first child of neighbor to the last child spot in
	* node.
	*/
	node->btc_children[hdr->bth_count] = neighbor->btc_children[0];
	node->btc_children[hdr->bth_count]->bth_parent = node;

	/* Move the first element of neighbor to the separator spot. */
	uint8_t *take_elem = neighbor->btc_elems;
	bcpy(take_elem, separator, size);
	r_hdr->bth_count--;

	/*
	* Shift the elements and children of neighbor to cover the
	* stolen elements.
	*/
	bt_shift_core_left(tree, neighbor, 1, r_hdr->bth_count,
	BSS_TRAPEZOID);
	zfs_btree_poison_node_at(tree, r_hdr, r_hdr->bth_count, 1);
	return;
	}

	/*
	* In this case, neither of our neighbors can spare an element, so we
	* need to merge with one of them. We prefer the left one,
	* arbitrarily. Move the separator into the leftmost merging node
	* (which may be us or the left neighbor), and then move the right
	* merging node's elements. Once that's done, we go back and delete
	* the element we're removing. Finally, go into the parent and delete
	* the right merging node and the separator. This may cause further
	* merging.
	*/
	zfs_btree_hdr_t new_rm_hdr, keep_hdr;
	uint32_t new_idx = idx;
	if (l_hdr != NULL) {
	keep_hdr = l_hdr;
	new_rm_hdr = hdr;
	new_idx += keep_hdr->bth_count + 1;
	} else {
	ASSERT3P(r_hdr, !=, NULL);
	keep_hdr = hdr;
	new_rm_hdr = r_hdr;
	parent_idx++;
	}

	ASSERT(zfs_btree_is_core(keep_hdr));
	ASSERT(zfs_btree_is_core(new_rm_hdr));

	zfs_btree_core_t keep = (zfs_btree_core_t )keep_hdr;
	zfs_btree_core_t rm = (zfs_btree_core_t )new_rm_hdr;

	if (zfs_btree_verify_intensity >= 5) {
	for (uint32_t i = 0; i < new_rm_hdr->bth_count + 1; i++) {
	zfs_btree_verify_poison_at(tree, keep_hdr,
	keep_hdr->bth_count + i);
	}
	}

	/* Move the separator into the left node. */
	uint8_t e_out = keep->btc_elems + keep_hdr->bth_count size;
	uint8_t separator = parent->btc_elems + (parent_idx - 1)
	size;
	bcpy(separator, e_out, size);
	keep_hdr->bth_count++;

	/* Move all our elements and children into the left node. */
	bt_transfer_core(tree, rm, 0, new_rm_hdr->bth_count, keep,
	keep_hdr->bth_count, BSS_TRAPEZOID);

	uint32_t old_count = keep_hdr->bth_count;

	/* Update bookkeeping */
	keep_hdr->bth_count += new_rm_hdr->bth_count;
	ASSERT3U(keep_hdr->bth_count, ==, (min_count * 2) + 1);

	/*
	* Shift the element and children to the right of rm_hdr to
	* the left by one spot.
	*/
	ASSERT3P(keep->btc_children[new_idx], ==, rm_hdr);
	bt_shift_core_left(tree, keep, new_idx, keep_hdr->bth_count - new_idx,
	BSS_PARALLELOGRAM);
	keep_hdr->bth_count--;

	/* Reparent all our children to point to the left node. */
	zfs_btree_hdr_t **new_start = keep->btc_children +
	old_count - 1;
	for (uint32_t i = 0; i < new_rm_hdr->bth_count + 1; i++)
	new_start[i]->bth_parent = keep;
	for (uint32_t i = 0; i <= keep_hdr->bth_count; i++) {
	ASSERT3P(keep->btc_children[i]->bth_parent, ==, keep);
	ASSERT3P(keep->btc_children[i], !=, rm_hdr);
	}
	zfs_btree_poison_node_at(tree, keep_hdr, keep_hdr->bth_count, 1);

	new_rm_hdr->bth_count = 0;
	zfs_btree_remove_from_node(tree, parent, new_rm_hdr);
	zfs_btree_node_destroy(tree, new_rm_hdr);
	}

	/* Remove the element at the specific location. */
	void
	zfs_btree_remove_idx(zfs_btree_t tree, zfs_btree_index_t where)
	{
	size_t size = tree->bt_elem_size;
	zfs_btree_hdr_t *hdr = where->bti_node;
	uint32_t idx = where->bti_offset;

	ASSERT(!where->bti_before);
	if (tree->bt_bulk != NULL) {
	/*
	* Leave bulk insert mode. Note that our index would be
	* invalid after we correct the tree, so we copy the value
	* we're planning to remove and find it again after
	* bulk_finish.
	*/
	uint8_t *value = zfs_btree_get(tree, where);
	uint8_t *tmp = kmem_alloc(size, KM_SLEEP);
	bcpy(value, tmp, size);
	zfs_btree_bulk_finish(tree);
	VERIFY3P(zfs_btree_find(tree, tmp, where), !=, NULL);
	kmem_free(tmp, size);
	hdr = where->bti_node;
	idx = where->bti_offset;
	}

	tree->bt_num_elems--;
	/*
	* If the element happens to be in a core node, we move a leaf node's
	* element into its place and then remove the leaf node element. This
	* makes the rebalance logic not need to be recursive both upwards and
	* downwards.
	*/
	if (zfs_btree_is_core(hdr)) {
	zfs_btree_core_t node = (zfs_btree_core_t )hdr;
	zfs_btree_hdr_t *left_subtree = node->btc_children[idx];
	void *new_value = zfs_btree_last_helper(tree, left_subtree,
	where);
	ASSERT3P(new_value, !=, NULL);

	bcpy(new_value, node->btc_elems + idx * size, size);

	hdr = where->bti_node;
	idx = where->bti_offset;
	ASSERT(!where->bti_before);
	}

	/*
	* First, we'll update the leaf's metadata. Then, we shift any
	* elements after the idx to the left. After that, we rebalance if
	* needed.
	*/
	ASSERT(!zfs_btree_is_core(hdr));
	zfs_btree_leaf_t leaf = (zfs_btree_leaf_t )hdr;
	ASSERT3U(hdr->bth_count, >, 0);

	uint32_t min_count = (tree->bt_leaf_cap / 2) - 1;

	/*
	* If we're over the minimum size or this is the root, just overwrite
	* the value and return.
	*/
	if (hdr->bth_count > min_count \|\| hdr->bth_parent == NULL) {
	bt_shrink_leaf(tree, leaf, idx, 1);
	if (hdr->bth_parent == NULL) {
	ASSERT0(tree->bt_height);
	if (hdr->bth_count == 0) {
	tree->bt_root = NULL;
	tree->bt_height--;
	zfs_btree_node_destroy(tree, &leaf->btl_hdr);
	}
	}
	zfs_btree_verify(tree);
	return;
	}
	ASSERT3U(hdr->bth_count, ==, min_count);

	/*
	* Now we try to take a node from a sibling. We check left, then
	* right. If they exist and have more than the minimum number of
	* elements, we move the separator between us and them to our node
	* and move their closest element (last for left, first for right) to
	* the separator. Along the way we need to collapse the gap made by
	* idx, and (for our right neighbor) the gap made by removing their
	* first element.
	*
	* Note: this logic currently doesn't support taking from a neighbor
	* that isn't a sibling. This isn't critical functionality, but may be
	* worth implementing in the future for completeness' sake.
	*/
	zfs_btree_core_t *parent = hdr->bth_parent;
	uint32_t parent_idx = zfs_btree_find_parent_idx(tree, hdr);

	zfs_btree_hdr_t *l_hdr = (parent_idx == 0 ? NULL :
	parent->btc_children[parent_idx - 1]);
	if (l_hdr != NULL && l_hdr->bth_count > min_count) {
	/* We can take a node from the left neighbor. */
	ASSERT(!zfs_btree_is_core(l_hdr));
	zfs_btree_leaf_t neighbor = (zfs_btree_leaf_t )l_hdr;

	/*
	* Move our elements back by one spot to make room for the
	* stolen element and overwrite the element being removed.
	*/
	bt_shift_leaf(tree, leaf, 0, idx, 1, BSD_RIGHT);

	/* Move the separator to our first spot. */
	uint8_t separator = parent->btc_elems + (parent_idx - 1)
	size;
	bcpy(separator, leaf->btl_elems + hdr->bth_first * size, size);

	/* Move our neighbor's last element to the separator. */
	uint8_t *take_elem = neighbor->btl_elems +
	(l_hdr->bth_first + l_hdr->bth_count - 1) * size;
	bcpy(take_elem, separator, size);

	/* Delete our neighbor's last element. */
	bt_shrink_leaf(tree, neighbor, l_hdr->bth_count - 1, 1);
	zfs_btree_verify(tree);
	return;
	}

	zfs_btree_hdr_t *r_hdr = (parent_idx == parent->btc_hdr.bth_count ?
	NULL : parent->btc_children[parent_idx + 1]);
	if (r_hdr != NULL && r_hdr->bth_count > min_count) {
	/* We can take a node from the right neighbor. */
	ASSERT(!zfs_btree_is_core(r_hdr));
	zfs_btree_leaf_t neighbor = (zfs_btree_leaf_t )r_hdr;

	/*
	* Move our elements after the element being removed forwards
	* by one spot to make room for the stolen element and
	* overwrite the element being removed.
	*/
	bt_shift_leaf(tree, leaf, idx + 1, hdr->bth_count - idx - 1,
	1, BSD_LEFT);

	/* Move the separator between us to our last spot. */
	uint8_t separator = parent->btc_elems + parent_idx size;
	bcpy(separator, leaf->btl_elems + (hdr->bth_first +
	hdr->bth_count - 1) * size, size);

	/* Move our neighbor's first element to the separator. */
	uint8_t *take_elem = neighbor->btl_elems +
	r_hdr->bth_first * size;
	bcpy(take_elem, separator, size);

	/* Delete our neighbor's first element. */
	bt_shrink_leaf(tree, neighbor, 0, 1);
	zfs_btree_verify(tree);
	return;
	}

	/*
	* In this case, neither of our neighbors can spare an element, so we
	* need to merge with one of them. We prefer the left one, arbitrarily.
	* After remove we move the separator into the leftmost merging node
	* (which may be us or the left neighbor), and then move the right
	* merging node's elements. Once that's done, we go back and delete
	* the element we're removing. Finally, go into the parent and delete
	* the right merging node and the separator. This may cause further
	* merging.
	*/
	zfs_btree_hdr_t rm_hdr, k_hdr;
	if (l_hdr != NULL) {
	k_hdr = l_hdr;
	rm_hdr = hdr;
	} else {
	ASSERT3P(r_hdr, !=, NULL);
	k_hdr = hdr;
	rm_hdr = r_hdr;
	parent_idx++;
	}
	ASSERT(!zfs_btree_is_core(k_hdr));
	ASSERT(!zfs_btree_is_core(rm_hdr));
	ASSERT3U(k_hdr->bth_count, ==, min_count);
	ASSERT3U(rm_hdr->bth_count, ==, min_count);
	zfs_btree_leaf_t keep = (zfs_btree_leaf_t )k_hdr;
	zfs_btree_leaf_t rm = (zfs_btree_leaf_t )rm_hdr;

	if (zfs_btree_verify_intensity >= 5) {
	for (uint32_t i = 0; i < rm_hdr->bth_count + 1; i++) {
	zfs_btree_verify_poison_at(tree, k_hdr,
	k_hdr->bth_count + i);
	}
	}

	/*
	* Remove the value from the node. It will go below the minimum,
	* but we'll fix it in no time.
	*/
	bt_shrink_leaf(tree, leaf, idx, 1);

	/* Prepare space for elements to be moved from the right. */
	uint32_t k_count = k_hdr->bth_count;
	bt_grow_leaf(tree, keep, k_count, 1 + rm_hdr->bth_count);
	ASSERT3U(k_hdr->bth_count, ==, min_count * 2);

	/* Move the separator into the first open spot. */
	uint8_t out = keep->btl_elems + (k_hdr->bth_first + k_count) size;
	uint8_t separator = parent->btc_elems + (parent_idx - 1) size;
	bcpy(separator, out, size);

	/* Move our elements to the left neighbor. */
	bt_transfer_leaf(tree, rm, 0, rm_hdr->bth_count, keep, k_count + 1);

	/* Remove the emptied node from the parent. */
	zfs_btree_remove_from_node(tree, parent, rm_hdr);
	zfs_btree_node_destroy(tree, rm_hdr);
	zfs_btree_verify(tree);
	}

	/* Remove the given value from the tree. */
	void
	zfs_btree_remove(zfs_btree_t tree, const void value)
	{
	zfs_btree_index_t where = {0};
	VERIFY3P(zfs_btree_find(tree, value, &where), !=, NULL);
	zfs_btree_remove_idx(tree, &where);
	}

	/* Return the number of elements in the tree. */
	ulong_t
	zfs_btree_numnodes(zfs_btree_t *tree)
	{
	return (tree->bt_num_elems);
	}

	/*
	* This function is used to visit all the elements in the tree before
	* destroying the tree. This allows the calling code to perform any cleanup it
	* needs to do. This is more efficient than just removing the first element
	* over and over, because it removes all rebalancing. Once the destroy_nodes()
	* function has been called, no other btree operations are valid until it
	* returns NULL, which point the only valid operation is zfs_btree_destroy().
	*
	* example:
	*
	* zfs_btree_index_t *cookie = NULL;
	* my_data_t *node;
	*
	* while ((node = zfs_btree_destroy_nodes(tree, &cookie)) != NULL)
	* free(node->ptr);
	* zfs_btree_destroy(tree);
	*
	*/
	void *
	zfs_btree_destroy_nodes(zfs_btree_t tree, zfs_btree_index_t *cookie)
	{
	if (*cookie == NULL) {
	if (tree->bt_height == -1)
	return (NULL);
	cookie = kmem_alloc(sizeof (*cookie), KM_SLEEP);
	return (zfs_btree_first(tree, *cookie));
	}

	void rval = zfs_btree_next_helper(tree, cookie, *cookie,
	zfs_btree_node_destroy);
	if (rval == NULL) {
	tree->bt_root = NULL;
	tree->bt_height = -1;
	tree->bt_num_elems = 0;
	kmem_free(cookie, sizeof (*cookie));
	tree->bt_bulk = NULL;
	}
	return (rval);
	}

	static void
	zfs_btree_clear_helper(zfs_btree_t tree, zfs_btree_hdr_t hdr)
	{
	if (zfs_btree_is_core(hdr)) {
	zfs_btree_core_t btc = (zfs_btree_core_t )hdr;
	for (uint32_t i = 0; i <= hdr->bth_count; i++)
	zfs_btree_clear_helper(tree, btc->btc_children[i]);
	}

	zfs_btree_node_destroy(tree, hdr);
	}

	void
	zfs_btree_clear(zfs_btree_t *tree)
	{
	if (tree->bt_root == NULL) {
	ASSERT0(tree->bt_num_elems);
	return;
	}

	zfs_btree_clear_helper(tree, tree->bt_root);
	tree->bt_num_elems = 0;
	tree->bt_root = NULL;
	tree->bt_num_nodes = 0;
	tree->bt_height = -1;
	tree->bt_bulk = NULL;
	}

	void
	zfs_btree_destroy(zfs_btree_t *tree)
	{
	ASSERT0(tree->bt_num_elems);
	ASSERT3P(tree->bt_root, ==, NULL);
	}

	/* Verify that every child of this node has the correct parent pointer. */
	static void
	zfs_btree_verify_pointers_helper(zfs_btree_t tree, zfs_btree_hdr_t hdr)
	{
	if (!zfs_btree_is_core(hdr))
	return;

	zfs_btree_core_t node = (zfs_btree_core_t )hdr;
	for (uint32_t i = 0; i <= hdr->bth_count; i++) {
	VERIFY3P(node->btc_children[i]->bth_parent, ==, hdr);
	zfs_btree_verify_pointers_helper(tree, node->btc_children[i]);
	}
	}

	/* Verify that every node has the correct parent pointer. */
	static void
	zfs_btree_verify_pointers(zfs_btree_t *tree)
	{
	if (tree->bt_height == -1) {
	VERIFY3P(tree->bt_root, ==, NULL);
	return;
	}
	VERIFY3P(tree->bt_root->bth_parent, ==, NULL);
	zfs_btree_verify_pointers_helper(tree, tree->bt_root);
	}

	/*
	* Verify that all the current node and its children satisfy the count
	* invariants, and return the total count in the subtree rooted in this node.
	*/
	static uint64_t
	zfs_btree_verify_counts_helper(zfs_btree_t tree, zfs_btree_hdr_t hdr)
	{
	if (!zfs_btree_is_core(hdr)) {
	if (tree->bt_root != hdr && tree->bt_bulk &&
	hdr != &tree->bt_bulk->btl_hdr) {
	VERIFY3U(hdr->bth_count, >=, tree->bt_leaf_cap / 2 - 1);
	}

	return (hdr->bth_count);
	} else {

	zfs_btree_core_t node = (zfs_btree_core_t )hdr;
	uint64_t ret = hdr->bth_count;
	if (tree->bt_root != hdr && tree->bt_bulk == NULL)
	VERIFY3P(hdr->bth_count, >=, BTREE_CORE_ELEMS / 2 - 1);
	for (uint32_t i = 0; i <= hdr->bth_count; i++) {
	ret += zfs_btree_verify_counts_helper(tree,
	node->btc_children[i]);
	}

	return (ret);
	}
	}

	/*
	* Verify that all nodes satisfy the invariants and that the total number of
	* elements is correct.
	*/
	static void
	zfs_btree_verify_counts(zfs_btree_t *tree)
	{
	EQUIV(tree->bt_num_elems == 0, tree->bt_height == -1);
	if (tree->bt_height == -1) {
	return;
	}
	VERIFY3P(zfs_btree_verify_counts_helper(tree, tree->bt_root), ==,
	tree->bt_num_elems);
	}

	/*
	* Check that the subtree rooted at this node has a uniform height. Returns
	* the number of nodes under this node, to help verify bt_num_nodes.
	*/
	static uint64_t
	zfs_btree_verify_height_helper(zfs_btree_t tree, zfs_btree_hdr_t hdr,
	int32_t height)
	{
	if (!zfs_btree_is_core(hdr)) {
	VERIFY0(height);
	return (1);
	}

	zfs_btree_core_t node = (zfs_btree_core_t )hdr;
	uint64_t ret = 1;
	for (uint32_t i = 0; i <= hdr->bth_count; i++) {
	ret += zfs_btree_verify_height_helper(tree,
	node->btc_children[i], height - 1);
	}
	return (ret);
	}

	/*
	* Check that the tree rooted at this node has a uniform height, and that the
	* bt_height in the tree is correct.
	*/
	static void
	zfs_btree_verify_height(zfs_btree_t *tree)
	{
	EQUIV(tree->bt_height == -1, tree->bt_root == NULL);
	if (tree->bt_height == -1) {
	return;
	}

	VERIFY3U(zfs_btree_verify_height_helper(tree, tree->bt_root,
	tree->bt_height), ==, tree->bt_num_nodes);
	}

	/*
	* Check that the elements in this node are sorted, and that if this is a core
	* node, the separators are properly between the subtrees they separaate and
	* that the children also satisfy this requirement.
	*/
	static void
	zfs_btree_verify_order_helper(zfs_btree_t tree, zfs_btree_hdr_t hdr)
	{
	size_t size = tree->bt_elem_size;
	if (!zfs_btree_is_core(hdr)) {
	zfs_btree_leaf_t leaf = (zfs_btree_leaf_t )hdr;
	for (uint32_t i = 1; i < hdr->bth_count; i++) {
	VERIFY3S(tree->bt_compar(leaf->btl_elems +
	(hdr->bth_first + i - 1) * size,
	leaf->btl_elems +
	(hdr->bth_first + i) * size), ==, -1);
	}
	return;
	}

	zfs_btree_core_t node = (zfs_btree_core_t )hdr;
	for (uint32_t i = 1; i < hdr->bth_count; i++) {
	VERIFY3S(tree->bt_compar(node->btc_elems + (i - 1) * size,
	node->btc_elems + i * size), ==, -1);
	}
	for (uint32_t i = 0; i < hdr->bth_count; i++) {
	uint8_t *left_child_last = NULL;
	zfs_btree_hdr_t *left_child_hdr = node->btc_children[i];
	if (zfs_btree_is_core(left_child_hdr)) {
	zfs_btree_core_t *left_child =
	(zfs_btree_core_t *)left_child_hdr;
	left_child_last = left_child->btc_elems +
	(left_child_hdr->bth_count - 1) * size;
	} else {
	zfs_btree_leaf_t *left_child =
	(zfs_btree_leaf_t *)left_child_hdr;
	left_child_last = left_child->btl_elems +
	(left_child_hdr->bth_first +
	left_child_hdr->bth_count - 1) * size;
	}
	int comp = tree->bt_compar(node->btc_elems + i * size,
	left_child_last);
	if (comp <= 0) {
	panic("btree: compar returned %d (expected 1) at "
	"%px %d: compar(%px, %px)", comp, node, i,
	node->btc_elems + i * size, left_child_last);
	}

	uint8_t *right_child_first = NULL;
	zfs_btree_hdr_t *right_child_hdr = node->btc_children[i + 1];
	if (zfs_btree_is_core(right_child_hdr)) {
	zfs_btree_core_t *right_child =
	(zfs_btree_core_t *)right_child_hdr;
	right_child_first = right_child->btc_elems;
	} else {
	zfs_btree_leaf_t *right_child =
	(zfs_btree_leaf_t *)right_child_hdr;
	right_child_first = right_child->btl_elems +
	right_child_hdr->bth_first * size;
	}
	comp = tree->bt_compar(node->btc_elems + i * size,
	right_child_first);
	if (comp >= 0) {
	panic("btree: compar returned %d (expected -1) at "
	"%px %d: compar(%px, %px)", comp, node, i,
	node->btc_elems + i * size, right_child_first);
	}
	}
	for (uint32_t i = 0; i <= hdr->bth_count; i++)
	zfs_btree_verify_order_helper(tree, node->btc_children[i]);
	}

	/* Check that all elements in the tree are in sorted order. */
	static void
	zfs_btree_verify_order(zfs_btree_t *tree)
	{
	EQUIV(tree->bt_height == -1, tree->bt_root == NULL);
	if (tree->bt_height == -1) {
	return;
	}

	zfs_btree_verify_order_helper(tree, tree->bt_root);
	}

	#ifdef ZFS_DEBUG
	/* Check that all unused memory is poisoned correctly. */
	static void
	zfs_btree_verify_poison_helper(zfs_btree_t tree, zfs_btree_hdr_t hdr)
	{
	size_t size = tree->bt_elem_size;
	if (!zfs_btree_is_core(hdr)) {
	zfs_btree_leaf_t leaf = (zfs_btree_leaf_t )hdr;
	for (size_t i = 0; i < hdr->bth_first * size; i++)
	VERIFY3U(leaf->btl_elems[i], ==, 0x0f);
	size_t esize = tree->bt_leaf_size -
	offsetof(zfs_btree_leaf_t, btl_elems);
	for (size_t i = (hdr->bth_first + hdr->bth_count) * size;
	i < esize; i++)
	VERIFY3U(leaf->btl_elems[i], ==, 0x0f);
	} else {
	zfs_btree_core_t node = (zfs_btree_core_t )hdr;
	for (size_t i = hdr->bth_count * size;
	i < BTREE_CORE_ELEMS * size; i++)
	VERIFY3U(node->btc_elems[i], ==, 0x0f);

	for (uint32_t i = hdr->bth_count + 1; i <= BTREE_CORE_ELEMS;
	i++) {
	VERIFY3P(node->btc_children[i], ==,
	(zfs_btree_hdr_t *)BTREE_POISON);
	}

	for (uint32_t i = 0; i <= hdr->bth_count; i++) {
	zfs_btree_verify_poison_helper(tree,
	node->btc_children[i]);
	}
	}
	}
	#endif

	/* Check that unused memory in the tree is still poisoned. */
	static void
	zfs_btree_verify_poison(zfs_btree_t *tree)
	{
	#ifdef ZFS_DEBUG
	if (tree->bt_height == -1)
	return;
	zfs_btree_verify_poison_helper(tree, tree->bt_root);
	#endif
	}

	void
	zfs_btree_verify(zfs_btree_t *tree)
	{
	if (zfs_btree_verify_intensity == 0)
	return;
	zfs_btree_verify_height(tree);
	if (zfs_btree_verify_intensity == 1)
	return;
	zfs_btree_verify_pointers(tree);
	if (zfs_btree_verify_intensity == 2)
	return;
	zfs_btree_verify_counts(tree);
	if (zfs_btree_verify_intensity == 3)
	return;
	zfs_btree_verify_order(tree);

	if (zfs_btree_verify_intensity == 4)
	return;
	zfs_btree_verify_poison(tree);
	}

	/* BEGIN CSTYLED */
	ZFS_MODULE_PARAM(zfs, zfs_, btree_verify_intensity, UINT, ZMOD_RW,
	"Enable btree verification. Levels above 4 require ZFS be built "
	"with debugging");
	/* END CSTYLED */
	diff --git a/sys/contrib/openzfs/module/zfs/dbuf.c b/sys/contrib/openzfs/module/zfs/dbuf.c
	index bb913f556374..56fe2c4dbe30 100644
	--- a/sys/contrib/openzfs/module/zfs/dbuf.c
	+++ b/sys/contrib/openzfs/module/zfs/dbuf.c
	@@ -1,5254 +1,5265 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2012, 2020 by Delphix. All rights reserved.
	* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
	* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
	* Copyright (c) 2019, Klara Inc.
	* Copyright (c) 2019, Allan Jude
	* Copyright (c) 2021, 2022 by Pawel Jakub Dawidek
	*/

	#include <sys/zfs_context.h>
	#include <sys/arc.h>
	#include <sys/dmu.h>
	#include <sys/dmu_send.h>
	#include <sys/dmu_impl.h>
	#include <sys/dbuf.h>
	#include <sys/dmu_objset.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_dir.h>
	#include <sys/dmu_tx.h>
	#include <sys/spa.h>
	#include <sys/zio.h>
	#include <sys/dmu_zfetch.h>
	#include <sys/sa.h>
	#include <sys/sa_impl.h>
	#include <sys/zfeature.h>
	#include <sys/blkptr.h>
	#include <sys/range_tree.h>
	#include <sys/trace_zfs.h>
	#include <sys/callb.h>
	#include <sys/abd.h>
	#include <sys/brt.h>
	#include <sys/vdev.h>
	#include <cityhash.h>
	#include <sys/spa_impl.h>
	#include <sys/wmsum.h>
	#include <sys/vdev_impl.h>

	static kstat_t *dbuf_ksp;

	typedef struct dbuf_stats {
	/*
	* Various statistics about the size of the dbuf cache.
	*/
	kstat_named_t cache_count;
	kstat_named_t cache_size_bytes;
	kstat_named_t cache_size_bytes_max;
	/*
	* Statistics regarding the bounds on the dbuf cache size.
	*/
	kstat_named_t cache_target_bytes;
	kstat_named_t cache_lowater_bytes;
	kstat_named_t cache_hiwater_bytes;
	/*
	* Total number of dbuf cache evictions that have occurred.
	*/
	kstat_named_t cache_total_evicts;
	/*
	* The distribution of dbuf levels in the dbuf cache and
	* the total size of all dbufs at each level.
	*/
	kstat_named_t cache_levels[DN_MAX_LEVELS];
	kstat_named_t cache_levels_bytes[DN_MAX_LEVELS];
	/*
	* Statistics about the dbuf hash table.
	*/
	kstat_named_t hash_hits;
	kstat_named_t hash_misses;
	kstat_named_t hash_collisions;
	kstat_named_t hash_elements;
	kstat_named_t hash_elements_max;
	/*
	* Number of sublists containing more than one dbuf in the dbuf
	* hash table. Keep track of the longest hash chain.
	*/
	kstat_named_t hash_chains;
	kstat_named_t hash_chain_max;
	/*
	* Number of times a dbuf_create() discovers that a dbuf was
	* already created and in the dbuf hash table.
	*/
	kstat_named_t hash_insert_race;
	/*
	* Number of entries in the hash table dbuf and mutex arrays.
	*/
	kstat_named_t hash_table_count;
	kstat_named_t hash_mutex_count;
	/*
	* Statistics about the size of the metadata dbuf cache.
	*/
	kstat_named_t metadata_cache_count;
	kstat_named_t metadata_cache_size_bytes;
	kstat_named_t metadata_cache_size_bytes_max;
	/*
	* For diagnostic purposes, this is incremented whenever we can't add
	* something to the metadata cache because it's full, and instead put
	* the data in the regular dbuf cache.
	*/
	kstat_named_t metadata_cache_overflow;
	} dbuf_stats_t;

	dbuf_stats_t dbuf_stats = {
	{ "cache_count", KSTAT_DATA_UINT64 },
	{ "cache_size_bytes", KSTAT_DATA_UINT64 },
	{ "cache_size_bytes_max", KSTAT_DATA_UINT64 },
	{ "cache_target_bytes", KSTAT_DATA_UINT64 },
	{ "cache_lowater_bytes", KSTAT_DATA_UINT64 },
	{ "cache_hiwater_bytes", KSTAT_DATA_UINT64 },
	{ "cache_total_evicts", KSTAT_DATA_UINT64 },
	{ { "cache_levels_N", KSTAT_DATA_UINT64 } },
	{ { "cache_levels_bytes_N", KSTAT_DATA_UINT64 } },
	{ "hash_hits", KSTAT_DATA_UINT64 },
	{ "hash_misses", KSTAT_DATA_UINT64 },
	{ "hash_collisions", KSTAT_DATA_UINT64 },
	{ "hash_elements", KSTAT_DATA_UINT64 },
	{ "hash_elements_max", KSTAT_DATA_UINT64 },
	{ "hash_chains", KSTAT_DATA_UINT64 },
	{ "hash_chain_max", KSTAT_DATA_UINT64 },
	{ "hash_insert_race", KSTAT_DATA_UINT64 },
	{ "hash_table_count", KSTAT_DATA_UINT64 },
	{ "hash_mutex_count", KSTAT_DATA_UINT64 },
	{ "metadata_cache_count", KSTAT_DATA_UINT64 },
	{ "metadata_cache_size_bytes", KSTAT_DATA_UINT64 },
	{ "metadata_cache_size_bytes_max", KSTAT_DATA_UINT64 },
	{ "metadata_cache_overflow", KSTAT_DATA_UINT64 }
	};

	struct {
	wmsum_t cache_count;
	wmsum_t cache_total_evicts;
	wmsum_t cache_levels[DN_MAX_LEVELS];
	wmsum_t cache_levels_bytes[DN_MAX_LEVELS];
	wmsum_t hash_hits;
	wmsum_t hash_misses;
	wmsum_t hash_collisions;
	wmsum_t hash_chains;
	wmsum_t hash_insert_race;
	wmsum_t metadata_cache_count;
	wmsum_t metadata_cache_overflow;
	} dbuf_sums;

	#define DBUF_STAT_INCR(stat, val) \
	wmsum_add(&dbuf_sums.stat, val)
	#define DBUF_STAT_DECR(stat, val) \
	DBUF_STAT_INCR(stat, -(val))
	#define DBUF_STAT_BUMP(stat) \
	DBUF_STAT_INCR(stat, 1)
	#define DBUF_STAT_BUMPDOWN(stat) \
	DBUF_STAT_INCR(stat, -1)
	#define DBUF_STAT_MAX(stat, v) { \
	uint64_t _m; \
	while ((v) > (_m = dbuf_stats.stat.value.ui64) && \
	(_m != atomic_cas_64(&dbuf_stats.stat.value.ui64, _m, (v))))\
	continue; \
	}

	static void dbuf_write(dbuf_dirty_record_t dr, arc_buf_t data, dmu_tx_t *tx);
	static void dbuf_sync_leaf_verify_bonus_dnode(dbuf_dirty_record_t *dr);

	/*
	* Global data structures and functions for the dbuf cache.
	*/
	static kmem_cache_t *dbuf_kmem_cache;
	static taskq_t *dbu_evict_taskq;

	static kthread_t *dbuf_cache_evict_thread;
	static kmutex_t dbuf_evict_lock;
	static kcondvar_t dbuf_evict_cv;
	static boolean_t dbuf_evict_thread_exit;

	/*
	* There are two dbuf caches; each dbuf can only be in one of them at a time.
	*
	* 1. Cache of metadata dbufs, to help make read-heavy administrative commands
	* from /sbin/zfs run faster. The "metadata cache" specifically stores dbufs
	* that represent the metadata that describes filesystems/snapshots/
	* bookmarks/properties/etc. We only evict from this cache when we export a
	* pool, to short-circuit as much I/O as possible for all administrative
	* commands that need the metadata. There is no eviction policy for this
	* cache, because we try to only include types in it which would occupy a
	* very small amount of space per object but create a large impact on the
	* performance of these commands. Instead, after it reaches a maximum size
	* (which should only happen on very small memory systems with a very large
	* number of filesystem objects), we stop taking new dbufs into the
	* metadata cache, instead putting them in the normal dbuf cache.
	*
	* 2. LRU cache of dbufs. The dbuf cache maintains a list of dbufs that
	* are not currently held but have been recently released. These dbufs
	* are not eligible for arc eviction until they are aged out of the cache.
	* Dbufs that are aged out of the cache will be immediately destroyed and
	* become eligible for arc eviction.
	*
	* Dbufs are added to these caches once the last hold is released. If a dbuf is
	* later accessed and still exists in the dbuf cache, then it will be removed
	* from the cache and later re-added to the head of the cache.
	*
	* If a given dbuf meets the requirements for the metadata cache, it will go
	* there, otherwise it will be considered for the generic LRU dbuf cache. The
	* caches and the refcounts tracking their sizes are stored in an array indexed
	* by those caches' matching enum values (from dbuf_cached_state_t).
	*/
	typedef struct dbuf_cache {
	multilist_t cache;
	zfs_refcount_t size ____cacheline_aligned;
	} dbuf_cache_t;
	dbuf_cache_t dbuf_caches[DB_CACHE_MAX];

	/* Size limits for the caches */
	static uint64_t dbuf_cache_max_bytes = UINT64_MAX;
	static uint64_t dbuf_metadata_cache_max_bytes = UINT64_MAX;

	/* Set the default sizes of the caches to log2 fraction of arc size */
	static uint_t dbuf_cache_shift = 5;
	static uint_t dbuf_metadata_cache_shift = 6;

	/* Set the dbuf hash mutex count as log2 shift (dynamic by default) */
	static uint_t dbuf_mutex_cache_shift = 0;

	static unsigned long dbuf_cache_target_bytes(void);
	static unsigned long dbuf_metadata_cache_target_bytes(void);

	/*
	* The LRU dbuf cache uses a three-stage eviction policy:
	* - A low water marker designates when the dbuf eviction thread
	* should stop evicting from the dbuf cache.
	* - When we reach the maximum size (aka mid water mark), we
	* signal the eviction thread to run.
	* - The high water mark indicates when the eviction thread
	* is unable to keep up with the incoming load and eviction must
	* happen in the context of the calling thread.
	*
	* The dbuf cache:
	* (max size)
	* low water mid water hi water
	* +----------------------------------------+----------+----------+
	* \| \| \| \|
	* \| \| \| \|
	* \| \| \| \|
	* \| \| \| \|
	* +----------------------------------------+----------+----------+
	* stop signal evict
	* evicting eviction directly
	* thread
	*
	* The high and low water marks indicate the operating range for the eviction
	* thread. The low water mark is, by default, 90% of the total size of the
	* cache and the high water mark is at 110% (both of these percentages can be
	* changed by setting dbuf_cache_lowater_pct and dbuf_cache_hiwater_pct,
	* respectively). The eviction thread will try to ensure that the cache remains
	* within this range by waking up every second and checking if the cache is
	* above the low water mark. The thread can also be woken up by callers adding
	* elements into the cache if the cache is larger than the mid water (i.e max
	* cache size). Once the eviction thread is woken up and eviction is required,
	* it will continue evicting buffers until it's able to reduce the cache size
	* to the low water mark. If the cache size continues to grow and hits the high
	* water mark, then callers adding elements to the cache will begin to evict
	* directly from the cache until the cache is no longer above the high water
	* mark.
	*/

	/*
	* The percentage above and below the maximum cache size.
	*/
	static uint_t dbuf_cache_hiwater_pct = 10;
	static uint_t dbuf_cache_lowater_pct = 10;

	static int
	dbuf_cons(void vdb, void unused, int kmflag)
	{
	(void) unused, (void) kmflag;
	dmu_buf_impl_t *db = vdb;
	memset(db, 0, sizeof (dmu_buf_impl_t));

	- mutex_init(&db->db_mtx, NULL, MUTEX_DEFAULT, NULL);
	- rw_init(&db->db_rwlock, NULL, RW_DEFAULT, NULL);
	+ mutex_init(&db->db_mtx, NULL, MUTEX_NOLOCKDEP, NULL);
	+ rw_init(&db->db_rwlock, NULL, RW_NOLOCKDEP, NULL);
	cv_init(&db->db_changed, NULL, CV_DEFAULT, NULL);
	multilist_link_init(&db->db_cache_link);
	zfs_refcount_create(&db->db_holds);

	return (0);
	}

	static void
	dbuf_dest(void vdb, void unused)
	{
	(void) unused;
	dmu_buf_impl_t *db = vdb;
	mutex_destroy(&db->db_mtx);
	rw_destroy(&db->db_rwlock);
	cv_destroy(&db->db_changed);
	ASSERT(!multilist_link_active(&db->db_cache_link));
	zfs_refcount_destroy(&db->db_holds);
	}

	/*
	* dbuf hash table routines
	*/
	static dbuf_hash_table_t dbuf_hash_table;

	/*
	* We use Cityhash for this. It's fast, and has good hash properties without
	* requiring any large static buffers.
	*/
	static uint64_t
	dbuf_hash(void *os, uint64_t obj, uint8_t lvl, uint64_t blkid)
	{
	return (cityhash4((uintptr_t)os, obj, (uint64_t)lvl, blkid));
	}

	#define DTRACE_SET_STATE(db, why) \
	DTRACE_PROBE2(dbuf__state_change, dmu_buf_impl_t *, db, \
	const char *, why)

	#define DBUF_EQUAL(dbuf, os, obj, level, blkid) \
	((dbuf)->db.db_object == (obj) && \
	(dbuf)->db_objset == (os) && \
	(dbuf)->db_level == (level) && \
	(dbuf)->db_blkid == (blkid))

	dmu_buf_impl_t *
	dbuf_find(objset_t *os, uint64_t obj, uint8_t level, uint64_t blkid,
	uint64_t *hash_out)
	{
	dbuf_hash_table_t *h = &dbuf_hash_table;
	uint64_t hv;
	uint64_t idx;
	dmu_buf_impl_t *db;

	hv = dbuf_hash(os, obj, level, blkid);
	idx = hv & h->hash_table_mask;

	mutex_enter(DBUF_HASH_MUTEX(h, idx));
	for (db = h->hash_table[idx]; db != NULL; db = db->db_hash_next) {
	if (DBUF_EQUAL(db, os, obj, level, blkid)) {
	mutex_enter(&db->db_mtx);
	if (db->db_state != DB_EVICTING) {
	mutex_exit(DBUF_HASH_MUTEX(h, idx));
	return (db);
	}
	mutex_exit(&db->db_mtx);
	}
	}
	mutex_exit(DBUF_HASH_MUTEX(h, idx));
	if (hash_out != NULL)
	*hash_out = hv;
	return (NULL);
	}

	static dmu_buf_impl_t *
	dbuf_find_bonus(objset_t *os, uint64_t object)
	{
	dnode_t *dn;
	dmu_buf_impl_t *db = NULL;

	if (dnode_hold(os, object, FTAG, &dn) == 0) {
	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	if (dn->dn_bonus != NULL) {
	db = dn->dn_bonus;
	mutex_enter(&db->db_mtx);
	}
	rw_exit(&dn->dn_struct_rwlock);
	dnode_rele(dn, FTAG);
	}
	return (db);
	}

	/*
	* Insert an entry into the hash table. If there is already an element
	* equal to elem in the hash table, then the already existing element
	* will be returned and the new element will not be inserted.
	* Otherwise returns NULL.
	*/
	static dmu_buf_impl_t *
	dbuf_hash_insert(dmu_buf_impl_t *db)
	{
	dbuf_hash_table_t *h = &dbuf_hash_table;
	objset_t *os = db->db_objset;
	uint64_t obj = db->db.db_object;
	int level = db->db_level;
	uint64_t blkid, idx;
	dmu_buf_impl_t *dbf;
	uint32_t i;

	blkid = db->db_blkid;
	ASSERT3U(dbuf_hash(os, obj, level, blkid), ==, db->db_hash);
	idx = db->db_hash & h->hash_table_mask;

	mutex_enter(DBUF_HASH_MUTEX(h, idx));
	for (dbf = h->hash_table[idx], i = 0; dbf != NULL;
	dbf = dbf->db_hash_next, i++) {
	if (DBUF_EQUAL(dbf, os, obj, level, blkid)) {
	mutex_enter(&dbf->db_mtx);
	if (dbf->db_state != DB_EVICTING) {
	mutex_exit(DBUF_HASH_MUTEX(h, idx));
	return (dbf);
	}
	mutex_exit(&dbf->db_mtx);
	}
	}

	if (i > 0) {
	DBUF_STAT_BUMP(hash_collisions);
	if (i == 1)
	DBUF_STAT_BUMP(hash_chains);

	DBUF_STAT_MAX(hash_chain_max, i);
	}

	mutex_enter(&db->db_mtx);
	db->db_hash_next = h->hash_table[idx];
	h->hash_table[idx] = db;
	mutex_exit(DBUF_HASH_MUTEX(h, idx));
	uint64_t he = atomic_inc_64_nv(&dbuf_stats.hash_elements.value.ui64);
	DBUF_STAT_MAX(hash_elements_max, he);

	return (NULL);
	}

	/*
	* This returns whether this dbuf should be stored in the metadata cache, which
	* is based on whether it's from one of the dnode types that store data related
	* to traversing dataset hierarchies.
	*/
	static boolean_t
	dbuf_include_in_metadata_cache(dmu_buf_impl_t *db)
	{
	DB_DNODE_ENTER(db);
	dmu_object_type_t type = DB_DNODE(db)->dn_type;
	DB_DNODE_EXIT(db);

	/* Check if this dbuf is one of the types we care about */
	if (DMU_OT_IS_METADATA_CACHED(type)) {
	/* If we hit this, then we set something up wrong in dmu_ot */
	ASSERT(DMU_OT_IS_METADATA(type));

	/*
	* Sanity check for small-memory systems: don't allocate too
	* much memory for this purpose.
	*/
	if (zfs_refcount_count(
	&dbuf_caches[DB_DBUF_METADATA_CACHE].size) >
	dbuf_metadata_cache_target_bytes()) {
	DBUF_STAT_BUMP(metadata_cache_overflow);
	return (B_FALSE);
	}

	return (B_TRUE);
	}

	return (B_FALSE);
	}

	/*
	* Remove an entry from the hash table. It must be in the EVICTING state.
	*/
	static void
	dbuf_hash_remove(dmu_buf_impl_t *db)
	{
	dbuf_hash_table_t *h = &dbuf_hash_table;
	uint64_t idx;
	dmu_buf_impl_t dbf, *dbp;

	ASSERT3U(dbuf_hash(db->db_objset, db->db.db_object, db->db_level,
	db->db_blkid), ==, db->db_hash);
	idx = db->db_hash & h->hash_table_mask;

	/*
	* We mustn't hold db_mtx to maintain lock ordering:
	* DBUF_HASH_MUTEX > db_mtx.
	*/
	ASSERT(zfs_refcount_is_zero(&db->db_holds));
	ASSERT(db->db_state == DB_EVICTING);
	ASSERT(!MUTEX_HELD(&db->db_mtx));

	mutex_enter(DBUF_HASH_MUTEX(h, idx));
	dbp = &h->hash_table[idx];
	while ((dbf = *dbp) != db) {
	dbp = &dbf->db_hash_next;
	ASSERT(dbf != NULL);
	}
	*dbp = db->db_hash_next;
	db->db_hash_next = NULL;
	if (h->hash_table[idx] &&
	h->hash_table[idx]->db_hash_next == NULL)
	DBUF_STAT_BUMPDOWN(hash_chains);
	mutex_exit(DBUF_HASH_MUTEX(h, idx));
	atomic_dec_64(&dbuf_stats.hash_elements.value.ui64);
	}

	typedef enum {
	DBVU_EVICTING,
	DBVU_NOT_EVICTING
	} dbvu_verify_type_t;

	static void
	dbuf_verify_user(dmu_buf_impl_t *db, dbvu_verify_type_t verify_type)
	{
	#ifdef ZFS_DEBUG
	int64_t holds;

	if (db->db_user == NULL)
	return;

	/* Only data blocks support the attachment of user data. */
	ASSERT(db->db_level == 0);

	/* Clients must resolve a dbuf before attaching user data. */
	ASSERT(db->db.db_data != NULL);
	ASSERT3U(db->db_state, ==, DB_CACHED);

	holds = zfs_refcount_count(&db->db_holds);
	if (verify_type == DBVU_EVICTING) {
	/*
	* Immediate eviction occurs when holds == dirtycnt.
	* For normal eviction buffers, holds is zero on
	* eviction, except when dbuf_fix_old_data() calls
	* dbuf_clear_data(). However, the hold count can grow
	* during eviction even though db_mtx is held (see
	* dmu_bonus_hold() for an example), so we can only
	* test the generic invariant that holds >= dirtycnt.
	*/
	ASSERT3U(holds, >=, db->db_dirtycnt);
	} else {
	if (db->db_user_immediate_evict == TRUE)
	ASSERT3U(holds, >=, db->db_dirtycnt);
	else
	ASSERT3U(holds, >, 0);
	}
	#endif
	}

	static void
	dbuf_evict_user(dmu_buf_impl_t *db)
	{
	dmu_buf_user_t *dbu = db->db_user;

	ASSERT(MUTEX_HELD(&db->db_mtx));

	if (dbu == NULL)
	return;

	dbuf_verify_user(db, DBVU_EVICTING);
	db->db_user = NULL;

	#ifdef ZFS_DEBUG
	if (dbu->dbu_clear_on_evict_dbufp != NULL)
	*dbu->dbu_clear_on_evict_dbufp = NULL;
	#endif

	if (db->db_caching_status != DB_NO_CACHE) {
	/*
	* This is a cached dbuf, so the size of the user data is
	* included in its cached amount. We adjust it here because the
	* user data has already been detached from the dbuf, and the
	* sync functions are not supposed to touch it (the dbuf might
	* not exist anymore by the time the sync functions run.
	*/
	uint64_t size = dbu->dbu_size;
	(void) zfs_refcount_remove_many(
	- &dbuf_caches[db->db_caching_status].size, size, db);
	+ &dbuf_caches[db->db_caching_status].size, size, dbu);
	if (db->db_caching_status == DB_DBUF_CACHE)
	DBUF_STAT_DECR(cache_levels_bytes[db->db_level], size);
	}

	/*
	* There are two eviction callbacks - one that we call synchronously
	* and one that we invoke via a taskq. The async one is useful for
	* avoiding lock order reversals and limiting stack depth.
	*
	* Note that if we have a sync callback but no async callback,
	* it's likely that the sync callback will free the structure
	* containing the dbu. In that case we need to take care to not
	* dereference dbu after calling the sync evict func.
	*/
	boolean_t has_async = (dbu->dbu_evict_func_async != NULL);

	if (dbu->dbu_evict_func_sync != NULL)
	dbu->dbu_evict_func_sync(dbu);

	if (has_async) {
	taskq_dispatch_ent(dbu_evict_taskq, dbu->dbu_evict_func_async,
	dbu, 0, &dbu->dbu_tqent);
	}
	}

	boolean_t
	dbuf_is_metadata(dmu_buf_impl_t *db)
	{
	/*
	* Consider indirect blocks and spill blocks to be meta data.
	*/
	if (db->db_level > 0 \|\| db->db_blkid == DMU_SPILL_BLKID) {
	return (B_TRUE);
	} else {
	boolean_t is_metadata;

	DB_DNODE_ENTER(db);
	is_metadata = DMU_OT_IS_METADATA(DB_DNODE(db)->dn_type);
	DB_DNODE_EXIT(db);

	return (is_metadata);
	}
	}

	/*
	* We want to exclude buffers that are on a special allocation class from
	* L2ARC.
	*/
	boolean_t
	dbuf_is_l2cacheable(dmu_buf_impl_t *db)
	{
	if (db->db_objset->os_secondary_cache == ZFS_CACHE_ALL \|\|
	(db->db_objset->os_secondary_cache ==
	ZFS_CACHE_METADATA && dbuf_is_metadata(db))) {
	if (l2arc_exclude_special == 0)
	return (B_TRUE);

	blkptr_t *bp = db->db_blkptr;
	if (bp == NULL \|\| BP_IS_HOLE(bp))
	return (B_FALSE);
	uint64_t vdev = DVA_GET_VDEV(bp->blk_dva);
	vdev_t *rvd = db->db_objset->os_spa->spa_root_vdev;
	vdev_t *vd = NULL;

	if (vdev < rvd->vdev_children)
	vd = rvd->vdev_child[vdev];

	if (vd == NULL)
	return (B_TRUE);

	if (vd->vdev_alloc_bias != VDEV_BIAS_SPECIAL &&
	vd->vdev_alloc_bias != VDEV_BIAS_DEDUP)
	return (B_TRUE);
	}
	return (B_FALSE);
	}

	static inline boolean_t
	dnode_level_is_l2cacheable(blkptr_t bp, dnode_t dn, int64_t level)
	{
	if (dn->dn_objset->os_secondary_cache == ZFS_CACHE_ALL \|\|
	(dn->dn_objset->os_secondary_cache == ZFS_CACHE_METADATA &&
	(level > 0 \|\|
	DMU_OT_IS_METADATA(dn->dn_handle->dnh_dnode->dn_type)))) {
	if (l2arc_exclude_special == 0)
	return (B_TRUE);

	if (bp == NULL \|\| BP_IS_HOLE(bp))
	return (B_FALSE);
	uint64_t vdev = DVA_GET_VDEV(bp->blk_dva);
	vdev_t *rvd = dn->dn_objset->os_spa->spa_root_vdev;
	vdev_t *vd = NULL;

	if (vdev < rvd->vdev_children)
	vd = rvd->vdev_child[vdev];

	if (vd == NULL)
	return (B_TRUE);

	if (vd->vdev_alloc_bias != VDEV_BIAS_SPECIAL &&
	vd->vdev_alloc_bias != VDEV_BIAS_DEDUP)
	return (B_TRUE);
	}
	return (B_FALSE);
	}


	/*
	* This function must return indices evenly distributed between all
	* sublists of the multilist. This is needed due to how the dbuf eviction
	* code is laid out; dbuf_evict_thread() assumes dbufs are evenly
	* distributed between all sublists and uses this assumption when
	* deciding which sublist to evict from and how much to evict from it.
	*/
	static unsigned int
	dbuf_cache_multilist_index_func(multilist_t ml, void obj)
	{
	dmu_buf_impl_t *db = obj;

	/*
	* The assumption here, is the hash value for a given
	* dmu_buf_impl_t will remain constant throughout it's lifetime
	* (i.e. it's objset, object, level and blkid fields don't change).
	* Thus, we don't need to store the dbuf's sublist index
	* on insertion, as this index can be recalculated on removal.
	*
	* Also, the low order bits of the hash value are thought to be
	* distributed evenly. Otherwise, in the case that the multilist
	* has a power of two number of sublists, each sublists' usage
	* would not be evenly distributed. In this context full 64bit
	* division would be a waste of time, so limit it to 32 bits.
	*/
	return ((unsigned int)dbuf_hash(db->db_objset, db->db.db_object,
	db->db_level, db->db_blkid) %
	multilist_get_num_sublists(ml));
	}

	/*
	* The target size of the dbuf cache can grow with the ARC target,
	* unless limited by the tunable dbuf_cache_max_bytes.
	*/
	static inline unsigned long
	dbuf_cache_target_bytes(void)
	{
	return (MIN(dbuf_cache_max_bytes,
	arc_target_bytes() >> dbuf_cache_shift));
	}

	/*
	* The target size of the dbuf metadata cache can grow with the ARC target,
	* unless limited by the tunable dbuf_metadata_cache_max_bytes.
	*/
	static inline unsigned long
	dbuf_metadata_cache_target_bytes(void)
	{
	return (MIN(dbuf_metadata_cache_max_bytes,
	arc_target_bytes() >> dbuf_metadata_cache_shift));
	}

	static inline uint64_t
	dbuf_cache_hiwater_bytes(void)
	{
	uint64_t dbuf_cache_target = dbuf_cache_target_bytes();
	return (dbuf_cache_target +
	(dbuf_cache_target * dbuf_cache_hiwater_pct) / 100);
	}

	static inline uint64_t
	dbuf_cache_lowater_bytes(void)
	{
	uint64_t dbuf_cache_target = dbuf_cache_target_bytes();
	return (dbuf_cache_target -
	(dbuf_cache_target * dbuf_cache_lowater_pct) / 100);
	}

	static inline boolean_t
	dbuf_cache_above_lowater(void)
	{
	return (zfs_refcount_count(&dbuf_caches[DB_DBUF_CACHE].size) >
	dbuf_cache_lowater_bytes());
	}

	/*
	* Evict the oldest eligible dbuf from the dbuf cache.
	*/
	static void
	dbuf_evict_one(void)
	{
	int idx = multilist_get_random_index(&dbuf_caches[DB_DBUF_CACHE].cache);
	multilist_sublist_t *mls = multilist_sublist_lock_idx(
	&dbuf_caches[DB_DBUF_CACHE].cache, idx);

	ASSERT(!MUTEX_HELD(&dbuf_evict_lock));

	dmu_buf_impl_t *db = multilist_sublist_tail(mls);
	while (db != NULL && mutex_tryenter(&db->db_mtx) == 0) {
	db = multilist_sublist_prev(mls, db);
	}

	DTRACE_PROBE2(dbuf__evict__one, dmu_buf_impl_t *, db,
	multilist_sublist_t *, mls);

	if (db != NULL) {
	multilist_sublist_remove(mls, db);
	multilist_sublist_unlock(mls);
	- uint64_t size = db->db.db_size + dmu_buf_user_size(&db->db);
	+ uint64_t size = db->db.db_size;
	+ uint64_t usize = dmu_buf_user_size(&db->db);
	(void) zfs_refcount_remove_many(
	&dbuf_caches[DB_DBUF_CACHE].size, size, db);
	+ (void) zfs_refcount_remove_many(
	+ &dbuf_caches[DB_DBUF_CACHE].size, usize, db->db_user);
	DBUF_STAT_BUMPDOWN(cache_levels[db->db_level]);
	DBUF_STAT_BUMPDOWN(cache_count);
	- DBUF_STAT_DECR(cache_levels_bytes[db->db_level], size);
	+ DBUF_STAT_DECR(cache_levels_bytes[db->db_level], size + usize);
	ASSERT3U(db->db_caching_status, ==, DB_DBUF_CACHE);
	db->db_caching_status = DB_NO_CACHE;
	dbuf_destroy(db);
	DBUF_STAT_BUMP(cache_total_evicts);
	} else {
	multilist_sublist_unlock(mls);
	}
	}

	/*
	* The dbuf evict thread is responsible for aging out dbufs from the
	* cache. Once the cache has reached it's maximum size, dbufs are removed
	* and destroyed. The eviction thread will continue running until the size
	* of the dbuf cache is at or below the maximum size. Once the dbuf is aged
	* out of the cache it is destroyed and becomes eligible for arc eviction.
	*/
	static __attribute__((noreturn)) void
	dbuf_evict_thread(void *unused)
	{
	(void) unused;
	callb_cpr_t cpr;

	CALLB_CPR_INIT(&cpr, &dbuf_evict_lock, callb_generic_cpr, FTAG);

	mutex_enter(&dbuf_evict_lock);
	while (!dbuf_evict_thread_exit) {
	while (!dbuf_cache_above_lowater() && !dbuf_evict_thread_exit) {
	CALLB_CPR_SAFE_BEGIN(&cpr);
	(void) cv_timedwait_idle_hires(&dbuf_evict_cv,
	&dbuf_evict_lock, SEC2NSEC(1), MSEC2NSEC(1), 0);
	CALLB_CPR_SAFE_END(&cpr, &dbuf_evict_lock);
	}
	mutex_exit(&dbuf_evict_lock);

	/*
	* Keep evicting as long as we're above the low water mark
	* for the cache. We do this without holding the locks to
	* minimize lock contention.
	*/
	while (dbuf_cache_above_lowater() && !dbuf_evict_thread_exit) {
	dbuf_evict_one();
	}

	mutex_enter(&dbuf_evict_lock);
	}

	dbuf_evict_thread_exit = B_FALSE;
	cv_broadcast(&dbuf_evict_cv);
	CALLB_CPR_EXIT(&cpr); /* drops dbuf_evict_lock */
	thread_exit();
	}

	/*
	* Wake up the dbuf eviction thread if the dbuf cache is at its max size.
	* If the dbuf cache is at its high water mark, then evict a dbuf from the
	* dbuf cache using the caller's context.
	*/
	static void
	dbuf_evict_notify(uint64_t size)
	{
	/*
	* We check if we should evict without holding the dbuf_evict_lock,
	* because it's OK to occasionally make the wrong decision here,
	* and grabbing the lock results in massive lock contention.
	*/
	if (size > dbuf_cache_target_bytes()) {
	if (size > dbuf_cache_hiwater_bytes())
	dbuf_evict_one();
	cv_signal(&dbuf_evict_cv);
	}
	}

	static int
	dbuf_kstat_update(kstat_t *ksp, int rw)
	{
	dbuf_stats_t *ds = ksp->ks_data;
	dbuf_hash_table_t *h = &dbuf_hash_table;

	if (rw == KSTAT_WRITE)
	return (SET_ERROR(EACCES));

	ds->cache_count.value.ui64 =
	wmsum_value(&dbuf_sums.cache_count);
	ds->cache_size_bytes.value.ui64 =
	zfs_refcount_count(&dbuf_caches[DB_DBUF_CACHE].size);
	ds->cache_target_bytes.value.ui64 = dbuf_cache_target_bytes();
	ds->cache_hiwater_bytes.value.ui64 = dbuf_cache_hiwater_bytes();
	ds->cache_lowater_bytes.value.ui64 = dbuf_cache_lowater_bytes();
	ds->cache_total_evicts.value.ui64 =
	wmsum_value(&dbuf_sums.cache_total_evicts);
	for (int i = 0; i < DN_MAX_LEVELS; i++) {
	ds->cache_levels[i].value.ui64 =
	wmsum_value(&dbuf_sums.cache_levels[i]);
	ds->cache_levels_bytes[i].value.ui64 =
	wmsum_value(&dbuf_sums.cache_levels_bytes[i]);
	}
	ds->hash_hits.value.ui64 =
	wmsum_value(&dbuf_sums.hash_hits);
	ds->hash_misses.value.ui64 =
	wmsum_value(&dbuf_sums.hash_misses);
	ds->hash_collisions.value.ui64 =
	wmsum_value(&dbuf_sums.hash_collisions);
	ds->hash_chains.value.ui64 =
	wmsum_value(&dbuf_sums.hash_chains);
	ds->hash_insert_race.value.ui64 =
	wmsum_value(&dbuf_sums.hash_insert_race);
	ds->hash_table_count.value.ui64 = h->hash_table_mask + 1;
	ds->hash_mutex_count.value.ui64 = h->hash_mutex_mask + 1;
	ds->metadata_cache_count.value.ui64 =
	wmsum_value(&dbuf_sums.metadata_cache_count);
	ds->metadata_cache_size_bytes.value.ui64 = zfs_refcount_count(
	&dbuf_caches[DB_DBUF_METADATA_CACHE].size);
	ds->metadata_cache_overflow.value.ui64 =
	wmsum_value(&dbuf_sums.metadata_cache_overflow);
	return (0);
	}

	void
	dbuf_init(void)
	{
	uint64_t hmsize, hsize = 1ULL << 16;
	dbuf_hash_table_t *h = &dbuf_hash_table;

	/*
	* The hash table is big enough to fill one eighth of physical memory
	* with an average block size of zfs_arc_average_blocksize (default 8K).
	* By default, the table will take up
	* totalmem * sizeof(void*) / 8K (1MB per GB with 8-byte pointers).
	*/
	while (hsize * zfs_arc_average_blocksize < arc_all_memory() / 8)
	hsize <<= 1;

	h->hash_table = NULL;
	while (h->hash_table == NULL) {
	h->hash_table_mask = hsize - 1;

	h->hash_table = vmem_zalloc(hsize * sizeof (void *), KM_SLEEP);
	if (h->hash_table == NULL)
	hsize >>= 1;

	ASSERT3U(hsize, >=, 1ULL << 10);
	}

	/*
	* The hash table buckets are protected by an array of mutexes where
	* each mutex is reponsible for protecting 128 buckets. A minimum
	* array size of 8192 is targeted to avoid contention.
	*/
	if (dbuf_mutex_cache_shift == 0)
	hmsize = MAX(hsize >> 7, 1ULL << 13);
	else
	hmsize = 1ULL << MIN(dbuf_mutex_cache_shift, 24);

	h->hash_mutexes = NULL;
	while (h->hash_mutexes == NULL) {
	h->hash_mutex_mask = hmsize - 1;

	h->hash_mutexes = vmem_zalloc(hmsize * sizeof (kmutex_t),
	KM_SLEEP);
	if (h->hash_mutexes == NULL)
	hmsize >>= 1;
	}

	dbuf_kmem_cache = kmem_cache_create("dmu_buf_impl_t",
	sizeof (dmu_buf_impl_t),
	0, dbuf_cons, dbuf_dest, NULL, NULL, NULL, 0);

	for (int i = 0; i < hmsize; i++)
	- mutex_init(&h->hash_mutexes[i], NULL, MUTEX_DEFAULT, NULL);
	+ mutex_init(&h->hash_mutexes[i], NULL, MUTEX_NOLOCKDEP, NULL);

	dbuf_stats_init(h);

	/*
	* All entries are queued via taskq_dispatch_ent(), so min/maxalloc
	* configuration is not required.
	*/
	dbu_evict_taskq = taskq_create("dbu_evict", 1, defclsyspri, 0, 0, 0);

	for (dbuf_cached_state_t dcs = 0; dcs < DB_CACHE_MAX; dcs++) {
	multilist_create(&dbuf_caches[dcs].cache,
	sizeof (dmu_buf_impl_t),
	offsetof(dmu_buf_impl_t, db_cache_link),
	dbuf_cache_multilist_index_func);
	zfs_refcount_create(&dbuf_caches[dcs].size);
	}

	dbuf_evict_thread_exit = B_FALSE;
	mutex_init(&dbuf_evict_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&dbuf_evict_cv, NULL, CV_DEFAULT, NULL);
	dbuf_cache_evict_thread = thread_create(NULL, 0, dbuf_evict_thread,
	NULL, 0, &p0, TS_RUN, minclsyspri);

	wmsum_init(&dbuf_sums.cache_count, 0);
	wmsum_init(&dbuf_sums.cache_total_evicts, 0);
	for (int i = 0; i < DN_MAX_LEVELS; i++) {
	wmsum_init(&dbuf_sums.cache_levels[i], 0);
	wmsum_init(&dbuf_sums.cache_levels_bytes[i], 0);
	}
	wmsum_init(&dbuf_sums.hash_hits, 0);
	wmsum_init(&dbuf_sums.hash_misses, 0);
	wmsum_init(&dbuf_sums.hash_collisions, 0);
	wmsum_init(&dbuf_sums.hash_chains, 0);
	wmsum_init(&dbuf_sums.hash_insert_race, 0);
	wmsum_init(&dbuf_sums.metadata_cache_count, 0);
	wmsum_init(&dbuf_sums.metadata_cache_overflow, 0);

	dbuf_ksp = kstat_create("zfs", 0, "dbufstats", "misc",
	KSTAT_TYPE_NAMED, sizeof (dbuf_stats) / sizeof (kstat_named_t),
	KSTAT_FLAG_VIRTUAL);
	if (dbuf_ksp != NULL) {
	for (int i = 0; i < DN_MAX_LEVELS; i++) {
	snprintf(dbuf_stats.cache_levels[i].name,
	KSTAT_STRLEN, "cache_level_%d", i);
	dbuf_stats.cache_levels[i].data_type =
	KSTAT_DATA_UINT64;
	snprintf(dbuf_stats.cache_levels_bytes[i].name,
	KSTAT_STRLEN, "cache_level_%d_bytes", i);
	dbuf_stats.cache_levels_bytes[i].data_type =
	KSTAT_DATA_UINT64;
	}
	dbuf_ksp->ks_data = &dbuf_stats;
	dbuf_ksp->ks_update = dbuf_kstat_update;
	kstat_install(dbuf_ksp);
	}
	}

	void
	dbuf_fini(void)
	{
	dbuf_hash_table_t *h = &dbuf_hash_table;

	dbuf_stats_destroy();

	for (int i = 0; i < (h->hash_mutex_mask + 1); i++)
	mutex_destroy(&h->hash_mutexes[i]);

	vmem_free(h->hash_table, (h->hash_table_mask + 1) * sizeof (void *));
	vmem_free(h->hash_mutexes, (h->hash_mutex_mask + 1) *
	sizeof (kmutex_t));

	kmem_cache_destroy(dbuf_kmem_cache);
	taskq_destroy(dbu_evict_taskq);

	mutex_enter(&dbuf_evict_lock);
	dbuf_evict_thread_exit = B_TRUE;
	while (dbuf_evict_thread_exit) {
	cv_signal(&dbuf_evict_cv);
	cv_wait(&dbuf_evict_cv, &dbuf_evict_lock);
	}
	mutex_exit(&dbuf_evict_lock);

	mutex_destroy(&dbuf_evict_lock);
	cv_destroy(&dbuf_evict_cv);

	for (dbuf_cached_state_t dcs = 0; dcs < DB_CACHE_MAX; dcs++) {
	zfs_refcount_destroy(&dbuf_caches[dcs].size);
	multilist_destroy(&dbuf_caches[dcs].cache);
	}

	if (dbuf_ksp != NULL) {
	kstat_delete(dbuf_ksp);
	dbuf_ksp = NULL;
	}

	wmsum_fini(&dbuf_sums.cache_count);
	wmsum_fini(&dbuf_sums.cache_total_evicts);
	for (int i = 0; i < DN_MAX_LEVELS; i++) {
	wmsum_fini(&dbuf_sums.cache_levels[i]);
	wmsum_fini(&dbuf_sums.cache_levels_bytes[i]);
	}
	wmsum_fini(&dbuf_sums.hash_hits);
	wmsum_fini(&dbuf_sums.hash_misses);
	wmsum_fini(&dbuf_sums.hash_collisions);
	wmsum_fini(&dbuf_sums.hash_chains);
	wmsum_fini(&dbuf_sums.hash_insert_race);
	wmsum_fini(&dbuf_sums.metadata_cache_count);
	wmsum_fini(&dbuf_sums.metadata_cache_overflow);
	}

	/*
	* Other stuff.
	*/

	#ifdef ZFS_DEBUG
	static void
	dbuf_verify(dmu_buf_impl_t *db)
	{
	dnode_t *dn;
	dbuf_dirty_record_t *dr;
	uint32_t txg_prev;

	ASSERT(MUTEX_HELD(&db->db_mtx));

	if (!(zfs_flags & ZFS_DEBUG_DBUF_VERIFY))
	return;

	ASSERT(db->db_objset != NULL);
	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	if (dn == NULL) {
	ASSERT(db->db_parent == NULL);
	ASSERT(db->db_blkptr == NULL);
	} else {
	ASSERT3U(db->db.db_object, ==, dn->dn_object);
	ASSERT3P(db->db_objset, ==, dn->dn_objset);
	ASSERT3U(db->db_level, <, dn->dn_nlevels);
	ASSERT(db->db_blkid == DMU_BONUS_BLKID \|\|
	db->db_blkid == DMU_SPILL_BLKID \|\|
	!avl_is_empty(&dn->dn_dbufs));
	}
	if (db->db_blkid == DMU_BONUS_BLKID) {
	ASSERT(dn != NULL);
	ASSERT3U(db->db.db_size, >=, dn->dn_bonuslen);
	ASSERT3U(db->db.db_offset, ==, DMU_BONUS_BLKID);
	} else if (db->db_blkid == DMU_SPILL_BLKID) {
	ASSERT(dn != NULL);
	ASSERT0(db->db.db_offset);
	} else {
	ASSERT3U(db->db.db_offset, ==, db->db_blkid * db->db.db_size);
	}

	if ((dr = list_head(&db->db_dirty_records)) != NULL) {
	ASSERT(dr->dr_dbuf == db);
	txg_prev = dr->dr_txg;
	for (dr = list_next(&db->db_dirty_records, dr); dr != NULL;
	dr = list_next(&db->db_dirty_records, dr)) {
	ASSERT(dr->dr_dbuf == db);
	ASSERT(txg_prev > dr->dr_txg);
	txg_prev = dr->dr_txg;
	}
	}

	/*
	* We can't assert that db_size matches dn_datablksz because it
	* can be momentarily different when another thread is doing
	* dnode_set_blksz().
	*/
	if (db->db_level == 0 && db->db.db_object == DMU_META_DNODE_OBJECT) {
	dr = db->db_data_pending;
	/*
	* It should only be modified in syncing context, so
	* make sure we only have one copy of the data.
	*/
	ASSERT(dr == NULL \|\| dr->dt.dl.dr_data == db->db_buf);
	}

	/* verify db->db_blkptr */
	if (db->db_blkptr) {
	if (db->db_parent == dn->dn_dbuf) {
	/* db is pointed to by the dnode */
	/* ASSERT3U(db->db_blkid, <, dn->dn_nblkptr); */
	if (DMU_OBJECT_IS_SPECIAL(db->db.db_object))
	ASSERT(db->db_parent == NULL);
	else
	ASSERT(db->db_parent != NULL);
	if (db->db_blkid != DMU_SPILL_BLKID)
	ASSERT3P(db->db_blkptr, ==,
	&dn->dn_phys->dn_blkptr[db->db_blkid]);
	} else {
	/* db is pointed to by an indirect block */
	int epb __maybe_unused = db->db_parent->db.db_size >>
	SPA_BLKPTRSHIFT;
	ASSERT3U(db->db_parent->db_level, ==, db->db_level+1);
	ASSERT3U(db->db_parent->db.db_object, ==,
	db->db.db_object);
	/*
	* dnode_grow_indblksz() can make this fail if we don't
	* have the parent's rwlock. XXX indblksz no longer
	* grows. safe to do this now?
	*/
	if (RW_LOCK_HELD(&db->db_parent->db_rwlock)) {
	ASSERT3P(db->db_blkptr, ==,
	((blkptr_t *)db->db_parent->db.db_data +
	db->db_blkid % epb));
	}
	}
	}
	if ((db->db_blkptr == NULL \|\| BP_IS_HOLE(db->db_blkptr)) &&
	(db->db_buf == NULL \|\| db->db_buf->b_data) &&
	db->db.db_data && db->db_blkid != DMU_BONUS_BLKID &&
	db->db_state != DB_FILL && (dn == NULL \|\| !dn->dn_free_txg)) {
	/*
	* If the blkptr isn't set but they have nonzero data,
	* it had better be dirty, otherwise we'll lose that
	* data when we evict this buffer.
	*
	* There is an exception to this rule for indirect blocks; in
	* this case, if the indirect block is a hole, we fill in a few
	* fields on each of the child blocks (importantly, birth time)
	* to prevent hole birth times from being lost when you
	* partially fill in a hole.
	*/
	if (db->db_dirtycnt == 0) {
	if (db->db_level == 0) {
	uint64_t *buf = db->db.db_data;
	int i;

	for (i = 0; i < db->db.db_size >> 3; i++) {
	ASSERT(buf[i] == 0);
	}
	} else {
	blkptr_t *bps = db->db.db_data;
	ASSERT3U(1 << DB_DNODE(db)->dn_indblkshift, ==,
	db->db.db_size);
	/*
	* We want to verify that all the blkptrs in the
	* indirect block are holes, but we may have
	* automatically set up a few fields for them.
	* We iterate through each blkptr and verify
	* they only have those fields set.
	*/
	for (int i = 0;
	i < db->db.db_size / sizeof (blkptr_t);
	i++) {
	blkptr_t *bp = &bps[i];
	ASSERT(ZIO_CHECKSUM_IS_ZERO(
	&bp->blk_cksum));
	ASSERT(
	DVA_IS_EMPTY(&bp->blk_dva[0]) &&
	DVA_IS_EMPTY(&bp->blk_dva[1]) &&
	DVA_IS_EMPTY(&bp->blk_dva[2]));
	ASSERT0(bp->blk_fill);
	ASSERT0(bp->blk_pad[0]);
	ASSERT0(bp->blk_pad[1]);
	ASSERT(!BP_IS_EMBEDDED(bp));
	ASSERT(BP_IS_HOLE(bp));
	ASSERT0(BP_GET_PHYSICAL_BIRTH(bp));
	}
	}
	}
	}
	DB_DNODE_EXIT(db);
	}
	#endif

	static void
	dbuf_clear_data(dmu_buf_impl_t *db)
	{
	ASSERT(MUTEX_HELD(&db->db_mtx));
	dbuf_evict_user(db);
	ASSERT3P(db->db_buf, ==, NULL);
	db->db.db_data = NULL;
	if (db->db_state != DB_NOFILL) {
	db->db_state = DB_UNCACHED;
	DTRACE_SET_STATE(db, "clear data");
	}
	}

	static void
	dbuf_set_data(dmu_buf_impl_t db, arc_buf_t buf)
	{
	ASSERT(MUTEX_HELD(&db->db_mtx));
	ASSERT(buf != NULL);

	db->db_buf = buf;
	ASSERT(buf->b_data != NULL);
	db->db.db_data = buf->b_data;
	}

	static arc_buf_t *
	dbuf_alloc_arcbuf(dmu_buf_impl_t *db)
	{
	spa_t *spa = db->db_objset->os_spa;

	return (arc_alloc_buf(spa, db, DBUF_GET_BUFC_TYPE(db), db->db.db_size));
	}

	/*
	* Loan out an arc_buf for read. Return the loaned arc_buf.
	*/
	arc_buf_t *
	dbuf_loan_arcbuf(dmu_buf_impl_t *db)
	{
	arc_buf_t *abuf;

	ASSERT(db->db_blkid != DMU_BONUS_BLKID);
	mutex_enter(&db->db_mtx);
	if (arc_released(db->db_buf) \|\| zfs_refcount_count(&db->db_holds) > 1) {
	int blksz = db->db.db_size;
	spa_t *spa = db->db_objset->os_spa;

	mutex_exit(&db->db_mtx);
	abuf = arc_loan_buf(spa, B_FALSE, blksz);
	memcpy(abuf->b_data, db->db.db_data, blksz);
	} else {
	abuf = db->db_buf;
	arc_loan_inuse_buf(abuf, db);
	db->db_buf = NULL;
	dbuf_clear_data(db);
	mutex_exit(&db->db_mtx);
	}
	return (abuf);
	}

	/*
	* Calculate which level n block references the data at the level 0 offset
	* provided.
	*/
	uint64_t
	dbuf_whichblock(const dnode_t *dn, const int64_t level, const uint64_t offset)
	{
	if (dn->dn_datablkshift != 0 && dn->dn_indblkshift != 0) {
	/*
	* The level n blkid is equal to the level 0 blkid divided by
	* the number of level 0s in a level n block.
	*
	* The level 0 blkid is offset >> datablkshift =
	* offset / 2^datablkshift.
	*
	* The number of level 0s in a level n is the number of block
	* pointers in an indirect block, raised to the power of level.
	* This is 2^(indblkshift - SPA_BLKPTRSHIFT)^level =
	* 2^(level*(indblkshift - SPA_BLKPTRSHIFT)).
	*
	* Thus, the level n blkid is: offset /
	* ((2^datablkshift)(2^(level(indblkshift-SPA_BLKPTRSHIFT))))
	* = offset / 2^(datablkshift + level *
	* (indblkshift - SPA_BLKPTRSHIFT))
	* = offset >> (datablkshift + level *
	* (indblkshift - SPA_BLKPTRSHIFT))
	*/

	const unsigned exp = dn->dn_datablkshift +
	level * (dn->dn_indblkshift - SPA_BLKPTRSHIFT);

	if (exp >= 8 * sizeof (offset)) {
	/* This only happens on the highest indirection level */
	ASSERT3U(level, ==, dn->dn_nlevels - 1);
	return (0);
	}

	ASSERT3U(exp, <, 8 * sizeof (offset));

	return (offset >> exp);
	} else {
	ASSERT3U(offset, <, dn->dn_datablksz);
	return (0);
	}
	}

	/*
	* This function is used to lock the parent of the provided dbuf. This should be
	* used when modifying or reading db_blkptr.
	*/
	db_lock_type_t
	dmu_buf_lock_parent(dmu_buf_impl_t db, krw_t rw, const void tag)
	{
	enum db_lock_type ret = DLT_NONE;
	if (db->db_parent != NULL) {
	rw_enter(&db->db_parent->db_rwlock, rw);
	ret = DLT_PARENT;
	} else if (dmu_objset_ds(db->db_objset) != NULL) {
	rrw_enter(&dmu_objset_ds(db->db_objset)->ds_bp_rwlock, rw,
	tag);
	ret = DLT_OBJSET;
	}
	/*
	* We only return a DLT_NONE lock when it's the top-most indirect block
	* of the meta-dnode of the MOS.
	*/
	return (ret);
	}

	/*
	* We need to pass the lock type in because it's possible that the block will
	* move from being the topmost indirect block in a dnode (and thus, have no
	* parent) to not the top-most via an indirection increase. This would cause a
	* panic if we didn't pass the lock type in.
	*/
	void
	dmu_buf_unlock_parent(dmu_buf_impl_t db, db_lock_type_t type, const void tag)
	{
	if (type == DLT_PARENT)
	rw_exit(&db->db_parent->db_rwlock);
	else if (type == DLT_OBJSET)
	rrw_exit(&dmu_objset_ds(db->db_objset)->ds_bp_rwlock, tag);
	}

	static void
	dbuf_read_done(zio_t zio, const zbookmark_phys_t zb, const blkptr_t *bp,
	arc_buf_t buf, void vdb)
	{
	(void) zb, (void) bp;
	dmu_buf_impl_t *db = vdb;

	mutex_enter(&db->db_mtx);
	ASSERT3U(db->db_state, ==, DB_READ);
	/*
	* All reads are synchronous, so we must have a hold on the dbuf
	*/
	ASSERT(zfs_refcount_count(&db->db_holds) > 0);
	ASSERT(db->db_buf == NULL);
	ASSERT(db->db.db_data == NULL);
	if (buf == NULL) {
	/* i/o error */
	ASSERT(zio == NULL \|\| zio->io_error != 0);
	ASSERT(db->db_blkid != DMU_BONUS_BLKID);
	ASSERT3P(db->db_buf, ==, NULL);
	db->db_state = DB_UNCACHED;
	DTRACE_SET_STATE(db, "i/o error");
	} else if (db->db_level == 0 && db->db_freed_in_flight) {
	/* freed in flight */
	ASSERT(zio == NULL \|\| zio->io_error == 0);
	arc_release(buf, db);
	memset(buf->b_data, 0, db->db.db_size);
	arc_buf_freeze(buf);
	db->db_freed_in_flight = FALSE;
	dbuf_set_data(db, buf);
	db->db_state = DB_CACHED;
	DTRACE_SET_STATE(db, "freed in flight");
	} else {
	/* success */
	ASSERT(zio == NULL \|\| zio->io_error == 0);
	dbuf_set_data(db, buf);
	db->db_state = DB_CACHED;
	DTRACE_SET_STATE(db, "successful read");
	}
	cv_broadcast(&db->db_changed);
	dbuf_rele_and_unlock(db, NULL, B_FALSE);
	}

	/*
	* Shortcut for performing reads on bonus dbufs. Returns
	* an error if we fail to verify the dnode associated with
	* a decrypted block. Otherwise success.
	*/
	static int
	dbuf_read_bonus(dmu_buf_impl_t db, dnode_t dn)
	{
	int bonuslen, max_bonuslen;

	bonuslen = MIN(dn->dn_bonuslen, dn->dn_phys->dn_bonuslen);
	max_bonuslen = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots);
	ASSERT(MUTEX_HELD(&db->db_mtx));
	ASSERT(DB_DNODE_HELD(db));
	ASSERT3U(bonuslen, <=, db->db.db_size);
	db->db.db_data = kmem_alloc(max_bonuslen, KM_SLEEP);
	arc_space_consume(max_bonuslen, ARC_SPACE_BONUS);
	if (bonuslen < max_bonuslen)
	memset(db->db.db_data, 0, max_bonuslen);
	if (bonuslen)
	memcpy(db->db.db_data, DN_BONUS(dn->dn_phys), bonuslen);
	db->db_state = DB_CACHED;
	DTRACE_SET_STATE(db, "bonus buffer filled");
	return (0);
	}

	static void
	dbuf_handle_indirect_hole(dmu_buf_impl_t db, dnode_t dn, blkptr_t *dbbp)
	{
	blkptr_t *bps = db->db.db_data;
	uint32_t indbs = 1ULL << dn->dn_indblkshift;
	int n_bps = indbs >> SPA_BLKPTRSHIFT;

	for (int i = 0; i < n_bps; i++) {
	blkptr_t *bp = &bps[i];

	ASSERT3U(BP_GET_LSIZE(dbbp), ==, indbs);
	BP_SET_LSIZE(bp, BP_GET_LEVEL(dbbp) == 1 ?
	dn->dn_datablksz : BP_GET_LSIZE(dbbp));
	BP_SET_TYPE(bp, BP_GET_TYPE(dbbp));
	BP_SET_LEVEL(bp, BP_GET_LEVEL(dbbp) - 1);
	BP_SET_BIRTH(bp, BP_GET_LOGICAL_BIRTH(dbbp), 0);
	}
	}

	/*
	* Handle reads on dbufs that are holes, if necessary. This function
	* requires that the dbuf's mutex is held. Returns success (0) if action
	* was taken, ENOENT if no action was taken.
	*/
	static int
	dbuf_read_hole(dmu_buf_impl_t db, dnode_t dn, blkptr_t *bp)
	{
	ASSERT(MUTEX_HELD(&db->db_mtx));

	int is_hole = bp == NULL \|\| BP_IS_HOLE(bp);
	/*
	* For level 0 blocks only, if the above check fails:
	* Recheck BP_IS_HOLE() after dnode_block_freed() in case dnode_sync()
	* processes the delete record and clears the bp while we are waiting
	* for the dn_mtx (resulting in a "no" from block_freed).
	*/
	if (!is_hole && db->db_level == 0)
	is_hole = dnode_block_freed(dn, db->db_blkid) \|\| BP_IS_HOLE(bp);

	if (is_hole) {
	dbuf_set_data(db, dbuf_alloc_arcbuf(db));
	memset(db->db.db_data, 0, db->db.db_size);

	if (bp != NULL && db->db_level > 0 && BP_IS_HOLE(bp) &&
	BP_GET_LOGICAL_BIRTH(bp) != 0) {
	dbuf_handle_indirect_hole(db, dn, bp);
	}
	db->db_state = DB_CACHED;
	DTRACE_SET_STATE(db, "hole read satisfied");
	return (0);
	}
	return (ENOENT);
	}

	/*
	* This function ensures that, when doing a decrypting read of a block,
	* we make sure we have decrypted the dnode associated with it. We must do
	* this so that we ensure we are fully authenticating the checksum-of-MACs
	* tree from the root of the objset down to this block. Indirect blocks are
	* always verified against their secure checksum-of-MACs assuming that the
	* dnode containing them is correct. Now that we are doing a decrypting read,
	* we can be sure that the key is loaded and verify that assumption. This is
	* especially important considering that we always read encrypted dnode
	* blocks as raw data (without verifying their MACs) to start, and
	* decrypt / authenticate them when we need to read an encrypted bonus buffer.
	*/
	static int
	dbuf_read_verify_dnode_crypt(dmu_buf_impl_t db, dnode_t dn, uint32_t flags)
	{
	objset_t *os = db->db_objset;
	dmu_buf_impl_t *dndb;
	arc_buf_t *dnbuf;
	zbookmark_phys_t zb;
	int err;

	if ((flags & DB_RF_NO_DECRYPT) != 0 \|\|
	!os->os_encrypted \|\| os->os_raw_receive \|\|
	(dndb = dn->dn_dbuf) == NULL)
	return (0);

	dnbuf = dndb->db_buf;
	if (!arc_is_encrypted(dnbuf))
	return (0);

	mutex_enter(&dndb->db_mtx);

	/*
	* Since dnode buffer is modified by sync process, there can be only
	* one copy of it. It means we can not modify (decrypt) it while it
	* is being written. I don't see how this may happen now, since
	* encrypted dnode writes by receive should be completed before any
	* plain-text reads due to txg wait, but better be safe than sorry.
	*/
	while (1) {
	if (!arc_is_encrypted(dnbuf)) {
	mutex_exit(&dndb->db_mtx);
	return (0);
	}
	dbuf_dirty_record_t *dr = dndb->db_data_pending;
	if (dr == NULL \|\| dr->dt.dl.dr_data != dnbuf)
	break;
	cv_wait(&dndb->db_changed, &dndb->db_mtx);
	};

	SET_BOOKMARK(&zb, dmu_objset_id(os),
	DMU_META_DNODE_OBJECT, 0, dndb->db_blkid);
	err = arc_untransform(dnbuf, os->os_spa, &zb, B_TRUE);

	/*
	* An error code of EACCES tells us that the key is still not
	* available. This is ok if we are only reading authenticated
	* (and therefore non-encrypted) blocks.
	*/
	if (err == EACCES && ((db->db_blkid != DMU_BONUS_BLKID &&
	!DMU_OT_IS_ENCRYPTED(dn->dn_type)) \|\|
	(db->db_blkid == DMU_BONUS_BLKID &&
	!DMU_OT_IS_ENCRYPTED(dn->dn_bonustype))))
	err = 0;

	mutex_exit(&dndb->db_mtx);

	return (err);
	}

	/*
	* Drops db_mtx and the parent lock specified by dblt and tag before
	* returning.
	*/
	static int
	dbuf_read_impl(dmu_buf_impl_t db, dnode_t dn, zio_t *zio, uint32_t flags,
	db_lock_type_t dblt, const void *tag)
	{
	zbookmark_phys_t zb;
	uint32_t aflags = ARC_FLAG_NOWAIT;
	int err, zio_flags;
	blkptr_t bp, *bpp = NULL;

	ASSERT(!zfs_refcount_is_zero(&db->db_holds));
	ASSERT(MUTEX_HELD(&db->db_mtx));
	ASSERT(db->db_state == DB_UNCACHED \|\| db->db_state == DB_NOFILL);
	ASSERT(db->db_buf == NULL);
	ASSERT(db->db_parent == NULL \|\|
	RW_LOCK_HELD(&db->db_parent->db_rwlock));

	if (db->db_blkid == DMU_BONUS_BLKID) {
	err = dbuf_read_bonus(db, dn);
	goto early_unlock;
	}

	/*
	* If we have a pending block clone, we don't want to read the
	* underlying block, but the content of the block being cloned,
	* pointed by the dirty record, so we have the most recent data.
	* If there is no dirty record, then we hit a race in a sync
	* process when the dirty record is already removed, while the
	* dbuf is not yet destroyed. Such case is equivalent to uncached.
	*/
	if (db->db_state == DB_NOFILL) {
	dbuf_dirty_record_t *dr = list_head(&db->db_dirty_records);
	if (dr != NULL) {
	if (!dr->dt.dl.dr_brtwrite) {
	err = EIO;
	goto early_unlock;
	}
	bp = dr->dt.dl.dr_overridden_by;
	bpp = &bp;
	}
	}

	if (bpp == NULL && db->db_blkptr != NULL) {
	bp = *db->db_blkptr;
	bpp = &bp;
	}

	err = dbuf_read_hole(db, dn, bpp);
	if (err == 0)
	goto early_unlock;

	ASSERT(bpp != NULL);

	/*
	* Any attempt to read a redacted block should result in an error. This
	* will never happen under normal conditions, but can be useful for
	* debugging purposes.
	*/
	if (BP_IS_REDACTED(bpp)) {
	ASSERT(dsl_dataset_feature_is_active(
	db->db_objset->os_dsl_dataset,
	SPA_FEATURE_REDACTED_DATASETS));
	err = SET_ERROR(EIO);
	goto early_unlock;
	}

	SET_BOOKMARK(&zb, dmu_objset_id(db->db_objset),
	db->db.db_object, db->db_level, db->db_blkid);

	/*
	* All bps of an encrypted os should have the encryption bit set.
	* If this is not true it indicates tampering and we report an error.
	*/
	if (db->db_objset->os_encrypted && !BP_USES_CRYPT(bpp)) {
	spa_log_error(db->db_objset->os_spa, &zb,
	BP_GET_LOGICAL_BIRTH(bpp));
	err = SET_ERROR(EIO);
	goto early_unlock;
	}

	db->db_state = DB_READ;
	DTRACE_SET_STATE(db, "read issued");
	mutex_exit(&db->db_mtx);

	if (!DBUF_IS_CACHEABLE(db))
	aflags \|= ARC_FLAG_UNCACHED;
	else if (dbuf_is_l2cacheable(db))
	aflags \|= ARC_FLAG_L2CACHE;

	dbuf_add_ref(db, NULL);

	zio_flags = (flags & DB_RF_CANFAIL) ?
	ZIO_FLAG_CANFAIL : ZIO_FLAG_MUSTSUCCEED;

	if ((flags & DB_RF_NO_DECRYPT) && BP_IS_PROTECTED(db->db_blkptr))
	zio_flags \|= ZIO_FLAG_RAW;
	/*
	* The zio layer will copy the provided blkptr later, but we have our
	* own copy so that we can release the parent's rwlock. We have to
	* do that so that if dbuf_read_done is called synchronously (on
	* an l1 cache hit) we don't acquire the db_mtx while holding the
	* parent's rwlock, which would be a lock ordering violation.
	*/
	dmu_buf_unlock_parent(db, dblt, tag);
	return (arc_read(zio, db->db_objset->os_spa, bpp,
	dbuf_read_done, db, ZIO_PRIORITY_SYNC_READ, zio_flags,
	&aflags, &zb));

	early_unlock:
	mutex_exit(&db->db_mtx);
	dmu_buf_unlock_parent(db, dblt, tag);
	return (err);
	}

	/*
	* This is our just-in-time copy function. It makes a copy of buffers that
	* have been modified in a previous transaction group before we access them in
	* the current active group.
	*
	* This function is used in three places: when we are dirtying a buffer for the
	* first time in a txg, when we are freeing a range in a dnode that includes
	* this buffer, and when we are accessing a buffer which was received compressed
	* and later referenced in a WRITE_BYREF record.
	*
	* Note that when we are called from dbuf_free_range() we do not put a hold on
	* the buffer, we just traverse the active dbuf list for the dnode.
	*/
	static void
	dbuf_fix_old_data(dmu_buf_impl_t *db, uint64_t txg)
	{
	dbuf_dirty_record_t *dr = list_head(&db->db_dirty_records);

	ASSERT(MUTEX_HELD(&db->db_mtx));
	ASSERT(db->db.db_data != NULL);
	ASSERT(db->db_level == 0);
	ASSERT(db->db.db_object != DMU_META_DNODE_OBJECT);

	if (dr == NULL \|\|
	(dr->dt.dl.dr_data !=
	((db->db_blkid == DMU_BONUS_BLKID) ? db->db.db_data : db->db_buf)))
	return;

	/*
	* If the last dirty record for this dbuf has not yet synced
	* and its referencing the dbuf data, either:
	* reset the reference to point to a new copy,
	* or (if there a no active holders)
	* just null out the current db_data pointer.
	*/
	ASSERT3U(dr->dr_txg, >=, txg - 2);
	if (db->db_blkid == DMU_BONUS_BLKID) {
	dnode_t *dn = DB_DNODE(db);
	int bonuslen = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots);
	dr->dt.dl.dr_data = kmem_alloc(bonuslen, KM_SLEEP);
	arc_space_consume(bonuslen, ARC_SPACE_BONUS);
	memcpy(dr->dt.dl.dr_data, db->db.db_data, bonuslen);
	} else if (zfs_refcount_count(&db->db_holds) > db->db_dirtycnt) {
	dnode_t *dn = DB_DNODE(db);
	int size = arc_buf_size(db->db_buf);
	arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db);
	spa_t *spa = db->db_objset->os_spa;
	enum zio_compress compress_type =
	arc_get_compression(db->db_buf);
	uint8_t complevel = arc_get_complevel(db->db_buf);

	if (arc_is_encrypted(db->db_buf)) {
	boolean_t byteorder;
	uint8_t salt[ZIO_DATA_SALT_LEN];
	uint8_t iv[ZIO_DATA_IV_LEN];
	uint8_t mac[ZIO_DATA_MAC_LEN];

	arc_get_raw_params(db->db_buf, &byteorder, salt,
	iv, mac);
	dr->dt.dl.dr_data = arc_alloc_raw_buf(spa, db,
	dmu_objset_id(dn->dn_objset), byteorder, salt, iv,
	mac, dn->dn_type, size, arc_buf_lsize(db->db_buf),
	compress_type, complevel);
	} else if (compress_type != ZIO_COMPRESS_OFF) {
	ASSERT3U(type, ==, ARC_BUFC_DATA);
	dr->dt.dl.dr_data = arc_alloc_compressed_buf(spa, db,
	size, arc_buf_lsize(db->db_buf), compress_type,
	complevel);
	} else {
	dr->dt.dl.dr_data = arc_alloc_buf(spa, db, type, size);
	}
	memcpy(dr->dt.dl.dr_data->b_data, db->db.db_data, size);
	} else {
	db->db_buf = NULL;
	dbuf_clear_data(db);
	}
	}

	int
	dbuf_read(dmu_buf_impl_t db, zio_t pio, uint32_t flags)
	{
	dnode_t *dn;
	boolean_t miss = B_TRUE, need_wait = B_FALSE, prefetch;
	int err;

	ASSERT(!zfs_refcount_is_zero(&db->db_holds));

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);

	/*
	* Ensure that this block's dnode has been decrypted if the caller
	* has requested decrypted data.
	*/
	err = dbuf_read_verify_dnode_crypt(db, dn, flags);
	if (err != 0)
	goto done;

	prefetch = db->db_level == 0 && db->db_blkid != DMU_BONUS_BLKID &&
	(flags & DB_RF_NOPREFETCH) == 0;

	mutex_enter(&db->db_mtx);
	if (flags & DB_RF_PARTIAL_FIRST)
	db->db_partial_read = B_TRUE;
	else if (!(flags & DB_RF_PARTIAL_MORE))
	db->db_partial_read = B_FALSE;
	miss = (db->db_state != DB_CACHED);

	if (db->db_state == DB_READ \|\| db->db_state == DB_FILL) {
	/*
	* Another reader came in while the dbuf was in flight between
	* UNCACHED and CACHED. Either a writer will finish filling
	* the buffer, sending the dbuf to CACHED, or the first reader's
	* request will reach the read_done callback and send the dbuf
	* to CACHED. Otherwise, a failure occurred and the dbuf will
	* be sent to UNCACHED.
	*/
	if (flags & DB_RF_NEVERWAIT) {
	mutex_exit(&db->db_mtx);
	DB_DNODE_EXIT(db);
	goto done;
	}
	do {
	ASSERT(db->db_state == DB_READ \|\|
	(flags & DB_RF_HAVESTRUCT) == 0);
	DTRACE_PROBE2(blocked__read, dmu_buf_impl_t *, db,
	zio_t *, pio);
	cv_wait(&db->db_changed, &db->db_mtx);
	} while (db->db_state == DB_READ \|\| db->db_state == DB_FILL);
	if (db->db_state == DB_UNCACHED) {
	err = SET_ERROR(EIO);
	mutex_exit(&db->db_mtx);
	DB_DNODE_EXIT(db);
	goto done;
	}
	}

	if (db->db_state == DB_CACHED) {
	/*
	* If the arc buf is compressed or encrypted and the caller
	* requested uncompressed data, we need to untransform it
	* before returning. We also call arc_untransform() on any
	* unauthenticated blocks, which will verify their MAC if
	* the key is now available.
	*/
	if ((flags & DB_RF_NO_DECRYPT) == 0 && db->db_buf != NULL &&
	(arc_is_encrypted(db->db_buf) \|\|
	arc_is_unauthenticated(db->db_buf) \|\|
	arc_get_compression(db->db_buf) != ZIO_COMPRESS_OFF)) {
	spa_t *spa = dn->dn_objset->os_spa;
	zbookmark_phys_t zb;

	SET_BOOKMARK(&zb, dmu_objset_id(db->db_objset),
	db->db.db_object, db->db_level, db->db_blkid);
	dbuf_fix_old_data(db, spa_syncing_txg(spa));
	err = arc_untransform(db->db_buf, spa, &zb, B_FALSE);
	dbuf_set_data(db, db->db_buf);
	}
	mutex_exit(&db->db_mtx);
	} else {
	ASSERT(db->db_state == DB_UNCACHED \|\|
	db->db_state == DB_NOFILL);
	db_lock_type_t dblt = dmu_buf_lock_parent(db, RW_READER, FTAG);
	if (pio == NULL && (db->db_state == DB_NOFILL \|\|
	(db->db_blkptr != NULL && !BP_IS_HOLE(db->db_blkptr)))) {
	spa_t *spa = dn->dn_objset->os_spa;
	pio = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL);
	need_wait = B_TRUE;
	}
	err = dbuf_read_impl(db, dn, pio, flags, dblt, FTAG);
	/* dbuf_read_impl drops db_mtx and parent's rwlock. */
	miss = (db->db_state != DB_CACHED);
	}

	if (err == 0 && prefetch) {
	dmu_zfetch(&dn->dn_zfetch, db->db_blkid, 1, B_TRUE, miss,
	flags & DB_RF_HAVESTRUCT);
	}
	DB_DNODE_EXIT(db);

	/*
	* If we created a zio we must execute it to avoid leaking it, even if
	* it isn't attached to any work due to an error in dbuf_read_impl().
	*/
	if (need_wait) {
	if (err == 0)
	err = zio_wait(pio);
	else
	(void) zio_wait(pio);
	pio = NULL;
	}

	done:
	if (miss)
	DBUF_STAT_BUMP(hash_misses);
	else
	DBUF_STAT_BUMP(hash_hits);
	if (pio && err != 0) {
	zio_t *zio = zio_null(pio, pio->io_spa, NULL, NULL, NULL,
	ZIO_FLAG_CANFAIL);
	zio->io_error = err;
	zio_nowait(zio);
	}

	return (err);
	}

	static void
	dbuf_noread(dmu_buf_impl_t *db)
	{
	ASSERT(!zfs_refcount_is_zero(&db->db_holds));
	ASSERT(db->db_blkid != DMU_BONUS_BLKID);
	mutex_enter(&db->db_mtx);
	while (db->db_state == DB_READ \|\| db->db_state == DB_FILL)
	cv_wait(&db->db_changed, &db->db_mtx);
	if (db->db_state == DB_UNCACHED) {
	ASSERT(db->db_buf == NULL);
	ASSERT(db->db.db_data == NULL);
	dbuf_set_data(db, dbuf_alloc_arcbuf(db));
	db->db_state = DB_FILL;
	DTRACE_SET_STATE(db, "assigning filled buffer");
	} else if (db->db_state == DB_NOFILL) {
	dbuf_clear_data(db);
	} else {
	ASSERT3U(db->db_state, ==, DB_CACHED);
	}
	mutex_exit(&db->db_mtx);
	}

	void
	dbuf_unoverride(dbuf_dirty_record_t *dr)
	{
	dmu_buf_impl_t *db = dr->dr_dbuf;
	blkptr_t *bp = &dr->dt.dl.dr_overridden_by;
	uint64_t txg = dr->dr_txg;

	ASSERT(MUTEX_HELD(&db->db_mtx));
	/*
	* This assert is valid because dmu_sync() expects to be called by
	* a zilog's get_data while holding a range lock. This call only
	* comes from dbuf_dirty() callers who must also hold a range lock.
	*/
	ASSERT(dr->dt.dl.dr_override_state != DR_IN_DMU_SYNC);
	ASSERT(db->db_level == 0);

	if (db->db_blkid == DMU_BONUS_BLKID \|\|
	dr->dt.dl.dr_override_state == DR_NOT_OVERRIDDEN)
	return;

	ASSERT(db->db_data_pending != dr);

	/* free this block */
	if (!BP_IS_HOLE(bp) && !dr->dt.dl.dr_nopwrite)
	zio_free(db->db_objset->os_spa, txg, bp);

	if (dr->dt.dl.dr_brtwrite) {
	ASSERT0P(dr->dt.dl.dr_data);
	dr->dt.dl.dr_data = db->db_buf;
	}
	dr->dt.dl.dr_override_state = DR_NOT_OVERRIDDEN;
	dr->dt.dl.dr_nopwrite = B_FALSE;
	dr->dt.dl.dr_brtwrite = B_FALSE;
	dr->dt.dl.dr_has_raw_params = B_FALSE;

	/*
	* Release the already-written buffer, so we leave it in
	* a consistent dirty state. Note that all callers are
	* modifying the buffer, so they will immediately do
	* another (redundant) arc_release(). Therefore, leave
	* the buf thawed to save the effort of freezing &
	* immediately re-thawing it.
	*/
	if (dr->dt.dl.dr_data)
	arc_release(dr->dt.dl.dr_data, db);
	}

	/*
	* Evict (if its unreferenced) or clear (if its referenced) any level-0
	* data blocks in the free range, so that any future readers will find
	* empty blocks.
	*/
	void
	dbuf_free_range(dnode_t *dn, uint64_t start_blkid, uint64_t end_blkid,
	dmu_tx_t *tx)
	{
	dmu_buf_impl_t *db_search;
	dmu_buf_impl_t db, db_next;
	uint64_t txg = tx->tx_txg;
	avl_index_t where;
	dbuf_dirty_record_t *dr;

	if (end_blkid > dn->dn_maxblkid &&
	!(start_blkid == DMU_SPILL_BLKID \|\| end_blkid == DMU_SPILL_BLKID))
	end_blkid = dn->dn_maxblkid;
	dprintf_dnode(dn, "start=%llu end=%llu\n", (u_longlong_t)start_blkid,
	(u_longlong_t)end_blkid);

	db_search = kmem_alloc(sizeof (dmu_buf_impl_t), KM_SLEEP);
	db_search->db_level = 0;
	db_search->db_blkid = start_blkid;
	db_search->db_state = DB_SEARCH;

	mutex_enter(&dn->dn_dbufs_mtx);
	db = avl_find(&dn->dn_dbufs, db_search, &where);
	ASSERT3P(db, ==, NULL);

	db = avl_nearest(&dn->dn_dbufs, where, AVL_AFTER);

	for (; db != NULL; db = db_next) {
	db_next = AVL_NEXT(&dn->dn_dbufs, db);
	ASSERT(db->db_blkid != DMU_BONUS_BLKID);

	if (db->db_level != 0 \|\| db->db_blkid > end_blkid) {
	break;
	}
	ASSERT3U(db->db_blkid, >=, start_blkid);

	/* found a level 0 buffer in the range */
	mutex_enter(&db->db_mtx);
	if (dbuf_undirty(db, tx)) {
	/* mutex has been dropped and dbuf destroyed */
	continue;
	}

	if (db->db_state == DB_UNCACHED \|\|
	db->db_state == DB_NOFILL \|\|
	db->db_state == DB_EVICTING) {
	ASSERT(db->db.db_data == NULL);
	mutex_exit(&db->db_mtx);
	continue;
	}
	if (db->db_state == DB_READ \|\| db->db_state == DB_FILL) {
	/* will be handled in dbuf_read_done or dbuf_rele */
	db->db_freed_in_flight = TRUE;
	mutex_exit(&db->db_mtx);
	continue;
	}
	if (zfs_refcount_count(&db->db_holds) == 0) {
	ASSERT(db->db_buf);
	dbuf_destroy(db);
	continue;
	}
	/* The dbuf is referenced */

	dr = list_head(&db->db_dirty_records);
	if (dr != NULL) {
	if (dr->dr_txg == txg) {
	/*
	* This buffer is "in-use", re-adjust the file
	* size to reflect that this buffer may
	* contain new data when we sync.
	*/
	if (db->db_blkid != DMU_SPILL_BLKID &&
	db->db_blkid > dn->dn_maxblkid)
	dn->dn_maxblkid = db->db_blkid;
	dbuf_unoverride(dr);
	} else {
	/*
	* This dbuf is not dirty in the open context.
	* Either uncache it (if its not referenced in
	* the open context) or reset its contents to
	* empty.
	*/
	dbuf_fix_old_data(db, txg);
	}
	}
	/* clear the contents if its cached */
	if (db->db_state == DB_CACHED) {
	ASSERT(db->db.db_data != NULL);
	arc_release(db->db_buf, db);
	rw_enter(&db->db_rwlock, RW_WRITER);
	memset(db->db.db_data, 0, db->db.db_size);
	rw_exit(&db->db_rwlock);
	arc_buf_freeze(db->db_buf);
	}

	mutex_exit(&db->db_mtx);
	}

	mutex_exit(&dn->dn_dbufs_mtx);
	kmem_free(db_search, sizeof (dmu_buf_impl_t));
	}

	void
	dbuf_new_size(dmu_buf_impl_t db, int size, dmu_tx_t tx)
	{
	arc_buf_t buf, old_buf;
	dbuf_dirty_record_t *dr;
	int osize = db->db.db_size;
	arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db);
	dnode_t *dn;

	ASSERT(db->db_blkid != DMU_BONUS_BLKID);

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);

	/*
	* XXX we should be doing a dbuf_read, checking the return
	* value and returning that up to our callers
	*/
	dmu_buf_will_dirty(&db->db, tx);

	/* create the data buffer for the new block */
	buf = arc_alloc_buf(dn->dn_objset->os_spa, db, type, size);

	/* copy old block data to the new block */
	old_buf = db->db_buf;
	memcpy(buf->b_data, old_buf->b_data, MIN(osize, size));
	/* zero the remainder */
	if (size > osize)
	memset((uint8_t *)buf->b_data + osize, 0, size - osize);

	mutex_enter(&db->db_mtx);
	dbuf_set_data(db, buf);
	arc_buf_destroy(old_buf, db);
	db->db.db_size = size;

	dr = list_head(&db->db_dirty_records);
	/* dirty record added by dmu_buf_will_dirty() */
	VERIFY(dr != NULL);
	if (db->db_level == 0)
	dr->dt.dl.dr_data = buf;
	ASSERT3U(dr->dr_txg, ==, tx->tx_txg);
	ASSERT3U(dr->dr_accounted, ==, osize);
	dr->dr_accounted = size;
	mutex_exit(&db->db_mtx);

	dmu_objset_willuse_space(dn->dn_objset, size - osize, tx);
	DB_DNODE_EXIT(db);
	}

	void
	dbuf_release_bp(dmu_buf_impl_t *db)
	{
	objset_t *os __maybe_unused = db->db_objset;

	ASSERT(dsl_pool_sync_context(dmu_objset_pool(os)));
	ASSERT(arc_released(os->os_phys_buf) \|\|
	list_link_active(&os->os_dsl_dataset->ds_synced_link));
	ASSERT(db->db_parent == NULL \|\| arc_released(db->db_parent->db_buf));

	(void) arc_release(db->db_buf, db);
	}

	/*
	* We already have a dirty record for this TXG, and we are being
	* dirtied again.
	*/
	static void
	dbuf_redirty(dbuf_dirty_record_t *dr)
	{
	dmu_buf_impl_t *db = dr->dr_dbuf;

	ASSERT(MUTEX_HELD(&db->db_mtx));

	if (db->db_level == 0 && db->db_blkid != DMU_BONUS_BLKID) {
	/*
	* If this buffer has already been written out,
	* we now need to reset its state.
	*/
	dbuf_unoverride(dr);
	if (db->db.db_object != DMU_META_DNODE_OBJECT &&
	db->db_state != DB_NOFILL) {
	/* Already released on initial dirty, so just thaw. */
	ASSERT(arc_released(db->db_buf));
	arc_buf_thaw(db->db_buf);
	}
	}
	}

	dbuf_dirty_record_t *
	dbuf_dirty_lightweight(dnode_t dn, uint64_t blkid, dmu_tx_t tx)
	{
	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	IMPLY(dn->dn_objset->os_raw_receive, dn->dn_maxblkid >= blkid);
	dnode_new_blkid(dn, blkid, tx, B_TRUE, B_FALSE);
	ASSERT(dn->dn_maxblkid >= blkid);

	dbuf_dirty_record_t dr = kmem_zalloc(sizeof (dr), KM_SLEEP);
	list_link_init(&dr->dr_dirty_node);
	list_link_init(&dr->dr_dbuf_node);
	dr->dr_dnode = dn;
	dr->dr_txg = tx->tx_txg;
	dr->dt.dll.dr_blkid = blkid;
	dr->dr_accounted = dn->dn_datablksz;

	/*
	* There should not be any dbuf for the block that we're dirtying.
	* Otherwise the buffer contents could be inconsistent between the
	* dbuf and the lightweight dirty record.
	*/
	ASSERT3P(NULL, ==, dbuf_find(dn->dn_objset, dn->dn_object, 0, blkid,
	NULL));

	mutex_enter(&dn->dn_mtx);
	int txgoff = tx->tx_txg & TXG_MASK;
	if (dn->dn_free_ranges[txgoff] != NULL) {
	range_tree_clear(dn->dn_free_ranges[txgoff], blkid, 1);
	}

	if (dn->dn_nlevels == 1) {
	ASSERT3U(blkid, <, dn->dn_nblkptr);
	list_insert_tail(&dn->dn_dirty_records[txgoff], dr);
	mutex_exit(&dn->dn_mtx);
	rw_exit(&dn->dn_struct_rwlock);
	dnode_setdirty(dn, tx);
	} else {
	mutex_exit(&dn->dn_mtx);

	int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
	dmu_buf_impl_t *parent_db = dbuf_hold_level(dn,
	1, blkid >> epbs, FTAG);
	rw_exit(&dn->dn_struct_rwlock);
	if (parent_db == NULL) {
	kmem_free(dr, sizeof (*dr));
	return (NULL);
	}
	int err = dbuf_read(parent_db, NULL,
	(DB_RF_NOPREFETCH \| DB_RF_CANFAIL));
	if (err != 0) {
	dbuf_rele(parent_db, FTAG);
	kmem_free(dr, sizeof (*dr));
	return (NULL);
	}

	dbuf_dirty_record_t *parent_dr = dbuf_dirty(parent_db, tx);
	dbuf_rele(parent_db, FTAG);
	mutex_enter(&parent_dr->dt.di.dr_mtx);
	ASSERT3U(parent_dr->dr_txg, ==, tx->tx_txg);
	list_insert_tail(&parent_dr->dt.di.dr_children, dr);
	mutex_exit(&parent_dr->dt.di.dr_mtx);
	dr->dr_parent = parent_dr;
	}

	dmu_objset_willuse_space(dn->dn_objset, dr->dr_accounted, tx);

	return (dr);
	}

	dbuf_dirty_record_t *
	dbuf_dirty(dmu_buf_impl_t db, dmu_tx_t tx)
	{
	dnode_t *dn;
	objset_t *os;
	dbuf_dirty_record_t dr, dr_next, *dr_head;
	int txgoff = tx->tx_txg & TXG_MASK;
	boolean_t drop_struct_rwlock = B_FALSE;

	ASSERT(tx->tx_txg != 0);
	ASSERT(!zfs_refcount_is_zero(&db->db_holds));
	DMU_TX_DIRTY_BUF(tx, db);

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	/*
	* Shouldn't dirty a regular buffer in syncing context. Private
	* objects may be dirtied in syncing context, but only if they
	* were already pre-dirtied in open context.
	*/
	#ifdef ZFS_DEBUG
	if (dn->dn_objset->os_dsl_dataset != NULL) {
	rrw_enter(&dn->dn_objset->os_dsl_dataset->ds_bp_rwlock,
	RW_READER, FTAG);
	}
	ASSERT(!dmu_tx_is_syncing(tx) \|\|
	BP_IS_HOLE(dn->dn_objset->os_rootbp) \|\|
	DMU_OBJECT_IS_SPECIAL(dn->dn_object) \|\|
	dn->dn_objset->os_dsl_dataset == NULL);
	if (dn->dn_objset->os_dsl_dataset != NULL)
	rrw_exit(&dn->dn_objset->os_dsl_dataset->ds_bp_rwlock, FTAG);
	#endif
	/*
	* We make this assert for private objects as well, but after we
	* check if we're already dirty. They are allowed to re-dirty
	* in syncing context.
	*/
	ASSERT(dn->dn_object == DMU_META_DNODE_OBJECT \|\|
	dn->dn_dirtyctx == DN_UNDIRTIED \|\| dn->dn_dirtyctx ==
	(dmu_tx_is_syncing(tx) ? DN_DIRTY_SYNC : DN_DIRTY_OPEN));

	mutex_enter(&db->db_mtx);
	/*
	* XXX make this true for indirects too? The problem is that
	* transactions created with dmu_tx_create_assigned() from
	* syncing context don't bother holding ahead.
	*/
	ASSERT(db->db_level != 0 \|\|
	db->db_state == DB_CACHED \|\| db->db_state == DB_FILL \|\|
	db->db_state == DB_NOFILL);

	mutex_enter(&dn->dn_mtx);
	dnode_set_dirtyctx(dn, tx, db);
	if (tx->tx_txg > dn->dn_dirty_txg)
	dn->dn_dirty_txg = tx->tx_txg;
	mutex_exit(&dn->dn_mtx);

	if (db->db_blkid == DMU_SPILL_BLKID)
	dn->dn_have_spill = B_TRUE;

	/*
	* If this buffer is already dirty, we're done.
	*/
	dr_head = list_head(&db->db_dirty_records);
	ASSERT(dr_head == NULL \|\| dr_head->dr_txg <= tx->tx_txg \|\|
	db->db.db_object == DMU_META_DNODE_OBJECT);
	dr_next = dbuf_find_dirty_lte(db, tx->tx_txg);
	if (dr_next && dr_next->dr_txg == tx->tx_txg) {
	DB_DNODE_EXIT(db);

	dbuf_redirty(dr_next);
	mutex_exit(&db->db_mtx);
	return (dr_next);
	}

	/*
	* Only valid if not already dirty.
	*/
	ASSERT(dn->dn_object == 0 \|\|
	dn->dn_dirtyctx == DN_UNDIRTIED \|\| dn->dn_dirtyctx ==
	(dmu_tx_is_syncing(tx) ? DN_DIRTY_SYNC : DN_DIRTY_OPEN));

	ASSERT3U(dn->dn_nlevels, >, db->db_level);

	/*
	* We should only be dirtying in syncing context if it's the
	* mos or we're initializing the os or it's a special object.
	* However, we are allowed to dirty in syncing context provided
	* we already dirtied it in open context. Hence we must make
	* this assertion only if we're not already dirty.
	*/
	os = dn->dn_objset;
	VERIFY3U(tx->tx_txg, <=, spa_final_dirty_txg(os->os_spa));
	#ifdef ZFS_DEBUG
	if (dn->dn_objset->os_dsl_dataset != NULL)
	rrw_enter(&os->os_dsl_dataset->ds_bp_rwlock, RW_READER, FTAG);
	ASSERT(!dmu_tx_is_syncing(tx) \|\| DMU_OBJECT_IS_SPECIAL(dn->dn_object) \|\|
	os->os_dsl_dataset == NULL \|\| BP_IS_HOLE(os->os_rootbp));
	if (dn->dn_objset->os_dsl_dataset != NULL)
	rrw_exit(&os->os_dsl_dataset->ds_bp_rwlock, FTAG);
	#endif
	ASSERT(db->db.db_size != 0);

	dprintf_dbuf(db, "size=%llx\n", (u_longlong_t)db->db.db_size);

	if (db->db_blkid != DMU_BONUS_BLKID && db->db_state != DB_NOFILL) {
	dmu_objset_willuse_space(os, db->db.db_size, tx);
	}

	/*
	* If this buffer is dirty in an old transaction group we need
	* to make a copy of it so that the changes we make in this
	* transaction group won't leak out when we sync the older txg.
	*/
	dr = kmem_zalloc(sizeof (dbuf_dirty_record_t), KM_SLEEP);
	list_link_init(&dr->dr_dirty_node);
	list_link_init(&dr->dr_dbuf_node);
	dr->dr_dnode = dn;
	if (db->db_level == 0) {
	void *data_old = db->db_buf;

	if (db->db_state != DB_NOFILL) {
	if (db->db_blkid == DMU_BONUS_BLKID) {
	dbuf_fix_old_data(db, tx->tx_txg);
	data_old = db->db.db_data;
	} else if (db->db.db_object != DMU_META_DNODE_OBJECT) {
	/*
	* Release the data buffer from the cache so
	* that we can modify it without impacting
	* possible other users of this cached data
	* block. Note that indirect blocks and
	* private objects are not released until the
	* syncing state (since they are only modified
	* then).
	*/
	arc_release(db->db_buf, db);
	dbuf_fix_old_data(db, tx->tx_txg);
	data_old = db->db_buf;
	}
	ASSERT(data_old != NULL);
	}
	dr->dt.dl.dr_data = data_old;
	} else {
	mutex_init(&dr->dt.di.dr_mtx, NULL, MUTEX_NOLOCKDEP, NULL);
	list_create(&dr->dt.di.dr_children,
	sizeof (dbuf_dirty_record_t),
	offsetof(dbuf_dirty_record_t, dr_dirty_node));
	}
	if (db->db_blkid != DMU_BONUS_BLKID && db->db_state != DB_NOFILL) {
	dr->dr_accounted = db->db.db_size;
	}
	dr->dr_dbuf = db;
	dr->dr_txg = tx->tx_txg;
	list_insert_before(&db->db_dirty_records, dr_next, dr);

	/*
	* We could have been freed_in_flight between the dbuf_noread
	* and dbuf_dirty. We win, as though the dbuf_noread() had
	* happened after the free.
	*/
	if (db->db_level == 0 && db->db_blkid != DMU_BONUS_BLKID &&
	db->db_blkid != DMU_SPILL_BLKID) {
	mutex_enter(&dn->dn_mtx);
	if (dn->dn_free_ranges[txgoff] != NULL) {
	range_tree_clear(dn->dn_free_ranges[txgoff],
	db->db_blkid, 1);
	}
	mutex_exit(&dn->dn_mtx);
	db->db_freed_in_flight = FALSE;
	}

	/*
	* This buffer is now part of this txg
	*/
	dbuf_add_ref(db, (void *)(uintptr_t)tx->tx_txg);
	db->db_dirtycnt += 1;
	ASSERT3U(db->db_dirtycnt, <=, 3);

	mutex_exit(&db->db_mtx);

	if (db->db_blkid == DMU_BONUS_BLKID \|\|
	db->db_blkid == DMU_SPILL_BLKID) {
	mutex_enter(&dn->dn_mtx);
	ASSERT(!list_link_active(&dr->dr_dirty_node));
	list_insert_tail(&dn->dn_dirty_records[txgoff], dr);
	mutex_exit(&dn->dn_mtx);
	dnode_setdirty(dn, tx);
	DB_DNODE_EXIT(db);
	return (dr);
	}

	if (!RW_WRITE_HELD(&dn->dn_struct_rwlock)) {
	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	drop_struct_rwlock = B_TRUE;
	}

	/*
	* If we are overwriting a dedup BP, then unless it is snapshotted,
	* when we get to syncing context we will need to decrement its
	* refcount in the DDT. Prefetch the relevant DDT block so that
	* syncing context won't have to wait for the i/o.
	*/
	if (db->db_blkptr != NULL) {
	db_lock_type_t dblt = dmu_buf_lock_parent(db, RW_READER, FTAG);
	ddt_prefetch(os->os_spa, db->db_blkptr);
	dmu_buf_unlock_parent(db, dblt, FTAG);
	}

	/*
	* We need to hold the dn_struct_rwlock to make this assertion,
	* because it protects dn_phys / dn_next_nlevels from changing.
	*/
	ASSERT((dn->dn_phys->dn_nlevels == 0 && db->db_level == 0) \|\|
	dn->dn_phys->dn_nlevels > db->db_level \|\|
	dn->dn_next_nlevels[txgoff] > db->db_level \|\|
	dn->dn_next_nlevels[(tx->tx_txg-1) & TXG_MASK] > db->db_level \|\|
	dn->dn_next_nlevels[(tx->tx_txg-2) & TXG_MASK] > db->db_level);


	if (db->db_level == 0) {
	ASSERT(!db->db_objset->os_raw_receive \|\|
	dn->dn_maxblkid >= db->db_blkid);
	dnode_new_blkid(dn, db->db_blkid, tx,
	drop_struct_rwlock, B_FALSE);
	ASSERT(dn->dn_maxblkid >= db->db_blkid);
	}

	if (db->db_level+1 < dn->dn_nlevels) {
	dmu_buf_impl_t *parent = db->db_parent;
	dbuf_dirty_record_t *di;
	int parent_held = FALSE;

	if (db->db_parent == NULL \|\| db->db_parent == dn->dn_dbuf) {
	int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
	parent = dbuf_hold_level(dn, db->db_level + 1,
	db->db_blkid >> epbs, FTAG);
	ASSERT(parent != NULL);
	parent_held = TRUE;
	}
	if (drop_struct_rwlock)
	rw_exit(&dn->dn_struct_rwlock);
	ASSERT3U(db->db_level + 1, ==, parent->db_level);
	di = dbuf_dirty(parent, tx);
	if (parent_held)
	dbuf_rele(parent, FTAG);

	mutex_enter(&db->db_mtx);
	/*
	* Since we've dropped the mutex, it's possible that
	* dbuf_undirty() might have changed this out from under us.
	*/
	if (list_head(&db->db_dirty_records) == dr \|\|
	dn->dn_object == DMU_META_DNODE_OBJECT) {
	mutex_enter(&di->dt.di.dr_mtx);
	ASSERT3U(di->dr_txg, ==, tx->tx_txg);
	ASSERT(!list_link_active(&dr->dr_dirty_node));
	list_insert_tail(&di->dt.di.dr_children, dr);
	mutex_exit(&di->dt.di.dr_mtx);
	dr->dr_parent = di;
	}
	mutex_exit(&db->db_mtx);
	} else {
	ASSERT(db->db_level + 1 == dn->dn_nlevels);
	ASSERT(db->db_blkid < dn->dn_nblkptr);
	ASSERT(db->db_parent == NULL \|\| db->db_parent == dn->dn_dbuf);
	mutex_enter(&dn->dn_mtx);
	ASSERT(!list_link_active(&dr->dr_dirty_node));
	list_insert_tail(&dn->dn_dirty_records[txgoff], dr);
	mutex_exit(&dn->dn_mtx);
	if (drop_struct_rwlock)
	rw_exit(&dn->dn_struct_rwlock);
	}

	dnode_setdirty(dn, tx);
	DB_DNODE_EXIT(db);
	return (dr);
	}

	static void
	dbuf_undirty_bonus(dbuf_dirty_record_t *dr)
	{
	dmu_buf_impl_t *db = dr->dr_dbuf;

	if (dr->dt.dl.dr_data != db->db.db_data) {
	struct dnode *dn = dr->dr_dnode;
	int max_bonuslen = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots);

	kmem_free(dr->dt.dl.dr_data, max_bonuslen);
	arc_space_return(max_bonuslen, ARC_SPACE_BONUS);
	}
	db->db_data_pending = NULL;
	ASSERT(list_next(&db->db_dirty_records, dr) == NULL);
	list_remove(&db->db_dirty_records, dr);
	if (dr->dr_dbuf->db_level != 0) {
	mutex_destroy(&dr->dt.di.dr_mtx);
	list_destroy(&dr->dt.di.dr_children);
	}
	kmem_free(dr, sizeof (dbuf_dirty_record_t));
	ASSERT3U(db->db_dirtycnt, >, 0);
	db->db_dirtycnt -= 1;
	}

	/*
	* Undirty a buffer in the transaction group referenced by the given
	* transaction. Return whether this evicted the dbuf.
	*/
	boolean_t
	dbuf_undirty(dmu_buf_impl_t db, dmu_tx_t tx)
	{
	uint64_t txg = tx->tx_txg;
	boolean_t brtwrite;

	ASSERT(txg != 0);

	/*
	* Due to our use of dn_nlevels below, this can only be called
	* in open context, unless we are operating on the MOS.
	* From syncing context, dn_nlevels may be different from the
	* dn_nlevels used when dbuf was dirtied.
	*/
	ASSERT(db->db_objset ==
	dmu_objset_pool(db->db_objset)->dp_meta_objset \|\|
	txg != spa_syncing_txg(dmu_objset_spa(db->db_objset)));
	ASSERT(db->db_blkid != DMU_BONUS_BLKID);
	ASSERT0(db->db_level);
	ASSERT(MUTEX_HELD(&db->db_mtx));

	/*
	* If this buffer is not dirty, we're done.
	*/
	dbuf_dirty_record_t *dr = dbuf_find_dirty_eq(db, txg);
	if (dr == NULL)
	return (B_FALSE);
	ASSERT(dr->dr_dbuf == db);

	brtwrite = dr->dt.dl.dr_brtwrite;
	if (brtwrite) {
	/*
	* We are freeing a block that we cloned in the same
	* transaction group.
	*/
	brt_pending_remove(dmu_objset_spa(db->db_objset),
	&dr->dt.dl.dr_overridden_by, tx);
	}

	dnode_t *dn = dr->dr_dnode;

	dprintf_dbuf(db, "size=%llx\n", (u_longlong_t)db->db.db_size);

	ASSERT(db->db.db_size != 0);

	dsl_pool_undirty_space(dmu_objset_pool(dn->dn_objset),
	dr->dr_accounted, txg);

	list_remove(&db->db_dirty_records, dr);

	/*
	* Note that there are three places in dbuf_dirty()
	* where this dirty record may be put on a list.
	* Make sure to do a list_remove corresponding to
	* every one of those list_insert calls.
	*/
	if (dr->dr_parent) {
	mutex_enter(&dr->dr_parent->dt.di.dr_mtx);
	list_remove(&dr->dr_parent->dt.di.dr_children, dr);
	mutex_exit(&dr->dr_parent->dt.di.dr_mtx);
	} else if (db->db_blkid == DMU_SPILL_BLKID \|\|
	db->db_level + 1 == dn->dn_nlevels) {
	ASSERT(db->db_blkptr == NULL \|\| db->db_parent == dn->dn_dbuf);
	mutex_enter(&dn->dn_mtx);
	list_remove(&dn->dn_dirty_records[txg & TXG_MASK], dr);
	mutex_exit(&dn->dn_mtx);
	}

	if (db->db_state != DB_NOFILL && !brtwrite) {
	dbuf_unoverride(dr);

	ASSERT(db->db_buf != NULL);
	ASSERT(dr->dt.dl.dr_data != NULL);
	if (dr->dt.dl.dr_data != db->db_buf)
	arc_buf_destroy(dr->dt.dl.dr_data, db);
	}

	kmem_free(dr, sizeof (dbuf_dirty_record_t));

	ASSERT(db->db_dirtycnt > 0);
	db->db_dirtycnt -= 1;

	if (zfs_refcount_remove(&db->db_holds, (void *)(uintptr_t)txg) == 0) {
	ASSERT(db->db_state == DB_NOFILL \|\| brtwrite \|\|
	arc_released(db->db_buf));
	dbuf_destroy(db);
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	static void
	dmu_buf_will_dirty_impl(dmu_buf_t db_fake, int flags, dmu_tx_t tx)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;
	boolean_t undirty = B_FALSE;

	ASSERT(tx->tx_txg != 0);
	ASSERT(!zfs_refcount_is_zero(&db->db_holds));

	/*
	* Quick check for dirtiness to improve performance for some workloads
	* (e.g. file deletion with indirect blocks cached).
	*/
	mutex_enter(&db->db_mtx);
	if (db->db_state == DB_CACHED \|\| db->db_state == DB_NOFILL) {
	/*
	* It's possible that the dbuf is already dirty but not cached,
	* because there are some calls to dbuf_dirty() that don't
	* go through dmu_buf_will_dirty().
	*/
	dbuf_dirty_record_t *dr = dbuf_find_dirty_eq(db, tx->tx_txg);
	if (dr != NULL) {
	if (db->db_level == 0 &&
	dr->dt.dl.dr_brtwrite) {
	/*
	* Block cloning: If we are dirtying a cloned
	* level 0 block, we cannot simply redirty it,
	* because this dr has no associated data.
	* We will go through a full undirtying below,
	* before dirtying it again.
	*/
	undirty = B_TRUE;
	} else {
	/* This dbuf is already dirty and cached. */
	dbuf_redirty(dr);
	mutex_exit(&db->db_mtx);
	return;
	}
	}
	}
	mutex_exit(&db->db_mtx);

	DB_DNODE_ENTER(db);
	if (RW_WRITE_HELD(&DB_DNODE(db)->dn_struct_rwlock))
	flags \|= DB_RF_HAVESTRUCT;
	DB_DNODE_EXIT(db);

	/*
	* Block cloning: Do the dbuf_read() before undirtying the dbuf, as we
	* want to make sure dbuf_read() will read the pending cloned block and
	* not the uderlying block that is being replaced. dbuf_undirty() will
	* do dbuf_unoverride(), so we will end up with cloned block content,
	* without overridden BP.
	*/
	(void) dbuf_read(db, NULL, flags);
	if (undirty) {
	mutex_enter(&db->db_mtx);
	VERIFY(!dbuf_undirty(db, tx));
	mutex_exit(&db->db_mtx);
	}
	(void) dbuf_dirty(db, tx);
	}

	void
	dmu_buf_will_dirty(dmu_buf_t db_fake, dmu_tx_t tx)
	{
	dmu_buf_will_dirty_impl(db_fake,
	DB_RF_MUST_SUCCEED \| DB_RF_NOPREFETCH, tx);
	}

	boolean_t
	dmu_buf_is_dirty(dmu_buf_t db_fake, dmu_tx_t tx)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;
	dbuf_dirty_record_t *dr;

	mutex_enter(&db->db_mtx);
	dr = dbuf_find_dirty_eq(db, tx->tx_txg);
	mutex_exit(&db->db_mtx);
	return (dr != NULL);
	}

	void
	dmu_buf_will_clone(dmu_buf_t db_fake, dmu_tx_t tx)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;

	/*
	* Block cloning: We are going to clone into this block, so undirty
	* modifications done to this block so far in this txg. This includes
	* writes and clones into this block.
	*/
	mutex_enter(&db->db_mtx);
	DBUF_VERIFY(db);
	VERIFY(!dbuf_undirty(db, tx));
	ASSERT0P(dbuf_find_dirty_eq(db, tx->tx_txg));
	if (db->db_buf != NULL) {
	arc_buf_destroy(db->db_buf, db);
	db->db_buf = NULL;
	dbuf_clear_data(db);
	}

	db->db_state = DB_NOFILL;
	DTRACE_SET_STATE(db, "allocating NOFILL buffer for clone");

	DBUF_VERIFY(db);
	mutex_exit(&db->db_mtx);

	dbuf_noread(db);
	(void) dbuf_dirty(db, tx);
	}

	void
	dmu_buf_will_not_fill(dmu_buf_t db_fake, dmu_tx_t tx)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;

	mutex_enter(&db->db_mtx);
	db->db_state = DB_NOFILL;
	DTRACE_SET_STATE(db, "allocating NOFILL buffer");
	mutex_exit(&db->db_mtx);

	dbuf_noread(db);
	(void) dbuf_dirty(db, tx);
	}

	void
	dmu_buf_will_fill(dmu_buf_t db_fake, dmu_tx_t tx, boolean_t canfail)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;

	ASSERT(db->db_blkid != DMU_BONUS_BLKID);
	ASSERT(tx->tx_txg != 0);
	ASSERT(db->db_level == 0);
	ASSERT(!zfs_refcount_is_zero(&db->db_holds));

	ASSERT(db->db.db_object != DMU_META_DNODE_OBJECT \|\|
	dmu_tx_private_ok(tx));

	mutex_enter(&db->db_mtx);
	if (db->db_state == DB_NOFILL) {
	/*
	* Block cloning: We will be completely overwriting a block
	* cloned in this transaction group, so let's undirty the
	* pending clone and mark the block as uncached. This will be
	* as if the clone was never done. But if the fill can fail
	* we should have a way to return back to the cloned data.
	*/
	if (canfail && dbuf_find_dirty_eq(db, tx->tx_txg) != NULL) {
	mutex_exit(&db->db_mtx);
	dmu_buf_will_dirty(db_fake, tx);
	return;
	}
	VERIFY(!dbuf_undirty(db, tx));
	db->db_state = DB_UNCACHED;
	}
	mutex_exit(&db->db_mtx);

	dbuf_noread(db);
	(void) dbuf_dirty(db, tx);
	}

	/*
	* This function is effectively the same as dmu_buf_will_dirty(), but
	* indicates the caller expects raw encrypted data in the db, and provides
	* the crypt params (byteorder, salt, iv, mac) which should be stored in the
	* blkptr_t when this dbuf is written. This is only used for blocks of
	* dnodes, during raw receive.
	*/
	void
	dmu_buf_set_crypt_params(dmu_buf_t *db_fake, boolean_t byteorder,
	const uint8_t salt, const uint8_t iv, const uint8_t mac, dmu_tx_t tx)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;
	dbuf_dirty_record_t *dr;

	/*
	* dr_has_raw_params is only processed for blocks of dnodes
	* (see dbuf_sync_dnode_leaf_crypt()).
	*/
	ASSERT3U(db->db.db_object, ==, DMU_META_DNODE_OBJECT);
	ASSERT3U(db->db_level, ==, 0);
	ASSERT(db->db_objset->os_raw_receive);

	dmu_buf_will_dirty_impl(db_fake,
	DB_RF_MUST_SUCCEED \| DB_RF_NOPREFETCH \| DB_RF_NO_DECRYPT, tx);

	dr = dbuf_find_dirty_eq(db, tx->tx_txg);

	ASSERT3P(dr, !=, NULL);

	dr->dt.dl.dr_has_raw_params = B_TRUE;
	dr->dt.dl.dr_byteorder = byteorder;
	memcpy(dr->dt.dl.dr_salt, salt, ZIO_DATA_SALT_LEN);
	memcpy(dr->dt.dl.dr_iv, iv, ZIO_DATA_IV_LEN);
	memcpy(dr->dt.dl.dr_mac, mac, ZIO_DATA_MAC_LEN);
	}

	static void
	dbuf_override_impl(dmu_buf_impl_t db, const blkptr_t bp, dmu_tx_t *tx)
	{
	struct dirty_leaf *dl;
	dbuf_dirty_record_t *dr;

	dr = list_head(&db->db_dirty_records);
	ASSERT3P(dr, !=, NULL);
	ASSERT3U(dr->dr_txg, ==, tx->tx_txg);
	dl = &dr->dt.dl;
	dl->dr_overridden_by = *bp;
	dl->dr_override_state = DR_OVERRIDDEN;
	BP_SET_LOGICAL_BIRTH(&dl->dr_overridden_by, dr->dr_txg);
	}

	boolean_t
	dmu_buf_fill_done(dmu_buf_t dbuf, dmu_tx_t tx, boolean_t failed)
	{
	(void) tx;
	dmu_buf_impl_t db = (dmu_buf_impl_t )dbuf;
	mutex_enter(&db->db_mtx);
	DBUF_VERIFY(db);

	if (db->db_state == DB_FILL) {
	if (db->db_level == 0 && db->db_freed_in_flight) {
	ASSERT(db->db_blkid != DMU_BONUS_BLKID);
	/* we were freed while filling */
	/* XXX dbuf_undirty? */
	memset(db->db.db_data, 0, db->db.db_size);
	db->db_freed_in_flight = FALSE;
	db->db_state = DB_CACHED;
	DTRACE_SET_STATE(db,
	"fill done handling freed in flight");
	failed = B_FALSE;
	} else if (failed) {
	VERIFY(!dbuf_undirty(db, tx));
	+ arc_buf_destroy(db->db_buf, db);
	db->db_buf = NULL;
	dbuf_clear_data(db);
	DTRACE_SET_STATE(db, "fill failed");
	} else {
	db->db_state = DB_CACHED;
	DTRACE_SET_STATE(db, "fill done");
	}
	cv_broadcast(&db->db_changed);
	} else {
	db->db_state = DB_CACHED;
	failed = B_FALSE;
	}
	mutex_exit(&db->db_mtx);
	return (failed);
	}

	void
	dmu_buf_write_embedded(dmu_buf_t dbuf, void data,
	bp_embedded_type_t etype, enum zio_compress comp,
	int uncompressed_size, int compressed_size, int byteorder,
	dmu_tx_t *tx)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )dbuf;
	struct dirty_leaf *dl;
	dmu_object_type_t type;
	dbuf_dirty_record_t *dr;

	if (etype == BP_EMBEDDED_TYPE_DATA) {
	ASSERT(spa_feature_is_active(dmu_objset_spa(db->db_objset),
	SPA_FEATURE_EMBEDDED_DATA));
	}

	DB_DNODE_ENTER(db);
	type = DB_DNODE(db)->dn_type;
	DB_DNODE_EXIT(db);

	ASSERT0(db->db_level);
	ASSERT(db->db_blkid != DMU_BONUS_BLKID);

	dmu_buf_will_not_fill(dbuf, tx);

	dr = list_head(&db->db_dirty_records);
	ASSERT3P(dr, !=, NULL);
	ASSERT3U(dr->dr_txg, ==, tx->tx_txg);
	dl = &dr->dt.dl;
	encode_embedded_bp_compressed(&dl->dr_overridden_by,
	data, comp, uncompressed_size, compressed_size);
	BPE_SET_ETYPE(&dl->dr_overridden_by, etype);
	BP_SET_TYPE(&dl->dr_overridden_by, type);
	BP_SET_LEVEL(&dl->dr_overridden_by, 0);
	BP_SET_BYTEORDER(&dl->dr_overridden_by, byteorder);

	dl->dr_override_state = DR_OVERRIDDEN;
	BP_SET_LOGICAL_BIRTH(&dl->dr_overridden_by, dr->dr_txg);
	}

	void
	dmu_buf_redact(dmu_buf_t dbuf, dmu_tx_t tx)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )dbuf;
	dmu_object_type_t type;
	ASSERT(dsl_dataset_feature_is_active(db->db_objset->os_dsl_dataset,
	SPA_FEATURE_REDACTED_DATASETS));

	DB_DNODE_ENTER(db);
	type = DB_DNODE(db)->dn_type;
	DB_DNODE_EXIT(db);

	ASSERT0(db->db_level);
	dmu_buf_will_not_fill(dbuf, tx);

	blkptr_t bp = { { { {0} } } };
	BP_SET_TYPE(&bp, type);
	BP_SET_LEVEL(&bp, 0);
	BP_SET_BIRTH(&bp, tx->tx_txg, 0);
	BP_SET_REDACTED(&bp);
	BPE_SET_LSIZE(&bp, dbuf->db_size);

	dbuf_override_impl(db, &bp, tx);
	}

	/*
	* Directly assign a provided arc buf to a given dbuf if it's not referenced
	* by anybody except our caller. Otherwise copy arcbuf's contents to dbuf.
	*/
	void
	dbuf_assign_arcbuf(dmu_buf_impl_t db, arc_buf_t buf, dmu_tx_t *tx)
	{
	ASSERT(!zfs_refcount_is_zero(&db->db_holds));
	ASSERT(db->db_blkid != DMU_BONUS_BLKID);
	ASSERT(db->db_level == 0);
	ASSERT3U(dbuf_is_metadata(db), ==, arc_is_metadata(buf));
	ASSERT(buf != NULL);
	ASSERT3U(arc_buf_lsize(buf), ==, db->db.db_size);
	ASSERT(tx->tx_txg != 0);

	arc_return_buf(buf, db);
	ASSERT(arc_released(buf));

	mutex_enter(&db->db_mtx);

	while (db->db_state == DB_READ \|\| db->db_state == DB_FILL)
	cv_wait(&db->db_changed, &db->db_mtx);

	ASSERT(db->db_state == DB_CACHED \|\| db->db_state == DB_UNCACHED \|\|
	db->db_state == DB_NOFILL);

	if (db->db_state == DB_CACHED &&
	zfs_refcount_count(&db->db_holds) - 1 > db->db_dirtycnt) {
	/*
	* In practice, we will never have a case where we have an
	* encrypted arc buffer while additional holds exist on the
	* dbuf. We don't handle this here so we simply assert that
	* fact instead.
	*/
	ASSERT(!arc_is_encrypted(buf));
	mutex_exit(&db->db_mtx);
	(void) dbuf_dirty(db, tx);
	memcpy(db->db.db_data, buf->b_data, db->db.db_size);
	arc_buf_destroy(buf, db);
	return;
	}

	if (db->db_state == DB_CACHED) {
	dbuf_dirty_record_t *dr = list_head(&db->db_dirty_records);

	ASSERT(db->db_buf != NULL);
	if (dr != NULL && dr->dr_txg == tx->tx_txg) {
	ASSERT(dr->dt.dl.dr_data == db->db_buf);

	if (!arc_released(db->db_buf)) {
	ASSERT(dr->dt.dl.dr_override_state ==
	DR_OVERRIDDEN);
	arc_release(db->db_buf, db);
	}
	dr->dt.dl.dr_data = buf;
	arc_buf_destroy(db->db_buf, db);
	} else if (dr == NULL \|\| dr->dt.dl.dr_data != db->db_buf) {
	arc_release(db->db_buf, db);
	arc_buf_destroy(db->db_buf, db);
	}
	db->db_buf = NULL;
	} else if (db->db_state == DB_NOFILL) {
	/*
	* We will be completely replacing the cloned block. In case
	* it was cloned in this transaction group, let's undirty the
	* pending clone and mark the block as uncached. This will be
	* as if the clone was never done.
	*/
	VERIFY(!dbuf_undirty(db, tx));
	db->db_state = DB_UNCACHED;
	}
	ASSERT(db->db_buf == NULL);
	dbuf_set_data(db, buf);
	db->db_state = DB_FILL;
	DTRACE_SET_STATE(db, "filling assigned arcbuf");
	mutex_exit(&db->db_mtx);
	(void) dbuf_dirty(db, tx);
	dmu_buf_fill_done(&db->db, tx, B_FALSE);
	}

	void
	dbuf_destroy(dmu_buf_impl_t *db)
	{
	dnode_t *dn;
	dmu_buf_impl_t *parent = db->db_parent;
	dmu_buf_impl_t *dndb;

	ASSERT(MUTEX_HELD(&db->db_mtx));
	ASSERT(zfs_refcount_is_zero(&db->db_holds));

	if (db->db_buf != NULL) {
	arc_buf_destroy(db->db_buf, db);
	db->db_buf = NULL;
	}

	if (db->db_blkid == DMU_BONUS_BLKID) {
	int slots = DB_DNODE(db)->dn_num_slots;
	int bonuslen = DN_SLOTS_TO_BONUSLEN(slots);
	if (db->db.db_data != NULL) {
	kmem_free(db->db.db_data, bonuslen);
	arc_space_return(bonuslen, ARC_SPACE_BONUS);
	db->db_state = DB_UNCACHED;
	DTRACE_SET_STATE(db, "buffer cleared");
	}
	}

	dbuf_clear_data(db);

	if (multilist_link_active(&db->db_cache_link)) {
	ASSERT(db->db_caching_status == DB_DBUF_CACHE \|\|
	db->db_caching_status == DB_DBUF_METADATA_CACHE);

	multilist_remove(&dbuf_caches[db->db_caching_status].cache, db);

	ASSERT0(dmu_buf_user_size(&db->db));
	(void) zfs_refcount_remove_many(
	&dbuf_caches[db->db_caching_status].size,
	db->db.db_size, db);

	if (db->db_caching_status == DB_DBUF_METADATA_CACHE) {
	DBUF_STAT_BUMPDOWN(metadata_cache_count);
	} else {
	DBUF_STAT_BUMPDOWN(cache_levels[db->db_level]);
	DBUF_STAT_BUMPDOWN(cache_count);
	DBUF_STAT_DECR(cache_levels_bytes[db->db_level],
	db->db.db_size);
	}
	db->db_caching_status = DB_NO_CACHE;
	}

	ASSERT(db->db_state == DB_UNCACHED \|\| db->db_state == DB_NOFILL);
	ASSERT(db->db_data_pending == NULL);
	ASSERT(list_is_empty(&db->db_dirty_records));

	db->db_state = DB_EVICTING;
	DTRACE_SET_STATE(db, "buffer eviction started");
	db->db_blkptr = NULL;

	/*
	* Now that db_state is DB_EVICTING, nobody else can find this via
	* the hash table. We can now drop db_mtx, which allows us to
	* acquire the dn_dbufs_mtx.
	*/
	mutex_exit(&db->db_mtx);

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	dndb = dn->dn_dbuf;
	if (db->db_blkid != DMU_BONUS_BLKID) {
	boolean_t needlock = !MUTEX_HELD(&dn->dn_dbufs_mtx);
	if (needlock)
	mutex_enter_nested(&dn->dn_dbufs_mtx,
	NESTED_SINGLE);
	avl_remove(&dn->dn_dbufs, db);
	membar_producer();
	DB_DNODE_EXIT(db);
	if (needlock)
	mutex_exit(&dn->dn_dbufs_mtx);
	/*
	* Decrementing the dbuf count means that the hold corresponding
	* to the removed dbuf is no longer discounted in dnode_move(),
	* so the dnode cannot be moved until after we release the hold.
	* The membar_producer() ensures visibility of the decremented
	* value in dnode_move(), since DB_DNODE_EXIT doesn't actually
	* release any lock.
	*/
	mutex_enter(&dn->dn_mtx);
	dnode_rele_and_unlock(dn, db, B_TRUE);
	db->db_dnode_handle = NULL;

	dbuf_hash_remove(db);
	} else {
	DB_DNODE_EXIT(db);
	}

	ASSERT(zfs_refcount_is_zero(&db->db_holds));

	db->db_parent = NULL;

	ASSERT(db->db_buf == NULL);
	ASSERT(db->db.db_data == NULL);
	ASSERT(db->db_hash_next == NULL);
	ASSERT(db->db_blkptr == NULL);
	ASSERT(db->db_data_pending == NULL);
	ASSERT3U(db->db_caching_status, ==, DB_NO_CACHE);
	ASSERT(!multilist_link_active(&db->db_cache_link));

	/*
	* If this dbuf is referenced from an indirect dbuf,
	* decrement the ref count on the indirect dbuf.
	*/
	if (parent && parent != dndb) {
	mutex_enter(&parent->db_mtx);
	dbuf_rele_and_unlock(parent, db, B_TRUE);
	}

	kmem_cache_free(dbuf_kmem_cache, db);
	arc_space_return(sizeof (dmu_buf_impl_t), ARC_SPACE_DBUF);
	}

	/*
	* Note: While bpp will always be updated if the function returns success,
	* parentp will not be updated if the dnode does not have dn_dbuf filled in;
	* this happens when the dnode is the meta-dnode, or {user\|group\|project}used
	* object.
	*/
	__attribute__((always_inline))
	static inline int
	dbuf_findbp(dnode_t *dn, int level, uint64_t blkid, int fail_sparse,
	dmu_buf_impl_t parentp, blkptr_t bpp)
	{
	*parentp = NULL;
	*bpp = NULL;

	ASSERT(blkid != DMU_BONUS_BLKID);

	if (blkid == DMU_SPILL_BLKID) {
	mutex_enter(&dn->dn_mtx);
	if (dn->dn_have_spill &&
	(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR))
	*bpp = DN_SPILL_BLKPTR(dn->dn_phys);
	else
	*bpp = NULL;
	dbuf_add_ref(dn->dn_dbuf, NULL);
	*parentp = dn->dn_dbuf;
	mutex_exit(&dn->dn_mtx);
	return (0);
	}

	int nlevels =
	(dn->dn_phys->dn_nlevels == 0) ? 1 : dn->dn_phys->dn_nlevels;
	int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;

	ASSERT3U(level * epbs, <, 64);
	ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock));
	/*
	* This assertion shouldn't trip as long as the max indirect block size
	* is less than 1M. The reason for this is that up to that point,
	* the number of levels required to address an entire object with blocks
	* of size SPA_MINBLOCKSIZE satisfies nlevels * epbs + 1 <= 64. In
	* other words, if N * epbs + 1 > 64, then if (N-1) * epbs + 1 > 55
	* (i.e. we can address the entire object), objects will all use at most
	* N-1 levels and the assertion won't overflow. However, once epbs is
	* 13, 4 * 13 + 1 = 53, but 5 * 13 + 1 = 66. Then, 4 levels will not be
	* enough to address an entire object, so objects will have 5 levels,
	* but then this assertion will overflow.
	*
	* All this is to say that if we ever increase DN_MAX_INDBLKSHIFT, we
	* need to redo this logic to handle overflows.
	*/
	ASSERT(level >= nlevels \|\|
	((nlevels - level - 1) * epbs) +
	highbit64(dn->dn_phys->dn_nblkptr) <= 64);
	if (level >= nlevels \|\|
	blkid >= ((uint64_t)dn->dn_phys->dn_nblkptr <<
	((nlevels - level - 1) * epbs)) \|\|
	(fail_sparse &&
	blkid > (dn->dn_phys->dn_maxblkid >> (level * epbs)))) {
	/* the buffer has no parent yet */
	return (SET_ERROR(ENOENT));
	} else if (level < nlevels-1) {
	/* this block is referenced from an indirect block */
	int err;

	err = dbuf_hold_impl(dn, level + 1,
	blkid >> epbs, fail_sparse, FALSE, NULL, parentp);

	if (err)
	return (err);
	err = dbuf_read(*parentp, NULL,
	(DB_RF_HAVESTRUCT \| DB_RF_NOPREFETCH \| DB_RF_CANFAIL));
	if (err) {
	dbuf_rele(*parentp, NULL);
	*parentp = NULL;
	return (err);
	}
	rw_enter(&(*parentp)->db_rwlock, RW_READER);
	bpp = ((blkptr_t )(*parentp)->db.db_data) +
	(blkid & ((1ULL << epbs) - 1));
	if (blkid > (dn->dn_phys->dn_maxblkid >> (level * epbs)))
	ASSERT(BP_IS_HOLE(*bpp));
	rw_exit(&(*parentp)->db_rwlock);
	return (0);
	} else {
	/* the block is referenced from the dnode */
	ASSERT3U(level, ==, nlevels-1);
	ASSERT(dn->dn_phys->dn_nblkptr == 0 \|\|
	blkid < dn->dn_phys->dn_nblkptr);
	if (dn->dn_dbuf) {
	dbuf_add_ref(dn->dn_dbuf, NULL);
	*parentp = dn->dn_dbuf;
	}
	*bpp = &dn->dn_phys->dn_blkptr[blkid];
	return (0);
	}
	}

	static dmu_buf_impl_t *
	dbuf_create(dnode_t *dn, uint8_t level, uint64_t blkid,
	dmu_buf_impl_t parent, blkptr_t blkptr, uint64_t hash)
	{
	objset_t *os = dn->dn_objset;
	dmu_buf_impl_t db, odb;

	ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock));
	ASSERT(dn->dn_type != DMU_OT_NONE);

	db = kmem_cache_alloc(dbuf_kmem_cache, KM_SLEEP);

	list_create(&db->db_dirty_records, sizeof (dbuf_dirty_record_t),
	offsetof(dbuf_dirty_record_t, dr_dbuf_node));

	db->db_objset = os;
	db->db.db_object = dn->dn_object;
	db->db_level = level;
	db->db_blkid = blkid;
	db->db_dirtycnt = 0;
	db->db_dnode_handle = dn->dn_handle;
	db->db_parent = parent;
	db->db_blkptr = blkptr;
	db->db_hash = hash;

	db->db_user = NULL;
	db->db_user_immediate_evict = FALSE;
	db->db_freed_in_flight = FALSE;
	db->db_pending_evict = FALSE;

	if (blkid == DMU_BONUS_BLKID) {
	ASSERT3P(parent, ==, dn->dn_dbuf);
	db->db.db_size = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots) -
	(dn->dn_nblkptr-1) * sizeof (blkptr_t);
	ASSERT3U(db->db.db_size, >=, dn->dn_bonuslen);
	db->db.db_offset = DMU_BONUS_BLKID;
	db->db_state = DB_UNCACHED;
	DTRACE_SET_STATE(db, "bonus buffer created");
	db->db_caching_status = DB_NO_CACHE;
	/* the bonus dbuf is not placed in the hash table */
	arc_space_consume(sizeof (dmu_buf_impl_t), ARC_SPACE_DBUF);
	return (db);
	} else if (blkid == DMU_SPILL_BLKID) {
	db->db.db_size = (blkptr != NULL) ?
	BP_GET_LSIZE(blkptr) : SPA_MINBLOCKSIZE;
	db->db.db_offset = 0;
	} else {
	int blocksize =
	db->db_level ? 1 << dn->dn_indblkshift : dn->dn_datablksz;
	db->db.db_size = blocksize;
	db->db.db_offset = db->db_blkid * blocksize;
	}

	/*
	* Hold the dn_dbufs_mtx while we get the new dbuf
	* in the hash table and added to the dbufs list.
	* This prevents a possible deadlock with someone
	* trying to look up this dbuf before it's added to the
	* dn_dbufs list.
	*/
	mutex_enter(&dn->dn_dbufs_mtx);
	db->db_state = DB_EVICTING; /* not worth logging this state change */
	if ((odb = dbuf_hash_insert(db)) != NULL) {
	/* someone else inserted it first */
	mutex_exit(&dn->dn_dbufs_mtx);
	kmem_cache_free(dbuf_kmem_cache, db);
	DBUF_STAT_BUMP(hash_insert_race);
	return (odb);
	}
	avl_add(&dn->dn_dbufs, db);

	db->db_state = DB_UNCACHED;
	DTRACE_SET_STATE(db, "regular buffer created");
	db->db_caching_status = DB_NO_CACHE;
	mutex_exit(&dn->dn_dbufs_mtx);
	arc_space_consume(sizeof (dmu_buf_impl_t), ARC_SPACE_DBUF);

	if (parent && parent != dn->dn_dbuf)
	dbuf_add_ref(parent, db);

	ASSERT(dn->dn_object == DMU_META_DNODE_OBJECT \|\|
	zfs_refcount_count(&dn->dn_holds) > 0);
	(void) zfs_refcount_add(&dn->dn_holds, db);

	dprintf_dbuf(db, "db=%p\n", db);

	return (db);
	}

	/*
	* This function returns a block pointer and information about the object,
	* given a dnode and a block. This is a publicly accessible version of
	* dbuf_findbp that only returns some information, rather than the
	* dbuf. Note that the dnode passed in must be held, and the dn_struct_rwlock
	* should be locked as (at least) a reader.
	*/
	int
	dbuf_dnode_findbp(dnode_t *dn, uint64_t level, uint64_t blkid,
	blkptr_t bp, uint16_t datablkszsec, uint8_t *indblkshift)
	{
	dmu_buf_impl_t *dbp = NULL;
	blkptr_t *bp2;
	int err = 0;
	ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock));

	err = dbuf_findbp(dn, level, blkid, B_FALSE, &dbp, &bp2);
	if (err == 0) {
	ASSERT3P(bp2, !=, NULL);
	bp = bp2;
	if (dbp != NULL)
	dbuf_rele(dbp, NULL);
	if (datablkszsec != NULL)
	*datablkszsec = dn->dn_phys->dn_datablkszsec;
	if (indblkshift != NULL)
	*indblkshift = dn->dn_phys->dn_indblkshift;
	}

	return (err);
	}

	typedef struct dbuf_prefetch_arg {
	spa_t dpa_spa; / The spa to issue the prefetch in. */
	zbookmark_phys_t dpa_zb; /* The target block to prefetch. */
	int dpa_epbs; /* Entries (blkptr_t's) Per Block Shift. */
	int dpa_curlevel; /* The current level that we're reading */
	dnode_t dpa_dnode; / The dnode associated with the prefetch */
	zio_priority_t dpa_prio; /* The priority I/Os should be issued at. */
	zio_t dpa_zio; / The parent zio_t for all prefetches. */
	arc_flags_t dpa_aflags; /* Flags to pass to the final prefetch. */
	dbuf_prefetch_fn dpa_cb; /* prefetch completion callback */
	void dpa_arg; / prefetch completion arg */
	} dbuf_prefetch_arg_t;

	static void
	dbuf_prefetch_fini(dbuf_prefetch_arg_t *dpa, boolean_t io_done)
	{
	if (dpa->dpa_cb != NULL) {
	dpa->dpa_cb(dpa->dpa_arg, dpa->dpa_zb.zb_level,
	dpa->dpa_zb.zb_blkid, io_done);
	}
	kmem_free(dpa, sizeof (*dpa));
	}

	static void
	dbuf_issue_final_prefetch_done(zio_t zio, const zbookmark_phys_t zb,
	const blkptr_t iobp, arc_buf_t abuf, void *private)
	{
	(void) zio, (void) zb, (void) iobp;
	dbuf_prefetch_arg_t *dpa = private;

	if (abuf != NULL)
	arc_buf_destroy(abuf, private);

	dbuf_prefetch_fini(dpa, B_TRUE);
	}

	/*
	* Actually issue the prefetch read for the block given.
	*/
	static void
	dbuf_issue_final_prefetch(dbuf_prefetch_arg_t dpa, blkptr_t bp)
	{
	ASSERT(!BP_IS_REDACTED(bp) \|\|
	dsl_dataset_feature_is_active(
	dpa->dpa_dnode->dn_objset->os_dsl_dataset,
	SPA_FEATURE_REDACTED_DATASETS));

	if (BP_IS_HOLE(bp) \|\| BP_IS_EMBEDDED(bp) \|\| BP_IS_REDACTED(bp))
	return (dbuf_prefetch_fini(dpa, B_FALSE));

	int zio_flags = ZIO_FLAG_CANFAIL \| ZIO_FLAG_SPECULATIVE;
	arc_flags_t aflags =
	dpa->dpa_aflags \| ARC_FLAG_NOWAIT \| ARC_FLAG_PREFETCH \|
	ARC_FLAG_NO_BUF;

	/* dnodes are always read as raw and then converted later */
	if (BP_GET_TYPE(bp) == DMU_OT_DNODE && BP_IS_PROTECTED(bp) &&
	dpa->dpa_curlevel == 0)
	zio_flags \|= ZIO_FLAG_RAW;

	ASSERT3U(dpa->dpa_curlevel, ==, BP_GET_LEVEL(bp));
	ASSERT3U(dpa->dpa_curlevel, ==, dpa->dpa_zb.zb_level);
	ASSERT(dpa->dpa_zio != NULL);
	(void) arc_read(dpa->dpa_zio, dpa->dpa_spa, bp,
	dbuf_issue_final_prefetch_done, dpa,
	dpa->dpa_prio, zio_flags, &aflags, &dpa->dpa_zb);
	}

	/*
	* Called when an indirect block above our prefetch target is read in. This
	* will either read in the next indirect block down the tree or issue the actual
	* prefetch if the next block down is our target.
	*/
	static void
	dbuf_prefetch_indirect_done(zio_t zio, const zbookmark_phys_t zb,
	const blkptr_t iobp, arc_buf_t abuf, void *private)
	{
	(void) zb, (void) iobp;
	dbuf_prefetch_arg_t *dpa = private;

	ASSERT3S(dpa->dpa_zb.zb_level, <, dpa->dpa_curlevel);
	ASSERT3S(dpa->dpa_curlevel, >, 0);

	if (abuf == NULL) {
	ASSERT(zio == NULL \|\| zio->io_error != 0);
	dbuf_prefetch_fini(dpa, B_TRUE);
	return;
	}
	ASSERT(zio == NULL \|\| zio->io_error == 0);

	/*
	* The dpa_dnode is only valid if we are called with a NULL
	* zio. This indicates that the arc_read() returned without
	* first calling zio_read() to issue a physical read. Once
	* a physical read is made the dpa_dnode must be invalidated
	* as the locks guarding it may have been dropped. If the
	* dpa_dnode is still valid, then we want to add it to the dbuf
	* cache. To do so, we must hold the dbuf associated with the block
	* we just prefetched, read its contents so that we associate it
	* with an arc_buf_t, and then release it.
	*/
	if (zio != NULL) {
	ASSERT3S(BP_GET_LEVEL(zio->io_bp), ==, dpa->dpa_curlevel);
	if (zio->io_flags & ZIO_FLAG_RAW_COMPRESS) {
	ASSERT3U(BP_GET_PSIZE(zio->io_bp), ==, zio->io_size);
	} else {
	ASSERT3U(BP_GET_LSIZE(zio->io_bp), ==, zio->io_size);
	}
	ASSERT3P(zio->io_spa, ==, dpa->dpa_spa);

	dpa->dpa_dnode = NULL;
	} else if (dpa->dpa_dnode != NULL) {
	uint64_t curblkid = dpa->dpa_zb.zb_blkid >>
	(dpa->dpa_epbs * (dpa->dpa_curlevel -
	dpa->dpa_zb.zb_level));
	dmu_buf_impl_t *db = dbuf_hold_level(dpa->dpa_dnode,
	dpa->dpa_curlevel, curblkid, FTAG);
	if (db == NULL) {
	arc_buf_destroy(abuf, private);
	dbuf_prefetch_fini(dpa, B_TRUE);
	return;
	}
	(void) dbuf_read(db, NULL,
	DB_RF_MUST_SUCCEED \| DB_RF_NOPREFETCH \| DB_RF_HAVESTRUCT);
	dbuf_rele(db, FTAG);
	}

	dpa->dpa_curlevel--;
	uint64_t nextblkid = dpa->dpa_zb.zb_blkid >>
	(dpa->dpa_epbs * (dpa->dpa_curlevel - dpa->dpa_zb.zb_level));
	blkptr_t bp = ((blkptr_t )abuf->b_data) +
	P2PHASE(nextblkid, 1ULL << dpa->dpa_epbs);

	ASSERT(!BP_IS_REDACTED(bp) \|\| (dpa->dpa_dnode &&
	dsl_dataset_feature_is_active(
	dpa->dpa_dnode->dn_objset->os_dsl_dataset,
	SPA_FEATURE_REDACTED_DATASETS)));
	if (BP_IS_HOLE(bp) \|\| BP_IS_REDACTED(bp)) {
	arc_buf_destroy(abuf, private);
	dbuf_prefetch_fini(dpa, B_TRUE);
	return;
	} else if (dpa->dpa_curlevel == dpa->dpa_zb.zb_level) {
	ASSERT3U(nextblkid, ==, dpa->dpa_zb.zb_blkid);
	dbuf_issue_final_prefetch(dpa, bp);
	} else {
	arc_flags_t iter_aflags = ARC_FLAG_NOWAIT;
	zbookmark_phys_t zb;

	/* flag if L2ARC eligible, l2arc_noprefetch then decides */
	if (dpa->dpa_aflags & ARC_FLAG_L2CACHE)
	iter_aflags \|= ARC_FLAG_L2CACHE;

	ASSERT3U(dpa->dpa_curlevel, ==, BP_GET_LEVEL(bp));

	SET_BOOKMARK(&zb, dpa->dpa_zb.zb_objset,
	dpa->dpa_zb.zb_object, dpa->dpa_curlevel, nextblkid);

	(void) arc_read(dpa->dpa_zio, dpa->dpa_spa,
	bp, dbuf_prefetch_indirect_done, dpa,
	ZIO_PRIORITY_SYNC_READ,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_SPECULATIVE,
	&iter_aflags, &zb);
	}

	arc_buf_destroy(abuf, private);
	}

	/*
	* Issue prefetch reads for the given block on the given level. If the indirect
	* blocks above that block are not in memory, we will read them in
	* asynchronously. As a result, this call never blocks waiting for a read to
	* complete. Note that the prefetch might fail if the dataset is encrypted and
	* the encryption key is unmapped before the IO completes.
	*/
	int
	dbuf_prefetch_impl(dnode_t *dn, int64_t level, uint64_t blkid,
	zio_priority_t prio, arc_flags_t aflags, dbuf_prefetch_fn cb,
	void *arg)
	{
	blkptr_t bp;
	int epbs, nlevels, curlevel;
	uint64_t curblkid;

	ASSERT(blkid != DMU_BONUS_BLKID);
	ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock));

	if (blkid > dn->dn_maxblkid)
	goto no_issue;

	if (level == 0 && dnode_block_freed(dn, blkid))
	goto no_issue;

	/*
	* This dnode hasn't been written to disk yet, so there's nothing to
	* prefetch.
	*/
	nlevels = dn->dn_phys->dn_nlevels;
	if (level >= nlevels \|\| dn->dn_phys->dn_nblkptr == 0)
	goto no_issue;

	epbs = dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT;
	if (dn->dn_phys->dn_maxblkid < blkid << (epbs * level))
	goto no_issue;

	dmu_buf_impl_t *db = dbuf_find(dn->dn_objset, dn->dn_object,
	level, blkid, NULL);
	if (db != NULL) {
	mutex_exit(&db->db_mtx);
	/*
	* This dbuf already exists. It is either CACHED, or
	* (we assume) about to be read or filled.
	*/
	goto no_issue;
	}

	/*
	* Find the closest ancestor (indirect block) of the target block
	* that is present in the cache. In this indirect block, we will
	* find the bp that is at curlevel, curblkid.
	*/
	curlevel = level;
	curblkid = blkid;
	while (curlevel < nlevels - 1) {
	int parent_level = curlevel + 1;
	uint64_t parent_blkid = curblkid >> epbs;
	dmu_buf_impl_t *db;

	if (dbuf_hold_impl(dn, parent_level, parent_blkid,
	FALSE, TRUE, FTAG, &db) == 0) {
	blkptr_t *bpp = db->db_buf->b_data;
	bp = bpp[P2PHASE(curblkid, 1 << epbs)];
	dbuf_rele(db, FTAG);
	break;
	}

	curlevel = parent_level;
	curblkid = parent_blkid;
	}

	if (curlevel == nlevels - 1) {
	/* No cached indirect blocks found. */
	ASSERT3U(curblkid, <, dn->dn_phys->dn_nblkptr);
	bp = dn->dn_phys->dn_blkptr[curblkid];
	}
	ASSERT(!BP_IS_REDACTED(&bp) \|\|
	dsl_dataset_feature_is_active(dn->dn_objset->os_dsl_dataset,
	SPA_FEATURE_REDACTED_DATASETS));
	if (BP_IS_HOLE(&bp) \|\| BP_IS_REDACTED(&bp))
	goto no_issue;

	ASSERT3U(curlevel, ==, BP_GET_LEVEL(&bp));

	zio_t *pio = zio_root(dmu_objset_spa(dn->dn_objset), NULL, NULL,
	ZIO_FLAG_CANFAIL);

	dbuf_prefetch_arg_t dpa = kmem_zalloc(sizeof (dpa), KM_SLEEP);
	dsl_dataset_t *ds = dn->dn_objset->os_dsl_dataset;
	SET_BOOKMARK(&dpa->dpa_zb, ds != NULL ? ds->ds_object : DMU_META_OBJSET,
	dn->dn_object, level, blkid);
	dpa->dpa_curlevel = curlevel;
	dpa->dpa_prio = prio;
	dpa->dpa_aflags = aflags;
	dpa->dpa_spa = dn->dn_objset->os_spa;
	dpa->dpa_dnode = dn;
	dpa->dpa_epbs = epbs;
	dpa->dpa_zio = pio;
	dpa->dpa_cb = cb;
	dpa->dpa_arg = arg;

	if (!DNODE_LEVEL_IS_CACHEABLE(dn, level))
	dpa->dpa_aflags \|= ARC_FLAG_UNCACHED;
	else if (dnode_level_is_l2cacheable(&bp, dn, level))
	dpa->dpa_aflags \|= ARC_FLAG_L2CACHE;

	/*
	* If we have the indirect just above us, no need to do the asynchronous
	* prefetch chain; we'll just run the last step ourselves. If we're at
	* a higher level, though, we want to issue the prefetches for all the
	* indirect blocks asynchronously, so we can go on with whatever we were
	* doing.
	*/
	if (curlevel == level) {
	ASSERT3U(curblkid, ==, blkid);
	dbuf_issue_final_prefetch(dpa, &bp);
	} else {
	arc_flags_t iter_aflags = ARC_FLAG_NOWAIT;
	zbookmark_phys_t zb;

	/* flag if L2ARC eligible, l2arc_noprefetch then decides */
	if (dnode_level_is_l2cacheable(&bp, dn, level))
	iter_aflags \|= ARC_FLAG_L2CACHE;

	SET_BOOKMARK(&zb, ds != NULL ? ds->ds_object : DMU_META_OBJSET,
	dn->dn_object, curlevel, curblkid);
	(void) arc_read(dpa->dpa_zio, dpa->dpa_spa,
	&bp, dbuf_prefetch_indirect_done, dpa,
	ZIO_PRIORITY_SYNC_READ,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_SPECULATIVE,
	&iter_aflags, &zb);
	}
	/*
	* We use pio here instead of dpa_zio since it's possible that
	* dpa may have already been freed.
	*/
	zio_nowait(pio);
	return (1);
	no_issue:
	if (cb != NULL)
	cb(arg, level, blkid, B_FALSE);
	return (0);
	}

	int
	dbuf_prefetch(dnode_t *dn, int64_t level, uint64_t blkid, zio_priority_t prio,
	arc_flags_t aflags)
	{

	return (dbuf_prefetch_impl(dn, level, blkid, prio, aflags, NULL, NULL));
	}

	/*
	* Helper function for dbuf_hold_impl() to copy a buffer. Handles
	* the case of encrypted, compressed and uncompressed buffers by
	* allocating the new buffer, respectively, with arc_alloc_raw_buf(),
	* arc_alloc_compressed_buf() or arc_alloc_buf().*
	*
	* NOTE: Declared noinline to avoid stack bloat in dbuf_hold_impl().
	*/
	noinline static void
	dbuf_hold_copy(dnode_t dn, dmu_buf_impl_t db)
	{
	dbuf_dirty_record_t *dr = db->db_data_pending;
	arc_buf_t *data = dr->dt.dl.dr_data;
	enum zio_compress compress_type = arc_get_compression(data);
	uint8_t complevel = arc_get_complevel(data);

	if (arc_is_encrypted(data)) {
	boolean_t byteorder;
	uint8_t salt[ZIO_DATA_SALT_LEN];
	uint8_t iv[ZIO_DATA_IV_LEN];
	uint8_t mac[ZIO_DATA_MAC_LEN];

	arc_get_raw_params(data, &byteorder, salt, iv, mac);
	dbuf_set_data(db, arc_alloc_raw_buf(dn->dn_objset->os_spa, db,
	dmu_objset_id(dn->dn_objset), byteorder, salt, iv, mac,
	dn->dn_type, arc_buf_size(data), arc_buf_lsize(data),
	compress_type, complevel));
	} else if (compress_type != ZIO_COMPRESS_OFF) {
	dbuf_set_data(db, arc_alloc_compressed_buf(
	dn->dn_objset->os_spa, db, arc_buf_size(data),
	arc_buf_lsize(data), compress_type, complevel));
	} else {
	dbuf_set_data(db, arc_alloc_buf(dn->dn_objset->os_spa, db,
	DBUF_GET_BUFC_TYPE(db), db->db.db_size));
	}

	rw_enter(&db->db_rwlock, RW_WRITER);
	memcpy(db->db.db_data, data->b_data, arc_buf_size(data));
	rw_exit(&db->db_rwlock);
	}

	/*
	* Returns with db_holds incremented, and db_mtx not held.
	* Note: dn_struct_rwlock must be held.
	*/
	int
	dbuf_hold_impl(dnode_t *dn, uint8_t level, uint64_t blkid,
	boolean_t fail_sparse, boolean_t fail_uncached,
	const void tag, dmu_buf_impl_t *dbp)
	{
	dmu_buf_impl_t db, parent = NULL;
	uint64_t hv;

	/* If the pool has been created, verify the tx_sync_lock is not held */
	spa_t *spa = dn->dn_objset->os_spa;
	dsl_pool_t *dp = spa->spa_dsl_pool;
	if (dp != NULL) {
	ASSERT(!MUTEX_HELD(&dp->dp_tx.tx_sync_lock));
	}

	ASSERT(blkid != DMU_BONUS_BLKID);
	ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock));
	ASSERT3U(dn->dn_nlevels, >, level);

	*dbp = NULL;

	/* dbuf_find() returns with db_mtx held */
	db = dbuf_find(dn->dn_objset, dn->dn_object, level, blkid, &hv);

	if (db == NULL) {
	blkptr_t *bp = NULL;
	int err;

	if (fail_uncached)
	return (SET_ERROR(ENOENT));

	ASSERT3P(parent, ==, NULL);
	err = dbuf_findbp(dn, level, blkid, fail_sparse, &parent, &bp);
	if (fail_sparse) {
	if (err == 0 && bp && BP_IS_HOLE(bp))
	err = SET_ERROR(ENOENT);
	if (err) {
	if (parent)
	dbuf_rele(parent, NULL);
	return (err);
	}
	}
	if (err && err != ENOENT)
	return (err);
	db = dbuf_create(dn, level, blkid, parent, bp, hv);
	}

	if (fail_uncached && db->db_state != DB_CACHED) {
	mutex_exit(&db->db_mtx);
	return (SET_ERROR(ENOENT));
	}

	if (db->db_buf != NULL) {
	arc_buf_access(db->db_buf);
	ASSERT3P(db->db.db_data, ==, db->db_buf->b_data);
	}

	ASSERT(db->db_buf == NULL \|\| arc_referenced(db->db_buf));

	/*
	* If this buffer is currently syncing out, and we are
	* still referencing it from db_data, we need to make a copy
	* of it in case we decide we want to dirty it again in this txg.
	*/
	if (db->db_level == 0 && db->db_blkid != DMU_BONUS_BLKID &&
	dn->dn_object != DMU_META_DNODE_OBJECT &&
	db->db_state == DB_CACHED && db->db_data_pending) {
	dbuf_dirty_record_t *dr = db->db_data_pending;
	if (dr->dt.dl.dr_data == db->db_buf) {
	ASSERT3P(db->db_buf, !=, NULL);
	dbuf_hold_copy(dn, db);
	}
	}

	if (multilist_link_active(&db->db_cache_link)) {
	ASSERT(zfs_refcount_is_zero(&db->db_holds));
	ASSERT(db->db_caching_status == DB_DBUF_CACHE \|\|
	db->db_caching_status == DB_DBUF_METADATA_CACHE);

	multilist_remove(&dbuf_caches[db->db_caching_status].cache, db);

	- uint64_t size = db->db.db_size + dmu_buf_user_size(&db->db);
	+ uint64_t size = db->db.db_size;
	+ uint64_t usize = dmu_buf_user_size(&db->db);
	(void) zfs_refcount_remove_many(
	&dbuf_caches[db->db_caching_status].size, size, db);
	+ (void) zfs_refcount_remove_many(
	+ &dbuf_caches[db->db_caching_status].size, usize,
	+ db->db_user);

	if (db->db_caching_status == DB_DBUF_METADATA_CACHE) {
	DBUF_STAT_BUMPDOWN(metadata_cache_count);
	} else {
	DBUF_STAT_BUMPDOWN(cache_levels[db->db_level]);
	DBUF_STAT_BUMPDOWN(cache_count);
	- DBUF_STAT_DECR(cache_levels_bytes[db->db_level], size);
	+ DBUF_STAT_DECR(cache_levels_bytes[db->db_level],
	+ size + usize);
	}
	db->db_caching_status = DB_NO_CACHE;
	}
	(void) zfs_refcount_add(&db->db_holds, tag);
	DBUF_VERIFY(db);
	mutex_exit(&db->db_mtx);

	/* NOTE: we can't rele the parent until after we drop the db_mtx */
	if (parent)
	dbuf_rele(parent, NULL);

	ASSERT3P(DB_DNODE(db), ==, dn);
	ASSERT3U(db->db_blkid, ==, blkid);
	ASSERT3U(db->db_level, ==, level);
	*dbp = db;

	return (0);
	}

	dmu_buf_impl_t *
	dbuf_hold(dnode_t dn, uint64_t blkid, const void tag)
	{
	return (dbuf_hold_level(dn, 0, blkid, tag));
	}

	dmu_buf_impl_t *
	dbuf_hold_level(dnode_t dn, int level, uint64_t blkid, const void tag)
	{
	dmu_buf_impl_t *db;
	int err = dbuf_hold_impl(dn, level, blkid, FALSE, FALSE, tag, &db);
	return (err ? NULL : db);
	}

	void
	dbuf_create_bonus(dnode_t *dn)
	{
	ASSERT(RW_WRITE_HELD(&dn->dn_struct_rwlock));

	ASSERT(dn->dn_bonus == NULL);
	dn->dn_bonus = dbuf_create(dn, 0, DMU_BONUS_BLKID, dn->dn_dbuf, NULL,
	dbuf_hash(dn->dn_objset, dn->dn_object, 0, DMU_BONUS_BLKID));
	}

	int
	dbuf_spill_set_blksz(dmu_buf_t db_fake, uint64_t blksz, dmu_tx_t tx)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;

	if (db->db_blkid != DMU_SPILL_BLKID)
	return (SET_ERROR(ENOTSUP));
	if (blksz == 0)
	blksz = SPA_MINBLOCKSIZE;
	ASSERT3U(blksz, <=, spa_maxblocksize(dmu_objset_spa(db->db_objset)));
	blksz = P2ROUNDUP(blksz, SPA_MINBLOCKSIZE);

	dbuf_new_size(db, blksz, tx);

	return (0);
	}

	void
	dbuf_rm_spill(dnode_t dn, dmu_tx_t tx)
	{
	dbuf_free_range(dn, DMU_SPILL_BLKID, DMU_SPILL_BLKID, tx);
	}

	#pragma weak dmu_buf_add_ref = dbuf_add_ref
	void
	dbuf_add_ref(dmu_buf_impl_t db, const void tag)
	{
	int64_t holds = zfs_refcount_add(&db->db_holds, tag);
	VERIFY3S(holds, >, 1);
	}

	#pragma weak dmu_buf_try_add_ref = dbuf_try_add_ref
	boolean_t
	dbuf_try_add_ref(dmu_buf_t db_fake, objset_t os, uint64_t obj, uint64_t blkid,
	const void *tag)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;
	dmu_buf_impl_t *found_db;
	boolean_t result = B_FALSE;

	if (blkid == DMU_BONUS_BLKID)
	found_db = dbuf_find_bonus(os, obj);
	else
	found_db = dbuf_find(os, obj, 0, blkid, NULL);

	if (found_db != NULL) {
	if (db == found_db && dbuf_refcount(db) > db->db_dirtycnt) {
	(void) zfs_refcount_add(&db->db_holds, tag);
	result = B_TRUE;
	}
	mutex_exit(&found_db->db_mtx);
	}
	return (result);
	}

	/*
	* If you call dbuf_rele() you had better not be referencing the dnode handle
	* unless you have some other direct or indirect hold on the dnode. (An indirect
	* hold is a hold on one of the dnode's dbufs, including the bonus buffer.)
	* Without that, the dbuf_rele() could lead to a dnode_rele() followed by the
	* dnode's parent dbuf evicting its dnode handles.
	*/
	void
	dbuf_rele(dmu_buf_impl_t db, const void tag)
	{
	mutex_enter(&db->db_mtx);
	dbuf_rele_and_unlock(db, tag, B_FALSE);
	}

	void
	dmu_buf_rele(dmu_buf_t db, const void tag)
	{
	dbuf_rele((dmu_buf_impl_t *)db, tag);
	}

	/*
	* dbuf_rele() for an already-locked dbuf. This is necessary to allow
	* db_dirtycnt and db_holds to be updated atomically. The 'evicting'
	* argument should be set if we are already in the dbuf-evicting code
	* path, in which case we don't want to recursively evict. This allows us to
	* avoid deeply nested stacks that would have a call flow similar to this:
	*
	* dbuf_rele()-->dbuf_rele_and_unlock()-->dbuf_evict_notify()
	* ^ \|
	* \| \|
	* +-----dbuf_destroy()<--dbuf_evict_one()<--------+
	*
	*/
	void
	dbuf_rele_and_unlock(dmu_buf_impl_t db, const void tag, boolean_t evicting)
	{
	int64_t holds;
	uint64_t size;

	ASSERT(MUTEX_HELD(&db->db_mtx));
	DBUF_VERIFY(db);

	/*
	* Remove the reference to the dbuf before removing its hold on the
	* dnode so we can guarantee in dnode_move() that a referenced bonus
	* buffer has a corresponding dnode hold.
	*/
	holds = zfs_refcount_remove(&db->db_holds, tag);
	ASSERT(holds >= 0);

	/*
	* We can't freeze indirects if there is a possibility that they
	* may be modified in the current syncing context.
	*/
	if (db->db_buf != NULL &&
	holds == (db->db_level == 0 ? db->db_dirtycnt : 0)) {
	arc_buf_freeze(db->db_buf);
	}

	if (holds == db->db_dirtycnt &&
	db->db_level == 0 && db->db_user_immediate_evict)
	dbuf_evict_user(db);

	if (holds == 0) {
	if (db->db_blkid == DMU_BONUS_BLKID) {
	dnode_t *dn;
	boolean_t evict_dbuf = db->db_pending_evict;

	/*
	* If the dnode moves here, we cannot cross this
	* barrier until the move completes.
	*/
	DB_DNODE_ENTER(db);

	dn = DB_DNODE(db);
	atomic_dec_32(&dn->dn_dbufs_count);

	/*
	* Decrementing the dbuf count means that the bonus
	* buffer's dnode hold is no longer discounted in
	* dnode_move(). The dnode cannot move until after
	* the dnode_rele() below.
	*/
	DB_DNODE_EXIT(db);

	/*
	* Do not reference db after its lock is dropped.
	* Another thread may evict it.
	*/
	mutex_exit(&db->db_mtx);

	if (evict_dbuf)
	dnode_evict_bonus(dn);

	dnode_rele(dn, db);
	} else if (db->db_buf == NULL) {
	/*
	* This is a special case: we never associated this
	* dbuf with any data allocated from the ARC.
	*/
	ASSERT(db->db_state == DB_UNCACHED \|\|
	db->db_state == DB_NOFILL);
	dbuf_destroy(db);
	} else if (arc_released(db->db_buf)) {
	/*
	* This dbuf has anonymous data associated with it.
	*/
	dbuf_destroy(db);
	} else if (!(DBUF_IS_CACHEABLE(db) \|\| db->db_partial_read) \|\|
	db->db_pending_evict) {
	dbuf_destroy(db);
	} else if (!multilist_link_active(&db->db_cache_link)) {
	ASSERT3U(db->db_caching_status, ==, DB_NO_CACHE);

	dbuf_cached_state_t dcs =
	dbuf_include_in_metadata_cache(db) ?
	DB_DBUF_METADATA_CACHE : DB_DBUF_CACHE;
	db->db_caching_status = dcs;

	multilist_insert(&dbuf_caches[dcs].cache, db);
	- uint64_t db_size = db->db.db_size +
	- dmu_buf_user_size(&db->db);
	- size = zfs_refcount_add_many(
	+ uint64_t db_size = db->db.db_size;
	+ uint64_t dbu_size = dmu_buf_user_size(&db->db);
	+ (void) zfs_refcount_add_many(
	&dbuf_caches[dcs].size, db_size, db);
	+ size = zfs_refcount_add_many(
	+ &dbuf_caches[dcs].size, dbu_size, db->db_user);
	uint8_t db_level = db->db_level;
	mutex_exit(&db->db_mtx);

	if (dcs == DB_DBUF_METADATA_CACHE) {
	DBUF_STAT_BUMP(metadata_cache_count);
	DBUF_STAT_MAX(metadata_cache_size_bytes_max,
	size);
	} else {
	DBUF_STAT_BUMP(cache_count);
	DBUF_STAT_MAX(cache_size_bytes_max, size);
	DBUF_STAT_BUMP(cache_levels[db_level]);
	DBUF_STAT_INCR(cache_levels_bytes[db_level],
	- db_size);
	+ db_size + dbu_size);
	}

	if (dcs == DB_DBUF_CACHE && !evicting)
	dbuf_evict_notify(size);
	}
	} else {
	mutex_exit(&db->db_mtx);
	}

	}

	#pragma weak dmu_buf_refcount = dbuf_refcount
	uint64_t
	dbuf_refcount(dmu_buf_impl_t *db)
	{
	return (zfs_refcount_count(&db->db_holds));
	}

	uint64_t
	dmu_buf_user_refcount(dmu_buf_t *db_fake)
	{
	uint64_t holds;
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;

	mutex_enter(&db->db_mtx);
	ASSERT3U(zfs_refcount_count(&db->db_holds), >=, db->db_dirtycnt);
	holds = zfs_refcount_count(&db->db_holds) - db->db_dirtycnt;
	mutex_exit(&db->db_mtx);

	return (holds);
	}

	void *
	dmu_buf_replace_user(dmu_buf_t db_fake, dmu_buf_user_t old_user,
	dmu_buf_user_t *new_user)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;

	mutex_enter(&db->db_mtx);
	dbuf_verify_user(db, DBVU_NOT_EVICTING);
	if (db->db_user == old_user)
	db->db_user = new_user;
	else
	old_user = db->db_user;
	dbuf_verify_user(db, DBVU_NOT_EVICTING);
	mutex_exit(&db->db_mtx);

	return (old_user);
	}

	void *
	dmu_buf_set_user(dmu_buf_t db_fake, dmu_buf_user_t user)
	{
	return (dmu_buf_replace_user(db_fake, NULL, user));
	}

	void *
	dmu_buf_set_user_ie(dmu_buf_t db_fake, dmu_buf_user_t user)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;

	db->db_user_immediate_evict = TRUE;
	return (dmu_buf_set_user(db_fake, user));
	}

	void *
	dmu_buf_remove_user(dmu_buf_t db_fake, dmu_buf_user_t user)
	{
	return (dmu_buf_replace_user(db_fake, user, NULL));
	}

	void *
	dmu_buf_get_user(dmu_buf_t *db_fake)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;

	dbuf_verify_user(db, DBVU_NOT_EVICTING);
	return (db->db_user);
	}

	uint64_t
	dmu_buf_user_size(dmu_buf_t *db_fake)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;
	if (db->db_user == NULL)
	return (0);
	return (atomic_load_64(&db->db_user->dbu_size));
	}

	void
	dmu_buf_add_user_size(dmu_buf_t *db_fake, uint64_t nadd)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;
	ASSERT3U(db->db_caching_status, ==, DB_NO_CACHE);
	ASSERT3P(db->db_user, !=, NULL);
	ASSERT3U(atomic_load_64(&db->db_user->dbu_size), <, UINT64_MAX - nadd);
	atomic_add_64(&db->db_user->dbu_size, nadd);
	}

	void
	dmu_buf_sub_user_size(dmu_buf_t *db_fake, uint64_t nsub)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;
	ASSERT3U(db->db_caching_status, ==, DB_NO_CACHE);
	ASSERT3P(db->db_user, !=, NULL);
	ASSERT3U(atomic_load_64(&db->db_user->dbu_size), >=, nsub);
	atomic_sub_64(&db->db_user->dbu_size, nsub);
	}

	void
	dmu_buf_user_evict_wait(void)
	{
	taskq_wait(dbu_evict_taskq);
	}

	blkptr_t *
	dmu_buf_get_blkptr(dmu_buf_t *db)
	{
	dmu_buf_impl_t dbi = (dmu_buf_impl_t )db;
	return (dbi->db_blkptr);
	}

	objset_t *
	dmu_buf_get_objset(dmu_buf_t *db)
	{
	dmu_buf_impl_t dbi = (dmu_buf_impl_t )db;
	return (dbi->db_objset);
	}

	static void
	dbuf_check_blkptr(dnode_t dn, dmu_buf_impl_t db)
	{
	/* ASSERT(dmu_tx_is_syncing(tx) */
	ASSERT(MUTEX_HELD(&db->db_mtx));

	if (db->db_blkptr != NULL)
	return;

	if (db->db_blkid == DMU_SPILL_BLKID) {
	db->db_blkptr = DN_SPILL_BLKPTR(dn->dn_phys);
	BP_ZERO(db->db_blkptr);
	return;
	}
	if (db->db_level == dn->dn_phys->dn_nlevels-1) {
	/*
	* This buffer was allocated at a time when there was
	* no available blkptrs from the dnode, or it was
	* inappropriate to hook it in (i.e., nlevels mismatch).
	*/
	ASSERT(db->db_blkid < dn->dn_phys->dn_nblkptr);
	ASSERT(db->db_parent == NULL);
	db->db_parent = dn->dn_dbuf;
	db->db_blkptr = &dn->dn_phys->dn_blkptr[db->db_blkid];
	DBUF_VERIFY(db);
	} else {
	dmu_buf_impl_t *parent = db->db_parent;
	int epbs = dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT;

	ASSERT(dn->dn_phys->dn_nlevels > 1);
	if (parent == NULL) {
	mutex_exit(&db->db_mtx);
	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	parent = dbuf_hold_level(dn, db->db_level + 1,
	db->db_blkid >> epbs, db);
	rw_exit(&dn->dn_struct_rwlock);
	mutex_enter(&db->db_mtx);
	db->db_parent = parent;
	}
	db->db_blkptr = (blkptr_t *)parent->db.db_data +
	(db->db_blkid & ((1ULL << epbs) - 1));
	DBUF_VERIFY(db);
	}
	}

	static void
	dbuf_sync_bonus(dbuf_dirty_record_t dr, dmu_tx_t tx)
	{
	dmu_buf_impl_t *db = dr->dr_dbuf;
	void *data = dr->dt.dl.dr_data;

	ASSERT0(db->db_level);
	ASSERT(MUTEX_HELD(&db->db_mtx));
	ASSERT(db->db_blkid == DMU_BONUS_BLKID);
	ASSERT(data != NULL);

	dnode_t *dn = dr->dr_dnode;
	ASSERT3U(DN_MAX_BONUS_LEN(dn->dn_phys), <=,
	DN_SLOTS_TO_BONUSLEN(dn->dn_phys->dn_extra_slots + 1));
	memcpy(DN_BONUS(dn->dn_phys), data, DN_MAX_BONUS_LEN(dn->dn_phys));

	dbuf_sync_leaf_verify_bonus_dnode(dr);

	dbuf_undirty_bonus(dr);
	dbuf_rele_and_unlock(db, (void *)(uintptr_t)tx->tx_txg, B_FALSE);
	}

	/*
	* When syncing out a blocks of dnodes, adjust the block to deal with
	* encryption. Normally, we make sure the block is decrypted before writing
	* it. If we have crypt params, then we are writing a raw (encrypted) block,
	* from a raw receive. In this case, set the ARC buf's crypt params so
	* that the BP will be filled with the correct byteorder, salt, iv, and mac.
	*/
	static void
	dbuf_prepare_encrypted_dnode_leaf(dbuf_dirty_record_t *dr)
	{
	int err;
	dmu_buf_impl_t *db = dr->dr_dbuf;

	ASSERT(MUTEX_HELD(&db->db_mtx));
	ASSERT3U(db->db.db_object, ==, DMU_META_DNODE_OBJECT);
	ASSERT3U(db->db_level, ==, 0);

	if (!db->db_objset->os_raw_receive && arc_is_encrypted(db->db_buf)) {
	zbookmark_phys_t zb;

	/*
	* Unfortunately, there is currently no mechanism for
	* syncing context to handle decryption errors. An error
	* here is only possible if an attacker maliciously
	* changed a dnode block and updated the associated
	* checksums going up the block tree.
	*/
	SET_BOOKMARK(&zb, dmu_objset_id(db->db_objset),
	db->db.db_object, db->db_level, db->db_blkid);
	err = arc_untransform(db->db_buf, db->db_objset->os_spa,
	&zb, B_TRUE);
	if (err)
	panic("Invalid dnode block MAC");
	} else if (dr->dt.dl.dr_has_raw_params) {
	(void) arc_release(dr->dt.dl.dr_data, db);
	arc_convert_to_raw(dr->dt.dl.dr_data,
	dmu_objset_id(db->db_objset),
	dr->dt.dl.dr_byteorder, DMU_OT_DNODE,
	dr->dt.dl.dr_salt, dr->dt.dl.dr_iv, dr->dt.dl.dr_mac);
	}
	}

	/*
	* dbuf_sync_indirect() is called recursively from dbuf_sync_list() so it
	* is critical the we not allow the compiler to inline this function in to
	* dbuf_sync_list() thereby drastically bloating the stack usage.
	*/
	noinline static void
	dbuf_sync_indirect(dbuf_dirty_record_t dr, dmu_tx_t tx)
	{
	dmu_buf_impl_t *db = dr->dr_dbuf;
	dnode_t *dn = dr->dr_dnode;

	ASSERT(dmu_tx_is_syncing(tx));

	dprintf_dbuf_bp(db, db->db_blkptr, "blkptr=%p", db->db_blkptr);

	mutex_enter(&db->db_mtx);

	ASSERT(db->db_level > 0);
	DBUF_VERIFY(db);

	/* Read the block if it hasn't been read yet. */
	if (db->db_buf == NULL) {
	mutex_exit(&db->db_mtx);
	(void) dbuf_read(db, NULL, DB_RF_MUST_SUCCEED);
	mutex_enter(&db->db_mtx);
	}
	ASSERT3U(db->db_state, ==, DB_CACHED);
	ASSERT(db->db_buf != NULL);

	/* Indirect block size must match what the dnode thinks it is. */
	ASSERT3U(db->db.db_size, ==, 1<<dn->dn_phys->dn_indblkshift);
	dbuf_check_blkptr(dn, db);

	/* Provide the pending dirty record to child dbufs */
	db->db_data_pending = dr;

	mutex_exit(&db->db_mtx);

	dbuf_write(dr, db->db_buf, tx);

	zio_t *zio = dr->dr_zio;
	mutex_enter(&dr->dt.di.dr_mtx);
	dbuf_sync_list(&dr->dt.di.dr_children, db->db_level - 1, tx);
	ASSERT(list_head(&dr->dt.di.dr_children) == NULL);
	mutex_exit(&dr->dt.di.dr_mtx);
	zio_nowait(zio);
	}

	/*
	* Verify that the size of the data in our bonus buffer does not exceed
	* its recorded size.
	*
	* The purpose of this verification is to catch any cases in development
	* where the size of a phys structure (i.e space_map_phys_t) grows and,
	* due to incorrect feature management, older pools expect to read more
	* data even though they didn't actually write it to begin with.
	*
	* For a example, this would catch an error in the feature logic where we
	* open an older pool and we expect to write the space map histogram of
	* a space map with size SPACE_MAP_SIZE_V0.
	*/
	static void
	dbuf_sync_leaf_verify_bonus_dnode(dbuf_dirty_record_t *dr)
	{
	#ifdef ZFS_DEBUG
	dnode_t *dn = dr->dr_dnode;

	/*
	* Encrypted bonus buffers can have data past their bonuslen.
	* Skip the verification of these blocks.
	*/
	if (DMU_OT_IS_ENCRYPTED(dn->dn_bonustype))
	return;

	uint16_t bonuslen = dn->dn_phys->dn_bonuslen;
	uint16_t maxbonuslen = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots);
	ASSERT3U(bonuslen, <=, maxbonuslen);

	arc_buf_t *datap = dr->dt.dl.dr_data;
	char datap_end = ((char )datap) + bonuslen;
	char datap_max = ((char )datap) + maxbonuslen;

	/* ensure that everything is zero after our data */
	for (; datap_end < datap_max; datap_end++)
	ASSERT(*datap_end == 0);
	#endif
	}

	static blkptr_t *
	dbuf_lightweight_bp(dbuf_dirty_record_t *dr)
	{
	/* This must be a lightweight dirty record. */
	ASSERT3P(dr->dr_dbuf, ==, NULL);
	dnode_t *dn = dr->dr_dnode;

	if (dn->dn_phys->dn_nlevels == 1) {
	VERIFY3U(dr->dt.dll.dr_blkid, <, dn->dn_phys->dn_nblkptr);
	return (&dn->dn_phys->dn_blkptr[dr->dt.dll.dr_blkid]);
	} else {
	dmu_buf_impl_t *parent_db = dr->dr_parent->dr_dbuf;
	int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
	VERIFY3U(parent_db->db_level, ==, 1);
	VERIFY3P(parent_db->db_dnode_handle->dnh_dnode, ==, dn);
	VERIFY3U(dr->dt.dll.dr_blkid >> epbs, ==, parent_db->db_blkid);
	blkptr_t *bp = parent_db->db.db_data;
	return (&bp[dr->dt.dll.dr_blkid & ((1 << epbs) - 1)]);
	}
	}

	static void
	dbuf_lightweight_ready(zio_t *zio)
	{
	dbuf_dirty_record_t *dr = zio->io_private;
	blkptr_t *bp = zio->io_bp;

	if (zio->io_error != 0)
	return;

	dnode_t *dn = dr->dr_dnode;

	blkptr_t *bp_orig = dbuf_lightweight_bp(dr);
	spa_t *spa = dmu_objset_spa(dn->dn_objset);
	int64_t delta = bp_get_dsize_sync(spa, bp) -
	bp_get_dsize_sync(spa, bp_orig);
	dnode_diduse_space(dn, delta);

	uint64_t blkid = dr->dt.dll.dr_blkid;
	mutex_enter(&dn->dn_mtx);
	if (blkid > dn->dn_phys->dn_maxblkid) {
	ASSERT0(dn->dn_objset->os_raw_receive);
	dn->dn_phys->dn_maxblkid = blkid;
	}
	mutex_exit(&dn->dn_mtx);

	if (!BP_IS_EMBEDDED(bp)) {
	uint64_t fill = BP_IS_HOLE(bp) ? 0 : 1;
	BP_SET_FILL(bp, fill);
	}

	dmu_buf_impl_t *parent_db;
	EQUIV(dr->dr_parent == NULL, dn->dn_phys->dn_nlevels == 1);
	if (dr->dr_parent == NULL) {
	parent_db = dn->dn_dbuf;
	} else {
	parent_db = dr->dr_parent->dr_dbuf;
	}
	rw_enter(&parent_db->db_rwlock, RW_WRITER);
	bp_orig = bp;
	rw_exit(&parent_db->db_rwlock);
	}

	static void
	dbuf_lightweight_done(zio_t *zio)
	{
	dbuf_dirty_record_t *dr = zio->io_private;

	VERIFY0(zio->io_error);

	objset_t *os = dr->dr_dnode->dn_objset;
	dmu_tx_t *tx = os->os_synctx;

	if (zio->io_flags & (ZIO_FLAG_IO_REWRITE \| ZIO_FLAG_NOPWRITE)) {
	ASSERT(BP_EQUAL(zio->io_bp, &zio->io_bp_orig));
	} else {
	dsl_dataset_t *ds = os->os_dsl_dataset;
	(void) dsl_dataset_block_kill(ds, &zio->io_bp_orig, tx, B_TRUE);
	dsl_dataset_block_born(ds, zio->io_bp, tx);
	}

	dsl_pool_undirty_space(dmu_objset_pool(os), dr->dr_accounted,
	zio->io_txg);

	abd_free(dr->dt.dll.dr_abd);
	kmem_free(dr, sizeof (*dr));
	}

	noinline static void
	dbuf_sync_lightweight(dbuf_dirty_record_t dr, dmu_tx_t tx)
	{
	dnode_t *dn = dr->dr_dnode;
	zio_t *pio;
	if (dn->dn_phys->dn_nlevels == 1) {
	pio = dn->dn_zio;
	} else {
	pio = dr->dr_parent->dr_zio;
	}

	zbookmark_phys_t zb = {
	.zb_objset = dmu_objset_id(dn->dn_objset),
	.zb_object = dn->dn_object,
	.zb_level = 0,
	.zb_blkid = dr->dt.dll.dr_blkid,
	};

	/*
	* See comment in dbuf_write(). This is so that zio->io_bp_orig
	* will have the old BP in dbuf_lightweight_done().
	*/
	dr->dr_bp_copy = *dbuf_lightweight_bp(dr);

	dr->dr_zio = zio_write(pio, dmu_objset_spa(dn->dn_objset),
	dmu_tx_get_txg(tx), &dr->dr_bp_copy, dr->dt.dll.dr_abd,
	dn->dn_datablksz, abd_get_size(dr->dt.dll.dr_abd),
	&dr->dt.dll.dr_props, dbuf_lightweight_ready, NULL,
	dbuf_lightweight_done, dr, ZIO_PRIORITY_ASYNC_WRITE,
	ZIO_FLAG_MUSTSUCCEED \| dr->dt.dll.dr_flags, &zb);

	zio_nowait(dr->dr_zio);
	}

	/*
	* dbuf_sync_leaf() is called recursively from dbuf_sync_list() so it is
	* critical the we not allow the compiler to inline this function in to
	* dbuf_sync_list() thereby drastically bloating the stack usage.
	*/
	noinline static void
	dbuf_sync_leaf(dbuf_dirty_record_t dr, dmu_tx_t tx)
	{
	arc_buf_t **datap = &dr->dt.dl.dr_data;
	dmu_buf_impl_t *db = dr->dr_dbuf;
	dnode_t *dn = dr->dr_dnode;
	objset_t *os;
	uint64_t txg = tx->tx_txg;

	ASSERT(dmu_tx_is_syncing(tx));

	dprintf_dbuf_bp(db, db->db_blkptr, "blkptr=%p", db->db_blkptr);

	mutex_enter(&db->db_mtx);
	/*
	* To be synced, we must be dirtied. But we
	* might have been freed after the dirty.
	*/
	if (db->db_state == DB_UNCACHED) {
	/* This buffer has been freed since it was dirtied */
	ASSERT(db->db.db_data == NULL);
	} else if (db->db_state == DB_FILL) {
	/* This buffer was freed and is now being re-filled */
	ASSERT(db->db.db_data != dr->dt.dl.dr_data);
	} else if (db->db_state == DB_READ) {
	/*
	* This buffer has a clone we need to write, and an in-flight
	* read on the BP we're about to clone. Its safe to issue the
	* write here because the read has already been issued and the
	* contents won't change.
	*/
	ASSERT(dr->dt.dl.dr_brtwrite &&
	dr->dt.dl.dr_override_state == DR_OVERRIDDEN);
	} else {
	ASSERT(db->db_state == DB_CACHED \|\| db->db_state == DB_NOFILL);
	}
	DBUF_VERIFY(db);

	if (db->db_blkid == DMU_SPILL_BLKID) {
	mutex_enter(&dn->dn_mtx);
	if (!(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR)) {
	/*
	* In the previous transaction group, the bonus buffer
	* was entirely used to store the attributes for the
	* dnode which overrode the dn_spill field. However,
	* when adding more attributes to the file a spill
	* block was required to hold the extra attributes.
	*
	* Make sure to clear the garbage left in the dn_spill
	* field from the previous attributes in the bonus
	* buffer. Otherwise, after writing out the spill
	* block to the new allocated dva, it will free
	* the old block pointed to by the invalid dn_spill.
	*/
	db->db_blkptr = NULL;
	}
	dn->dn_phys->dn_flags \|= DNODE_FLAG_SPILL_BLKPTR;
	mutex_exit(&dn->dn_mtx);
	}

	/*
	* If this is a bonus buffer, simply copy the bonus data into the
	* dnode. It will be written out when the dnode is synced (and it
	* will be synced, since it must have been dirty for dbuf_sync to
	* be called).
	*/
	if (db->db_blkid == DMU_BONUS_BLKID) {
	ASSERT(dr->dr_dbuf == db);
	dbuf_sync_bonus(dr, tx);
	return;
	}

	os = dn->dn_objset;

	/*
	* This function may have dropped the db_mtx lock allowing a dmu_sync
	* operation to sneak in. As a result, we need to ensure that we
	* don't check the dr_override_state until we have returned from
	* dbuf_check_blkptr.
	*/
	dbuf_check_blkptr(dn, db);

	/*
	* If this buffer is in the middle of an immediate write,
	* wait for the synchronous IO to complete.
	*/
	while (dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC) {
	ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT);
	cv_wait(&db->db_changed, &db->db_mtx);
	}

	/*
	* If this is a dnode block, ensure it is appropriately encrypted
	* or decrypted, depending on what we are writing to it this txg.
	*/
	if (os->os_encrypted && dn->dn_object == DMU_META_DNODE_OBJECT)
	dbuf_prepare_encrypted_dnode_leaf(dr);

	if (datap != NULL && datap == db->db_buf &&
	dn->dn_object != DMU_META_DNODE_OBJECT &&
	zfs_refcount_count(&db->db_holds) > 1 &&
	dr->dt.dl.dr_override_state != DR_OVERRIDDEN) {
	/*
	* If this buffer is currently "in use" (i.e., there
	* are active holds and db_data still references it),
	* then make a copy before we start the write so that
	* any modifications from the open txg will not leak
	* into this write.
	*
	* NOTE: this copy does not need to be made for
	* objects only modified in the syncing context (e.g.
	* DNONE_DNODE blocks).
	*/
	int psize = arc_buf_size(*datap);
	int lsize = arc_buf_lsize(*datap);
	arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db);
	enum zio_compress compress_type = arc_get_compression(*datap);
	uint8_t complevel = arc_get_complevel(*datap);

	if (arc_is_encrypted(*datap)) {
	boolean_t byteorder;
	uint8_t salt[ZIO_DATA_SALT_LEN];
	uint8_t iv[ZIO_DATA_IV_LEN];
	uint8_t mac[ZIO_DATA_MAC_LEN];

	arc_get_raw_params(*datap, &byteorder, salt, iv, mac);
	*datap = arc_alloc_raw_buf(os->os_spa, db,
	dmu_objset_id(os), byteorder, salt, iv, mac,
	dn->dn_type, psize, lsize, compress_type,
	complevel);
	} else if (compress_type != ZIO_COMPRESS_OFF) {
	ASSERT3U(type, ==, ARC_BUFC_DATA);
	*datap = arc_alloc_compressed_buf(os->os_spa, db,
	psize, lsize, compress_type, complevel);
	} else {
	*datap = arc_alloc_buf(os->os_spa, db, type, psize);
	}
	memcpy((*datap)->b_data, db->db.db_data, psize);
	}
	db->db_data_pending = dr;

	mutex_exit(&db->db_mtx);

	dbuf_write(dr, *datap, tx);

	ASSERT(!list_link_active(&dr->dr_dirty_node));
	if (dn->dn_object == DMU_META_DNODE_OBJECT) {
	list_insert_tail(&dn->dn_dirty_records[txg & TXG_MASK], dr);
	} else {
	zio_nowait(dr->dr_zio);
	}
	}

	/*
	* Syncs out a range of dirty records for indirect or leaf dbufs. May be
	* called recursively from dbuf_sync_indirect().
	*/
	void
	dbuf_sync_list(list_t list, int level, dmu_tx_t tx)
	{
	dbuf_dirty_record_t *dr;

	while ((dr = list_head(list))) {
	if (dr->dr_zio != NULL) {
	/*
	* If we find an already initialized zio then we
	* are processing the meta-dnode, and we have finished.
	* The dbufs for all dnodes are put back on the list
	* during processing, so that we can zio_wait()
	* these IOs after initiating all child IOs.
	*/
	ASSERT3U(dr->dr_dbuf->db.db_object, ==,
	DMU_META_DNODE_OBJECT);
	break;
	}
	list_remove(list, dr);
	if (dr->dr_dbuf == NULL) {
	dbuf_sync_lightweight(dr, tx);
	} else {
	if (dr->dr_dbuf->db_blkid != DMU_BONUS_BLKID &&
	dr->dr_dbuf->db_blkid != DMU_SPILL_BLKID) {
	VERIFY3U(dr->dr_dbuf->db_level, ==, level);
	}
	if (dr->dr_dbuf->db_level > 0)
	dbuf_sync_indirect(dr, tx);
	else
	dbuf_sync_leaf(dr, tx);
	}
	}
	}

	static void
	dbuf_write_ready(zio_t zio, arc_buf_t buf, void *vdb)
	{
	(void) buf;
	dmu_buf_impl_t *db = vdb;
	dnode_t *dn;
	blkptr_t *bp = zio->io_bp;
	blkptr_t *bp_orig = &zio->io_bp_orig;
	spa_t *spa = zio->io_spa;
	int64_t delta;
	uint64_t fill = 0;
	int i;

	ASSERT3P(db->db_blkptr, !=, NULL);
	ASSERT3P(&db->db_data_pending->dr_bp_copy, ==, bp);

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	delta = bp_get_dsize_sync(spa, bp) - bp_get_dsize_sync(spa, bp_orig);
	dnode_diduse_space(dn, delta - zio->io_prev_space_delta);
	zio->io_prev_space_delta = delta;

	if (BP_GET_LOGICAL_BIRTH(bp) != 0) {
	ASSERT((db->db_blkid != DMU_SPILL_BLKID &&
	BP_GET_TYPE(bp) == dn->dn_type) \|\|
	(db->db_blkid == DMU_SPILL_BLKID &&
	BP_GET_TYPE(bp) == dn->dn_bonustype) \|\|
	BP_IS_EMBEDDED(bp));
	ASSERT(BP_GET_LEVEL(bp) == db->db_level);
	}

	mutex_enter(&db->db_mtx);

	#ifdef ZFS_DEBUG
	if (db->db_blkid == DMU_SPILL_BLKID) {
	ASSERT(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR);
	ASSERT(!(BP_IS_HOLE(bp)) &&
	db->db_blkptr == DN_SPILL_BLKPTR(dn->dn_phys));
	}
	#endif

	if (db->db_level == 0) {
	mutex_enter(&dn->dn_mtx);
	if (db->db_blkid > dn->dn_phys->dn_maxblkid &&
	db->db_blkid != DMU_SPILL_BLKID) {
	ASSERT0(db->db_objset->os_raw_receive);
	dn->dn_phys->dn_maxblkid = db->db_blkid;
	}
	mutex_exit(&dn->dn_mtx);

	if (dn->dn_type == DMU_OT_DNODE) {
	i = 0;
	while (i < db->db.db_size) {
	dnode_phys_t *dnp =
	(void )(((char )db->db.db_data) + i);

	i += DNODE_MIN_SIZE;
	if (dnp->dn_type != DMU_OT_NONE) {
	fill++;
	for (int j = 0; j < dnp->dn_nblkptr;
	j++) {
	(void) zfs_blkptr_verify(spa,
	&dnp->dn_blkptr[j],
	BLK_CONFIG_SKIP,
	BLK_VERIFY_HALT);
	}
	if (dnp->dn_flags &
	DNODE_FLAG_SPILL_BLKPTR) {
	(void) zfs_blkptr_verify(spa,
	DN_SPILL_BLKPTR(dnp),
	BLK_CONFIG_SKIP,
	BLK_VERIFY_HALT);
	}
	i += dnp->dn_extra_slots *
	DNODE_MIN_SIZE;
	}
	}
	} else {
	if (BP_IS_HOLE(bp)) {
	fill = 0;
	} else {
	fill = 1;
	}
	}
	} else {
	blkptr_t *ibp = db->db.db_data;
	ASSERT3U(db->db.db_size, ==, 1<<dn->dn_phys->dn_indblkshift);
	for (i = db->db.db_size >> SPA_BLKPTRSHIFT; i > 0; i--, ibp++) {
	if (BP_IS_HOLE(ibp))
	continue;
	(void) zfs_blkptr_verify(spa, ibp,
	BLK_CONFIG_SKIP, BLK_VERIFY_HALT);
	fill += BP_GET_FILL(ibp);
	}
	}
	DB_DNODE_EXIT(db);

	if (!BP_IS_EMBEDDED(bp))
	BP_SET_FILL(bp, fill);

	mutex_exit(&db->db_mtx);

	db_lock_type_t dblt = dmu_buf_lock_parent(db, RW_WRITER, FTAG);
	db->db_blkptr = bp;
	dmu_buf_unlock_parent(db, dblt, FTAG);
	}

	/*
	* This function gets called just prior to running through the compression
	* stage of the zio pipeline. If we're an indirect block comprised of only
	* holes, then we want this indirect to be compressed away to a hole. In
	* order to do that we must zero out any information about the holes that
	* this indirect points to prior to before we try to compress it.
	*/
	static void
	dbuf_write_children_ready(zio_t zio, arc_buf_t buf, void *vdb)
	{
	(void) zio, (void) buf;
	dmu_buf_impl_t *db = vdb;
	dnode_t *dn;
	blkptr_t *bp;
	unsigned int epbs, i;

	ASSERT3U(db->db_level, >, 0);
	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	epbs = dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT;
	ASSERT3U(epbs, <, 31);

	/* Determine if all our children are holes */
	for (i = 0, bp = db->db.db_data; i < 1ULL << epbs; i++, bp++) {
	if (!BP_IS_HOLE(bp))
	break;
	}

	/*
	* If all the children are holes, then zero them all out so that
	* we may get compressed away.
	*/
	if (i == 1ULL << epbs) {
	/*
	* We only found holes. Grab the rwlock to prevent
	* anybody from reading the blocks we're about to
	* zero out.
	*/
	rw_enter(&db->db_rwlock, RW_WRITER);
	memset(db->db.db_data, 0, db->db.db_size);
	rw_exit(&db->db_rwlock);
	}
	DB_DNODE_EXIT(db);
	}

	static void
	dbuf_write_done(zio_t zio, arc_buf_t buf, void *vdb)
	{
	(void) buf;
	dmu_buf_impl_t *db = vdb;
	blkptr_t *bp_orig = &zio->io_bp_orig;
	blkptr_t *bp = db->db_blkptr;
	objset_t *os = db->db_objset;
	dmu_tx_t *tx = os->os_synctx;

	ASSERT0(zio->io_error);
	ASSERT(db->db_blkptr == bp);

	/*
	* For nopwrites and rewrites we ensure that the bp matches our
	* original and bypass all the accounting.
	*/
	if (zio->io_flags & (ZIO_FLAG_IO_REWRITE \| ZIO_FLAG_NOPWRITE)) {
	ASSERT(BP_EQUAL(bp, bp_orig));
	} else {
	dsl_dataset_t *ds = os->os_dsl_dataset;
	(void) dsl_dataset_block_kill(ds, bp_orig, tx, B_TRUE);
	dsl_dataset_block_born(ds, bp, tx);
	}

	mutex_enter(&db->db_mtx);

	DBUF_VERIFY(db);

	dbuf_dirty_record_t *dr = db->db_data_pending;
	dnode_t *dn = dr->dr_dnode;
	ASSERT(!list_link_active(&dr->dr_dirty_node));
	ASSERT(dr->dr_dbuf == db);
	ASSERT(list_next(&db->db_dirty_records, dr) == NULL);
	list_remove(&db->db_dirty_records, dr);

	#ifdef ZFS_DEBUG
	if (db->db_blkid == DMU_SPILL_BLKID) {
	ASSERT(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR);
	ASSERT(!(BP_IS_HOLE(db->db_blkptr)) &&
	db->db_blkptr == DN_SPILL_BLKPTR(dn->dn_phys));
	}
	#endif

	if (db->db_level == 0) {
	ASSERT(db->db_blkid != DMU_BONUS_BLKID);
	ASSERT(dr->dt.dl.dr_override_state == DR_NOT_OVERRIDDEN);
	if (dr->dt.dl.dr_data != NULL &&
	dr->dt.dl.dr_data != db->db_buf) {
	arc_buf_destroy(dr->dt.dl.dr_data, db);
	}
	} else {
	ASSERT(list_head(&dr->dt.di.dr_children) == NULL);
	ASSERT3U(db->db.db_size, ==, 1 << dn->dn_phys->dn_indblkshift);
	if (!BP_IS_HOLE(db->db_blkptr)) {
	int epbs __maybe_unused = dn->dn_phys->dn_indblkshift -
	SPA_BLKPTRSHIFT;
	ASSERT3U(db->db_blkid, <=,
	dn->dn_phys->dn_maxblkid >> (db->db_level * epbs));
	ASSERT3U(BP_GET_LSIZE(db->db_blkptr), ==,
	db->db.db_size);
	}
	mutex_destroy(&dr->dt.di.dr_mtx);
	list_destroy(&dr->dt.di.dr_children);
	}

	cv_broadcast(&db->db_changed);
	ASSERT(db->db_dirtycnt > 0);
	db->db_dirtycnt -= 1;
	db->db_data_pending = NULL;
	dbuf_rele_and_unlock(db, (void *)(uintptr_t)tx->tx_txg, B_FALSE);

	dsl_pool_undirty_space(dmu_objset_pool(os), dr->dr_accounted,
	zio->io_txg);

	kmem_free(dr, sizeof (dbuf_dirty_record_t));
	}

	static void
	dbuf_write_nofill_ready(zio_t *zio)
	{
	dbuf_write_ready(zio, NULL, zio->io_private);
	}

	static void
	dbuf_write_nofill_done(zio_t *zio)
	{
	dbuf_write_done(zio, NULL, zio->io_private);
	}

	static void
	dbuf_write_override_ready(zio_t *zio)
	{
	dbuf_dirty_record_t *dr = zio->io_private;
	dmu_buf_impl_t *db = dr->dr_dbuf;

	dbuf_write_ready(zio, NULL, db);
	}

	static void
	dbuf_write_override_done(zio_t *zio)
	{
	dbuf_dirty_record_t *dr = zio->io_private;
	dmu_buf_impl_t *db = dr->dr_dbuf;
	blkptr_t *obp = &dr->dt.dl.dr_overridden_by;

	mutex_enter(&db->db_mtx);
	if (!BP_EQUAL(zio->io_bp, obp)) {
	if (!BP_IS_HOLE(obp))
	dsl_free(spa_get_dsl(zio->io_spa), zio->io_txg, obp);
	arc_release(dr->dt.dl.dr_data, db);
	}
	mutex_exit(&db->db_mtx);

	dbuf_write_done(zio, NULL, db);

	if (zio->io_abd != NULL)
	abd_free(zio->io_abd);
	}

	typedef struct dbuf_remap_impl_callback_arg {
	objset_t *drica_os;
	uint64_t drica_blk_birth;
	dmu_tx_t *drica_tx;
	} dbuf_remap_impl_callback_arg_t;

	static void
	dbuf_remap_impl_callback(uint64_t vdev, uint64_t offset, uint64_t size,
	void *arg)
	{
	dbuf_remap_impl_callback_arg_t *drica = arg;
	objset_t *os = drica->drica_os;
	spa_t *spa = dmu_objset_spa(os);
	dmu_tx_t *tx = drica->drica_tx;

	ASSERT(dsl_pool_sync_context(spa_get_dsl(spa)));

	if (os == spa_meta_objset(spa)) {
	spa_vdev_indirect_mark_obsolete(spa, vdev, offset, size, tx);
	} else {
	dsl_dataset_block_remapped(dmu_objset_ds(os), vdev, offset,
	size, drica->drica_blk_birth, tx);
	}
	}

	static void
	dbuf_remap_impl(dnode_t dn, blkptr_t bp, krwlock_t rw, dmu_tx_t tx)
	{
	blkptr_t bp_copy = *bp;
	spa_t *spa = dmu_objset_spa(dn->dn_objset);
	dbuf_remap_impl_callback_arg_t drica;

	ASSERT(dsl_pool_sync_context(spa_get_dsl(spa)));

	drica.drica_os = dn->dn_objset;
	drica.drica_blk_birth = BP_GET_LOGICAL_BIRTH(bp);
	drica.drica_tx = tx;
	if (spa_remap_blkptr(spa, &bp_copy, dbuf_remap_impl_callback,
	&drica)) {
	/*
	* If the blkptr being remapped is tracked by a livelist,
	* then we need to make sure the livelist reflects the update.
	* First, cancel out the old blkptr by appending a 'FREE'
	* entry. Next, add an 'ALLOC' to track the new version. This
	* way we avoid trying to free an inaccurate blkptr at delete.
	* Note that embedded blkptrs are not tracked in livelists.
	*/
	if (dn->dn_objset != spa_meta_objset(spa)) {
	dsl_dataset_t *ds = dmu_objset_ds(dn->dn_objset);
	if (dsl_deadlist_is_open(&ds->ds_dir->dd_livelist) &&
	BP_GET_LOGICAL_BIRTH(bp) >
	ds->ds_dir->dd_origin_txg) {
	ASSERT(!BP_IS_EMBEDDED(bp));
	ASSERT(dsl_dir_is_clone(ds->ds_dir));
	ASSERT(spa_feature_is_enabled(spa,
	SPA_FEATURE_LIVELIST));
	bplist_append(&ds->ds_dir->dd_pending_frees,
	bp);
	bplist_append(&ds->ds_dir->dd_pending_allocs,
	&bp_copy);
	}
	}

	/*
	* The db_rwlock prevents dbuf_read_impl() from
	* dereferencing the BP while we are changing it. To
	* avoid lock contention, only grab it when we are actually
	* changing the BP.
	*/
	if (rw != NULL)
	rw_enter(rw, RW_WRITER);
	*bp = bp_copy;
	if (rw != NULL)
	rw_exit(rw);
	}
	}

	/*
	* Remap any existing BP's to concrete vdevs, if possible.
	*/
	static void
	dbuf_remap(dnode_t dn, dmu_buf_impl_t db, dmu_tx_t *tx)
	{
	spa_t *spa = dmu_objset_spa(db->db_objset);
	ASSERT(dsl_pool_sync_context(spa_get_dsl(spa)));

	if (!spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REMOVAL))
	return;

	if (db->db_level > 0) {
	blkptr_t *bp = db->db.db_data;
	for (int i = 0; i < db->db.db_size >> SPA_BLKPTRSHIFT; i++) {
	dbuf_remap_impl(dn, &bp[i], &db->db_rwlock, tx);
	}
	} else if (db->db.db_object == DMU_META_DNODE_OBJECT) {
	dnode_phys_t *dnp = db->db.db_data;
	ASSERT3U(db->db_dnode_handle->dnh_dnode->dn_type, ==,
	DMU_OT_DNODE);
	for (int i = 0; i < db->db.db_size >> DNODE_SHIFT;
	i += dnp[i].dn_extra_slots + 1) {
	for (int j = 0; j < dnp[i].dn_nblkptr; j++) {
	krwlock_t *lock = (dn->dn_dbuf == NULL ? NULL :
	&dn->dn_dbuf->db_rwlock);
	dbuf_remap_impl(dn, &dnp[i].dn_blkptr[j], lock,
	tx);
	}
	}
	}
	}


	/*
	* Populate dr->dr_zio with a zio to commit a dirty buffer to disk.
	* Caller is responsible for issuing the zio_[no]wait(dr->dr_zio).
	*/
	static void
	dbuf_write(dbuf_dirty_record_t dr, arc_buf_t data, dmu_tx_t *tx)
	{
	dmu_buf_impl_t *db = dr->dr_dbuf;
	dnode_t *dn = dr->dr_dnode;
	objset_t *os;
	dmu_buf_impl_t *parent = db->db_parent;
	uint64_t txg = tx->tx_txg;
	zbookmark_phys_t zb;
	zio_prop_t zp;
	zio_t pio; / parent I/O */
	int wp_flag = 0;

	ASSERT(dmu_tx_is_syncing(tx));

	os = dn->dn_objset;

	if (db->db_level > 0 \|\| dn->dn_type == DMU_OT_DNODE) {
	/*
	* Private object buffers are released here rather than in
	* dbuf_dirty() since they are only modified in the syncing
	* context and we don't want the overhead of making multiple
	* copies of the data.
	*/
	if (BP_IS_HOLE(db->db_blkptr))
	arc_buf_thaw(data);
	else
	dbuf_release_bp(db);
	dbuf_remap(dn, db, tx);
	}

	if (parent != dn->dn_dbuf) {
	/* Our parent is an indirect block. */
	/* We have a dirty parent that has been scheduled for write. */
	ASSERT(parent && parent->db_data_pending);
	/* Our parent's buffer is one level closer to the dnode. */
	ASSERT(db->db_level == parent->db_level-1);
	/*
	* We're about to modify our parent's db_data by modifying
	* our block pointer, so the parent must be released.
	*/
	ASSERT(arc_released(parent->db_buf));
	pio = parent->db_data_pending->dr_zio;
	} else {
	/* Our parent is the dnode itself. */
	ASSERT((db->db_level == dn->dn_phys->dn_nlevels-1 &&
	db->db_blkid != DMU_SPILL_BLKID) \|\|
	(db->db_blkid == DMU_SPILL_BLKID && db->db_level == 0));
	if (db->db_blkid != DMU_SPILL_BLKID)
	ASSERT3P(db->db_blkptr, ==,
	&dn->dn_phys->dn_blkptr[db->db_blkid]);
	pio = dn->dn_zio;
	}

	ASSERT(db->db_level == 0 \|\| data == db->db_buf);
	ASSERT3U(BP_GET_LOGICAL_BIRTH(db->db_blkptr), <=, txg);
	ASSERT(pio);

	SET_BOOKMARK(&zb, os->os_dsl_dataset ?
	os->os_dsl_dataset->ds_object : DMU_META_OBJSET,
	db->db.db_object, db->db_level, db->db_blkid);

	if (db->db_blkid == DMU_SPILL_BLKID)
	wp_flag = WP_SPILL;
	wp_flag \|= (data == NULL) ? WP_NOFILL : 0;

	dmu_write_policy(os, dn, db->db_level, wp_flag, &zp);

	/*
	* We copy the blkptr now (rather than when we instantiate the dirty
	* record), because its value can change between open context and
	* syncing context. We do not need to hold dn_struct_rwlock to read
	* db_blkptr because we are in syncing context.
	*/
	dr->dr_bp_copy = *db->db_blkptr;

	if (db->db_level == 0 &&
	dr->dt.dl.dr_override_state == DR_OVERRIDDEN) {
	/*
	* The BP for this block has been provided by open context
	* (by dmu_sync() or dmu_buf_write_embedded()).
	*/
	abd_t *contents = (data != NULL) ?
	abd_get_from_buf(data->b_data, arc_buf_size(data)) : NULL;

	dr->dr_zio = zio_write(pio, os->os_spa, txg, &dr->dr_bp_copy,
	contents, db->db.db_size, db->db.db_size, &zp,
	dbuf_write_override_ready, NULL,
	dbuf_write_override_done,
	dr, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb);
	mutex_enter(&db->db_mtx);
	dr->dt.dl.dr_override_state = DR_NOT_OVERRIDDEN;
	zio_write_override(dr->dr_zio, &dr->dt.dl.dr_overridden_by,
	dr->dt.dl.dr_copies, dr->dt.dl.dr_nopwrite,
	dr->dt.dl.dr_brtwrite);
	mutex_exit(&db->db_mtx);
	} else if (data == NULL) {
	ASSERT(zp.zp_checksum == ZIO_CHECKSUM_OFF \|\|
	zp.zp_checksum == ZIO_CHECKSUM_NOPARITY);
	dr->dr_zio = zio_write(pio, os->os_spa, txg,
	&dr->dr_bp_copy, NULL, db->db.db_size, db->db.db_size, &zp,
	dbuf_write_nofill_ready, NULL,
	dbuf_write_nofill_done, db,
	ZIO_PRIORITY_ASYNC_WRITE,
	ZIO_FLAG_MUSTSUCCEED \| ZIO_FLAG_NODATA, &zb);
	} else {
	ASSERT(arc_released(data));

	/*
	* For indirect blocks, we want to setup the children
	* ready callback so that we can properly handle an indirect
	* block that only contains holes.
	*/
	arc_write_done_func_t *children_ready_cb = NULL;
	if (db->db_level != 0)
	children_ready_cb = dbuf_write_children_ready;

	dr->dr_zio = arc_write(pio, os->os_spa, txg,
	&dr->dr_bp_copy, data, !DBUF_IS_CACHEABLE(db),
	dbuf_is_l2cacheable(db), &zp, dbuf_write_ready,
	children_ready_cb, dbuf_write_done, db,
	ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb);
	}
	}

	EXPORT_SYMBOL(dbuf_find);
	EXPORT_SYMBOL(dbuf_is_metadata);
	EXPORT_SYMBOL(dbuf_destroy);
	EXPORT_SYMBOL(dbuf_loan_arcbuf);
	EXPORT_SYMBOL(dbuf_whichblock);
	EXPORT_SYMBOL(dbuf_read);
	EXPORT_SYMBOL(dbuf_unoverride);
	EXPORT_SYMBOL(dbuf_free_range);
	EXPORT_SYMBOL(dbuf_new_size);
	EXPORT_SYMBOL(dbuf_release_bp);
	EXPORT_SYMBOL(dbuf_dirty);
	EXPORT_SYMBOL(dmu_buf_set_crypt_params);
	EXPORT_SYMBOL(dmu_buf_will_dirty);
	EXPORT_SYMBOL(dmu_buf_is_dirty);
	EXPORT_SYMBOL(dmu_buf_will_clone);
	EXPORT_SYMBOL(dmu_buf_will_not_fill);
	EXPORT_SYMBOL(dmu_buf_will_fill);
	EXPORT_SYMBOL(dmu_buf_fill_done);
	EXPORT_SYMBOL(dmu_buf_rele);
	EXPORT_SYMBOL(dbuf_assign_arcbuf);
	EXPORT_SYMBOL(dbuf_prefetch);
	EXPORT_SYMBOL(dbuf_hold_impl);
	EXPORT_SYMBOL(dbuf_hold);
	EXPORT_SYMBOL(dbuf_hold_level);
	EXPORT_SYMBOL(dbuf_create_bonus);
	EXPORT_SYMBOL(dbuf_spill_set_blksz);
	EXPORT_SYMBOL(dbuf_rm_spill);
	EXPORT_SYMBOL(dbuf_add_ref);
	EXPORT_SYMBOL(dbuf_rele);
	EXPORT_SYMBOL(dbuf_rele_and_unlock);
	EXPORT_SYMBOL(dbuf_refcount);
	EXPORT_SYMBOL(dbuf_sync_list);
	EXPORT_SYMBOL(dmu_buf_set_user);
	EXPORT_SYMBOL(dmu_buf_set_user_ie);
	EXPORT_SYMBOL(dmu_buf_get_user);
	EXPORT_SYMBOL(dmu_buf_get_blkptr);

	ZFS_MODULE_PARAM(zfs_dbuf_cache, dbuf_cache_, max_bytes, U64, ZMOD_RW,
	"Maximum size in bytes of the dbuf cache.");

	ZFS_MODULE_PARAM(zfs_dbuf_cache, dbuf_cache_, hiwater_pct, UINT, ZMOD_RW,
	"Percentage over dbuf_cache_max_bytes for direct dbuf eviction.");

	ZFS_MODULE_PARAM(zfs_dbuf_cache, dbuf_cache_, lowater_pct, UINT, ZMOD_RW,
	"Percentage below dbuf_cache_max_bytes when dbuf eviction stops.");

	ZFS_MODULE_PARAM(zfs_dbuf, dbuf_, metadata_cache_max_bytes, U64, ZMOD_RW,
	"Maximum size in bytes of dbuf metadata cache.");

	ZFS_MODULE_PARAM(zfs_dbuf, dbuf_, cache_shift, UINT, ZMOD_RW,
	"Set size of dbuf cache to log2 fraction of arc size.");

	ZFS_MODULE_PARAM(zfs_dbuf, dbuf_, metadata_cache_shift, UINT, ZMOD_RW,
	"Set size of dbuf metadata cache to log2 fraction of arc size.");

	ZFS_MODULE_PARAM(zfs_dbuf, dbuf_, mutex_cache_shift, UINT, ZMOD_RD,
	"Set size of dbuf cache mutex array as log2 shift.");
	diff --git a/sys/contrib/openzfs/module/zfs/dmu.c b/sys/contrib/openzfs/module/zfs/dmu.c
	index 6ef149aab9a6..8b440aafba43 100644
	--- a/sys/contrib/openzfs/module/zfs/dmu.c
	+++ b/sys/contrib/openzfs/module/zfs/dmu.c
	@@ -1,2618 +1,2619 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
	* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
	* Copyright (c) 2013, Joyent, Inc. All rights reserved.
	* Copyright (c) 2016, Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2015 by Chunwei Chen. All rights reserved.
	* Copyright (c) 2019 Datto Inc.
	* Copyright (c) 2019, Klara Inc.
	* Copyright (c) 2019, Allan Jude
	* Copyright (c) 2022 Hewlett Packard Enterprise Development LP.
	* Copyright (c) 2021, 2022 by Pawel Jakub Dawidek
	*/

	#include <sys/dmu.h>
	#include <sys/dmu_impl.h>
	#include <sys/dmu_tx.h>
	#include <sys/dbuf.h>
	#include <sys/dnode.h>
	#include <sys/zfs_context.h>
	#include <sys/dmu_objset.h>
	#include <sys/dmu_traverse.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_dir.h>
	#include <sys/dsl_pool.h>
	#include <sys/dsl_synctask.h>
	#include <sys/dsl_prop.h>
	#include <sys/dmu_zfetch.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/zap.h>
	#include <sys/zio_checksum.h>
	#include <sys/zio_compress.h>
	#include <sys/sa.h>
	#include <sys/zfeature.h>
	#include <sys/abd.h>
	#include <sys/brt.h>
	#include <sys/trace_zfs.h>
	#include <sys/zfs_racct.h>
	#include <sys/zfs_rlock.h>
	#ifdef _KERNEL
	#include <sys/vmsystm.h>
	#include <sys/zfs_znode.h>
	#endif

	/*
	* Enable/disable nopwrite feature.
	*/
	static int zfs_nopwrite_enabled = 1;

	/*
	* Tunable to control percentage of dirtied L1 blocks from frees allowed into
	* one TXG. After this threshold is crossed, additional dirty blocks from frees
	* will wait until the next TXG.
	* A value of zero will disable this throttle.
	*/
	static uint_t zfs_per_txg_dirty_frees_percent = 30;

	/*
	* Enable/disable forcing txg sync when dirty checking for holes with lseek().
	* By default this is enabled to ensure accurate hole reporting, it can result
	* in a significant performance penalty for lseek(SEEK_HOLE) heavy workloads.
	* Disabling this option will result in holes never being reported in dirty
	* files which is always safe.
	*/
	static int zfs_dmu_offset_next_sync = 1;

	/*
	* Limit the amount we can prefetch with one call to this amount. This
	* helps to limit the amount of memory that can be used by prefetching.
	* Larger objects should be prefetched a bit at a time.
	*/
	#ifdef _ILP32
	uint_t dmu_prefetch_max = 8 * 1024 * 1024;
	#else
	uint_t dmu_prefetch_max = 8 * SPA_MAXBLOCKSIZE;
	#endif

	const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES] = {
	{DMU_BSWAP_UINT8, TRUE, FALSE, FALSE, "unallocated" },
	{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "object directory" },
	{DMU_BSWAP_UINT64, TRUE, TRUE, FALSE, "object array" },
	{DMU_BSWAP_UINT8, TRUE, FALSE, FALSE, "packed nvlist" },
	{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "packed nvlist size" },
	{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "bpobj" },
	{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "bpobj header" },
	{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "SPA space map header" },
	{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "SPA space map" },
	{DMU_BSWAP_UINT64, TRUE, FALSE, TRUE, "ZIL intent log" },
	{DMU_BSWAP_DNODE, TRUE, FALSE, TRUE, "DMU dnode" },
	{DMU_BSWAP_OBJSET, TRUE, TRUE, FALSE, "DMU objset" },
	{DMU_BSWAP_UINT64, TRUE, TRUE, FALSE, "DSL directory" },
	{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "DSL directory child map"},
	{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "DSL dataset snap map" },
	{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "DSL props" },
	{DMU_BSWAP_UINT64, TRUE, TRUE, FALSE, "DSL dataset" },
	{DMU_BSWAP_ZNODE, TRUE, FALSE, FALSE, "ZFS znode" },
	{DMU_BSWAP_OLDACL, TRUE, FALSE, TRUE, "ZFS V0 ACL" },
	{DMU_BSWAP_UINT8, FALSE, FALSE, TRUE, "ZFS plain file" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, TRUE, "ZFS directory" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "ZFS master node" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, TRUE, "ZFS delete queue" },
	{DMU_BSWAP_UINT8, FALSE, FALSE, TRUE, "zvol object" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "zvol prop" },
	{DMU_BSWAP_UINT8, FALSE, FALSE, TRUE, "other uint8[]" },
	{DMU_BSWAP_UINT64, FALSE, FALSE, TRUE, "other uint64[]" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "other ZAP" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "persistent error log" },
	{DMU_BSWAP_UINT8, TRUE, FALSE, FALSE, "SPA history" },
	{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "SPA history offsets" },
	{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "Pool properties" },
	{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "DSL permissions" },
	{DMU_BSWAP_ACL, TRUE, FALSE, TRUE, "ZFS ACL" },
	{DMU_BSWAP_UINT8, TRUE, FALSE, TRUE, "ZFS SYSACL" },
	{DMU_BSWAP_UINT8, TRUE, FALSE, TRUE, "FUID table" },
	{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "FUID table size" },
	{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "DSL dataset next clones"},
	{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "scan work queue" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, TRUE, "ZFS user/group/project used" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, TRUE, "ZFS user/group/project quota"},
	{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "snapshot refcount tags"},
	{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "DDT ZAP algorithm" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "DDT statistics" },
	{DMU_BSWAP_UINT8, TRUE, FALSE, TRUE, "System attributes" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, TRUE, "SA master node" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, TRUE, "SA attr registration" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, TRUE, "SA attr layouts" },
	{DMU_BSWAP_ZAP, TRUE, FALSE, FALSE, "scan translations" },
	{DMU_BSWAP_UINT8, FALSE, FALSE, TRUE, "deduplicated block" },
	{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "DSL deadlist map" },
	{DMU_BSWAP_UINT64, TRUE, TRUE, FALSE, "DSL deadlist map hdr" },
	{DMU_BSWAP_ZAP, TRUE, TRUE, FALSE, "DSL dir clones" },
	{DMU_BSWAP_UINT64, TRUE, FALSE, FALSE, "bpobj subobj" }
	};

	dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS] = {
	{ byteswap_uint8_array, "uint8" },
	{ byteswap_uint16_array, "uint16" },
	{ byteswap_uint32_array, "uint32" },
	{ byteswap_uint64_array, "uint64" },
	{ zap_byteswap, "zap" },
	{ dnode_buf_byteswap, "dnode" },
	{ dmu_objset_byteswap, "objset" },
	{ zfs_znode_byteswap, "znode" },
	{ zfs_oldacl_byteswap, "oldacl" },
	{ zfs_acl_byteswap, "acl" }
	};

	int
	dmu_buf_hold_noread_by_dnode(dnode_t *dn, uint64_t offset,
	const void tag, dmu_buf_t *dbp)
	{
	uint64_t blkid;
	dmu_buf_impl_t *db;

	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	blkid = dbuf_whichblock(dn, 0, offset);
	db = dbuf_hold(dn, blkid, tag);
	rw_exit(&dn->dn_struct_rwlock);

	if (db == NULL) {
	*dbp = NULL;
	return (SET_ERROR(EIO));
	}

	*dbp = &db->db;
	return (0);
	}

	int
	dmu_buf_hold_noread(objset_t *os, uint64_t object, uint64_t offset,
	const void tag, dmu_buf_t *dbp)
	{
	dnode_t *dn;
	uint64_t blkid;
	dmu_buf_impl_t *db;
	int err;

	err = dnode_hold(os, object, FTAG, &dn);
	if (err)
	return (err);
	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	blkid = dbuf_whichblock(dn, 0, offset);
	db = dbuf_hold(dn, blkid, tag);
	rw_exit(&dn->dn_struct_rwlock);
	dnode_rele(dn, FTAG);

	if (db == NULL) {
	*dbp = NULL;
	return (SET_ERROR(EIO));
	}

	*dbp = &db->db;
	return (err);
	}

	int
	dmu_buf_hold_by_dnode(dnode_t *dn, uint64_t offset,
	const void tag, dmu_buf_t *dbp, int flags)
	{
	int err;
	int db_flags = DB_RF_CANFAIL;

	if (flags & DMU_READ_NO_PREFETCH)
	db_flags \|= DB_RF_NOPREFETCH;
	if (flags & DMU_READ_NO_DECRYPT)
	db_flags \|= DB_RF_NO_DECRYPT;

	err = dmu_buf_hold_noread_by_dnode(dn, offset, tag, dbp);
	if (err == 0) {
	dmu_buf_impl_t db = (dmu_buf_impl_t )(*dbp);
	err = dbuf_read(db, NULL, db_flags);
	if (err != 0) {
	dbuf_rele(db, tag);
	*dbp = NULL;
	}
	}

	return (err);
	}

	int
	dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
	const void tag, dmu_buf_t *dbp, int flags)
	{
	int err;
	int db_flags = DB_RF_CANFAIL;

	if (flags & DMU_READ_NO_PREFETCH)
	db_flags \|= DB_RF_NOPREFETCH;
	if (flags & DMU_READ_NO_DECRYPT)
	db_flags \|= DB_RF_NO_DECRYPT;

	err = dmu_buf_hold_noread(os, object, offset, tag, dbp);
	if (err == 0) {
	dmu_buf_impl_t db = (dmu_buf_impl_t )(*dbp);
	err = dbuf_read(db, NULL, db_flags);
	if (err != 0) {
	dbuf_rele(db, tag);
	*dbp = NULL;
	}
	}

	return (err);
	}

	int
	dmu_bonus_max(void)
	{
	return (DN_OLD_MAX_BONUSLEN);
	}

	int
	dmu_set_bonus(dmu_buf_t db_fake, int newsize, dmu_tx_t tx)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;
	dnode_t *dn;
	int error;

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);

	if (dn->dn_bonus != db) {
	error = SET_ERROR(EINVAL);
	} else if (newsize < 0 \|\| newsize > db_fake->db_size) {
	error = SET_ERROR(EINVAL);
	} else {
	dnode_setbonuslen(dn, newsize, tx);
	error = 0;
	}

	DB_DNODE_EXIT(db);
	return (error);
	}

	int
	dmu_set_bonustype(dmu_buf_t db_fake, dmu_object_type_t type, dmu_tx_t tx)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;
	dnode_t *dn;
	int error;

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);

	if (!DMU_OT_IS_VALID(type)) {
	error = SET_ERROR(EINVAL);
	} else if (dn->dn_bonus != db) {
	error = SET_ERROR(EINVAL);
	} else {
	dnode_setbonus_type(dn, type, tx);
	error = 0;
	}

	DB_DNODE_EXIT(db);
	return (error);
	}

	dmu_object_type_t
	dmu_get_bonustype(dmu_buf_t *db_fake)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;
	dnode_t *dn;
	dmu_object_type_t type;

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	type = dn->dn_bonustype;
	DB_DNODE_EXIT(db);

	return (type);
	}

	int
	dmu_rm_spill(objset_t os, uint64_t object, dmu_tx_t tx)
	{
	dnode_t *dn;
	int error;

	error = dnode_hold(os, object, FTAG, &dn);
	dbuf_rm_spill(dn, tx);
	rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
	dnode_rm_spill(dn, tx);
	rw_exit(&dn->dn_struct_rwlock);
	dnode_rele(dn, FTAG);
	return (error);
	}

	/*
	* Lookup and hold the bonus buffer for the provided dnode. If the dnode
	* has not yet been allocated a new bonus dbuf a will be allocated.
	* Returns ENOENT, EIO, or 0.
	*/
	int dmu_bonus_hold_by_dnode(dnode_t dn, const void tag, dmu_buf_t **dbp,
	uint32_t flags)
	{
	dmu_buf_impl_t *db;
	int error;
	uint32_t db_flags = DB_RF_MUST_SUCCEED;

	if (flags & DMU_READ_NO_PREFETCH)
	db_flags \|= DB_RF_NOPREFETCH;
	if (flags & DMU_READ_NO_DECRYPT)
	db_flags \|= DB_RF_NO_DECRYPT;

	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	if (dn->dn_bonus == NULL) {
	if (!rw_tryupgrade(&dn->dn_struct_rwlock)) {
	rw_exit(&dn->dn_struct_rwlock);
	rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
	}
	if (dn->dn_bonus == NULL)
	dbuf_create_bonus(dn);
	}
	db = dn->dn_bonus;

	/* as long as the bonus buf is held, the dnode will be held */
	if (zfs_refcount_add(&db->db_holds, tag) == 1) {
	VERIFY(dnode_add_ref(dn, db));
	atomic_inc_32(&dn->dn_dbufs_count);
	}

	/*
	* Wait to drop dn_struct_rwlock until after adding the bonus dbuf's
	* hold and incrementing the dbuf count to ensure that dnode_move() sees
	* a dnode hold for every dbuf.
	*/
	rw_exit(&dn->dn_struct_rwlock);

	error = dbuf_read(db, NULL, db_flags);
	if (error) {
	dnode_evict_bonus(dn);
	dbuf_rele(db, tag);
	*dbp = NULL;
	return (error);
	}

	*dbp = &db->db;
	return (0);
	}

	int
	dmu_bonus_hold(objset_t os, uint64_t object, const void tag, dmu_buf_t **dbp)
	{
	dnode_t *dn;
	int error;

	error = dnode_hold(os, object, FTAG, &dn);
	if (error)
	return (error);

	error = dmu_bonus_hold_by_dnode(dn, tag, dbp, DMU_READ_NO_PREFETCH);
	dnode_rele(dn, FTAG);

	return (error);
	}

	/*
	* returns ENOENT, EIO, or 0.
	*
	* This interface will allocate a blank spill dbuf when a spill blk
	* doesn't already exist on the dnode.
	*
	* if you only want to find an already existing spill db, then
	* dmu_spill_hold_existing() should be used.
	*/
	int
	dmu_spill_hold_by_dnode(dnode_t dn, uint32_t flags, const void tag,
	dmu_buf_t **dbp)
	{
	dmu_buf_impl_t *db = NULL;
	int err;

	if ((flags & DB_RF_HAVESTRUCT) == 0)
	rw_enter(&dn->dn_struct_rwlock, RW_READER);

	db = dbuf_hold(dn, DMU_SPILL_BLKID, tag);

	if ((flags & DB_RF_HAVESTRUCT) == 0)
	rw_exit(&dn->dn_struct_rwlock);

	if (db == NULL) {
	*dbp = NULL;
	return (SET_ERROR(EIO));
	}
	err = dbuf_read(db, NULL, flags);
	if (err == 0)
	*dbp = &db->db;
	else {
	dbuf_rele(db, tag);
	*dbp = NULL;
	}
	return (err);
	}

	int
	dmu_spill_hold_existing(dmu_buf_t bonus, const void tag, dmu_buf_t **dbp)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )bonus;
	dnode_t *dn;
	int err;

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);

	if (spa_version(dn->dn_objset->os_spa) < SPA_VERSION_SA) {
	err = SET_ERROR(EINVAL);
	} else {
	rw_enter(&dn->dn_struct_rwlock, RW_READER);

	if (!dn->dn_have_spill) {
	err = SET_ERROR(ENOENT);
	} else {
	err = dmu_spill_hold_by_dnode(dn,
	DB_RF_HAVESTRUCT \| DB_RF_CANFAIL, tag, dbp);
	}

	rw_exit(&dn->dn_struct_rwlock);
	}

	DB_DNODE_EXIT(db);
	return (err);
	}

	int
	dmu_spill_hold_by_bonus(dmu_buf_t bonus, uint32_t flags, const void tag,
	dmu_buf_t **dbp)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )bonus;
	dnode_t *dn;
	int err;
	uint32_t db_flags = DB_RF_CANFAIL;

	if (flags & DMU_READ_NO_DECRYPT)
	db_flags \|= DB_RF_NO_DECRYPT;

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	err = dmu_spill_hold_by_dnode(dn, db_flags, tag, dbp);
	DB_DNODE_EXIT(db);

	return (err);
	}

	/*
	* Note: longer-term, we should modify all of the dmu_buf_*() interfaces
	* to take a held dnode rather than <os, object> -- the lookup is wasteful,
	* and can induce severe lock contention when writing to several files
	* whose dnodes are in the same block.
	*/
	int
	dmu_buf_hold_array_by_dnode(dnode_t *dn, uint64_t offset, uint64_t length,
	boolean_t read, const void tag, int numbufsp, dmu_buf_t ***dbpp,
	uint32_t flags)
	{
	dmu_buf_t **dbp;
	zstream_t *zs = NULL;
	uint64_t blkid, nblks, i;
	uint32_t dbuf_flags;
	int err;
	zio_t *zio = NULL;
	boolean_t missed = B_FALSE;

	ASSERT(!read \|\| length <= DMU_MAX_ACCESS);

	/*
	* Note: We directly notify the prefetch code of this read, so that
	* we can tell it about the multi-block read. dbuf_read() only knows
	* about the one block it is accessing.
	*/
	dbuf_flags = DB_RF_CANFAIL \| DB_RF_NEVERWAIT \| DB_RF_HAVESTRUCT \|
	DB_RF_NOPREFETCH;

	if ((flags & DMU_READ_NO_DECRYPT) != 0)
	dbuf_flags \|= DB_RF_NO_DECRYPT;

	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	if (dn->dn_datablkshift) {
	int blkshift = dn->dn_datablkshift;
	nblks = (P2ROUNDUP(offset + length, 1ULL << blkshift) -
	- P2ALIGN(offset, 1ULL << blkshift)) >> blkshift;
	+ P2ALIGN_TYPED(offset, 1ULL << blkshift, uint64_t))
	+ >> blkshift;
	} else {
	if (offset + length > dn->dn_datablksz) {
	zfs_panic_recover("zfs: accessing past end of object "
	"%llx/%llx (size=%u access=%llu+%llu)",
	(longlong_t)dn->dn_objset->
	os_dsl_dataset->ds_object,
	(longlong_t)dn->dn_object, dn->dn_datablksz,
	(longlong_t)offset, (longlong_t)length);
	rw_exit(&dn->dn_struct_rwlock);
	return (SET_ERROR(EIO));
	}
	nblks = 1;
	}
	dbp = kmem_zalloc(sizeof (dmu_buf_t ) nblks, KM_SLEEP);

	if (read)
	zio = zio_root(dn->dn_objset->os_spa, NULL, NULL,
	ZIO_FLAG_CANFAIL);
	blkid = dbuf_whichblock(dn, 0, offset);
	if ((flags & DMU_READ_NO_PREFETCH) == 0) {
	/*
	* Prepare the zfetch before initiating the demand reads, so
	* that if multiple threads block on same indirect block, we
	* base predictions on the original less racy request order.
	*/
	zs = dmu_zfetch_prepare(&dn->dn_zfetch, blkid, nblks, read,
	B_TRUE);
	}
	for (i = 0; i < nblks; i++) {
	dmu_buf_impl_t *db = dbuf_hold(dn, blkid + i, tag);
	if (db == NULL) {
	if (zs) {
	dmu_zfetch_run(&dn->dn_zfetch, zs, missed,
	B_TRUE);
	}
	rw_exit(&dn->dn_struct_rwlock);
	dmu_buf_rele_array(dbp, nblks, tag);
	if (read)
	zio_nowait(zio);
	return (SET_ERROR(EIO));
	}

	/*
	* Initiate async demand data read.
	* We check the db_state after calling dbuf_read() because
	* (1) dbuf_read() may change the state to CACHED due to a
	* hit in the ARC, and (2) on a cache miss, a child will
	* have been added to "zio" but not yet completed, so the
	* state will not yet be CACHED.
	*/
	if (read) {
	if (i == nblks - 1 && blkid + i < dn->dn_maxblkid &&
	offset + length < db->db.db_offset +
	db->db.db_size) {
	if (offset <= db->db.db_offset)
	dbuf_flags \|= DB_RF_PARTIAL_FIRST;
	else
	dbuf_flags \|= DB_RF_PARTIAL_MORE;
	}
	(void) dbuf_read(db, zio, dbuf_flags);
	if (db->db_state != DB_CACHED)
	missed = B_TRUE;
	}
	dbp[i] = &db->db;
	}

	if (!read)
	zfs_racct_write(length, nblks);

	if (zs)
	dmu_zfetch_run(&dn->dn_zfetch, zs, missed, B_TRUE);
	rw_exit(&dn->dn_struct_rwlock);

	if (read) {
	/* wait for async read i/o */
	err = zio_wait(zio);
	if (err) {
	dmu_buf_rele_array(dbp, nblks, tag);
	return (err);
	}

	/* wait for other io to complete */
	for (i = 0; i < nblks; i++) {
	dmu_buf_impl_t db = (dmu_buf_impl_t )dbp[i];
	mutex_enter(&db->db_mtx);
	while (db->db_state == DB_READ \|\|
	db->db_state == DB_FILL)
	cv_wait(&db->db_changed, &db->db_mtx);
	if (db->db_state == DB_UNCACHED)
	err = SET_ERROR(EIO);
	mutex_exit(&db->db_mtx);
	if (err) {
	dmu_buf_rele_array(dbp, nblks, tag);
	return (err);
	}
	}
	}

	*numbufsp = nblks;
	*dbpp = dbp;
	return (0);
	}

	int
	dmu_buf_hold_array(objset_t *os, uint64_t object, uint64_t offset,
	uint64_t length, int read, const void tag, int numbufsp,
	dmu_buf_t ***dbpp)
	{
	dnode_t *dn;
	int err;

	err = dnode_hold(os, object, FTAG, &dn);
	if (err)
	return (err);

	err = dmu_buf_hold_array_by_dnode(dn, offset, length, read, tag,
	numbufsp, dbpp, DMU_READ_PREFETCH);

	dnode_rele(dn, FTAG);

	return (err);
	}

	int
	dmu_buf_hold_array_by_bonus(dmu_buf_t *db_fake, uint64_t offset,
	uint64_t length, boolean_t read, const void tag, int numbufsp,
	dmu_buf_t ***dbpp)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;
	dnode_t *dn;
	int err;

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	err = dmu_buf_hold_array_by_dnode(dn, offset, length, read, tag,
	numbufsp, dbpp, DMU_READ_PREFETCH);
	DB_DNODE_EXIT(db);

	return (err);
	}

	void
	dmu_buf_rele_array(dmu_buf_t *dbp_fake, int numbufs, const void tag)
	{
	int i;
	dmu_buf_impl_t dbp = (dmu_buf_impl_t )dbp_fake;

	if (numbufs == 0)
	return;

	for (i = 0; i < numbufs; i++) {
	if (dbp[i])
	dbuf_rele(dbp[i], tag);
	}

	kmem_free(dbp, sizeof (dmu_buf_t ) numbufs);
	}

	/*
	* Issue prefetch I/Os for the given blocks. If level is greater than 0, the
	* indirect blocks prefetched will be those that point to the blocks containing
	* the data starting at offset, and continuing to offset + len. If the range
	* it too long, prefetch the first dmu_prefetch_max bytes as requested, while
	* for the rest only a higher level, also fitting within dmu_prefetch_max. It
	* should primarily help random reads, since for long sequential reads there is
	* a speculative prefetcher.
	*
	* Note that if the indirect blocks above the blocks being prefetched are not
	* in cache, they will be asynchronously read in. Dnode read by dnode_hold()
	* is currently synchronous.
	*/
	void
	dmu_prefetch(objset_t *os, uint64_t object, int64_t level, uint64_t offset,
	uint64_t len, zio_priority_t pri)
	{
	dnode_t *dn;

	if (dmu_prefetch_max == 0 \|\| len == 0) {
	dmu_prefetch_dnode(os, object, pri);
	return;
	}

	if (dnode_hold(os, object, FTAG, &dn) != 0)
	return;

	dmu_prefetch_by_dnode(dn, level, offset, len, pri);

	dnode_rele(dn, FTAG);
	}

	void
	dmu_prefetch_by_dnode(dnode_t *dn, int64_t level, uint64_t offset,
	uint64_t len, zio_priority_t pri)
	{
	int64_t level2 = level;
	uint64_t start, end, start2, end2;

	/*
	* Depending on len we may do two prefetches: blocks [start, end) at
	* level, and following blocks [start2, end2) at higher level2.
	*/
	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	if (dn->dn_datablkshift != 0) {
	/*
	* The object has multiple blocks. Calculate the full range
	* of blocks [start, end2) and then split it into two parts,
	* so that the first [start, end) fits into dmu_prefetch_max.
	*/
	start = dbuf_whichblock(dn, level, offset);
	end2 = dbuf_whichblock(dn, level, offset + len - 1) + 1;
	uint8_t ibs = dn->dn_indblkshift;
	uint8_t bs = (level == 0) ? dn->dn_datablkshift : ibs;
	uint_t limit = P2ROUNDUP(dmu_prefetch_max, 1 << bs) >> bs;
	start2 = end = MIN(end2, start + limit);

	/*
	* Find level2 where [start2, end2) fits into dmu_prefetch_max.
	*/
	uint8_t ibps = ibs - SPA_BLKPTRSHIFT;
	limit = P2ROUNDUP(dmu_prefetch_max, 1 << ibs) >> ibs;
	do {
	level2++;
	start2 = P2ROUNDUP(start2, 1 << ibps) >> ibps;
	end2 = P2ROUNDUP(end2, 1 << ibps) >> ibps;
	} while (end2 - start2 > limit);
	} else {
	/* There is only one block. Prefetch it or nothing. */
	start = start2 = end2 = 0;
	end = start + (level == 0 && offset < dn->dn_datablksz);
	}

	for (uint64_t i = start; i < end; i++)
	dbuf_prefetch(dn, level, i, pri, 0);
	for (uint64_t i = start2; i < end2; i++)
	dbuf_prefetch(dn, level2, i, pri, 0);
	rw_exit(&dn->dn_struct_rwlock);
	}

	/*
	* Issue prefetch I/Os for the given object's dnode.
	*/
	void
	dmu_prefetch_dnode(objset_t *os, uint64_t object, zio_priority_t pri)
	{
	if (object == 0 \|\| object >= DN_MAX_OBJECT)
	return;

	dnode_t *dn = DMU_META_DNODE(os);
	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	uint64_t blkid = dbuf_whichblock(dn, 0, object * sizeof (dnode_phys_t));
	dbuf_prefetch(dn, 0, blkid, pri, 0);
	rw_exit(&dn->dn_struct_rwlock);
	}

	/*
	* Get the next "chunk" of file data to free. We traverse the file from
	* the end so that the file gets shorter over time (if we crashes in the
	* middle, this will leave us in a better state). We find allocated file
	* data by simply searching the allocated level 1 indirects.
	*
	* On input, *start should be the first offset that does not need to be
	* freed (e.g. "offset + length"). On return, *start will be the first
	* offset that should be freed and l1blks is set to the number of level 1
	* indirect blocks found within the chunk.
	*/
	static int
	get_next_chunk(dnode_t dn, uint64_t start, uint64_t minimum, uint64_t *l1blks)
	{
	uint64_t blks;
	uint64_t maxblks = DMU_MAX_ACCESS >> (dn->dn_indblkshift + 1);
	/* bytes of data covered by a level-1 indirect block */
	uint64_t iblkrange = (uint64_t)dn->dn_datablksz *
	EPB(dn->dn_indblkshift, SPA_BLKPTRSHIFT);

	ASSERT3U(minimum, <=, *start);

	/*
	* Check if we can free the entire range assuming that all of the
	* L1 blocks in this range have data. If we can, we use this
	* worst case value as an estimate so we can avoid having to look
	* at the object's actual data.
	*/
	uint64_t total_l1blks =
	(roundup(start, iblkrange) - (minimum / iblkrange iblkrange)) /
	iblkrange;
	if (total_l1blks <= maxblks) {
	*l1blks = total_l1blks;
	*start = minimum;
	return (0);
	}
	ASSERT(ISP2(iblkrange));

	for (blks = 0; *start > minimum && blks < maxblks; blks++) {
	int err;

	/*
	* dnode_next_offset(BACKWARDS) will find an allocated L1
	* indirect block at or before the input offset. We must
	* decrement *start so that it is at the end of the region
	* to search.
	*/
	(*start)--;

	err = dnode_next_offset(dn,
	DNODE_FIND_BACKWARDS, start, 2, 1, 0);

	/* if there are no indirect blocks before start, we are done */
	if (err == ESRCH) {
	*start = minimum;
	break;
	} else if (err != 0) {
	*l1blks = blks;
	return (err);
	}

	/* set start to the beginning of this L1 indirect */
	- start = P2ALIGN(start, iblkrange);
	+ start = P2ALIGN_TYPED(start, iblkrange, uint64_t);
	}
	if (*start < minimum)
	*start = minimum;
	*l1blks = blks;

	return (0);
	}

	/*
	* If this objset is of type OST_ZFS return true if vfs's unmounted flag is set,
	* otherwise return false.
	* Used below in dmu_free_long_range_impl() to enable abort when unmounting
	*/
	static boolean_t
	dmu_objset_zfs_unmounting(objset_t *os)
	{
	#ifdef _KERNEL
	if (dmu_objset_type(os) == DMU_OST_ZFS)
	return (zfs_get_vfs_flag_unmounted(os));
	#else
	(void) os;
	#endif
	return (B_FALSE);
	}

	static int
	dmu_free_long_range_impl(objset_t os, dnode_t dn, uint64_t offset,
	uint64_t length)
	{
	uint64_t object_size;
	int err;
	uint64_t dirty_frees_threshold;
	dsl_pool_t *dp = dmu_objset_pool(os);

	if (dn == NULL)
	return (SET_ERROR(EINVAL));

	object_size = (dn->dn_maxblkid + 1) * dn->dn_datablksz;
	if (offset >= object_size)
	return (0);

	if (zfs_per_txg_dirty_frees_percent <= 100)
	dirty_frees_threshold =
	zfs_per_txg_dirty_frees_percent * zfs_dirty_data_max / 100;
	else
	dirty_frees_threshold = zfs_dirty_data_max / 20;

	if (length == DMU_OBJECT_END \|\| offset + length > object_size)
	length = object_size - offset;

	while (length != 0) {
	uint64_t chunk_end, chunk_begin, chunk_len;
	uint64_t l1blks;
	dmu_tx_t *tx;

	if (dmu_objset_zfs_unmounting(dn->dn_objset))
	return (SET_ERROR(EINTR));

	chunk_end = chunk_begin = offset + length;

	/* move chunk_begin backwards to the beginning of this chunk */
	err = get_next_chunk(dn, &chunk_begin, offset, &l1blks);
	if (err)
	return (err);
	ASSERT3U(chunk_begin, >=, offset);
	ASSERT3U(chunk_begin, <=, chunk_end);

	chunk_len = chunk_end - chunk_begin;

	tx = dmu_tx_create(os);
	dmu_tx_hold_free(tx, dn->dn_object, chunk_begin, chunk_len);

	/*
	* Mark this transaction as typically resulting in a net
	* reduction in space used.
	*/
	dmu_tx_mark_netfree(tx);
	err = dmu_tx_assign(tx, TXG_WAIT);
	if (err) {
	dmu_tx_abort(tx);
	return (err);
	}

	uint64_t txg = dmu_tx_get_txg(tx);

	mutex_enter(&dp->dp_lock);
	uint64_t long_free_dirty =
	dp->dp_long_free_dirty_pertxg[txg & TXG_MASK];
	mutex_exit(&dp->dp_lock);

	/*
	* To avoid filling up a TXG with just frees, wait for
	* the next TXG to open before freeing more chunks if
	* we have reached the threshold of frees.
	*/
	if (dirty_frees_threshold != 0 &&
	long_free_dirty >= dirty_frees_threshold) {
	DMU_TX_STAT_BUMP(dmu_tx_dirty_frees_delay);
	dmu_tx_commit(tx);
	txg_wait_open(dp, 0, B_TRUE);
	continue;
	}

	/*
	* In order to prevent unnecessary write throttling, for each
	* TXG, we track the cumulative size of L1 blocks being dirtied
	* in dnode_free_range() below. We compare this number to a
	* tunable threshold, past which we prevent new L1 dirty freeing
	* blocks from being added into the open TXG. See
	* dmu_free_long_range_impl() for details. The threshold
	* prevents write throttle activation due to dirty freeing L1
	* blocks taking up a large percentage of zfs_dirty_data_max.
	*/
	mutex_enter(&dp->dp_lock);
	dp->dp_long_free_dirty_pertxg[txg & TXG_MASK] +=
	l1blks << dn->dn_indblkshift;
	mutex_exit(&dp->dp_lock);
	DTRACE_PROBE3(free__long__range,
	uint64_t, long_free_dirty, uint64_t, chunk_len,
	uint64_t, txg);
	dnode_free_range(dn, chunk_begin, chunk_len, tx);

	dmu_tx_commit(tx);

	length -= chunk_len;
	}
	return (0);
	}

	int
	dmu_free_long_range(objset_t *os, uint64_t object,
	uint64_t offset, uint64_t length)
	{
	dnode_t *dn;
	int err;

	err = dnode_hold(os, object, FTAG, &dn);
	if (err != 0)
	return (err);
	err = dmu_free_long_range_impl(os, dn, offset, length);

	/*
	* It is important to zero out the maxblkid when freeing the entire
	* file, so that (a) subsequent calls to dmu_free_long_range_impl()
	* will take the fast path, and (b) dnode_reallocate() can verify
	* that the entire file has been freed.
	*/
	if (err == 0 && offset == 0 && length == DMU_OBJECT_END)
	dn->dn_maxblkid = 0;

	dnode_rele(dn, FTAG);
	return (err);
	}

	int
	dmu_free_long_object(objset_t *os, uint64_t object)
	{
	dmu_tx_t *tx;
	int err;

	err = dmu_free_long_range(os, object, 0, DMU_OBJECT_END);
	if (err != 0)
	return (err);

	tx = dmu_tx_create(os);
	dmu_tx_hold_bonus(tx, object);
	dmu_tx_hold_free(tx, object, 0, DMU_OBJECT_END);
	dmu_tx_mark_netfree(tx);
	err = dmu_tx_assign(tx, TXG_WAIT);
	if (err == 0) {
	err = dmu_object_free(os, object, tx);
	dmu_tx_commit(tx);
	} else {
	dmu_tx_abort(tx);
	}

	return (err);
	}

	int
	dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
	uint64_t size, dmu_tx_t *tx)
	{
	dnode_t *dn;
	int err = dnode_hold(os, object, FTAG, &dn);
	if (err)
	return (err);
	ASSERT(offset < UINT64_MAX);
	ASSERT(size == DMU_OBJECT_END \|\| size <= UINT64_MAX - offset);
	dnode_free_range(dn, offset, size, tx);
	dnode_rele(dn, FTAG);
	return (0);
	}

	static int
	dmu_read_impl(dnode_t *dn, uint64_t offset, uint64_t size,
	void *buf, uint32_t flags)
	{
	dmu_buf_t **dbp;
	int numbufs, err = 0;

	/*
	* Deal with odd block sizes, where there can't be data past the first
	* block. If we ever do the tail block optimization, we will need to
	* handle that here as well.
	*/
	if (dn->dn_maxblkid == 0) {
	uint64_t newsz = offset > dn->dn_datablksz ? 0 :
	MIN(size, dn->dn_datablksz - offset);
	memset((char *)buf + newsz, 0, size - newsz);
	size = newsz;
	}

	while (size > 0) {
	uint64_t mylen = MIN(size, DMU_MAX_ACCESS / 2);
	int i;

	/*
	* NB: we could do this block-at-a-time, but it's nice
	* to be reading in parallel.
	*/
	err = dmu_buf_hold_array_by_dnode(dn, offset, mylen,
	TRUE, FTAG, &numbufs, &dbp, flags);
	if (err)
	break;

	for (i = 0; i < numbufs; i++) {
	uint64_t tocpy;
	int64_t bufoff;
	dmu_buf_t *db = dbp[i];

	ASSERT(size > 0);

	bufoff = offset - db->db_offset;
	tocpy = MIN(db->db_size - bufoff, size);

	(void) memcpy(buf, (char *)db->db_data + bufoff, tocpy);

	offset += tocpy;
	size -= tocpy;
	buf = (char *)buf + tocpy;
	}
	dmu_buf_rele_array(dbp, numbufs, FTAG);
	}
	return (err);
	}

	int
	dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
	void *buf, uint32_t flags)
	{
	dnode_t *dn;
	int err;

	err = dnode_hold(os, object, FTAG, &dn);
	if (err != 0)
	return (err);

	err = dmu_read_impl(dn, offset, size, buf, flags);
	dnode_rele(dn, FTAG);
	return (err);
	}

	int
	dmu_read_by_dnode(dnode_t dn, uint64_t offset, uint64_t size, void buf,
	uint32_t flags)
	{
	return (dmu_read_impl(dn, offset, size, buf, flags));
	}

	static void
	dmu_write_impl(dmu_buf_t **dbp, int numbufs, uint64_t offset, uint64_t size,
	const void buf, dmu_tx_t tx)
	{
	int i;

	for (i = 0; i < numbufs; i++) {
	uint64_t tocpy;
	int64_t bufoff;
	dmu_buf_t *db = dbp[i];

	ASSERT(size > 0);

	bufoff = offset - db->db_offset;
	tocpy = MIN(db->db_size - bufoff, size);

	ASSERT(i == 0 \|\| i == numbufs-1 \|\| tocpy == db->db_size);

	if (tocpy == db->db_size)
	dmu_buf_will_fill(db, tx, B_FALSE);
	else
	dmu_buf_will_dirty(db, tx);

	(void) memcpy((char *)db->db_data + bufoff, buf, tocpy);

	if (tocpy == db->db_size)
	dmu_buf_fill_done(db, tx, B_FALSE);

	offset += tocpy;
	size -= tocpy;
	buf = (char *)buf + tocpy;
	}
	}

	void
	dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
	const void buf, dmu_tx_t tx)
	{
	dmu_buf_t **dbp;
	int numbufs;

	if (size == 0)
	return;

	VERIFY0(dmu_buf_hold_array(os, object, offset, size,
	FALSE, FTAG, &numbufs, &dbp));
	dmu_write_impl(dbp, numbufs, offset, size, buf, tx);
	dmu_buf_rele_array(dbp, numbufs, FTAG);
	}

	/*
	* Note: Lustre is an external consumer of this interface.
	*/
	void
	dmu_write_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size,
	const void buf, dmu_tx_t tx)
	{
	dmu_buf_t **dbp;
	int numbufs;

	if (size == 0)
	return;

	VERIFY0(dmu_buf_hold_array_by_dnode(dn, offset, size,
	FALSE, FTAG, &numbufs, &dbp, DMU_READ_PREFETCH));
	dmu_write_impl(dbp, numbufs, offset, size, buf, tx);
	dmu_buf_rele_array(dbp, numbufs, FTAG);
	}

	void
	dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
	dmu_tx_t *tx)
	{
	dmu_buf_t **dbp;
	int numbufs, i;

	if (size == 0)
	return;

	VERIFY(0 == dmu_buf_hold_array(os, object, offset, size,
	FALSE, FTAG, &numbufs, &dbp));

	for (i = 0; i < numbufs; i++) {
	dmu_buf_t *db = dbp[i];

	dmu_buf_will_not_fill(db, tx);
	}
	dmu_buf_rele_array(dbp, numbufs, FTAG);
	}

	void
	dmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset,
	void *data, uint8_t etype, uint8_t comp, int uncompressed_size,
	int compressed_size, int byteorder, dmu_tx_t *tx)
	{
	dmu_buf_t *db;

	ASSERT3U(etype, <, NUM_BP_EMBEDDED_TYPES);
	ASSERT3U(comp, <, ZIO_COMPRESS_FUNCTIONS);
	VERIFY0(dmu_buf_hold_noread(os, object, offset,
	FTAG, &db));

	dmu_buf_write_embedded(db,
	data, (bp_embedded_type_t)etype, (enum zio_compress)comp,
	uncompressed_size, compressed_size, byteorder, tx);

	dmu_buf_rele(db, FTAG);
	}

	void
	dmu_redact(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
	dmu_tx_t *tx)
	{
	int numbufs, i;
	dmu_buf_t **dbp;

	VERIFY0(dmu_buf_hold_array(os, object, offset, size, FALSE, FTAG,
	&numbufs, &dbp));
	for (i = 0; i < numbufs; i++)
	dmu_buf_redact(dbp[i], tx);
	dmu_buf_rele_array(dbp, numbufs, FTAG);
	}

	#ifdef _KERNEL
	int
	dmu_read_uio_dnode(dnode_t dn, zfs_uio_t uio, uint64_t size)
	{
	dmu_buf_t **dbp;
	int numbufs, i, err;

	/*
	* NB: we could do this block-at-a-time, but it's nice
	* to be reading in parallel.
	*/
	err = dmu_buf_hold_array_by_dnode(dn, zfs_uio_offset(uio), size,
	TRUE, FTAG, &numbufs, &dbp, 0);
	if (err)
	return (err);

	for (i = 0; i < numbufs; i++) {
	uint64_t tocpy;
	int64_t bufoff;
	dmu_buf_t *db = dbp[i];

	ASSERT(size > 0);

	bufoff = zfs_uio_offset(uio) - db->db_offset;
	tocpy = MIN(db->db_size - bufoff, size);

	err = zfs_uio_fault_move((char *)db->db_data + bufoff, tocpy,
	UIO_READ, uio);

	if (err)
	break;

	size -= tocpy;
	}
	dmu_buf_rele_array(dbp, numbufs, FTAG);

	return (err);
	}

	/*
	* Read 'size' bytes into the uio buffer.
	* From object zdb->db_object.
	* Starting at zfs_uio_offset(uio).
	*
	* If the caller already has a dbuf in the target object
	* (e.g. its bonus buffer), this routine is faster than dmu_read_uio(),
	* because we don't have to find the dnode_t for the object.
	*/
	int
	dmu_read_uio_dbuf(dmu_buf_t zdb, zfs_uio_t uio, uint64_t size)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )zdb;
	dnode_t *dn;
	int err;

	if (size == 0)
	return (0);

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	err = dmu_read_uio_dnode(dn, uio, size);
	DB_DNODE_EXIT(db);

	return (err);
	}

	/*
	* Read 'size' bytes into the uio buffer.
	* From the specified object
	* Starting at offset zfs_uio_offset(uio).
	*/
	int
	dmu_read_uio(objset_t os, uint64_t object, zfs_uio_t uio, uint64_t size)
	{
	dnode_t *dn;
	int err;

	if (size == 0)
	return (0);

	err = dnode_hold(os, object, FTAG, &dn);
	if (err)
	return (err);

	err = dmu_read_uio_dnode(dn, uio, size);

	dnode_rele(dn, FTAG);

	return (err);
	}

	int
	dmu_write_uio_dnode(dnode_t dn, zfs_uio_t uio, uint64_t size, dmu_tx_t *tx)
	{
	dmu_buf_t **dbp;
	int numbufs;
	int err = 0;
	int i;

	err = dmu_buf_hold_array_by_dnode(dn, zfs_uio_offset(uio), size,
	FALSE, FTAG, &numbufs, &dbp, DMU_READ_PREFETCH);
	if (err)
	return (err);

	for (i = 0; i < numbufs; i++) {
	uint64_t tocpy;
	int64_t bufoff;
	dmu_buf_t *db = dbp[i];

	ASSERT(size > 0);

	offset_t off = zfs_uio_offset(uio);
	bufoff = off - db->db_offset;
	tocpy = MIN(db->db_size - bufoff, size);

	ASSERT(i == 0 \|\| i == numbufs-1 \|\| tocpy == db->db_size);

	if (tocpy == db->db_size)
	dmu_buf_will_fill(db, tx, B_TRUE);
	else
	dmu_buf_will_dirty(db, tx);

	err = zfs_uio_fault_move((char *)db->db_data + bufoff,
	tocpy, UIO_WRITE, uio);

	if (tocpy == db->db_size && dmu_buf_fill_done(db, tx, err)) {
	/* The fill was reverted. Undo any uio progress. */
	zfs_uio_advance(uio, off - zfs_uio_offset(uio));
	}

	if (err)
	break;

	size -= tocpy;
	}

	dmu_buf_rele_array(dbp, numbufs, FTAG);
	return (err);
	}

	/*
	* Write 'size' bytes from the uio buffer.
	* To object zdb->db_object.
	* Starting at offset zfs_uio_offset(uio).
	*
	* If the caller already has a dbuf in the target object
	* (e.g. its bonus buffer), this routine is faster than dmu_write_uio(),
	* because we don't have to find the dnode_t for the object.
	*/
	int
	dmu_write_uio_dbuf(dmu_buf_t zdb, zfs_uio_t uio, uint64_t size,
	dmu_tx_t *tx)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )zdb;
	dnode_t *dn;
	int err;

	if (size == 0)
	return (0);

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	err = dmu_write_uio_dnode(dn, uio, size, tx);
	DB_DNODE_EXIT(db);

	return (err);
	}

	/*
	* Write 'size' bytes from the uio buffer.
	* To the specified object.
	* Starting at offset zfs_uio_offset(uio).
	*/
	int
	dmu_write_uio(objset_t os, uint64_t object, zfs_uio_t uio, uint64_t size,
	dmu_tx_t *tx)
	{
	dnode_t *dn;
	int err;

	if (size == 0)
	return (0);

	err = dnode_hold(os, object, FTAG, &dn);
	if (err)
	return (err);

	err = dmu_write_uio_dnode(dn, uio, size, tx);

	dnode_rele(dn, FTAG);

	return (err);
	}
	#endif /* _KERNEL */

	/*
	* Allocate a loaned anonymous arc buffer.
	*/
	arc_buf_t *
	dmu_request_arcbuf(dmu_buf_t *handle, int size)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )handle;

	return (arc_loan_buf(db->db_objset->os_spa, B_FALSE, size));
	}

	/*
	* Free a loaned arc buffer.
	*/
	void
	dmu_return_arcbuf(arc_buf_t *buf)
	{
	arc_return_buf(buf, FTAG);
	arc_buf_destroy(buf, FTAG);
	}

	/*
	* A "lightweight" write is faster than a regular write (e.g.
	* dmu_write_by_dnode() or dmu_assign_arcbuf_by_dnode()), because it avoids the
	* CPU cost of creating a dmu_buf_impl_t and arc_buf_[hdr_]_t. However, the
	* data can not be read or overwritten until the transaction's txg has been
	* synced. This makes it appropriate for workloads that are known to be
	* (temporarily) write-only, like "zfs receive".
	*
	* A single block is written, starting at the specified offset in bytes. If
	* the call is successful, it returns 0 and the provided abd has been
	* consumed (the caller should not free it).
	*/
	int
	dmu_lightweight_write_by_dnode(dnode_t dn, uint64_t offset, abd_t abd,
	const zio_prop_t zp, zio_flag_t flags, dmu_tx_t tx)
	{
	dbuf_dirty_record_t *dr =
	dbuf_dirty_lightweight(dn, dbuf_whichblock(dn, 0, offset), tx);
	if (dr == NULL)
	return (SET_ERROR(EIO));
	dr->dt.dll.dr_abd = abd;
	dr->dt.dll.dr_props = *zp;
	dr->dt.dll.dr_flags = flags;
	return (0);
	}

	/*
	* When possible directly assign passed loaned arc buffer to a dbuf.
	* If this is not possible copy the contents of passed arc buf via
	* dmu_write().
	*/
	int
	dmu_assign_arcbuf_by_dnode(dnode_t dn, uint64_t offset, arc_buf_t buf,
	dmu_tx_t *tx)
	{
	dmu_buf_impl_t *db;
	objset_t *os = dn->dn_objset;
	uint64_t object = dn->dn_object;
	uint32_t blksz = (uint32_t)arc_buf_lsize(buf);
	uint64_t blkid;

	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	blkid = dbuf_whichblock(dn, 0, offset);
	db = dbuf_hold(dn, blkid, FTAG);
	rw_exit(&dn->dn_struct_rwlock);
	if (db == NULL)
	return (SET_ERROR(EIO));

	/*
	* We can only assign if the offset is aligned and the arc buf is the
	* same size as the dbuf.
	*/
	if (offset == db->db.db_offset && blksz == db->db.db_size) {
	zfs_racct_write(blksz, 1);
	dbuf_assign_arcbuf(db, buf, tx);
	dbuf_rele(db, FTAG);
	} else {
	/* compressed bufs must always be assignable to their dbuf */
	ASSERT3U(arc_get_compression(buf), ==, ZIO_COMPRESS_OFF);
	ASSERT(!(buf->b_flags & ARC_BUF_FLAG_COMPRESSED));

	dbuf_rele(db, FTAG);
	dmu_write(os, object, offset, blksz, buf->b_data, tx);
	dmu_return_arcbuf(buf);
	}

	return (0);
	}

	int
	dmu_assign_arcbuf_by_dbuf(dmu_buf_t handle, uint64_t offset, arc_buf_t buf,
	dmu_tx_t *tx)
	{
	int err;
	dmu_buf_impl_t dbuf = (dmu_buf_impl_t )handle;

	DB_DNODE_ENTER(dbuf);
	err = dmu_assign_arcbuf_by_dnode(DB_DNODE(dbuf), offset, buf, tx);
	DB_DNODE_EXIT(dbuf);

	return (err);
	}

	typedef struct {
	dbuf_dirty_record_t *dsa_dr;
	dmu_sync_cb_t *dsa_done;
	zgd_t *dsa_zgd;
	dmu_tx_t *dsa_tx;
	} dmu_sync_arg_t;

	static void
	dmu_sync_ready(zio_t zio, arc_buf_t buf, void *varg)
	{
	(void) buf;
	dmu_sync_arg_t *dsa = varg;
	dmu_buf_t *db = dsa->dsa_zgd->zgd_db;
	blkptr_t *bp = zio->io_bp;

	if (zio->io_error == 0) {
	if (BP_IS_HOLE(bp)) {
	/*
	* A block of zeros may compress to a hole, but the
	* block size still needs to be known for replay.
	*/
	BP_SET_LSIZE(bp, db->db_size);
	} else if (!BP_IS_EMBEDDED(bp)) {
	ASSERT(BP_GET_LEVEL(bp) == 0);
	BP_SET_FILL(bp, 1);
	}
	}
	}

	static void
	dmu_sync_late_arrival_ready(zio_t *zio)
	{
	dmu_sync_ready(zio, NULL, zio->io_private);
	}

	static void
	dmu_sync_done(zio_t zio, arc_buf_t buf, void *varg)
	{
	(void) buf;
	dmu_sync_arg_t *dsa = varg;
	dbuf_dirty_record_t *dr = dsa->dsa_dr;
	dmu_buf_impl_t *db = dr->dr_dbuf;
	zgd_t *zgd = dsa->dsa_zgd;

	/*
	* Record the vdev(s) backing this blkptr so they can be flushed after
	* the writes for the lwb have completed.
	*/
	if (zio->io_error == 0) {
	zil_lwb_add_block(zgd->zgd_lwb, zgd->zgd_bp);
	}

	mutex_enter(&db->db_mtx);
	ASSERT(dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC);
	if (zio->io_error == 0) {
	dr->dt.dl.dr_nopwrite = !!(zio->io_flags & ZIO_FLAG_NOPWRITE);
	if (dr->dt.dl.dr_nopwrite) {
	blkptr_t *bp = zio->io_bp;
	blkptr_t *bp_orig = &zio->io_bp_orig;
	uint8_t chksum = BP_GET_CHECKSUM(bp_orig);

	ASSERT(BP_EQUAL(bp, bp_orig));
	VERIFY(BP_EQUAL(bp, db->db_blkptr));
	ASSERT(zio->io_prop.zp_compress != ZIO_COMPRESS_OFF);
	VERIFY(zio_checksum_table[chksum].ci_flags &
	ZCHECKSUM_FLAG_NOPWRITE);
	}
	dr->dt.dl.dr_overridden_by = *zio->io_bp;
	dr->dt.dl.dr_override_state = DR_OVERRIDDEN;
	dr->dt.dl.dr_copies = zio->io_prop.zp_copies;

	/*
	* Old style holes are filled with all zeros, whereas
	* new-style holes maintain their lsize, type, level,
	* and birth time (see zio_write_compress). While we
	* need to reset the BP_SET_LSIZE() call that happened
	* in dmu_sync_ready for old style holes, we do not
	* want to wipe out the information contained in new
	* style holes. Thus, only zero out the block pointer if
	* it's an old style hole.
	*/
	if (BP_IS_HOLE(&dr->dt.dl.dr_overridden_by) &&
	BP_GET_LOGICAL_BIRTH(&dr->dt.dl.dr_overridden_by) == 0)
	BP_ZERO(&dr->dt.dl.dr_overridden_by);
	} else {
	dr->dt.dl.dr_override_state = DR_NOT_OVERRIDDEN;
	}
	cv_broadcast(&db->db_changed);
	mutex_exit(&db->db_mtx);

	dsa->dsa_done(dsa->dsa_zgd, zio->io_error);

	kmem_free(dsa, sizeof (*dsa));
	}

	static void
	dmu_sync_late_arrival_done(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;
	dmu_sync_arg_t *dsa = zio->io_private;
	zgd_t *zgd = dsa->dsa_zgd;

	if (zio->io_error == 0) {
	/*
	* Record the vdev(s) backing this blkptr so they can be
	* flushed after the writes for the lwb have completed.
	*/
	zil_lwb_add_block(zgd->zgd_lwb, zgd->zgd_bp);

	if (!BP_IS_HOLE(bp)) {
	blkptr_t *bp_orig __maybe_unused = &zio->io_bp_orig;
	ASSERT(!(zio->io_flags & ZIO_FLAG_NOPWRITE));
	ASSERT(BP_IS_HOLE(bp_orig) \|\| !BP_EQUAL(bp, bp_orig));
	ASSERT(BP_GET_LOGICAL_BIRTH(zio->io_bp) == zio->io_txg);
	ASSERT(zio->io_txg > spa_syncing_txg(zio->io_spa));
	zio_free(zio->io_spa, zio->io_txg, zio->io_bp);
	}
	}

	dmu_tx_commit(dsa->dsa_tx);

	dsa->dsa_done(dsa->dsa_zgd, zio->io_error);

	abd_free(zio->io_abd);
	kmem_free(dsa, sizeof (*dsa));
	}

	static int
	dmu_sync_late_arrival(zio_t pio, objset_t os, dmu_sync_cb_t done, zgd_t zgd,
	zio_prop_t zp, zbookmark_phys_t zb)
	{
	dmu_sync_arg_t *dsa;
	dmu_tx_t *tx;
	int error;

	error = dbuf_read((dmu_buf_impl_t *)zgd->zgd_db, NULL,
	DB_RF_CANFAIL \| DB_RF_NOPREFETCH);
	if (error != 0)
	return (error);

	tx = dmu_tx_create(os);
	dmu_tx_hold_space(tx, zgd->zgd_db->db_size);
	/*
	* This transaction does not produce any dirty data or log blocks, so
	* it should not be throttled. All other cases wait for TXG sync, by
	* which time the log block we are writing will be obsolete, so we can
	* skip waiting and just return error here instead.
	*/
	if (dmu_tx_assign(tx, TXG_NOWAIT \| TXG_NOTHROTTLE) != 0) {
	dmu_tx_abort(tx);
	/* Make zl_get_data do txg_waited_synced() */
	return (SET_ERROR(EIO));
	}

	/*
	* In order to prevent the zgd's lwb from being free'd prior to
	* dmu_sync_late_arrival_done() being called, we have to ensure
	* the lwb's "max txg" takes this tx's txg into account.
	*/
	zil_lwb_add_txg(zgd->zgd_lwb, dmu_tx_get_txg(tx));

	dsa = kmem_alloc(sizeof (dmu_sync_arg_t), KM_SLEEP);
	dsa->dsa_dr = NULL;
	dsa->dsa_done = done;
	dsa->dsa_zgd = zgd;
	dsa->dsa_tx = tx;

	/*
	* Since we are currently syncing this txg, it's nontrivial to
	* determine what BP to nopwrite against, so we disable nopwrite.
	*
	* When syncing, the db_blkptr is initially the BP of the previous
	* txg. We can not nopwrite against it because it will be changed
	* (this is similar to the non-late-arrival case where the dbuf is
	* dirty in a future txg).
	*
	* Then dbuf_write_ready() sets bp_blkptr to the location we will write.
	* We can not nopwrite against it because although the BP will not
	* (typically) be changed, the data has not yet been persisted to this
	* location.
	*
	* Finally, when dbuf_write_done() is called, it is theoretically
	* possible to always nopwrite, because the data that was written in
	* this txg is the same data that we are trying to write. However we
	* would need to check that this dbuf is not dirty in any future
	* txg's (as we do in the normal dmu_sync() path). For simplicity, we
	* don't nopwrite in this case.
	*/
	zp->zp_nopwrite = B_FALSE;

	zio_nowait(zio_write(pio, os->os_spa, dmu_tx_get_txg(tx), zgd->zgd_bp,
	abd_get_from_buf(zgd->zgd_db->db_data, zgd->zgd_db->db_size),
	zgd->zgd_db->db_size, zgd->zgd_db->db_size, zp,
	dmu_sync_late_arrival_ready, NULL, dmu_sync_late_arrival_done,
	dsa, ZIO_PRIORITY_SYNC_WRITE, ZIO_FLAG_CANFAIL, zb));

	return (0);
	}

	/*
	* Intent log support: sync the block associated with db to disk.
	* N.B. and XXX: the caller is responsible for making sure that the
	* data isn't changing while dmu_sync() is writing it.
	*
	* Return values:
	*
	* EEXIST: this txg has already been synced, so there's nothing to do.
	* The caller should not log the write.
	*
	* ENOENT: the block was dbuf_free_range()'d, so there's nothing to do.
	* The caller should not log the write.
	*
	* EALREADY: this block is already in the process of being synced.
	* The caller should track its progress (somehow).
	*
	* EIO: could not do the I/O.
	* The caller should do a txg_wait_synced().
	*
	* 0: the I/O has been initiated.
	* The caller should log this blkptr in the done callback.
	* It is possible that the I/O will fail, in which case
	* the error will be reported to the done callback and
	* propagated to pio from zio_done().
	*/
	int
	dmu_sync(zio_t pio, uint64_t txg, dmu_sync_cb_t done, zgd_t *zgd)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )zgd->zgd_db;
	objset_t *os = db->db_objset;
	dsl_dataset_t *ds = os->os_dsl_dataset;
	dbuf_dirty_record_t dr, dr_next;
	dmu_sync_arg_t *dsa;
	zbookmark_phys_t zb;
	zio_prop_t zp;
	dnode_t *dn;

	ASSERT(pio != NULL);
	ASSERT(txg != 0);

	SET_BOOKMARK(&zb, ds->ds_object,
	db->db.db_object, db->db_level, db->db_blkid);

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	dmu_write_policy(os, dn, db->db_level, WP_DMU_SYNC, &zp);
	DB_DNODE_EXIT(db);

	/*
	* If we're frozen (running ziltest), we always need to generate a bp.
	*/
	if (txg > spa_freeze_txg(os->os_spa))
	return (dmu_sync_late_arrival(pio, os, done, zgd, &zp, &zb));

	/*
	* Grabbing db_mtx now provides a barrier between dbuf_sync_leaf()
	* and us. If we determine that this txg is not yet syncing,
	* but it begins to sync a moment later, that's OK because the
	* sync thread will block in dbuf_sync_leaf() until we drop db_mtx.
	*/
	mutex_enter(&db->db_mtx);

	if (txg <= spa_last_synced_txg(os->os_spa)) {
	/*
	* This txg has already synced. There's nothing to do.
	*/
	mutex_exit(&db->db_mtx);
	return (SET_ERROR(EEXIST));
	}

	if (txg <= spa_syncing_txg(os->os_spa)) {
	/*
	* This txg is currently syncing, so we can't mess with
	* the dirty record anymore; just write a new log block.
	*/
	mutex_exit(&db->db_mtx);
	return (dmu_sync_late_arrival(pio, os, done, zgd, &zp, &zb));
	}

	dr = dbuf_find_dirty_eq(db, txg);

	if (dr == NULL) {
	/*
	* There's no dr for this dbuf, so it must have been freed.
	* There's no need to log writes to freed blocks, so we're done.
	*/
	mutex_exit(&db->db_mtx);
	return (SET_ERROR(ENOENT));
	}

	dr_next = list_next(&db->db_dirty_records, dr);
	ASSERT(dr_next == NULL \|\| dr_next->dr_txg < txg);

	if (db->db_blkptr != NULL) {
	/*
	* We need to fill in zgd_bp with the current blkptr so that
	* the nopwrite code can check if we're writing the same
	* data that's already on disk. We can only nopwrite if we
	* are sure that after making the copy, db_blkptr will not
	* change until our i/o completes. We ensure this by
	* holding the db_mtx, and only allowing nopwrite if the
	* block is not already dirty (see below). This is verified
	* by dmu_sync_done(), which VERIFYs that the db_blkptr has
	* not changed.
	*/
	zgd->zgd_bp = db->db_blkptr;
	}

	/*
	* Assume the on-disk data is X, the current syncing data (in
	* txg - 1) is Y, and the current in-memory data is Z (currently
	* in dmu_sync).
	*
	* We usually want to perform a nopwrite if X and Z are the
	* same. However, if Y is different (i.e. the BP is going to
	* change before this write takes effect), then a nopwrite will
	* be incorrect - we would override with X, which could have
	* been freed when Y was written.
	*
	* (Note that this is not a concern when we are nop-writing from
	* syncing context, because X and Y must be identical, because
	* all previous txgs have been synced.)
	*
	* Therefore, we disable nopwrite if the current BP could change
	* before this TXG. There are two ways it could change: by
	* being dirty (dr_next is non-NULL), or by being freed
	* (dnode_block_freed()). This behavior is verified by
	* zio_done(), which VERIFYs that the override BP is identical
	* to the on-disk BP.
	*/
	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	if (dr_next != NULL \|\| dnode_block_freed(dn, db->db_blkid))
	zp.zp_nopwrite = B_FALSE;
	DB_DNODE_EXIT(db);

	ASSERT(dr->dr_txg == txg);
	if (dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC \|\|
	dr->dt.dl.dr_override_state == DR_OVERRIDDEN) {
	/*
	* We have already issued a sync write for this buffer,
	* or this buffer has already been synced. It could not
	* have been dirtied since, or we would have cleared the state.
	*/
	mutex_exit(&db->db_mtx);
	return (SET_ERROR(EALREADY));
	}

	ASSERT(dr->dt.dl.dr_override_state == DR_NOT_OVERRIDDEN);
	dr->dt.dl.dr_override_state = DR_IN_DMU_SYNC;
	mutex_exit(&db->db_mtx);

	dsa = kmem_alloc(sizeof (dmu_sync_arg_t), KM_SLEEP);
	dsa->dsa_dr = dr;
	dsa->dsa_done = done;
	dsa->dsa_zgd = zgd;
	dsa->dsa_tx = NULL;

	zio_nowait(arc_write(pio, os->os_spa, txg, zgd->zgd_bp,
	dr->dt.dl.dr_data, !DBUF_IS_CACHEABLE(db), dbuf_is_l2cacheable(db),
	&zp, dmu_sync_ready, NULL, dmu_sync_done, dsa,
	ZIO_PRIORITY_SYNC_WRITE, ZIO_FLAG_CANFAIL, &zb));

	return (0);
	}

	int
	dmu_object_set_nlevels(objset_t os, uint64_t object, int nlevels, dmu_tx_t tx)
	{
	dnode_t *dn;
	int err;

	err = dnode_hold(os, object, FTAG, &dn);
	if (err)
	return (err);
	err = dnode_set_nlevels(dn, nlevels, tx);
	dnode_rele(dn, FTAG);
	return (err);
	}

	int
	dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size, int ibs,
	dmu_tx_t *tx)
	{
	dnode_t *dn;
	int err;

	err = dnode_hold(os, object, FTAG, &dn);
	if (err)
	return (err);
	err = dnode_set_blksz(dn, size, ibs, tx);
	dnode_rele(dn, FTAG);
	return (err);
	}

	int
	dmu_object_set_maxblkid(objset_t *os, uint64_t object, uint64_t maxblkid,
	dmu_tx_t *tx)
	{
	dnode_t *dn;
	int err;

	err = dnode_hold(os, object, FTAG, &dn);
	if (err)
	return (err);
	rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
	dnode_new_blkid(dn, maxblkid, tx, B_FALSE, B_TRUE);
	rw_exit(&dn->dn_struct_rwlock);
	dnode_rele(dn, FTAG);
	return (0);
	}

	void
	dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
	dmu_tx_t *tx)
	{
	dnode_t *dn;

	/*
	* Send streams include each object's checksum function. This
	* check ensures that the receiving system can understand the
	* checksum function transmitted.
	*/
	ASSERT3U(checksum, <, ZIO_CHECKSUM_LEGACY_FUNCTIONS);

	VERIFY0(dnode_hold(os, object, FTAG, &dn));
	ASSERT3U(checksum, <, ZIO_CHECKSUM_FUNCTIONS);
	dn->dn_checksum = checksum;
	dnode_setdirty(dn, tx);
	dnode_rele(dn, FTAG);
	}

	void
	dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
	dmu_tx_t *tx)
	{
	dnode_t *dn;

	/*
	* Send streams include each object's compression function. This
	* check ensures that the receiving system can understand the
	* compression function transmitted.
	*/
	ASSERT3U(compress, <, ZIO_COMPRESS_LEGACY_FUNCTIONS);

	VERIFY0(dnode_hold(os, object, FTAG, &dn));
	dn->dn_compress = compress;
	dnode_setdirty(dn, tx);
	dnode_rele(dn, FTAG);
	}

	/*
	* When the "redundant_metadata" property is set to "most", only indirect
	* blocks of this level and higher will have an additional ditto block.
	*/
	static const int zfs_redundant_metadata_most_ditto_level = 2;

	void
	dmu_write_policy(objset_t os, dnode_t dn, int level, int wp, zio_prop_t *zp)
	{
	dmu_object_type_t type = dn ? dn->dn_type : DMU_OT_OBJSET;
	boolean_t ismd = (level > 0 \|\| DMU_OT_IS_METADATA(type) \|\|
	(wp & WP_SPILL));
	enum zio_checksum checksum = os->os_checksum;
	enum zio_compress compress = os->os_compress;
	uint8_t complevel = os->os_complevel;
	enum zio_checksum dedup_checksum = os->os_dedup_checksum;
	boolean_t dedup = B_FALSE;
	boolean_t nopwrite = B_FALSE;
	boolean_t dedup_verify = os->os_dedup_verify;
	boolean_t encrypt = B_FALSE;
	int copies = os->os_copies;

	/*
	* We maintain different write policies for each of the following
	* types of data:
	* 1. metadata
	* 2. preallocated blocks (i.e. level-0 blocks of a dump device)
	* 3. all other level 0 blocks
	*/
	if (ismd) {
	/*
	* XXX -- we should design a compression algorithm
	* that specializes in arrays of bps.
	*/
	compress = zio_compress_select(os->os_spa,
	ZIO_COMPRESS_ON, ZIO_COMPRESS_ON);

	/*
	* Metadata always gets checksummed. If the data
	* checksum is multi-bit correctable, and it's not a
	* ZBT-style checksum, then it's suitable for metadata
	* as well. Otherwise, the metadata checksum defaults
	* to fletcher4.
	*/
	if (!(zio_checksum_table[checksum].ci_flags &
	ZCHECKSUM_FLAG_METADATA) \|\|
	(zio_checksum_table[checksum].ci_flags &
	ZCHECKSUM_FLAG_EMBEDDED))
	checksum = ZIO_CHECKSUM_FLETCHER_4;

	switch (os->os_redundant_metadata) {
	case ZFS_REDUNDANT_METADATA_ALL:
	copies++;
	break;
	case ZFS_REDUNDANT_METADATA_MOST:
	if (level >= zfs_redundant_metadata_most_ditto_level \|\|
	DMU_OT_IS_METADATA(type) \|\| (wp & WP_SPILL))
	copies++;
	break;
	case ZFS_REDUNDANT_METADATA_SOME:
	if (DMU_OT_IS_CRITICAL(type))
	copies++;
	break;
	case ZFS_REDUNDANT_METADATA_NONE:
	break;
	}
	} else if (wp & WP_NOFILL) {
	ASSERT(level == 0);

	/*
	* If we're writing preallocated blocks, we aren't actually
	* writing them so don't set any policy properties. These
	* blocks are currently only used by an external subsystem
	* outside of zfs (i.e. dump) and not written by the zio
	* pipeline.
	*/
	compress = ZIO_COMPRESS_OFF;
	checksum = ZIO_CHECKSUM_OFF;
	} else {
	compress = zio_compress_select(os->os_spa, dn->dn_compress,
	compress);
	complevel = zio_complevel_select(os->os_spa, compress,
	complevel, complevel);

	checksum = (dedup_checksum == ZIO_CHECKSUM_OFF) ?
	zio_checksum_select(dn->dn_checksum, checksum) :
	dedup_checksum;

	/*
	* Determine dedup setting. If we are in dmu_sync(),
	* we won't actually dedup now because that's all
	* done in syncing context; but we do want to use the
	* dedup checksum. If the checksum is not strong
	* enough to ensure unique signatures, force
	* dedup_verify.
	*/
	if (dedup_checksum != ZIO_CHECKSUM_OFF) {
	dedup = (wp & WP_DMU_SYNC) ? B_FALSE : B_TRUE;
	if (!(zio_checksum_table[checksum].ci_flags &
	ZCHECKSUM_FLAG_DEDUP))
	dedup_verify = B_TRUE;
	}

	/*
	* Enable nopwrite if we have secure enough checksum
	* algorithm (see comment in zio_nop_write) and
	* compression is enabled. We don't enable nopwrite if
	* dedup is enabled as the two features are mutually
	* exclusive.
	*/
	nopwrite = (!dedup && (zio_checksum_table[checksum].ci_flags &
	ZCHECKSUM_FLAG_NOPWRITE) &&
	compress != ZIO_COMPRESS_OFF && zfs_nopwrite_enabled);
	}

	/*
	* All objects in an encrypted objset are protected from modification
	* via a MAC. Encrypted objects store their IV and salt in the last DVA
	* in the bp, so we cannot use all copies. Encrypted objects are also
	* not subject to nopwrite since writing the same data will still
	* result in a new ciphertext. Only encrypted blocks can be dedup'd
	* to avoid ambiguity in the dedup code since the DDT does not store
	* object types.
	*/
	if (os->os_encrypted && (wp & WP_NOFILL) == 0) {
	encrypt = B_TRUE;

	if (DMU_OT_IS_ENCRYPTED(type)) {
	copies = MIN(copies, SPA_DVAS_PER_BP - 1);
	nopwrite = B_FALSE;
	} else {
	dedup = B_FALSE;
	}

	if (level <= 0 &&
	(type == DMU_OT_DNODE \|\| type == DMU_OT_OBJSET)) {
	compress = ZIO_COMPRESS_EMPTY;
	}
	}

	zp->zp_compress = compress;
	zp->zp_complevel = complevel;
	zp->zp_checksum = checksum;
	zp->zp_type = (wp & WP_SPILL) ? dn->dn_bonustype : type;
	zp->zp_level = level;
	zp->zp_copies = MIN(copies, spa_max_replication(os->os_spa));
	zp->zp_dedup = dedup;
	zp->zp_dedup_verify = dedup && dedup_verify;
	zp->zp_nopwrite = nopwrite;
	zp->zp_encrypt = encrypt;
	zp->zp_byteorder = ZFS_HOST_BYTEORDER;
	memset(zp->zp_salt, 0, ZIO_DATA_SALT_LEN);
	memset(zp->zp_iv, 0, ZIO_DATA_IV_LEN);
	memset(zp->zp_mac, 0, ZIO_DATA_MAC_LEN);
	zp->zp_zpl_smallblk = DMU_OT_IS_FILE(zp->zp_type) ?
	os->os_zpl_special_smallblock : 0;

	ASSERT3U(zp->zp_compress, !=, ZIO_COMPRESS_INHERIT);
	}

	/*
	* Reports the location of data and holes in an object. In order to
	* accurately report holes all dirty data must be synced to disk. This
	* causes extremely poor performance when seeking for holes in a dirty file.
	* As a compromise, only provide hole data when the dnode is clean. When
	* a dnode is dirty report the dnode as having no holes by returning EBUSY
	* which is always safe to do.
	*/
	int
	dmu_offset_next(objset_t os, uint64_t object, boolean_t hole, uint64_t off)
	{
	dnode_t *dn;
	int restarted = 0, err;

	restart:
	err = dnode_hold(os, object, FTAG, &dn);
	if (err)
	return (err);

	rw_enter(&dn->dn_struct_rwlock, RW_READER);

	if (dnode_is_dirty(dn)) {
	/*
	* If the zfs_dmu_offset_next_sync module option is enabled
	* then hole reporting has been requested. Dirty dnodes
	* must be synced to disk to accurately report holes.
	*
	* Provided a RL_READER rangelock spanning 0-UINT64_MAX is
	* held by the caller only a single restart will be required.
	* We tolerate callers which do not hold the rangelock by
	* returning EBUSY and not reporting holes after one restart.
	*/
	if (zfs_dmu_offset_next_sync) {
	rw_exit(&dn->dn_struct_rwlock);
	dnode_rele(dn, FTAG);

	if (restarted)
	return (SET_ERROR(EBUSY));

	txg_wait_synced(dmu_objset_pool(os), 0);
	restarted = 1;
	goto restart;
	}

	err = SET_ERROR(EBUSY);
	} else {
	err = dnode_next_offset(dn, DNODE_FIND_HAVELOCK \|
	(hole ? DNODE_FIND_HOLE : 0), off, 1, 1, 0);
	}

	rw_exit(&dn->dn_struct_rwlock);
	dnode_rele(dn, FTAG);

	return (err);
	}

	int
	dmu_read_l0_bps(objset_t *os, uint64_t object, uint64_t offset, uint64_t length,
	blkptr_t bps, size_t nbpsp)
	{
	dmu_buf_t *dbp, dbuf;
	dmu_buf_impl_t *db;
	blkptr_t *bp;
	int error, numbufs;

	error = dmu_buf_hold_array(os, object, offset, length, FALSE, FTAG,
	&numbufs, &dbp);
	if (error != 0) {
	if (error == ESRCH) {
	error = SET_ERROR(ENXIO);
	}
	return (error);
	}

	ASSERT3U(numbufs, <=, *nbpsp);

	for (int i = 0; i < numbufs; i++) {
	dbuf = dbp[i];
	db = (dmu_buf_impl_t *)dbuf;

	mutex_enter(&db->db_mtx);

	if (!list_is_empty(&db->db_dirty_records)) {
	dbuf_dirty_record_t *dr;

	dr = list_head(&db->db_dirty_records);
	if (dr->dt.dl.dr_brtwrite) {
	/*
	* This is very special case where we clone a
	* block and in the same transaction group we
	* read its BP (most likely to clone the clone).
	*/
	bp = &dr->dt.dl.dr_overridden_by;
	} else {
	/*
	* The block was modified in the same
	* transaction group.
	*/
	mutex_exit(&db->db_mtx);
	error = SET_ERROR(EAGAIN);
	goto out;
	}
	} else {
	bp = db->db_blkptr;
	}

	mutex_exit(&db->db_mtx);

	if (bp == NULL) {
	/*
	* The file size was increased, but the block was never
	* written, otherwise we would either have the block
	* pointer or the dirty record and would not get here.
	* It is effectively a hole, so report it as such.
	*/
	BP_ZERO(&bps[i]);
	continue;
	}
	/*
	* Make sure we clone only data blocks.
	*/
	if (BP_IS_METADATA(bp) && !BP_IS_HOLE(bp)) {
	error = SET_ERROR(EINVAL);
	goto out;
	}

	/*
	* If the block was allocated in transaction group that is not
	* yet synced, we could clone it, but we couldn't write this
	* operation into ZIL, or it may be impossible to replay, since
	* the block may appear not yet allocated at that point.
	*/
	if (BP_GET_BIRTH(bp) > spa_freeze_txg(os->os_spa)) {
	error = SET_ERROR(EINVAL);
	goto out;
	}
	if (BP_GET_BIRTH(bp) > spa_last_synced_txg(os->os_spa)) {
	error = SET_ERROR(EAGAIN);
	goto out;
	}

	bps[i] = *bp;
	}

	*nbpsp = numbufs;
	out:
	dmu_buf_rele_array(dbp, numbufs, FTAG);

	return (error);
	}

	int
	dmu_brt_clone(objset_t *os, uint64_t object, uint64_t offset, uint64_t length,
	dmu_tx_t tx, const blkptr_t bps, size_t nbps)
	{
	spa_t *spa;
	dmu_buf_t *dbp, dbuf;
	dmu_buf_impl_t *db;
	struct dirty_leaf *dl;
	dbuf_dirty_record_t *dr;
	const blkptr_t *bp;
	int error = 0, i, numbufs;

	spa = os->os_spa;

	VERIFY0(dmu_buf_hold_array(os, object, offset, length, FALSE, FTAG,
	&numbufs, &dbp));
	ASSERT3U(nbps, ==, numbufs);

	/*
	* Before we start cloning make sure that the dbufs sizes match new BPs
	* sizes. If they don't, that's a no-go, as we are not able to shrink
	* dbufs.
	*/
	for (i = 0; i < numbufs; i++) {
	dbuf = dbp[i];
	db = (dmu_buf_impl_t *)dbuf;
	bp = &bps[i];

	ASSERT0(db->db_level);
	ASSERT(db->db_blkid != DMU_BONUS_BLKID);
	ASSERT(db->db_blkid != DMU_SPILL_BLKID);

	if (!BP_IS_HOLE(bp) && BP_GET_LSIZE(bp) != dbuf->db_size) {
	error = SET_ERROR(EXDEV);
	goto out;
	}
	}

	for (i = 0; i < numbufs; i++) {
	dbuf = dbp[i];
	db = (dmu_buf_impl_t *)dbuf;
	bp = &bps[i];

	ASSERT0(db->db_level);
	ASSERT(db->db_blkid != DMU_BONUS_BLKID);
	ASSERT(db->db_blkid != DMU_SPILL_BLKID);
	ASSERT(BP_IS_HOLE(bp) \|\| dbuf->db_size == BP_GET_LSIZE(bp));

	dmu_buf_will_clone(dbuf, tx);

	mutex_enter(&db->db_mtx);

	dr = list_head(&db->db_dirty_records);
	VERIFY(dr != NULL);
	ASSERT3U(dr->dr_txg, ==, tx->tx_txg);
	dl = &dr->dt.dl;
	dl->dr_overridden_by = *bp;
	if (!BP_IS_HOLE(bp) \|\| BP_GET_LOGICAL_BIRTH(bp) != 0) {
	if (!BP_IS_EMBEDDED(bp)) {
	BP_SET_BIRTH(&dl->dr_overridden_by, dr->dr_txg,
	BP_GET_BIRTH(bp));
	} else {
	BP_SET_LOGICAL_BIRTH(&dl->dr_overridden_by,
	dr->dr_txg);
	}
	}
	dl->dr_brtwrite = B_TRUE;
	dl->dr_override_state = DR_OVERRIDDEN;

	mutex_exit(&db->db_mtx);

	/*
	* When data in embedded into BP there is no need to create
	* BRT entry as there is no data block. Just copy the BP as
	* it contains the data.
	*/
	if (!BP_IS_HOLE(bp) && !BP_IS_EMBEDDED(bp)) {
	brt_pending_add(spa, bp, tx);
	}
	}
	out:
	dmu_buf_rele_array(dbp, numbufs, FTAG);

	return (error);
	}

	void
	__dmu_object_info_from_dnode(dnode_t dn, dmu_object_info_t doi)
	{
	dnode_phys_t *dnp = dn->dn_phys;

	doi->doi_data_block_size = dn->dn_datablksz;
	doi->doi_metadata_block_size = dn->dn_indblkshift ?
	1ULL << dn->dn_indblkshift : 0;
	doi->doi_type = dn->dn_type;
	doi->doi_bonus_type = dn->dn_bonustype;
	doi->doi_bonus_size = dn->dn_bonuslen;
	doi->doi_dnodesize = dn->dn_num_slots << DNODE_SHIFT;
	doi->doi_indirection = dn->dn_nlevels;
	doi->doi_checksum = dn->dn_checksum;
	doi->doi_compress = dn->dn_compress;
	doi->doi_nblkptr = dn->dn_nblkptr;
	doi->doi_physical_blocks_512 = (DN_USED_BYTES(dnp) + 256) >> 9;
	doi->doi_max_offset = (dn->dn_maxblkid + 1) * dn->dn_datablksz;
	doi->doi_fill_count = 0;
	for (int i = 0; i < dnp->dn_nblkptr; i++)
	doi->doi_fill_count += BP_GET_FILL(&dnp->dn_blkptr[i]);
	}

	void
	dmu_object_info_from_dnode(dnode_t dn, dmu_object_info_t doi)
	{
	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	mutex_enter(&dn->dn_mtx);

	__dmu_object_info_from_dnode(dn, doi);

	mutex_exit(&dn->dn_mtx);
	rw_exit(&dn->dn_struct_rwlock);
	}

	/*
	* Get information on a DMU object.
	* If doi is NULL, just indicates whether the object exists.
	*/
	int
	dmu_object_info(objset_t os, uint64_t object, dmu_object_info_t doi)
	{
	dnode_t *dn;
	int err = dnode_hold(os, object, FTAG, &dn);

	if (err)
	return (err);

	if (doi != NULL)
	dmu_object_info_from_dnode(dn, doi);

	dnode_rele(dn, FTAG);
	return (0);
	}

	/*
	* As above, but faster; can be used when you have a held dbuf in hand.
	*/
	void
	dmu_object_info_from_db(dmu_buf_t db_fake, dmu_object_info_t doi)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;

	DB_DNODE_ENTER(db);
	dmu_object_info_from_dnode(DB_DNODE(db), doi);
	DB_DNODE_EXIT(db);
	}

	/*
	* Faster still when you only care about the size.
	*/
	void
	dmu_object_size_from_db(dmu_buf_t db_fake, uint32_t blksize,
	u_longlong_t *nblk512)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;
	dnode_t *dn;

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);

	*blksize = dn->dn_datablksz;
	/* add in number of slots used for the dnode itself */
	*nblk512 = ((DN_USED_BYTES(dn->dn_phys) + SPA_MINBLOCKSIZE/2) >>
	SPA_MINBLOCKSHIFT) + dn->dn_num_slots;
	DB_DNODE_EXIT(db);
	}

	void
	dmu_object_dnsize_from_db(dmu_buf_t db_fake, int dnsize)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )db_fake;
	dnode_t *dn;

	DB_DNODE_ENTER(db);
	dn = DB_DNODE(db);
	*dnsize = dn->dn_num_slots << DNODE_SHIFT;
	DB_DNODE_EXIT(db);
	}

	void
	byteswap_uint64_array(void *vbuf, size_t size)
	{
	uint64_t *buf = vbuf;
	size_t count = size >> 3;
	int i;

	ASSERT((size & 7) == 0);

	for (i = 0; i < count; i++)
	buf[i] = BSWAP_64(buf[i]);
	}

	void
	byteswap_uint32_array(void *vbuf, size_t size)
	{
	uint32_t *buf = vbuf;
	size_t count = size >> 2;
	int i;

	ASSERT((size & 3) == 0);

	for (i = 0; i < count; i++)
	buf[i] = BSWAP_32(buf[i]);
	}

	void
	byteswap_uint16_array(void *vbuf, size_t size)
	{
	uint16_t *buf = vbuf;
	size_t count = size >> 1;
	int i;

	ASSERT((size & 1) == 0);

	for (i = 0; i < count; i++)
	buf[i] = BSWAP_16(buf[i]);
	}

	void
	byteswap_uint8_array(void *vbuf, size_t size)
	{
	(void) vbuf, (void) size;
	}

	void
	dmu_init(void)
	{
	abd_init();
	zfs_dbgmsg_init();
	sa_cache_init();
	dmu_objset_init();
	dnode_init();
	zfetch_init();
	dmu_tx_init();
	l2arc_init();
	arc_init();
	dbuf_init();
	}

	void
	dmu_fini(void)
	{
	arc_fini(); /* arc depends on l2arc, so arc must go first */
	l2arc_fini();
	dmu_tx_fini();
	zfetch_fini();
	dbuf_fini();
	dnode_fini();
	dmu_objset_fini();
	sa_cache_fini();
	zfs_dbgmsg_fini();
	abd_fini();
	}

	EXPORT_SYMBOL(dmu_bonus_hold);
	EXPORT_SYMBOL(dmu_bonus_hold_by_dnode);
	EXPORT_SYMBOL(dmu_buf_hold_array_by_bonus);
	EXPORT_SYMBOL(dmu_buf_rele_array);
	EXPORT_SYMBOL(dmu_prefetch);
	EXPORT_SYMBOL(dmu_prefetch_by_dnode);
	EXPORT_SYMBOL(dmu_prefetch_dnode);
	EXPORT_SYMBOL(dmu_free_range);
	EXPORT_SYMBOL(dmu_free_long_range);
	EXPORT_SYMBOL(dmu_free_long_object);
	EXPORT_SYMBOL(dmu_read);
	EXPORT_SYMBOL(dmu_read_by_dnode);
	EXPORT_SYMBOL(dmu_write);
	EXPORT_SYMBOL(dmu_write_by_dnode);
	EXPORT_SYMBOL(dmu_prealloc);
	EXPORT_SYMBOL(dmu_object_info);
	EXPORT_SYMBOL(dmu_object_info_from_dnode);
	EXPORT_SYMBOL(dmu_object_info_from_db);
	EXPORT_SYMBOL(dmu_object_size_from_db);
	EXPORT_SYMBOL(dmu_object_dnsize_from_db);
	EXPORT_SYMBOL(dmu_object_set_nlevels);
	EXPORT_SYMBOL(dmu_object_set_blocksize);
	EXPORT_SYMBOL(dmu_object_set_maxblkid);
	EXPORT_SYMBOL(dmu_object_set_checksum);
	EXPORT_SYMBOL(dmu_object_set_compress);
	EXPORT_SYMBOL(dmu_offset_next);
	EXPORT_SYMBOL(dmu_write_policy);
	EXPORT_SYMBOL(dmu_sync);
	EXPORT_SYMBOL(dmu_request_arcbuf);
	EXPORT_SYMBOL(dmu_return_arcbuf);
	EXPORT_SYMBOL(dmu_assign_arcbuf_by_dnode);
	EXPORT_SYMBOL(dmu_assign_arcbuf_by_dbuf);
	EXPORT_SYMBOL(dmu_buf_hold);
	EXPORT_SYMBOL(dmu_ot);

	ZFS_MODULE_PARAM(zfs, zfs_, nopwrite_enabled, INT, ZMOD_RW,
	"Enable NOP writes");

	ZFS_MODULE_PARAM(zfs, zfs_, per_txg_dirty_frees_percent, UINT, ZMOD_RW,
	"Percentage of dirtied blocks from frees in one TXG");

	ZFS_MODULE_PARAM(zfs, zfs_, dmu_offset_next_sync, INT, ZMOD_RW,
	"Enable forcing txg sync to find holes");

	/* CSTYLED */
	ZFS_MODULE_PARAM(zfs, , dmu_prefetch_max, UINT, ZMOD_RW,
	"Limit one prefetch call to this size");
	diff --git a/sys/contrib/openzfs/module/zfs/dmu_diff.c b/sys/contrib/openzfs/module/zfs/dmu_diff.c
	index a2b1a27c88e9..0def0956beb8 100644
	--- a/sys/contrib/openzfs/module/zfs/dmu_diff.c
	+++ b/sys/contrib/openzfs/module/zfs/dmu_diff.c
	@@ -1,240 +1,240 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2018 by Delphix. All rights reserved.
	* Copyright (c) 2019, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
	*/

	#include <sys/dmu.h>
	#include <sys/dmu_impl.h>
	#include <sys/dmu_tx.h>
	#include <sys/dbuf.h>
	#include <sys/dnode.h>
	#include <sys/zfs_context.h>
	#include <sys/dmu_objset.h>
	#include <sys/dmu_traverse.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_dir.h>
	#include <sys/dsl_pool.h>
	#include <sys/dsl_synctask.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/zap.h>
	#include <sys/zio_checksum.h>
	#include <sys/zfs_znode.h>
	#include <sys/zfs_file.h>


	typedef struct dmu_diffarg {
	zfs_file_t da_fp; / file to which we are reporting */
	offset_t *da_offp;
	int da_err; /* error that stopped diff search */
	dmu_diff_record_t da_ddr;
	} dmu_diffarg_t;

	static int
	write_record(dmu_diffarg_t *da)
	{
	zfs_file_t *fp;
	ssize_t resid;

	if (da->da_ddr.ddr_type == DDR_NONE) {
	da->da_err = 0;
	return (0);
	}

	fp = da->da_fp;
	da->da_err = zfs_file_write(fp, (caddr_t)&da->da_ddr,
	sizeof (da->da_ddr), &resid);
	*da->da_offp += sizeof (da->da_ddr);
	return (da->da_err);
	}

	static int
	report_free_dnode_range(dmu_diffarg_t *da, uint64_t first, uint64_t last)
	{
	ASSERT(first <= last);
	if (da->da_ddr.ddr_type != DDR_FREE \|\|
	first != da->da_ddr.ddr_last + 1) {
	if (write_record(da) != 0)
	return (da->da_err);
	da->da_ddr.ddr_type = DDR_FREE;
	da->da_ddr.ddr_first = first;
	da->da_ddr.ddr_last = last;
	return (0);
	}
	da->da_ddr.ddr_last = last;
	return (0);
	}

	static int
	report_dnode(dmu_diffarg_t da, uint64_t object, dnode_phys_t dnp)
	{
	ASSERT(dnp != NULL);
	if (dnp->dn_type == DMU_OT_NONE)
	return (report_free_dnode_range(da, object, object));

	if (da->da_ddr.ddr_type != DDR_INUSE \|\|
	object != da->da_ddr.ddr_last + 1) {
	if (write_record(da) != 0)
	return (da->da_err);
	da->da_ddr.ddr_type = DDR_INUSE;
	da->da_ddr.ddr_first = da->da_ddr.ddr_last = object;
	return (0);
	}
	da->da_ddr.ddr_last = object;
	return (0);
	}

	#define DBP_SPAN(dnp, level) \
	(((uint64_t)dnp->dn_datablkszsec) << (SPA_MINBLOCKSHIFT + \
	(level) * (dnp->dn_indblkshift - SPA_BLKPTRSHIFT)))

	static int
	diff_cb(spa_t spa, zilog_t zilog, const blkptr_t *bp,
	const zbookmark_phys_t zb, const dnode_phys_t dnp, void *arg)
	{
	(void) zilog;
	dmu_diffarg_t *da = arg;
	int err = 0;

	- if (issig(JUSTLOOKING) && issig(FORREAL))
	+ if (issig())
	return (SET_ERROR(EINTR));

	if (zb->zb_level == ZB_DNODE_LEVEL \|\|
	zb->zb_object != DMU_META_DNODE_OBJECT)
	return (0);

	if (BP_IS_HOLE(bp)) {
	uint64_t span = DBP_SPAN(dnp, zb->zb_level);
	uint64_t dnobj = (zb->zb_blkid * span) >> DNODE_SHIFT;

	err = report_free_dnode_range(da, dnobj,
	dnobj + (span >> DNODE_SHIFT) - 1);
	if (err)
	return (err);
	} else if (zb->zb_level == 0) {
	dnode_phys_t *blk;
	arc_buf_t *abuf;
	arc_flags_t aflags = ARC_FLAG_WAIT;
	int epb = BP_GET_LSIZE(bp) >> DNODE_SHIFT;
	int zio_flags = ZIO_FLAG_CANFAIL;
	int i;

	if (BP_IS_PROTECTED(bp))
	zio_flags \|= ZIO_FLAG_RAW;

	if (arc_read(NULL, spa, bp, arc_getbuf_func, &abuf,
	ZIO_PRIORITY_ASYNC_READ, zio_flags, &aflags, zb) != 0)
	return (SET_ERROR(EIO));

	blk = abuf->b_data;
	for (i = 0; i < epb; i += blk[i].dn_extra_slots + 1) {
	uint64_t dnobj = (zb->zb_blkid <<
	(DNODE_BLOCK_SHIFT - DNODE_SHIFT)) + i;
	err = report_dnode(da, dnobj, blk+i);
	if (err)
	break;
	}
	arc_buf_destroy(abuf, &abuf);
	if (err)
	return (err);
	/* Don't care about the data blocks */
	return (TRAVERSE_VISIT_NO_CHILDREN);
	}
	return (0);
	}

	int
	dmu_diff(const char tosnap_name, const char fromsnap_name,
	zfs_file_t fp, offset_t offp)
	{
	dmu_diffarg_t da;
	dsl_dataset_t *fromsnap;
	dsl_dataset_t *tosnap;
	dsl_pool_t *dp;
	int error;
	uint64_t fromtxg;

	if (strchr(tosnap_name, '@') == NULL \|\|
	strchr(fromsnap_name, '@') == NULL)
	return (SET_ERROR(EINVAL));

	error = dsl_pool_hold(tosnap_name, FTAG, &dp);
	if (error != 0)
	return (error);

	error = dsl_dataset_hold(dp, tosnap_name, FTAG, &tosnap);
	if (error != 0) {
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	error = dsl_dataset_hold(dp, fromsnap_name, FTAG, &fromsnap);
	if (error != 0) {
	dsl_dataset_rele(tosnap, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	if (!dsl_dataset_is_before(tosnap, fromsnap, 0)) {
	dsl_dataset_rele(fromsnap, FTAG);
	dsl_dataset_rele(tosnap, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (SET_ERROR(EXDEV));
	}

	fromtxg = dsl_dataset_phys(fromsnap)->ds_creation_txg;
	dsl_dataset_rele(fromsnap, FTAG);

	dsl_dataset_long_hold(tosnap, FTAG);
	dsl_pool_rele(dp, FTAG);

	da.da_fp = fp;
	da.da_offp = offp;
	da.da_ddr.ddr_type = DDR_NONE;
	da.da_ddr.ddr_first = da.da_ddr.ddr_last = 0;
	da.da_err = 0;

	/*
	* Since zfs diff only looks at dnodes which are stored in plaintext
	* (other than bonus buffers), we don't technically need to decrypt
	* the dataset to perform this operation. However, the command line
	* utility will still fail if the keys are not loaded because the
	* dataset isn't mounted and because it will fail when it attempts to
	* call the ZFS_IOC_OBJ_TO_STATS ioctl.
	*/
	error = traverse_dataset(tosnap, fromtxg,
	TRAVERSE_PRE \| TRAVERSE_PREFETCH_METADATA \| TRAVERSE_NO_DECRYPT,
	diff_cb, &da);

	if (error != 0) {
	da.da_err = error;
	} else {
	/* we set the da.da_err we return as side-effect */
	(void) write_record(&da);
	}

	dsl_dataset_long_rele(tosnap, FTAG);
	dsl_dataset_rele(tosnap, FTAG);

	return (da.da_err);
	}
	diff --git a/sys/contrib/openzfs/module/zfs/dmu_object.c b/sys/contrib/openzfs/module/zfs/dmu_object.c
	index d0e39a423bb0..56986ea43446 100644
	--- a/sys/contrib/openzfs/module/zfs/dmu_object.c
	+++ b/sys/contrib/openzfs/module/zfs/dmu_object.c
	@@ -1,525 +1,525 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2013, 2017 by Delphix. All rights reserved.
	* Copyright 2014 HybridCluster. All rights reserved.
	*/

	#include <sys/dbuf.h>
	#include <sys/dmu.h>
	#include <sys/dmu_impl.h>
	#include <sys/dmu_objset.h>
	#include <sys/dmu_tx.h>
	#include <sys/dnode.h>
	#include <sys/zap.h>
	#include <sys/zfeature.h>
	#include <sys/dsl_dataset.h>

	/*
	* Each of the concurrent object allocators will grab
	* 2^dmu_object_alloc_chunk_shift dnode slots at a time. The default is to
	* grab 128 slots, which is 4 blocks worth. This was experimentally
	* determined to be the lowest value that eliminates the measurable effect
	* of lock contention from this code path.
	*/
	uint_t dmu_object_alloc_chunk_shift = 7;

	static uint64_t
	dmu_object_alloc_impl(objset_t *os, dmu_object_type_t ot, int blocksize,
	int indirect_blockshift, dmu_object_type_t bonustype, int bonuslen,
	int dnodesize, dnode_t *allocated_dnode, const void tag, dmu_tx_t *tx)
	{
	uint64_t object;
	uint64_t L1_dnode_count = DNODES_PER_BLOCK <<
	(DMU_META_DNODE(os)->dn_indblkshift - SPA_BLKPTRSHIFT);
	dnode_t *dn = NULL;
	int dn_slots = dnodesize >> DNODE_SHIFT;
	boolean_t restarted = B_FALSE;
	uint64_t *cpuobj = NULL;
	uint_t dnodes_per_chunk = 1 << dmu_object_alloc_chunk_shift;
	int error;

	cpuobj = &os->os_obj_next_percpu[CPU_SEQID_UNSTABLE %
	os->os_obj_next_percpu_len];

	if (dn_slots == 0) {
	dn_slots = DNODE_MIN_SLOTS;
	} else {
	ASSERT3S(dn_slots, >=, DNODE_MIN_SLOTS);
	ASSERT3S(dn_slots, <=, DNODE_MAX_SLOTS);
	}

	/*
	* The "chunk" of dnodes that is assigned to a CPU-specific
	* allocator needs to be at least one block's worth, to avoid
	* lock contention on the dbuf. It can be at most one L1 block's
	* worth, so that the "rescan after polishing off a L1's worth"
	* logic below will be sure to kick in.
	*/
	if (dnodes_per_chunk < DNODES_PER_BLOCK)
	dnodes_per_chunk = DNODES_PER_BLOCK;
	if (dnodes_per_chunk > L1_dnode_count)
	dnodes_per_chunk = L1_dnode_count;

	/*
	* The caller requested the dnode be returned as a performance
	* optimization in order to avoid releasing the hold only to
	* immediately reacquire it. Since they caller is responsible
	* for releasing the hold they must provide the tag.
	*/
	if (allocated_dnode != NULL) {
	ASSERT3P(tag, !=, NULL);
	} else {
	ASSERT3P(tag, ==, NULL);
	tag = FTAG;
	}

	object = *cpuobj;
	for (;;) {
	/*
	* If we finished a chunk of dnodes, get a new one from
	* the global allocator.
	*/
	if ((P2PHASE(object, dnodes_per_chunk) == 0) \|\|
	(P2PHASE(object + dn_slots - 1, dnodes_per_chunk) <
	dn_slots)) {
	DNODE_STAT_BUMP(dnode_alloc_next_chunk);
	mutex_enter(&os->os_obj_lock);
	ASSERT0(P2PHASE(os->os_obj_next_chunk,
	dnodes_per_chunk));
	object = os->os_obj_next_chunk;

	/*
	* Each time we polish off a L1 bp worth of dnodes
	* (2^12 objects), move to another L1 bp that's
	* still reasonably sparse (at most 1/4 full). Look
	* from the beginning at most once per txg. If we
	* still can't allocate from that L1 block, search
	* for an empty L0 block, which will quickly skip
	* to the end of the metadnode if no nearby L0
	* blocks are empty. This fallback avoids a
	* pathology where full dnode blocks containing
	* large dnodes appear sparse because they have a
	* low blk_fill, leading to many failed allocation
	* attempts. In the long term a better mechanism to
	* search for sparse metadnode regions, such as
	* spacemaps, could be implemented.
	*
	* os_scan_dnodes is set during txg sync if enough
	* objects have been freed since the previous
	* rescan to justify backfilling again.
	*
	* Note that dmu_traverse depends on the behavior
	* that we use multiple blocks of the dnode object
	* before going back to reuse objects. Any change
	* to this algorithm should preserve that property
	* or find another solution to the issues described
	* in traverse_visitbp.
	*/
	if (P2PHASE(object, L1_dnode_count) == 0) {
	uint64_t offset;
	uint64_t blkfill;
	int minlvl;
	if (os->os_rescan_dnodes) {
	offset = 0;
	os->os_rescan_dnodes = B_FALSE;
	} else {
	offset = object << DNODE_SHIFT;
	}
	blkfill = restarted ? 1 : DNODES_PER_BLOCK >> 2;
	minlvl = restarted ? 1 : 2;
	restarted = B_TRUE;
	error = dnode_next_offset(DMU_META_DNODE(os),
	DNODE_FIND_HOLE, &offset, minlvl,
	blkfill, 0);
	if (error == 0) {
	object = offset >> DNODE_SHIFT;
	}
	}
	/*
	* Note: if "restarted", we may find a L0 that
	* is not suitably aligned.
	*/
	os->os_obj_next_chunk =
	- P2ALIGN(object, dnodes_per_chunk) +
	+ P2ALIGN_TYPED(object, dnodes_per_chunk, uint64_t) +
	dnodes_per_chunk;
	(void) atomic_swap_64(cpuobj, object);
	mutex_exit(&os->os_obj_lock);
	}

	/*
	* The value of (*cpuobj) before adding dn_slots is the object
	* ID assigned to us. The value afterwards is the object ID
	* assigned to whoever wants to do an allocation next.
	*/
	object = atomic_add_64_nv(cpuobj, dn_slots) - dn_slots;

	/*
	* XXX We should check for an i/o error here and return
	* up to our caller. Actually we should pre-read it in
	* dmu_tx_assign(), but there is currently no mechanism
	* to do so.
	*/
	error = dnode_hold_impl(os, object, DNODE_MUST_BE_FREE,
	dn_slots, tag, &dn);
	if (error == 0) {
	rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
	/*
	* Another thread could have allocated it; check
	* again now that we have the struct lock.
	*/
	if (dn->dn_type == DMU_OT_NONE) {
	dnode_allocate(dn, ot, blocksize,
	indirect_blockshift, bonustype,
	bonuslen, dn_slots, tx);
	rw_exit(&dn->dn_struct_rwlock);
	dmu_tx_add_new_object(tx, dn);

	/*
	* Caller requested the allocated dnode be
	* returned and is responsible for the hold.
	*/
	if (allocated_dnode != NULL)
	*allocated_dnode = dn;
	else
	dnode_rele(dn, tag);

	return (object);
	}
	rw_exit(&dn->dn_struct_rwlock);
	dnode_rele(dn, tag);
	DNODE_STAT_BUMP(dnode_alloc_race);
	}

	/*
	* Skip to next known valid starting point on error. This
	* is the start of the next block of dnodes.
	*/
	if (dmu_object_next(os, &object, B_TRUE, 0) != 0) {
	object = P2ROUNDUP(object + 1, DNODES_PER_BLOCK);
	DNODE_STAT_BUMP(dnode_alloc_next_block);
	}
	(void) atomic_swap_64(cpuobj, object);
	}
	}

	uint64_t
	dmu_object_alloc(objset_t *os, dmu_object_type_t ot, int blocksize,
	dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx)
	{
	return dmu_object_alloc_impl(os, ot, blocksize, 0, bonustype,
	bonuslen, 0, NULL, NULL, tx);
	}

	uint64_t
	dmu_object_alloc_ibs(objset_t *os, dmu_object_type_t ot, int blocksize,
	int indirect_blockshift, dmu_object_type_t bonustype, int bonuslen,
	dmu_tx_t *tx)
	{
	return dmu_object_alloc_impl(os, ot, blocksize, indirect_blockshift,
	bonustype, bonuslen, 0, NULL, NULL, tx);
	}

	uint64_t
	dmu_object_alloc_dnsize(objset_t *os, dmu_object_type_t ot, int blocksize,
	dmu_object_type_t bonustype, int bonuslen, int dnodesize, dmu_tx_t *tx)
	{
	return (dmu_object_alloc_impl(os, ot, blocksize, 0, bonustype,
	bonuslen, dnodesize, NULL, NULL, tx));
	}

	/*
	* Allocate a new object and return a pointer to the newly allocated dnode
	* via the allocated_dnode argument. The returned dnode will be held and
	* the caller is responsible for releasing the hold by calling dnode_rele().
	*/
	uint64_t
	dmu_object_alloc_hold(objset_t *os, dmu_object_type_t ot, int blocksize,
	int indirect_blockshift, dmu_object_type_t bonustype, int bonuslen,
	int dnodesize, dnode_t *allocated_dnode, const void tag, dmu_tx_t *tx)
	{
	return (dmu_object_alloc_impl(os, ot, blocksize, indirect_blockshift,
	bonustype, bonuslen, dnodesize, allocated_dnode, tag, tx));
	}

	int
	dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
	int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx)
	{
	return (dmu_object_claim_dnsize(os, object, ot, blocksize, bonustype,
	bonuslen, 0, tx));
	}

	int
	dmu_object_claim_dnsize(objset_t *os, uint64_t object, dmu_object_type_t ot,
	int blocksize, dmu_object_type_t bonustype, int bonuslen,
	int dnodesize, dmu_tx_t *tx)
	{
	dnode_t *dn;
	int dn_slots = dnodesize >> DNODE_SHIFT;
	int err;

	if (dn_slots == 0)
	dn_slots = DNODE_MIN_SLOTS;
	ASSERT3S(dn_slots, >=, DNODE_MIN_SLOTS);
	ASSERT3S(dn_slots, <=, DNODE_MAX_SLOTS);

	if (object == DMU_META_DNODE_OBJECT && !dmu_tx_private_ok(tx))
	return (SET_ERROR(EBADF));

	err = dnode_hold_impl(os, object, DNODE_MUST_BE_FREE, dn_slots,
	FTAG, &dn);
	if (err)
	return (err);

	dnode_allocate(dn, ot, blocksize, 0, bonustype, bonuslen, dn_slots, tx);
	dmu_tx_add_new_object(tx, dn);

	dnode_rele(dn, FTAG);

	return (0);
	}

	int
	dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
	int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx)
	{
	return (dmu_object_reclaim_dnsize(os, object, ot, blocksize, bonustype,
	bonuslen, DNODE_MIN_SIZE, B_FALSE, tx));
	}

	int
	dmu_object_reclaim_dnsize(objset_t *os, uint64_t object, dmu_object_type_t ot,
	int blocksize, dmu_object_type_t bonustype, int bonuslen, int dnodesize,
	boolean_t keep_spill, dmu_tx_t *tx)
	{
	dnode_t *dn;
	int dn_slots = dnodesize >> DNODE_SHIFT;
	int err;

	if (dn_slots == 0)
	dn_slots = DNODE_MIN_SLOTS;

	if (object == DMU_META_DNODE_OBJECT)
	return (SET_ERROR(EBADF));

	err = dnode_hold_impl(os, object, DNODE_MUST_BE_ALLOCATED, 0,
	FTAG, &dn);
	if (err)
	return (err);

	dnode_reallocate(dn, ot, blocksize, bonustype, bonuslen, dn_slots,
	keep_spill, tx);

	dnode_rele(dn, FTAG);
	return (err);
	}

	int
	dmu_object_rm_spill(objset_t os, uint64_t object, dmu_tx_t tx)
	{
	dnode_t *dn;
	int err;

	err = dnode_hold_impl(os, object, DNODE_MUST_BE_ALLOCATED, 0,
	FTAG, &dn);
	if (err)
	return (err);

	rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
	if (dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR) {
	dbuf_rm_spill(dn, tx);
	dnode_rm_spill(dn, tx);
	}
	rw_exit(&dn->dn_struct_rwlock);

	dnode_rele(dn, FTAG);
	return (err);
	}

	int
	dmu_object_free(objset_t os, uint64_t object, dmu_tx_t tx)
	{
	dnode_t *dn;
	int err;

	ASSERT(object != DMU_META_DNODE_OBJECT \|\| dmu_tx_private_ok(tx));

	err = dnode_hold_impl(os, object, DNODE_MUST_BE_ALLOCATED, 0,
	FTAG, &dn);
	if (err)
	return (err);

	ASSERT(dn->dn_type != DMU_OT_NONE);
	/*
	* If we don't create this free range, we'll leak indirect blocks when
	* we get to freeing the dnode in syncing context.
	*/
	dnode_free_range(dn, 0, DMU_OBJECT_END, tx);
	dnode_free(dn, tx);
	dnode_rele(dn, FTAG);

	return (0);
	}

	/*
	* Return (in *objectp) the next object which is allocated (or a hole)
	* after *object, taking into account only objects that may have been modified
	* after the specified txg.
	*/
	int
	dmu_object_next(objset_t os, uint64_t objectp, boolean_t hole, uint64_t txg)
	{
	uint64_t offset;
	uint64_t start_obj;
	struct dsl_dataset *ds = os->os_dsl_dataset;
	int error;

	if (*objectp == 0) {
	start_obj = 1;
	} else if (ds && dsl_dataset_feature_is_active(ds,
	SPA_FEATURE_LARGE_DNODE)) {
	uint64_t i = *objectp + 1;
	uint64_t last_obj = *objectp \| (DNODES_PER_BLOCK - 1);
	dmu_object_info_t doi;

	/*
	* Scan through the remaining meta dnode block. The contents
	* of each slot in the block are known so it can be quickly
	* checked. If the block is exhausted without a match then
	* hand off to dnode_next_offset() for further scanning.
	*/
	while (i <= last_obj) {
	if (i == 0)
	return (SET_ERROR(ESRCH));
	error = dmu_object_info(os, i, &doi);
	if (error == ENOENT) {
	if (hole) {
	*objectp = i;
	return (0);
	} else {
	i++;
	}
	} else if (error == EEXIST) {
	i++;
	} else if (error == 0) {
	if (hole) {
	i += doi.doi_dnodesize >> DNODE_SHIFT;
	} else {
	*objectp = i;
	return (0);
	}
	} else {
	return (error);
	}
	}

	start_obj = i;
	} else {
	start_obj = *objectp + 1;
	}

	offset = start_obj << DNODE_SHIFT;

	error = dnode_next_offset(DMU_META_DNODE(os),
	(hole ? DNODE_FIND_HOLE : 0), &offset, 0, DNODES_PER_BLOCK, txg);

	*objectp = offset >> DNODE_SHIFT;

	return (error);
	}

	/*
	* Turn this object from old_type into DMU_OTN_ZAP_METADATA, and bump the
	* refcount on SPA_FEATURE_EXTENSIBLE_DATASET.
	*
	* Only for use from syncing context, on MOS objects.
	*/
	void
	dmu_object_zapify(objset_t *mos, uint64_t object, dmu_object_type_t old_type,
	dmu_tx_t *tx)
	{
	dnode_t *dn;

	ASSERT(dmu_tx_is_syncing(tx));

	VERIFY0(dnode_hold(mos, object, FTAG, &dn));
	if (dn->dn_type == DMU_OTN_ZAP_METADATA) {
	dnode_rele(dn, FTAG);
	return;
	}
	ASSERT3U(dn->dn_type, ==, old_type);
	ASSERT0(dn->dn_maxblkid);

	/*
	* We must initialize the ZAP data before changing the type,
	* so that concurrent calls to *_is_zapified() can determine if
	* the object has been completely zapified by checking the type.
	*/
	mzap_create_impl(dn, 0, 0, tx);

	dn->dn_next_type[tx->tx_txg & TXG_MASK] = dn->dn_type =
	DMU_OTN_ZAP_METADATA;
	dnode_setdirty(dn, tx);
	dnode_rele(dn, FTAG);

	spa_feature_incr(dmu_objset_spa(mos),
	SPA_FEATURE_EXTENSIBLE_DATASET, tx);
	}

	void
	dmu_object_free_zapified(objset_t mos, uint64_t object, dmu_tx_t tx)
	{
	dnode_t *dn;
	dmu_object_type_t t;

	ASSERT(dmu_tx_is_syncing(tx));

	VERIFY0(dnode_hold(mos, object, FTAG, &dn));
	t = dn->dn_type;
	dnode_rele(dn, FTAG);

	if (t == DMU_OTN_ZAP_METADATA) {
	spa_feature_decr(dmu_objset_spa(mos),
	SPA_FEATURE_EXTENSIBLE_DATASET, tx);
	}
	VERIFY0(dmu_object_free(mos, object, tx));
	}

	EXPORT_SYMBOL(dmu_object_alloc);
	EXPORT_SYMBOL(dmu_object_alloc_ibs);
	EXPORT_SYMBOL(dmu_object_alloc_dnsize);
	EXPORT_SYMBOL(dmu_object_alloc_hold);
	EXPORT_SYMBOL(dmu_object_claim);
	EXPORT_SYMBOL(dmu_object_claim_dnsize);
	EXPORT_SYMBOL(dmu_object_reclaim);
	EXPORT_SYMBOL(dmu_object_reclaim_dnsize);
	EXPORT_SYMBOL(dmu_object_rm_spill);
	EXPORT_SYMBOL(dmu_object_free);
	EXPORT_SYMBOL(dmu_object_next);
	EXPORT_SYMBOL(dmu_object_zapify);
	EXPORT_SYMBOL(dmu_object_free_zapified);

	/* BEGIN CSTYLED */
	ZFS_MODULE_PARAM(zfs, , dmu_object_alloc_chunk_shift, UINT, ZMOD_RW,
	"CPU-specific allocator grabs 2^N objects at once");
	/* END CSTYLED */
	diff --git a/sys/contrib/openzfs/module/zfs/dmu_objset.c b/sys/contrib/openzfs/module/zfs/dmu_objset.c
	index f1818ae155bd..8f4fefa4f4dd 100644
	--- a/sys/contrib/openzfs/module/zfs/dmu_objset.c
	+++ b/sys/contrib/openzfs/module/zfs/dmu_objset.c
	@@ -1,3174 +1,3174 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2020 by Delphix. All rights reserved.
	* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
	* Copyright (c) 2013, Joyent, Inc. All rights reserved.
	* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
	* Copyright (c) 2015, STRATO AG, Inc. All rights reserved.
	* Copyright (c) 2016 Actifio, Inc. All rights reserved.
	* Copyright 2017 Nexenta Systems, Inc.
	* Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
	* Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
	* Copyright (c) 2019, Klara Inc.
	* Copyright (c) 2019, Allan Jude
	* Copyright (c) 2022 Hewlett Packard Enterprise Development LP.
	*/

	/* Portions Copyright 2010 Robert Milkowski */

	#include <sys/cred.h>
	#include <sys/zfs_context.h>
	#include <sys/dmu_objset.h>
	#include <sys/dsl_dir.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_prop.h>
	#include <sys/dsl_pool.h>
	#include <sys/dsl_synctask.h>
	#include <sys/dsl_deleg.h>
	#include <sys/dnode.h>
	#include <sys/dbuf.h>
	#include <sys/zvol.h>
	#include <sys/dmu_tx.h>
	#include <sys/zap.h>
	#include <sys/zil.h>
	#include <sys/dmu_impl.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/sa.h>
	#include <sys/zfs_onexit.h>
	#include <sys/dsl_destroy.h>
	#include <sys/vdev.h>
	#include <sys/zfeature.h>
	#include <sys/policy.h>
	#include <sys/spa_impl.h>
	#include <sys/dmu_recv.h>
	#include <sys/zfs_project.h>
	#include "zfs_namecheck.h"
	#include <sys/vdev_impl.h>
	#include <sys/arc.h>

	/*
	* Needed to close a window in dnode_move() that allows the objset to be freed
	* before it can be safely accessed.
	*/
	krwlock_t os_lock;

	/*
	* Tunable to overwrite the maximum number of threads for the parallelization
	* of dmu_objset_find_dp, needed to speed up the import of pools with many
	* datasets.
	* Default is 4 times the number of leaf vdevs.
	*/
	static const int dmu_find_threads = 0;

	/*
	* Backfill lower metadnode objects after this many have been freed.
	* Backfilling negatively impacts object creation rates, so only do it
	* if there are enough holes to fill.
	*/
	static const int dmu_rescan_dnode_threshold = 1 << DN_MAX_INDBLKSHIFT;

	static const char *upgrade_tag = "upgrade_tag";

	static void dmu_objset_find_dp_cb(void *arg);

	static void dmu_objset_upgrade(objset_t *os, dmu_objset_upgrade_cb_t cb);
	static void dmu_objset_upgrade_stop(objset_t *os);

	void
	dmu_objset_init(void)
	{
	rw_init(&os_lock, NULL, RW_DEFAULT, NULL);
	}

	void
	dmu_objset_fini(void)
	{
	rw_destroy(&os_lock);
	}

	spa_t *
	dmu_objset_spa(objset_t *os)
	{
	return (os->os_spa);
	}

	zilog_t *
	dmu_objset_zil(objset_t *os)
	{
	return (os->os_zil);
	}

	dsl_pool_t *
	dmu_objset_pool(objset_t *os)
	{
	dsl_dataset_t *ds;

	if ((ds = os->os_dsl_dataset) != NULL && ds->ds_dir)
	return (ds->ds_dir->dd_pool);
	else
	return (spa_get_dsl(os->os_spa));
	}

	dsl_dataset_t *
	dmu_objset_ds(objset_t *os)
	{
	return (os->os_dsl_dataset);
	}

	dmu_objset_type_t
	dmu_objset_type(objset_t *os)
	{
	return (os->os_phys->os_type);
	}

	void
	dmu_objset_name(objset_t os, char buf)
	{
	dsl_dataset_name(os->os_dsl_dataset, buf);
	}

	uint64_t
	dmu_objset_id(objset_t *os)
	{
	dsl_dataset_t *ds = os->os_dsl_dataset;

	return (ds ? ds->ds_object : 0);
	}

	uint64_t
	dmu_objset_dnodesize(objset_t *os)
	{
	return (os->os_dnodesize);
	}

	zfs_sync_type_t
	dmu_objset_syncprop(objset_t *os)
	{
	return (os->os_sync);
	}

	zfs_logbias_op_t
	dmu_objset_logbias(objset_t *os)
	{
	return (os->os_logbias);
	}

	static void
	checksum_changed_cb(void *arg, uint64_t newval)
	{
	objset_t *os = arg;

	/*
	* Inheritance should have been done by now.
	*/
	ASSERT(newval != ZIO_CHECKSUM_INHERIT);

	os->os_checksum = zio_checksum_select(newval, ZIO_CHECKSUM_ON_VALUE);
	}

	static void
	compression_changed_cb(void *arg, uint64_t newval)
	{
	objset_t *os = arg;

	/*
	* Inheritance and range checking should have been done by now.
	*/
	ASSERT(newval != ZIO_COMPRESS_INHERIT);

	os->os_compress = zio_compress_select(os->os_spa,
	ZIO_COMPRESS_ALGO(newval), ZIO_COMPRESS_ON);
	os->os_complevel = zio_complevel_select(os->os_spa, os->os_compress,
	ZIO_COMPRESS_LEVEL(newval), ZIO_COMPLEVEL_DEFAULT);
	}

	static void
	copies_changed_cb(void *arg, uint64_t newval)
	{
	objset_t *os = arg;

	/*
	* Inheritance and range checking should have been done by now.
	*/
	ASSERT(newval > 0);
	ASSERT(newval <= spa_max_replication(os->os_spa));

	os->os_copies = newval;
	}

	static void
	dedup_changed_cb(void *arg, uint64_t newval)
	{
	objset_t *os = arg;
	spa_t *spa = os->os_spa;
	enum zio_checksum checksum;

	/*
	* Inheritance should have been done by now.
	*/
	ASSERT(newval != ZIO_CHECKSUM_INHERIT);

	checksum = zio_checksum_dedup_select(spa, newval, ZIO_CHECKSUM_OFF);

	os->os_dedup_checksum = checksum & ZIO_CHECKSUM_MASK;
	os->os_dedup_verify = !!(checksum & ZIO_CHECKSUM_VERIFY);
	}

	static void
	primary_cache_changed_cb(void *arg, uint64_t newval)
	{
	objset_t *os = arg;

	/*
	* Inheritance and range checking should have been done by now.
	*/
	ASSERT(newval == ZFS_CACHE_ALL \|\| newval == ZFS_CACHE_NONE \|\|
	newval == ZFS_CACHE_METADATA);

	os->os_primary_cache = newval;
	}

	static void
	secondary_cache_changed_cb(void *arg, uint64_t newval)
	{
	objset_t *os = arg;

	/*
	* Inheritance and range checking should have been done by now.
	*/
	ASSERT(newval == ZFS_CACHE_ALL \|\| newval == ZFS_CACHE_NONE \|\|
	newval == ZFS_CACHE_METADATA);

	os->os_secondary_cache = newval;
	}

	static void
	prefetch_changed_cb(void *arg, uint64_t newval)
	{
	objset_t *os = arg;

	/*
	* Inheritance should have been done by now.
	*/
	ASSERT(newval == ZFS_PREFETCH_ALL \|\| newval == ZFS_PREFETCH_NONE \|\|
	newval == ZFS_PREFETCH_METADATA);
	os->os_prefetch = newval;
	}

	static void
	sync_changed_cb(void *arg, uint64_t newval)
	{
	objset_t *os = arg;

	/*
	* Inheritance and range checking should have been done by now.
	*/
	ASSERT(newval == ZFS_SYNC_STANDARD \|\| newval == ZFS_SYNC_ALWAYS \|\|
	newval == ZFS_SYNC_DISABLED);

	os->os_sync = newval;
	if (os->os_zil)
	zil_set_sync(os->os_zil, newval);
	}

	static void
	redundant_metadata_changed_cb(void *arg, uint64_t newval)
	{
	objset_t *os = arg;

	/*
	* Inheritance and range checking should have been done by now.
	*/
	ASSERT(newval == ZFS_REDUNDANT_METADATA_ALL \|\|
	newval == ZFS_REDUNDANT_METADATA_MOST \|\|
	newval == ZFS_REDUNDANT_METADATA_SOME \|\|
	newval == ZFS_REDUNDANT_METADATA_NONE);

	os->os_redundant_metadata = newval;
	}

	static void
	dnodesize_changed_cb(void *arg, uint64_t newval)
	{
	objset_t *os = arg;

	switch (newval) {
	case ZFS_DNSIZE_LEGACY:
	os->os_dnodesize = DNODE_MIN_SIZE;
	break;
	case ZFS_DNSIZE_AUTO:
	/*
	* Choose a dnode size that will work well for most
	* workloads if the user specified "auto". Future code
	* improvements could dynamically select a dnode size
	* based on observed workload patterns.
	*/
	os->os_dnodesize = DNODE_MIN_SIZE * 2;
	break;
	case ZFS_DNSIZE_1K:
	case ZFS_DNSIZE_2K:
	case ZFS_DNSIZE_4K:
	case ZFS_DNSIZE_8K:
	case ZFS_DNSIZE_16K:
	os->os_dnodesize = newval;
	break;
	}
	}

	static void
	smallblk_changed_cb(void *arg, uint64_t newval)
	{
	objset_t *os = arg;

	/*
	* Inheritance and range checking should have been done by now.
	*/
	ASSERT(newval <= SPA_MAXBLOCKSIZE);
	ASSERT(ISP2(newval));

	os->os_zpl_special_smallblock = newval;
	}

	static void
	logbias_changed_cb(void *arg, uint64_t newval)
	{
	objset_t *os = arg;

	ASSERT(newval == ZFS_LOGBIAS_LATENCY \|\|
	newval == ZFS_LOGBIAS_THROUGHPUT);
	os->os_logbias = newval;
	if (os->os_zil)
	zil_set_logbias(os->os_zil, newval);
	}

	static void
	recordsize_changed_cb(void *arg, uint64_t newval)
	{
	objset_t *os = arg;

	os->os_recordsize = newval;
	}

	void
	dmu_objset_byteswap(void *buf, size_t size)
	{
	objset_phys_t *osp = buf;

	ASSERT(size == OBJSET_PHYS_SIZE_V1 \|\| size == OBJSET_PHYS_SIZE_V2 \|\|
	size == sizeof (objset_phys_t));
	dnode_byteswap(&osp->os_meta_dnode);
	byteswap_uint64_array(&osp->os_zil_header, sizeof (zil_header_t));
	osp->os_type = BSWAP_64(osp->os_type);
	osp->os_flags = BSWAP_64(osp->os_flags);
	if (size >= OBJSET_PHYS_SIZE_V2) {
	dnode_byteswap(&osp->os_userused_dnode);
	dnode_byteswap(&osp->os_groupused_dnode);
	if (size >= sizeof (objset_phys_t))
	dnode_byteswap(&osp->os_projectused_dnode);
	}
	}

	/*
	* The hash is a CRC-based hash of the objset_t pointer and the object number.
	*/
	static uint64_t
	dnode_hash(const objset_t *os, uint64_t obj)
	{
	uintptr_t osv = (uintptr_t)os;
	uint64_t crc = -1ULL;

	ASSERT(zfs_crc64_table[128] == ZFS_CRC64_POLY);
	/*
	* The lower 11 bits of the pointer don't have much entropy, because
	* the objset_t is more than 1KB long and so likely aligned to 2KB.
	*/
	crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (osv >> 11)) & 0xFF];
	crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 0)) & 0xFF];
	crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 8)) & 0xFF];
	crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 16)) & 0xFF];

	crc ^= (osv>>14) ^ (obj>>24);

	return (crc);
	}

	static unsigned int
	dnode_multilist_index_func(multilist_t ml, void obj)
	{
	dnode_t *dn = obj;

	/*
	* The low order bits of the hash value are thought to be
	* distributed evenly. Otherwise, in the case that the multilist
	* has a power of two number of sublists, each sublists' usage
	* would not be evenly distributed. In this context full 64bit
	* division would be a waste of time, so limit it to 32 bits.
	*/
	return ((unsigned int)dnode_hash(dn->dn_objset, dn->dn_object) %
	multilist_get_num_sublists(ml));
	}

	static inline boolean_t
	dmu_os_is_l2cacheable(objset_t *os)
	{
	if (os->os_secondary_cache == ZFS_CACHE_ALL \|\|
	os->os_secondary_cache == ZFS_CACHE_METADATA) {
	if (l2arc_exclude_special == 0)
	return (B_TRUE);

	blkptr_t *bp = os->os_rootbp;
	if (bp == NULL \|\| BP_IS_HOLE(bp))
	return (B_FALSE);
	uint64_t vdev = DVA_GET_VDEV(bp->blk_dva);
	vdev_t *rvd = os->os_spa->spa_root_vdev;
	vdev_t *vd = NULL;

	if (vdev < rvd->vdev_children)
	vd = rvd->vdev_child[vdev];

	if (vd == NULL)
	return (B_TRUE);

	if (vd->vdev_alloc_bias != VDEV_BIAS_SPECIAL &&
	vd->vdev_alloc_bias != VDEV_BIAS_DEDUP)
	return (B_TRUE);
	}
	return (B_FALSE);
	}

	/*
	* Instantiates the objset_t in-memory structure corresponding to the
	* objset_phys_t that's pointed to by the specified blkptr_t.
	*/
	int
	dmu_objset_open_impl(spa_t spa, dsl_dataset_t ds, blkptr_t *bp,
	objset_t **osp)
	{
	objset_t *os;
	int i, err;

	ASSERT(ds == NULL \|\| MUTEX_HELD(&ds->ds_opening_lock));
	ASSERT(!BP_IS_REDACTED(bp));

	/*
	* We need the pool config lock to get properties.
	*/
	ASSERT(ds == NULL \|\| dsl_pool_config_held(ds->ds_dir->dd_pool));

	/*
	* The $ORIGIN dataset (if it exists) doesn't have an associated
	* objset, so there's no reason to open it. The $ORIGIN dataset
	* will not exist on pools older than SPA_VERSION_ORIGIN.
	*/
	if (ds != NULL && spa_get_dsl(spa) != NULL &&
	spa_get_dsl(spa)->dp_origin_snap != NULL) {
	ASSERT3P(ds->ds_dir, !=,
	spa_get_dsl(spa)->dp_origin_snap->ds_dir);
	}

	os = kmem_zalloc(sizeof (objset_t), KM_SLEEP);
	os->os_dsl_dataset = ds;
	os->os_spa = spa;
	os->os_rootbp = bp;
	if (!BP_IS_HOLE(os->os_rootbp)) {
	arc_flags_t aflags = ARC_FLAG_WAIT;
	zbookmark_phys_t zb;
	int size;
	zio_flag_t zio_flags = ZIO_FLAG_CANFAIL;
	SET_BOOKMARK(&zb, ds ? ds->ds_object : DMU_META_OBJSET,
	ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID);

	if (dmu_os_is_l2cacheable(os))
	aflags \|= ARC_FLAG_L2CACHE;

	if (ds != NULL && ds->ds_dir->dd_crypto_obj != 0) {
	ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF);
	ASSERT(BP_IS_AUTHENTICATED(bp));
	zio_flags \|= ZIO_FLAG_RAW;
	}

	dprintf_bp(os->os_rootbp, "reading %s", "");
	err = arc_read(NULL, spa, os->os_rootbp,
	arc_getbuf_func, &os->os_phys_buf,
	ZIO_PRIORITY_SYNC_READ, zio_flags, &aflags, &zb);
	if (err != 0) {
	kmem_free(os, sizeof (objset_t));
	/* convert checksum errors into IO errors */
	if (err == ECKSUM)
	err = SET_ERROR(EIO);
	return (err);
	}

	if (spa_version(spa) < SPA_VERSION_USERSPACE)
	size = OBJSET_PHYS_SIZE_V1;
	else if (!spa_feature_is_enabled(spa,
	SPA_FEATURE_PROJECT_QUOTA))
	size = OBJSET_PHYS_SIZE_V2;
	else
	size = sizeof (objset_phys_t);

	/* Increase the blocksize if we are permitted. */
	if (arc_buf_size(os->os_phys_buf) < size) {
	arc_buf_t *buf = arc_alloc_buf(spa, &os->os_phys_buf,
	ARC_BUFC_METADATA, size);
	memset(buf->b_data, 0, size);
	memcpy(buf->b_data, os->os_phys_buf->b_data,
	arc_buf_size(os->os_phys_buf));
	arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf);
	os->os_phys_buf = buf;
	}

	os->os_phys = os->os_phys_buf->b_data;
	os->os_flags = os->os_phys->os_flags;
	} else {
	int size = spa_version(spa) >= SPA_VERSION_USERSPACE ?
	sizeof (objset_phys_t) : OBJSET_PHYS_SIZE_V1;
	os->os_phys_buf = arc_alloc_buf(spa, &os->os_phys_buf,
	ARC_BUFC_METADATA, size);
	os->os_phys = os->os_phys_buf->b_data;
	memset(os->os_phys, 0, size);
	}
	/*
	* These properties will be filled in by the logic in zfs_get_zplprop()
	* when they are queried for the first time.
	*/
	os->os_version = OBJSET_PROP_UNINITIALIZED;
	os->os_normalization = OBJSET_PROP_UNINITIALIZED;
	os->os_utf8only = OBJSET_PROP_UNINITIALIZED;
	os->os_casesensitivity = OBJSET_PROP_UNINITIALIZED;

	/*
	* Note: the changed_cb will be called once before the register
	* func returns, thus changing the checksum/compression from the
	* default (fletcher2/off). Snapshots don't need to know about
	* checksum/compression/copies.
	*/
	if (ds != NULL) {
	os->os_encrypted = (ds->ds_dir->dd_crypto_obj != 0);

	err = dsl_prop_register(ds,
	zfs_prop_to_name(ZFS_PROP_PRIMARYCACHE),
	primary_cache_changed_cb, os);
	if (err == 0) {
	err = dsl_prop_register(ds,
	zfs_prop_to_name(ZFS_PROP_SECONDARYCACHE),
	secondary_cache_changed_cb, os);
	}
	if (err == 0) {
	err = dsl_prop_register(ds,
	zfs_prop_to_name(ZFS_PROP_PREFETCH),
	prefetch_changed_cb, os);
	}
	if (!ds->ds_is_snapshot) {
	if (err == 0) {
	err = dsl_prop_register(ds,
	zfs_prop_to_name(ZFS_PROP_CHECKSUM),
	checksum_changed_cb, os);
	}
	if (err == 0) {
	err = dsl_prop_register(ds,
	zfs_prop_to_name(ZFS_PROP_COMPRESSION),
	compression_changed_cb, os);
	}
	if (err == 0) {
	err = dsl_prop_register(ds,
	zfs_prop_to_name(ZFS_PROP_COPIES),
	copies_changed_cb, os);
	}
	if (err == 0) {
	err = dsl_prop_register(ds,
	zfs_prop_to_name(ZFS_PROP_DEDUP),
	dedup_changed_cb, os);
	}
	if (err == 0) {
	err = dsl_prop_register(ds,
	zfs_prop_to_name(ZFS_PROP_LOGBIAS),
	logbias_changed_cb, os);
	}
	if (err == 0) {
	err = dsl_prop_register(ds,
	zfs_prop_to_name(ZFS_PROP_SYNC),
	sync_changed_cb, os);
	}
	if (err == 0) {
	err = dsl_prop_register(ds,
	zfs_prop_to_name(
	ZFS_PROP_REDUNDANT_METADATA),
	redundant_metadata_changed_cb, os);
	}
	if (err == 0) {
	err = dsl_prop_register(ds,
	zfs_prop_to_name(ZFS_PROP_RECORDSIZE),
	recordsize_changed_cb, os);
	}
	if (err == 0) {
	err = dsl_prop_register(ds,
	zfs_prop_to_name(ZFS_PROP_DNODESIZE),
	dnodesize_changed_cb, os);
	}
	if (err == 0) {
	err = dsl_prop_register(ds,
	zfs_prop_to_name(
	ZFS_PROP_SPECIAL_SMALL_BLOCKS),
	smallblk_changed_cb, os);
	}
	}
	if (err != 0) {
	arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf);
	kmem_free(os, sizeof (objset_t));
	return (err);
	}
	} else {
	/* It's the meta-objset. */
	os->os_checksum = ZIO_CHECKSUM_FLETCHER_4;
	os->os_compress = ZIO_COMPRESS_ON;
	os->os_complevel = ZIO_COMPLEVEL_DEFAULT;
	os->os_encrypted = B_FALSE;
	os->os_copies = spa_max_replication(spa);
	os->os_dedup_checksum = ZIO_CHECKSUM_OFF;
	os->os_dedup_verify = B_FALSE;
	os->os_logbias = ZFS_LOGBIAS_LATENCY;
	os->os_sync = ZFS_SYNC_STANDARD;
	os->os_primary_cache = ZFS_CACHE_ALL;
	os->os_secondary_cache = ZFS_CACHE_ALL;
	os->os_dnodesize = DNODE_MIN_SIZE;
	os->os_prefetch = ZFS_PREFETCH_ALL;
	}

	if (ds == NULL \|\| !ds->ds_is_snapshot)
	os->os_zil_header = os->os_phys->os_zil_header;
	os->os_zil = zil_alloc(os, &os->os_zil_header);

	for (i = 0; i < TXG_SIZE; i++) {
	multilist_create(&os->os_dirty_dnodes[i], sizeof (dnode_t),
	offsetof(dnode_t, dn_dirty_link[i]),
	dnode_multilist_index_func);
	}
	list_create(&os->os_dnodes, sizeof (dnode_t),
	offsetof(dnode_t, dn_link));
	list_create(&os->os_downgraded_dbufs, sizeof (dmu_buf_impl_t),
	offsetof(dmu_buf_impl_t, db_link));

	list_link_init(&os->os_evicting_node);

	mutex_init(&os->os_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&os->os_userused_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&os->os_obj_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&os->os_user_ptr_lock, NULL, MUTEX_DEFAULT, NULL);
	os->os_obj_next_percpu_len = boot_ncpus;
	os->os_obj_next_percpu = kmem_zalloc(os->os_obj_next_percpu_len *
	sizeof (os->os_obj_next_percpu[0]), KM_SLEEP);

	dnode_special_open(os, &os->os_phys->os_meta_dnode,
	DMU_META_DNODE_OBJECT, &os->os_meta_dnode);
	if (OBJSET_BUF_HAS_USERUSED(os->os_phys_buf)) {
	dnode_special_open(os, &os->os_phys->os_userused_dnode,
	DMU_USERUSED_OBJECT, &os->os_userused_dnode);
	dnode_special_open(os, &os->os_phys->os_groupused_dnode,
	DMU_GROUPUSED_OBJECT, &os->os_groupused_dnode);
	if (OBJSET_BUF_HAS_PROJECTUSED(os->os_phys_buf))
	dnode_special_open(os,
	&os->os_phys->os_projectused_dnode,
	DMU_PROJECTUSED_OBJECT, &os->os_projectused_dnode);
	}

	mutex_init(&os->os_upgrade_lock, NULL, MUTEX_DEFAULT, NULL);

	*osp = os;
	return (0);
	}

	int
	dmu_objset_from_ds(dsl_dataset_t ds, objset_t *osp)
	{
	int err = 0;

	/*
	* We need the pool_config lock to manipulate the dsl_dataset_t.
	* Even if the dataset is long-held, we need the pool_config lock
	* to open the objset, as it needs to get properties.
	*/
	ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool));

	mutex_enter(&ds->ds_opening_lock);
	if (ds->ds_objset == NULL) {
	objset_t *os;
	rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
	err = dmu_objset_open_impl(dsl_dataset_get_spa(ds),
	ds, dsl_dataset_get_blkptr(ds), &os);
	rrw_exit(&ds->ds_bp_rwlock, FTAG);

	if (err == 0) {
	mutex_enter(&ds->ds_lock);
	ASSERT(ds->ds_objset == NULL);
	ds->ds_objset = os;
	mutex_exit(&ds->ds_lock);
	}
	}
	*osp = ds->ds_objset;
	mutex_exit(&ds->ds_opening_lock);
	return (err);
	}

	/*
	* Holds the pool while the objset is held. Therefore only one objset
	* can be held at a time.
	*/
	int
	dmu_objset_hold_flags(const char name, boolean_t decrypt, const void tag,
	objset_t **osp)
	{
	dsl_pool_t *dp;
	dsl_dataset_t *ds;
	int err;
	ds_hold_flags_t flags;

	flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : DS_HOLD_FLAG_NONE;
	err = dsl_pool_hold(name, tag, &dp);
	if (err != 0)
	return (err);
	err = dsl_dataset_hold_flags(dp, name, flags, tag, &ds);
	if (err != 0) {
	dsl_pool_rele(dp, tag);
	return (err);
	}

	err = dmu_objset_from_ds(ds, osp);
	if (err != 0) {
	dsl_dataset_rele(ds, tag);
	dsl_pool_rele(dp, tag);
	}

	return (err);
	}

	int
	dmu_objset_hold(const char name, const void tag, objset_t **osp)
	{
	return (dmu_objset_hold_flags(name, B_FALSE, tag, osp));
	}

	static int
	dmu_objset_own_impl(dsl_dataset_t *ds, dmu_objset_type_t type,
	boolean_t readonly, boolean_t decrypt, const void tag, objset_t *osp)
	{
	(void) tag;

	int err = dmu_objset_from_ds(ds, osp);
	if (err != 0) {
	return (err);
	} else if (type != DMU_OST_ANY && type != (*osp)->os_phys->os_type) {
	return (SET_ERROR(EINVAL));
	} else if (!readonly && dsl_dataset_is_snapshot(ds)) {
	return (SET_ERROR(EROFS));
	} else if (!readonly && decrypt &&
	dsl_dir_incompatible_encryption_version(ds->ds_dir)) {
	return (SET_ERROR(EROFS));
	}

	/* if we are decrypting, we can now check MACs in os->os_phys_buf */
	if (decrypt && arc_is_unauthenticated((*osp)->os_phys_buf)) {
	zbookmark_phys_t zb;

	SET_BOOKMARK(&zb, ds->ds_object, ZB_ROOT_OBJECT,
	ZB_ROOT_LEVEL, ZB_ROOT_BLKID);
	err = arc_untransform((osp)->os_phys_buf, (osp)->os_spa,
	&zb, B_FALSE);
	if (err != 0)
	return (err);

	ASSERT0(arc_is_unauthenticated((*osp)->os_phys_buf));
	}

	return (0);
	}

	/*
	* dsl_pool must not be held when this is called.
	* Upon successful return, there will be a longhold on the dataset,
	* and the dsl_pool will not be held.
	*/
	int
	dmu_objset_own(const char *name, dmu_objset_type_t type,
	boolean_t readonly, boolean_t decrypt, const void tag, objset_t *osp)
	{
	dsl_pool_t *dp;
	dsl_dataset_t *ds;
	int err;
	ds_hold_flags_t flags;

	flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : DS_HOLD_FLAG_NONE;
	err = dsl_pool_hold(name, FTAG, &dp);
	if (err != 0)
	return (err);
	err = dsl_dataset_own(dp, name, flags, tag, &ds);
	if (err != 0) {
	dsl_pool_rele(dp, FTAG);
	return (err);
	}
	err = dmu_objset_own_impl(ds, type, readonly, decrypt, tag, osp);
	if (err != 0) {
	dsl_dataset_disown(ds, flags, tag);
	dsl_pool_rele(dp, FTAG);
	return (err);
	}

	/*
	* User accounting requires the dataset to be decrypted and rw.
	* We also don't begin user accounting during claiming to help
	* speed up pool import times and to keep this txg reserved
	* completely for recovery work.
	*/
	if (!readonly && !dp->dp_spa->spa_claiming &&
	(ds->ds_dir->dd_crypto_obj == 0 \|\| decrypt)) {
	if (dmu_objset_userobjspace_upgradable(*osp) \|\|
	dmu_objset_projectquota_upgradable(*osp)) {
	dmu_objset_id_quota_upgrade(*osp);
	} else if (dmu_objset_userused_enabled(*osp)) {
	dmu_objset_userspace_upgrade(*osp);
	}
	}

	dsl_pool_rele(dp, FTAG);
	return (0);
	}

	int
	dmu_objset_own_obj(dsl_pool_t *dp, uint64_t obj, dmu_objset_type_t type,
	boolean_t readonly, boolean_t decrypt, const void tag, objset_t *osp)
	{
	dsl_dataset_t *ds;
	int err;
	ds_hold_flags_t flags;

	flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : DS_HOLD_FLAG_NONE;
	err = dsl_dataset_own_obj(dp, obj, flags, tag, &ds);
	if (err != 0)
	return (err);

	err = dmu_objset_own_impl(ds, type, readonly, decrypt, tag, osp);
	if (err != 0) {
	dsl_dataset_disown(ds, flags, tag);
	return (err);
	}

	return (0);
	}

	void
	dmu_objset_rele_flags(objset_t os, boolean_t decrypt, const void tag)
	{
	ds_hold_flags_t flags;
	dsl_pool_t *dp = dmu_objset_pool(os);

	flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : DS_HOLD_FLAG_NONE;
	dsl_dataset_rele_flags(os->os_dsl_dataset, flags, tag);
	dsl_pool_rele(dp, tag);
	}

	void
	dmu_objset_rele(objset_t os, const void tag)
	{
	dmu_objset_rele_flags(os, B_FALSE, tag);
	}

	/*
	* When we are called, os MUST refer to an objset associated with a dataset
	* that is owned by 'tag'; that is, is held and long held by 'tag' and ds_owner
	* == tag. We will then release and reacquire ownership of the dataset while
	* holding the pool config_rwlock to avoid intervening namespace or ownership
	* changes may occur.
	*
	* This exists solely to accommodate zfs_ioc_userspace_upgrade()'s desire to
	* release the hold on its dataset and acquire a new one on the dataset of the
	* same name so that it can be partially torn down and reconstructed.
	*/
	void
	dmu_objset_refresh_ownership(dsl_dataset_t ds, dsl_dataset_t *newds,
	boolean_t decrypt, const void *tag)
	{
	dsl_pool_t *dp;
	char name[ZFS_MAX_DATASET_NAME_LEN];
	ds_hold_flags_t flags;

	flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : DS_HOLD_FLAG_NONE;
	VERIFY3P(ds, !=, NULL);
	VERIFY3P(ds->ds_owner, ==, tag);
	VERIFY(dsl_dataset_long_held(ds));

	dsl_dataset_name(ds, name);
	dp = ds->ds_dir->dd_pool;
	dsl_pool_config_enter(dp, FTAG);
	dsl_dataset_disown(ds, flags, tag);
	VERIFY0(dsl_dataset_own(dp, name, flags, tag, newds));
	dsl_pool_config_exit(dp, FTAG);
	}

	void
	dmu_objset_disown(objset_t os, boolean_t decrypt, const void tag)
	{
	ds_hold_flags_t flags;

	flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : DS_HOLD_FLAG_NONE;
	/*
	* Stop upgrading thread
	*/
	dmu_objset_upgrade_stop(os);
	dsl_dataset_disown(os->os_dsl_dataset, flags, tag);
	}

	void
	dmu_objset_evict_dbufs(objset_t *os)
	{
	dnode_t *dn_marker;
	dnode_t *dn;

	dn_marker = kmem_alloc(sizeof (dnode_t), KM_SLEEP);

	mutex_enter(&os->os_lock);
	dn = list_head(&os->os_dnodes);
	while (dn != NULL) {
	/*
	* Skip dnodes without holds. We have to do this dance
	* because dnode_add_ref() only works if there is already a
	* hold. If the dnode has no holds, then it has no dbufs.
	*/
	if (dnode_add_ref(dn, FTAG)) {
	list_insert_after(&os->os_dnodes, dn, dn_marker);
	mutex_exit(&os->os_lock);

	dnode_evict_dbufs(dn);
	dnode_rele(dn, FTAG);

	mutex_enter(&os->os_lock);
	dn = list_next(&os->os_dnodes, dn_marker);
	list_remove(&os->os_dnodes, dn_marker);
	} else {
	dn = list_next(&os->os_dnodes, dn);
	}
	}
	mutex_exit(&os->os_lock);

	kmem_free(dn_marker, sizeof (dnode_t));

	if (DMU_USERUSED_DNODE(os) != NULL) {
	if (DMU_PROJECTUSED_DNODE(os) != NULL)
	dnode_evict_dbufs(DMU_PROJECTUSED_DNODE(os));
	dnode_evict_dbufs(DMU_GROUPUSED_DNODE(os));
	dnode_evict_dbufs(DMU_USERUSED_DNODE(os));
	}
	dnode_evict_dbufs(DMU_META_DNODE(os));
	}

	/*
	* Objset eviction processing is split into into two pieces.
	* The first marks the objset as evicting, evicts any dbufs that
	* have a refcount of zero, and then queues up the objset for the
	* second phase of eviction. Once os->os_dnodes has been cleared by
	* dnode_buf_pageout()->dnode_destroy(), the second phase is executed.
	* The second phase closes the special dnodes, dequeues the objset from
	* the list of those undergoing eviction, and finally frees the objset.
	*
	* NOTE: Due to asynchronous eviction processing (invocation of
	* dnode_buf_pageout()), it is possible for the meta dnode for the
	* objset to have no holds even though os->os_dnodes is not empty.
	*/
	void
	dmu_objset_evict(objset_t *os)
	{
	dsl_dataset_t *ds = os->os_dsl_dataset;

	for (int t = 0; t < TXG_SIZE; t++)
	ASSERT(!dmu_objset_is_dirty(os, t));

	if (ds)
	dsl_prop_unregister_all(ds, os);

	if (os->os_sa)
	sa_tear_down(os);

	dmu_objset_evict_dbufs(os);

	mutex_enter(&os->os_lock);
	spa_evicting_os_register(os->os_spa, os);
	if (list_is_empty(&os->os_dnodes)) {
	mutex_exit(&os->os_lock);
	dmu_objset_evict_done(os);
	} else {
	mutex_exit(&os->os_lock);
	}


	}

	void
	dmu_objset_evict_done(objset_t *os)
	{
	ASSERT3P(list_head(&os->os_dnodes), ==, NULL);

	dnode_special_close(&os->os_meta_dnode);
	if (DMU_USERUSED_DNODE(os)) {
	if (DMU_PROJECTUSED_DNODE(os))
	dnode_special_close(&os->os_projectused_dnode);
	dnode_special_close(&os->os_userused_dnode);
	dnode_special_close(&os->os_groupused_dnode);
	}
	zil_free(os->os_zil);

	arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf);

	/*
	* This is a barrier to prevent the objset from going away in
	* dnode_move() until we can safely ensure that the objset is still in
	* use. We consider the objset valid before the barrier and invalid
	* after the barrier.
	*/
	rw_enter(&os_lock, RW_READER);
	rw_exit(&os_lock);

	kmem_free(os->os_obj_next_percpu,
	os->os_obj_next_percpu_len * sizeof (os->os_obj_next_percpu[0]));

	mutex_destroy(&os->os_lock);
	mutex_destroy(&os->os_userused_lock);
	mutex_destroy(&os->os_obj_lock);
	mutex_destroy(&os->os_user_ptr_lock);
	mutex_destroy(&os->os_upgrade_lock);
	for (int i = 0; i < TXG_SIZE; i++)
	multilist_destroy(&os->os_dirty_dnodes[i]);
	spa_evicting_os_deregister(os->os_spa, os);
	kmem_free(os, sizeof (objset_t));
	}

	inode_timespec_t
	dmu_objset_snap_cmtime(objset_t *os)
	{
	return (dsl_dir_snap_cmtime(os->os_dsl_dataset->ds_dir));
	}

	objset_t *
	dmu_objset_create_impl_dnstats(spa_t spa, dsl_dataset_t ds, blkptr_t *bp,
	dmu_objset_type_t type, int levels, int blksz, int ibs, dmu_tx_t *tx)
	{
	objset_t *os;
	dnode_t *mdn;

	ASSERT(dmu_tx_is_syncing(tx));

	if (blksz == 0)
	blksz = DNODE_BLOCK_SIZE;
	if (ibs == 0)
	ibs = DN_MAX_INDBLKSHIFT;

	if (ds != NULL)
	VERIFY0(dmu_objset_from_ds(ds, &os));
	else
	VERIFY0(dmu_objset_open_impl(spa, NULL, bp, &os));

	mdn = DMU_META_DNODE(os);

	dnode_allocate(mdn, DMU_OT_DNODE, blksz, ibs, DMU_OT_NONE, 0,
	DNODE_MIN_SLOTS, tx);

	/*
	* We don't want to have to increase the meta-dnode's nlevels
	* later, because then we could do it in quiescing context while
	* we are also accessing it in open context.
	*
	* This precaution is not necessary for the MOS (ds == NULL),
	* because the MOS is only updated in syncing context.
	* This is most fortunate: the MOS is the only objset that
	* needs to be synced multiple times as spa_sync() iterates
	* to convergence, so minimizing its dn_nlevels matters.
	*/
	if (ds != NULL) {
	if (levels == 0) {
	levels = 1;

	/*
	* Determine the number of levels necessary for the
	* meta-dnode to contain DN_MAX_OBJECT dnodes. Note
	* that in order to ensure that we do not overflow
	* 64 bits, there has to be a nlevels that gives us a
	* number of blocks > DN_MAX_OBJECT but < 2^64.
	* Therefore, (mdn->dn_indblkshift - SPA_BLKPTRSHIFT)
	* (10) must be less than (64 - log2(DN_MAX_OBJECT))
	* (16).
	*/
	while ((uint64_t)mdn->dn_nblkptr <<
	(mdn->dn_datablkshift - DNODE_SHIFT + (levels - 1) *
	(mdn->dn_indblkshift - SPA_BLKPTRSHIFT)) <
	DN_MAX_OBJECT)
	levels++;
	}

	mdn->dn_next_nlevels[tx->tx_txg & TXG_MASK] =
	mdn->dn_nlevels = levels;
	}

	ASSERT(type != DMU_OST_NONE);
	ASSERT(type != DMU_OST_ANY);
	ASSERT(type < DMU_OST_NUMTYPES);
	os->os_phys->os_type = type;

	/*
	* Enable user accounting if it is enabled and this is not an
	* encrypted receive.
	*/
	if (dmu_objset_userused_enabled(os) &&
	(!os->os_encrypted \|\| !dmu_objset_is_receiving(os))) {
	os->os_phys->os_flags \|= OBJSET_FLAG_USERACCOUNTING_COMPLETE;
	if (dmu_objset_userobjused_enabled(os)) {
	ASSERT3P(ds, !=, NULL);
	ds->ds_feature_activation[
	SPA_FEATURE_USEROBJ_ACCOUNTING] = (void *)B_TRUE;
	os->os_phys->os_flags \|=
	OBJSET_FLAG_USEROBJACCOUNTING_COMPLETE;
	}
	if (dmu_objset_projectquota_enabled(os)) {
	ASSERT3P(ds, !=, NULL);
	ds->ds_feature_activation[
	SPA_FEATURE_PROJECT_QUOTA] = (void *)B_TRUE;
	os->os_phys->os_flags \|=
	OBJSET_FLAG_PROJECTQUOTA_COMPLETE;
	}
	os->os_flags = os->os_phys->os_flags;
	}

	dsl_dataset_dirty(ds, tx);

	return (os);
	}

	/* called from dsl for meta-objset */
	objset_t *
	dmu_objset_create_impl(spa_t spa, dsl_dataset_t ds, blkptr_t *bp,
	dmu_objset_type_t type, dmu_tx_t *tx)
	{
	return (dmu_objset_create_impl_dnstats(spa, ds, bp, type, 0, 0, 0, tx));
	}

	typedef struct dmu_objset_create_arg {
	const char *doca_name;
	cred_t *doca_cred;
	proc_t *doca_proc;
	void (doca_userfunc)(objset_t os, void *arg,
	cred_t cr, dmu_tx_t tx);
	void *doca_userarg;
	dmu_objset_type_t doca_type;
	uint64_t doca_flags;
	dsl_crypto_params_t *doca_dcp;
	} dmu_objset_create_arg_t;

	static int
	dmu_objset_create_check(void arg, dmu_tx_t tx)
	{
	dmu_objset_create_arg_t *doca = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dir_t *pdd;
	dsl_dataset_t *parentds;
	objset_t *parentos;
	const char *tail;
	int error;

	if (strchr(doca->doca_name, '@') != NULL)
	return (SET_ERROR(EINVAL));

	if (strlen(doca->doca_name) >= ZFS_MAX_DATASET_NAME_LEN)
	return (SET_ERROR(ENAMETOOLONG));

	if (dataset_nestcheck(doca->doca_name) != 0)
	return (SET_ERROR(ENAMETOOLONG));

	error = dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail);
	if (error != 0)
	return (error);
	if (tail == NULL) {
	dsl_dir_rele(pdd, FTAG);
	return (SET_ERROR(EEXIST));
	}

	error = dmu_objset_create_crypt_check(pdd, doca->doca_dcp, NULL);
	if (error != 0) {
	dsl_dir_rele(pdd, FTAG);
	return (error);
	}

	error = dsl_fs_ss_limit_check(pdd, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL,
	doca->doca_cred, doca->doca_proc);
	if (error != 0) {
	dsl_dir_rele(pdd, FTAG);
	return (error);
	}

	/* can't create below anything but filesystems (eg. no ZVOLs) */
	error = dsl_dataset_hold_obj(pdd->dd_pool,
	dsl_dir_phys(pdd)->dd_head_dataset_obj, FTAG, &parentds);
	if (error != 0) {
	dsl_dir_rele(pdd, FTAG);
	return (error);
	}
	error = dmu_objset_from_ds(parentds, &parentos);
	if (error != 0) {
	dsl_dataset_rele(parentds, FTAG);
	dsl_dir_rele(pdd, FTAG);
	return (error);
	}
	if (dmu_objset_type(parentos) != DMU_OST_ZFS) {
	dsl_dataset_rele(parentds, FTAG);
	dsl_dir_rele(pdd, FTAG);
	return (SET_ERROR(ZFS_ERR_WRONG_PARENT));
	}
	dsl_dataset_rele(parentds, FTAG);
	dsl_dir_rele(pdd, FTAG);

	return (error);
	}

	static void
	dmu_objset_create_sync(void arg, dmu_tx_t tx)
	{
	dmu_objset_create_arg_t *doca = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	spa_t *spa = dp->dp_spa;
	dsl_dir_t *pdd;
	const char *tail;
	dsl_dataset_t *ds;
	uint64_t obj;
	blkptr_t *bp;
	objset_t *os;
	zio_t *rzio;

	VERIFY0(dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail));

	obj = dsl_dataset_create_sync(pdd, tail, NULL, doca->doca_flags,
	doca->doca_cred, doca->doca_dcp, tx);

	VERIFY0(dsl_dataset_hold_obj_flags(pdd->dd_pool, obj,
	DS_HOLD_FLAG_DECRYPT, FTAG, &ds));
	rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
	bp = dsl_dataset_get_blkptr(ds);
	os = dmu_objset_create_impl(spa, ds, bp, doca->doca_type, tx);
	rrw_exit(&ds->ds_bp_rwlock, FTAG);

	if (doca->doca_userfunc != NULL) {
	doca->doca_userfunc(os, doca->doca_userarg,
	doca->doca_cred, tx);
	}

	/*
	* The doca_userfunc() may write out some data that needs to be
	* encrypted if the dataset is encrypted (specifically the root
	* directory). This data must be written out before the encryption
	* key mapping is removed by dsl_dataset_rele_flags(). Force the
	* I/O to occur immediately by invoking the relevant sections of
	* dsl_pool_sync().
	*/
	if (os->os_encrypted) {
	dsl_dataset_t *tmpds = NULL;
	boolean_t need_sync_done = B_FALSE;

	mutex_enter(&ds->ds_lock);
	ds->ds_owner = FTAG;
	mutex_exit(&ds->ds_lock);

	rzio = zio_root(spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
	tmpds = txg_list_remove_this(&dp->dp_dirty_datasets, ds,
	tx->tx_txg);
	if (tmpds != NULL) {
	dsl_dataset_sync(ds, rzio, tx);
	need_sync_done = B_TRUE;
	}
	VERIFY0(zio_wait(rzio));

	dmu_objset_sync_done(os, tx);
	taskq_wait(dp->dp_sync_taskq);
	if (txg_list_member(&dp->dp_dirty_datasets, ds, tx->tx_txg)) {
	ASSERT3P(ds->ds_key_mapping, !=, NULL);
	key_mapping_rele(spa, ds->ds_key_mapping, ds);
	}

	rzio = zio_root(spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
	tmpds = txg_list_remove_this(&dp->dp_dirty_datasets, ds,
	tx->tx_txg);
	if (tmpds != NULL) {
	dmu_buf_rele(ds->ds_dbuf, ds);
	dsl_dataset_sync(ds, rzio, tx);
	}
	VERIFY0(zio_wait(rzio));

	if (need_sync_done) {
	ASSERT3P(ds->ds_key_mapping, !=, NULL);
	key_mapping_rele(spa, ds->ds_key_mapping, ds);
	dsl_dataset_sync_done(ds, tx);
	dmu_buf_rele(ds->ds_dbuf, ds);
	}

	mutex_enter(&ds->ds_lock);
	ds->ds_owner = NULL;
	mutex_exit(&ds->ds_lock);
	}

	spa_history_log_internal_ds(ds, "create", tx, " ");

	dsl_dataset_rele_flags(ds, DS_HOLD_FLAG_DECRYPT, FTAG);
	dsl_dir_rele(pdd, FTAG);
	}

	int
	dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
	dsl_crypto_params_t dcp, dmu_objset_create_sync_func_t func, void arg)
	{
	dmu_objset_create_arg_t doca;
	dsl_crypto_params_t tmp_dcp = { 0 };

	doca.doca_name = name;
	doca.doca_cred = CRED();
	doca.doca_proc = curproc;
	doca.doca_flags = flags;
	doca.doca_userfunc = func;
	doca.doca_userarg = arg;
	doca.doca_type = type;

	/*
	* Some callers (mostly for testing) do not provide a dcp on their
	* own but various code inside the sync task will require it to be
	* allocated. Rather than adding NULL checks throughout this code
	* or adding dummy dcp's to all of the callers we simply create a
	* dummy one here and use that. This zero dcp will have the same
	* effect as asking for inheritance of all encryption params.
	*/
	doca.doca_dcp = (dcp != NULL) ? dcp : &tmp_dcp;

	int rv = dsl_sync_task(name,
	dmu_objset_create_check, dmu_objset_create_sync, &doca,
	6, ZFS_SPACE_CHECK_NORMAL);

	if (rv == 0)
	zvol_create_minor(name);
	return (rv);
	}

	typedef struct dmu_objset_clone_arg {
	const char *doca_clone;
	const char *doca_origin;
	cred_t *doca_cred;
	proc_t *doca_proc;
	} dmu_objset_clone_arg_t;

	static int
	dmu_objset_clone_check(void arg, dmu_tx_t tx)
	{
	dmu_objset_clone_arg_t *doca = arg;
	dsl_dir_t *pdd;
	const char *tail;
	int error;
	dsl_dataset_t *origin;
	dsl_pool_t *dp = dmu_tx_pool(tx);

	if (strchr(doca->doca_clone, '@') != NULL)
	return (SET_ERROR(EINVAL));

	if (strlen(doca->doca_clone) >= ZFS_MAX_DATASET_NAME_LEN)
	return (SET_ERROR(ENAMETOOLONG));

	error = dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail);
	if (error != 0)
	return (error);
	if (tail == NULL) {
	dsl_dir_rele(pdd, FTAG);
	return (SET_ERROR(EEXIST));
	}

	error = dsl_fs_ss_limit_check(pdd, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL,
	doca->doca_cred, doca->doca_proc);
	if (error != 0) {
	dsl_dir_rele(pdd, FTAG);
	return (SET_ERROR(EDQUOT));
	}

	error = dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin);
	if (error != 0) {
	dsl_dir_rele(pdd, FTAG);
	return (error);
	}

	/* You can only clone snapshots, not the head datasets. */
	if (!origin->ds_is_snapshot) {
	dsl_dataset_rele(origin, FTAG);
	dsl_dir_rele(pdd, FTAG);
	return (SET_ERROR(EINVAL));
	}

	dsl_dataset_rele(origin, FTAG);
	dsl_dir_rele(pdd, FTAG);

	return (0);
	}

	static void
	dmu_objset_clone_sync(void arg, dmu_tx_t tx)
	{
	dmu_objset_clone_arg_t *doca = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	dsl_dir_t *pdd;
	const char *tail;
	dsl_dataset_t origin, ds;
	uint64_t obj;
	char namebuf[ZFS_MAX_DATASET_NAME_LEN];

	VERIFY0(dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail));
	VERIFY0(dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin));

	obj = dsl_dataset_create_sync(pdd, tail, origin, 0,
	doca->doca_cred, NULL, tx);

	VERIFY0(dsl_dataset_hold_obj(pdd->dd_pool, obj, FTAG, &ds));
	dsl_dataset_name(origin, namebuf);
	spa_history_log_internal_ds(ds, "clone", tx,
	"origin=%s (%llu)", namebuf, (u_longlong_t)origin->ds_object);
	dsl_dataset_rele(ds, FTAG);
	dsl_dataset_rele(origin, FTAG);
	dsl_dir_rele(pdd, FTAG);
	}

	int
	dmu_objset_clone(const char clone, const char origin)
	{
	dmu_objset_clone_arg_t doca;

	doca.doca_clone = clone;
	doca.doca_origin = origin;
	doca.doca_cred = CRED();
	doca.doca_proc = curproc;

	int rv = dsl_sync_task(clone,
	dmu_objset_clone_check, dmu_objset_clone_sync, &doca,
	6, ZFS_SPACE_CHECK_NORMAL);

	if (rv == 0)
	zvol_create_minor(clone);

	return (rv);
	}

	int
	dmu_objset_snapshot_one(const char fsname, const char snapname)
	{
	int err;
	char *longsnap = kmem_asprintf("%s@%s", fsname, snapname);
	nvlist_t *snaps = fnvlist_alloc();

	fnvlist_add_boolean(snaps, longsnap);
	kmem_strfree(longsnap);
	err = dsl_dataset_snapshot(snaps, NULL, NULL);
	fnvlist_free(snaps);
	return (err);
	}

	static void
	dmu_objset_upgrade_task_cb(void *data)
	{
	objset_t *os = data;

	mutex_enter(&os->os_upgrade_lock);
	os->os_upgrade_status = EINTR;
	if (!os->os_upgrade_exit) {
	int status;

	mutex_exit(&os->os_upgrade_lock);

	status = os->os_upgrade_cb(os);

	mutex_enter(&os->os_upgrade_lock);

	os->os_upgrade_status = status;
	}
	os->os_upgrade_exit = B_TRUE;
	os->os_upgrade_id = 0;
	mutex_exit(&os->os_upgrade_lock);
	dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag);
	}

	static void
	dmu_objset_upgrade(objset_t *os, dmu_objset_upgrade_cb_t cb)
	{
	if (os->os_upgrade_id != 0)
	return;

	ASSERT(dsl_pool_config_held(dmu_objset_pool(os)));
	dsl_dataset_long_hold(dmu_objset_ds(os), upgrade_tag);

	mutex_enter(&os->os_upgrade_lock);
	if (os->os_upgrade_id == 0 && os->os_upgrade_status == 0) {
	os->os_upgrade_exit = B_FALSE;
	os->os_upgrade_cb = cb;
	os->os_upgrade_id = taskq_dispatch(
	os->os_spa->spa_upgrade_taskq,
	dmu_objset_upgrade_task_cb, os, TQ_SLEEP);
	if (os->os_upgrade_id == TASKQID_INVALID) {
	dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag);
	os->os_upgrade_status = ENOMEM;
	}
	} else {
	dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag);
	}
	mutex_exit(&os->os_upgrade_lock);
	}

	static void
	dmu_objset_upgrade_stop(objset_t *os)
	{
	mutex_enter(&os->os_upgrade_lock);
	os->os_upgrade_exit = B_TRUE;
	if (os->os_upgrade_id != 0) {
	taskqid_t id = os->os_upgrade_id;

	os->os_upgrade_id = 0;
	mutex_exit(&os->os_upgrade_lock);

	if ((taskq_cancel_id(os->os_spa->spa_upgrade_taskq, id)) == 0) {
	dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag);
	}
	txg_wait_synced(os->os_spa->spa_dsl_pool, 0);
	} else {
	mutex_exit(&os->os_upgrade_lock);
	}
	}

	static void
	dmu_objset_sync_dnodes(multilist_sublist_t list, dmu_tx_t tx)
	{
	dnode_t *dn;

	while ((dn = multilist_sublist_head(list)) != NULL) {
	ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT);
	ASSERT(dn->dn_dbuf->db_data_pending);
	/*
	* Initialize dn_zio outside dnode_sync() because the
	* meta-dnode needs to set it outside dnode_sync().
	*/
	dn->dn_zio = dn->dn_dbuf->db_data_pending->dr_zio;
	ASSERT(dn->dn_zio);

	ASSERT3U(dn->dn_nlevels, <=, DN_MAX_LEVELS);
	multilist_sublist_remove(list, dn);

	/*
	* See the comment above dnode_rele_task() for an explanation
	* of why this dnode hold is always needed (even when not
	* doing user accounting).
	*/
	multilist_t *newlist = &dn->dn_objset->os_synced_dnodes;
	(void) dnode_add_ref(dn, newlist);
	multilist_insert(newlist, dn);

	dnode_sync(dn, tx);
	}
	}

	static void
	dmu_objset_write_ready(zio_t zio, arc_buf_t abuf, void *arg)
	{
	(void) abuf;
	blkptr_t *bp = zio->io_bp;
	objset_t *os = arg;
	dnode_phys_t *dnp = &os->os_phys->os_meta_dnode;
	uint64_t fill = 0;

	ASSERT(!BP_IS_EMBEDDED(bp));
	ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_OBJSET);
	ASSERT0(BP_GET_LEVEL(bp));

	/*
	* Update rootbp fill count: it should be the number of objects
	* allocated in the object set (not counting the "special"
	* objects that are stored in the objset_phys_t -- the meta
	* dnode and user/group/project accounting objects).
	*/
	for (int i = 0; i < dnp->dn_nblkptr; i++)
	fill += BP_GET_FILL(&dnp->dn_blkptr[i]);

	BP_SET_FILL(bp, fill);

	if (os->os_dsl_dataset != NULL)
	rrw_enter(&os->os_dsl_dataset->ds_bp_rwlock, RW_WRITER, FTAG);
	os->os_rootbp = bp;
	if (os->os_dsl_dataset != NULL)
	rrw_exit(&os->os_dsl_dataset->ds_bp_rwlock, FTAG);
	}

	static void
	dmu_objset_write_done(zio_t zio, arc_buf_t abuf, void *arg)
	{
	(void) abuf;
	blkptr_t *bp = zio->io_bp;
	blkptr_t *bp_orig = &zio->io_bp_orig;
	objset_t *os = arg;

	if (zio->io_flags & ZIO_FLAG_IO_REWRITE) {
	ASSERT(BP_EQUAL(bp, bp_orig));
	} else {
	dsl_dataset_t *ds = os->os_dsl_dataset;
	dmu_tx_t *tx = os->os_synctx;

	(void) dsl_dataset_block_kill(ds, bp_orig, tx, B_TRUE);
	dsl_dataset_block_born(ds, bp, tx);
	}
	kmem_free(bp, sizeof (*bp));
	}

	typedef struct sync_objset_arg {
	zio_t *soa_zio;
	objset_t *soa_os;
	dmu_tx_t *soa_tx;
	kmutex_t soa_mutex;
	int soa_count;
	taskq_ent_t soa_tq_ent;
	} sync_objset_arg_t;

	typedef struct sync_dnodes_arg {
	multilist_t *sda_list;
	int sda_sublist_idx;
	multilist_t *sda_newlist;
	sync_objset_arg_t *sda_soa;
	} sync_dnodes_arg_t;

	static void sync_meta_dnode_task(void *arg);

	static void
	sync_dnodes_task(void *arg)
	{
	sync_dnodes_arg_t *sda = arg;
	sync_objset_arg_t *soa = sda->sda_soa;
	objset_t *os = soa->soa_os;

	uint_t allocator = spa_acq_allocator(os->os_spa);
	multilist_sublist_t *ms =
	multilist_sublist_lock_idx(sda->sda_list, sda->sda_sublist_idx);

	dmu_objset_sync_dnodes(ms, soa->soa_tx);

	multilist_sublist_unlock(ms);
	spa_rel_allocator(os->os_spa, allocator);

	kmem_free(sda, sizeof (*sda));

	mutex_enter(&soa->soa_mutex);
	ASSERT(soa->soa_count != 0);
	if (--soa->soa_count != 0) {
	mutex_exit(&soa->soa_mutex);
	return;
	}
	mutex_exit(&soa->soa_mutex);

	taskq_dispatch_ent(dmu_objset_pool(os)->dp_sync_taskq,
	sync_meta_dnode_task, soa, TQ_FRONT, &soa->soa_tq_ent);
	}

	/*
	* Issue the zio_nowait() for all dirty record zios on the meta dnode,
	* then trigger the callback for the zil_sync. This runs once for each
	* objset, only after any/all sublists in the objset have been synced.
	*/
	static void
	sync_meta_dnode_task(void *arg)
	{
	sync_objset_arg_t *soa = arg;
	objset_t *os = soa->soa_os;
	dmu_tx_t *tx = soa->soa_tx;
	int txgoff = tx->tx_txg & TXG_MASK;
	dbuf_dirty_record_t *dr;

	ASSERT0(soa->soa_count);

	list_t *list = &DMU_META_DNODE(os)->dn_dirty_records[txgoff];
	while ((dr = list_remove_head(list)) != NULL) {
	ASSERT0(dr->dr_dbuf->db_level);
	zio_nowait(dr->dr_zio);
	}

	/* Enable dnode backfill if enough objects have been freed. */
	if (os->os_freed_dnodes >= dmu_rescan_dnode_threshold) {
	os->os_rescan_dnodes = B_TRUE;
	os->os_freed_dnodes = 0;
	}

	/*
	* Free intent log blocks up to this tx.
	*/
	zil_sync(os->os_zil, tx);
	os->os_phys->os_zil_header = os->os_zil_header;
	zio_nowait(soa->soa_zio);

	mutex_destroy(&soa->soa_mutex);
	kmem_free(soa, sizeof (*soa));
	}

	/* called from dsl */
	void
	dmu_objset_sync(objset_t os, zio_t pio, dmu_tx_t *tx)
	{
	int txgoff;
	zbookmark_phys_t zb;
	zio_prop_t zp;
	zio_t *zio;
	int num_sublists;
	multilist_t *ml;
	blkptr_t blkptr_copy = kmem_alloc(sizeof (os->os_rootbp), KM_SLEEP);
	blkptr_copy = os->os_rootbp;

	dprintf_ds(os->os_dsl_dataset, "txg=%llu\n", (u_longlong_t)tx->tx_txg);

	ASSERT(dmu_tx_is_syncing(tx));
	/* XXX the write_done callback should really give us the tx... */
	os->os_synctx = tx;

	if (os->os_dsl_dataset == NULL) {
	/*
	* This is the MOS. If we have upgraded,
	* spa_max_replication() could change, so reset
	* os_copies here.
	*/
	os->os_copies = spa_max_replication(os->os_spa);
	}

	/*
	* Create the root block IO
	*/
	SET_BOOKMARK(&zb, os->os_dsl_dataset ?
	os->os_dsl_dataset->ds_object : DMU_META_OBJSET,
	ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID);
	arc_release(os->os_phys_buf, &os->os_phys_buf);

	dmu_write_policy(os, NULL, 0, 0, &zp);

	/*
	* If we are either claiming the ZIL or doing a raw receive, write
	* out the os_phys_buf raw. Neither of these actions will effect the
	* MAC at this point.
	*/
	if (os->os_raw_receive \|\|
	os->os_next_write_raw[tx->tx_txg & TXG_MASK]) {
	ASSERT(os->os_encrypted);
	arc_convert_to_raw(os->os_phys_buf,
	os->os_dsl_dataset->ds_object, ZFS_HOST_BYTEORDER,
	DMU_OT_OBJSET, NULL, NULL, NULL);
	}

	zio = arc_write(pio, os->os_spa, tx->tx_txg,
	blkptr_copy, os->os_phys_buf, B_FALSE, dmu_os_is_l2cacheable(os),
	&zp, dmu_objset_write_ready, NULL, dmu_objset_write_done,
	os, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb);

	/*
	* Sync special dnodes - the parent IO for the sync is the root block
	*/
	DMU_META_DNODE(os)->dn_zio = zio;
	dnode_sync(DMU_META_DNODE(os), tx);

	os->os_phys->os_flags = os->os_flags;

	if (DMU_USERUSED_DNODE(os) &&
	DMU_USERUSED_DNODE(os)->dn_type != DMU_OT_NONE) {
	DMU_USERUSED_DNODE(os)->dn_zio = zio;
	dnode_sync(DMU_USERUSED_DNODE(os), tx);
	DMU_GROUPUSED_DNODE(os)->dn_zio = zio;
	dnode_sync(DMU_GROUPUSED_DNODE(os), tx);
	}

	if (DMU_PROJECTUSED_DNODE(os) &&
	DMU_PROJECTUSED_DNODE(os)->dn_type != DMU_OT_NONE) {
	DMU_PROJECTUSED_DNODE(os)->dn_zio = zio;
	dnode_sync(DMU_PROJECTUSED_DNODE(os), tx);
	}

	txgoff = tx->tx_txg & TXG_MASK;

	/*
	* We must create the list here because it uses the
	* dn_dirty_link[] of this txg. But it may already
	* exist because we call dsl_dataset_sync() twice per txg.
	*/
	if (os->os_synced_dnodes.ml_sublists == NULL) {
	multilist_create(&os->os_synced_dnodes, sizeof (dnode_t),
	offsetof(dnode_t, dn_dirty_link[txgoff]),
	dnode_multilist_index_func);
	} else {
	ASSERT3U(os->os_synced_dnodes.ml_offset, ==,
	offsetof(dnode_t, dn_dirty_link[txgoff]));
	}

	/*
	* zio_nowait(zio) is done after any/all sublist and meta dnode
	* zios have been nowaited, and the zil_sync() has been performed.
	* The soa is freed at the end of sync_meta_dnode_task.
	*/
	sync_objset_arg_t soa = kmem_alloc(sizeof (soa), KM_SLEEP);
	soa->soa_zio = zio;
	soa->soa_os = os;
	soa->soa_tx = tx;
	taskq_init_ent(&soa->soa_tq_ent);
	mutex_init(&soa->soa_mutex, NULL, MUTEX_DEFAULT, NULL);

	ml = &os->os_dirty_dnodes[txgoff];
	soa->soa_count = num_sublists = multilist_get_num_sublists(ml);

	for (int i = 0; i < num_sublists; i++) {
	if (multilist_sublist_is_empty_idx(ml, i))
	soa->soa_count--;
	}

	if (soa->soa_count == 0) {
	taskq_dispatch_ent(dmu_objset_pool(os)->dp_sync_taskq,
	sync_meta_dnode_task, soa, TQ_FRONT, &soa->soa_tq_ent);
	} else {
	/*
	* Sync sublists in parallel. The last to finish
	* (i.e., when soa->soa_count reaches zero) must
	* dispatch sync_meta_dnode_task.
	*/
	for (int i = 0; i < num_sublists; i++) {
	if (multilist_sublist_is_empty_idx(ml, i))
	continue;
	sync_dnodes_arg_t *sda =
	kmem_alloc(sizeof (*sda), KM_SLEEP);
	sda->sda_list = ml;
	sda->sda_sublist_idx = i;
	sda->sda_soa = soa;
	(void) taskq_dispatch(
	dmu_objset_pool(os)->dp_sync_taskq,
	sync_dnodes_task, sda, 0);
	/* sync_dnodes_task frees sda */
	}
	}
	}

	boolean_t
	dmu_objset_is_dirty(objset_t *os, uint64_t txg)
	{
	return (!multilist_is_empty(&os->os_dirty_dnodes[txg & TXG_MASK]));
	}

	static file_info_cb_t *file_cbs[DMU_OST_NUMTYPES];

	void
	dmu_objset_register_type(dmu_objset_type_t ost, file_info_cb_t *cb)
	{
	file_cbs[ost] = cb;
	}

	int
	dmu_get_file_info(objset_t os, dmu_object_type_t bonustype, const void data,
	zfs_file_info_t *zfi)
	{
	file_info_cb_t *cb = file_cbs[os->os_phys->os_type];
	if (cb == NULL)
	return (EINVAL);
	return (cb(bonustype, data, zfi));
	}

	boolean_t
	dmu_objset_userused_enabled(objset_t *os)
	{
	return (spa_version(os->os_spa) >= SPA_VERSION_USERSPACE &&
	file_cbs[os->os_phys->os_type] != NULL &&
	DMU_USERUSED_DNODE(os) != NULL);
	}

	boolean_t
	dmu_objset_userobjused_enabled(objset_t *os)
	{
	return (dmu_objset_userused_enabled(os) &&
	spa_feature_is_enabled(os->os_spa, SPA_FEATURE_USEROBJ_ACCOUNTING));
	}

	boolean_t
	dmu_objset_projectquota_enabled(objset_t *os)
	{
	return (file_cbs[os->os_phys->os_type] != NULL &&
	DMU_PROJECTUSED_DNODE(os) != NULL &&
	spa_feature_is_enabled(os->os_spa, SPA_FEATURE_PROJECT_QUOTA));
	}

	typedef struct userquota_node {
	/* must be in the first filed, see userquota_update_cache() */
	char uqn_id[20 + DMU_OBJACCT_PREFIX_LEN];
	int64_t uqn_delta;
	avl_node_t uqn_node;
	} userquota_node_t;

	typedef struct userquota_cache {
	avl_tree_t uqc_user_deltas;
	avl_tree_t uqc_group_deltas;
	avl_tree_t uqc_project_deltas;
	} userquota_cache_t;

	static int
	userquota_compare(const void l, const void r)
	{
	const userquota_node_t *luqn = l;
	const userquota_node_t *ruqn = r;
	int rv;

	/*
	* NB: can only access uqn_id because userquota_update_cache() doesn't
	* pass in an entire userquota_node_t.
	*/
	rv = strcmp(luqn->uqn_id, ruqn->uqn_id);

	return (TREE_ISIGN(rv));
	}

	static void
	do_userquota_cacheflush(objset_t os, userquota_cache_t cache, dmu_tx_t *tx)
	{
	void *cookie;
	userquota_node_t *uqn;

	ASSERT(dmu_tx_is_syncing(tx));

	cookie = NULL;
	while ((uqn = avl_destroy_nodes(&cache->uqc_user_deltas,
	&cookie)) != NULL) {
	/*
	* os_userused_lock protects against concurrent calls to
	* zap_increment_int(). It's needed because zap_increment_int()
	* is not thread-safe (i.e. not atomic).
	*/
	mutex_enter(&os->os_userused_lock);
	VERIFY0(zap_increment(os, DMU_USERUSED_OBJECT,
	uqn->uqn_id, uqn->uqn_delta, tx));
	mutex_exit(&os->os_userused_lock);
	kmem_free(uqn, sizeof (*uqn));
	}
	avl_destroy(&cache->uqc_user_deltas);

	cookie = NULL;
	while ((uqn = avl_destroy_nodes(&cache->uqc_group_deltas,
	&cookie)) != NULL) {
	mutex_enter(&os->os_userused_lock);
	VERIFY0(zap_increment(os, DMU_GROUPUSED_OBJECT,
	uqn->uqn_id, uqn->uqn_delta, tx));
	mutex_exit(&os->os_userused_lock);
	kmem_free(uqn, sizeof (*uqn));
	}
	avl_destroy(&cache->uqc_group_deltas);

	if (dmu_objset_projectquota_enabled(os)) {
	cookie = NULL;
	while ((uqn = avl_destroy_nodes(&cache->uqc_project_deltas,
	&cookie)) != NULL) {
	mutex_enter(&os->os_userused_lock);
	VERIFY0(zap_increment(os, DMU_PROJECTUSED_OBJECT,
	uqn->uqn_id, uqn->uqn_delta, tx));
	mutex_exit(&os->os_userused_lock);
	kmem_free(uqn, sizeof (*uqn));
	}
	avl_destroy(&cache->uqc_project_deltas);
	}
	}

	static void
	userquota_update_cache(avl_tree_t avl, const char id, int64_t delta)
	{
	userquota_node_t *uqn;
	avl_index_t idx;

	ASSERT(strlen(id) < sizeof (uqn->uqn_id));
	/*
	* Use id directly for searching because uqn_id is the first field of
	* userquota_node_t and fields after uqn_id won't be accessed in
	* avl_find().
	*/
	uqn = avl_find(avl, (const void *)id, &idx);
	if (uqn == NULL) {
	uqn = kmem_zalloc(sizeof (*uqn), KM_SLEEP);
	strlcpy(uqn->uqn_id, id, sizeof (uqn->uqn_id));
	avl_insert(avl, uqn, idx);
	}
	uqn->uqn_delta += delta;
	}

	static void
	do_userquota_update(objset_t os, userquota_cache_t cache, uint64_t used,
	uint64_t flags, uint64_t user, uint64_t group, uint64_t project,
	boolean_t subtract)
	{
	if (flags & DNODE_FLAG_USERUSED_ACCOUNTED) {
	int64_t delta = DNODE_MIN_SIZE + used;
	char name[20];

	if (subtract)
	delta = -delta;

	(void) snprintf(name, sizeof (name), "%llx", (longlong_t)user);
	userquota_update_cache(&cache->uqc_user_deltas, name, delta);

	(void) snprintf(name, sizeof (name), "%llx", (longlong_t)group);
	userquota_update_cache(&cache->uqc_group_deltas, name, delta);

	if (dmu_objset_projectquota_enabled(os)) {
	(void) snprintf(name, sizeof (name), "%llx",
	(longlong_t)project);
	userquota_update_cache(&cache->uqc_project_deltas,
	name, delta);
	}
	}
	}

	static void
	do_userobjquota_update(objset_t os, userquota_cache_t cache, uint64_t flags,
	uint64_t user, uint64_t group, uint64_t project, boolean_t subtract)
	{
	if (flags & DNODE_FLAG_USEROBJUSED_ACCOUNTED) {
	char name[20 + DMU_OBJACCT_PREFIX_LEN];
	int delta = subtract ? -1 : 1;

	(void) snprintf(name, sizeof (name), DMU_OBJACCT_PREFIX "%llx",
	(longlong_t)user);
	userquota_update_cache(&cache->uqc_user_deltas, name, delta);

	(void) snprintf(name, sizeof (name), DMU_OBJACCT_PREFIX "%llx",
	(longlong_t)group);
	userquota_update_cache(&cache->uqc_group_deltas, name, delta);

	if (dmu_objset_projectquota_enabled(os)) {
	(void) snprintf(name, sizeof (name),
	DMU_OBJACCT_PREFIX "%llx", (longlong_t)project);
	userquota_update_cache(&cache->uqc_project_deltas,
	name, delta);
	}
	}
	}

	typedef struct userquota_updates_arg {
	objset_t *uua_os;
	int uua_sublist_idx;
	dmu_tx_t *uua_tx;
	} userquota_updates_arg_t;

	static void
	userquota_updates_task(void *arg)
	{
	userquota_updates_arg_t *uua = arg;
	objset_t *os = uua->uua_os;
	dmu_tx_t *tx = uua->uua_tx;
	dnode_t *dn;
	userquota_cache_t cache = { { 0 } };

	multilist_sublist_t *list = multilist_sublist_lock_idx(
	&os->os_synced_dnodes, uua->uua_sublist_idx);

	ASSERT(multilist_sublist_head(list) == NULL \|\|
	dmu_objset_userused_enabled(os));
	avl_create(&cache.uqc_user_deltas, userquota_compare,
	sizeof (userquota_node_t), offsetof(userquota_node_t, uqn_node));
	avl_create(&cache.uqc_group_deltas, userquota_compare,
	sizeof (userquota_node_t), offsetof(userquota_node_t, uqn_node));
	if (dmu_objset_projectquota_enabled(os))
	avl_create(&cache.uqc_project_deltas, userquota_compare,
	sizeof (userquota_node_t), offsetof(userquota_node_t,
	uqn_node));

	while ((dn = multilist_sublist_head(list)) != NULL) {
	int flags;
	ASSERT(!DMU_OBJECT_IS_SPECIAL(dn->dn_object));
	ASSERT(dn->dn_phys->dn_type == DMU_OT_NONE \|\|
	dn->dn_phys->dn_flags &
	DNODE_FLAG_USERUSED_ACCOUNTED);

	flags = dn->dn_id_flags;
	ASSERT(flags);
	if (flags & DN_ID_OLD_EXIST) {
	do_userquota_update(os, &cache, dn->dn_oldused,
	dn->dn_oldflags, dn->dn_olduid, dn->dn_oldgid,
	dn->dn_oldprojid, B_TRUE);
	do_userobjquota_update(os, &cache, dn->dn_oldflags,
	dn->dn_olduid, dn->dn_oldgid,
	dn->dn_oldprojid, B_TRUE);
	}
	if (flags & DN_ID_NEW_EXIST) {
	do_userquota_update(os, &cache,
	DN_USED_BYTES(dn->dn_phys), dn->dn_phys->dn_flags,
	dn->dn_newuid, dn->dn_newgid,
	dn->dn_newprojid, B_FALSE);
	do_userobjquota_update(os, &cache,
	dn->dn_phys->dn_flags, dn->dn_newuid, dn->dn_newgid,
	dn->dn_newprojid, B_FALSE);
	}

	mutex_enter(&dn->dn_mtx);
	dn->dn_oldused = 0;
	dn->dn_oldflags = 0;
	if (dn->dn_id_flags & DN_ID_NEW_EXIST) {
	dn->dn_olduid = dn->dn_newuid;
	dn->dn_oldgid = dn->dn_newgid;
	dn->dn_oldprojid = dn->dn_newprojid;
	dn->dn_id_flags \|= DN_ID_OLD_EXIST;
	if (dn->dn_bonuslen == 0)
	dn->dn_id_flags \|= DN_ID_CHKED_SPILL;
	else
	dn->dn_id_flags \|= DN_ID_CHKED_BONUS;
	}
	dn->dn_id_flags &= ~(DN_ID_NEW_EXIST);
	mutex_exit(&dn->dn_mtx);

	multilist_sublist_remove(list, dn);
	dnode_rele(dn, &os->os_synced_dnodes);
	}
	do_userquota_cacheflush(os, &cache, tx);
	multilist_sublist_unlock(list);
	kmem_free(uua, sizeof (*uua));
	}

	/*
	* Release dnode holds from dmu_objset_sync_dnodes(). When the dnode is being
	* synced (i.e. we have issued the zio's for blocks in the dnode), it can't be
	* evicted because the block containing the dnode can't be evicted until it is
	* written out. However, this hold is necessary to prevent the dnode_t from
	* being moved (via dnode_move()) while it's still referenced by
	* dbuf_dirty_record_t:dr_dnode. And dr_dnode is needed for
	* dirty_lightweight_leaf-type dirty records.
	*
	* If we are doing user-object accounting, the dnode_rele() happens from
	* userquota_updates_task() instead.
	*/
	static void
	dnode_rele_task(void *arg)
	{
	userquota_updates_arg_t *uua = arg;
	objset_t *os = uua->uua_os;

	multilist_sublist_t *list = multilist_sublist_lock_idx(
	&os->os_synced_dnodes, uua->uua_sublist_idx);

	dnode_t *dn;
	while ((dn = multilist_sublist_head(list)) != NULL) {
	multilist_sublist_remove(list, dn);
	dnode_rele(dn, &os->os_synced_dnodes);
	}
	multilist_sublist_unlock(list);
	kmem_free(uua, sizeof (*uua));
	}

	/*
	* Return TRUE if userquota updates are needed.
	*/
	static boolean_t
	dmu_objset_do_userquota_updates_prep(objset_t os, dmu_tx_t tx)
	{
	if (!dmu_objset_userused_enabled(os))
	return (B_FALSE);

	/*
	* If this is a raw receive just return and handle accounting
	* later when we have the keys loaded. We also don't do user
	* accounting during claiming since the datasets are not owned
	* for the duration of claiming and this txg should only be
	* used for recovery.
	*/
	if (os->os_encrypted && dmu_objset_is_receiving(os))
	return (B_FALSE);

	if (tx->tx_txg <= os->os_spa->spa_claim_max_txg)
	return (B_FALSE);

	/* Allocate the user/group/project used objects if necessary. */
	if (DMU_USERUSED_DNODE(os)->dn_type == DMU_OT_NONE) {
	VERIFY0(zap_create_claim(os,
	DMU_USERUSED_OBJECT,
	DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx));
	VERIFY0(zap_create_claim(os,
	DMU_GROUPUSED_OBJECT,
	DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx));
	}

	if (dmu_objset_projectquota_enabled(os) &&
	DMU_PROJECTUSED_DNODE(os)->dn_type == DMU_OT_NONE) {
	VERIFY0(zap_create_claim(os, DMU_PROJECTUSED_OBJECT,
	DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx));
	}
	return (B_TRUE);
	}

	/*
	* Dispatch taskq tasks to dp_sync_taskq to update the user accounting, and
	* also release the holds on the dnodes from dmu_objset_sync_dnodes().
	* The caller must taskq_wait(dp_sync_taskq).
	*/
	void
	dmu_objset_sync_done(objset_t os, dmu_tx_t tx)
	{
	boolean_t need_userquota = dmu_objset_do_userquota_updates_prep(os, tx);

	int num_sublists = multilist_get_num_sublists(&os->os_synced_dnodes);
	for (int i = 0; i < num_sublists; i++) {
	userquota_updates_arg_t *uua =
	kmem_alloc(sizeof (*uua), KM_SLEEP);
	uua->uua_os = os;
	uua->uua_sublist_idx = i;
	uua->uua_tx = tx;

	/*
	* If we don't need to update userquotas, use
	* dnode_rele_task() to call dnode_rele()
	*/
	(void) taskq_dispatch(dmu_objset_pool(os)->dp_sync_taskq,
	need_userquota ? userquota_updates_task : dnode_rele_task,
	uua, 0);
	/* callback frees uua */
	}
	}


	/*
	* Returns a pointer to data to find uid/gid from
	*
	* If a dirty record for transaction group that is syncing can't
	* be found then NULL is returned. In the NULL case it is assumed
	* the uid/gid aren't changing.
	*/
	static void *
	dmu_objset_userquota_find_data(dmu_buf_impl_t db, dmu_tx_t tx)
	{
	dbuf_dirty_record_t *dr;
	void *data;

	if (db->db_dirtycnt == 0)
	return (db->db.db_data); /* Nothing is changing */

	dr = dbuf_find_dirty_eq(db, tx->tx_txg);

	if (dr == NULL) {
	data = NULL;
	} else {
	if (dr->dr_dnode->dn_bonuslen == 0 &&
	dr->dr_dbuf->db_blkid == DMU_SPILL_BLKID)
	data = dr->dt.dl.dr_data->b_data;
	else
	data = dr->dt.dl.dr_data;
	}

	return (data);
	}

	void
	dmu_objset_userquota_get_ids(dnode_t dn, boolean_t before, dmu_tx_t tx)
	{
	objset_t *os = dn->dn_objset;
	void *data = NULL;
	dmu_buf_impl_t *db = NULL;
	int flags = dn->dn_id_flags;
	int error;
	boolean_t have_spill = B_FALSE;

	if (!dmu_objset_userused_enabled(dn->dn_objset))
	return;

	/*
	* Raw receives introduce a problem with user accounting. Raw
	* receives cannot update the user accounting info because the
	* user ids and the sizes are encrypted. To guarantee that we
	* never end up with bad user accounting, we simply disable it
	* during raw receives. We also disable this for normal receives
	* so that an incremental raw receive may be done on top of an
	* existing non-raw receive.
	*/
	if (os->os_encrypted && dmu_objset_is_receiving(os))
	return;

	if (before && (flags & (DN_ID_CHKED_BONUS\|DN_ID_OLD_EXIST\|
	DN_ID_CHKED_SPILL)))
	return;

	if (before && dn->dn_bonuslen != 0)
	data = DN_BONUS(dn->dn_phys);
	else if (!before && dn->dn_bonuslen != 0) {
	if (dn->dn_bonus) {
	db = dn->dn_bonus;
	mutex_enter(&db->db_mtx);
	data = dmu_objset_userquota_find_data(db, tx);
	} else {
	data = DN_BONUS(dn->dn_phys);
	}
	} else if (dn->dn_bonuslen == 0 && dn->dn_bonustype == DMU_OT_SA) {
	int rf = 0;

	if (RW_WRITE_HELD(&dn->dn_struct_rwlock))
	rf \|= DB_RF_HAVESTRUCT;
	error = dmu_spill_hold_by_dnode(dn,
	rf \| DB_RF_MUST_SUCCEED,
	FTAG, (dmu_buf_t **)&db);
	ASSERT(error == 0);
	mutex_enter(&db->db_mtx);
	data = (before) ? db->db.db_data :
	dmu_objset_userquota_find_data(db, tx);
	have_spill = B_TRUE;
	} else {
	mutex_enter(&dn->dn_mtx);
	dn->dn_id_flags \|= DN_ID_CHKED_BONUS;
	mutex_exit(&dn->dn_mtx);
	return;
	}

	/*
	* Must always call the callback in case the object
	* type has changed and that type isn't an object type to track
	*/
	zfs_file_info_t zfi;
	error = file_cbs[os->os_phys->os_type](dn->dn_bonustype, data, &zfi);

	if (before) {
	ASSERT(data);
	dn->dn_olduid = zfi.zfi_user;
	dn->dn_oldgid = zfi.zfi_group;
	dn->dn_oldprojid = zfi.zfi_project;
	} else if (data) {
	dn->dn_newuid = zfi.zfi_user;
	dn->dn_newgid = zfi.zfi_group;
	dn->dn_newprojid = zfi.zfi_project;
	}

	/*
	* Preserve existing uid/gid when the callback can't determine
	* what the new uid/gid are and the callback returned EEXIST.
	* The EEXIST error tells us to just use the existing uid/gid.
	* If we don't know what the old values are then just assign
	* them to 0, since that is a new file being created.
	*/
	if (!before && data == NULL && error == EEXIST) {
	if (flags & DN_ID_OLD_EXIST) {
	dn->dn_newuid = dn->dn_olduid;
	dn->dn_newgid = dn->dn_oldgid;
	dn->dn_newprojid = dn->dn_oldprojid;
	} else {
	dn->dn_newuid = 0;
	dn->dn_newgid = 0;
	dn->dn_newprojid = ZFS_DEFAULT_PROJID;
	}
	error = 0;
	}

	if (db)
	mutex_exit(&db->db_mtx);

	mutex_enter(&dn->dn_mtx);
	if (error == 0 && before)
	dn->dn_id_flags \|= DN_ID_OLD_EXIST;
	if (error == 0 && !before)
	dn->dn_id_flags \|= DN_ID_NEW_EXIST;

	if (have_spill) {
	dn->dn_id_flags \|= DN_ID_CHKED_SPILL;
	} else {
	dn->dn_id_flags \|= DN_ID_CHKED_BONUS;
	}
	mutex_exit(&dn->dn_mtx);
	if (have_spill)
	dmu_buf_rele((dmu_buf_t *)db, FTAG);
	}

	boolean_t
	dmu_objset_userspace_present(objset_t *os)
	{
	return (os->os_phys->os_flags &
	OBJSET_FLAG_USERACCOUNTING_COMPLETE);
	}

	boolean_t
	dmu_objset_userobjspace_present(objset_t *os)
	{
	return (os->os_phys->os_flags &
	OBJSET_FLAG_USEROBJACCOUNTING_COMPLETE);
	}

	boolean_t
	dmu_objset_projectquota_present(objset_t *os)
	{
	return (os->os_phys->os_flags &
	OBJSET_FLAG_PROJECTQUOTA_COMPLETE);
	}

	static int
	dmu_objset_space_upgrade(objset_t *os)
	{
	uint64_t obj;
	int err = 0;

	/*
	* We simply need to mark every object dirty, so that it will be
	* synced out and now accounted. If this is called
	* concurrently, or if we already did some work before crashing,
	* that's fine, since we track each object's accounted state
	* independently.
	*/

	for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) {
	dmu_tx_t *tx;
	dmu_buf_t *db;
	int objerr;

	mutex_enter(&os->os_upgrade_lock);
	if (os->os_upgrade_exit)
	err = SET_ERROR(EINTR);
	mutex_exit(&os->os_upgrade_lock);
	if (err != 0)
	return (err);

	- if (issig(JUSTLOOKING) && issig(FORREAL))
	+ if (issig())
	return (SET_ERROR(EINTR));

	objerr = dmu_bonus_hold(os, obj, FTAG, &db);
	if (objerr != 0)
	continue;
	tx = dmu_tx_create(os);
	dmu_tx_hold_bonus(tx, obj);
	objerr = dmu_tx_assign(tx, TXG_WAIT);
	if (objerr != 0) {
	dmu_buf_rele(db, FTAG);
	dmu_tx_abort(tx);
	continue;
	}
	dmu_buf_will_dirty(db, tx);
	dmu_buf_rele(db, FTAG);
	dmu_tx_commit(tx);
	}
	return (0);
	}

	static int
	dmu_objset_userspace_upgrade_cb(objset_t *os)
	{
	int err = 0;

	if (dmu_objset_userspace_present(os))
	return (0);
	if (dmu_objset_is_snapshot(os))
	return (SET_ERROR(EINVAL));
	if (!dmu_objset_userused_enabled(os))
	return (SET_ERROR(ENOTSUP));

	err = dmu_objset_space_upgrade(os);
	if (err)
	return (err);

	os->os_flags \|= OBJSET_FLAG_USERACCOUNTING_COMPLETE;
	txg_wait_synced(dmu_objset_pool(os), 0);
	return (0);
	}

	void
	dmu_objset_userspace_upgrade(objset_t *os)
	{
	dmu_objset_upgrade(os, dmu_objset_userspace_upgrade_cb);
	}

	static int
	dmu_objset_id_quota_upgrade_cb(objset_t *os)
	{
	int err = 0;

	if (dmu_objset_userobjspace_present(os) &&
	dmu_objset_projectquota_present(os))
	return (0);
	if (dmu_objset_is_snapshot(os))
	return (SET_ERROR(EINVAL));
	if (!dmu_objset_userused_enabled(os))
	return (SET_ERROR(ENOTSUP));
	if (!dmu_objset_projectquota_enabled(os) &&
	dmu_objset_userobjspace_present(os))
	return (SET_ERROR(ENOTSUP));

	if (dmu_objset_userobjused_enabled(os))
	dmu_objset_ds(os)->ds_feature_activation[
	SPA_FEATURE_USEROBJ_ACCOUNTING] = (void *)B_TRUE;
	if (dmu_objset_projectquota_enabled(os))
	dmu_objset_ds(os)->ds_feature_activation[
	SPA_FEATURE_PROJECT_QUOTA] = (void *)B_TRUE;

	err = dmu_objset_space_upgrade(os);
	if (err)
	return (err);

	os->os_flags \|= OBJSET_FLAG_USERACCOUNTING_COMPLETE;
	if (dmu_objset_userobjused_enabled(os))
	os->os_flags \|= OBJSET_FLAG_USEROBJACCOUNTING_COMPLETE;
	if (dmu_objset_projectquota_enabled(os))
	os->os_flags \|= OBJSET_FLAG_PROJECTQUOTA_COMPLETE;

	txg_wait_synced(dmu_objset_pool(os), 0);
	return (0);
	}

	void
	dmu_objset_id_quota_upgrade(objset_t *os)
	{
	dmu_objset_upgrade(os, dmu_objset_id_quota_upgrade_cb);
	}

	boolean_t
	dmu_objset_userobjspace_upgradable(objset_t *os)
	{
	return (dmu_objset_type(os) == DMU_OST_ZFS &&
	!dmu_objset_is_snapshot(os) &&
	dmu_objset_userobjused_enabled(os) &&
	!dmu_objset_userobjspace_present(os) &&
	spa_writeable(dmu_objset_spa(os)));
	}

	boolean_t
	dmu_objset_projectquota_upgradable(objset_t *os)
	{
	return (dmu_objset_type(os) == DMU_OST_ZFS &&
	!dmu_objset_is_snapshot(os) &&
	dmu_objset_projectquota_enabled(os) &&
	!dmu_objset_projectquota_present(os) &&
	spa_writeable(dmu_objset_spa(os)));
	}

	void
	dmu_objset_space(objset_t os, uint64_t refdbytesp, uint64_t *availbytesp,
	uint64_t usedobjsp, uint64_t availobjsp)
	{
	dsl_dataset_space(os->os_dsl_dataset, refdbytesp, availbytesp,
	usedobjsp, availobjsp);
	}

	uint64_t
	dmu_objset_fsid_guid(objset_t *os)
	{
	return (dsl_dataset_fsid_guid(os->os_dsl_dataset));
	}

	void
	dmu_objset_fast_stat(objset_t os, dmu_objset_stats_t stat)
	{
	stat->dds_type = os->os_phys->os_type;
	if (os->os_dsl_dataset)
	dsl_dataset_fast_stat(os->os_dsl_dataset, stat);
	}

	void
	dmu_objset_stats(objset_t os, nvlist_t nv)
	{
	ASSERT(os->os_dsl_dataset \|\|
	os->os_phys->os_type == DMU_OST_META);

	if (os->os_dsl_dataset != NULL)
	dsl_dataset_stats(os->os_dsl_dataset, nv);

	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_TYPE,
	os->os_phys->os_type);
	dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USERACCOUNTING,
	dmu_objset_userspace_present(os));
	}

	int
	dmu_objset_is_snapshot(objset_t *os)
	{
	if (os->os_dsl_dataset != NULL)
	return (os->os_dsl_dataset->ds_is_snapshot);
	else
	return (B_FALSE);
	}

	int
	dmu_snapshot_realname(objset_t os, const char name, char *real, int maxlen,
	boolean_t *conflict)
	{
	dsl_dataset_t *ds = os->os_dsl_dataset;
	uint64_t ignored;

	if (dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0)
	return (SET_ERROR(ENOENT));

	return (zap_lookup_norm(ds->ds_dir->dd_pool->dp_meta_objset,
	dsl_dataset_phys(ds)->ds_snapnames_zapobj, name, 8, 1, &ignored,
	MT_NORMALIZE, real, maxlen, conflict));
	}

	int
	dmu_snapshot_list_next(objset_t os, int namelen, char name,
	uint64_t idp, uint64_t offp, boolean_t *case_conflict)
	{
	dsl_dataset_t *ds = os->os_dsl_dataset;
	zap_cursor_t cursor;
	zap_attribute_t attr;

	ASSERT(dsl_pool_config_held(dmu_objset_pool(os)));

	if (dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0)
	return (SET_ERROR(ENOENT));

	zap_cursor_init_serialized(&cursor,
	ds->ds_dir->dd_pool->dp_meta_objset,
	dsl_dataset_phys(ds)->ds_snapnames_zapobj, *offp);

	if (zap_cursor_retrieve(&cursor, &attr) != 0) {
	zap_cursor_fini(&cursor);
	return (SET_ERROR(ENOENT));
	}

	if (strlen(attr.za_name) + 1 > namelen) {
	zap_cursor_fini(&cursor);
	return (SET_ERROR(ENAMETOOLONG));
	}

	(void) strlcpy(name, attr.za_name, namelen);
	if (idp)
	*idp = attr.za_first_integer;
	if (case_conflict)
	*case_conflict = attr.za_normalization_conflict;
	zap_cursor_advance(&cursor);
	*offp = zap_cursor_serialize(&cursor);
	zap_cursor_fini(&cursor);

	return (0);
	}

	int
	dmu_snapshot_lookup(objset_t os, const char name, uint64_t *value)
	{
	return (dsl_dataset_snap_lookup(os->os_dsl_dataset, name, value));
	}

	int
	dmu_dir_list_next(objset_t os, int namelen, char name,
	uint64_t idp, uint64_t offp)
	{
	dsl_dir_t *dd = os->os_dsl_dataset->ds_dir;
	zap_cursor_t cursor;
	zap_attribute_t attr;

	/* there is no next dir on a snapshot! */
	if (os->os_dsl_dataset->ds_object !=
	dsl_dir_phys(dd)->dd_head_dataset_obj)
	return (SET_ERROR(ENOENT));

	zap_cursor_init_serialized(&cursor,
	dd->dd_pool->dp_meta_objset,
	dsl_dir_phys(dd)->dd_child_dir_zapobj, *offp);

	if (zap_cursor_retrieve(&cursor, &attr) != 0) {
	zap_cursor_fini(&cursor);
	return (SET_ERROR(ENOENT));
	}

	if (strlen(attr.za_name) + 1 > namelen) {
	zap_cursor_fini(&cursor);
	return (SET_ERROR(ENAMETOOLONG));
	}

	(void) strlcpy(name, attr.za_name, namelen);
	if (idp)
	*idp = attr.za_first_integer;
	zap_cursor_advance(&cursor);
	*offp = zap_cursor_serialize(&cursor);
	zap_cursor_fini(&cursor);

	return (0);
	}

	typedef struct dmu_objset_find_ctx {
	taskq_t *dc_tq;
	dsl_pool_t *dc_dp;
	uint64_t dc_ddobj;
	char dc_ddname; / last component of ddobj's name */
	int (dc_func)(dsl_pool_t , dsl_dataset_t , void );
	void *dc_arg;
	int dc_flags;
	kmutex_t *dc_error_lock;
	int *dc_error;
	} dmu_objset_find_ctx_t;

	static void
	dmu_objset_find_dp_impl(dmu_objset_find_ctx_t *dcp)
	{
	dsl_pool_t *dp = dcp->dc_dp;
	dsl_dir_t *dd;
	dsl_dataset_t *ds;
	zap_cursor_t zc;
	zap_attribute_t *attr;
	uint64_t thisobj;
	int err = 0;

	/* don't process if there already was an error */
	if (*dcp->dc_error != 0)
	goto out;

	/*
	* Note: passing the name (dc_ddname) here is optional, but it
	* improves performance because we don't need to call
	* zap_value_search() to determine the name.
	*/
	err = dsl_dir_hold_obj(dp, dcp->dc_ddobj, dcp->dc_ddname, FTAG, &dd);
	if (err != 0)
	goto out;

	/* Don't visit hidden ($MOS & $ORIGIN) objsets. */
	if (dd->dd_myname[0] == '$') {
	dsl_dir_rele(dd, FTAG);
	goto out;
	}

	thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj;
	attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP);

	/*
	* Iterate over all children.
	*/
	if (dcp->dc_flags & DS_FIND_CHILDREN) {
	for (zap_cursor_init(&zc, dp->dp_meta_objset,
	dsl_dir_phys(dd)->dd_child_dir_zapobj);
	zap_cursor_retrieve(&zc, attr) == 0;
	(void) zap_cursor_advance(&zc)) {
	ASSERT3U(attr->za_integer_length, ==,
	sizeof (uint64_t));
	ASSERT3U(attr->za_num_integers, ==, 1);

	dmu_objset_find_ctx_t *child_dcp =
	kmem_alloc(sizeof (*child_dcp), KM_SLEEP);
	child_dcp = dcp;
	child_dcp->dc_ddobj = attr->za_first_integer;
	child_dcp->dc_ddname = spa_strdup(attr->za_name);
	if (dcp->dc_tq != NULL)
	(void) taskq_dispatch(dcp->dc_tq,
	dmu_objset_find_dp_cb, child_dcp, TQ_SLEEP);
	else
	dmu_objset_find_dp_impl(child_dcp);
	}
	zap_cursor_fini(&zc);
	}

	/*
	* Iterate over all snapshots.
	*/
	if (dcp->dc_flags & DS_FIND_SNAPSHOTS) {
	dsl_dataset_t *ds;
	err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds);

	if (err == 0) {
	uint64_t snapobj;

	snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj;
	dsl_dataset_rele(ds, FTAG);

	for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj);
	zap_cursor_retrieve(&zc, attr) == 0;
	(void) zap_cursor_advance(&zc)) {
	ASSERT3U(attr->za_integer_length, ==,
	sizeof (uint64_t));
	ASSERT3U(attr->za_num_integers, ==, 1);

	err = dsl_dataset_hold_obj(dp,
	attr->za_first_integer, FTAG, &ds);
	if (err != 0)
	break;
	err = dcp->dc_func(dp, ds, dcp->dc_arg);
	dsl_dataset_rele(ds, FTAG);
	if (err != 0)
	break;
	}
	zap_cursor_fini(&zc);
	}
	}

	kmem_free(attr, sizeof (zap_attribute_t));

	if (err != 0) {
	dsl_dir_rele(dd, FTAG);
	goto out;
	}

	/*
	* Apply to self.
	*/
	err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds);

	/*
	* Note: we hold the dir while calling dsl_dataset_hold_obj() so
	* that the dir will remain cached, and we won't have to re-instantiate
	* it (which could be expensive due to finding its name via
	* zap_value_search()).
	*/
	dsl_dir_rele(dd, FTAG);
	if (err != 0)
	goto out;
	err = dcp->dc_func(dp, ds, dcp->dc_arg);
	dsl_dataset_rele(ds, FTAG);

	out:
	if (err != 0) {
	mutex_enter(dcp->dc_error_lock);
	/* only keep first error */
	if (*dcp->dc_error == 0)
	*dcp->dc_error = err;
	mutex_exit(dcp->dc_error_lock);
	}

	if (dcp->dc_ddname != NULL)
	spa_strfree(dcp->dc_ddname);
	kmem_free(dcp, sizeof (*dcp));
	}

	static void
	dmu_objset_find_dp_cb(void *arg)
	{
	dmu_objset_find_ctx_t *dcp = arg;
	dsl_pool_t *dp = dcp->dc_dp;

	/*
	* We need to get a pool_config_lock here, as there are several
	* assert(pool_config_held) down the stack. Getting a lock via
	* dsl_pool_config_enter is risky, as it might be stalled by a
	* pending writer. This would deadlock, as the write lock can
	* only be granted when our parent thread gives up the lock.
	* The _prio interface gives us priority over a pending writer.
	*/
	dsl_pool_config_enter_prio(dp, FTAG);

	dmu_objset_find_dp_impl(dcp);

	dsl_pool_config_exit(dp, FTAG);
	}

	/*
	* Find objsets under and including ddobj, call func(ds) on each.
	* The order for the enumeration is completely undefined.
	* func is called with dsl_pool_config held.
	*/
	int
	dmu_objset_find_dp(dsl_pool_t *dp, uint64_t ddobj,
	int func(dsl_pool_t , dsl_dataset_t , void ), void arg, int flags)
	{
	int error = 0;
	taskq_t *tq = NULL;
	int ntasks;
	dmu_objset_find_ctx_t *dcp;
	kmutex_t err_lock;

	mutex_init(&err_lock, NULL, MUTEX_DEFAULT, NULL);
	dcp = kmem_alloc(sizeof (*dcp), KM_SLEEP);
	dcp->dc_tq = NULL;
	dcp->dc_dp = dp;
	dcp->dc_ddobj = ddobj;
	dcp->dc_ddname = NULL;
	dcp->dc_func = func;
	dcp->dc_arg = arg;
	dcp->dc_flags = flags;
	dcp->dc_error_lock = &err_lock;
	dcp->dc_error = &error;

	if ((flags & DS_FIND_SERIALIZE) \|\| dsl_pool_config_held_writer(dp)) {
	/*
	* In case a write lock is held we can't make use of
	* parallelism, as down the stack of the worker threads
	* the lock is asserted via dsl_pool_config_held.
	* In case of a read lock this is solved by getting a read
	* lock in each worker thread, which isn't possible in case
	* of a writer lock. So we fall back to the synchronous path
	* here.
	* In the future it might be possible to get some magic into
	* dsl_pool_config_held in a way that it returns true for
	* the worker threads so that a single lock held from this
	* thread suffices. For now, stay single threaded.
	*/
	dmu_objset_find_dp_impl(dcp);
	mutex_destroy(&err_lock);

	return (error);
	}

	ntasks = dmu_find_threads;
	if (ntasks == 0)
	ntasks = vdev_count_leaves(dp->dp_spa) * 4;
	tq = taskq_create("dmu_objset_find", ntasks, maxclsyspri, ntasks,
	INT_MAX, 0);
	if (tq == NULL) {
	kmem_free(dcp, sizeof (*dcp));
	mutex_destroy(&err_lock);

	return (SET_ERROR(ENOMEM));
	}
	dcp->dc_tq = tq;

	/* dcp will be freed by task */
	(void) taskq_dispatch(tq, dmu_objset_find_dp_cb, dcp, TQ_SLEEP);

	/*
	* PORTING: this code relies on the property of taskq_wait to wait
	* until no more tasks are queued and no more tasks are active. As
	* we always queue new tasks from within other tasks, task_wait
	* reliably waits for the full recursion to finish, even though we
	* enqueue new tasks after taskq_wait has been called.
	* On platforms other than illumos, taskq_wait may not have this
	* property.
	*/
	taskq_wait(tq);
	taskq_destroy(tq);
	mutex_destroy(&err_lock);

	return (error);
	}

	/*
	* Find all objsets under name, and for each, call 'func(child_name, arg)'.
	* The dp_config_rwlock must not be held when this is called, and it
	* will not be held when the callback is called.
	* Therefore this function should only be used when the pool is not changing
	* (e.g. in syncing context), or the callback can deal with the possible races.
	*/
	static int
	dmu_objset_find_impl(spa_t spa, const char name,
	int func(const char , void ), void *arg, int flags)
	{
	dsl_dir_t *dd;
	dsl_pool_t *dp = spa_get_dsl(spa);
	dsl_dataset_t *ds;
	zap_cursor_t zc;
	zap_attribute_t *attr;
	char *child;
	uint64_t thisobj;
	int err;

	dsl_pool_config_enter(dp, FTAG);

	err = dsl_dir_hold(dp, name, FTAG, &dd, NULL);
	if (err != 0) {
	dsl_pool_config_exit(dp, FTAG);
	return (err);
	}

	/* Don't visit hidden ($MOS & $ORIGIN) objsets. */
	if (dd->dd_myname[0] == '$') {
	dsl_dir_rele(dd, FTAG);
	dsl_pool_config_exit(dp, FTAG);
	return (0);
	}

	thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj;
	attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP);

	/*
	* Iterate over all children.
	*/
	if (flags & DS_FIND_CHILDREN) {
	for (zap_cursor_init(&zc, dp->dp_meta_objset,
	dsl_dir_phys(dd)->dd_child_dir_zapobj);
	zap_cursor_retrieve(&zc, attr) == 0;
	(void) zap_cursor_advance(&zc)) {
	ASSERT3U(attr->za_integer_length, ==,
	sizeof (uint64_t));
	ASSERT3U(attr->za_num_integers, ==, 1);

	child = kmem_asprintf("%s/%s", name, attr->za_name);
	dsl_pool_config_exit(dp, FTAG);
	err = dmu_objset_find_impl(spa, child,
	func, arg, flags);
	dsl_pool_config_enter(dp, FTAG);
	kmem_strfree(child);
	if (err != 0)
	break;
	}
	zap_cursor_fini(&zc);

	if (err != 0) {
	dsl_dir_rele(dd, FTAG);
	dsl_pool_config_exit(dp, FTAG);
	kmem_free(attr, sizeof (zap_attribute_t));
	return (err);
	}
	}

	/*
	* Iterate over all snapshots.
	*/
	if (flags & DS_FIND_SNAPSHOTS) {
	err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds);

	if (err == 0) {
	uint64_t snapobj;

	snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj;
	dsl_dataset_rele(ds, FTAG);

	for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj);
	zap_cursor_retrieve(&zc, attr) == 0;
	(void) zap_cursor_advance(&zc)) {
	ASSERT3U(attr->za_integer_length, ==,
	sizeof (uint64_t));
	ASSERT3U(attr->za_num_integers, ==, 1);

	child = kmem_asprintf("%s@%s",
	name, attr->za_name);
	dsl_pool_config_exit(dp, FTAG);
	err = func(child, arg);
	dsl_pool_config_enter(dp, FTAG);
	kmem_strfree(child);
	if (err != 0)
	break;
	}
	zap_cursor_fini(&zc);
	}
	}

	dsl_dir_rele(dd, FTAG);
	kmem_free(attr, sizeof (zap_attribute_t));
	dsl_pool_config_exit(dp, FTAG);

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

	/* Apply to self. */
	return (func(name, arg));
	}

	/*
	* See comment above dmu_objset_find_impl().
	*/
	int
	dmu_objset_find(const char name, int func(const char , void ), void arg,
	int flags)
	{
	spa_t *spa;
	int error;

	error = spa_open(name, &spa, FTAG);
	if (error != 0)
	return (error);
	error = dmu_objset_find_impl(spa, name, func, arg, flags);
	spa_close(spa, FTAG);
	return (error);
	}

	boolean_t
	dmu_objset_incompatible_encryption_version(objset_t *os)
	{
	return (dsl_dir_incompatible_encryption_version(
	os->os_dsl_dataset->ds_dir));
	}

	void
	dmu_objset_set_user(objset_t os, void user_ptr)
	{
	ASSERT(MUTEX_HELD(&os->os_user_ptr_lock));
	os->os_user_ptr = user_ptr;
	}

	void *
	dmu_objset_get_user(objset_t *os)
	{
	ASSERT(MUTEX_HELD(&os->os_user_ptr_lock));
	return (os->os_user_ptr);
	}

	/*
	* Determine name of filesystem, given name of snapshot.
	* buf must be at least ZFS_MAX_DATASET_NAME_LEN bytes
	*/
	int
	dmu_fsname(const char snapname, char buf)
	{
	char *atp = strchr(snapname, '@');
	if (atp == NULL)
	return (SET_ERROR(EINVAL));
	if (atp - snapname >= ZFS_MAX_DATASET_NAME_LEN)
	return (SET_ERROR(ENAMETOOLONG));
	(void) strlcpy(buf, snapname, atp - snapname + 1);
	return (0);
	}

	/*
	* Call when we think we're going to write/free space in open context
	* to track the amount of dirty data in the open txg, which is also the
	* amount of memory that can not be evicted until this txg syncs.
	*
	* Note that there are two conditions where this can be called from
	* syncing context:
	*
	* [1] When we just created the dataset, in which case we go on with
	* updating any accounting of dirty data as usual.
	* [2] When we are dirtying MOS data, in which case we only update the
	* pool's accounting of dirty data.
	*/
	void
	dmu_objset_willuse_space(objset_t os, int64_t space, dmu_tx_t tx)
	{
	dsl_dataset_t *ds = os->os_dsl_dataset;
	int64_t aspace = spa_get_worst_case_asize(os->os_spa, space);

	if (ds != NULL) {
	dsl_dir_willuse_space(ds->ds_dir, aspace, tx);
	}

	dsl_pool_dirty_space(dmu_tx_pool(tx), space, tx);
	}

	#if defined(_KERNEL)
	EXPORT_SYMBOL(dmu_objset_zil);
	EXPORT_SYMBOL(dmu_objset_pool);
	EXPORT_SYMBOL(dmu_objset_ds);
	EXPORT_SYMBOL(dmu_objset_type);
	EXPORT_SYMBOL(dmu_objset_name);
	EXPORT_SYMBOL(dmu_objset_hold);
	EXPORT_SYMBOL(dmu_objset_hold_flags);
	EXPORT_SYMBOL(dmu_objset_own);
	EXPORT_SYMBOL(dmu_objset_rele);
	EXPORT_SYMBOL(dmu_objset_rele_flags);
	EXPORT_SYMBOL(dmu_objset_disown);
	EXPORT_SYMBOL(dmu_objset_from_ds);
	EXPORT_SYMBOL(dmu_objset_create);
	EXPORT_SYMBOL(dmu_objset_clone);
	EXPORT_SYMBOL(dmu_objset_stats);
	EXPORT_SYMBOL(dmu_objset_fast_stat);
	EXPORT_SYMBOL(dmu_objset_spa);
	EXPORT_SYMBOL(dmu_objset_space);
	EXPORT_SYMBOL(dmu_objset_fsid_guid);
	EXPORT_SYMBOL(dmu_objset_find);
	EXPORT_SYMBOL(dmu_objset_byteswap);
	EXPORT_SYMBOL(dmu_objset_evict_dbufs);
	EXPORT_SYMBOL(dmu_objset_snap_cmtime);
	EXPORT_SYMBOL(dmu_objset_dnodesize);

	EXPORT_SYMBOL(dmu_objset_sync);
	EXPORT_SYMBOL(dmu_objset_is_dirty);
	EXPORT_SYMBOL(dmu_objset_create_impl_dnstats);
	EXPORT_SYMBOL(dmu_objset_create_impl);
	EXPORT_SYMBOL(dmu_objset_open_impl);
	EXPORT_SYMBOL(dmu_objset_evict);
	EXPORT_SYMBOL(dmu_objset_register_type);
	EXPORT_SYMBOL(dmu_objset_sync_done);
	EXPORT_SYMBOL(dmu_objset_userquota_get_ids);
	EXPORT_SYMBOL(dmu_objset_userused_enabled);
	EXPORT_SYMBOL(dmu_objset_userspace_upgrade);
	EXPORT_SYMBOL(dmu_objset_userspace_present);
	EXPORT_SYMBOL(dmu_objset_userobjused_enabled);
	EXPORT_SYMBOL(dmu_objset_userobjspace_upgradable);
	EXPORT_SYMBOL(dmu_objset_userobjspace_present);
	EXPORT_SYMBOL(dmu_objset_projectquota_enabled);
	EXPORT_SYMBOL(dmu_objset_projectquota_present);
	EXPORT_SYMBOL(dmu_objset_projectquota_upgradable);
	EXPORT_SYMBOL(dmu_objset_id_quota_upgrade);
	#endif
	diff --git a/sys/contrib/openzfs/module/zfs/dmu_recv.c b/sys/contrib/openzfs/module/zfs/dmu_recv.c
	index 680aed4513bc..0119191d7920 100644
	--- a/sys/contrib/openzfs/module/zfs/dmu_recv.c
	+++ b/sys/contrib/openzfs/module/zfs/dmu_recv.c
	@@ -1,3809 +1,3809 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2011, 2020 by Delphix. All rights reserved.
	* Copyright (c) 2014, Joyent, Inc. All rights reserved.
	* Copyright 2014 HybridCluster. All rights reserved.
	* Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
	* Copyright (c) 2019, Klara Inc.
	* Copyright (c) 2019, Allan Jude
	* Copyright (c) 2019 Datto Inc.
	* Copyright (c) 2022 Axcient.
	*/

	#include <sys/arc.h>
	#include <sys/spa_impl.h>
	#include <sys/dmu.h>
	#include <sys/dmu_impl.h>
	#include <sys/dmu_send.h>
	#include <sys/dmu_recv.h>
	#include <sys/dmu_tx.h>
	#include <sys/dbuf.h>
	#include <sys/dnode.h>
	#include <sys/zfs_context.h>
	#include <sys/dmu_objset.h>
	#include <sys/dmu_traverse.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_dir.h>
	#include <sys/dsl_prop.h>
	#include <sys/dsl_pool.h>
	#include <sys/dsl_synctask.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/zap.h>
	#include <sys/zvol.h>
	#include <sys/zio_checksum.h>
	#include <sys/zfs_znode.h>
	#include <zfs_fletcher.h>
	#include <sys/avl.h>
	#include <sys/ddt.h>
	#include <sys/zfs_onexit.h>
	#include <sys/dsl_destroy.h>
	#include <sys/blkptr.h>
	#include <sys/dsl_bookmark.h>
	#include <sys/zfeature.h>
	#include <sys/bqueue.h>
	#include <sys/objlist.h>
	#ifdef _KERNEL
	#include <sys/zfs_vfsops.h>
	#endif
	#include <sys/zfs_file.h>

	static uint_t zfs_recv_queue_length = SPA_MAXBLOCKSIZE;
	static uint_t zfs_recv_queue_ff = 20;
	static uint_t zfs_recv_write_batch_size = 1024 * 1024;
	static int zfs_recv_best_effort_corrective = 0;

	static const void *const dmu_recv_tag = "dmu_recv_tag";
	const char *const recv_clone_name = "%recv";

	typedef enum {
	ORNS_NO,
	ORNS_YES,
	ORNS_MAYBE
	} or_need_sync_t;

	static int receive_read_payload_and_next_header(dmu_recv_cookie_t *ra, int len,
	void *buf);

	struct receive_record_arg {
	dmu_replay_record_t header;
	void payload; / Pointer to a buffer containing the payload */
	/*
	* If the record is a WRITE or SPILL, pointer to the abd containing the
	* payload.
	*/
	abd_t *abd;
	int payload_size;
	uint64_t bytes_read; /* bytes read from stream when record created */
	boolean_t eos_marker; /* Marks the end of the stream */
	bqueue_node_t node;
	};

	struct receive_writer_arg {
	objset_t *os;
	boolean_t byteswap;
	bqueue_t q;

	/*
	* These three members are used to signal to the main thread when
	* we're done.
	*/
	kmutex_t mutex;
	kcondvar_t cv;
	boolean_t done;

	int err;
	const char *tofs;
	boolean_t heal;
	boolean_t resumable;
	boolean_t raw; /* DMU_BACKUP_FEATURE_RAW set */
	boolean_t spill; /* DRR_FLAG_SPILL_BLOCK set */
	boolean_t full; /* this is a full send stream */
	uint64_t last_object;
	uint64_t last_offset;
	uint64_t max_object; /* highest object ID referenced in stream */
	uint64_t bytes_read; /* bytes read when current record created */

	list_t write_batch;

	/* Encryption parameters for the last received DRR_OBJECT_RANGE */
	boolean_t or_crypt_params_present;
	uint64_t or_firstobj;
	uint64_t or_numslots;
	uint8_t or_salt[ZIO_DATA_SALT_LEN];
	uint8_t or_iv[ZIO_DATA_IV_LEN];
	uint8_t or_mac[ZIO_DATA_MAC_LEN];
	boolean_t or_byteorder;
	zio_t *heal_pio;

	/* Keep track of DRR_FREEOBJECTS right after DRR_OBJECT_RANGE */
	or_need_sync_t or_need_sync;
	};

	typedef struct dmu_recv_begin_arg {
	const char *drba_origin;
	dmu_recv_cookie_t *drba_cookie;
	cred_t *drba_cred;
	proc_t *drba_proc;
	dsl_crypto_params_t *drba_dcp;
	} dmu_recv_begin_arg_t;

	static void
	byteswap_record(dmu_replay_record_t *drr)
	{
	#define DO64(X) (drr->drr_u.X = BSWAP_64(drr->drr_u.X))
	#define DO32(X) (drr->drr_u.X = BSWAP_32(drr->drr_u.X))
	drr->drr_type = BSWAP_32(drr->drr_type);
	drr->drr_payloadlen = BSWAP_32(drr->drr_payloadlen);

	switch (drr->drr_type) {
	case DRR_BEGIN:
	DO64(drr_begin.drr_magic);
	DO64(drr_begin.drr_versioninfo);
	DO64(drr_begin.drr_creation_time);
	DO32(drr_begin.drr_type);
	DO32(drr_begin.drr_flags);
	DO64(drr_begin.drr_toguid);
	DO64(drr_begin.drr_fromguid);
	break;
	case DRR_OBJECT:
	DO64(drr_object.drr_object);
	DO32(drr_object.drr_type);
	DO32(drr_object.drr_bonustype);
	DO32(drr_object.drr_blksz);
	DO32(drr_object.drr_bonuslen);
	DO32(drr_object.drr_raw_bonuslen);
	DO64(drr_object.drr_toguid);
	DO64(drr_object.drr_maxblkid);
	break;
	case DRR_FREEOBJECTS:
	DO64(drr_freeobjects.drr_firstobj);
	DO64(drr_freeobjects.drr_numobjs);
	DO64(drr_freeobjects.drr_toguid);
	break;
	case DRR_WRITE:
	DO64(drr_write.drr_object);
	DO32(drr_write.drr_type);
	DO64(drr_write.drr_offset);
	DO64(drr_write.drr_logical_size);
	DO64(drr_write.drr_toguid);
	ZIO_CHECKSUM_BSWAP(&drr->drr_u.drr_write.drr_key.ddk_cksum);
	DO64(drr_write.drr_key.ddk_prop);
	DO64(drr_write.drr_compressed_size);
	break;
	case DRR_WRITE_EMBEDDED:
	DO64(drr_write_embedded.drr_object);
	DO64(drr_write_embedded.drr_offset);
	DO64(drr_write_embedded.drr_length);
	DO64(drr_write_embedded.drr_toguid);
	DO32(drr_write_embedded.drr_lsize);
	DO32(drr_write_embedded.drr_psize);
	break;
	case DRR_FREE:
	DO64(drr_free.drr_object);
	DO64(drr_free.drr_offset);
	DO64(drr_free.drr_length);
	DO64(drr_free.drr_toguid);
	break;
	case DRR_SPILL:
	DO64(drr_spill.drr_object);
	DO64(drr_spill.drr_length);
	DO64(drr_spill.drr_toguid);
	DO64(drr_spill.drr_compressed_size);
	DO32(drr_spill.drr_type);
	break;
	case DRR_OBJECT_RANGE:
	DO64(drr_object_range.drr_firstobj);
	DO64(drr_object_range.drr_numslots);
	DO64(drr_object_range.drr_toguid);
	break;
	case DRR_REDACT:
	DO64(drr_redact.drr_object);
	DO64(drr_redact.drr_offset);
	DO64(drr_redact.drr_length);
	DO64(drr_redact.drr_toguid);
	break;
	case DRR_END:
	DO64(drr_end.drr_toguid);
	ZIO_CHECKSUM_BSWAP(&drr->drr_u.drr_end.drr_checksum);
	break;
	default:
	break;
	}

	if (drr->drr_type != DRR_BEGIN) {
	ZIO_CHECKSUM_BSWAP(&drr->drr_u.drr_checksum.drr_checksum);
	}

	#undef DO64
	#undef DO32
	}

	static boolean_t
	redact_snaps_contains(uint64_t *snaps, uint64_t num_snaps, uint64_t guid)
	{
	for (int i = 0; i < num_snaps; i++) {
	if (snaps[i] == guid)
	return (B_TRUE);
	}
	return (B_FALSE);
	}

	/*
	* Check that the new stream we're trying to receive is redacted with respect to
	* a subset of the snapshots that the origin was redacted with respect to. For
	* the reasons behind this, see the man page on redacted zfs sends and receives.
	*/
	static boolean_t
	compatible_redact_snaps(uint64_t *origin_snaps, uint64_t origin_num_snaps,
	uint64_t *redact_snaps, uint64_t num_redact_snaps)
	{
	/*
	* Short circuit the comparison; if we are redacted with respect to
	* more snapshots than the origin, we can't be redacted with respect
	* to a subset.
	*/
	if (num_redact_snaps > origin_num_snaps) {
	return (B_FALSE);
	}

	for (int i = 0; i < num_redact_snaps; i++) {
	if (!redact_snaps_contains(origin_snaps, origin_num_snaps,
	redact_snaps[i])) {
	return (B_FALSE);
	}
	}
	return (B_TRUE);
	}

	static boolean_t
	redact_check(dmu_recv_begin_arg_t drba, dsl_dataset_t origin)
	{
	uint64_t *origin_snaps;
	uint64_t origin_num_snaps;
	dmu_recv_cookie_t *drc = drba->drba_cookie;
	struct drr_begin *drrb = drc->drc_drrb;
	int featureflags = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo);
	int err = 0;
	boolean_t ret = B_TRUE;
	uint64_t *redact_snaps;
	uint_t numredactsnaps;

	/*
	* If this is a full send stream, we're safe no matter what.
	*/
	if (drrb->drr_fromguid == 0)
	return (ret);

	VERIFY(dsl_dataset_get_uint64_array_feature(origin,
	SPA_FEATURE_REDACTED_DATASETS, &origin_num_snaps, &origin_snaps));

	if (nvlist_lookup_uint64_array(drc->drc_begin_nvl,
	BEGINNV_REDACT_FROM_SNAPS, &redact_snaps, &numredactsnaps) ==
	0) {
	/*
	* If the send stream was sent from the redaction bookmark or
	* the redacted version of the dataset, then we're safe. Verify
	* that this is from the a compatible redaction bookmark or
	* redacted dataset.
	*/
	if (!compatible_redact_snaps(origin_snaps, origin_num_snaps,
	redact_snaps, numredactsnaps)) {
	err = EINVAL;
	}
	} else if (featureflags & DMU_BACKUP_FEATURE_REDACTED) {
	/*
	* If the stream is redacted, it must be redacted with respect
	* to a subset of what the origin is redacted with respect to.
	* See case number 2 in the zfs man page section on redacted zfs
	* send.
	*/
	err = nvlist_lookup_uint64_array(drc->drc_begin_nvl,
	BEGINNV_REDACT_SNAPS, &redact_snaps, &numredactsnaps);

	if (err != 0 \|\| !compatible_redact_snaps(origin_snaps,
	origin_num_snaps, redact_snaps, numredactsnaps)) {
	err = EINVAL;
	}
	} else if (!redact_snaps_contains(origin_snaps, origin_num_snaps,
	drrb->drr_toguid)) {
	/*
	* If the stream isn't redacted but the origin is, this must be
	* one of the snapshots the origin is redacted with respect to.
	* See case number 1 in the zfs man page section on redacted zfs
	* send.
	*/
	err = EINVAL;
	}

	if (err != 0)
	ret = B_FALSE;
	return (ret);
	}

	/*
	* If we previously received a stream with --large-block, we don't support
	* receiving an incremental on top of it without --large-block. This avoids
	* forcing a read-modify-write or trying to re-aggregate a string of WRITE
	* records.
	*/
	static int
	recv_check_large_blocks(dsl_dataset_t *ds, uint64_t featureflags)
	{
	if (dsl_dataset_feature_is_active(ds, SPA_FEATURE_LARGE_BLOCKS) &&
	!(featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS))
	return (SET_ERROR(ZFS_ERR_STREAM_LARGE_BLOCK_MISMATCH));
	return (0);
	}

	static int
	recv_begin_check_existing_impl(dmu_recv_begin_arg_t drba, dsl_dataset_t ds,
	uint64_t fromguid, uint64_t featureflags)
	{
	uint64_t obj;
	uint64_t children;
	int error;
	dsl_dataset_t *snap;
	dsl_pool_t *dp = ds->ds_dir->dd_pool;
	boolean_t encrypted = ds->ds_dir->dd_crypto_obj != 0;
	boolean_t raw = (featureflags & DMU_BACKUP_FEATURE_RAW) != 0;
	boolean_t embed = (featureflags & DMU_BACKUP_FEATURE_EMBED_DATA) != 0;

	/* Temporary clone name must not exist. */
	error = zap_lookup(dp->dp_meta_objset,
	dsl_dir_phys(ds->ds_dir)->dd_child_dir_zapobj, recv_clone_name,
	8, 1, &obj);
	if (error != ENOENT)
	return (error == 0 ? SET_ERROR(EBUSY) : error);

	/* Resume state must not be set. */
	if (dsl_dataset_has_resume_receive_state(ds))
	return (SET_ERROR(EBUSY));

	/* New snapshot name must not exist if we're not healing it. */
	error = zap_lookup(dp->dp_meta_objset,
	dsl_dataset_phys(ds)->ds_snapnames_zapobj,
	drba->drba_cookie->drc_tosnap, 8, 1, &obj);
	if (drba->drba_cookie->drc_heal) {
	if (error != 0)
	return (error);
	} else if (error != ENOENT) {
	return (error == 0 ? SET_ERROR(EEXIST) : error);
	}

	/* Must not have children if receiving a ZVOL. */
	error = zap_count(dp->dp_meta_objset,
	dsl_dir_phys(ds->ds_dir)->dd_child_dir_zapobj, &children);
	if (error != 0)
	return (error);
	if (drba->drba_cookie->drc_drrb->drr_type != DMU_OST_ZFS &&
	children > 0)
	return (SET_ERROR(ZFS_ERR_WRONG_PARENT));

	/*
	* Check snapshot limit before receiving. We'll recheck again at the
	* end, but might as well abort before receiving if we're already over
	* the limit.
	*
	* Note that we do not check the file system limit with
	* dsl_dir_fscount_check because the temporary %clones don't count
	* against that limit.
	*/
	error = dsl_fs_ss_limit_check(ds->ds_dir, 1, ZFS_PROP_SNAPSHOT_LIMIT,
	NULL, drba->drba_cred, drba->drba_proc);
	if (error != 0)
	return (error);

	if (drba->drba_cookie->drc_heal) {
	/* Encryption is incompatible with embedded data. */
	if (encrypted && embed)
	return (SET_ERROR(EINVAL));

	/* Healing is not supported when in 'force' mode. */
	if (drba->drba_cookie->drc_force)
	return (SET_ERROR(EINVAL));

	/* Must have keys loaded if doing encrypted non-raw recv. */
	if (encrypted && !raw) {
	if (spa_keystore_lookup_key(dp->dp_spa, ds->ds_object,
	NULL, NULL) != 0)
	return (SET_ERROR(EACCES));
	}

	error = dsl_dataset_hold_obj(dp, obj, FTAG, &snap);
	if (error != 0)
	return (error);

	/*
	* When not doing best effort corrective recv healing can only
	* be done if the send stream is for the same snapshot as the
	* one we are trying to heal.
	*/
	if (zfs_recv_best_effort_corrective == 0 &&
	drba->drba_cookie->drc_drrb->drr_toguid !=
	dsl_dataset_phys(snap)->ds_guid) {
	dsl_dataset_rele(snap, FTAG);
	return (SET_ERROR(ENOTSUP));
	}
	dsl_dataset_rele(snap, FTAG);
	} else if (fromguid != 0) {
	/* Sanity check the incremental recv */
	uint64_t obj = dsl_dataset_phys(ds)->ds_prev_snap_obj;

	/* Can't perform a raw receive on top of a non-raw receive */
	if (!encrypted && raw)
	return (SET_ERROR(EINVAL));

	/* Encryption is incompatible with embedded data */
	if (encrypted && embed)
	return (SET_ERROR(EINVAL));

	/* Find snapshot in this dir that matches fromguid. */
	while (obj != 0) {
	error = dsl_dataset_hold_obj(dp, obj, FTAG,
	&snap);
	if (error != 0)
	return (SET_ERROR(ENODEV));
	if (snap->ds_dir != ds->ds_dir) {
	dsl_dataset_rele(snap, FTAG);
	return (SET_ERROR(ENODEV));
	}
	if (dsl_dataset_phys(snap)->ds_guid == fromguid)
	break;
	obj = dsl_dataset_phys(snap)->ds_prev_snap_obj;
	dsl_dataset_rele(snap, FTAG);
	}
	if (obj == 0)
	return (SET_ERROR(ENODEV));

	if (drba->drba_cookie->drc_force) {
	drba->drba_cookie->drc_fromsnapobj = obj;
	} else {
	/*
	* If we are not forcing, there must be no
	* changes since fromsnap. Raw sends have an
	* additional constraint that requires that
	* no "noop" snapshots exist between fromsnap
	* and tosnap for the IVset checking code to
	* work properly.
	*/
	if (dsl_dataset_modified_since_snap(ds, snap) \|\|
	(raw &&
	dsl_dataset_phys(ds)->ds_prev_snap_obj !=
	snap->ds_object)) {
	dsl_dataset_rele(snap, FTAG);
	return (SET_ERROR(ETXTBSY));
	}
	drba->drba_cookie->drc_fromsnapobj =
	ds->ds_prev->ds_object;
	}

	if (dsl_dataset_feature_is_active(snap,
	SPA_FEATURE_REDACTED_DATASETS) && !redact_check(drba,
	snap)) {
	dsl_dataset_rele(snap, FTAG);
	return (SET_ERROR(EINVAL));
	}

	error = recv_check_large_blocks(snap, featureflags);
	if (error != 0) {
	dsl_dataset_rele(snap, FTAG);
	return (error);
	}

	dsl_dataset_rele(snap, FTAG);
	} else {
	/* If full and not healing then must be forced. */
	if (!drba->drba_cookie->drc_force)
	return (SET_ERROR(EEXIST));

	/*
	* We don't support using zfs recv -F to blow away
	* encrypted filesystems. This would require the
	* dsl dir to point to the old encryption key and
	* the new one at the same time during the receive.
	*/
	if ((!encrypted && raw) \|\| encrypted)
	return (SET_ERROR(EINVAL));

	/*
	* Perform the same encryption checks we would if
	* we were creating a new dataset from scratch.
	*/
	if (!raw) {
	boolean_t will_encrypt;

	error = dmu_objset_create_crypt_check(
	ds->ds_dir->dd_parent, drba->drba_dcp,
	&will_encrypt);
	if (error != 0)
	return (error);

	if (will_encrypt && embed)
	return (SET_ERROR(EINVAL));
	}
	}

	return (0);
	}

	/*
	* Check that any feature flags used in the data stream we're receiving are
	* supported by the pool we are receiving into.
	*
	* Note that some of the features we explicitly check here have additional
	* (implicit) features they depend on, but those dependencies are enforced
	* through the zfeature_register() calls declaring the features that we
	* explicitly check.
	*/
	static int
	recv_begin_check_feature_flags_impl(uint64_t featureflags, spa_t *spa)
	{
	/*
	* Check if there are any unsupported feature flags.
	*/
	if (!DMU_STREAM_SUPPORTED(featureflags)) {
	return (SET_ERROR(ZFS_ERR_UNKNOWN_SEND_STREAM_FEATURE));
	}

	/* Verify pool version supports SA if SA_SPILL feature set */
	if ((featureflags & DMU_BACKUP_FEATURE_SA_SPILL) &&
	spa_version(spa) < SPA_VERSION_SA)
	return (SET_ERROR(ENOTSUP));

	/*
	* LZ4 compressed, ZSTD compressed, embedded, mooched, large blocks,
	* and large_dnodes in the stream can only be used if those pool
	* features are enabled because we don't attempt to decompress /
	* un-embed / un-mooch / split up the blocks / dnodes during the
	* receive process.
	*/
	if ((featureflags & DMU_BACKUP_FEATURE_LZ4) &&
	!spa_feature_is_enabled(spa, SPA_FEATURE_LZ4_COMPRESS))
	return (SET_ERROR(ENOTSUP));
	if ((featureflags & DMU_BACKUP_FEATURE_ZSTD) &&
	!spa_feature_is_enabled(spa, SPA_FEATURE_ZSTD_COMPRESS))
	return (SET_ERROR(ENOTSUP));
	if ((featureflags & DMU_BACKUP_FEATURE_EMBED_DATA) &&
	!spa_feature_is_enabled(spa, SPA_FEATURE_EMBEDDED_DATA))
	return (SET_ERROR(ENOTSUP));
	if ((featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS) &&
	!spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS))
	return (SET_ERROR(ENOTSUP));
	if ((featureflags & DMU_BACKUP_FEATURE_LARGE_DNODE) &&
	!spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_DNODE))
	return (SET_ERROR(ENOTSUP));

	/*
	* Receiving redacted streams requires that redacted datasets are
	* enabled.
	*/
	if ((featureflags & DMU_BACKUP_FEATURE_REDACTED) &&
	!spa_feature_is_enabled(spa, SPA_FEATURE_REDACTED_DATASETS))
	return (SET_ERROR(ENOTSUP));

	return (0);
	}

	static int
	dmu_recv_begin_check(void arg, dmu_tx_t tx)
	{
	dmu_recv_begin_arg_t *drba = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	struct drr_begin *drrb = drba->drba_cookie->drc_drrb;
	uint64_t fromguid = drrb->drr_fromguid;
	int flags = drrb->drr_flags;
	ds_hold_flags_t dsflags = DS_HOLD_FLAG_NONE;
	int error;
	uint64_t featureflags = drba->drba_cookie->drc_featureflags;
	dsl_dataset_t *ds;
	const char *tofs = drba->drba_cookie->drc_tofs;

	/* already checked */
	ASSERT3U(drrb->drr_magic, ==, DMU_BACKUP_MAGIC);
	ASSERT(!(featureflags & DMU_BACKUP_FEATURE_RESUMING));

	if (DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo) ==
	DMU_COMPOUNDSTREAM \|\|
	drrb->drr_type >= DMU_OST_NUMTYPES \|\|
	((flags & DRR_FLAG_CLONE) && drba->drba_origin == NULL))
	return (SET_ERROR(EINVAL));

	error = recv_begin_check_feature_flags_impl(featureflags, dp->dp_spa);
	if (error != 0)
	return (error);

	/* Resumable receives require extensible datasets */
	if (drba->drba_cookie->drc_resumable &&
	!spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_EXTENSIBLE_DATASET))
	return (SET_ERROR(ENOTSUP));

	if (featureflags & DMU_BACKUP_FEATURE_RAW) {
	/* raw receives require the encryption feature */
	if (!spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_ENCRYPTION))
	return (SET_ERROR(ENOTSUP));

	/* embedded data is incompatible with encryption and raw recv */
	if (featureflags & DMU_BACKUP_FEATURE_EMBED_DATA)
	return (SET_ERROR(EINVAL));

	/* raw receives require spill block allocation flag */
	if (!(flags & DRR_FLAG_SPILL_BLOCK))
	return (SET_ERROR(ZFS_ERR_SPILL_BLOCK_FLAG_MISSING));
	} else {
	/*
	* We support unencrypted datasets below encrypted ones now,
	* so add the DS_HOLD_FLAG_DECRYPT flag only if we are dealing
	* with a dataset we may encrypt.
	*/
	if (drba->drba_dcp == NULL \|\|
	drba->drba_dcp->cp_crypt != ZIO_CRYPT_OFF) {
	dsflags \|= DS_HOLD_FLAG_DECRYPT;
	}
	}

	error = dsl_dataset_hold_flags(dp, tofs, dsflags, FTAG, &ds);
	if (error == 0) {
	/* target fs already exists; recv into temp clone */

	/* Can't recv a clone into an existing fs */
	if (flags & DRR_FLAG_CLONE \|\| drba->drba_origin) {
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (SET_ERROR(EINVAL));
	}

	error = recv_begin_check_existing_impl(drba, ds, fromguid,
	featureflags);
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	} else if (error == ENOENT) {
	/* target fs does not exist; must be a full backup or clone */
	char buf[ZFS_MAX_DATASET_NAME_LEN];
	objset_t *os;

	/* healing recv must be done "into" an existing snapshot */
	if (drba->drba_cookie->drc_heal == B_TRUE)
	return (SET_ERROR(ENOTSUP));

	/*
	* If it's a non-clone incremental, we are missing the
	* target fs, so fail the recv.
	*/
	if (fromguid != 0 && !((flags & DRR_FLAG_CLONE) \|\|
	drba->drba_origin))
	return (SET_ERROR(ENOENT));

	/*
	* If we're receiving a full send as a clone, and it doesn't
	* contain all the necessary free records and freeobject
	* records, reject it.
	*/
	if (fromguid == 0 && drba->drba_origin != NULL &&
	!(flags & DRR_FLAG_FREERECORDS))
	return (SET_ERROR(EINVAL));

	/* Open the parent of tofs */
	ASSERT3U(strlen(tofs), <, sizeof (buf));
	(void) strlcpy(buf, tofs, strrchr(tofs, '/') - tofs + 1);
	error = dsl_dataset_hold(dp, buf, FTAG, &ds);
	if (error != 0)
	return (error);

	if ((featureflags & DMU_BACKUP_FEATURE_RAW) == 0 &&
	drba->drba_origin == NULL) {
	boolean_t will_encrypt;

	/*
	* Check that we aren't breaking any encryption rules
	* and that we have all the parameters we need to
	* create an encrypted dataset if necessary. If we are
	* making an encrypted dataset the stream can't have
	* embedded data.
	*/
	error = dmu_objset_create_crypt_check(ds->ds_dir,
	drba->drba_dcp, &will_encrypt);
	if (error != 0) {
	dsl_dataset_rele(ds, FTAG);
	return (error);
	}

	if (will_encrypt &&
	(featureflags & DMU_BACKUP_FEATURE_EMBED_DATA)) {
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(EINVAL));
	}
	}

	/*
	* Check filesystem and snapshot limits before receiving. We'll
	* recheck snapshot limits again at the end (we create the
	* filesystems and increment those counts during begin_sync).
	*/
	error = dsl_fs_ss_limit_check(ds->ds_dir, 1,
	ZFS_PROP_FILESYSTEM_LIMIT, NULL,
	drba->drba_cred, drba->drba_proc);
	if (error != 0) {
	dsl_dataset_rele(ds, FTAG);
	return (error);
	}

	error = dsl_fs_ss_limit_check(ds->ds_dir, 1,
	ZFS_PROP_SNAPSHOT_LIMIT, NULL,
	drba->drba_cred, drba->drba_proc);
	if (error != 0) {
	dsl_dataset_rele(ds, FTAG);
	return (error);
	}

	/* can't recv below anything but filesystems (eg. no ZVOLs) */
	error = dmu_objset_from_ds(ds, &os);
	if (error != 0) {
	dsl_dataset_rele(ds, FTAG);
	return (error);
	}
	if (dmu_objset_type(os) != DMU_OST_ZFS) {
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(ZFS_ERR_WRONG_PARENT));
	}

	if (drba->drba_origin != NULL) {
	dsl_dataset_t *origin;
	error = dsl_dataset_hold_flags(dp, drba->drba_origin,
	dsflags, FTAG, &origin);
	if (error != 0) {
	dsl_dataset_rele(ds, FTAG);
	return (error);
	}
	if (!origin->ds_is_snapshot) {
	dsl_dataset_rele_flags(origin, dsflags, FTAG);
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(EINVAL));
	}
	if (dsl_dataset_phys(origin)->ds_guid != fromguid &&
	fromguid != 0) {
	dsl_dataset_rele_flags(origin, dsflags, FTAG);
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(ENODEV));
	}

	if (origin->ds_dir->dd_crypto_obj != 0 &&
	(featureflags & DMU_BACKUP_FEATURE_EMBED_DATA)) {
	dsl_dataset_rele_flags(origin, dsflags, FTAG);
	dsl_dataset_rele(ds, FTAG);
	return (SET_ERROR(EINVAL));
	}

	/*
	* If the origin is redacted we need to verify that this
	* send stream can safely be received on top of the
	* origin.
	*/
	if (dsl_dataset_feature_is_active(origin,
	SPA_FEATURE_REDACTED_DATASETS)) {
	if (!redact_check(drba, origin)) {
	dsl_dataset_rele_flags(origin, dsflags,
	FTAG);
	dsl_dataset_rele_flags(ds, dsflags,
	FTAG);
	return (SET_ERROR(EINVAL));
	}
	}

	error = recv_check_large_blocks(ds, featureflags);
	if (error != 0) {
	dsl_dataset_rele_flags(origin, dsflags, FTAG);
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (error);
	}

	dsl_dataset_rele_flags(origin, dsflags, FTAG);
	}

	dsl_dataset_rele(ds, FTAG);
	error = 0;
	}
	return (error);
	}

	static void
	dmu_recv_begin_sync(void arg, dmu_tx_t tx)
	{
	dmu_recv_begin_arg_t *drba = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	objset_t *mos = dp->dp_meta_objset;
	dmu_recv_cookie_t *drc = drba->drba_cookie;
	struct drr_begin *drrb = drc->drc_drrb;
	const char *tofs = drc->drc_tofs;
	uint64_t featureflags = drc->drc_featureflags;
	dsl_dataset_t ds, newds;
	objset_t *os;
	uint64_t dsobj;
	ds_hold_flags_t dsflags = DS_HOLD_FLAG_NONE;
	int error;
	uint64_t crflags = 0;
	dsl_crypto_params_t dummy_dcp = { 0 };
	dsl_crypto_params_t *dcp = drba->drba_dcp;

	if (drrb->drr_flags & DRR_FLAG_CI_DATA)
	crflags \|= DS_FLAG_CI_DATASET;

	if ((featureflags & DMU_BACKUP_FEATURE_RAW) == 0)
	dsflags \|= DS_HOLD_FLAG_DECRYPT;

	/*
	* Raw, non-incremental recvs always use a dummy dcp with
	* the raw cmd set. Raw incremental recvs do not use a dcp
	* since the encryption parameters are already set in stone.
	*/
	if (dcp == NULL && drrb->drr_fromguid == 0 &&
	drba->drba_origin == NULL) {
	ASSERT3P(dcp, ==, NULL);
	dcp = &dummy_dcp;

	if (featureflags & DMU_BACKUP_FEATURE_RAW)
	dcp->cp_cmd = DCP_CMD_RAW_RECV;
	}

	error = dsl_dataset_hold_flags(dp, tofs, dsflags, FTAG, &ds);
	if (error == 0) {
	/* Create temporary clone unless we're doing corrective recv */
	dsl_dataset_t *snap = NULL;

	if (drba->drba_cookie->drc_fromsnapobj != 0) {
	VERIFY0(dsl_dataset_hold_obj(dp,
	drba->drba_cookie->drc_fromsnapobj, FTAG, &snap));
	ASSERT3P(dcp, ==, NULL);
	}
	if (drc->drc_heal) {
	/* When healing we want to use the provided snapshot */
	VERIFY0(dsl_dataset_snap_lookup(ds, drc->drc_tosnap,
	&dsobj));
	} else {
	dsobj = dsl_dataset_create_sync(ds->ds_dir,
	recv_clone_name, snap, crflags, drba->drba_cred,
	dcp, tx);
	}
	if (drba->drba_cookie->drc_fromsnapobj != 0)
	dsl_dataset_rele(snap, FTAG);
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	} else {
	dsl_dir_t *dd;
	const char *tail;
	dsl_dataset_t *origin = NULL;

	VERIFY0(dsl_dir_hold(dp, tofs, FTAG, &dd, &tail));

	if (drba->drba_origin != NULL) {
	VERIFY0(dsl_dataset_hold(dp, drba->drba_origin,
	FTAG, &origin));
	ASSERT3P(dcp, ==, NULL);
	}

	/* Create new dataset. */
	dsobj = dsl_dataset_create_sync(dd, strrchr(tofs, '/') + 1,
	origin, crflags, drba->drba_cred, dcp, tx);
	if (origin != NULL)
	dsl_dataset_rele(origin, FTAG);
	dsl_dir_rele(dd, FTAG);
	drc->drc_newfs = B_TRUE;
	}
	VERIFY0(dsl_dataset_own_obj_force(dp, dsobj, dsflags, dmu_recv_tag,
	&newds));
	if (dsl_dataset_feature_is_active(newds,
	SPA_FEATURE_REDACTED_DATASETS)) {
	/*
	* If the origin dataset is redacted, the child will be redacted
	* when we create it. We clear the new dataset's
	* redaction info; if it should be redacted, we'll fill
	* in its information later.
	*/
	dsl_dataset_deactivate_feature(newds,
	SPA_FEATURE_REDACTED_DATASETS, tx);
	}
	VERIFY0(dmu_objset_from_ds(newds, &os));

	if (drc->drc_resumable) {
	dsl_dataset_zapify(newds, tx);
	if (drrb->drr_fromguid != 0) {
	VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_FROMGUID,
	8, 1, &drrb->drr_fromguid, tx));
	}
	VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_TOGUID,
	8, 1, &drrb->drr_toguid, tx));
	VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_TONAME,
	1, strlen(drrb->drr_toname) + 1, drrb->drr_toname, tx));
	uint64_t one = 1;
	uint64_t zero = 0;
	VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_OBJECT,
	8, 1, &one, tx));
	VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_OFFSET,
	8, 1, &zero, tx));
	VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_BYTES,
	8, 1, &zero, tx));
	if (featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS) {
	VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_LARGEBLOCK,
	8, 1, &one, tx));
	}
	if (featureflags & DMU_BACKUP_FEATURE_EMBED_DATA) {
	VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_EMBEDOK,
	8, 1, &one, tx));
	}
	if (featureflags & DMU_BACKUP_FEATURE_COMPRESSED) {
	VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_COMPRESSOK,
	8, 1, &one, tx));
	}
	if (featureflags & DMU_BACKUP_FEATURE_RAW) {
	VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_RAWOK,
	8, 1, &one, tx));
	}

	uint64_t *redact_snaps;
	uint_t numredactsnaps;
	if (nvlist_lookup_uint64_array(drc->drc_begin_nvl,
	BEGINNV_REDACT_FROM_SNAPS, &redact_snaps,
	&numredactsnaps) == 0) {
	VERIFY0(zap_add(mos, dsobj,
	DS_FIELD_RESUME_REDACT_BOOKMARK_SNAPS,
	sizeof (*redact_snaps), numredactsnaps,
	redact_snaps, tx));
	}
	}

	/*
	* Usually the os->os_encrypted value is tied to the presence of a
	* DSL Crypto Key object in the dd. However, that will not be received
	* until dmu_recv_stream(), so we set the value manually for now.
	*/
	if (featureflags & DMU_BACKUP_FEATURE_RAW) {
	os->os_encrypted = B_TRUE;
	drba->drba_cookie->drc_raw = B_TRUE;
	}

	if (featureflags & DMU_BACKUP_FEATURE_REDACTED) {
	uint64_t *redact_snaps;
	uint_t numredactsnaps;
	VERIFY0(nvlist_lookup_uint64_array(drc->drc_begin_nvl,
	BEGINNV_REDACT_SNAPS, &redact_snaps, &numredactsnaps));
	dsl_dataset_activate_redaction(newds, redact_snaps,
	numredactsnaps, tx);
	}

	dmu_buf_will_dirty(newds->ds_dbuf, tx);
	dsl_dataset_phys(newds)->ds_flags \|= DS_FLAG_INCONSISTENT;

	/*
	* If we actually created a non-clone, we need to create the objset
	* in our new dataset. If this is a raw send we postpone this until
	* dmu_recv_stream() so that we can allocate the metadnode with the
	* properties from the DRR_BEGIN payload.
	*/
	rrw_enter(&newds->ds_bp_rwlock, RW_READER, FTAG);
	if (BP_IS_HOLE(dsl_dataset_get_blkptr(newds)) &&
	(featureflags & DMU_BACKUP_FEATURE_RAW) == 0 &&
	!drc->drc_heal) {
	(void) dmu_objset_create_impl(dp->dp_spa,
	newds, dsl_dataset_get_blkptr(newds), drrb->drr_type, tx);
	}
	rrw_exit(&newds->ds_bp_rwlock, FTAG);

	drba->drba_cookie->drc_ds = newds;
	drba->drba_cookie->drc_os = os;

	spa_history_log_internal_ds(newds, "receive", tx, " ");
	}

	static int
	dmu_recv_resume_begin_check(void arg, dmu_tx_t tx)
	{
	dmu_recv_begin_arg_t *drba = arg;
	dmu_recv_cookie_t *drc = drba->drba_cookie;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	struct drr_begin *drrb = drc->drc_drrb;
	int error;
	ds_hold_flags_t dsflags = DS_HOLD_FLAG_NONE;
	dsl_dataset_t *ds;
	const char *tofs = drc->drc_tofs;

	/* already checked */
	ASSERT3U(drrb->drr_magic, ==, DMU_BACKUP_MAGIC);
	ASSERT(drc->drc_featureflags & DMU_BACKUP_FEATURE_RESUMING);

	if (DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo) ==
	DMU_COMPOUNDSTREAM \|\|
	drrb->drr_type >= DMU_OST_NUMTYPES)
	return (SET_ERROR(EINVAL));

	/*
	* This is mostly a sanity check since we should have already done these
	* checks during a previous attempt to receive the data.
	*/
	error = recv_begin_check_feature_flags_impl(drc->drc_featureflags,
	dp->dp_spa);
	if (error != 0)
	return (error);

	/* 6 extra bytes for /%recv */
	char recvname[ZFS_MAX_DATASET_NAME_LEN + 6];

	(void) snprintf(recvname, sizeof (recvname), "%s/%s",
	tofs, recv_clone_name);

	if (drc->drc_featureflags & DMU_BACKUP_FEATURE_RAW) {
	/* raw receives require spill block allocation flag */
	if (!(drrb->drr_flags & DRR_FLAG_SPILL_BLOCK))
	return (SET_ERROR(ZFS_ERR_SPILL_BLOCK_FLAG_MISSING));
	} else {
	dsflags \|= DS_HOLD_FLAG_DECRYPT;
	}

	boolean_t recvexist = B_TRUE;
	if (dsl_dataset_hold_flags(dp, recvname, dsflags, FTAG, &ds) != 0) {
	/* %recv does not exist; continue in tofs */
	recvexist = B_FALSE;
	error = dsl_dataset_hold_flags(dp, tofs, dsflags, FTAG, &ds);
	if (error != 0)
	return (error);
	}

	/*
	* Resume of full/newfs recv on existing dataset should be done with
	* force flag
	*/
	if (recvexist && drrb->drr_fromguid == 0 && !drc->drc_force) {
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (SET_ERROR(ZFS_ERR_RESUME_EXISTS));
	}

	/* check that ds is marked inconsistent */
	if (!DS_IS_INCONSISTENT(ds)) {
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (SET_ERROR(EINVAL));
	}

	/* check that there is resuming data, and that the toguid matches */
	if (!dsl_dataset_is_zapified(ds)) {
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (SET_ERROR(EINVAL));
	}
	uint64_t val;
	error = zap_lookup(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_TOGUID, sizeof (val), 1, &val);
	if (error != 0 \|\| drrb->drr_toguid != val) {
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (SET_ERROR(EINVAL));
	}

	/*
	* Check if the receive is still running. If so, it will be owned.
	* Note that nothing else can own the dataset (e.g. after the receive
	* fails) because it will be marked inconsistent.
	*/
	if (dsl_dataset_has_owner(ds)) {
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (SET_ERROR(EBUSY));
	}

	/* There should not be any snapshots of this fs yet. */
	if (ds->ds_prev != NULL && ds->ds_prev->ds_dir == ds->ds_dir) {
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (SET_ERROR(EINVAL));
	}

	/*
	* Note: resume point will be checked when we process the first WRITE
	* record.
	*/

	/* check that the origin matches */
	val = 0;
	(void) zap_lookup(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_FROMGUID, sizeof (val), 1, &val);
	if (drrb->drr_fromguid != val) {
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (SET_ERROR(EINVAL));
	}

	if (ds->ds_prev != NULL && drrb->drr_fromguid != 0)
	drc->drc_fromsnapobj = ds->ds_prev->ds_object;

	/*
	* If we're resuming, and the send is redacted, then the original send
	* must have been redacted, and must have been redacted with respect to
	* the same snapshots.
	*/
	if (drc->drc_featureflags & DMU_BACKUP_FEATURE_REDACTED) {
	uint64_t num_ds_redact_snaps;
	uint64_t *ds_redact_snaps;

	uint_t num_stream_redact_snaps;
	uint64_t *stream_redact_snaps;

	if (nvlist_lookup_uint64_array(drc->drc_begin_nvl,
	BEGINNV_REDACT_SNAPS, &stream_redact_snaps,
	&num_stream_redact_snaps) != 0) {
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (SET_ERROR(EINVAL));
	}

	if (!dsl_dataset_get_uint64_array_feature(ds,
	SPA_FEATURE_REDACTED_DATASETS, &num_ds_redact_snaps,
	&ds_redact_snaps)) {
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (SET_ERROR(EINVAL));
	}

	for (int i = 0; i < num_ds_redact_snaps; i++) {
	if (!redact_snaps_contains(ds_redact_snaps,
	num_ds_redact_snaps, stream_redact_snaps[i])) {
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (SET_ERROR(EINVAL));
	}
	}
	}

	error = recv_check_large_blocks(ds, drc->drc_featureflags);
	if (error != 0) {
	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (error);
	}

	dsl_dataset_rele_flags(ds, dsflags, FTAG);
	return (0);
	}

	static void
	dmu_recv_resume_begin_sync(void arg, dmu_tx_t tx)
	{
	dmu_recv_begin_arg_t *drba = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	const char *tofs = drba->drba_cookie->drc_tofs;
	uint64_t featureflags = drba->drba_cookie->drc_featureflags;
	dsl_dataset_t *ds;
	ds_hold_flags_t dsflags = DS_HOLD_FLAG_NONE;
	/* 6 extra bytes for /%recv */
	char recvname[ZFS_MAX_DATASET_NAME_LEN + 6];

	(void) snprintf(recvname, sizeof (recvname), "%s/%s", tofs,
	recv_clone_name);

	if (featureflags & DMU_BACKUP_FEATURE_RAW) {
	drba->drba_cookie->drc_raw = B_TRUE;
	} else {
	dsflags \|= DS_HOLD_FLAG_DECRYPT;
	}

	if (dsl_dataset_own_force(dp, recvname, dsflags, dmu_recv_tag, &ds)
	!= 0) {
	/* %recv does not exist; continue in tofs */
	VERIFY0(dsl_dataset_own_force(dp, tofs, dsflags, dmu_recv_tag,
	&ds));
	drba->drba_cookie->drc_newfs = B_TRUE;
	}

	ASSERT(DS_IS_INCONSISTENT(ds));
	rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
	ASSERT(!BP_IS_HOLE(dsl_dataset_get_blkptr(ds)) \|\|
	drba->drba_cookie->drc_raw);
	rrw_exit(&ds->ds_bp_rwlock, FTAG);

	drba->drba_cookie->drc_ds = ds;
	VERIFY0(dmu_objset_from_ds(ds, &drba->drba_cookie->drc_os));
	drba->drba_cookie->drc_should_save = B_TRUE;

	spa_history_log_internal_ds(ds, "resume receive", tx, " ");
	}

	/*
	* NB: callers MUST call dmu_recv_stream() if dmu_recv_begin()
	* succeeds; otherwise we will leak the holds on the datasets.
	*/
	int
	dmu_recv_begin(const char tofs, const char tosnap,
	dmu_replay_record_t *drr_begin, boolean_t force, boolean_t heal,
	boolean_t resumable, nvlist_t localprops, nvlist_t hidden_args,
	const char origin, dmu_recv_cookie_t drc, zfs_file_t *fp,
	offset_t *voffp)
	{
	dmu_recv_begin_arg_t drba = { 0 };
	int err = 0;

	memset(drc, 0, sizeof (dmu_recv_cookie_t));
	drc->drc_drr_begin = drr_begin;
	drc->drc_drrb = &drr_begin->drr_u.drr_begin;
	drc->drc_tosnap = tosnap;
	drc->drc_tofs = tofs;
	drc->drc_force = force;
	drc->drc_heal = heal;
	drc->drc_resumable = resumable;
	drc->drc_cred = CRED();
	drc->drc_proc = curproc;
	drc->drc_clone = (origin != NULL);

	if (drc->drc_drrb->drr_magic == BSWAP_64(DMU_BACKUP_MAGIC)) {
	drc->drc_byteswap = B_TRUE;
	(void) fletcher_4_incremental_byteswap(drr_begin,
	sizeof (dmu_replay_record_t), &drc->drc_cksum);
	byteswap_record(drr_begin);
	} else if (drc->drc_drrb->drr_magic == DMU_BACKUP_MAGIC) {
	(void) fletcher_4_incremental_native(drr_begin,
	sizeof (dmu_replay_record_t), &drc->drc_cksum);
	} else {
	return (SET_ERROR(EINVAL));
	}

	drc->drc_fp = fp;
	drc->drc_voff = *voffp;
	drc->drc_featureflags =
	DMU_GET_FEATUREFLAGS(drc->drc_drrb->drr_versioninfo);

	uint32_t payloadlen = drc->drc_drr_begin->drr_payloadlen;

	/*
	* Since OpenZFS 2.0.0, we have enforced a 64MB limit in userspace
	* configurable via ZFS_SENDRECV_MAX_NVLIST. We enforce 256MB as a hard
	* upper limit. Systems with less than 1GB of RAM will see a lower
	* limit from `arc_all_memory() / 4`.
	*/
	if (payloadlen > (MIN((1U << 28), arc_all_memory() / 4)))
	return (E2BIG);


	if (payloadlen != 0) {
	void *payload = vmem_alloc(payloadlen, KM_SLEEP);
	/*
	* For compatibility with recursive send streams, we don't do
	* this here if the stream could be part of a package. Instead,
	* we'll do it in dmu_recv_stream. If we pull the next header
	* too early, and it's the END record, we break the `recv_skip`
	* logic.
	*/

	err = receive_read_payload_and_next_header(drc, payloadlen,
	payload);
	if (err != 0) {
	vmem_free(payload, payloadlen);
	return (err);
	}
	err = nvlist_unpack(payload, payloadlen, &drc->drc_begin_nvl,
	KM_SLEEP);
	vmem_free(payload, payloadlen);
	if (err != 0) {
	kmem_free(drc->drc_next_rrd,
	sizeof (*drc->drc_next_rrd));
	return (err);
	}
	}

	if (drc->drc_drrb->drr_flags & DRR_FLAG_SPILL_BLOCK)
	drc->drc_spill = B_TRUE;

	drba.drba_origin = origin;
	drba.drba_cookie = drc;
	drba.drba_cred = CRED();
	drba.drba_proc = curproc;

	if (drc->drc_featureflags & DMU_BACKUP_FEATURE_RESUMING) {
	err = dsl_sync_task(tofs,
	dmu_recv_resume_begin_check, dmu_recv_resume_begin_sync,
	&drba, 5, ZFS_SPACE_CHECK_NORMAL);
	} else {
	/*
	* For non-raw, non-incremental, non-resuming receives the
	* user can specify encryption parameters on the command line
	* with "zfs recv -o". For these receives we create a dcp and
	* pass it to the sync task. Creating the dcp will implicitly
	* remove the encryption params from the localprops nvlist,
	* which avoids errors when trying to set these normally
	* read-only properties. Any other kind of receive that
	* attempts to set these properties will fail as a result.
	*/
	if ((DMU_GET_FEATUREFLAGS(drc->drc_drrb->drr_versioninfo) &
	DMU_BACKUP_FEATURE_RAW) == 0 &&
	origin == NULL && drc->drc_drrb->drr_fromguid == 0) {
	err = dsl_crypto_params_create_nvlist(DCP_CMD_NONE,
	localprops, hidden_args, &drba.drba_dcp);
	}

	if (err == 0) {
	err = dsl_sync_task(tofs,
	dmu_recv_begin_check, dmu_recv_begin_sync,
	&drba, 5, ZFS_SPACE_CHECK_NORMAL);
	dsl_crypto_params_free(drba.drba_dcp, !!err);
	}
	}

	if (err != 0) {
	kmem_free(drc->drc_next_rrd, sizeof (*drc->drc_next_rrd));
	nvlist_free(drc->drc_begin_nvl);
	}
	return (err);
	}

	/*
	* Holds data need for corrective recv callback
	*/
	typedef struct cr_cb_data {
	uint64_t size;
	zbookmark_phys_t zb;
	spa_t *spa;
	} cr_cb_data_t;

	static void
	corrective_read_done(zio_t *zio)
	{
	cr_cb_data_t *data = zio->io_private;
	/* Corruption corrected; update error log if needed */
	if (zio->io_error == 0) {
	spa_remove_error(data->spa, &data->zb,
	BP_GET_LOGICAL_BIRTH(zio->io_bp));
	}
	kmem_free(data, sizeof (cr_cb_data_t));
	abd_free(zio->io_abd);
	}

	/*
	* zio_rewrite the data pointed to by bp with the data from the rrd's abd.
	*/
	static int
	do_corrective_recv(struct receive_writer_arg rwa, struct drr_write drrw,
	struct receive_record_arg rrd, blkptr_t bp)
	{
	int err;
	zio_t *io;
	zbookmark_phys_t zb;
	dnode_t *dn;
	abd_t *abd = rrd->abd;
	zio_cksum_t bp_cksum = bp->blk_cksum;
	zio_flag_t flags = ZIO_FLAG_SPECULATIVE \| ZIO_FLAG_DONT_RETRY \|
	ZIO_FLAG_CANFAIL;

	if (rwa->raw)
	flags \|= ZIO_FLAG_RAW;

	err = dnode_hold(rwa->os, drrw->drr_object, FTAG, &dn);
	if (err != 0)
	return (err);
	SET_BOOKMARK(&zb, dmu_objset_id(rwa->os), drrw->drr_object, 0,
	dbuf_whichblock(dn, 0, drrw->drr_offset));
	dnode_rele(dn, FTAG);

	if (!rwa->raw && DRR_WRITE_COMPRESSED(drrw)) {
	/* Decompress the stream data */
	abd_t *dabd = abd_alloc_linear(
	drrw->drr_logical_size, B_FALSE);
	err = zio_decompress_data(drrw->drr_compressiontype,
	abd, abd_to_buf(dabd), abd_get_size(abd),
	abd_get_size(dabd), NULL);

	if (err != 0) {
	abd_free(dabd);
	return (err);
	}
	/* Swap in the newly decompressed data into the abd */
	abd_free(abd);
	abd = dabd;
	}

	if (!rwa->raw && BP_GET_COMPRESS(bp) != ZIO_COMPRESS_OFF) {
	/* Recompress the data */
	abd_t *cabd = abd_alloc_linear(BP_GET_PSIZE(bp),
	B_FALSE);
	void *buf = abd_to_buf(cabd);
	uint64_t csize = zio_compress_data(BP_GET_COMPRESS(bp),
	abd, &buf, abd_get_size(abd),
	rwa->os->os_complevel);
	abd_zero_off(cabd, csize, BP_GET_PSIZE(bp) - csize);
	/* Swap in newly compressed data into the abd */
	abd_free(abd);
	abd = cabd;
	flags \|= ZIO_FLAG_RAW_COMPRESS;
	}

	/*
	* The stream is not encrypted but the data on-disk is.
	* We need to re-encrypt the buf using the same
	* encryption type, salt, iv, and mac that was used to encrypt
	* the block previosly.
	*/
	if (!rwa->raw && BP_USES_CRYPT(bp)) {
	dsl_dataset_t *ds;
	dsl_crypto_key_t *dck = NULL;
	uint8_t salt[ZIO_DATA_SALT_LEN];
	uint8_t iv[ZIO_DATA_IV_LEN];
	uint8_t mac[ZIO_DATA_MAC_LEN];
	boolean_t no_crypt = B_FALSE;
	dsl_pool_t *dp = dmu_objset_pool(rwa->os);
	abd_t *eabd = abd_alloc_linear(BP_GET_PSIZE(bp), B_FALSE);

	zio_crypt_decode_params_bp(bp, salt, iv);
	zio_crypt_decode_mac_bp(bp, mac);

	dsl_pool_config_enter(dp, FTAG);
	err = dsl_dataset_hold_flags(dp, rwa->tofs,
	DS_HOLD_FLAG_DECRYPT, FTAG, &ds);
	if (err != 0) {
	dsl_pool_config_exit(dp, FTAG);
	abd_free(eabd);
	return (SET_ERROR(EACCES));
	}

	/* Look up the key from the spa's keystore */
	err = spa_keystore_lookup_key(rwa->os->os_spa,
	zb.zb_objset, FTAG, &dck);
	if (err != 0) {
	dsl_dataset_rele_flags(ds, DS_HOLD_FLAG_DECRYPT,
	FTAG);
	dsl_pool_config_exit(dp, FTAG);
	abd_free(eabd);
	return (SET_ERROR(EACCES));
	}

	err = zio_do_crypt_abd(B_TRUE, &dck->dck_key,
	BP_GET_TYPE(bp), BP_SHOULD_BYTESWAP(bp), salt, iv,
	mac, abd_get_size(abd), abd, eabd, &no_crypt);

	spa_keystore_dsl_key_rele(rwa->os->os_spa, dck, FTAG);
	dsl_dataset_rele_flags(ds, DS_HOLD_FLAG_DECRYPT, FTAG);
	dsl_pool_config_exit(dp, FTAG);

	ASSERT0(no_crypt);
	if (err != 0) {
	abd_free(eabd);
	return (err);
	}
	/* Swap in the newly encrypted data into the abd */
	abd_free(abd);
	abd = eabd;

	/*
	* We want to prevent zio_rewrite() from trying to
	* encrypt the data again
	*/
	flags \|= ZIO_FLAG_RAW_ENCRYPT;
	}
	rrd->abd = abd;

	io = zio_rewrite(NULL, rwa->os->os_spa, BP_GET_LOGICAL_BIRTH(bp), bp,
	abd, BP_GET_PSIZE(bp), NULL, NULL, ZIO_PRIORITY_SYNC_WRITE, flags,
	&zb);

	ASSERT(abd_get_size(abd) == BP_GET_LSIZE(bp) \|\|
	abd_get_size(abd) == BP_GET_PSIZE(bp));

	/* compute new bp checksum value and make sure it matches the old one */
	zio_checksum_compute(io, BP_GET_CHECKSUM(bp), abd, abd_get_size(abd));
	if (!ZIO_CHECKSUM_EQUAL(bp_cksum, io->io_bp->blk_cksum)) {
	zio_destroy(io);
	if (zfs_recv_best_effort_corrective != 0)
	return (0);
	return (SET_ERROR(ECKSUM));
	}

	/* Correct the corruption in place */
	err = zio_wait(io);
	if (err == 0) {
	cr_cb_data_t *cb_data =
	kmem_alloc(sizeof (cr_cb_data_t), KM_SLEEP);
	cb_data->spa = rwa->os->os_spa;
	cb_data->size = drrw->drr_logical_size;
	cb_data->zb = zb;
	/* Test if healing worked by re-reading the bp */
	err = zio_wait(zio_read(rwa->heal_pio, rwa->os->os_spa, bp,
	abd_alloc_for_io(drrw->drr_logical_size, B_FALSE),
	drrw->drr_logical_size, corrective_read_done,
	cb_data, ZIO_PRIORITY_ASYNC_READ, flags, NULL));
	}
	if (err != 0 && zfs_recv_best_effort_corrective != 0)
	err = 0;

	return (err);
	}

	static int
	receive_read(dmu_recv_cookie_t drc, int len, void buf)
	{
	int done = 0;

	/*
	* The code doesn't rely on this (lengths being multiples of 8). See
	* comment in dump_bytes.
	*/
	ASSERT(len % 8 == 0 \|\|
	(drc->drc_featureflags & DMU_BACKUP_FEATURE_RAW) != 0);

	while (done < len) {
	ssize_t resid = len - done;
	zfs_file_t *fp = drc->drc_fp;
	int err = zfs_file_read(fp, (char *)buf + done,
	len - done, &resid);
	if (err == 0 && resid == len - done) {
	/*
	* Note: ECKSUM or ZFS_ERR_STREAM_TRUNCATED indicates
	* that the receive was interrupted and can
	* potentially be resumed.
	*/
	err = SET_ERROR(ZFS_ERR_STREAM_TRUNCATED);
	}
	drc->drc_voff += len - done - resid;
	done = len - resid;
	if (err != 0)
	return (err);
	}

	drc->drc_bytes_read += len;

	ASSERT3U(done, ==, len);
	return (0);
	}

	static inline uint8_t
	deduce_nblkptr(dmu_object_type_t bonus_type, uint64_t bonus_size)
	{
	if (bonus_type == DMU_OT_SA) {
	return (1);
	} else {
	return (1 +
	((DN_OLD_MAX_BONUSLEN -
	MIN(DN_OLD_MAX_BONUSLEN, bonus_size)) >> SPA_BLKPTRSHIFT));
	}
	}

	static void
	save_resume_state(struct receive_writer_arg *rwa,
	uint64_t object, uint64_t offset, dmu_tx_t *tx)
	{
	int txgoff = dmu_tx_get_txg(tx) & TXG_MASK;

	if (!rwa->resumable)
	return;

	/*
	* We use ds_resume_bytes[] != 0 to indicate that we need to
	* update this on disk, so it must not be 0.
	*/
	ASSERT(rwa->bytes_read != 0);

	/*
	* We only resume from write records, which have a valid
	* (non-meta-dnode) object number.
	*/
	ASSERT(object != 0);

	/*
	* For resuming to work correctly, we must receive records in order,
	* sorted by object,offset. This is checked by the callers, but
	* assert it here for good measure.
	*/
	ASSERT3U(object, >=, rwa->os->os_dsl_dataset->ds_resume_object[txgoff]);
	ASSERT(object != rwa->os->os_dsl_dataset->ds_resume_object[txgoff] \|\|
	offset >= rwa->os->os_dsl_dataset->ds_resume_offset[txgoff]);
	ASSERT3U(rwa->bytes_read, >=,
	rwa->os->os_dsl_dataset->ds_resume_bytes[txgoff]);

	rwa->os->os_dsl_dataset->ds_resume_object[txgoff] = object;
	rwa->os->os_dsl_dataset->ds_resume_offset[txgoff] = offset;
	rwa->os->os_dsl_dataset->ds_resume_bytes[txgoff] = rwa->bytes_read;
	}

	static int
	receive_object_is_same_generation(objset_t *os, uint64_t object,
	dmu_object_type_t old_bonus_type, dmu_object_type_t new_bonus_type,
	const void new_bonus, boolean_t samegenp)
	{
	zfs_file_info_t zoi;
	int err;

	dmu_buf_t *old_bonus_dbuf;
	err = dmu_bonus_hold(os, object, FTAG, &old_bonus_dbuf);
	if (err != 0)
	return (err);
	err = dmu_get_file_info(os, old_bonus_type, old_bonus_dbuf->db_data,
	&zoi);
	dmu_buf_rele(old_bonus_dbuf, FTAG);
	if (err != 0)
	return (err);
	uint64_t old_gen = zoi.zfi_generation;

	err = dmu_get_file_info(os, new_bonus_type, new_bonus, &zoi);
	if (err != 0)
	return (err);
	uint64_t new_gen = zoi.zfi_generation;

	*samegenp = (old_gen == new_gen);
	return (0);
	}

	static int
	receive_handle_existing_object(const struct receive_writer_arg *rwa,
	const struct drr_object drro, const dmu_object_info_t doi,
	const void *bonus_data,
	uint64_t object_to_hold, uint32_t new_blksz)
	{
	uint32_t indblksz = drro->drr_indblkshift ?
	1ULL << drro->drr_indblkshift : 0;
	int nblkptr = deduce_nblkptr(drro->drr_bonustype,
	drro->drr_bonuslen);
	uint8_t dn_slots = drro->drr_dn_slots != 0 ?
	drro->drr_dn_slots : DNODE_MIN_SLOTS;
	boolean_t do_free_range = B_FALSE;
	int err;

	*object_to_hold = drro->drr_object;

	/* nblkptr should be bounded by the bonus size and type */
	if (rwa->raw && nblkptr != drro->drr_nblkptr)
	return (SET_ERROR(EINVAL));

	/*
	* After the previous send stream, the sending system may
	* have freed this object, and then happened to re-allocate
	* this object number in a later txg. In this case, we are
	* receiving a different logical file, and the block size may
	* appear to be different. i.e. we may have a different
	* block size for this object than what the send stream says.
	* In this case we need to remove the object's contents,
	* so that its structure can be changed and then its contents
	* entirely replaced by subsequent WRITE records.
	*
	* If this is a -L (--large-block) incremental stream, and
	* the previous stream was not -L, the block size may appear
	* to increase. i.e. we may have a smaller block size for
	* this object than what the send stream says. In this case
	* we need to keep the object's contents and block size
	* intact, so that we don't lose parts of the object's
	* contents that are not changed by this incremental send
	* stream.
	*
	* We can distinguish between the two above cases by using
	* the ZPL's generation number (see
	* receive_object_is_same_generation()). However, we only
	* want to rely on the generation number when absolutely
	* necessary, because with raw receives, the generation is
	* encrypted. We also want to minimize dependence on the
	* ZPL, so that other types of datasets can also be received
	* (e.g. ZVOLs, although note that ZVOLS currently do not
	* reallocate their objects or change their structure).
	* Therefore, we check a number of different cases where we
	* know it is safe to discard the object's contents, before
	* using the ZPL's generation number to make the above
	* distinction.
	*/
	if (drro->drr_blksz != doi->doi_data_block_size) {
	if (rwa->raw) {
	/*
	* RAW streams always have large blocks, so
	* we are sure that the data is not needed
	* due to changing --large-block to be on.
	* Which is fortunate since the bonus buffer
	* (which contains the ZPL generation) is
	* encrypted, and the key might not be
	* loaded.
	*/
	do_free_range = B_TRUE;
	} else if (rwa->full) {
	/*
	* This is a full send stream, so it always
	* replaces what we have. Even if the
	* generation numbers happen to match, this
	* can not actually be the same logical file.
	* This is relevant when receiving a full
	* send as a clone.
	*/
	do_free_range = B_TRUE;
	} else if (drro->drr_type !=
	DMU_OT_PLAIN_FILE_CONTENTS \|\|
	doi->doi_type != DMU_OT_PLAIN_FILE_CONTENTS) {
	/*
	* PLAIN_FILE_CONTENTS are the only type of
	* objects that have ever been stored with
	* large blocks, so we don't need the special
	* logic below. ZAP blocks can shrink (when
	* there's only one block), so we don't want
	* to hit the error below about block size
	* only increasing.
	*/
	do_free_range = B_TRUE;
	} else if (doi->doi_max_offset <=
	doi->doi_data_block_size) {
	/*
	* There is only one block. We can free it,
	* because its contents will be replaced by a
	* WRITE record. This can not be the no-L ->
	* -L case, because the no-L case would have
	* resulted in multiple blocks. If we
	* supported -L -> no-L, it would not be safe
	* to free the file's contents. Fortunately,
	* that is not allowed (see
	* recv_check_large_blocks()).
	*/
	do_free_range = B_TRUE;
	} else {
	boolean_t is_same_gen;
	err = receive_object_is_same_generation(rwa->os,
	drro->drr_object, doi->doi_bonus_type,
	drro->drr_bonustype, bonus_data, &is_same_gen);
	if (err != 0)
	return (SET_ERROR(EINVAL));

	if (is_same_gen) {
	/*
	* This is the same logical file, and
	* the block size must be increasing.
	* It could only decrease if
	* --large-block was changed to be
	* off, which is checked in
	* recv_check_large_blocks().
	*/
	if (drro->drr_blksz <=
	doi->doi_data_block_size)
	return (SET_ERROR(EINVAL));
	/*
	* We keep the existing blocksize and
	* contents.
	*/
	*new_blksz =
	doi->doi_data_block_size;
	} else {
	do_free_range = B_TRUE;
	}
	}
	}

	/* nblkptr can only decrease if the object was reallocated */
	if (nblkptr < doi->doi_nblkptr)
	do_free_range = B_TRUE;

	/* number of slots can only change on reallocation */
	if (dn_slots != doi->doi_dnodesize >> DNODE_SHIFT)
	do_free_range = B_TRUE;

	/*
	* For raw sends we also check a few other fields to
	* ensure we are preserving the objset structure exactly
	* as it was on the receive side:
	* - A changed indirect block size
	* - A smaller nlevels
	*/
	if (rwa->raw) {
	if (indblksz != doi->doi_metadata_block_size)
	do_free_range = B_TRUE;
	if (drro->drr_nlevels < doi->doi_indirection)
	do_free_range = B_TRUE;
	}

	if (do_free_range) {
	err = dmu_free_long_range(rwa->os, drro->drr_object,
	0, DMU_OBJECT_END);
	if (err != 0)
	return (SET_ERROR(EINVAL));
	}

	/*
	* The dmu does not currently support decreasing nlevels or changing
	* indirect block size if there is already one, same as changing the
	* number of of dnode slots on an object. For non-raw sends this
	* does not matter and the new object can just use the previous one's
	* parameters. For raw sends, however, the structure of the received
	* dnode (including indirects and dnode slots) must match that of the
	* send side. Therefore, instead of using dmu_object_reclaim(), we
	* must free the object completely and call dmu_object_claim_dnsize()
	* instead.
	*/
	if ((rwa->raw && ((doi->doi_indirection > 1 &&
	indblksz != doi->doi_metadata_block_size) \|\|
	drro->drr_nlevels < doi->doi_indirection)) \|\|
	dn_slots != doi->doi_dnodesize >> DNODE_SHIFT) {
	err = dmu_free_long_object(rwa->os, drro->drr_object);
	if (err != 0)
	return (SET_ERROR(EINVAL));

	txg_wait_synced(dmu_objset_pool(rwa->os), 0);
	*object_to_hold = DMU_NEW_OBJECT;
	}

	/*
	* For raw receives, free everything beyond the new incoming
	* maxblkid. Normally this would be done with a DRR_FREE
	* record that would come after this DRR_OBJECT record is
	* processed. However, for raw receives we manually set the
	* maxblkid from the drr_maxblkid and so we must first free
	* everything above that blkid to ensure the DMU is always
	* consistent with itself. We will never free the first block
	* of the object here because a maxblkid of 0 could indicate
	* an object with a single block or one with no blocks. This
	* free may be skipped when dmu_free_long_range() was called
	* above since it covers the entire object's contents.
	*/
	if (rwa->raw && *object_to_hold != DMU_NEW_OBJECT && !do_free_range) {
	err = dmu_free_long_range(rwa->os, drro->drr_object,
	(drro->drr_maxblkid + 1) * doi->doi_data_block_size,
	DMU_OBJECT_END);
	if (err != 0)
	return (SET_ERROR(EINVAL));
	}
	return (0);
	}

	noinline static int
	receive_object(struct receive_writer_arg rwa, struct drr_object drro,
	void *data)
	{
	dmu_object_info_t doi;
	dmu_tx_t *tx;
	int err;
	uint32_t new_blksz = drro->drr_blksz;
	uint8_t dn_slots = drro->drr_dn_slots != 0 ?
	drro->drr_dn_slots : DNODE_MIN_SLOTS;

	if (drro->drr_type == DMU_OT_NONE \|\|
	!DMU_OT_IS_VALID(drro->drr_type) \|\|
	!DMU_OT_IS_VALID(drro->drr_bonustype) \|\|
	drro->drr_checksumtype >= ZIO_CHECKSUM_FUNCTIONS \|\|
	drro->drr_compress >= ZIO_COMPRESS_FUNCTIONS \|\|
	P2PHASE(drro->drr_blksz, SPA_MINBLOCKSIZE) \|\|
	drro->drr_blksz < SPA_MINBLOCKSIZE \|\|
	drro->drr_blksz > spa_maxblocksize(dmu_objset_spa(rwa->os)) \|\|
	drro->drr_bonuslen >
	DN_BONUS_SIZE(spa_maxdnodesize(dmu_objset_spa(rwa->os))) \|\|
	dn_slots >
	(spa_maxdnodesize(dmu_objset_spa(rwa->os)) >> DNODE_SHIFT)) {
	return (SET_ERROR(EINVAL));
	}

	if (rwa->raw) {
	/*
	* We should have received a DRR_OBJECT_RANGE record
	* containing this block and stored it in rwa.
	*/
	if (drro->drr_object < rwa->or_firstobj \|\|
	drro->drr_object >= rwa->or_firstobj + rwa->or_numslots \|\|
	drro->drr_raw_bonuslen < drro->drr_bonuslen \|\|
	drro->drr_indblkshift > SPA_MAXBLOCKSHIFT \|\|
	drro->drr_nlevels > DN_MAX_LEVELS \|\|
	drro->drr_nblkptr > DN_MAX_NBLKPTR \|\|
	DN_SLOTS_TO_BONUSLEN(dn_slots) <
	drro->drr_raw_bonuslen)
	return (SET_ERROR(EINVAL));
	} else {
	/*
	* The DRR_OBJECT_SPILL flag is valid when the DRR_BEGIN
	* record indicates this by setting DRR_FLAG_SPILL_BLOCK.
	*/
	if (((drro->drr_flags & ~(DRR_OBJECT_SPILL))) \|\|
	(!rwa->spill && DRR_OBJECT_HAS_SPILL(drro->drr_flags))) {
	return (SET_ERROR(EINVAL));
	}

	if (drro->drr_raw_bonuslen != 0 \|\| drro->drr_nblkptr != 0 \|\|
	drro->drr_indblkshift != 0 \|\| drro->drr_nlevels != 0) {
	return (SET_ERROR(EINVAL));
	}
	}

	err = dmu_object_info(rwa->os, drro->drr_object, &doi);

	if (err != 0 && err != ENOENT && err != EEXIST)
	return (SET_ERROR(EINVAL));

	if (drro->drr_object > rwa->max_object)
	rwa->max_object = drro->drr_object;

	/*
	* If we are losing blkptrs or changing the block size this must
	* be a new file instance. We must clear out the previous file
	* contents before we can change this type of metadata in the dnode.
	* Raw receives will also check that the indirect structure of the
	* dnode hasn't changed.
	*/
	uint64_t object_to_hold;
	if (err == 0) {
	err = receive_handle_existing_object(rwa, drro, &doi, data,
	&object_to_hold, &new_blksz);
	if (err != 0)
	return (err);
	} else if (err == EEXIST) {
	/*
	* The object requested is currently an interior slot of a
	* multi-slot dnode. This will be resolved when the next txg
	* is synced out, since the send stream will have told us
	* to free this slot when we freed the associated dnode
	* earlier in the stream.
	*/
	txg_wait_synced(dmu_objset_pool(rwa->os), 0);

	if (dmu_object_info(rwa->os, drro->drr_object, NULL) != ENOENT)
	return (SET_ERROR(EINVAL));

	/* object was freed and we are about to allocate a new one */
	object_to_hold = DMU_NEW_OBJECT;
	} else {
	/*
	* If the only record in this range so far was DRR_FREEOBJECTS
	* with at least one actually freed object, it's possible that
	* the block will now be converted to a hole. We need to wait
	* for the txg to sync to prevent races.
	*/
	if (rwa->or_need_sync == ORNS_YES)
	txg_wait_synced(dmu_objset_pool(rwa->os), 0);

	/* object is free and we are about to allocate a new one */
	object_to_hold = DMU_NEW_OBJECT;
	}

	/* Only relevant for the first object in the range */
	rwa->or_need_sync = ORNS_NO;

	/*
	* If this is a multi-slot dnode there is a chance that this
	* object will expand into a slot that is already used by
	* another object from the previous snapshot. We must free
	* these objects before we attempt to allocate the new dnode.
	*/
	if (dn_slots > 1) {
	boolean_t need_sync = B_FALSE;

	for (uint64_t slot = drro->drr_object + 1;
	slot < drro->drr_object + dn_slots;
	slot++) {
	dmu_object_info_t slot_doi;

	err = dmu_object_info(rwa->os, slot, &slot_doi);
	if (err == ENOENT \|\| err == EEXIST)
	continue;
	else if (err != 0)
	return (err);

	err = dmu_free_long_object(rwa->os, slot);
	if (err != 0)
	return (err);

	need_sync = B_TRUE;
	}

	if (need_sync)
	txg_wait_synced(dmu_objset_pool(rwa->os), 0);
	}

	tx = dmu_tx_create(rwa->os);
	dmu_tx_hold_bonus(tx, object_to_hold);
	dmu_tx_hold_write(tx, object_to_hold, 0, 0);
	err = dmu_tx_assign(tx, TXG_WAIT);
	if (err != 0) {
	dmu_tx_abort(tx);
	return (err);
	}

	if (object_to_hold == DMU_NEW_OBJECT) {
	/* Currently free, wants to be allocated */
	err = dmu_object_claim_dnsize(rwa->os, drro->drr_object,
	drro->drr_type, new_blksz,
	drro->drr_bonustype, drro->drr_bonuslen,
	dn_slots << DNODE_SHIFT, tx);
	} else if (drro->drr_type != doi.doi_type \|\|
	new_blksz != doi.doi_data_block_size \|\|
	drro->drr_bonustype != doi.doi_bonus_type \|\|
	drro->drr_bonuslen != doi.doi_bonus_size) {
	/* Currently allocated, but with different properties */
	err = dmu_object_reclaim_dnsize(rwa->os, drro->drr_object,
	drro->drr_type, new_blksz,
	drro->drr_bonustype, drro->drr_bonuslen,
	dn_slots << DNODE_SHIFT, rwa->spill ?
	DRR_OBJECT_HAS_SPILL(drro->drr_flags) : B_FALSE, tx);
	} else if (rwa->spill && !DRR_OBJECT_HAS_SPILL(drro->drr_flags)) {
	/*
	* Currently allocated, the existing version of this object
	* may reference a spill block that is no longer allocated
	* at the source and needs to be freed.
	*/
	err = dmu_object_rm_spill(rwa->os, drro->drr_object, tx);
	}

	if (err != 0) {
	dmu_tx_commit(tx);
	return (SET_ERROR(EINVAL));
	}

	if (rwa->or_crypt_params_present) {
	/*
	* Set the crypt params for the buffer associated with this
	* range of dnodes. This causes the blkptr_t to have the
	* same crypt params (byteorder, salt, iv, mac) as on the
	* sending side.
	*
	* Since we are committing this tx now, it is possible for
	* the dnode block to end up on-disk with the incorrect MAC,
	* if subsequent objects in this block are received in a
	* different txg. However, since the dataset is marked as
	* inconsistent, no code paths will do a non-raw read (or
	* decrypt the block / verify the MAC). The receive code and
	* scrub code can safely do raw reads and verify the
	* checksum. They don't need to verify the MAC.
	*/
	dmu_buf_t *db = NULL;
	uint64_t offset = rwa->or_firstobj * DNODE_MIN_SIZE;

	err = dmu_buf_hold_by_dnode(DMU_META_DNODE(rwa->os),
	offset, FTAG, &db, DMU_READ_PREFETCH \| DMU_READ_NO_DECRYPT);
	if (err != 0) {
	dmu_tx_commit(tx);
	return (SET_ERROR(EINVAL));
	}

	dmu_buf_set_crypt_params(db, rwa->or_byteorder,
	rwa->or_salt, rwa->or_iv, rwa->or_mac, tx);

	dmu_buf_rele(db, FTAG);

	rwa->or_crypt_params_present = B_FALSE;
	}

	dmu_object_set_checksum(rwa->os, drro->drr_object,
	drro->drr_checksumtype, tx);
	dmu_object_set_compress(rwa->os, drro->drr_object,
	drro->drr_compress, tx);

	/* handle more restrictive dnode structuring for raw recvs */
	if (rwa->raw) {
	/*
	* Set the indirect block size, block shift, nlevels.
	* This will not fail because we ensured all of the
	* blocks were freed earlier if this is a new object.
	* For non-new objects block size and indirect block
	* shift cannot change and nlevels can only increase.
	*/
	ASSERT3U(new_blksz, ==, drro->drr_blksz);
	VERIFY0(dmu_object_set_blocksize(rwa->os, drro->drr_object,
	drro->drr_blksz, drro->drr_indblkshift, tx));
	VERIFY0(dmu_object_set_nlevels(rwa->os, drro->drr_object,
	drro->drr_nlevels, tx));

	/*
	* Set the maxblkid. This will always succeed because
	* we freed all blocks beyond the new maxblkid above.
	*/
	VERIFY0(dmu_object_set_maxblkid(rwa->os, drro->drr_object,
	drro->drr_maxblkid, tx));
	}

	if (data != NULL) {
	dmu_buf_t *db;
	dnode_t *dn;
	uint32_t flags = DMU_READ_NO_PREFETCH;

	if (rwa->raw)
	flags \|= DMU_READ_NO_DECRYPT;

	VERIFY0(dnode_hold(rwa->os, drro->drr_object, FTAG, &dn));
	VERIFY0(dmu_bonus_hold_by_dnode(dn, FTAG, &db, flags));

	dmu_buf_will_dirty(db, tx);

	ASSERT3U(db->db_size, >=, drro->drr_bonuslen);
	memcpy(db->db_data, data, DRR_OBJECT_PAYLOAD_SIZE(drro));

	/*
	* Raw bonus buffers have their byteorder determined by the
	* DRR_OBJECT_RANGE record.
	*/
	if (rwa->byteswap && !rwa->raw) {
	dmu_object_byteswap_t byteswap =
	DMU_OT_BYTESWAP(drro->drr_bonustype);
	dmu_ot_byteswap[byteswap].ob_func(db->db_data,
	DRR_OBJECT_PAYLOAD_SIZE(drro));
	}
	dmu_buf_rele(db, FTAG);
	dnode_rele(dn, FTAG);
	}

	/*
	* If the receive fails, we want the resume stream to start with the
	* same record that we last successfully received. There is no way to
	* request resume from the object record, but we can benefit from the
	* fact that sender always sends object record before anything else,
	* after which it will "resend" data at offset 0 and resume normally.
	*/
	save_resume_state(rwa, drro->drr_object, 0, tx);

	dmu_tx_commit(tx);

	return (0);
	}

	noinline static int
	receive_freeobjects(struct receive_writer_arg *rwa,
	struct drr_freeobjects *drrfo)
	{
	uint64_t obj;
	int next_err = 0;

	if (drrfo->drr_firstobj + drrfo->drr_numobjs < drrfo->drr_firstobj)
	return (SET_ERROR(EINVAL));

	for (obj = drrfo->drr_firstobj == 0 ? 1 : drrfo->drr_firstobj;
	obj < drrfo->drr_firstobj + drrfo->drr_numobjs &&
	obj < DN_MAX_OBJECT && next_err == 0;
	next_err = dmu_object_next(rwa->os, &obj, FALSE, 0)) {
	dmu_object_info_t doi;
	int err;

	err = dmu_object_info(rwa->os, obj, &doi);
	if (err == ENOENT)
	continue;
	else if (err != 0)
	return (err);

	err = dmu_free_long_object(rwa->os, obj);

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

	if (rwa->or_need_sync == ORNS_MAYBE)
	rwa->or_need_sync = ORNS_YES;
	}
	if (next_err != ESRCH)
	return (next_err);
	return (0);
	}

	/*
	* Note: if this fails, the caller will clean up any records left on the
	* rwa->write_batch list.
	*/
	static int
	flush_write_batch_impl(struct receive_writer_arg *rwa)
	{
	dnode_t *dn;
	int err;

	if (dnode_hold(rwa->os, rwa->last_object, FTAG, &dn) != 0)
	return (SET_ERROR(EINVAL));

	struct receive_record_arg *last_rrd = list_tail(&rwa->write_batch);
	struct drr_write *last_drrw = &last_rrd->header.drr_u.drr_write;

	struct receive_record_arg *first_rrd = list_head(&rwa->write_batch);
	struct drr_write *first_drrw = &first_rrd->header.drr_u.drr_write;

	ASSERT3U(rwa->last_object, ==, last_drrw->drr_object);
	ASSERT3U(rwa->last_offset, ==, last_drrw->drr_offset);

	dmu_tx_t *tx = dmu_tx_create(rwa->os);
	dmu_tx_hold_write_by_dnode(tx, dn, first_drrw->drr_offset,
	last_drrw->drr_offset - first_drrw->drr_offset +
	last_drrw->drr_logical_size);
	err = dmu_tx_assign(tx, TXG_WAIT);
	if (err != 0) {
	dmu_tx_abort(tx);
	dnode_rele(dn, FTAG);
	return (err);
	}

	struct receive_record_arg *rrd;
	while ((rrd = list_head(&rwa->write_batch)) != NULL) {
	struct drr_write *drrw = &rrd->header.drr_u.drr_write;
	abd_t *abd = rrd->abd;

	ASSERT3U(drrw->drr_object, ==, rwa->last_object);

	if (drrw->drr_logical_size != dn->dn_datablksz) {
	/*
	* The WRITE record is larger than the object's block
	* size. We must be receiving an incremental
	* large-block stream into a dataset that previously did
	* a non-large-block receive. Lightweight writes must
	* be exactly one block, so we need to decompress the
	* data (if compressed) and do a normal dmu_write().
	*/
	ASSERT3U(drrw->drr_logical_size, >, dn->dn_datablksz);
	if (DRR_WRITE_COMPRESSED(drrw)) {
	abd_t *decomp_abd =
	abd_alloc_linear(drrw->drr_logical_size,
	B_FALSE);

	err = zio_decompress_data(
	drrw->drr_compressiontype,
	abd, abd_to_buf(decomp_abd),
	abd_get_size(abd),
	abd_get_size(decomp_abd), NULL);

	if (err == 0) {
	dmu_write_by_dnode(dn,
	drrw->drr_offset,
	drrw->drr_logical_size,
	abd_to_buf(decomp_abd), tx);
	}
	abd_free(decomp_abd);
	} else {
	dmu_write_by_dnode(dn,
	drrw->drr_offset,
	drrw->drr_logical_size,
	abd_to_buf(abd), tx);
	}
	if (err == 0)
	abd_free(abd);
	} else {
	zio_prop_t zp = {0};
	dmu_write_policy(rwa->os, dn, 0, 0, &zp);

	zio_flag_t zio_flags = 0;

	if (rwa->raw) {
	zp.zp_encrypt = B_TRUE;
	zp.zp_compress = drrw->drr_compressiontype;
	zp.zp_byteorder = ZFS_HOST_BYTEORDER ^
	!!DRR_IS_RAW_BYTESWAPPED(drrw->drr_flags) ^
	rwa->byteswap;
	memcpy(zp.zp_salt, drrw->drr_salt,
	ZIO_DATA_SALT_LEN);
	memcpy(zp.zp_iv, drrw->drr_iv,
	ZIO_DATA_IV_LEN);
	memcpy(zp.zp_mac, drrw->drr_mac,
	ZIO_DATA_MAC_LEN);
	if (DMU_OT_IS_ENCRYPTED(zp.zp_type)) {
	zp.zp_nopwrite = B_FALSE;
	zp.zp_copies = MIN(zp.zp_copies,
	SPA_DVAS_PER_BP - 1);
	}
	zio_flags \|= ZIO_FLAG_RAW;
	} else if (DRR_WRITE_COMPRESSED(drrw)) {
	ASSERT3U(drrw->drr_compressed_size, >, 0);
	ASSERT3U(drrw->drr_logical_size, >=,
	drrw->drr_compressed_size);
	zp.zp_compress = drrw->drr_compressiontype;
	zio_flags \|= ZIO_FLAG_RAW_COMPRESS;
	} else if (rwa->byteswap) {
	/*
	* Note: compressed blocks never need to be
	* byteswapped, because WRITE records for
	* metadata blocks are never compressed. The
	* exception is raw streams, which are written
	* in the original byteorder, and the byteorder
	* bit is preserved in the BP by setting
	* zp_byteorder above.
	*/
	dmu_object_byteswap_t byteswap =
	DMU_OT_BYTESWAP(drrw->drr_type);
	dmu_ot_byteswap[byteswap].ob_func(
	abd_to_buf(abd),
	DRR_WRITE_PAYLOAD_SIZE(drrw));
	}

	/*
	* Since this data can't be read until the receive
	* completes, we can do a "lightweight" write for
	* improved performance.
	*/
	err = dmu_lightweight_write_by_dnode(dn,
	drrw->drr_offset, abd, &zp, zio_flags, tx);
	}

	if (err != 0) {
	/*
	* This rrd is left on the list, so the caller will
	* free it (and the abd).
	*/
	break;
	}

	/*
	* Note: If the receive fails, we want the resume stream to
	* start with the same record that we last successfully
	* received (as opposed to the next record), so that we can
	* verify that we are resuming from the correct location.
	*/
	save_resume_state(rwa, drrw->drr_object, drrw->drr_offset, tx);

	list_remove(&rwa->write_batch, rrd);
	kmem_free(rrd, sizeof (*rrd));
	}

	dmu_tx_commit(tx);
	dnode_rele(dn, FTAG);
	return (err);
	}

	noinline static int
	flush_write_batch(struct receive_writer_arg *rwa)
	{
	if (list_is_empty(&rwa->write_batch))
	return (0);
	int err = rwa->err;
	if (err == 0)
	err = flush_write_batch_impl(rwa);
	if (err != 0) {
	struct receive_record_arg *rrd;
	while ((rrd = list_remove_head(&rwa->write_batch)) != NULL) {
	abd_free(rrd->abd);
	kmem_free(rrd, sizeof (*rrd));
	}
	}
	ASSERT(list_is_empty(&rwa->write_batch));
	return (err);
	}

	noinline static int
	receive_process_write_record(struct receive_writer_arg *rwa,
	struct receive_record_arg *rrd)
	{
	int err = 0;

	ASSERT3U(rrd->header.drr_type, ==, DRR_WRITE);
	struct drr_write *drrw = &rrd->header.drr_u.drr_write;

	if (drrw->drr_offset + drrw->drr_logical_size < drrw->drr_offset \|\|
	!DMU_OT_IS_VALID(drrw->drr_type))
	return (SET_ERROR(EINVAL));

	if (rwa->heal) {
	blkptr_t *bp;
	dmu_buf_t *dbp;
	int flags = DB_RF_CANFAIL;

	if (rwa->raw)
	flags \|= DB_RF_NO_DECRYPT;

	if (rwa->byteswap) {
	dmu_object_byteswap_t byteswap =
	DMU_OT_BYTESWAP(drrw->drr_type);
	dmu_ot_byteswap[byteswap].ob_func(abd_to_buf(rrd->abd),
	DRR_WRITE_PAYLOAD_SIZE(drrw));
	}

	err = dmu_buf_hold_noread(rwa->os, drrw->drr_object,
	drrw->drr_offset, FTAG, &dbp);
	if (err != 0)
	return (err);

	/* Try to read the object to see if it needs healing */
	err = dbuf_read((dmu_buf_impl_t *)dbp, NULL, flags);
	/*
	* We only try to heal when dbuf_read() returns a ECKSUMs.
	* Other errors (even EIO) get returned to caller.
	* EIO indicates that the device is not present/accessible,
	* so writing to it will likely fail.
	* If the block is healthy, we don't want to overwrite it
	* unnecessarily.
	*/
	if (err != ECKSUM) {
	dmu_buf_rele(dbp, FTAG);
	return (err);
	}
	/* Make sure the on-disk block and recv record sizes match */
	if (drrw->drr_logical_size != dbp->db_size) {
	err = ENOTSUP;
	dmu_buf_rele(dbp, FTAG);
	return (err);
	}
	/* Get the block pointer for the corrupted block */
	bp = dmu_buf_get_blkptr(dbp);
	err = do_corrective_recv(rwa, drrw, rrd, bp);
	dmu_buf_rele(dbp, FTAG);
	return (err);
	}

	/*
	* For resuming to work, records must be in increasing order
	* by (object, offset).
	*/
	if (drrw->drr_object < rwa->last_object \|\|
	(drrw->drr_object == rwa->last_object &&
	drrw->drr_offset < rwa->last_offset)) {
	return (SET_ERROR(EINVAL));
	}

	struct receive_record_arg *first_rrd = list_head(&rwa->write_batch);
	struct drr_write *first_drrw = &first_rrd->header.drr_u.drr_write;
	uint64_t batch_size =
	MIN(zfs_recv_write_batch_size, DMU_MAX_ACCESS / 2);
	if (first_rrd != NULL &&
	(drrw->drr_object != first_drrw->drr_object \|\|
	drrw->drr_offset >= first_drrw->drr_offset + batch_size)) {
	err = flush_write_batch(rwa);
	if (err != 0)
	return (err);
	}

	rwa->last_object = drrw->drr_object;
	rwa->last_offset = drrw->drr_offset;

	if (rwa->last_object > rwa->max_object)
	rwa->max_object = rwa->last_object;

	list_insert_tail(&rwa->write_batch, rrd);
	/*
	* Return EAGAIN to indicate that we will use this rrd again,
	* so the caller should not free it
	*/
	return (EAGAIN);
	}

	static int
	receive_write_embedded(struct receive_writer_arg *rwa,
	struct drr_write_embedded drrwe, void data)
	{
	dmu_tx_t *tx;
	int err;

	if (drrwe->drr_offset + drrwe->drr_length < drrwe->drr_offset)
	return (SET_ERROR(EINVAL));

	if (drrwe->drr_psize > BPE_PAYLOAD_SIZE)
	return (SET_ERROR(EINVAL));

	if (drrwe->drr_etype >= NUM_BP_EMBEDDED_TYPES)
	return (SET_ERROR(EINVAL));
	if (drrwe->drr_compression >= ZIO_COMPRESS_FUNCTIONS)
	return (SET_ERROR(EINVAL));
	if (rwa->raw)
	return (SET_ERROR(EINVAL));

	if (drrwe->drr_object > rwa->max_object)
	rwa->max_object = drrwe->drr_object;

	tx = dmu_tx_create(rwa->os);

	dmu_tx_hold_write(tx, drrwe->drr_object,
	drrwe->drr_offset, drrwe->drr_length);
	err = dmu_tx_assign(tx, TXG_WAIT);
	if (err != 0) {
	dmu_tx_abort(tx);
	return (err);
	}

	dmu_write_embedded(rwa->os, drrwe->drr_object,
	drrwe->drr_offset, data, drrwe->drr_etype,
	drrwe->drr_compression, drrwe->drr_lsize, drrwe->drr_psize,
	rwa->byteswap ^ ZFS_HOST_BYTEORDER, tx);

	/* See comment in restore_write. */
	save_resume_state(rwa, drrwe->drr_object, drrwe->drr_offset, tx);
	dmu_tx_commit(tx);
	return (0);
	}

	static int
	receive_spill(struct receive_writer_arg rwa, struct drr_spill drrs,
	abd_t *abd)
	{
	dmu_buf_t db, db_spill;
	int err;

	if (drrs->drr_length < SPA_MINBLOCKSIZE \|\|
	drrs->drr_length > spa_maxblocksize(dmu_objset_spa(rwa->os)))
	return (SET_ERROR(EINVAL));

	/*
	* This is an unmodified spill block which was added to the stream
	* to resolve an issue with incorrectly removing spill blocks. It
	* should be ignored by current versions of the code which support
	* the DRR_FLAG_SPILL_BLOCK flag.
	*/
	if (rwa->spill && DRR_SPILL_IS_UNMODIFIED(drrs->drr_flags)) {
	abd_free(abd);
	return (0);
	}

	if (rwa->raw) {
	if (!DMU_OT_IS_VALID(drrs->drr_type) \|\|
	drrs->drr_compressiontype >= ZIO_COMPRESS_FUNCTIONS \|\|
	drrs->drr_compressed_size == 0)
	return (SET_ERROR(EINVAL));
	}

	if (dmu_object_info(rwa->os, drrs->drr_object, NULL) != 0)
	return (SET_ERROR(EINVAL));

	if (drrs->drr_object > rwa->max_object)
	rwa->max_object = drrs->drr_object;

	VERIFY0(dmu_bonus_hold(rwa->os, drrs->drr_object, FTAG, &db));
	if ((err = dmu_spill_hold_by_bonus(db, DMU_READ_NO_DECRYPT, FTAG,
	&db_spill)) != 0) {
	dmu_buf_rele(db, FTAG);
	return (err);
	}

	dmu_tx_t *tx = dmu_tx_create(rwa->os);

	dmu_tx_hold_spill(tx, db->db_object);

	err = dmu_tx_assign(tx, TXG_WAIT);
	if (err != 0) {
	dmu_buf_rele(db, FTAG);
	dmu_buf_rele(db_spill, FTAG);
	dmu_tx_abort(tx);
	return (err);
	}

	/*
	* Spill blocks may both grow and shrink. When a change in size
	* occurs any existing dbuf must be updated to match the logical
	* size of the provided arc_buf_t.
	*/
	if (db_spill->db_size != drrs->drr_length) {
	dmu_buf_will_fill(db_spill, tx, B_FALSE);
	VERIFY0(dbuf_spill_set_blksz(db_spill,
	drrs->drr_length, tx));
	}

	arc_buf_t *abuf;
	if (rwa->raw) {
	boolean_t byteorder = ZFS_HOST_BYTEORDER ^
	!!DRR_IS_RAW_BYTESWAPPED(drrs->drr_flags) ^
	rwa->byteswap;

	abuf = arc_loan_raw_buf(dmu_objset_spa(rwa->os),
	drrs->drr_object, byteorder, drrs->drr_salt,
	drrs->drr_iv, drrs->drr_mac, drrs->drr_type,
	drrs->drr_compressed_size, drrs->drr_length,
	drrs->drr_compressiontype, 0);
	} else {
	abuf = arc_loan_buf(dmu_objset_spa(rwa->os),
	DMU_OT_IS_METADATA(drrs->drr_type),
	drrs->drr_length);
	if (rwa->byteswap) {
	dmu_object_byteswap_t byteswap =
	DMU_OT_BYTESWAP(drrs->drr_type);
	dmu_ot_byteswap[byteswap].ob_func(abd_to_buf(abd),
	DRR_SPILL_PAYLOAD_SIZE(drrs));
	}
	}

	memcpy(abuf->b_data, abd_to_buf(abd), DRR_SPILL_PAYLOAD_SIZE(drrs));
	abd_free(abd);
	dbuf_assign_arcbuf((dmu_buf_impl_t *)db_spill, abuf, tx);

	dmu_buf_rele(db, FTAG);
	dmu_buf_rele(db_spill, FTAG);

	dmu_tx_commit(tx);
	return (0);
	}

	noinline static int
	receive_free(struct receive_writer_arg rwa, struct drr_free drrf)
	{
	int err;

	if (drrf->drr_length != -1ULL &&
	drrf->drr_offset + drrf->drr_length < drrf->drr_offset)
	return (SET_ERROR(EINVAL));

	if (dmu_object_info(rwa->os, drrf->drr_object, NULL) != 0)
	return (SET_ERROR(EINVAL));

	if (drrf->drr_object > rwa->max_object)
	rwa->max_object = drrf->drr_object;

	err = dmu_free_long_range(rwa->os, drrf->drr_object,
	drrf->drr_offset, drrf->drr_length);

	return (err);
	}

	static int
	receive_object_range(struct receive_writer_arg *rwa,
	struct drr_object_range *drror)
	{
	/*
	* By default, we assume this block is in our native format
	* (ZFS_HOST_BYTEORDER). We then take into account whether
	* the send stream is byteswapped (rwa->byteswap). Finally,
	* we need to byteswap again if this particular block was
	* in non-native format on the send side.
	*/
	boolean_t byteorder = ZFS_HOST_BYTEORDER ^ rwa->byteswap ^
	!!DRR_IS_RAW_BYTESWAPPED(drror->drr_flags);

	/*
	* Since dnode block sizes are constant, we should not need to worry
	* about making sure that the dnode block size is the same on the
	* sending and receiving sides for the time being. For non-raw sends,
	* this does not matter (and in fact we do not send a DRR_OBJECT_RANGE
	* record at all). Raw sends require this record type because the
	* encryption parameters are used to protect an entire block of bonus
	* buffers. If the size of dnode blocks ever becomes variable,
	* handling will need to be added to ensure that dnode block sizes
	* match on the sending and receiving side.
	*/
	if (drror->drr_numslots != DNODES_PER_BLOCK \|\|
	P2PHASE(drror->drr_firstobj, DNODES_PER_BLOCK) != 0 \|\|
	!rwa->raw)
	return (SET_ERROR(EINVAL));

	if (drror->drr_firstobj > rwa->max_object)
	rwa->max_object = drror->drr_firstobj;

	/*
	* The DRR_OBJECT_RANGE handling must be deferred to receive_object()
	* so that the block of dnodes is not written out when it's empty,
	* and converted to a HOLE BP.
	*/
	rwa->or_crypt_params_present = B_TRUE;
	rwa->or_firstobj = drror->drr_firstobj;
	rwa->or_numslots = drror->drr_numslots;
	memcpy(rwa->or_salt, drror->drr_salt, ZIO_DATA_SALT_LEN);
	memcpy(rwa->or_iv, drror->drr_iv, ZIO_DATA_IV_LEN);
	memcpy(rwa->or_mac, drror->drr_mac, ZIO_DATA_MAC_LEN);
	rwa->or_byteorder = byteorder;

	rwa->or_need_sync = ORNS_MAYBE;

	return (0);
	}

	/*
	* Until we have the ability to redact large ranges of data efficiently, we
	* process these records as frees.
	*/
	noinline static int
	receive_redact(struct receive_writer_arg rwa, struct drr_redact drrr)
	{
	struct drr_free drrf = {0};
	drrf.drr_length = drrr->drr_length;
	drrf.drr_object = drrr->drr_object;
	drrf.drr_offset = drrr->drr_offset;
	drrf.drr_toguid = drrr->drr_toguid;
	return (receive_free(rwa, &drrf));
	}

	/* used to destroy the drc_ds on error */
	static void
	dmu_recv_cleanup_ds(dmu_recv_cookie_t *drc)
	{
	dsl_dataset_t *ds = drc->drc_ds;
	ds_hold_flags_t dsflags;

	dsflags = (drc->drc_raw) ? DS_HOLD_FLAG_NONE : DS_HOLD_FLAG_DECRYPT;
	/*
	* Wait for the txg sync before cleaning up the receive. For
	* resumable receives, this ensures that our resume state has
	* been written out to disk. For raw receives, this ensures
	* that the user accounting code will not attempt to do anything
	* after we stopped receiving the dataset.
	*/
	txg_wait_synced(ds->ds_dir->dd_pool, 0);
	ds->ds_objset->os_raw_receive = B_FALSE;

	rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
	if (drc->drc_resumable && drc->drc_should_save &&
	!BP_IS_HOLE(dsl_dataset_get_blkptr(ds))) {
	rrw_exit(&ds->ds_bp_rwlock, FTAG);
	dsl_dataset_disown(ds, dsflags, dmu_recv_tag);
	} else {
	char name[ZFS_MAX_DATASET_NAME_LEN];
	rrw_exit(&ds->ds_bp_rwlock, FTAG);
	dsl_dataset_name(ds, name);
	dsl_dataset_disown(ds, dsflags, dmu_recv_tag);
	if (!drc->drc_heal)
	(void) dsl_destroy_head(name);
	}
	}

	static void
	receive_cksum(dmu_recv_cookie_t drc, int len, void buf)
	{
	if (drc->drc_byteswap) {
	(void) fletcher_4_incremental_byteswap(buf, len,
	&drc->drc_cksum);
	} else {
	(void) fletcher_4_incremental_native(buf, len, &drc->drc_cksum);
	}
	}

	/*
	* Read the payload into a buffer of size len, and update the current record's
	* payload field.
	* Allocate drc->drc_next_rrd and read the next record's header into
	* drc->drc_next_rrd->header.
	* Verify checksum of payload and next record.
	*/
	static int
	receive_read_payload_and_next_header(dmu_recv_cookie_t drc, int len, void buf)
	{
	int err;

	if (len != 0) {
	ASSERT3U(len, <=, SPA_MAXBLOCKSIZE);
	err = receive_read(drc, len, buf);
	if (err != 0)
	return (err);
	receive_cksum(drc, len, buf);

	/* note: rrd is NULL when reading the begin record's payload */
	if (drc->drc_rrd != NULL) {
	drc->drc_rrd->payload = buf;
	drc->drc_rrd->payload_size = len;
	drc->drc_rrd->bytes_read = drc->drc_bytes_read;
	}
	} else {
	ASSERT3P(buf, ==, NULL);
	}

	drc->drc_prev_cksum = drc->drc_cksum;

	drc->drc_next_rrd = kmem_zalloc(sizeof (*drc->drc_next_rrd), KM_SLEEP);
	err = receive_read(drc, sizeof (drc->drc_next_rrd->header),
	&drc->drc_next_rrd->header);
	drc->drc_next_rrd->bytes_read = drc->drc_bytes_read;

	if (err != 0) {
	kmem_free(drc->drc_next_rrd, sizeof (*drc->drc_next_rrd));
	drc->drc_next_rrd = NULL;
	return (err);
	}
	if (drc->drc_next_rrd->header.drr_type == DRR_BEGIN) {
	kmem_free(drc->drc_next_rrd, sizeof (*drc->drc_next_rrd));
	drc->drc_next_rrd = NULL;
	return (SET_ERROR(EINVAL));
	}

	/*
	* Note: checksum is of everything up to but not including the
	* checksum itself.
	*/
	ASSERT3U(offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum),
	==, sizeof (dmu_replay_record_t) - sizeof (zio_cksum_t));
	receive_cksum(drc,
	offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum),
	&drc->drc_next_rrd->header);

	zio_cksum_t cksum_orig =
	drc->drc_next_rrd->header.drr_u.drr_checksum.drr_checksum;
	zio_cksum_t *cksump =
	&drc->drc_next_rrd->header.drr_u.drr_checksum.drr_checksum;

	if (drc->drc_byteswap)
	byteswap_record(&drc->drc_next_rrd->header);

	if ((!ZIO_CHECKSUM_IS_ZERO(cksump)) &&
	!ZIO_CHECKSUM_EQUAL(drc->drc_cksum, *cksump)) {
	kmem_free(drc->drc_next_rrd, sizeof (*drc->drc_next_rrd));
	drc->drc_next_rrd = NULL;
	return (SET_ERROR(ECKSUM));
	}

	receive_cksum(drc, sizeof (cksum_orig), &cksum_orig);

	return (0);
	}

	/*
	* Issue the prefetch reads for any necessary indirect blocks.
	*
	* We use the object ignore list to tell us whether or not to issue prefetches
	* for a given object. We do this for both correctness (in case the blocksize
	* of an object has changed) and performance (if the object doesn't exist, don't
	* needlessly try to issue prefetches). We also trim the list as we go through
	* the stream to prevent it from growing to an unbounded size.
	*
	* The object numbers within will always be in sorted order, and any write
	* records we see will also be in sorted order, but they're not sorted with
	* respect to each other (i.e. we can get several object records before
	* receiving each object's write records). As a result, once we've reached a
	* given object number, we can safely remove any reference to lower object
	* numbers in the ignore list. In practice, we receive up to 32 object records
	* before receiving write records, so the list can have up to 32 nodes in it.
	*/
	static void
	receive_read_prefetch(dmu_recv_cookie_t *drc, uint64_t object, uint64_t offset,
	uint64_t length)
	{
	if (!objlist_exists(drc->drc_ignore_objlist, object)) {
	dmu_prefetch(drc->drc_os, object, 1, offset, length,
	ZIO_PRIORITY_SYNC_READ);
	}
	}

	/*
	* Read records off the stream, issuing any necessary prefetches.
	*/
	static int
	receive_read_record(dmu_recv_cookie_t *drc)
	{
	int err;

	switch (drc->drc_rrd->header.drr_type) {
	case DRR_OBJECT:
	{
	struct drr_object *drro =
	&drc->drc_rrd->header.drr_u.drr_object;
	uint32_t size = DRR_OBJECT_PAYLOAD_SIZE(drro);
	void *buf = NULL;
	dmu_object_info_t doi;

	if (size != 0)
	buf = kmem_zalloc(size, KM_SLEEP);

	err = receive_read_payload_and_next_header(drc, size, buf);
	if (err != 0) {
	kmem_free(buf, size);
	return (err);
	}
	err = dmu_object_info(drc->drc_os, drro->drr_object, &doi);
	/*
	* See receive_read_prefetch for an explanation why we're
	* storing this object in the ignore_obj_list.
	*/
	if (err == ENOENT \|\| err == EEXIST \|\|
	(err == 0 && doi.doi_data_block_size != drro->drr_blksz)) {
	objlist_insert(drc->drc_ignore_objlist,
	drro->drr_object);
	err = 0;
	}
	return (err);
	}
	case DRR_FREEOBJECTS:
	{
	err = receive_read_payload_and_next_header(drc, 0, NULL);
	return (err);
	}
	case DRR_WRITE:
	{
	struct drr_write *drrw = &drc->drc_rrd->header.drr_u.drr_write;
	int size = DRR_WRITE_PAYLOAD_SIZE(drrw);
	abd_t *abd = abd_alloc_linear(size, B_FALSE);
	err = receive_read_payload_and_next_header(drc, size,
	abd_to_buf(abd));
	if (err != 0) {
	abd_free(abd);
	return (err);
	}
	drc->drc_rrd->abd = abd;
	receive_read_prefetch(drc, drrw->drr_object, drrw->drr_offset,
	drrw->drr_logical_size);
	return (err);
	}
	case DRR_WRITE_EMBEDDED:
	{
	struct drr_write_embedded *drrwe =
	&drc->drc_rrd->header.drr_u.drr_write_embedded;
	uint32_t size = P2ROUNDUP(drrwe->drr_psize, 8);
	void *buf = kmem_zalloc(size, KM_SLEEP);

	err = receive_read_payload_and_next_header(drc, size, buf);
	if (err != 0) {
	kmem_free(buf, size);
	return (err);
	}

	receive_read_prefetch(drc, drrwe->drr_object, drrwe->drr_offset,
	drrwe->drr_length);
	return (err);
	}
	case DRR_FREE:
	case DRR_REDACT:
	{
	/*
	* It might be beneficial to prefetch indirect blocks here, but
	* we don't really have the data to decide for sure.
	*/
	err = receive_read_payload_and_next_header(drc, 0, NULL);
	return (err);
	}
	case DRR_END:
	{
	struct drr_end *drre = &drc->drc_rrd->header.drr_u.drr_end;
	if (!ZIO_CHECKSUM_EQUAL(drc->drc_prev_cksum,
	drre->drr_checksum))
	return (SET_ERROR(ECKSUM));
	return (0);
	}
	case DRR_SPILL:
	{
	struct drr_spill *drrs = &drc->drc_rrd->header.drr_u.drr_spill;
	int size = DRR_SPILL_PAYLOAD_SIZE(drrs);
	abd_t *abd = abd_alloc_linear(size, B_FALSE);
	err = receive_read_payload_and_next_header(drc, size,
	abd_to_buf(abd));
	if (err != 0)
	abd_free(abd);
	else
	drc->drc_rrd->abd = abd;
	return (err);
	}
	case DRR_OBJECT_RANGE:
	{
	err = receive_read_payload_and_next_header(drc, 0, NULL);
	return (err);

	}
	default:
	return (SET_ERROR(EINVAL));
	}
	}



	static void
	dprintf_drr(struct receive_record_arg *rrd, int err)
	{
	#ifdef ZFS_DEBUG
	switch (rrd->header.drr_type) {
	case DRR_OBJECT:
	{
	struct drr_object *drro = &rrd->header.drr_u.drr_object;
	dprintf("drr_type = OBJECT obj = %llu type = %u "
	"bonustype = %u blksz = %u bonuslen = %u cksumtype = %u "
	"compress = %u dn_slots = %u err = %d\n",
	(u_longlong_t)drro->drr_object, drro->drr_type,
	drro->drr_bonustype, drro->drr_blksz, drro->drr_bonuslen,
	drro->drr_checksumtype, drro->drr_compress,
	drro->drr_dn_slots, err);
	break;
	}
	case DRR_FREEOBJECTS:
	{
	struct drr_freeobjects *drrfo =
	&rrd->header.drr_u.drr_freeobjects;
	dprintf("drr_type = FREEOBJECTS firstobj = %llu "
	"numobjs = %llu err = %d\n",
	(u_longlong_t)drrfo->drr_firstobj,
	(u_longlong_t)drrfo->drr_numobjs, err);
	break;
	}
	case DRR_WRITE:
	{
	struct drr_write *drrw = &rrd->header.drr_u.drr_write;
	dprintf("drr_type = WRITE obj = %llu type = %u offset = %llu "
	"lsize = %llu cksumtype = %u flags = %u "
	"compress = %u psize = %llu err = %d\n",
	(u_longlong_t)drrw->drr_object, drrw->drr_type,
	(u_longlong_t)drrw->drr_offset,
	(u_longlong_t)drrw->drr_logical_size,
	drrw->drr_checksumtype, drrw->drr_flags,
	drrw->drr_compressiontype,
	(u_longlong_t)drrw->drr_compressed_size, err);
	break;
	}
	case DRR_WRITE_BYREF:
	{
	struct drr_write_byref *drrwbr =
	&rrd->header.drr_u.drr_write_byref;
	dprintf("drr_type = WRITE_BYREF obj = %llu offset = %llu "
	"length = %llu toguid = %llx refguid = %llx "
	"refobject = %llu refoffset = %llu cksumtype = %u "
	"flags = %u err = %d\n",
	(u_longlong_t)drrwbr->drr_object,
	(u_longlong_t)drrwbr->drr_offset,
	(u_longlong_t)drrwbr->drr_length,
	(u_longlong_t)drrwbr->drr_toguid,
	(u_longlong_t)drrwbr->drr_refguid,
	(u_longlong_t)drrwbr->drr_refobject,
	(u_longlong_t)drrwbr->drr_refoffset,
	drrwbr->drr_checksumtype, drrwbr->drr_flags, err);
	break;
	}
	case DRR_WRITE_EMBEDDED:
	{
	struct drr_write_embedded *drrwe =
	&rrd->header.drr_u.drr_write_embedded;
	dprintf("drr_type = WRITE_EMBEDDED obj = %llu offset = %llu "
	"length = %llu compress = %u etype = %u lsize = %u "
	"psize = %u err = %d\n",
	(u_longlong_t)drrwe->drr_object,
	(u_longlong_t)drrwe->drr_offset,
	(u_longlong_t)drrwe->drr_length,
	drrwe->drr_compression, drrwe->drr_etype,
	drrwe->drr_lsize, drrwe->drr_psize, err);
	break;
	}
	case DRR_FREE:
	{
	struct drr_free *drrf = &rrd->header.drr_u.drr_free;
	dprintf("drr_type = FREE obj = %llu offset = %llu "
	"length = %lld err = %d\n",
	(u_longlong_t)drrf->drr_object,
	(u_longlong_t)drrf->drr_offset,
	(longlong_t)drrf->drr_length,
	err);
	break;
	}
	case DRR_SPILL:
	{
	struct drr_spill *drrs = &rrd->header.drr_u.drr_spill;
	dprintf("drr_type = SPILL obj = %llu length = %llu "
	"err = %d\n", (u_longlong_t)drrs->drr_object,
	(u_longlong_t)drrs->drr_length, err);
	break;
	}
	case DRR_OBJECT_RANGE:
	{
	struct drr_object_range *drror =
	&rrd->header.drr_u.drr_object_range;
	dprintf("drr_type = OBJECT_RANGE firstobj = %llu "
	"numslots = %llu flags = %u err = %d\n",
	(u_longlong_t)drror->drr_firstobj,
	(u_longlong_t)drror->drr_numslots,
	drror->drr_flags, err);
	break;
	}
	default:
	return;
	}
	#endif
	}

	/*
	* Commit the records to the pool.
	*/
	static int
	receive_process_record(struct receive_writer_arg *rwa,
	struct receive_record_arg *rrd)
	{
	int err;

	/* Processing in order, therefore bytes_read should be increasing. */
	ASSERT3U(rrd->bytes_read, >=, rwa->bytes_read);
	rwa->bytes_read = rrd->bytes_read;

	/* We can only heal write records; other ones get ignored */
	if (rwa->heal && rrd->header.drr_type != DRR_WRITE) {
	if (rrd->abd != NULL) {
	abd_free(rrd->abd);
	rrd->abd = NULL;
	} else if (rrd->payload != NULL) {
	kmem_free(rrd->payload, rrd->payload_size);
	rrd->payload = NULL;
	}
	return (0);
	}

	if (!rwa->heal && rrd->header.drr_type != DRR_WRITE) {
	err = flush_write_batch(rwa);
	if (err != 0) {
	if (rrd->abd != NULL) {
	abd_free(rrd->abd);
	rrd->abd = NULL;
	rrd->payload = NULL;
	} else if (rrd->payload != NULL) {
	kmem_free(rrd->payload, rrd->payload_size);
	rrd->payload = NULL;
	}

	return (err);
	}
	}

	switch (rrd->header.drr_type) {
	case DRR_OBJECT:
	{
	struct drr_object *drro = &rrd->header.drr_u.drr_object;
	err = receive_object(rwa, drro, rrd->payload);
	kmem_free(rrd->payload, rrd->payload_size);
	rrd->payload = NULL;
	break;
	}
	case DRR_FREEOBJECTS:
	{
	struct drr_freeobjects *drrfo =
	&rrd->header.drr_u.drr_freeobjects;
	err = receive_freeobjects(rwa, drrfo);
	break;
	}
	case DRR_WRITE:
	{
	err = receive_process_write_record(rwa, rrd);
	if (rwa->heal) {
	/*
	* If healing - always free the abd after processing
	*/
	abd_free(rrd->abd);
	rrd->abd = NULL;
	} else if (err != EAGAIN) {
	/*
	* On success, a non-healing
	* receive_process_write_record() returns
	* EAGAIN to indicate that we do not want to free
	* the rrd or arc_buf.
	*/
	ASSERT(err != 0);
	abd_free(rrd->abd);
	rrd->abd = NULL;
	}
	break;
	}
	case DRR_WRITE_EMBEDDED:
	{
	struct drr_write_embedded *drrwe =
	&rrd->header.drr_u.drr_write_embedded;
	err = receive_write_embedded(rwa, drrwe, rrd->payload);
	kmem_free(rrd->payload, rrd->payload_size);
	rrd->payload = NULL;
	break;
	}
	case DRR_FREE:
	{
	struct drr_free *drrf = &rrd->header.drr_u.drr_free;
	err = receive_free(rwa, drrf);
	break;
	}
	case DRR_SPILL:
	{
	struct drr_spill *drrs = &rrd->header.drr_u.drr_spill;
	err = receive_spill(rwa, drrs, rrd->abd);
	if (err != 0)
	abd_free(rrd->abd);
	rrd->abd = NULL;
	rrd->payload = NULL;
	break;
	}
	case DRR_OBJECT_RANGE:
	{
	struct drr_object_range *drror =
	&rrd->header.drr_u.drr_object_range;
	err = receive_object_range(rwa, drror);
	break;
	}
	case DRR_REDACT:
	{
	struct drr_redact *drrr = &rrd->header.drr_u.drr_redact;
	err = receive_redact(rwa, drrr);
	break;
	}
	default:
	err = (SET_ERROR(EINVAL));
	}

	if (err != 0)
	dprintf_drr(rrd, err);

	return (err);
	}

	/*
	* dmu_recv_stream's worker thread; pull records off the queue, and then call
	* receive_process_record When we're done, signal the main thread and exit.
	*/
	static __attribute__((noreturn)) void
	receive_writer_thread(void *arg)
	{
	struct receive_writer_arg *rwa = arg;
	struct receive_record_arg *rrd;
	fstrans_cookie_t cookie = spl_fstrans_mark();

	for (rrd = bqueue_dequeue(&rwa->q); !rrd->eos_marker;
	rrd = bqueue_dequeue(&rwa->q)) {
	/*
	* If there's an error, the main thread will stop putting things
	* on the queue, but we need to clear everything in it before we
	* can exit.
	*/
	int err = 0;
	if (rwa->err == 0) {
	err = receive_process_record(rwa, rrd);
	} else if (rrd->abd != NULL) {
	abd_free(rrd->abd);
	rrd->abd = NULL;
	rrd->payload = NULL;
	} else if (rrd->payload != NULL) {
	kmem_free(rrd->payload, rrd->payload_size);
	rrd->payload = NULL;
	}
	/*
	* EAGAIN indicates that this record has been saved (on
	* raw->write_batch), and will be used again, so we don't
	* free it.
	* When healing data we always need to free the record.
	*/
	if (err != EAGAIN \|\| rwa->heal) {
	if (rwa->err == 0)
	rwa->err = err;
	kmem_free(rrd, sizeof (*rrd));
	}
	}
	kmem_free(rrd, sizeof (*rrd));

	if (rwa->heal) {
	zio_wait(rwa->heal_pio);
	} else {
	int err = flush_write_batch(rwa);
	if (rwa->err == 0)
	rwa->err = err;
	}
	mutex_enter(&rwa->mutex);
	rwa->done = B_TRUE;
	cv_signal(&rwa->cv);
	mutex_exit(&rwa->mutex);
	spl_fstrans_unmark(cookie);
	thread_exit();
	}

	static int
	resume_check(dmu_recv_cookie_t drc, nvlist_t begin_nvl)
	{
	uint64_t val;
	objset_t *mos = dmu_objset_pool(drc->drc_os)->dp_meta_objset;
	uint64_t dsobj = dmu_objset_id(drc->drc_os);
	uint64_t resume_obj, resume_off;

	if (nvlist_lookup_uint64(begin_nvl,
	"resume_object", &resume_obj) != 0 \|\|
	nvlist_lookup_uint64(begin_nvl,
	"resume_offset", &resume_off) != 0) {
	return (SET_ERROR(EINVAL));
	}
	VERIFY0(zap_lookup(mos, dsobj,
	DS_FIELD_RESUME_OBJECT, sizeof (val), 1, &val));
	if (resume_obj != val)
	return (SET_ERROR(EINVAL));
	VERIFY0(zap_lookup(mos, dsobj,
	DS_FIELD_RESUME_OFFSET, sizeof (val), 1, &val));
	if (resume_off != val)
	return (SET_ERROR(EINVAL));

	return (0);
	}

	/*
	* Read in the stream's records, one by one, and apply them to the pool. There
	* are two threads involved; the thread that calls this function will spin up a
	* worker thread, read the records off the stream one by one, and issue
	* prefetches for any necessary indirect blocks. It will then push the records
	* onto an internal blocking queue. The worker thread will pull the records off
	* the queue, and actually write the data into the DMU. This way, the worker
	* thread doesn't have to wait for reads to complete, since everything it needs
	* (the indirect blocks) will be prefetched.
	*
	* NB: callers must call dmu_recv_end() if this succeeds.
	*/
	int
	dmu_recv_stream(dmu_recv_cookie_t drc, offset_t voffp)
	{
	int err = 0;
	struct receive_writer_arg rwa = kmem_zalloc(sizeof (rwa), KM_SLEEP);

	if (dsl_dataset_has_resume_receive_state(drc->drc_ds)) {
	uint64_t bytes = 0;
	(void) zap_lookup(drc->drc_ds->ds_dir->dd_pool->dp_meta_objset,
	drc->drc_ds->ds_object, DS_FIELD_RESUME_BYTES,
	sizeof (bytes), 1, &bytes);
	drc->drc_bytes_read += bytes;
	}

	drc->drc_ignore_objlist = objlist_create();

	/* these were verified in dmu_recv_begin */
	ASSERT3U(DMU_GET_STREAM_HDRTYPE(drc->drc_drrb->drr_versioninfo), ==,
	DMU_SUBSTREAM);
	ASSERT3U(drc->drc_drrb->drr_type, <, DMU_OST_NUMTYPES);

	ASSERT(dsl_dataset_phys(drc->drc_ds)->ds_flags & DS_FLAG_INCONSISTENT);
	ASSERT0(drc->drc_os->os_encrypted &&
	(drc->drc_featureflags & DMU_BACKUP_FEATURE_EMBED_DATA));

	/* handle DSL encryption key payload */
	if (drc->drc_featureflags & DMU_BACKUP_FEATURE_RAW) {
	nvlist_t *keynvl = NULL;

	ASSERT(drc->drc_os->os_encrypted);
	ASSERT(drc->drc_raw);

	err = nvlist_lookup_nvlist(drc->drc_begin_nvl, "crypt_keydata",
	&keynvl);
	if (err != 0)
	goto out;

	if (!drc->drc_heal) {
	/*
	* If this is a new dataset we set the key immediately.
	* Otherwise we don't want to change the key until we
	* are sure the rest of the receive succeeded so we
	* stash the keynvl away until then.
	*/
	err = dsl_crypto_recv_raw(spa_name(drc->drc_os->os_spa),
	drc->drc_ds->ds_object, drc->drc_fromsnapobj,
	drc->drc_drrb->drr_type, keynvl, drc->drc_newfs);
	if (err != 0)
	goto out;
	}

	/* see comment in dmu_recv_end_sync() */
	drc->drc_ivset_guid = 0;
	(void) nvlist_lookup_uint64(keynvl, "to_ivset_guid",
	&drc->drc_ivset_guid);

	if (!drc->drc_newfs)
	drc->drc_keynvl = fnvlist_dup(keynvl);
	}

	if (drc->drc_featureflags & DMU_BACKUP_FEATURE_RESUMING) {
	err = resume_check(drc, drc->drc_begin_nvl);
	if (err != 0)
	goto out;
	}

	/*
	* For compatibility with recursive send streams, we do this here,
	* rather than in dmu_recv_begin. If we pull the next header too
	* early, and it's the END record, we break the `recv_skip` logic.
	*/
	if (drc->drc_drr_begin->drr_payloadlen == 0) {
	err = receive_read_payload_and_next_header(drc, 0, NULL);
	if (err != 0)
	goto out;
	}

	/*
	* If we failed before this point we will clean up any new resume
	* state that was created. Now that we've gotten past the initial
	* checks we are ok to retain that resume state.
	*/
	drc->drc_should_save = B_TRUE;

	(void) bqueue_init(&rwa->q, zfs_recv_queue_ff,
	MAX(zfs_recv_queue_length, 2 * zfs_max_recordsize),
	offsetof(struct receive_record_arg, node));
	cv_init(&rwa->cv, NULL, CV_DEFAULT, NULL);
	mutex_init(&rwa->mutex, NULL, MUTEX_DEFAULT, NULL);
	rwa->os = drc->drc_os;
	rwa->byteswap = drc->drc_byteswap;
	rwa->heal = drc->drc_heal;
	rwa->tofs = drc->drc_tofs;
	rwa->resumable = drc->drc_resumable;
	rwa->raw = drc->drc_raw;
	rwa->spill = drc->drc_spill;
	rwa->full = (drc->drc_drr_begin->drr_u.drr_begin.drr_fromguid == 0);
	rwa->os->os_raw_receive = drc->drc_raw;
	if (drc->drc_heal) {
	rwa->heal_pio = zio_root(drc->drc_os->os_spa, NULL, NULL,
	ZIO_FLAG_GODFATHER);
	}
	list_create(&rwa->write_batch, sizeof (struct receive_record_arg),
	offsetof(struct receive_record_arg, node.bqn_node));

	(void) thread_create(NULL, 0, receive_writer_thread, rwa, 0, curproc,
	TS_RUN, minclsyspri);
	/*
	* We're reading rwa->err without locks, which is safe since we are the
	* only reader, and the worker thread is the only writer. It's ok if we
	* miss a write for an iteration or two of the loop, since the writer
	* thread will keep freeing records we send it until we send it an eos
	* marker.
	*
	* We can leave this loop in 3 ways: First, if rwa->err is
	* non-zero. In that case, the writer thread will free the rrd we just
	* pushed. Second, if we're interrupted; in that case, either it's the
	* first loop and drc->drc_rrd was never allocated, or it's later, and
	* drc->drc_rrd has been handed off to the writer thread who will free
	* it. Finally, if receive_read_record fails or we're at the end of the
	* stream, then we free drc->drc_rrd and exit.
	*/
	while (rwa->err == 0) {
	- if (issig(JUSTLOOKING) && issig(FORREAL)) {
	+ if (issig()) {
	err = SET_ERROR(EINTR);
	break;
	}

	ASSERT3P(drc->drc_rrd, ==, NULL);
	drc->drc_rrd = drc->drc_next_rrd;
	drc->drc_next_rrd = NULL;
	/* Allocates and loads header into drc->drc_next_rrd */
	err = receive_read_record(drc);

	if (drc->drc_rrd->header.drr_type == DRR_END \|\| err != 0) {
	kmem_free(drc->drc_rrd, sizeof (*drc->drc_rrd));
	drc->drc_rrd = NULL;
	break;
	}

	bqueue_enqueue(&rwa->q, drc->drc_rrd,
	sizeof (struct receive_record_arg) +
	drc->drc_rrd->payload_size);
	drc->drc_rrd = NULL;
	}

	ASSERT3P(drc->drc_rrd, ==, NULL);
	drc->drc_rrd = kmem_zalloc(sizeof (*drc->drc_rrd), KM_SLEEP);
	drc->drc_rrd->eos_marker = B_TRUE;
	bqueue_enqueue_flush(&rwa->q, drc->drc_rrd, 1);

	mutex_enter(&rwa->mutex);
	while (!rwa->done) {
	/*
	* We need to use cv_wait_sig() so that any process that may
	* be sleeping here can still fork.
	*/
	(void) cv_wait_sig(&rwa->cv, &rwa->mutex);
	}
	mutex_exit(&rwa->mutex);

	/*
	* If we are receiving a full stream as a clone, all object IDs which
	* are greater than the maximum ID referenced in the stream are
	* by definition unused and must be freed.
	*/
	if (drc->drc_clone && drc->drc_drrb->drr_fromguid == 0) {
	uint64_t obj = rwa->max_object + 1;
	int free_err = 0;
	int next_err = 0;

	while (next_err == 0) {
	free_err = dmu_free_long_object(rwa->os, obj);
	if (free_err != 0 && free_err != ENOENT)
	break;

	next_err = dmu_object_next(rwa->os, &obj, FALSE, 0);
	}

	if (err == 0) {
	if (free_err != 0 && free_err != ENOENT)
	err = free_err;
	else if (next_err != ESRCH)
	err = next_err;
	}
	}

	cv_destroy(&rwa->cv);
	mutex_destroy(&rwa->mutex);
	bqueue_destroy(&rwa->q);
	list_destroy(&rwa->write_batch);
	if (err == 0)
	err = rwa->err;

	out:
	/*
	* If we hit an error before we started the receive_writer_thread
	* we need to clean up the next_rrd we create by processing the
	* DRR_BEGIN record.
	*/
	if (drc->drc_next_rrd != NULL)
	kmem_free(drc->drc_next_rrd, sizeof (*drc->drc_next_rrd));

	/*
	* The objset will be invalidated by dmu_recv_end() when we do
	* dsl_dataset_clone_swap_sync_impl().
	*/
	drc->drc_os = NULL;

	kmem_free(rwa, sizeof (*rwa));
	nvlist_free(drc->drc_begin_nvl);

	if (err != 0) {
	/*
	* Clean up references. If receive is not resumable,
	* destroy what we created, so we don't leave it in
	* the inconsistent state.
	*/
	dmu_recv_cleanup_ds(drc);
	nvlist_free(drc->drc_keynvl);
	}

	objlist_destroy(drc->drc_ignore_objlist);
	drc->drc_ignore_objlist = NULL;
	*voffp = drc->drc_voff;
	return (err);
	}

	static int
	dmu_recv_end_check(void arg, dmu_tx_t tx)
	{
	dmu_recv_cookie_t *drc = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	int error;

	ASSERT3P(drc->drc_ds->ds_owner, ==, dmu_recv_tag);

	if (drc->drc_heal) {
	error = 0;
	} else if (!drc->drc_newfs) {
	dsl_dataset_t *origin_head;

	error = dsl_dataset_hold(dp, drc->drc_tofs, FTAG, &origin_head);
	if (error != 0)
	return (error);
	if (drc->drc_force) {
	/*
	* We will destroy any snapshots in tofs (i.e. before
	* origin_head) that are after the origin (which is
	* the snap before drc_ds, because drc_ds can not
	* have any snaps of its own).
	*/
	uint64_t obj;

	obj = dsl_dataset_phys(origin_head)->ds_prev_snap_obj;
	while (obj !=
	dsl_dataset_phys(drc->drc_ds)->ds_prev_snap_obj) {
	dsl_dataset_t *snap;
	error = dsl_dataset_hold_obj(dp, obj, FTAG,
	&snap);
	if (error != 0)
	break;
	if (snap->ds_dir != origin_head->ds_dir)
	error = SET_ERROR(EINVAL);
	if (error == 0) {
	error = dsl_destroy_snapshot_check_impl(
	snap, B_FALSE);
	}
	obj = dsl_dataset_phys(snap)->ds_prev_snap_obj;
	dsl_dataset_rele(snap, FTAG);
	if (error != 0)
	break;
	}
	if (error != 0) {
	dsl_dataset_rele(origin_head, FTAG);
	return (error);
	}
	}
	if (drc->drc_keynvl != NULL) {
	error = dsl_crypto_recv_raw_key_check(drc->drc_ds,
	drc->drc_keynvl, tx);
	if (error != 0) {
	dsl_dataset_rele(origin_head, FTAG);
	return (error);
	}
	}

	error = dsl_dataset_clone_swap_check_impl(drc->drc_ds,
	origin_head, drc->drc_force, drc->drc_owner, tx);
	if (error != 0) {
	dsl_dataset_rele(origin_head, FTAG);
	return (error);
	}
	error = dsl_dataset_snapshot_check_impl(origin_head,
	drc->drc_tosnap, tx, B_TRUE, 1,
	drc->drc_cred, drc->drc_proc);
	dsl_dataset_rele(origin_head, FTAG);
	if (error != 0)
	return (error);

	error = dsl_destroy_head_check_impl(drc->drc_ds, 1);
	} else {
	error = dsl_dataset_snapshot_check_impl(drc->drc_ds,
	drc->drc_tosnap, tx, B_TRUE, 1,
	drc->drc_cred, drc->drc_proc);
	}
	return (error);
	}

	static void
	dmu_recv_end_sync(void arg, dmu_tx_t tx)
	{
	dmu_recv_cookie_t *drc = arg;
	dsl_pool_t *dp = dmu_tx_pool(tx);
	boolean_t encrypted = drc->drc_ds->ds_dir->dd_crypto_obj != 0;
	uint64_t newsnapobj = 0;

	spa_history_log_internal_ds(drc->drc_ds, "finish receiving",
	tx, "snap=%s", drc->drc_tosnap);
	drc->drc_ds->ds_objset->os_raw_receive = B_FALSE;

	if (drc->drc_heal) {
	if (drc->drc_keynvl != NULL) {
	nvlist_free(drc->drc_keynvl);
	drc->drc_keynvl = NULL;
	}
	} else if (!drc->drc_newfs) {
	dsl_dataset_t *origin_head;

	VERIFY0(dsl_dataset_hold(dp, drc->drc_tofs, FTAG,
	&origin_head));

	if (drc->drc_force) {
	/*
	* Destroy any snapshots of drc_tofs (origin_head)
	* after the origin (the snap before drc_ds).
	*/
	uint64_t obj;

	obj = dsl_dataset_phys(origin_head)->ds_prev_snap_obj;
	while (obj !=
	dsl_dataset_phys(drc->drc_ds)->ds_prev_snap_obj) {
	dsl_dataset_t *snap;
	VERIFY0(dsl_dataset_hold_obj(dp, obj, FTAG,
	&snap));
	ASSERT3P(snap->ds_dir, ==, origin_head->ds_dir);
	obj = dsl_dataset_phys(snap)->ds_prev_snap_obj;
	dsl_destroy_snapshot_sync_impl(snap,
	B_FALSE, tx);
	dsl_dataset_rele(snap, FTAG);
	}
	}
	if (drc->drc_keynvl != NULL) {
	dsl_crypto_recv_raw_key_sync(drc->drc_ds,
	drc->drc_keynvl, tx);
	nvlist_free(drc->drc_keynvl);
	drc->drc_keynvl = NULL;
	}

	VERIFY3P(drc->drc_ds->ds_prev, ==,
	origin_head->ds_prev);

	dsl_dataset_clone_swap_sync_impl(drc->drc_ds,
	origin_head, tx);
	/*
	* The objset was evicted by dsl_dataset_clone_swap_sync_impl,
	* so drc_os is no longer valid.
	*/
	drc->drc_os = NULL;

	dsl_dataset_snapshot_sync_impl(origin_head,
	drc->drc_tosnap, tx);

	/* set snapshot's creation time and guid */
	dmu_buf_will_dirty(origin_head->ds_prev->ds_dbuf, tx);
	dsl_dataset_phys(origin_head->ds_prev)->ds_creation_time =
	drc->drc_drrb->drr_creation_time;
	dsl_dataset_phys(origin_head->ds_prev)->ds_guid =
	drc->drc_drrb->drr_toguid;
	dsl_dataset_phys(origin_head->ds_prev)->ds_flags &=
	~DS_FLAG_INCONSISTENT;

	dmu_buf_will_dirty(origin_head->ds_dbuf, tx);
	dsl_dataset_phys(origin_head)->ds_flags &=
	~DS_FLAG_INCONSISTENT;

	newsnapobj =
	dsl_dataset_phys(origin_head)->ds_prev_snap_obj;

	dsl_dataset_rele(origin_head, FTAG);
	dsl_destroy_head_sync_impl(drc->drc_ds, tx);

	if (drc->drc_owner != NULL)
	VERIFY3P(origin_head->ds_owner, ==, drc->drc_owner);
	} else {
	dsl_dataset_t *ds = drc->drc_ds;

	dsl_dataset_snapshot_sync_impl(ds, drc->drc_tosnap, tx);

	/* set snapshot's creation time and guid */
	dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx);
	dsl_dataset_phys(ds->ds_prev)->ds_creation_time =
	drc->drc_drrb->drr_creation_time;
	dsl_dataset_phys(ds->ds_prev)->ds_guid =
	drc->drc_drrb->drr_toguid;
	dsl_dataset_phys(ds->ds_prev)->ds_flags &=
	~DS_FLAG_INCONSISTENT;

	dmu_buf_will_dirty(ds->ds_dbuf, tx);
	dsl_dataset_phys(ds)->ds_flags &= ~DS_FLAG_INCONSISTENT;
	if (dsl_dataset_has_resume_receive_state(ds)) {
	(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_FROMGUID, tx);
	(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_OBJECT, tx);
	(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_OFFSET, tx);
	(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_BYTES, tx);
	(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_TOGUID, tx);
	(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_TONAME, tx);
	(void) zap_remove(dp->dp_meta_objset, ds->ds_object,
	DS_FIELD_RESUME_REDACT_BOOKMARK_SNAPS, tx);
	}
	newsnapobj =
	dsl_dataset_phys(drc->drc_ds)->ds_prev_snap_obj;
	}

	/*
	* If this is a raw receive, the crypt_keydata nvlist will include
	* a to_ivset_guid for us to set on the new snapshot. This value
	* will override the value generated by the snapshot code. However,
	* this value may not be present, because older implementations of
	* the raw send code did not include this value, and we are still
	* allowed to receive them if the zfs_disable_ivset_guid_check
	* tunable is set, in which case we will leave the newly-generated
	* value.
	*/
	if (!drc->drc_heal && drc->drc_raw && drc->drc_ivset_guid != 0) {
	dmu_object_zapify(dp->dp_meta_objset, newsnapobj,
	DMU_OT_DSL_DATASET, tx);
	VERIFY0(zap_update(dp->dp_meta_objset, newsnapobj,
	DS_FIELD_IVSET_GUID, sizeof (uint64_t), 1,
	&drc->drc_ivset_guid, tx));
	}

	/*
	* Release the hold from dmu_recv_begin. This must be done before
	* we return to open context, so that when we free the dataset's dnode
	* we can evict its bonus buffer. Since the dataset may be destroyed
	* at this point (and therefore won't have a valid pointer to the spa)
	* we release the key mapping manually here while we do have a valid
	* pointer, if it exists.
	*/
	if (!drc->drc_raw && encrypted) {
	(void) spa_keystore_remove_mapping(dmu_tx_pool(tx)->dp_spa,
	drc->drc_ds->ds_object, drc->drc_ds);
	}
	dsl_dataset_disown(drc->drc_ds, 0, dmu_recv_tag);
	drc->drc_ds = NULL;
	}

	static int dmu_recv_end_modified_blocks = 3;

	static int
	dmu_recv_existing_end(dmu_recv_cookie_t *drc)
	{
	#ifdef _KERNEL
	/*
	* We will be destroying the ds; make sure its origin is unmounted if
	* necessary.
	*/
	char name[ZFS_MAX_DATASET_NAME_LEN];
	dsl_dataset_name(drc->drc_ds, name);
	zfs_destroy_unmount_origin(name);
	#endif

	return (dsl_sync_task(drc->drc_tofs,
	dmu_recv_end_check, dmu_recv_end_sync, drc,
	dmu_recv_end_modified_blocks, ZFS_SPACE_CHECK_NORMAL));
	}

	static int
	dmu_recv_new_end(dmu_recv_cookie_t *drc)
	{
	return (dsl_sync_task(drc->drc_tofs,
	dmu_recv_end_check, dmu_recv_end_sync, drc,
	dmu_recv_end_modified_blocks, ZFS_SPACE_CHECK_NORMAL));
	}

	int
	dmu_recv_end(dmu_recv_cookie_t drc, void owner)
	{
	int error;

	drc->drc_owner = owner;

	if (drc->drc_newfs)
	error = dmu_recv_new_end(drc);
	else
	error = dmu_recv_existing_end(drc);

	if (error != 0) {
	dmu_recv_cleanup_ds(drc);
	nvlist_free(drc->drc_keynvl);
	} else if (!drc->drc_heal) {
	if (drc->drc_newfs) {
	zvol_create_minor(drc->drc_tofs);
	}
	char *snapname = kmem_asprintf("%s@%s",
	drc->drc_tofs, drc->drc_tosnap);
	zvol_create_minor(snapname);
	kmem_strfree(snapname);
	}
	return (error);
	}

	/*
	* Return TRUE if this objset is currently being received into.
	*/
	boolean_t
	dmu_objset_is_receiving(objset_t *os)
	{
	return (os->os_dsl_dataset != NULL &&
	os->os_dsl_dataset->ds_owner == dmu_recv_tag);
	}

	ZFS_MODULE_PARAM(zfs_recv, zfs_recv_, queue_length, UINT, ZMOD_RW,
	"Maximum receive queue length");

	ZFS_MODULE_PARAM(zfs_recv, zfs_recv_, queue_ff, UINT, ZMOD_RW,
	"Receive queue fill fraction");

	ZFS_MODULE_PARAM(zfs_recv, zfs_recv_, write_batch_size, UINT, ZMOD_RW,
	"Maximum amount of writes to batch into one transaction");

	ZFS_MODULE_PARAM(zfs_recv, zfs_recv_, best_effort_corrective, INT, ZMOD_RW,
	"Ignore errors during corrective receive");
	/* END CSTYLED */
	diff --git a/sys/contrib/openzfs/module/zfs/dmu_redact.c b/sys/contrib/openzfs/module/zfs/dmu_redact.c
	index 5ac14edfca12..1feba0ba83de 100644
	--- a/sys/contrib/openzfs/module/zfs/dmu_redact.c
	+++ b/sys/contrib/openzfs/module/zfs/dmu_redact.c
	@@ -1,1199 +1,1199 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2017, 2018 by Delphix. All rights reserved.
	*/

	#include <sys/zfs_context.h>
	#include <sys/txg.h>
	#include <sys/dmu_objset.h>
	#include <sys/dmu_traverse.h>
	#include <sys/dmu_redact.h>
	#include <sys/bqueue.h>
	#include <sys/objlist.h>
	#include <sys/dmu_tx.h>
	#ifdef _KERNEL
	#include <sys/zfs_vfsops.h>
	#include <sys/zap.h>
	#include <sys/zfs_znode.h>
	#endif

	/*
	* This controls the number of entries in the buffer the redaction_list_update
	* synctask uses to buffer writes to the redaction list.
	*/
	static const int redact_sync_bufsize = 1024;

	/*
	* Controls how often to update the redaction list when creating a redaction
	* list.
	*/
	static const uint64_t redaction_list_update_interval_ns =
	1000 * 1000 * 1000ULL; /* 1s */

	/*
	* This tunable controls the length of the queues that zfs redact worker threads
	* use to communicate. If the dmu_redact_snap thread is blocking on these
	* queues, this variable may need to be increased. If there is a significant
	* slowdown at the start of a redact operation as these threads consume all the
	* available IO resources, or the queues are consuming too much memory, this
	* variable may need to be decreased.
	*/
	static const int zfs_redact_queue_length = 1024 * 1024;

	/*
	* These tunables control the fill fraction of the queues by zfs redact. The
	* fill fraction controls the frequency with which threads have to be
	* cv_signaled. If a lot of cpu time is being spent on cv_signal, then these
	* should be tuned down. If the queues empty before the signalled thread can
	* catch up, then these should be tuned up.
	*/
	static const uint64_t zfs_redact_queue_ff = 20;

	struct redact_record {
	bqueue_node_t ln;
	boolean_t eos_marker; /* Marks the end of the stream */
	uint64_t start_object;
	uint64_t start_blkid;
	uint64_t end_object;
	uint64_t end_blkid;
	uint8_t indblkshift;
	uint32_t datablksz;
	};

	struct redact_thread_arg {
	bqueue_t q;
	objset_t os; / Objset to traverse */
	dsl_dataset_t ds; / Dataset to traverse */
	struct redact_record *current_record;
	int error_code;
	boolean_t cancel;
	zbookmark_phys_t resume;
	objlist_t *deleted_objs;
	uint64_t *num_blocks_visited;
	uint64_t ignore_object; /* ignore further callbacks on this */
	uint64_t txg; /* txg to traverse since */
	};

	/*
	* The redaction node is a wrapper around the redaction record that is used
	* by the redaction merging thread to sort the records and determine overlaps.
	*
	* It contains two nodes; one sorts the records by their start_zb, and the other
	* sorts the records by their end_zb.
	*/
	struct redact_node {
	avl_node_t avl_node_start;
	avl_node_t avl_node_end;
	struct redact_record *record;
	struct redact_thread_arg *rt_arg;
	uint32_t thread_num;
	};

	struct merge_data {
	list_t md_redact_block_pending;
	redact_block_phys_t md_coalesce_block;
	uint64_t md_last_time;
	redact_block_phys_t md_furthest[TXG_SIZE];
	/* Lists of struct redact_block_list_node. */
	list_t md_blocks[TXG_SIZE];
	boolean_t md_synctask_txg[TXG_SIZE];
	uint64_t md_latest_synctask_txg;
	redaction_list_t *md_redaction_list;
	};

	/*
	* A wrapper around struct redact_block so it can be stored in a list_t.
	*/
	struct redact_block_list_node {
	redact_block_phys_t block;
	list_node_t node;
	};

	/*
	* We've found a new redaction candidate. In order to improve performance, we
	* coalesce these blocks when they're adjacent to each other. This function
	* handles that. If the new candidate block range is immediately after the
	* range we're building, coalesce it into the range we're building. Otherwise,
	* put the record we're building on the queue, and update the build pointer to
	* point to the new record.
	*/
	static void
	record_merge_enqueue(bqueue_t q, struct redact_record *build,
	struct redact_record *new)
	{
	if (new->eos_marker) {
	if (*build != NULL)
	bqueue_enqueue(q, build, sizeof (*build));
	bqueue_enqueue_flush(q, new, sizeof (*new));
	return;
	}
	if (*build == NULL) {
	*build = new;
	return;
	}
	struct redact_record curbuild = build;
	if ((curbuild->end_object == new->start_object &&
	curbuild->end_blkid + 1 == new->start_blkid &&
	curbuild->end_blkid != UINT64_MAX) \|\|
	(curbuild->end_object + 1 == new->start_object &&
	curbuild->end_blkid == UINT64_MAX && new->start_blkid == 0)) {
	curbuild->end_object = new->end_object;
	curbuild->end_blkid = new->end_blkid;
	kmem_free(new, sizeof (*new));
	} else {
	bqueue_enqueue(q, curbuild, sizeof (*curbuild));
	*build = new;
	}
	}
	#ifdef _KERNEL
	struct objnode {
	avl_node_t node;
	uint64_t obj;
	};

	static int
	objnode_compare(const void o1, const void o2)
	{
	const struct objnode *obj1 = o1;
	const struct objnode *obj2 = o2;
	if (obj1->obj < obj2->obj)
	return (-1);
	if (obj1->obj > obj2->obj)
	return (1);
	return (0);
	}


	static objlist_t *
	zfs_get_deleteq(objset_t *os)
	{
	objlist_t *deleteq_objlist = objlist_create();
	uint64_t deleteq_obj;
	zap_cursor_t zc;
	zap_attribute_t za;
	dmu_object_info_t doi;

	ASSERT3U(os->os_phys->os_type, ==, DMU_OST_ZFS);
	VERIFY0(dmu_object_info(os, MASTER_NODE_OBJ, &doi));
	ASSERT3U(doi.doi_type, ==, DMU_OT_MASTER_NODE);

	VERIFY0(zap_lookup(os, MASTER_NODE_OBJ,
	ZFS_UNLINKED_SET, sizeof (uint64_t), 1, &deleteq_obj));

	/*
	* In order to insert objects into the objlist, they must be in sorted
	* order. We don't know what order we'll get them out of the ZAP in, so
	* we insert them into and remove them from an avl_tree_t to sort them.
	*/
	avl_tree_t at;
	avl_create(&at, objnode_compare, sizeof (struct objnode),
	offsetof(struct objnode, node));

	for (zap_cursor_init(&zc, os, deleteq_obj);
	zap_cursor_retrieve(&zc, &za) == 0; zap_cursor_advance(&zc)) {
	struct objnode obj = kmem_zalloc(sizeof (obj), KM_SLEEP);
	obj->obj = za.za_first_integer;
	avl_add(&at, obj);
	}
	zap_cursor_fini(&zc);

	struct objnode next, found = avl_first(&at);
	while (found != NULL) {
	next = AVL_NEXT(&at, found);
	objlist_insert(deleteq_objlist, found->obj);
	found = next;
	}

	void *cookie = NULL;
	while ((found = avl_destroy_nodes(&at, &cookie)) != NULL)
	kmem_free(found, sizeof (*found));
	avl_destroy(&at);
	return (deleteq_objlist);
	}
	#endif

	/*
	* This is the callback function to traverse_dataset for the redaction threads
	* for dmu_redact_snap. This thread is responsible for creating redaction
	* records for all the data that is modified by the snapshots we're redacting
	* with respect to. Redaction records represent ranges of data that have been
	* modified by one of the redaction snapshots, and are stored in the
	* redact_record struct. We need to create redaction records for three
	* cases:
	*
	* First, if there's a normal write, we need to create a redaction record for
	* that block.
	*
	* Second, if there's a hole, we need to create a redaction record that covers
	* the whole range of the hole. If the hole is in the meta-dnode, it must cover
	* every block in all of the objects in the hole.
	*
	* Third, if there is a deleted object, we need to create a redaction record for
	* all of the blocks in that object.
	*/
	static int
	redact_cb(spa_t spa, zilog_t zilog, const blkptr_t *bp,
	const zbookmark_phys_t zb, const struct dnode_phys dnp, void *arg)
	{
	(void) spa, (void) zilog;
	struct redact_thread_arg *rta = arg;
	struct redact_record *record;

	ASSERT(zb->zb_object == DMU_META_DNODE_OBJECT \|\|
	zb->zb_object >= rta->resume.zb_object);

	if (rta->cancel)
	return (SET_ERROR(EINTR));

	if (rta->ignore_object == zb->zb_object)
	return (0);

	/*
	* If we're visiting a dnode, we need to handle the case where the
	* object has been deleted.
	*/
	if (zb->zb_level == ZB_DNODE_LEVEL) {
	ASSERT3U(zb->zb_level, ==, ZB_DNODE_LEVEL);

	if (zb->zb_object == 0)
	return (0);

	/*
	* If the object has been deleted, redact all of the blocks in
	* it.
	*/
	if (dnp->dn_type == DMU_OT_NONE \|\|
	objlist_exists(rta->deleted_objs, zb->zb_object)) {
	rta->ignore_object = zb->zb_object;
	record = kmem_zalloc(sizeof (struct redact_record),
	KM_SLEEP);

	record->eos_marker = B_FALSE;
	record->start_object = record->end_object =
	zb->zb_object;
	record->start_blkid = 0;
	record->end_blkid = UINT64_MAX;
	record_merge_enqueue(&rta->q,
	&rta->current_record, record);
	}
	return (0);
	} else if (zb->zb_level < 0) {
	return (0);
	} else if (zb->zb_level > 0 && !BP_IS_HOLE(bp)) {
	/*
	* If this is an indirect block, but not a hole, it doesn't
	* provide any useful information for redaction, so ignore it.
	*/
	return (0);
	}

	/*
	* At this point, there are two options left for the type of block we're
	* looking at. Either this is a hole (which could be in the dnode or
	* the meta-dnode), or it's a level 0 block of some sort. If it's a
	* hole, we create a redaction record that covers the whole range. If
	* the hole is in a dnode, we need to redact all the blocks in that
	* hole. If the hole is in the meta-dnode, we instead need to redact
	* all blocks in every object covered by that hole. If it's a level 0
	* block, we only need to redact that single block.
	*/
	record = kmem_zalloc(sizeof (struct redact_record), KM_SLEEP);
	record->eos_marker = B_FALSE;

	record->start_object = record->end_object = zb->zb_object;
	if (BP_IS_HOLE(bp)) {
	record->start_blkid = zb->zb_blkid *
	bp_span_in_blocks(dnp->dn_indblkshift, zb->zb_level);

	record->end_blkid = ((zb->zb_blkid + 1) *
	bp_span_in_blocks(dnp->dn_indblkshift, zb->zb_level)) - 1;

	if (zb->zb_object == DMU_META_DNODE_OBJECT) {
	record->start_object = record->start_blkid *
	((SPA_MINBLOCKSIZE * dnp->dn_datablkszsec) /
	sizeof (dnode_phys_t));
	record->start_blkid = 0;
	record->end_object = ((record->end_blkid +
	1) * ((SPA_MINBLOCKSIZE * dnp->dn_datablkszsec) /
	sizeof (dnode_phys_t))) - 1;
	record->end_blkid = UINT64_MAX;
	}
	} else if (zb->zb_level != 0 \|\|
	zb->zb_object == DMU_META_DNODE_OBJECT) {
	kmem_free(record, sizeof (*record));
	return (0);
	} else {
	record->start_blkid = record->end_blkid = zb->zb_blkid;
	}
	record->indblkshift = dnp->dn_indblkshift;
	record->datablksz = dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT;
	record_merge_enqueue(&rta->q, &rta->current_record, record);

	return (0);
	}

	static __attribute__((noreturn)) void
	redact_traverse_thread(void *arg)
	{
	struct redact_thread_arg *rt_arg = arg;
	int err;
	struct redact_record *data;
	#ifdef _KERNEL
	if (rt_arg->os->os_phys->os_type == DMU_OST_ZFS)
	rt_arg->deleted_objs = zfs_get_deleteq(rt_arg->os);
	else
	rt_arg->deleted_objs = objlist_create();
	#else
	rt_arg->deleted_objs = objlist_create();
	#endif

	err = traverse_dataset_resume(rt_arg->ds, rt_arg->txg,
	&rt_arg->resume, TRAVERSE_PRE \| TRAVERSE_PREFETCH_METADATA,
	redact_cb, rt_arg);

	if (err != EINTR)
	rt_arg->error_code = err;
	objlist_destroy(rt_arg->deleted_objs);
	data = kmem_zalloc(sizeof (*data), KM_SLEEP);
	data->eos_marker = B_TRUE;
	record_merge_enqueue(&rt_arg->q, &rt_arg->current_record, data);
	thread_exit();
	}

	static inline void
	create_zbookmark_from_obj_off(zbookmark_phys_t *zb, uint64_t object,
	uint64_t blkid)
	{
	zb->zb_object = object;
	zb->zb_level = 0;
	zb->zb_blkid = blkid;
	}

	/*
	* This is a utility function that can do the comparison for the start or ends
	* of the ranges in a redact_record.
	*/
	static int
	redact_range_compare(uint64_t obj1, uint64_t off1, uint32_t dbss1,
	uint64_t obj2, uint64_t off2, uint32_t dbss2)
	{
	zbookmark_phys_t z1, z2;
	create_zbookmark_from_obj_off(&z1, obj1, off1);
	create_zbookmark_from_obj_off(&z2, obj2, off2);

	return (zbookmark_compare(dbss1 >> SPA_MINBLOCKSHIFT, 0,
	dbss2 >> SPA_MINBLOCKSHIFT, 0, &z1, &z2));
	}

	/*
	* Compare two redaction records by their range's start location. Also makes
	* eos records always compare last. We use the thread number in the redact_node
	* to ensure that records do not compare equal (which is not allowed in our avl
	* trees).
	*/
	static int
	redact_node_compare_start(const void arg1, const void arg2)
	{
	const struct redact_node *rn1 = arg1;
	const struct redact_node *rn2 = arg2;
	const struct redact_record *rr1 = rn1->record;
	const struct redact_record *rr2 = rn2->record;
	if (rr1->eos_marker)
	return (1);
	if (rr2->eos_marker)
	return (-1);

	int cmp = redact_range_compare(rr1->start_object, rr1->start_blkid,
	rr1->datablksz, rr2->start_object, rr2->start_blkid,
	rr2->datablksz);
	if (cmp == 0)
	cmp = (rn1->thread_num < rn2->thread_num ? -1 : 1);
	return (cmp);
	}

	/*
	* Compare two redaction records by their range's end location. Also makes
	* eos records always compare last. We use the thread number in the redact_node
	* to ensure that records do not compare equal (which is not allowed in our avl
	* trees).
	*/
	static int
	redact_node_compare_end(const void arg1, const void arg2)
	{
	const struct redact_node *rn1 = arg1;
	const struct redact_node *rn2 = arg2;
	const struct redact_record *srr1 = rn1->record;
	const struct redact_record *srr2 = rn2->record;
	if (srr1->eos_marker)
	return (1);
	if (srr2->eos_marker)
	return (-1);

	int cmp = redact_range_compare(srr1->end_object, srr1->end_blkid,
	srr1->datablksz, srr2->end_object, srr2->end_blkid,
	srr2->datablksz);
	if (cmp == 0)
	cmp = (rn1->thread_num < rn2->thread_num ? -1 : 1);
	return (cmp);
	}

	/*
	* Utility function that compares two redaction records to determine if any part
	* of the "from" record is before any part of the "to" record. Also causes End
	* of Stream redaction records to compare after all others, so that the
	* redaction merging logic can stay simple.
	*/
	static boolean_t
	redact_record_before(const struct redact_record *from,
	const struct redact_record *to)
	{
	if (from->eos_marker == B_TRUE)
	return (B_FALSE);
	else if (to->eos_marker == B_TRUE)
	return (B_TRUE);
	return (redact_range_compare(from->start_object, from->start_blkid,
	from->datablksz, to->end_object, to->end_blkid,
	to->datablksz) <= 0);
	}

	/*
	* Pop a new redaction record off the queue, check that the records are in the
	* right order, and free the old data.
	*/
	static struct redact_record *
	get_next_redact_record(bqueue_t bq, struct redact_record prev)
	{
	struct redact_record *next = bqueue_dequeue(bq);
	ASSERT(redact_record_before(prev, next));
	kmem_free(prev, sizeof (*prev));
	return (next);
	}

	/*
	* Remove the given redaction node from both trees, pull a new redaction record
	* off the queue, free the old redaction record, update the redaction node, and
	* reinsert the node into the trees.
	*/
	static int
	update_avl_trees(avl_tree_t start_tree, avl_tree_t end_tree,
	struct redact_node *redact_node)
	{
	avl_remove(start_tree, redact_node);
	avl_remove(end_tree, redact_node);
	redact_node->record = get_next_redact_record(&redact_node->rt_arg->q,
	redact_node->record);
	avl_add(end_tree, redact_node);
	avl_add(start_tree, redact_node);
	return (redact_node->rt_arg->error_code);
	}

	/*
	* Synctask for updating redaction lists. We first take this txg's list of
	* redacted blocks and append those to the redaction list. We then update the
	* redaction list's bonus buffer. We store the furthest blocks we visited and
	* the list of snapshots that we're redacting with respect to. We need these so
	* that redacted sends and receives can be correctly resumed.
	*/
	static void
	redaction_list_update_sync(void arg, dmu_tx_t tx)
	{
	struct merge_data *md = arg;
	uint64_t txg = dmu_tx_get_txg(tx);
	list_t *list = &md->md_blocks[txg & TXG_MASK];
	redact_block_phys_t *furthest_visited =
	&md->md_furthest[txg & TXG_MASK];
	objset_t *mos = tx->tx_pool->dp_meta_objset;
	redaction_list_t *rl = md->md_redaction_list;
	int bufsize = redact_sync_bufsize;
	redact_block_phys_t buf = kmem_alloc(bufsize sizeof (*buf),
	KM_SLEEP);
	int index = 0;

	dmu_buf_will_dirty(rl->rl_dbuf, tx);

	for (struct redact_block_list_node *rbln = list_remove_head(list);
	rbln != NULL; rbln = list_remove_head(list)) {
	ASSERT3U(rbln->block.rbp_object, <=,
	furthest_visited->rbp_object);
	ASSERT(rbln->block.rbp_object < furthest_visited->rbp_object \|\|
	rbln->block.rbp_blkid <= furthest_visited->rbp_blkid);
	buf[index] = rbln->block;
	index++;
	if (index == bufsize) {
	dmu_write(mos, rl->rl_object,
	rl->rl_phys->rlp_num_entries * sizeof (*buf),
	bufsize * sizeof (*buf), buf, tx);
	rl->rl_phys->rlp_num_entries += bufsize;
	index = 0;
	}
	kmem_free(rbln, sizeof (*rbln));
	}
	if (index > 0) {
	dmu_write(mos, rl->rl_object, rl->rl_phys->rlp_num_entries *
	sizeof (buf), index sizeof (*buf), buf, tx);
	rl->rl_phys->rlp_num_entries += index;
	}
	kmem_free(buf, bufsize * sizeof (*buf));

	md->md_synctask_txg[txg & TXG_MASK] = B_FALSE;
	rl->rl_phys->rlp_last_object = furthest_visited->rbp_object;
	rl->rl_phys->rlp_last_blkid = furthest_visited->rbp_blkid;
	}

	static void
	commit_rl_updates(objset_t os, struct merge_data md, uint64_t object,
	uint64_t blkid)
	{
	dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(os->os_spa)->dp_mos_dir);
	dmu_tx_hold_space(tx, sizeof (struct redact_block_list_node));
	VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
	uint64_t txg = dmu_tx_get_txg(tx);
	if (!md->md_synctask_txg[txg & TXG_MASK]) {
	dsl_sync_task_nowait(dmu_tx_pool(tx),
	redaction_list_update_sync, md, tx);
	md->md_synctask_txg[txg & TXG_MASK] = B_TRUE;
	md->md_latest_synctask_txg = txg;
	}
	md->md_furthest[txg & TXG_MASK].rbp_object = object;
	md->md_furthest[txg & TXG_MASK].rbp_blkid = blkid;
	list_move_tail(&md->md_blocks[txg & TXG_MASK],
	&md->md_redact_block_pending);
	dmu_tx_commit(tx);
	md->md_last_time = gethrtime();
	}

	/*
	* We want to store the list of blocks that we're redacting in the bookmark's
	* redaction list. However, this list is stored in the MOS, which means it can
	* only be written to in syncing context. To get around this, we create a
	* synctask that will write to the mos for us. We tell it what to write by
	* a linked list for each current transaction group; every time we decide to
	* redact a block, we append it to the transaction group that is currently in
	* open context. We also update some progress information that the synctask
	* will store to enable resumable redacted sends.
	*/
	static void
	update_redaction_list(struct merge_data md, objset_t os,
	uint64_t object, uint64_t blkid, uint64_t endblkid, uint32_t blksz)
	{
	boolean_t enqueue = B_FALSE;
	redact_block_phys_t cur = {0};
	uint64_t count = endblkid - blkid + 1;
	while (count > REDACT_BLOCK_MAX_COUNT) {
	update_redaction_list(md, os, object, blkid,
	blkid + REDACT_BLOCK_MAX_COUNT - 1, blksz);
	blkid += REDACT_BLOCK_MAX_COUNT;
	count -= REDACT_BLOCK_MAX_COUNT;
	}
	redact_block_phys_t *coalesce = &md->md_coalesce_block;
	boolean_t new;
	if (coalesce->rbp_size_count == 0) {
	new = B_TRUE;
	enqueue = B_FALSE;
	} else {
	uint64_t old_count = redact_block_get_count(coalesce);
	if (coalesce->rbp_object == object &&
	coalesce->rbp_blkid + old_count == blkid &&
	old_count + count <= REDACT_BLOCK_MAX_COUNT) {
	ASSERT3U(redact_block_get_size(coalesce), ==, blksz);
	redact_block_set_count(coalesce, old_count + count);
	new = B_FALSE;
	enqueue = B_FALSE;
	} else {
	new = B_TRUE;
	enqueue = B_TRUE;
	}
	}

	if (new) {
	cur = *coalesce;
	coalesce->rbp_blkid = blkid;
	coalesce->rbp_object = object;

	redact_block_set_count(coalesce, count);
	redact_block_set_size(coalesce, blksz);
	}

	if (enqueue && redact_block_get_size(&cur) != 0) {
	struct redact_block_list_node *rbln =
	kmem_alloc(sizeof (struct redact_block_list_node),
	KM_SLEEP);
	rbln->block = cur;
	list_insert_tail(&md->md_redact_block_pending, rbln);
	}

	if (gethrtime() > md->md_last_time +
	redaction_list_update_interval_ns) {
	commit_rl_updates(os, md, object, blkid);
	}
	}

	/*
	* This thread merges all the redaction records provided by the worker threads,
	* and determines which blocks are redacted by all the snapshots. The algorithm
	* for doing so is similar to performing a merge in mergesort with n sub-lists
	* instead of 2, with some added complexity due to the fact that the entries are
	* ranges, not just single blocks. This algorithm relies on the fact that the
	* queues are sorted, which is ensured by the fact that traverse_dataset
	* traverses the dataset in a consistent order. We pull one entry off the front
	* of the queues of each secure dataset traversal thread. Then we repeat the
	* following: each record represents a range of blocks modified by one of the
	* redaction snapshots, and each block in that range may need to be redacted in
	* the send stream. Find the record with the latest start of its range, and the
	* record with the earliest end of its range. If the last start is before the
	* first end, then we know that the blocks in the range [last_start, first_end]
	* are covered by all of the ranges at the front of the queues, which means
	* every thread redacts that whole range. For example, let's say the ranges on
	* each queue look like this:
	*
	* Block Id 1 2 3 4 5 6 7 8 9 10 11
	* Thread 1 \| [====================]
	* Thread 2 \| [========]
	* Thread 3 \| [=================]
	*
	* Thread 3 has the last start (5), and the thread 2 has the last end (6). All
	* three threads modified the range [5,6], so that data should not be sent over
	* the wire. After we've determined whether or not to redact anything, we take
	* the record with the first end. We discard that record, and pull a new one
	* off the front of the queue it came from. In the above example, we would
	* discard Thread 2's record, and pull a new one. Let's say the next record we
	* pulled from Thread 2 covered range [10,11]. The new layout would look like
	* this:
	*
	* Block Id 1 2 3 4 5 6 7 8 9 10 11
	* Thread 1 \| [====================]
	* Thread 2 \| [==]
	* Thread 3 \| [=================]
	*
	* When we compare the last start (10, from Thread 2) and the first end (9, from
	* Thread 1), we see that the last start is greater than the first end.
	* Therefore, we do not redact anything from these records. We'll iterate by
	* replacing the record from Thread 1.
	*
	* We iterate by replacing the record with the lowest end because we know
	* that the record with the lowest end has helped us as much as it can. All the
	* ranges before it that we will ever redact have been redacted. In addition,
	* by replacing the one with the lowest end, we guarantee we catch all ranges
	* that need to be redacted. For example, if in the case above we had replaced
	* the record from Thread 1 instead, we might have ended up with the following:
	*
	* Block Id 1 2 3 4 5 6 7 8 9 10 11 12
	* Thread 1 \| [==]
	* Thread 2 \| [========]
	* Thread 3 \| [=================]
	*
	* If the next record from Thread 2 had been [8,10], for example, we should have
	* redacted part of that range, but because we updated Thread 1's record, we
	* missed it.
	*
	* We implement this algorithm by using two trees. The first sorts the
	* redaction records by their start_zb, and the second sorts them by their
	* end_zb. We use these to find the record with the last start and the record
	* with the first end. We create a record with that start and end, and send it
	* on. The overall runtime of this implementation is O(n log m), where n is the
	* total number of redaction records from all the different redaction snapshots,
	* and m is the number of redaction snapshots.
	*
	* If we redact with respect to zero snapshots, we create a redaction
	* record with the start object and blkid to 0, and the end object and blkid to
	* UINT64_MAX. This will result in us redacting every block.
	*/
	static int
	perform_thread_merge(bqueue_t *q, uint32_t num_threads,
	struct redact_thread_arg thread_args, boolean_t cancel)
	{
	struct redact_node *redact_nodes = NULL;
	avl_tree_t start_tree, end_tree;
	struct redact_record *record;
	struct redact_record *current_record = NULL;
	int err = 0;
	struct merge_data md = { {0} };
	list_create(&md.md_redact_block_pending,
	sizeof (struct redact_block_list_node),
	offsetof(struct redact_block_list_node, node));

	/*
	* If we're redacting with respect to zero snapshots, then no data is
	* permitted to be sent. We enqueue a record that redacts all blocks,
	* and an eos marker.
	*/
	if (num_threads == 0) {
	record = kmem_zalloc(sizeof (struct redact_record),
	KM_SLEEP);
	// We can't redact object 0, so don't try.
	record->start_object = 1;
	record->start_blkid = 0;
	record->end_object = record->end_blkid = UINT64_MAX;
	bqueue_enqueue(q, record, sizeof (*record));
	return (0);
	}
	redact_nodes = vmem_zalloc(num_threads *
	sizeof (*redact_nodes), KM_SLEEP);

	avl_create(&start_tree, redact_node_compare_start,
	sizeof (struct redact_node),
	offsetof(struct redact_node, avl_node_start));
	avl_create(&end_tree, redact_node_compare_end,
	sizeof (struct redact_node),
	offsetof(struct redact_node, avl_node_end));

	for (int i = 0; i < num_threads; i++) {
	struct redact_node *node = &redact_nodes[i];
	struct redact_thread_arg *targ = &thread_args[i];
	node->record = bqueue_dequeue(&targ->q);
	node->rt_arg = targ;
	node->thread_num = i;
	avl_add(&start_tree, node);
	avl_add(&end_tree, node);
	}

	/*
	* Once the first record in the end tree has returned EOS, every record
	* must be an EOS record, so we should stop.
	*/
	while (err == 0 && !((struct redact_node *)avl_first(&end_tree))->
	record->eos_marker) {
	if (*cancel) {
	err = EINTR;
	break;
	}
	struct redact_node *last_start = avl_last(&start_tree);
	struct redact_node *first_end = avl_first(&end_tree);

	/*
	* If the last start record is before the first end record,
	* then we have blocks that are redacted by all threads.
	* Therefore, we should redact them. Copy the record, and send
	* it to the main thread.
	*/
	if (redact_record_before(last_start->record,
	first_end->record)) {
	record = kmem_zalloc(sizeof (struct redact_record),
	KM_SLEEP);
	record = first_end->record;
	record->start_object = last_start->record->start_object;
	record->start_blkid = last_start->record->start_blkid;
	record_merge_enqueue(q, &current_record,
	record);
	}
	err = update_avl_trees(&start_tree, &end_tree, first_end);
	}

	/*
	* We're done; if we were cancelled, we need to cancel our workers and
	* clear out their queues. Either way, we need to remove every thread's
	* redact_node struct from the avl trees.
	*/
	for (int i = 0; i < num_threads; i++) {
	if (err != 0) {
	thread_args[i].cancel = B_TRUE;
	while (!redact_nodes[i].record->eos_marker) {
	(void) update_avl_trees(&start_tree, &end_tree,
	&redact_nodes[i]);
	}
	}
	avl_remove(&start_tree, &redact_nodes[i]);
	avl_remove(&end_tree, &redact_nodes[i]);
	kmem_free(redact_nodes[i].record,
	sizeof (struct redact_record));
	bqueue_destroy(&thread_args[i].q);
	}

	avl_destroy(&start_tree);
	avl_destroy(&end_tree);
	vmem_free(redact_nodes, num_threads * sizeof (*redact_nodes));
	if (current_record != NULL)
	bqueue_enqueue(q, current_record, sizeof (*current_record));
	return (err);
	}

	struct redact_merge_thread_arg {
	bqueue_t q;
	spa_t *spa;
	int numsnaps;
	struct redact_thread_arg *thr_args;
	boolean_t cancel;
	int error_code;
	};

	static __attribute__((noreturn)) void
	redact_merge_thread(void *arg)
	{
	struct redact_merge_thread_arg *rmta = arg;
	rmta->error_code = perform_thread_merge(&rmta->q,
	rmta->numsnaps, rmta->thr_args, &rmta->cancel);
	struct redact_record rec = kmem_zalloc(sizeof (rec), KM_SLEEP);
	rec->eos_marker = B_TRUE;
	bqueue_enqueue_flush(&rmta->q, rec, 1);
	thread_exit();
	}

	/*
	* Find the next object in or after the redaction range passed in, and hold
	* its dnode with the provided tag. Also update *object to contain the new
	* object number.
	*/
	static int
	hold_next_object(objset_t os, struct redact_record rec, const void *tag,
	uint64_t object, dnode_t *dn)
	{
	int err = 0;
	if (*dn != NULL)
	dnode_rele(*dn, tag);
	*dn = NULL;
	if (*object < rec->start_object) {
	*object = rec->start_object - 1;
	}
	err = dmu_object_next(os, object, B_FALSE, 0);
	if (err != 0)
	return (err);

	err = dnode_hold(os, *object, tag, dn);
	while (err == 0 && (*object < rec->start_object \|\|
	DMU_OT_IS_METADATA((*dn)->dn_type))) {
	dnode_rele(*dn, tag);
	*dn = NULL;
	err = dmu_object_next(os, object, B_FALSE, 0);
	if (err != 0)
	break;
	err = dnode_hold(os, *object, tag, dn);
	}
	return (err);
	}

	static int
	perform_redaction(objset_t os, redaction_list_t rl,
	struct redact_merge_thread_arg *rmta)
	{
	int err = 0;
	bqueue_t *q = &rmta->q;
	struct redact_record *rec = NULL;
	struct merge_data md = { {0} };

	list_create(&md.md_redact_block_pending,
	sizeof (struct redact_block_list_node),
	offsetof(struct redact_block_list_node, node));
	md.md_redaction_list = rl;

	for (int i = 0; i < TXG_SIZE; i++) {
	list_create(&md.md_blocks[i],
	sizeof (struct redact_block_list_node),
	offsetof(struct redact_block_list_node, node));
	}
	dnode_t *dn = NULL;
	uint64_t prev_obj = 0;
	for (rec = bqueue_dequeue(q); !rec->eos_marker && err == 0;
	rec = get_next_redact_record(q, rec)) {
	ASSERT3U(rec->start_object, !=, 0);
	uint64_t object;
	if (prev_obj != rec->start_object) {
	object = rec->start_object - 1;
	err = hold_next_object(os, rec, FTAG, &object, &dn);
	} else {
	object = prev_obj;
	}
	while (err == 0 && object <= rec->end_object) {
	- if (issig(JUSTLOOKING) && issig(FORREAL)) {
	+ if (issig()) {
	err = EINTR;
	break;
	}
	/*
	* Part of the current object is contained somewhere in
	* the range covered by rec.
	*/
	uint64_t startblkid;
	uint64_t endblkid;
	uint64_t maxblkid = dn->dn_phys->dn_maxblkid;

	if (rec->start_object < object)
	startblkid = 0;
	else if (rec->start_blkid > maxblkid)
	break;
	else
	startblkid = rec->start_blkid;

	if (rec->end_object > object \|\| rec->end_blkid >
	maxblkid) {
	endblkid = maxblkid;
	} else {
	endblkid = rec->end_blkid;
	}
	update_redaction_list(&md, os, object, startblkid,
	endblkid, dn->dn_datablksz);

	if (object == rec->end_object)
	break;
	err = hold_next_object(os, rec, FTAG, &object, &dn);
	}
	if (err == ESRCH)
	err = 0;
	if (dn != NULL)
	prev_obj = object;
	}
	if (err == 0 && dn != NULL)
	dnode_rele(dn, FTAG);

	if (err == ESRCH)
	err = 0;
	rmta->cancel = B_TRUE;
	while (!rec->eos_marker)
	rec = get_next_redact_record(q, rec);
	kmem_free(rec, sizeof (*rec));

	/*
	* There may be a block that's being coalesced, sync that out before we
	* return.
	*/
	if (err == 0 && md.md_coalesce_block.rbp_size_count != 0) {
	struct redact_block_list_node *rbln =
	kmem_alloc(sizeof (struct redact_block_list_node),
	KM_SLEEP);
	rbln->block = md.md_coalesce_block;
	list_insert_tail(&md.md_redact_block_pending, rbln);
	}
	commit_rl_updates(os, &md, UINT64_MAX, UINT64_MAX);

	/*
	* Wait for all the redaction info to sync out before we return, so that
	* anyone who attempts to resume this redaction will have all the data
	* they need.
	*/
	dsl_pool_t *dp = spa_get_dsl(os->os_spa);
	if (md.md_latest_synctask_txg != 0)
	txg_wait_synced(dp, md.md_latest_synctask_txg);
	for (int i = 0; i < TXG_SIZE; i++)
	list_destroy(&md.md_blocks[i]);
	return (err);
	}

	static boolean_t
	redact_snaps_contains(uint64_t *snaps, uint64_t num_snaps, uint64_t guid)
	{
	for (int i = 0; i < num_snaps; i++) {
	if (snaps[i] == guid)
	return (B_TRUE);
	}
	return (B_FALSE);
	}

	int
	dmu_redact_snap(const char snapname, nvlist_t redactnvl,
	const char *redactbook)
	{
	int err = 0;
	dsl_pool_t *dp = NULL;
	dsl_dataset_t *ds = NULL;
	int numsnaps = 0;
	objset_t *os;
	struct redact_thread_arg *args = NULL;
	redaction_list_t *new_rl = NULL;
	char *newredactbook;

	if ((err = dsl_pool_hold(snapname, FTAG, &dp)) != 0)
	return (err);

	newredactbook = kmem_zalloc(sizeof (char) * ZFS_MAX_DATASET_NAME_LEN,
	KM_SLEEP);

	if ((err = dsl_dataset_hold_flags(dp, snapname, DS_HOLD_FLAG_DECRYPT,
	FTAG, &ds)) != 0) {
	goto out;
	}
	dsl_dataset_long_hold(ds, FTAG);
	if (!ds->ds_is_snapshot \|\| dmu_objset_from_ds(ds, &os) != 0) {
	err = EINVAL;
	goto out;
	}
	if (dsl_dataset_feature_is_active(ds, SPA_FEATURE_REDACTED_DATASETS)) {
	err = EALREADY;
	goto out;
	}

	numsnaps = fnvlist_num_pairs(redactnvl);
	if (numsnaps > 0)
	args = vmem_zalloc(numsnaps * sizeof (*args), KM_SLEEP);

	nvpair_t *pair = NULL;
	for (int i = 0; i < numsnaps; i++) {
	pair = nvlist_next_nvpair(redactnvl, pair);
	const char *name = nvpair_name(pair);
	struct redact_thread_arg *rta = &args[i];
	err = dsl_dataset_hold_flags(dp, name, DS_HOLD_FLAG_DECRYPT,
	FTAG, &rta->ds);
	if (err != 0)
	break;
	/*
	* We want to do the long hold before we can get any other
	* errors, because the cleanup code will release the long
	* hold if rta->ds is filled in.
	*/
	dsl_dataset_long_hold(rta->ds, FTAG);

	err = dmu_objset_from_ds(rta->ds, &rta->os);
	if (err != 0)
	break;
	if (!dsl_dataset_is_before(rta->ds, ds, 0)) {
	err = EINVAL;
	break;
	}
	if (dsl_dataset_feature_is_active(rta->ds,
	SPA_FEATURE_REDACTED_DATASETS)) {
	err = EALREADY;
	break;

	}
	}
	if (err != 0)
	goto out;
	VERIFY3P(nvlist_next_nvpair(redactnvl, pair), ==, NULL);

	boolean_t resuming = B_FALSE;
	zfs_bookmark_phys_t bookmark;

	(void) strlcpy(newredactbook, snapname, ZFS_MAX_DATASET_NAME_LEN);
	char *c = strchr(newredactbook, '@');
	ASSERT3P(c, !=, NULL);
	int n = snprintf(c, ZFS_MAX_DATASET_NAME_LEN - (c - newredactbook),
	"#%s", redactbook);
	if (n >= ZFS_MAX_DATASET_NAME_LEN - (c - newredactbook)) {
	dsl_pool_rele(dp, FTAG);
	kmem_free(newredactbook,
	sizeof (char) * ZFS_MAX_DATASET_NAME_LEN);
	if (args != NULL)
	vmem_free(args, numsnaps * sizeof (*args));
	return (SET_ERROR(ENAMETOOLONG));
	}
	err = dsl_bookmark_lookup(dp, newredactbook, NULL, &bookmark);
	if (err == 0) {
	resuming = B_TRUE;
	if (bookmark.zbm_redaction_obj == 0) {
	err = EEXIST;
	goto out;
	}
	err = dsl_redaction_list_hold_obj(dp,
	bookmark.zbm_redaction_obj, FTAG, &new_rl);
	if (err != 0) {
	err = EIO;
	goto out;
	}
	dsl_redaction_list_long_hold(dp, new_rl, FTAG);
	if (new_rl->rl_phys->rlp_num_snaps != numsnaps) {
	err = ESRCH;
	goto out;
	}
	for (int i = 0; i < numsnaps; i++) {
	struct redact_thread_arg *rta = &args[i];
	if (!redact_snaps_contains(new_rl->rl_phys->rlp_snaps,
	new_rl->rl_phys->rlp_num_snaps,
	dsl_dataset_phys(rta->ds)->ds_guid)) {
	err = ESRCH;
	goto out;
	}
	}
	if (new_rl->rl_phys->rlp_last_blkid == UINT64_MAX &&
	new_rl->rl_phys->rlp_last_object == UINT64_MAX) {
	err = EEXIST;
	goto out;
	}
	dsl_pool_rele(dp, FTAG);
	dp = NULL;
	} else {
	uint64_t *guids = NULL;
	if (numsnaps > 0) {
	guids = vmem_zalloc(numsnaps * sizeof (uint64_t),
	KM_SLEEP);
	}
	for (int i = 0; i < numsnaps; i++) {
	struct redact_thread_arg *rta = &args[i];
	guids[i] = dsl_dataset_phys(rta->ds)->ds_guid;
	}

	dsl_pool_rele(dp, FTAG);
	dp = NULL;
	err = dsl_bookmark_create_redacted(newredactbook, snapname,
	numsnaps, guids, FTAG, &new_rl);
	vmem_free(guids, numsnaps * sizeof (uint64_t));
	if (err != 0)
	goto out;
	}

	for (int i = 0; i < numsnaps; i++) {
	struct redact_thread_arg *rta = &args[i];
	(void) bqueue_init(&rta->q, zfs_redact_queue_ff,
	zfs_redact_queue_length,
	offsetof(struct redact_record, ln));
	if (resuming) {
	rta->resume.zb_blkid =
	new_rl->rl_phys->rlp_last_blkid;
	rta->resume.zb_object =
	new_rl->rl_phys->rlp_last_object;
	}
	rta->txg = dsl_dataset_phys(ds)->ds_creation_txg;
	(void) thread_create(NULL, 0, redact_traverse_thread, rta,
	0, curproc, TS_RUN, minclsyspri);
	}

	struct redact_merge_thread_arg *rmta;
	rmta = kmem_zalloc(sizeof (struct redact_merge_thread_arg), KM_SLEEP);

	(void) bqueue_init(&rmta->q, zfs_redact_queue_ff,
	zfs_redact_queue_length, offsetof(struct redact_record, ln));
	rmta->numsnaps = numsnaps;
	rmta->spa = os->os_spa;
	rmta->thr_args = args;
	(void) thread_create(NULL, 0, redact_merge_thread, rmta, 0, curproc,
	TS_RUN, minclsyspri);
	err = perform_redaction(os, new_rl, rmta);
	bqueue_destroy(&rmta->q);
	kmem_free(rmta, sizeof (struct redact_merge_thread_arg));

	out:
	kmem_free(newredactbook, sizeof (char) * ZFS_MAX_DATASET_NAME_LEN);

	if (new_rl != NULL) {
	dsl_redaction_list_long_rele(new_rl, FTAG);
	dsl_redaction_list_rele(new_rl, FTAG);
	}
	for (int i = 0; i < numsnaps; i++) {
	struct redact_thread_arg *rta = &args[i];
	/*
	* rta->ds may be NULL if we got an error while filling
	* it in.
	*/
	if (rta->ds != NULL) {
	dsl_dataset_long_rele(rta->ds, FTAG);
	dsl_dataset_rele_flags(rta->ds,
	DS_HOLD_FLAG_DECRYPT, FTAG);
	}
	}

	if (args != NULL)
	vmem_free(args, numsnaps * sizeof (*args));
	if (dp != NULL)
	dsl_pool_rele(dp, FTAG);
	if (ds != NULL) {
	dsl_dataset_long_rele(ds, FTAG);
	dsl_dataset_rele_flags(ds, DS_HOLD_FLAG_DECRYPT, FTAG);
	}
	return (SET_ERROR(err));

	}
	diff --git a/sys/contrib/openzfs/module/zfs/dmu_send.c b/sys/contrib/openzfs/module/zfs/dmu_send.c
	index b6cc2f0a5e91..cb2b62fed313 100644
	--- a/sys/contrib/openzfs/module/zfs/dmu_send.c
	+++ b/sys/contrib/openzfs/module/zfs/dmu_send.c
	@@ -1,3122 +1,3122 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2011, 2018 by Delphix. All rights reserved.
	* Copyright (c) 2014, Joyent, Inc. All rights reserved.
	* Copyright 2014 HybridCluster. All rights reserved.
	* Copyright 2016 RackTop Systems.
	* Copyright (c) 2016 Actifio, Inc. All rights reserved.
	* Copyright (c) 2019, Klara Inc.
	* Copyright (c) 2019, Allan Jude
	*/

	#include <sys/dmu.h>
	#include <sys/dmu_impl.h>
	#include <sys/dmu_tx.h>
	#include <sys/dbuf.h>
	#include <sys/dnode.h>
	#include <sys/zfs_context.h>
	#include <sys/dmu_objset.h>
	#include <sys/dmu_traverse.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_dir.h>
	#include <sys/dsl_prop.h>
	#include <sys/dsl_pool.h>
	#include <sys/dsl_synctask.h>
	#include <sys/spa_impl.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/zap.h>
	#include <sys/zio_checksum.h>
	#include <sys/zfs_znode.h>
	#include <zfs_fletcher.h>
	#include <sys/avl.h>
	#include <sys/ddt.h>
	#include <sys/zfs_onexit.h>
	#include <sys/dmu_send.h>
	#include <sys/dmu_recv.h>
	#include <sys/dsl_destroy.h>
	#include <sys/blkptr.h>
	#include <sys/dsl_bookmark.h>
	#include <sys/zfeature.h>
	#include <sys/bqueue.h>
	#include <sys/zvol.h>
	#include <sys/policy.h>
	#include <sys/objlist.h>
	#ifdef _KERNEL
	#include <sys/zfs_vfsops.h>
	#endif

	/* Set this tunable to TRUE to replace corrupt data with 0x2f5baddb10c */
	static int zfs_send_corrupt_data = B_FALSE;
	/*
	* This tunable controls the amount of data (measured in bytes) that will be
	* prefetched by zfs send. If the main thread is blocking on reads that haven't
	* completed, this variable might need to be increased. If instead the main
	* thread is issuing new reads because the prefetches have fallen out of the
	* cache, this may need to be decreased.
	*/
	static uint_t zfs_send_queue_length = SPA_MAXBLOCKSIZE;
	/*
	* This tunable controls the length of the queues that zfs send worker threads
	* use to communicate. If the send_main_thread is blocking on these queues,
	* this variable may need to be increased. If there is a significant slowdown
	* at the start of a send as these threads consume all the available IO
	* resources, this variable may need to be decreased.
	*/
	static uint_t zfs_send_no_prefetch_queue_length = 1024 * 1024;
	/*
	* These tunables control the fill fraction of the queues by zfs send. The fill
	* fraction controls the frequency with which threads have to be cv_signaled.
	* If a lot of cpu time is being spent on cv_signal, then these should be tuned
	* down. If the queues empty before the signalled thread can catch up, then
	* these should be tuned up.
	*/
	static uint_t zfs_send_queue_ff = 20;
	static uint_t zfs_send_no_prefetch_queue_ff = 20;

	/*
	* Use this to override the recordsize calculation for fast zfs send estimates.
	*/
	static uint_t zfs_override_estimate_recordsize = 0;

	/* Set this tunable to FALSE to disable setting of DRR_FLAG_FREERECORDS */
	static const boolean_t zfs_send_set_freerecords_bit = B_TRUE;

	/* Set this tunable to FALSE is disable sending unmodified spill blocks. */
	static int zfs_send_unmodified_spill_blocks = B_TRUE;

	static inline boolean_t
	overflow_multiply(uint64_t a, uint64_t b, uint64_t *c)
	{
	uint64_t temp = a * b;
	if (b != 0 && temp / b != a)
	return (B_FALSE);
	*c = temp;
	return (B_TRUE);
	}

	struct send_thread_arg {
	bqueue_t q;
	objset_t os; / Objset to traverse */
	uint64_t fromtxg; /* Traverse from this txg */
	int flags; /* flags to pass to traverse_dataset */
	int error_code;
	boolean_t cancel;
	zbookmark_phys_t resume;
	uint64_t *num_blocks_visited;
	};

	struct redact_list_thread_arg {
	boolean_t cancel;
	bqueue_t q;
	zbookmark_phys_t resume;
	redaction_list_t *rl;
	boolean_t mark_redact;
	int error_code;
	uint64_t *num_blocks_visited;
	};

	struct send_merge_thread_arg {
	bqueue_t q;
	objset_t *os;
	struct redact_list_thread_arg *from_arg;
	struct send_thread_arg *to_arg;
	struct redact_list_thread_arg *redact_arg;
	int error;
	boolean_t cancel;
	};

	struct send_range {
	boolean_t eos_marker; /* Marks the end of the stream */
	uint64_t object;
	uint64_t start_blkid;
	uint64_t end_blkid;
	bqueue_node_t ln;
	enum type {DATA, HOLE, OBJECT, OBJECT_RANGE, REDACT,
	PREVIOUSLY_REDACTED} type;
	union {
	struct srd {
	dmu_object_type_t obj_type;
	uint32_t datablksz; // logical size
	uint32_t datasz; // payload size
	blkptr_t bp;
	arc_buf_t *abuf;
	abd_t *abd;
	kmutex_t lock;
	kcondvar_t cv;
	boolean_t io_outstanding;
	boolean_t io_compressed;
	int io_err;
	} data;
	struct srh {
	uint32_t datablksz;
	} hole;
	struct sro {
	/*
	* This is a pointer because embedding it in the
	* struct causes these structures to be massively larger
	* for all range types; this makes the code much less
	* memory efficient.
	*/
	dnode_phys_t *dnp;
	blkptr_t bp;
	} object;
	struct srr {
	uint32_t datablksz;
	} redact;
	struct sror {
	blkptr_t bp;
	} object_range;
	} sru;
	};

	/*
	* The list of data whose inclusion in a send stream can be pending from
	* one call to backup_cb to another. Multiple calls to dump_free(),
	* dump_freeobjects(), and dump_redact() can be aggregated into a single
	* DRR_FREE, DRR_FREEOBJECTS, or DRR_REDACT replay record.
	*/
	typedef enum {
	PENDING_NONE,
	PENDING_FREE,
	PENDING_FREEOBJECTS,
	PENDING_REDACT
	} dmu_pendop_t;

	typedef struct dmu_send_cookie {
	dmu_replay_record_t *dsc_drr;
	dmu_send_outparams_t *dsc_dso;
	offset_t *dsc_off;
	objset_t *dsc_os;
	zio_cksum_t dsc_zc;
	uint64_t dsc_toguid;
	uint64_t dsc_fromtxg;
	int dsc_err;
	dmu_pendop_t dsc_pending_op;
	uint64_t dsc_featureflags;
	uint64_t dsc_last_data_object;
	uint64_t dsc_last_data_offset;
	uint64_t dsc_resume_object;
	uint64_t dsc_resume_offset;
	boolean_t dsc_sent_begin;
	boolean_t dsc_sent_end;
	} dmu_send_cookie_t;

	static int do_dump(dmu_send_cookie_t dscp, struct send_range range);

	static void
	range_free(struct send_range *range)
	{
	if (range->type == OBJECT) {
	size_t size = sizeof (dnode_phys_t) *
	(range->sru.object.dnp->dn_extra_slots + 1);
	kmem_free(range->sru.object.dnp, size);
	} else if (range->type == DATA) {
	mutex_enter(&range->sru.data.lock);
	while (range->sru.data.io_outstanding)
	cv_wait(&range->sru.data.cv, &range->sru.data.lock);
	if (range->sru.data.abd != NULL)
	abd_free(range->sru.data.abd);
	if (range->sru.data.abuf != NULL) {
	arc_buf_destroy(range->sru.data.abuf,
	&range->sru.data.abuf);
	}
	mutex_exit(&range->sru.data.lock);

	cv_destroy(&range->sru.data.cv);
	mutex_destroy(&range->sru.data.lock);
	}
	kmem_free(range, sizeof (*range));
	}

	/*
	* For all record types except BEGIN, fill in the checksum (overlaid in
	* drr_u.drr_checksum.drr_checksum). The checksum verifies everything
	* up to the start of the checksum itself.
	*/
	static int
	dump_record(dmu_send_cookie_t dscp, void payload, int payload_len)
	{
	dmu_send_outparams_t *dso = dscp->dsc_dso;
	ASSERT3U(offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum),
	==, sizeof (dmu_replay_record_t) - sizeof (zio_cksum_t));
	(void) fletcher_4_incremental_native(dscp->dsc_drr,
	offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum),
	&dscp->dsc_zc);
	if (dscp->dsc_drr->drr_type == DRR_BEGIN) {
	dscp->dsc_sent_begin = B_TRUE;
	} else {
	ASSERT(ZIO_CHECKSUM_IS_ZERO(&dscp->dsc_drr->drr_u.
	drr_checksum.drr_checksum));
	dscp->dsc_drr->drr_u.drr_checksum.drr_checksum = dscp->dsc_zc;
	}
	if (dscp->dsc_drr->drr_type == DRR_END) {
	dscp->dsc_sent_end = B_TRUE;
	}
	(void) fletcher_4_incremental_native(&dscp->dsc_drr->
	drr_u.drr_checksum.drr_checksum,
	sizeof (zio_cksum_t), &dscp->dsc_zc);
	*dscp->dsc_off += sizeof (dmu_replay_record_t);
	dscp->dsc_err = dso->dso_outfunc(dscp->dsc_os, dscp->dsc_drr,
	sizeof (dmu_replay_record_t), dso->dso_arg);
	if (dscp->dsc_err != 0)
	return (SET_ERROR(EINTR));
	if (payload_len != 0) {
	*dscp->dsc_off += payload_len;
	/*
	* payload is null when dso_dryrun == B_TRUE (i.e. when we're
	* doing a send size calculation)
	*/
	if (payload != NULL) {
	(void) fletcher_4_incremental_native(
	payload, payload_len, &dscp->dsc_zc);
	}

	/*
	* The code does not rely on this (len being a multiple of 8).
	* We keep this assertion because of the corresponding assertion
	* in receive_read(). Keeping this assertion ensures that we do
	* not inadvertently break backwards compatibility (causing the
	* assertion in receive_read() to trigger on old software).
	*
	* Raw sends cannot be received on old software, and so can
	* bypass this assertion.
	*/

	ASSERT((payload_len % 8 == 0) \|\|
	(dscp->dsc_featureflags & DMU_BACKUP_FEATURE_RAW));

	dscp->dsc_err = dso->dso_outfunc(dscp->dsc_os, payload,
	payload_len, dso->dso_arg);
	if (dscp->dsc_err != 0)
	return (SET_ERROR(EINTR));
	}
	return (0);
	}

	/*
	* Fill in the drr_free struct, or perform aggregation if the previous record is
	* also a free record, and the two are adjacent.
	*
	* Note that we send free records even for a full send, because we want to be
	* able to receive a full send as a clone, which requires a list of all the free
	* and freeobject records that were generated on the source.
	*/
	static int
	dump_free(dmu_send_cookie_t *dscp, uint64_t object, uint64_t offset,
	uint64_t length)
	{
	struct drr_free *drrf = &(dscp->dsc_drr->drr_u.drr_free);

	/*
	* When we receive a free record, dbuf_free_range() assumes
	* that the receiving system doesn't have any dbufs in the range
	* being freed. This is always true because there is a one-record
	* constraint: we only send one WRITE record for any given
	* object,offset. We know that the one-record constraint is
	* true because we always send data in increasing order by
	* object,offset.
	*
	* If the increasing-order constraint ever changes, we should find
	* another way to assert that the one-record constraint is still
	* satisfied.
	*/
	ASSERT(object > dscp->dsc_last_data_object \|\|
	(object == dscp->dsc_last_data_object &&
	offset > dscp->dsc_last_data_offset));

	/*
	* If there is a pending op, but it's not PENDING_FREE, push it out,
	* since free block aggregation can only be done for blocks of the
	* same type (i.e., DRR_FREE records can only be aggregated with
	* other DRR_FREE records. DRR_FREEOBJECTS records can only be
	* aggregated with other DRR_FREEOBJECTS records).
	*/
	if (dscp->dsc_pending_op != PENDING_NONE &&
	dscp->dsc_pending_op != PENDING_FREE) {
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	dscp->dsc_pending_op = PENDING_NONE;
	}

	if (dscp->dsc_pending_op == PENDING_FREE) {
	/*
	* Check to see whether this free block can be aggregated
	* with pending one.
	*/
	if (drrf->drr_object == object && drrf->drr_offset +
	drrf->drr_length == offset) {
	if (offset + length < offset \|\| length == UINT64_MAX)
	drrf->drr_length = UINT64_MAX;
	else
	drrf->drr_length += length;
	return (0);
	} else {
	/* not a continuation. Push out pending record */
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	dscp->dsc_pending_op = PENDING_NONE;
	}
	}
	/* create a FREE record and make it pending */
	memset(dscp->dsc_drr, 0, sizeof (dmu_replay_record_t));
	dscp->dsc_drr->drr_type = DRR_FREE;
	drrf->drr_object = object;
	drrf->drr_offset = offset;
	if (offset + length < offset)
	drrf->drr_length = DMU_OBJECT_END;
	else
	drrf->drr_length = length;
	drrf->drr_toguid = dscp->dsc_toguid;
	if (length == DMU_OBJECT_END) {
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	} else {
	dscp->dsc_pending_op = PENDING_FREE;
	}

	return (0);
	}

	/*
	* Fill in the drr_redact struct, or perform aggregation if the previous record
	* is also a redaction record, and the two are adjacent.
	*/
	static int
	dump_redact(dmu_send_cookie_t *dscp, uint64_t object, uint64_t offset,
	uint64_t length)
	{
	struct drr_redact *drrr = &dscp->dsc_drr->drr_u.drr_redact;

	/*
	* If there is a pending op, but it's not PENDING_REDACT, push it out,
	* since free block aggregation can only be done for blocks of the
	* same type (i.e., DRR_REDACT records can only be aggregated with
	* other DRR_REDACT records).
	*/
	if (dscp->dsc_pending_op != PENDING_NONE &&
	dscp->dsc_pending_op != PENDING_REDACT) {
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	dscp->dsc_pending_op = PENDING_NONE;
	}

	if (dscp->dsc_pending_op == PENDING_REDACT) {
	/*
	* Check to see whether this redacted block can be aggregated
	* with pending one.
	*/
	if (drrr->drr_object == object && drrr->drr_offset +
	drrr->drr_length == offset) {
	drrr->drr_length += length;
	return (0);
	} else {
	/* not a continuation. Push out pending record */
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	dscp->dsc_pending_op = PENDING_NONE;
	}
	}
	/* create a REDACT record and make it pending */
	memset(dscp->dsc_drr, 0, sizeof (dmu_replay_record_t));
	dscp->dsc_drr->drr_type = DRR_REDACT;
	drrr->drr_object = object;
	drrr->drr_offset = offset;
	drrr->drr_length = length;
	drrr->drr_toguid = dscp->dsc_toguid;
	dscp->dsc_pending_op = PENDING_REDACT;

	return (0);
	}

	static int
	dmu_dump_write(dmu_send_cookie_t *dscp, dmu_object_type_t type, uint64_t object,
	uint64_t offset, int lsize, int psize, const blkptr_t *bp,
	boolean_t io_compressed, void *data)
	{
	uint64_t payload_size;
	boolean_t raw = (dscp->dsc_featureflags & DMU_BACKUP_FEATURE_RAW);
	struct drr_write *drrw = &(dscp->dsc_drr->drr_u.drr_write);

	/*
	* We send data in increasing object, offset order.
	* See comment in dump_free() for details.
	*/
	ASSERT(object > dscp->dsc_last_data_object \|\|
	(object == dscp->dsc_last_data_object &&
	offset > dscp->dsc_last_data_offset));
	dscp->dsc_last_data_object = object;
	dscp->dsc_last_data_offset = offset + lsize - 1;

	/*
	* If there is any kind of pending aggregation (currently either
	* a grouping of free objects or free blocks), push it out to
	* the stream, since aggregation can't be done across operations
	* of different types.
	*/
	if (dscp->dsc_pending_op != PENDING_NONE) {
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	dscp->dsc_pending_op = PENDING_NONE;
	}
	/* write a WRITE record */
	memset(dscp->dsc_drr, 0, sizeof (dmu_replay_record_t));
	dscp->dsc_drr->drr_type = DRR_WRITE;
	drrw->drr_object = object;
	drrw->drr_type = type;
	drrw->drr_offset = offset;
	drrw->drr_toguid = dscp->dsc_toguid;
	drrw->drr_logical_size = lsize;

	/* only set the compression fields if the buf is compressed or raw */
	boolean_t compressed =
	(bp != NULL ? BP_GET_COMPRESS(bp) != ZIO_COMPRESS_OFF &&
	io_compressed : lsize != psize);
	if (raw \|\| compressed) {
	ASSERT(bp != NULL);
	ASSERT(raw \|\| dscp->dsc_featureflags &
	DMU_BACKUP_FEATURE_COMPRESSED);
	ASSERT(!BP_IS_EMBEDDED(bp));
	ASSERT3S(psize, >, 0);

	if (raw) {
	ASSERT(BP_IS_PROTECTED(bp));

	/*
	* This is a raw protected block so we need to pass
	* along everything the receiving side will need to
	* interpret this block, including the byteswap, salt,
	* IV, and MAC.
	*/
	if (BP_SHOULD_BYTESWAP(bp))
	drrw->drr_flags \|= DRR_RAW_BYTESWAP;
	zio_crypt_decode_params_bp(bp, drrw->drr_salt,
	drrw->drr_iv);
	zio_crypt_decode_mac_bp(bp, drrw->drr_mac);
	} else {
	/* this is a compressed block */
	ASSERT(dscp->dsc_featureflags &
	DMU_BACKUP_FEATURE_COMPRESSED);
	ASSERT(!BP_SHOULD_BYTESWAP(bp));
	ASSERT(!DMU_OT_IS_METADATA(BP_GET_TYPE(bp)));
	ASSERT3U(BP_GET_COMPRESS(bp), !=, ZIO_COMPRESS_OFF);
	ASSERT3S(lsize, >=, psize);
	}

	/* set fields common to compressed and raw sends */
	drrw->drr_compressiontype = BP_GET_COMPRESS(bp);
	drrw->drr_compressed_size = psize;
	payload_size = drrw->drr_compressed_size;
	} else {
	payload_size = drrw->drr_logical_size;
	}

	if (bp == NULL \|\| BP_IS_EMBEDDED(bp) \|\| (BP_IS_PROTECTED(bp) && !raw)) {
	/*
	* There's no pre-computed checksum for partial-block writes,
	* embedded BP's, or encrypted BP's that are being sent as
	* plaintext, so (like fletcher4-checksummed blocks) userland
	* will have to compute a dedup-capable checksum itself.
	*/
	drrw->drr_checksumtype = ZIO_CHECKSUM_OFF;
	} else {
	drrw->drr_checksumtype = BP_GET_CHECKSUM(bp);
	if (zio_checksum_table[drrw->drr_checksumtype].ci_flags &
	ZCHECKSUM_FLAG_DEDUP)
	drrw->drr_flags \|= DRR_CHECKSUM_DEDUP;
	DDK_SET_LSIZE(&drrw->drr_key, BP_GET_LSIZE(bp));
	DDK_SET_PSIZE(&drrw->drr_key, BP_GET_PSIZE(bp));
	DDK_SET_COMPRESS(&drrw->drr_key, BP_GET_COMPRESS(bp));
	DDK_SET_CRYPT(&drrw->drr_key, BP_IS_PROTECTED(bp));
	drrw->drr_key.ddk_cksum = bp->blk_cksum;
	}

	if (dump_record(dscp, data, payload_size) != 0)
	return (SET_ERROR(EINTR));
	return (0);
	}

	static int
	dump_write_embedded(dmu_send_cookie_t *dscp, uint64_t object, uint64_t offset,
	int blksz, const blkptr_t *bp)
	{
	char buf[BPE_PAYLOAD_SIZE];
	struct drr_write_embedded *drrw =
	&(dscp->dsc_drr->drr_u.drr_write_embedded);

	if (dscp->dsc_pending_op != PENDING_NONE) {
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	dscp->dsc_pending_op = PENDING_NONE;
	}

	ASSERT(BP_IS_EMBEDDED(bp));

	memset(dscp->dsc_drr, 0, sizeof (dmu_replay_record_t));
	dscp->dsc_drr->drr_type = DRR_WRITE_EMBEDDED;
	drrw->drr_object = object;
	drrw->drr_offset = offset;
	drrw->drr_length = blksz;
	drrw->drr_toguid = dscp->dsc_toguid;
	drrw->drr_compression = BP_GET_COMPRESS(bp);
	drrw->drr_etype = BPE_GET_ETYPE(bp);
	drrw->drr_lsize = BPE_GET_LSIZE(bp);
	drrw->drr_psize = BPE_GET_PSIZE(bp);

	decode_embedded_bp_compressed(bp, buf);

	uint32_t psize = drrw->drr_psize;
	uint32_t rsize = P2ROUNDUP(psize, 8);

	if (psize != rsize)
	memset(buf + psize, 0, rsize - psize);

	if (dump_record(dscp, buf, rsize) != 0)
	return (SET_ERROR(EINTR));
	return (0);
	}

	static int
	dump_spill(dmu_send_cookie_t dscp, const blkptr_t bp, uint64_t object,
	void *data)
	{
	struct drr_spill *drrs = &(dscp->dsc_drr->drr_u.drr_spill);
	uint64_t blksz = BP_GET_LSIZE(bp);
	uint64_t payload_size = blksz;

	if (dscp->dsc_pending_op != PENDING_NONE) {
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	dscp->dsc_pending_op = PENDING_NONE;
	}

	/* write a SPILL record */
	memset(dscp->dsc_drr, 0, sizeof (dmu_replay_record_t));
	dscp->dsc_drr->drr_type = DRR_SPILL;
	drrs->drr_object = object;
	drrs->drr_length = blksz;
	drrs->drr_toguid = dscp->dsc_toguid;

	/* See comment in dump_dnode() for full details */
	if (zfs_send_unmodified_spill_blocks &&
	(BP_GET_LOGICAL_BIRTH(bp) <= dscp->dsc_fromtxg)) {
	drrs->drr_flags \|= DRR_SPILL_UNMODIFIED;
	}

	/* handle raw send fields */
	if (dscp->dsc_featureflags & DMU_BACKUP_FEATURE_RAW) {
	ASSERT(BP_IS_PROTECTED(bp));

	if (BP_SHOULD_BYTESWAP(bp))
	drrs->drr_flags \|= DRR_RAW_BYTESWAP;
	drrs->drr_compressiontype = BP_GET_COMPRESS(bp);
	drrs->drr_compressed_size = BP_GET_PSIZE(bp);
	zio_crypt_decode_params_bp(bp, drrs->drr_salt, drrs->drr_iv);
	zio_crypt_decode_mac_bp(bp, drrs->drr_mac);
	payload_size = drrs->drr_compressed_size;
	}

	if (dump_record(dscp, data, payload_size) != 0)
	return (SET_ERROR(EINTR));
	return (0);
	}

	static int
	dump_freeobjects(dmu_send_cookie_t *dscp, uint64_t firstobj, uint64_t numobjs)
	{
	struct drr_freeobjects *drrfo = &(dscp->dsc_drr->drr_u.drr_freeobjects);
	uint64_t maxobj = DNODES_PER_BLOCK *
	(DMU_META_DNODE(dscp->dsc_os)->dn_maxblkid + 1);

	/*
	* ZoL < 0.7 does not handle large FREEOBJECTS records correctly,
	* leading to zfs recv never completing. to avoid this issue, don't
	* send FREEOBJECTS records for object IDs which cannot exist on the
	* receiving side.
	*/
	if (maxobj > 0) {
	if (maxobj <= firstobj)
	return (0);

	if (maxobj < firstobj + numobjs)
	numobjs = maxobj - firstobj;
	}

	/*
	* If there is a pending op, but it's not PENDING_FREEOBJECTS,
	* push it out, since free block aggregation can only be done for
	* blocks of the same type (i.e., DRR_FREE records can only be
	* aggregated with other DRR_FREE records. DRR_FREEOBJECTS records
	* can only be aggregated with other DRR_FREEOBJECTS records).
	*/
	if (dscp->dsc_pending_op != PENDING_NONE &&
	dscp->dsc_pending_op != PENDING_FREEOBJECTS) {
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	dscp->dsc_pending_op = PENDING_NONE;
	}

	if (dscp->dsc_pending_op == PENDING_FREEOBJECTS) {
	/*
	* See whether this free object array can be aggregated
	* with pending one
	*/
	if (drrfo->drr_firstobj + drrfo->drr_numobjs == firstobj) {
	drrfo->drr_numobjs += numobjs;
	return (0);
	} else {
	/* can't be aggregated. Push out pending record */
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	dscp->dsc_pending_op = PENDING_NONE;
	}
	}

	/* write a FREEOBJECTS record */
	memset(dscp->dsc_drr, 0, sizeof (dmu_replay_record_t));
	dscp->dsc_drr->drr_type = DRR_FREEOBJECTS;
	drrfo->drr_firstobj = firstobj;
	drrfo->drr_numobjs = numobjs;
	drrfo->drr_toguid = dscp->dsc_toguid;

	dscp->dsc_pending_op = PENDING_FREEOBJECTS;

	return (0);
	}

	static int
	dump_dnode(dmu_send_cookie_t dscp, const blkptr_t bp, uint64_t object,
	dnode_phys_t *dnp)
	{
	struct drr_object *drro = &(dscp->dsc_drr->drr_u.drr_object);
	int bonuslen;

	if (object < dscp->dsc_resume_object) {
	/*
	* Note: when resuming, we will visit all the dnodes in
	* the block of dnodes that we are resuming from. In
	* this case it's unnecessary to send the dnodes prior to
	* the one we are resuming from. We should be at most one
	* block's worth of dnodes behind the resume point.
	*/
	ASSERT3U(dscp->dsc_resume_object - object, <,
	1 << (DNODE_BLOCK_SHIFT - DNODE_SHIFT));
	return (0);
	}

	if (dnp == NULL \|\| dnp->dn_type == DMU_OT_NONE)
	return (dump_freeobjects(dscp, object, 1));

	if (dscp->dsc_pending_op != PENDING_NONE) {
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	dscp->dsc_pending_op = PENDING_NONE;
	}

	/* write an OBJECT record */
	memset(dscp->dsc_drr, 0, sizeof (dmu_replay_record_t));
	dscp->dsc_drr->drr_type = DRR_OBJECT;
	drro->drr_object = object;
	drro->drr_type = dnp->dn_type;
	drro->drr_bonustype = dnp->dn_bonustype;
	drro->drr_blksz = dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT;
	drro->drr_bonuslen = dnp->dn_bonuslen;
	drro->drr_dn_slots = dnp->dn_extra_slots + 1;
	drro->drr_checksumtype = dnp->dn_checksum;
	drro->drr_compress = dnp->dn_compress;
	drro->drr_toguid = dscp->dsc_toguid;

	if (!(dscp->dsc_featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS) &&
	drro->drr_blksz > SPA_OLD_MAXBLOCKSIZE)
	drro->drr_blksz = SPA_OLD_MAXBLOCKSIZE;

	bonuslen = P2ROUNDUP(dnp->dn_bonuslen, 8);

	if ((dscp->dsc_featureflags & DMU_BACKUP_FEATURE_RAW)) {
	ASSERT(BP_IS_ENCRYPTED(bp));

	if (BP_SHOULD_BYTESWAP(bp))
	drro->drr_flags \|= DRR_RAW_BYTESWAP;

	/* needed for reconstructing dnp on recv side */
	drro->drr_maxblkid = dnp->dn_maxblkid;
	drro->drr_indblkshift = dnp->dn_indblkshift;
	drro->drr_nlevels = dnp->dn_nlevels;
	drro->drr_nblkptr = dnp->dn_nblkptr;

	/*
	* Since we encrypt the entire bonus area, the (raw) part
	* beyond the bonuslen is actually nonzero, so we need
	* to send it.
	*/
	if (bonuslen != 0) {
	if (drro->drr_bonuslen > DN_MAX_BONUS_LEN(dnp))
	return (SET_ERROR(EINVAL));
	drro->drr_raw_bonuslen = DN_MAX_BONUS_LEN(dnp);
	bonuslen = drro->drr_raw_bonuslen;
	}
	}

	/*
	* DRR_OBJECT_SPILL is set for every dnode which references a
	* spill block. This allows the receiving pool to definitively
	* determine when a spill block should be kept or freed.
	*/
	if (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR)
	drro->drr_flags \|= DRR_OBJECT_SPILL;

	if (dump_record(dscp, DN_BONUS(dnp), bonuslen) != 0)
	return (SET_ERROR(EINTR));

	/* Free anything past the end of the file. */
	if (dump_free(dscp, object, (dnp->dn_maxblkid + 1) *
	(dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT), DMU_OBJECT_END) != 0)
	return (SET_ERROR(EINTR));

	/*
	* Send DRR_SPILL records for unmodified spill blocks. This is useful
	* because changing certain attributes of the object (e.g. blocksize)
	* can cause old versions of ZFS to incorrectly remove a spill block.
	* Including these records in the stream forces an up to date version
	* to always be written ensuring they're never lost. Current versions
	* of the code which understand the DRR_FLAG_SPILL_BLOCK feature can
	* ignore these unmodified spill blocks.
	*/
	if (zfs_send_unmodified_spill_blocks &&
	(dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) &&
	(BP_GET_LOGICAL_BIRTH(DN_SPILL_BLKPTR(dnp)) <= dscp->dsc_fromtxg)) {
	struct send_range record;
	blkptr_t *bp = DN_SPILL_BLKPTR(dnp);

	memset(&record, 0, sizeof (struct send_range));
	record.type = DATA;
	record.object = object;
	record.eos_marker = B_FALSE;
	record.start_blkid = DMU_SPILL_BLKID;
	record.end_blkid = record.start_blkid + 1;
	record.sru.data.bp = *bp;
	record.sru.data.obj_type = dnp->dn_type;
	record.sru.data.datablksz = BP_GET_LSIZE(bp);

	if (do_dump(dscp, &record) != 0)
	return (SET_ERROR(EINTR));
	}

	if (dscp->dsc_err != 0)
	return (SET_ERROR(EINTR));

	return (0);
	}

	static int
	dump_object_range(dmu_send_cookie_t dscp, const blkptr_t bp,
	uint64_t firstobj, uint64_t numslots)
	{
	struct drr_object_range *drror =
	&(dscp->dsc_drr->drr_u.drr_object_range);

	/* we only use this record type for raw sends */
	ASSERT(BP_IS_PROTECTED(bp));
	ASSERT(dscp->dsc_featureflags & DMU_BACKUP_FEATURE_RAW);
	ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF);
	ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_DNODE);
	ASSERT0(BP_GET_LEVEL(bp));

	if (dscp->dsc_pending_op != PENDING_NONE) {
	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	dscp->dsc_pending_op = PENDING_NONE;
	}

	memset(dscp->dsc_drr, 0, sizeof (dmu_replay_record_t));
	dscp->dsc_drr->drr_type = DRR_OBJECT_RANGE;
	drror->drr_firstobj = firstobj;
	drror->drr_numslots = numslots;
	drror->drr_toguid = dscp->dsc_toguid;
	if (BP_SHOULD_BYTESWAP(bp))
	drror->drr_flags \|= DRR_RAW_BYTESWAP;
	zio_crypt_decode_params_bp(bp, drror->drr_salt, drror->drr_iv);
	zio_crypt_decode_mac_bp(bp, drror->drr_mac);

	if (dump_record(dscp, NULL, 0) != 0)
	return (SET_ERROR(EINTR));
	return (0);
	}

	static boolean_t
	send_do_embed(const blkptr_t *bp, uint64_t featureflags)
	{
	if (!BP_IS_EMBEDDED(bp))
	return (B_FALSE);

	/*
	* Compression function must be legacy, or explicitly enabled.
	*/
	if ((BP_GET_COMPRESS(bp) >= ZIO_COMPRESS_LEGACY_FUNCTIONS &&
	!(featureflags & DMU_BACKUP_FEATURE_LZ4)))
	return (B_FALSE);

	/*
	* If we have not set the ZSTD feature flag, we can't send ZSTD
	* compressed embedded blocks, as the receiver may not support them.
	*/
	if ((BP_GET_COMPRESS(bp) == ZIO_COMPRESS_ZSTD &&
	!(featureflags & DMU_BACKUP_FEATURE_ZSTD)))
	return (B_FALSE);

	/*
	* Embed type must be explicitly enabled.
	*/
	switch (BPE_GET_ETYPE(bp)) {
	case BP_EMBEDDED_TYPE_DATA:
	if (featureflags & DMU_BACKUP_FEATURE_EMBED_DATA)
	return (B_TRUE);
	break;
	default:
	return (B_FALSE);
	}
	return (B_FALSE);
	}

	/*
	* This function actually handles figuring out what kind of record needs to be
	* dumped, and calling the appropriate helper function. In most cases,
	* the data has already been read by send_reader_thread().
	*/
	static int
	do_dump(dmu_send_cookie_t dscp, struct send_range range)
	{
	int err = 0;
	switch (range->type) {
	case OBJECT:
	err = dump_dnode(dscp, &range->sru.object.bp, range->object,
	range->sru.object.dnp);
	return (err);
	case OBJECT_RANGE: {
	ASSERT3U(range->start_blkid + 1, ==, range->end_blkid);
	if (!(dscp->dsc_featureflags & DMU_BACKUP_FEATURE_RAW)) {
	return (0);
	}
	uint64_t epb = BP_GET_LSIZE(&range->sru.object_range.bp) >>
	DNODE_SHIFT;
	uint64_t firstobj = range->start_blkid * epb;
	err = dump_object_range(dscp, &range->sru.object_range.bp,
	firstobj, epb);
	break;
	}
	case REDACT: {
	struct srr *srrp = &range->sru.redact;
	err = dump_redact(dscp, range->object, range->start_blkid *
	srrp->datablksz, (range->end_blkid - range->start_blkid) *
	srrp->datablksz);
	return (err);
	}
	case DATA: {
	struct srd *srdp = &range->sru.data;
	blkptr_t *bp = &srdp->bp;
	spa_t *spa =
	dmu_objset_spa(dscp->dsc_os);

	ASSERT3U(srdp->datablksz, ==, BP_GET_LSIZE(bp));
	ASSERT3U(range->start_blkid + 1, ==, range->end_blkid);
	if (BP_GET_TYPE(bp) == DMU_OT_SA) {
	arc_flags_t aflags = ARC_FLAG_WAIT;
	zio_flag_t zioflags = ZIO_FLAG_CANFAIL;

	if (dscp->dsc_featureflags & DMU_BACKUP_FEATURE_RAW) {
	ASSERT(BP_IS_PROTECTED(bp));
	zioflags \|= ZIO_FLAG_RAW;
	}

	zbookmark_phys_t zb;
	ASSERT3U(range->start_blkid, ==, DMU_SPILL_BLKID);
	zb.zb_objset = dmu_objset_id(dscp->dsc_os);
	zb.zb_object = range->object;
	zb.zb_level = 0;
	zb.zb_blkid = range->start_blkid;

	arc_buf_t *abuf = NULL;
	if (!dscp->dsc_dso->dso_dryrun && arc_read(NULL, spa,
	bp, arc_getbuf_func, &abuf, ZIO_PRIORITY_ASYNC_READ,
	zioflags, &aflags, &zb) != 0)
	return (SET_ERROR(EIO));

	err = dump_spill(dscp, bp, zb.zb_object,
	(abuf == NULL ? NULL : abuf->b_data));
	if (abuf != NULL)
	arc_buf_destroy(abuf, &abuf);
	return (err);
	}
	if (send_do_embed(bp, dscp->dsc_featureflags)) {
	err = dump_write_embedded(dscp, range->object,
	range->start_blkid * srdp->datablksz,
	srdp->datablksz, bp);
	return (err);
	}
	ASSERT(range->object > dscp->dsc_resume_object \|\|
	(range->object == dscp->dsc_resume_object &&
	range->start_blkid * srdp->datablksz >=
	dscp->dsc_resume_offset));
	/* it's a level-0 block of a regular object */

	mutex_enter(&srdp->lock);
	while (srdp->io_outstanding)
	cv_wait(&srdp->cv, &srdp->lock);
	err = srdp->io_err;
	mutex_exit(&srdp->lock);

	if (err != 0) {
	if (zfs_send_corrupt_data &&
	!dscp->dsc_dso->dso_dryrun) {
	/*
	* Send a block filled with 0x"zfs badd bloc"
	*/
	srdp->abuf = arc_alloc_buf(spa, &srdp->abuf,
	ARC_BUFC_DATA, srdp->datablksz);
	uint64_t *ptr;
	for (ptr = srdp->abuf->b_data;
	(char )ptr < (char )srdp->abuf->b_data +
	srdp->datablksz; ptr++)
	*ptr = 0x2f5baddb10cULL;
	} else {
	return (SET_ERROR(EIO));
	}
	}

	ASSERT(dscp->dsc_dso->dso_dryrun \|\|
	srdp->abuf != NULL \|\| srdp->abd != NULL);

	uint64_t offset = range->start_blkid * srdp->datablksz;

	char *data = NULL;
	if (srdp->abd != NULL) {
	data = abd_to_buf(srdp->abd);
	ASSERT3P(srdp->abuf, ==, NULL);
	} else if (srdp->abuf != NULL) {
	data = srdp->abuf->b_data;
	}

	/*
	* If we have large blocks stored on disk but the send flags
	* don't allow us to send large blocks, we split the data from
	* the arc buf into chunks.
	*/
	if (srdp->datablksz > SPA_OLD_MAXBLOCKSIZE &&
	!(dscp->dsc_featureflags &
	DMU_BACKUP_FEATURE_LARGE_BLOCKS)) {
	while (srdp->datablksz > 0 && err == 0) {
	int n = MIN(srdp->datablksz,
	SPA_OLD_MAXBLOCKSIZE);
	err = dmu_dump_write(dscp, srdp->obj_type,
	range->object, offset, n, n, NULL, B_FALSE,
	data);
	offset += n;
	/*
	* When doing dry run, data==NULL is used as a
	* sentinel value by
	* dmu_dump_write()->dump_record().
	*/
	if (data != NULL)
	data += n;
	srdp->datablksz -= n;
	}
	} else {
	err = dmu_dump_write(dscp, srdp->obj_type,
	range->object, offset,
	srdp->datablksz, srdp->datasz, bp,
	srdp->io_compressed, data);
	}
	return (err);
	}
	case HOLE: {
	struct srh *srhp = &range->sru.hole;
	if (range->object == DMU_META_DNODE_OBJECT) {
	uint32_t span = srhp->datablksz >> DNODE_SHIFT;
	uint64_t first_obj = range->start_blkid * span;
	uint64_t numobj = range->end_blkid * span - first_obj;
	return (dump_freeobjects(dscp, first_obj, numobj));
	}
	uint64_t offset = 0;

	/*
	* If this multiply overflows, we don't need to send this block.
	* Even if it has a birth time, it can never not be a hole, so
	* we don't need to send records for it.
	*/
	if (!overflow_multiply(range->start_blkid, srhp->datablksz,
	&offset)) {
	return (0);
	}
	uint64_t len = 0;

	if (!overflow_multiply(range->end_blkid, srhp->datablksz, &len))
	len = UINT64_MAX;
	len = len - offset;
	return (dump_free(dscp, range->object, offset, len));
	}
	default:
	panic("Invalid range type in do_dump: %d", range->type);
	}
	return (err);
	}

	static struct send_range *
	range_alloc(enum type type, uint64_t object, uint64_t start_blkid,
	uint64_t end_blkid, boolean_t eos)
	{
	struct send_range range = kmem_alloc(sizeof (range), KM_SLEEP);
	range->type = type;
	range->object = object;
	range->start_blkid = start_blkid;
	range->end_blkid = end_blkid;
	range->eos_marker = eos;
	if (type == DATA) {
	range->sru.data.abd = NULL;
	range->sru.data.abuf = NULL;
	mutex_init(&range->sru.data.lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&range->sru.data.cv, NULL, CV_DEFAULT, NULL);
	range->sru.data.io_outstanding = 0;
	range->sru.data.io_err = 0;
	range->sru.data.io_compressed = B_FALSE;
	}
	return (range);
	}

	/*
	* This is the callback function to traverse_dataset that acts as a worker
	* thread for dmu_send_impl.
	*/
	static int
	send_cb(spa_t spa, zilog_t zilog, const blkptr_t *bp,
	const zbookmark_phys_t zb, const struct dnode_phys dnp, void *arg)
	{
	(void) zilog;
	struct send_thread_arg *sta = arg;
	struct send_range *record;

	ASSERT(zb->zb_object == DMU_META_DNODE_OBJECT \|\|
	zb->zb_object >= sta->resume.zb_object);

	/*
	* All bps of an encrypted os should have the encryption bit set.
	* If this is not true it indicates tampering and we report an error.
	*/
	if (sta->os->os_encrypted &&
	!BP_IS_HOLE(bp) && !BP_USES_CRYPT(bp)) {
	spa_log_error(spa, zb, BP_GET_LOGICAL_BIRTH(bp));
	return (SET_ERROR(EIO));
	}

	if (sta->cancel)
	return (SET_ERROR(EINTR));
	if (zb->zb_object != DMU_META_DNODE_OBJECT &&
	DMU_OBJECT_IS_SPECIAL(zb->zb_object))
	return (0);
	atomic_inc_64(sta->num_blocks_visited);

	if (zb->zb_level == ZB_DNODE_LEVEL) {
	if (zb->zb_object == DMU_META_DNODE_OBJECT)
	return (0);
	record = range_alloc(OBJECT, zb->zb_object, 0, 0, B_FALSE);
	record->sru.object.bp = *bp;
	size_t size = sizeof (dnp) (dnp->dn_extra_slots + 1);
	record->sru.object.dnp = kmem_alloc(size, KM_SLEEP);
	memcpy(record->sru.object.dnp, dnp, size);
	bqueue_enqueue(&sta->q, record, sizeof (*record));
	return (0);
	}
	if (zb->zb_level == 0 && zb->zb_object == DMU_META_DNODE_OBJECT &&
	!BP_IS_HOLE(bp)) {
	record = range_alloc(OBJECT_RANGE, 0, zb->zb_blkid,
	zb->zb_blkid + 1, B_FALSE);
	record->sru.object_range.bp = *bp;
	bqueue_enqueue(&sta->q, record, sizeof (*record));
	return (0);
	}
	if (zb->zb_level < 0 \|\| (zb->zb_level > 0 && !BP_IS_HOLE(bp)))
	return (0);
	if (zb->zb_object == DMU_META_DNODE_OBJECT && !BP_IS_HOLE(bp))
	return (0);

	uint64_t span = bp_span_in_blocks(dnp->dn_indblkshift, zb->zb_level);
	uint64_t start;

	/*
	* If this multiply overflows, we don't need to send this block.
	* Even if it has a birth time, it can never not be a hole, so
	* we don't need to send records for it.
	*/
	if (!overflow_multiply(span, zb->zb_blkid, &start) \|\| (!(zb->zb_blkid ==
	DMU_SPILL_BLKID \|\| DMU_OT_IS_METADATA(dnp->dn_type)) &&
	span * zb->zb_blkid > dnp->dn_maxblkid)) {
	ASSERT(BP_IS_HOLE(bp));
	return (0);
	}

	if (zb->zb_blkid == DMU_SPILL_BLKID)
	ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_SA);

	enum type record_type = DATA;
	if (BP_IS_HOLE(bp))
	record_type = HOLE;
	else if (BP_IS_REDACTED(bp))
	record_type = REDACT;
	else
	record_type = DATA;

	record = range_alloc(record_type, zb->zb_object, start,
	(start + span < start ? 0 : start + span), B_FALSE);

	uint64_t datablksz = (zb->zb_blkid == DMU_SPILL_BLKID ?
	BP_GET_LSIZE(bp) : dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT);

	if (BP_IS_HOLE(bp)) {
	record->sru.hole.datablksz = datablksz;
	} else if (BP_IS_REDACTED(bp)) {
	record->sru.redact.datablksz = datablksz;
	} else {
	record->sru.data.datablksz = datablksz;
	record->sru.data.obj_type = dnp->dn_type;
	record->sru.data.bp = *bp;
	}

	bqueue_enqueue(&sta->q, record, sizeof (*record));
	return (0);
	}

	struct redact_list_cb_arg {
	uint64_t *num_blocks_visited;
	bqueue_t *q;
	boolean_t *cancel;
	boolean_t mark_redact;
	};

	static int
	redact_list_cb(redact_block_phys_t rb, void arg)
	{
	struct redact_list_cb_arg *rlcap = arg;

	atomic_inc_64(rlcap->num_blocks_visited);
	if (*rlcap->cancel)
	return (-1);

	struct send_range *data = range_alloc(REDACT, rb->rbp_object,
	rb->rbp_blkid, rb->rbp_blkid + redact_block_get_count(rb), B_FALSE);
	ASSERT3U(data->end_blkid, >, rb->rbp_blkid);
	if (rlcap->mark_redact) {
	data->type = REDACT;
	data->sru.redact.datablksz = redact_block_get_size(rb);
	} else {
	data->type = PREVIOUSLY_REDACTED;
	}
	bqueue_enqueue(rlcap->q, data, sizeof (*data));

	return (0);
	}

	/*
	* This function kicks off the traverse_dataset. It also handles setting the
	* error code of the thread in case something goes wrong, and pushes the End of
	* Stream record when the traverse_dataset call has finished.
	*/
	static __attribute__((noreturn)) void
	send_traverse_thread(void *arg)
	{
	struct send_thread_arg *st_arg = arg;
	int err = 0;
	struct send_range *data;
	fstrans_cookie_t cookie = spl_fstrans_mark();

	err = traverse_dataset_resume(st_arg->os->os_dsl_dataset,
	st_arg->fromtxg, &st_arg->resume,
	st_arg->flags, send_cb, st_arg);

	if (err != EINTR)
	st_arg->error_code = err;
	data = range_alloc(DATA, 0, 0, 0, B_TRUE);
	bqueue_enqueue_flush(&st_arg->q, data, sizeof (*data));
	spl_fstrans_unmark(cookie);
	thread_exit();
	}

	/*
	* Utility function that causes End of Stream records to compare after of all
	* others, so that other threads' comparison logic can stay simple.
	*/
	static int __attribute__((unused))
	send_range_after(const struct send_range from, const struct send_range to)
	{
	if (from->eos_marker == B_TRUE)
	return (1);
	if (to->eos_marker == B_TRUE)
	return (-1);

	uint64_t from_obj = from->object;
	uint64_t from_end_obj = from->object + 1;
	uint64_t to_obj = to->object;
	uint64_t to_end_obj = to->object + 1;
	if (from_obj == 0) {
	ASSERT(from->type == HOLE \|\| from->type == OBJECT_RANGE);
	from_obj = from->start_blkid << DNODES_PER_BLOCK_SHIFT;
	from_end_obj = from->end_blkid << DNODES_PER_BLOCK_SHIFT;
	}
	if (to_obj == 0) {
	ASSERT(to->type == HOLE \|\| to->type == OBJECT_RANGE);
	to_obj = to->start_blkid << DNODES_PER_BLOCK_SHIFT;
	to_end_obj = to->end_blkid << DNODES_PER_BLOCK_SHIFT;
	}

	if (from_end_obj <= to_obj)
	return (-1);
	if (from_obj >= to_end_obj)
	return (1);
	int64_t cmp = TREE_CMP(to->type == OBJECT_RANGE, from->type ==
	OBJECT_RANGE);
	if (unlikely(cmp))
	return (cmp);
	cmp = TREE_CMP(to->type == OBJECT, from->type == OBJECT);
	if (unlikely(cmp))
	return (cmp);
	if (from->end_blkid <= to->start_blkid)
	return (-1);
	if (from->start_blkid >= to->end_blkid)
	return (1);
	return (0);
	}

	/*
	* Pop the new data off the queue, check that the records we receive are in
	* the right order, but do not free the old data. This is used so that the
	* records can be sent on to the main thread without copying the data.
	*/
	static struct send_range *
	get_next_range_nofree(bqueue_t bq, struct send_range prev)
	{
	struct send_range *next = bqueue_dequeue(bq);
	ASSERT3S(send_range_after(prev, next), ==, -1);
	return (next);
	}

	/*
	* Pop the new data off the queue, check that the records we receive are in
	* the right order, and free the old data.
	*/
	static struct send_range *
	get_next_range(bqueue_t bq, struct send_range prev)
	{
	struct send_range *next = get_next_range_nofree(bq, prev);
	range_free(prev);
	return (next);
	}

	static __attribute__((noreturn)) void
	redact_list_thread(void *arg)
	{
	struct redact_list_thread_arg *rlt_arg = arg;
	struct send_range *record;
	fstrans_cookie_t cookie = spl_fstrans_mark();
	if (rlt_arg->rl != NULL) {
	struct redact_list_cb_arg rlcba = {0};
	rlcba.cancel = &rlt_arg->cancel;
	rlcba.q = &rlt_arg->q;
	rlcba.num_blocks_visited = rlt_arg->num_blocks_visited;
	rlcba.mark_redact = rlt_arg->mark_redact;
	int err = dsl_redaction_list_traverse(rlt_arg->rl,
	&rlt_arg->resume, redact_list_cb, &rlcba);
	if (err != EINTR)
	rlt_arg->error_code = err;
	}
	record = range_alloc(DATA, 0, 0, 0, B_TRUE);
	bqueue_enqueue_flush(&rlt_arg->q, record, sizeof (*record));
	spl_fstrans_unmark(cookie);

	thread_exit();
	}

	/*
	* Compare the start point of the two provided ranges. End of stream ranges
	* compare last, objects compare before any data or hole inside that object and
	* multi-object holes that start at the same object.
	*/
	static int
	send_range_start_compare(struct send_range r1, struct send_range r2)
	{
	uint64_t r1_objequiv = r1->object;
	uint64_t r1_l0equiv = r1->start_blkid;
	uint64_t r2_objequiv = r2->object;
	uint64_t r2_l0equiv = r2->start_blkid;
	int64_t cmp = TREE_CMP(r1->eos_marker, r2->eos_marker);
	if (unlikely(cmp))
	return (cmp);
	if (r1->object == 0) {
	r1_objequiv = r1->start_blkid * DNODES_PER_BLOCK;
	r1_l0equiv = 0;
	}
	if (r2->object == 0) {
	r2_objequiv = r2->start_blkid * DNODES_PER_BLOCK;
	r2_l0equiv = 0;
	}

	cmp = TREE_CMP(r1_objequiv, r2_objequiv);
	if (likely(cmp))
	return (cmp);
	cmp = TREE_CMP(r2->type == OBJECT_RANGE, r1->type == OBJECT_RANGE);
	if (unlikely(cmp))
	return (cmp);
	cmp = TREE_CMP(r2->type == OBJECT, r1->type == OBJECT);
	if (unlikely(cmp))
	return (cmp);

	return (TREE_CMP(r1_l0equiv, r2_l0equiv));
	}

	enum q_idx {
	REDACT_IDX = 0,
	TO_IDX,
	FROM_IDX,
	NUM_THREADS
	};

	/*
	* This function returns the next range the send_merge_thread should operate on.
	* The inputs are two arrays; the first one stores the range at the front of the
	* queues stored in the second one. The ranges are sorted in descending
	* priority order; the metadata from earlier ranges overrules metadata from
	* later ranges. out_mask is used to return which threads the ranges came from;
	* bit i is set if ranges[i] started at the same place as the returned range.
	*
	* This code is not hardcoded to compare a specific number of threads; it could
	* be used with any number, just by changing the q_idx enum.
	*
	* The "next range" is the one with the earliest start; if two starts are equal,
	* the highest-priority range is the next to operate on. If a higher-priority
	* range starts in the middle of the first range, then the first range will be
	* truncated to end where the higher-priority range starts, and we will operate
	* on that one next time. In this way, we make sure that each block covered by
	* some range gets covered by a returned range, and each block covered is
	* returned using the metadata of the highest-priority range it appears in.
	*
	* For example, if the three ranges at the front of the queues were [2,4),
	* [3,5), and [1,3), then the ranges returned would be [1,2) with the metadata
	* from the third range, [2,4) with the metadata from the first range, and then
	* [4,5) with the metadata from the second.
	*/
	static struct send_range *
	find_next_range(struct send_range ranges, bqueue_t qs, uint64_t *out_mask)
	{
	int idx = 0; // index of the range with the earliest start
	int i;
	uint64_t bmask = 0;
	for (i = 1; i < NUM_THREADS; i++) {
	if (send_range_start_compare(ranges[i], ranges[idx]) < 0)
	idx = i;
	}
	if (ranges[idx]->eos_marker) {
	struct send_range *ret = range_alloc(DATA, 0, 0, 0, B_TRUE);
	*out_mask = 0;
	return (ret);
	}
	/*
	* Find all the ranges that start at that same point.
	*/
	for (i = 0; i < NUM_THREADS; i++) {
	if (send_range_start_compare(ranges[i], ranges[idx]) == 0)
	bmask \|= 1 << i;
	}
	*out_mask = bmask;
	/*
	* OBJECT_RANGE records only come from the TO thread, and should always
	* be treated as overlapping with nothing and sent on immediately. They
	* are only used in raw sends, and are never redacted.
	*/
	if (ranges[idx]->type == OBJECT_RANGE) {
	ASSERT3U(idx, ==, TO_IDX);
	ASSERT3U(*out_mask, ==, 1 << TO_IDX);
	struct send_range *ret = ranges[idx];
	ranges[idx] = get_next_range_nofree(qs[idx], ranges[idx]);
	return (ret);
	}
	/*
	* Find the first start or end point after the start of the first range.
	*/
	uint64_t first_change = ranges[idx]->end_blkid;
	for (i = 0; i < NUM_THREADS; i++) {
	if (i == idx \|\| ranges[i]->eos_marker \|\|
	ranges[i]->object > ranges[idx]->object \|\|
	ranges[i]->object == DMU_META_DNODE_OBJECT)
	continue;
	ASSERT3U(ranges[i]->object, ==, ranges[idx]->object);
	if (first_change > ranges[i]->start_blkid &&
	(bmask & (1 << i)) == 0)
	first_change = ranges[i]->start_blkid;
	else if (first_change > ranges[i]->end_blkid)
	first_change = ranges[i]->end_blkid;
	}
	/*
	* Update all ranges to no longer overlap with the range we're
	* returning. All such ranges must start at the same place as the range
	* being returned, and end at or after first_change. Thus we update
	* their start to first_change. If that makes them size 0, then free
	* them and pull a new range from that thread.
	*/
	for (i = 0; i < NUM_THREADS; i++) {
	if (i == idx \|\| (bmask & (1 << i)) == 0)
	continue;
	ASSERT3U(first_change, >, ranges[i]->start_blkid);
	ranges[i]->start_blkid = first_change;
	ASSERT3U(ranges[i]->start_blkid, <=, ranges[i]->end_blkid);
	if (ranges[i]->start_blkid == ranges[i]->end_blkid)
	ranges[i] = get_next_range(qs[i], ranges[i]);
	}
	/*
	* Short-circuit the simple case; if the range doesn't overlap with
	* anything else, or it only overlaps with things that start at the same
	* place and are longer, send it on.
	*/
	if (first_change == ranges[idx]->end_blkid) {
	struct send_range *ret = ranges[idx];
	ranges[idx] = get_next_range_nofree(qs[idx], ranges[idx]);
	return (ret);
	}

	/*
	* Otherwise, return a truncated copy of ranges[idx] and move the start
	* of ranges[idx] back to first_change.
	*/
	struct send_range ret = kmem_alloc(sizeof (ret), KM_SLEEP);
	ret = ranges[idx];
	ret->end_blkid = first_change;
	ranges[idx]->start_blkid = first_change;
	return (ret);
	}

	#define FROM_AND_REDACT_BITS ((1 << REDACT_IDX) \| (1 << FROM_IDX))

	/*
	* Merge the results from the from thread and the to thread, and then hand the
	* records off to send_prefetch_thread to prefetch them. If this is not a
	* send from a redaction bookmark, the from thread will push an end of stream
	* record and stop, and we'll just send everything that was changed in the
	* to_ds since the ancestor's creation txg. If it is, then since
	* traverse_dataset has a canonical order, we can compare each change as
	* they're pulled off the queues. That will give us a stream that is
	* appropriately sorted, and covers all records. In addition, we pull the
	* data from the redact_list_thread and use that to determine which blocks
	* should be redacted.
	*/
	static __attribute__((noreturn)) void
	send_merge_thread(void *arg)
	{
	struct send_merge_thread_arg *smt_arg = arg;
	struct send_range *front_ranges[NUM_THREADS];
	bqueue_t *queues[NUM_THREADS];
	int err = 0;
	fstrans_cookie_t cookie = spl_fstrans_mark();

	if (smt_arg->redact_arg == NULL) {
	front_ranges[REDACT_IDX] =
	kmem_zalloc(sizeof (struct send_range), KM_SLEEP);
	front_ranges[REDACT_IDX]->eos_marker = B_TRUE;
	front_ranges[REDACT_IDX]->type = REDACT;
	queues[REDACT_IDX] = NULL;
	} else {
	front_ranges[REDACT_IDX] =
	bqueue_dequeue(&smt_arg->redact_arg->q);
	queues[REDACT_IDX] = &smt_arg->redact_arg->q;
	}
	front_ranges[TO_IDX] = bqueue_dequeue(&smt_arg->to_arg->q);
	queues[TO_IDX] = &smt_arg->to_arg->q;
	front_ranges[FROM_IDX] = bqueue_dequeue(&smt_arg->from_arg->q);
	queues[FROM_IDX] = &smt_arg->from_arg->q;
	uint64_t mask = 0;
	struct send_range *range;
	for (range = find_next_range(front_ranges, queues, &mask);
	!range->eos_marker && err == 0 && !smt_arg->cancel;
	range = find_next_range(front_ranges, queues, &mask)) {
	/*
	* If the range in question was in both the from redact bookmark
	* and the bookmark we're using to redact, then don't send it.
	* It's already redacted on the receiving system, so a redaction
	* record would be redundant.
	*/
	if ((mask & FROM_AND_REDACT_BITS) == FROM_AND_REDACT_BITS) {
	ASSERT3U(range->type, ==, REDACT);
	range_free(range);
	continue;
	}
	bqueue_enqueue(&smt_arg->q, range, sizeof (*range));

	if (smt_arg->to_arg->error_code != 0) {
	err = smt_arg->to_arg->error_code;
	} else if (smt_arg->from_arg->error_code != 0) {
	err = smt_arg->from_arg->error_code;
	} else if (smt_arg->redact_arg != NULL &&
	smt_arg->redact_arg->error_code != 0) {
	err = smt_arg->redact_arg->error_code;
	}
	}
	if (smt_arg->cancel && err == 0)
	err = SET_ERROR(EINTR);
	smt_arg->error = err;
	if (smt_arg->error != 0) {
	smt_arg->to_arg->cancel = B_TRUE;
	smt_arg->from_arg->cancel = B_TRUE;
	if (smt_arg->redact_arg != NULL)
	smt_arg->redact_arg->cancel = B_TRUE;
	}
	for (int i = 0; i < NUM_THREADS; i++) {
	while (!front_ranges[i]->eos_marker) {
	front_ranges[i] = get_next_range(queues[i],
	front_ranges[i]);
	}
	range_free(front_ranges[i]);
	}
	range->eos_marker = B_TRUE;
	bqueue_enqueue_flush(&smt_arg->q, range, 1);
	spl_fstrans_unmark(cookie);
	thread_exit();
	}

	struct send_reader_thread_arg {
	struct send_merge_thread_arg *smta;
	bqueue_t q;
	boolean_t cancel;
	boolean_t issue_reads;
	uint64_t featureflags;
	int error;
	};

	static void
	dmu_send_read_done(zio_t *zio)
	{
	struct send_range *range = zio->io_private;

	mutex_enter(&range->sru.data.lock);
	if (zio->io_error != 0) {
	abd_free(range->sru.data.abd);
	range->sru.data.abd = NULL;
	range->sru.data.io_err = zio->io_error;
	}

	ASSERT(range->sru.data.io_outstanding);
	range->sru.data.io_outstanding = B_FALSE;
	cv_broadcast(&range->sru.data.cv);
	mutex_exit(&range->sru.data.lock);
	}

	static void
	issue_data_read(struct send_reader_thread_arg srta, struct send_range range)
	{
	struct srd *srdp = &range->sru.data;
	blkptr_t *bp = &srdp->bp;
	objset_t *os = srta->smta->os;

	ASSERT3U(range->type, ==, DATA);
	ASSERT3U(range->start_blkid + 1, ==, range->end_blkid);
	/*
	* If we have large blocks stored on disk but
	* the send flags don't allow us to send large
	* blocks, we split the data from the arc buf
	* into chunks.
	*/
	boolean_t split_large_blocks =
	srdp->datablksz > SPA_OLD_MAXBLOCKSIZE &&
	!(srta->featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS);
	/*
	* We should only request compressed data from the ARC if all
	* the following are true:
	* - stream compression was requested
	* - we aren't splitting large blocks into smaller chunks
	* - the data won't need to be byteswapped before sending
	* - this isn't an embedded block
	* - this isn't metadata (if receiving on a different endian
	* system it can be byteswapped more easily)
	*/
	boolean_t request_compressed =
	(srta->featureflags & DMU_BACKUP_FEATURE_COMPRESSED) &&
	!split_large_blocks && !BP_SHOULD_BYTESWAP(bp) &&
	!BP_IS_EMBEDDED(bp) && !DMU_OT_IS_METADATA(BP_GET_TYPE(bp));

	zio_flag_t zioflags = ZIO_FLAG_CANFAIL;

	if (srta->featureflags & DMU_BACKUP_FEATURE_RAW) {
	zioflags \|= ZIO_FLAG_RAW;
	srdp->io_compressed = B_TRUE;
	} else if (request_compressed) {
	zioflags \|= ZIO_FLAG_RAW_COMPRESS;
	srdp->io_compressed = B_TRUE;
	}

	srdp->datasz = (zioflags & ZIO_FLAG_RAW_COMPRESS) ?
	BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp);

	if (!srta->issue_reads)
	return;
	if (BP_IS_REDACTED(bp))
	return;
	if (send_do_embed(bp, srta->featureflags))
	return;

	zbookmark_phys_t zb = {
	.zb_objset = dmu_objset_id(os),
	.zb_object = range->object,
	.zb_level = 0,
	.zb_blkid = range->start_blkid,
	};

	arc_flags_t aflags = ARC_FLAG_CACHED_ONLY;

	int arc_err = arc_read(NULL, os->os_spa, bp,
	arc_getbuf_func, &srdp->abuf, ZIO_PRIORITY_ASYNC_READ,
	zioflags, &aflags, &zb);
	/*
	* If the data is not already cached in the ARC, we read directly
	* from zio. This avoids the performance overhead of adding a new
	* entry to the ARC, and we also avoid polluting the ARC cache with
	* data that is not likely to be used in the future.
	*/
	if (arc_err != 0) {
	srdp->abd = abd_alloc_linear(srdp->datasz, B_FALSE);
	srdp->io_outstanding = B_TRUE;
	zio_nowait(zio_read(NULL, os->os_spa, bp, srdp->abd,
	srdp->datasz, dmu_send_read_done, range,
	ZIO_PRIORITY_ASYNC_READ, zioflags, &zb));
	}
	}

	/*
	* Create a new record with the given values.
	*/
	static void
	enqueue_range(struct send_reader_thread_arg srta, bqueue_t q, dnode_t *dn,
	uint64_t blkid, uint64_t count, const blkptr_t *bp, uint32_t datablksz)
	{
	enum type range_type = (bp == NULL \|\| BP_IS_HOLE(bp) ? HOLE :
	(BP_IS_REDACTED(bp) ? REDACT : DATA));

	struct send_range *range = range_alloc(range_type, dn->dn_object,
	blkid, blkid + count, B_FALSE);

	if (blkid == DMU_SPILL_BLKID) {
	ASSERT3P(bp, !=, NULL);
	ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_SA);
	}

	switch (range_type) {
	case HOLE:
	range->sru.hole.datablksz = datablksz;
	break;
	case DATA:
	ASSERT3U(count, ==, 1);
	range->sru.data.datablksz = datablksz;
	range->sru.data.obj_type = dn->dn_type;
	range->sru.data.bp = *bp;
	issue_data_read(srta, range);
	break;
	case REDACT:
	range->sru.redact.datablksz = datablksz;
	break;
	default:
	break;
	}
	bqueue_enqueue(q, range, datablksz);
	}

	/*
	* This thread is responsible for two things: First, it retrieves the correct
	* blkptr in the to ds if we need to send the data because of something from
	* the from thread. As a result of this, we're the first ones to discover that
	* some indirect blocks can be discarded because they're not holes. Second,
	* it issues prefetches for the data we need to send.
	*/
	static __attribute__((noreturn)) void
	send_reader_thread(void *arg)
	{
	struct send_reader_thread_arg *srta = arg;
	struct send_merge_thread_arg *smta = srta->smta;
	bqueue_t *inq = &smta->q;
	bqueue_t *outq = &srta->q;
	objset_t *os = smta->os;
	fstrans_cookie_t cookie = spl_fstrans_mark();
	struct send_range *range = bqueue_dequeue(inq);
	int err = 0;

	/*
	* If the record we're analyzing is from a redaction bookmark from the
	* fromds, then we need to know whether or not it exists in the tods so
	* we know whether to create records for it or not. If it does, we need
	* the datablksz so we can generate an appropriate record for it.
	* Finally, if it isn't redacted, we need the blkptr so that we can send
	* a WRITE record containing the actual data.
	*/
	uint64_t last_obj = UINT64_MAX;
	uint64_t last_obj_exists = B_TRUE;
	while (!range->eos_marker && !srta->cancel && smta->error == 0 &&
	err == 0) {
	switch (range->type) {
	case DATA:
	issue_data_read(srta, range);
	bqueue_enqueue(outq, range, range->sru.data.datablksz);
	range = get_next_range_nofree(inq, range);
	break;
	case HOLE:
	case OBJECT:
	case OBJECT_RANGE:
	case REDACT: // Redacted blocks must exist
	bqueue_enqueue(outq, range, sizeof (*range));
	range = get_next_range_nofree(inq, range);
	break;
	case PREVIOUSLY_REDACTED: {
	/*
	* This entry came from the "from bookmark" when
	* sending from a bookmark that has a redaction
	* list. We need to check if this object/blkid
	* exists in the target ("to") dataset, and if
	* not then we drop this entry. We also need
	* to fill in the block pointer so that we know
	* what to prefetch.
	*
	* To accomplish the above, we first cache whether or
	* not the last object we examined exists. If it
	* doesn't, we can drop this record. If it does, we hold
	* the dnode and use it to call dbuf_dnode_findbp. We do
	* this instead of dbuf_bookmark_findbp because we will
	* often operate on large ranges, and holding the dnode
	* once is more efficient.
	*/
	boolean_t object_exists = B_TRUE;
	/*
	* If the data is redacted, we only care if it exists,
	* so that we don't send records for objects that have
	* been deleted.
	*/
	dnode_t *dn;
	if (range->object == last_obj && !last_obj_exists) {
	/*
	* If we're still examining the same object as
	* previously, and it doesn't exist, we don't
	* need to call dbuf_bookmark_findbp.
	*/
	object_exists = B_FALSE;
	} else {
	err = dnode_hold(os, range->object, FTAG, &dn);
	if (err == ENOENT) {
	object_exists = B_FALSE;
	err = 0;
	}
	last_obj = range->object;
	last_obj_exists = object_exists;
	}

	if (err != 0) {
	break;
	} else if (!object_exists) {
	/*
	* The block was modified, but doesn't
	* exist in the to dataset; if it was
	* deleted in the to dataset, then we'll
	* visit the hole bp for it at some point.
	*/
	range = get_next_range(inq, range);
	continue;
	}
	uint64_t file_max =
	MIN(dn->dn_maxblkid, range->end_blkid);
	/*
	* The object exists, so we need to try to find the
	* blkptr for each block in the range we're processing.
	*/
	rw_enter(&dn->dn_struct_rwlock, RW_READER);
	for (uint64_t blkid = range->start_blkid;
	blkid < file_max; blkid++) {
	blkptr_t bp;
	uint32_t datablksz =
	dn->dn_phys->dn_datablkszsec <<
	SPA_MINBLOCKSHIFT;
	uint64_t offset = blkid * datablksz;
	/*
	* This call finds the next non-hole block in
	* the object. This is to prevent a
	* performance problem where we're unredacting
	* a large hole. Using dnode_next_offset to
	* skip over the large hole avoids iterating
	* over every block in it.
	*/
	err = dnode_next_offset(dn, DNODE_FIND_HAVELOCK,
	&offset, 1, 1, 0);
	if (err == ESRCH) {
	offset = UINT64_MAX;
	err = 0;
	} else if (err != 0) {
	break;
	}
	if (offset != blkid * datablksz) {
	/*
	* if there is a hole from here
	* (blkid) to offset
	*/
	offset = MIN(offset, file_max *
	datablksz);
	uint64_t nblks = (offset / datablksz) -
	blkid;
	enqueue_range(srta, outq, dn, blkid,
	nblks, NULL, datablksz);
	blkid += nblks;
	}
	if (blkid >= file_max)
	break;
	err = dbuf_dnode_findbp(dn, 0, blkid, &bp,
	NULL, NULL);
	if (err != 0)
	break;
	ASSERT(!BP_IS_HOLE(&bp));
	enqueue_range(srta, outq, dn, blkid, 1, &bp,
	datablksz);
	}
	rw_exit(&dn->dn_struct_rwlock);
	dnode_rele(dn, FTAG);
	range = get_next_range(inq, range);
	}
	}
	}
	if (srta->cancel \|\| err != 0) {
	smta->cancel = B_TRUE;
	srta->error = err;
	} else if (smta->error != 0) {
	srta->error = smta->error;
	}
	while (!range->eos_marker)
	range = get_next_range(inq, range);

	bqueue_enqueue_flush(outq, range, 1);
	spl_fstrans_unmark(cookie);
	thread_exit();
	}

	#define NUM_SNAPS_NOT_REDACTED UINT64_MAX

	struct dmu_send_params {
	/* Pool args */
	const void *tag; // Tag dp was held with, will be used to release dp.
	dsl_pool_t *dp;
	/* To snapshot args */
	const char *tosnap;
	dsl_dataset_t *to_ds;
	/* From snapshot args */
	zfs_bookmark_phys_t ancestor_zb;
	uint64_t *fromredactsnaps;
	/* NUM_SNAPS_NOT_REDACTED if not sending from redaction bookmark */
	uint64_t numfromredactsnaps;
	/* Stream params */
	boolean_t is_clone;
	boolean_t embedok;
	boolean_t large_block_ok;
	boolean_t compressok;
	boolean_t rawok;
	boolean_t savedok;
	uint64_t resumeobj;
	uint64_t resumeoff;
	uint64_t saved_guid;
	zfs_bookmark_phys_t *redactbook;
	/* Stream output params */
	dmu_send_outparams_t *dso;

	/* Stream progress params */
	offset_t *off;
	int outfd;
	char saved_toname[MAXNAMELEN];
	};

	static int
	setup_featureflags(struct dmu_send_params dspp, objset_t os,
	uint64_t *featureflags)
	{
	dsl_dataset_t *to_ds = dspp->to_ds;
	dsl_pool_t *dp = dspp->dp;

	if (dmu_objset_type(os) == DMU_OST_ZFS) {
	uint64_t version;
	if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &version) != 0)
	return (SET_ERROR(EINVAL));

	if (version >= ZPL_VERSION_SA)
	*featureflags \|= DMU_BACKUP_FEATURE_SA_SPILL;
	}

	/* raw sends imply large_block_ok */
	if ((dspp->rawok \|\| dspp->large_block_ok) &&
	dsl_dataset_feature_is_active(to_ds, SPA_FEATURE_LARGE_BLOCKS)) {
	*featureflags \|= DMU_BACKUP_FEATURE_LARGE_BLOCKS;
	}

	/* encrypted datasets will not have embedded blocks */
	if ((dspp->embedok \|\| dspp->rawok) && !os->os_encrypted &&
	spa_feature_is_active(dp->dp_spa, SPA_FEATURE_EMBEDDED_DATA)) {
	*featureflags \|= DMU_BACKUP_FEATURE_EMBED_DATA;
	}

	/* raw send implies compressok */
	if (dspp->compressok \|\| dspp->rawok)
	*featureflags \|= DMU_BACKUP_FEATURE_COMPRESSED;

	if (dspp->rawok && os->os_encrypted)
	*featureflags \|= DMU_BACKUP_FEATURE_RAW;

	if ((*featureflags &
	(DMU_BACKUP_FEATURE_EMBED_DATA \| DMU_BACKUP_FEATURE_COMPRESSED \|
	DMU_BACKUP_FEATURE_RAW)) != 0 &&
	spa_feature_is_active(dp->dp_spa, SPA_FEATURE_LZ4_COMPRESS)) {
	*featureflags \|= DMU_BACKUP_FEATURE_LZ4;
	}

	/*
	* We specifically do not include DMU_BACKUP_FEATURE_EMBED_DATA here to
	* allow sending ZSTD compressed datasets to a receiver that does not
	* support ZSTD
	*/
	if ((*featureflags &
	(DMU_BACKUP_FEATURE_COMPRESSED \| DMU_BACKUP_FEATURE_RAW)) != 0 &&
	dsl_dataset_feature_is_active(to_ds, SPA_FEATURE_ZSTD_COMPRESS)) {
	*featureflags \|= DMU_BACKUP_FEATURE_ZSTD;
	}

	if (dspp->resumeobj != 0 \|\| dspp->resumeoff != 0) {
	*featureflags \|= DMU_BACKUP_FEATURE_RESUMING;
	}

	if (dspp->redactbook != NULL) {
	*featureflags \|= DMU_BACKUP_FEATURE_REDACTED;
	}

	if (dsl_dataset_feature_is_active(to_ds, SPA_FEATURE_LARGE_DNODE)) {
	*featureflags \|= DMU_BACKUP_FEATURE_LARGE_DNODE;
	}
	return (0);
	}

	static dmu_replay_record_t *
	create_begin_record(struct dmu_send_params dspp, objset_t os,
	uint64_t featureflags)
	{
	dmu_replay_record_t *drr = kmem_zalloc(sizeof (dmu_replay_record_t),
	KM_SLEEP);
	drr->drr_type = DRR_BEGIN;

	struct drr_begin *drrb = &drr->drr_u.drr_begin;
	dsl_dataset_t *to_ds = dspp->to_ds;

	drrb->drr_magic = DMU_BACKUP_MAGIC;
	drrb->drr_creation_time = dsl_dataset_phys(to_ds)->ds_creation_time;
	drrb->drr_type = dmu_objset_type(os);
	drrb->drr_toguid = dsl_dataset_phys(to_ds)->ds_guid;
	drrb->drr_fromguid = dspp->ancestor_zb.zbm_guid;

	DMU_SET_STREAM_HDRTYPE(drrb->drr_versioninfo, DMU_SUBSTREAM);
	DMU_SET_FEATUREFLAGS(drrb->drr_versioninfo, featureflags);

	if (dspp->is_clone)
	drrb->drr_flags \|= DRR_FLAG_CLONE;
	if (dsl_dataset_phys(dspp->to_ds)->ds_flags & DS_FLAG_CI_DATASET)
	drrb->drr_flags \|= DRR_FLAG_CI_DATA;
	if (zfs_send_set_freerecords_bit)
	drrb->drr_flags \|= DRR_FLAG_FREERECORDS;
	drr->drr_u.drr_begin.drr_flags \|= DRR_FLAG_SPILL_BLOCK;

	if (dspp->savedok) {
	drrb->drr_toguid = dspp->saved_guid;
	strlcpy(drrb->drr_toname, dspp->saved_toname,
	sizeof (drrb->drr_toname));
	} else {
	dsl_dataset_name(to_ds, drrb->drr_toname);
	if (!to_ds->ds_is_snapshot) {
	(void) strlcat(drrb->drr_toname, "@--head--",
	sizeof (drrb->drr_toname));
	}
	}
	return (drr);
	}

	static void
	setup_to_thread(struct send_thread_arg to_arg, objset_t to_os,
	dmu_sendstatus_t *dssp, uint64_t fromtxg, boolean_t rawok)
	{
	VERIFY0(bqueue_init(&to_arg->q, zfs_send_no_prefetch_queue_ff,
	MAX(zfs_send_no_prefetch_queue_length, 2 * zfs_max_recordsize),
	offsetof(struct send_range, ln)));
	to_arg->error_code = 0;
	to_arg->cancel = B_FALSE;
	to_arg->os = to_os;
	to_arg->fromtxg = fromtxg;
	to_arg->flags = TRAVERSE_PRE \| TRAVERSE_PREFETCH_METADATA;
	if (rawok)
	to_arg->flags \|= TRAVERSE_NO_DECRYPT;
	if (zfs_send_corrupt_data)
	to_arg->flags \|= TRAVERSE_HARD;
	to_arg->num_blocks_visited = &dssp->dss_blocks;
	(void) thread_create(NULL, 0, send_traverse_thread, to_arg, 0,
	curproc, TS_RUN, minclsyspri);
	}

	static void
	setup_from_thread(struct redact_list_thread_arg *from_arg,
	redaction_list_t from_rl, dmu_sendstatus_t dssp)
	{
	VERIFY0(bqueue_init(&from_arg->q, zfs_send_no_prefetch_queue_ff,
	MAX(zfs_send_no_prefetch_queue_length, 2 * zfs_max_recordsize),
	offsetof(struct send_range, ln)));
	from_arg->error_code = 0;
	from_arg->cancel = B_FALSE;
	from_arg->rl = from_rl;
	from_arg->mark_redact = B_FALSE;
	from_arg->num_blocks_visited = &dssp->dss_blocks;
	/*
	* If from_ds is null, send_traverse_thread just returns success and
	* enqueues an eos marker.
	*/
	(void) thread_create(NULL, 0, redact_list_thread, from_arg, 0,
	curproc, TS_RUN, minclsyspri);
	}

	static void
	setup_redact_list_thread(struct redact_list_thread_arg *rlt_arg,
	struct dmu_send_params dspp, redaction_list_t rl, dmu_sendstatus_t *dssp)
	{
	if (dspp->redactbook == NULL)
	return;

	rlt_arg->cancel = B_FALSE;
	VERIFY0(bqueue_init(&rlt_arg->q, zfs_send_no_prefetch_queue_ff,
	MAX(zfs_send_no_prefetch_queue_length, 2 * zfs_max_recordsize),
	offsetof(struct send_range, ln)));
	rlt_arg->error_code = 0;
	rlt_arg->mark_redact = B_TRUE;
	rlt_arg->rl = rl;
	rlt_arg->num_blocks_visited = &dssp->dss_blocks;

	(void) thread_create(NULL, 0, redact_list_thread, rlt_arg, 0,
	curproc, TS_RUN, minclsyspri);
	}

	static void
	setup_merge_thread(struct send_merge_thread_arg *smt_arg,
	struct dmu_send_params dspp, struct redact_list_thread_arg from_arg,
	struct send_thread_arg to_arg, struct redact_list_thread_arg rlt_arg,
	objset_t *os)
	{
	VERIFY0(bqueue_init(&smt_arg->q, zfs_send_no_prefetch_queue_ff,
	MAX(zfs_send_no_prefetch_queue_length, 2 * zfs_max_recordsize),
	offsetof(struct send_range, ln)));
	smt_arg->cancel = B_FALSE;
	smt_arg->error = 0;
	smt_arg->from_arg = from_arg;
	smt_arg->to_arg = to_arg;
	if (dspp->redactbook != NULL)
	smt_arg->redact_arg = rlt_arg;

	smt_arg->os = os;
	(void) thread_create(NULL, 0, send_merge_thread, smt_arg, 0, curproc,
	TS_RUN, minclsyspri);
	}

	static void
	setup_reader_thread(struct send_reader_thread_arg *srt_arg,
	struct dmu_send_params dspp, struct send_merge_thread_arg smt_arg,
	uint64_t featureflags)
	{
	VERIFY0(bqueue_init(&srt_arg->q, zfs_send_queue_ff,
	MAX(zfs_send_queue_length, 2 * zfs_max_recordsize),
	offsetof(struct send_range, ln)));
	srt_arg->smta = smt_arg;
	srt_arg->issue_reads = !dspp->dso->dso_dryrun;
	srt_arg->featureflags = featureflags;
	(void) thread_create(NULL, 0, send_reader_thread, srt_arg, 0,
	curproc, TS_RUN, minclsyspri);
	}

	static int
	setup_resume_points(struct dmu_send_params *dspp,
	struct send_thread_arg to_arg, struct redact_list_thread_arg from_arg,
	struct redact_list_thread_arg *rlt_arg,
	struct send_merge_thread_arg smt_arg, boolean_t resuming, objset_t os,
	redaction_list_t redact_rl, nvlist_t nvl)
	{
	(void) smt_arg;
	dsl_dataset_t *to_ds = dspp->to_ds;
	int err = 0;

	uint64_t obj = 0;
	uint64_t blkid = 0;
	if (resuming) {
	obj = dspp->resumeobj;
	dmu_object_info_t to_doi;
	err = dmu_object_info(os, obj, &to_doi);
	if (err != 0)
	return (err);

	blkid = dspp->resumeoff / to_doi.doi_data_block_size;
	}
	/*
	* If we're resuming a redacted send, we can skip to the appropriate
	* point in the redaction bookmark by binary searching through it.
	*/
	if (redact_rl != NULL) {
	SET_BOOKMARK(&rlt_arg->resume, to_ds->ds_object, obj, 0, blkid);
	}

	SET_BOOKMARK(&to_arg->resume, to_ds->ds_object, obj, 0, blkid);
	if (nvlist_exists(nvl, BEGINNV_REDACT_FROM_SNAPS)) {
	uint64_t objset = dspp->ancestor_zb.zbm_redaction_obj;
	/*
	* Note: If the resume point is in an object whose
	* blocksize is different in the from vs to snapshots,
	* we will have divided by the "wrong" blocksize.
	* However, in this case fromsnap's send_cb() will
	* detect that the blocksize has changed and therefore
	* ignore this object.
	*
	* If we're resuming a send from a redaction bookmark,
	* we still cannot accidentally suggest blocks behind
	* the to_ds. In addition, we know that any blocks in
	* the object in the to_ds will have to be sent, since
	* the size changed. Therefore, we can't cause any harm
	* this way either.
	*/
	SET_BOOKMARK(&from_arg->resume, objset, obj, 0, blkid);
	}
	if (resuming) {
	fnvlist_add_uint64(nvl, BEGINNV_RESUME_OBJECT, dspp->resumeobj);
	fnvlist_add_uint64(nvl, BEGINNV_RESUME_OFFSET, dspp->resumeoff);
	}
	return (0);
	}

	static dmu_sendstatus_t *
	setup_send_progress(struct dmu_send_params *dspp)
	{
	dmu_sendstatus_t dssp = kmem_zalloc(sizeof (dssp), KM_SLEEP);
	dssp->dss_outfd = dspp->outfd;
	dssp->dss_off = dspp->off;
	dssp->dss_proc = curproc;
	mutex_enter(&dspp->to_ds->ds_sendstream_lock);
	list_insert_head(&dspp->to_ds->ds_sendstreams, dssp);
	mutex_exit(&dspp->to_ds->ds_sendstream_lock);
	return (dssp);
	}

	/*
	* Actually do the bulk of the work in a zfs send.
	*
	* The idea is that we want to do a send from ancestor_zb to to_ds. We also
	* want to not send any data that has been modified by all the datasets in
	* redactsnaparr, and store the list of blocks that are redacted in this way in
	* a bookmark named redactbook, created on the to_ds. We do this by creating
	* several worker threads, whose function is described below.
	*
	* There are three cases.
	* The first case is a redacted zfs send. In this case there are 5 threads.
	* The first thread is the to_ds traversal thread: it calls dataset_traverse on
	* the to_ds and finds all the blocks that have changed since ancestor_zb (if
	* it's a full send, that's all blocks in the dataset). It then sends those
	* blocks on to the send merge thread. The redact list thread takes the data
	* from the redaction bookmark and sends those blocks on to the send merge
	* thread. The send merge thread takes the data from the to_ds traversal
	* thread, and combines it with the redaction records from the redact list
	* thread. If a block appears in both the to_ds's data and the redaction data,
	* the send merge thread will mark it as redacted and send it on to the prefetch
	* thread. Otherwise, the send merge thread will send the block on to the
	* prefetch thread unchanged. The prefetch thread will issue prefetch reads for
	* any data that isn't redacted, and then send the data on to the main thread.
	* The main thread behaves the same as in a normal send case, issuing demand
	* reads for data blocks and sending out records over the network
	*
	* The graphic below diagrams the flow of data in the case of a redacted zfs
	* send. Each box represents a thread, and each line represents the flow of
	* data.
	*
	* Records from the \|
	* redaction bookmark \|
	* +--------------------+ \| +---------------------------+
	* \| \| v \| Send Merge Thread \|
	* \| Redact List Thread +----------> Apply redaction marks to \|
	* \| \| \| records as specified by \|
	* +--------------------+ \| redaction ranges \|
	* +----^---------------+------+
	* \| \| Merged data
	* \| \|
	* \| +------------v--------+
	* \| \| Prefetch Thread \|
	* +--------------------+ \| \| Issues prefetch \|
	* \| to_ds Traversal \| \| \| reads of data blocks\|
	* \| Thread (finds +---------------+ +------------+--------+
	* \| candidate blocks) \| Blocks modified \| Prefetched data
	* +--------------------+ by to_ds since \|
	* ancestor_zb +------------v----+
	* \| Main Thread \| File Descriptor
	* \| Sends data over +->(to zfs receive)
	* \| wire \|
	* +-----------------+
	*
	* The second case is an incremental send from a redaction bookmark. The to_ds
	* traversal thread and the main thread behave the same as in the redacted
	* send case. The new thread is the from bookmark traversal thread. It
	* iterates over the redaction list in the redaction bookmark, and enqueues
	* records for each block that was redacted in the original send. The send
	* merge thread now has to merge the data from the two threads. For details
	* about that process, see the header comment of send_merge_thread(). Any data
	* it decides to send on will be prefetched by the prefetch thread. Note that
	* you can perform a redacted send from a redaction bookmark; in that case,
	* the data flow behaves very similarly to the flow in the redacted send case,
	* except with the addition of the bookmark traversal thread iterating over the
	* redaction bookmark. The send_merge_thread also has to take on the
	* responsibility of merging the redact list thread's records, the bookmark
	* traversal thread's records, and the to_ds records.
	*
	* +---------------------+
	* \| \|
	* \| Redact List Thread +--------------+
	* \| \| \|
	* +---------------------+ \|
	* Blocks in redaction list \| Ranges modified by every secure snap
	* of from bookmark \| (or EOS if not readcted)
	* \|
	* +---------------------+ \| +----v----------------------+
	* \| bookmark Traversal \| v \| Send Merge Thread \|
	* \| Thread (finds +---------> Merges bookmark, rlt, and \|
	* \| candidate blocks) \| \| to_ds send records \|
	* +---------------------+ +----^---------------+------+
	* \| \| Merged data
	* \| +------------v--------+
	* \| \| Prefetch Thread \|
	* +--------------------+ \| \| Issues prefetch \|
	* \| to_ds Traversal \| \| \| reads of data blocks\|
	* \| Thread (finds +---------------+ +------------+--------+
	* \| candidate blocks) \| Blocks modified \| Prefetched data
	* +--------------------+ by to_ds since +------------v----+
	* ancestor_zb \| Main Thread \| File Descriptor
	* \| Sends data over +->(to zfs receive)
	* \| wire \|
	* +-----------------+
	*
	* The final case is a simple zfs full or incremental send. The to_ds traversal
	* thread behaves the same as always. The redact list thread is never started.
	* The send merge thread takes all the blocks that the to_ds traversal thread
	* sends it, prefetches the data, and sends the blocks on to the main thread.
	* The main thread sends the data over the wire.
	*
	* To keep performance acceptable, we want to prefetch the data in the worker
	* threads. While the to_ds thread could simply use the TRAVERSE_PREFETCH
	* feature built into traverse_dataset, the combining and deletion of records
	* due to redaction and sends from redaction bookmarks mean that we could
	* issue many unnecessary prefetches. As a result, we only prefetch data
	* after we've determined that the record is not going to be redacted. To
	* prevent the prefetching from getting too far ahead of the main thread, the
	* blocking queues that are used for communication are capped not by the
	* number of entries in the queue, but by the sum of the size of the
	* prefetches associated with them. The limit on the amount of data that the
	* thread can prefetch beyond what the main thread has reached is controlled
	* by the global variable zfs_send_queue_length. In addition, to prevent poor
	* performance in the beginning of a send, we also limit the distance ahead
	* that the traversal threads can be. That distance is controlled by the
	* zfs_send_no_prefetch_queue_length tunable.
	*
	* Note: Releases dp using the specified tag.
	*/
	static int
	dmu_send_impl(struct dmu_send_params *dspp)
	{
	objset_t *os;
	dmu_replay_record_t *drr;
	dmu_sendstatus_t *dssp;
	dmu_send_cookie_t dsc = {0};
	int err;
	uint64_t fromtxg = dspp->ancestor_zb.zbm_creation_txg;
	uint64_t featureflags = 0;
	struct redact_list_thread_arg *from_arg;
	struct send_thread_arg *to_arg;
	struct redact_list_thread_arg *rlt_arg;
	struct send_merge_thread_arg *smt_arg;
	struct send_reader_thread_arg *srt_arg;
	struct send_range *range;
	redaction_list_t *from_rl = NULL;
	redaction_list_t *redact_rl = NULL;
	boolean_t resuming = (dspp->resumeobj != 0 \|\| dspp->resumeoff != 0);
	boolean_t book_resuming = resuming;

	dsl_dataset_t *to_ds = dspp->to_ds;
	zfs_bookmark_phys_t *ancestor_zb = &dspp->ancestor_zb;
	dsl_pool_t *dp = dspp->dp;
	const void *tag = dspp->tag;

	err = dmu_objset_from_ds(to_ds, &os);
	if (err != 0) {
	dsl_pool_rele(dp, tag);
	return (err);
	}

	/*
	* If this is a non-raw send of an encrypted ds, we can ensure that
	* the objset_phys_t is authenticated. This is safe because this is
	* either a snapshot or we have owned the dataset, ensuring that
	* it can't be modified.
	*/
	if (!dspp->rawok && os->os_encrypted &&
	arc_is_unauthenticated(os->os_phys_buf)) {
	zbookmark_phys_t zb;

	SET_BOOKMARK(&zb, to_ds->ds_object, ZB_ROOT_OBJECT,
	ZB_ROOT_LEVEL, ZB_ROOT_BLKID);
	err = arc_untransform(os->os_phys_buf, os->os_spa,
	&zb, B_FALSE);
	if (err != 0) {
	dsl_pool_rele(dp, tag);
	return (err);
	}

	ASSERT0(arc_is_unauthenticated(os->os_phys_buf));
	}

	if ((err = setup_featureflags(dspp, os, &featureflags)) != 0) {
	dsl_pool_rele(dp, tag);
	return (err);
	}

	/*
	* If we're doing a redacted send, hold the bookmark's redaction list.
	*/
	if (dspp->redactbook != NULL) {
	err = dsl_redaction_list_hold_obj(dp,
	dspp->redactbook->zbm_redaction_obj, FTAG,
	&redact_rl);
	if (err != 0) {
	dsl_pool_rele(dp, tag);
	return (SET_ERROR(EINVAL));
	}
	dsl_redaction_list_long_hold(dp, redact_rl, FTAG);
	}

	/*
	* If we're sending from a redaction bookmark, hold the redaction list
	* so that we can consider sending the redacted blocks.
	*/
	if (ancestor_zb->zbm_redaction_obj != 0) {
	err = dsl_redaction_list_hold_obj(dp,
	ancestor_zb->zbm_redaction_obj, FTAG, &from_rl);
	if (err != 0) {
	if (redact_rl != NULL) {
	dsl_redaction_list_long_rele(redact_rl, FTAG);
	dsl_redaction_list_rele(redact_rl, FTAG);
	}
	dsl_pool_rele(dp, tag);
	return (SET_ERROR(EINVAL));
	}
	dsl_redaction_list_long_hold(dp, from_rl, FTAG);
	}

	dsl_dataset_long_hold(to_ds, FTAG);

	from_arg = kmem_zalloc(sizeof (*from_arg), KM_SLEEP);
	to_arg = kmem_zalloc(sizeof (*to_arg), KM_SLEEP);
	rlt_arg = kmem_zalloc(sizeof (*rlt_arg), KM_SLEEP);
	smt_arg = kmem_zalloc(sizeof (*smt_arg), KM_SLEEP);
	srt_arg = kmem_zalloc(sizeof (*srt_arg), KM_SLEEP);

	drr = create_begin_record(dspp, os, featureflags);
	dssp = setup_send_progress(dspp);

	dsc.dsc_drr = drr;
	dsc.dsc_dso = dspp->dso;
	dsc.dsc_os = os;
	dsc.dsc_off = dspp->off;
	dsc.dsc_toguid = dsl_dataset_phys(to_ds)->ds_guid;
	dsc.dsc_fromtxg = fromtxg;
	dsc.dsc_pending_op = PENDING_NONE;
	dsc.dsc_featureflags = featureflags;
	dsc.dsc_resume_object = dspp->resumeobj;
	dsc.dsc_resume_offset = dspp->resumeoff;

	dsl_pool_rele(dp, tag);

	void *payload = NULL;
	size_t payload_len = 0;
	nvlist_t *nvl = fnvlist_alloc();

	/*
	* If we're doing a redacted send, we include the snapshots we're
	* redacted with respect to so that the target system knows what send
	* streams can be correctly received on top of this dataset. If we're
	* instead sending a redacted dataset, we include the snapshots that the
	* dataset was created with respect to.
	*/
	if (dspp->redactbook != NULL) {
	fnvlist_add_uint64_array(nvl, BEGINNV_REDACT_SNAPS,
	redact_rl->rl_phys->rlp_snaps,
	redact_rl->rl_phys->rlp_num_snaps);
	} else if (dsl_dataset_feature_is_active(to_ds,
	SPA_FEATURE_REDACTED_DATASETS)) {
	uint64_t *tods_guids;
	uint64_t length;
	VERIFY(dsl_dataset_get_uint64_array_feature(to_ds,
	SPA_FEATURE_REDACTED_DATASETS, &length, &tods_guids));
	fnvlist_add_uint64_array(nvl, BEGINNV_REDACT_SNAPS, tods_guids,
	length);
	}

	/*
	* If we're sending from a redaction bookmark, then we should retrieve
	* the guids of that bookmark so we can send them over the wire.
	*/
	if (from_rl != NULL) {
	fnvlist_add_uint64_array(nvl, BEGINNV_REDACT_FROM_SNAPS,
	from_rl->rl_phys->rlp_snaps,
	from_rl->rl_phys->rlp_num_snaps);
	}

	/*
	* If the snapshot we're sending from is redacted, include the redaction
	* list in the stream.
	*/
	if (dspp->numfromredactsnaps != NUM_SNAPS_NOT_REDACTED) {
	ASSERT3P(from_rl, ==, NULL);
	fnvlist_add_uint64_array(nvl, BEGINNV_REDACT_FROM_SNAPS,
	dspp->fromredactsnaps, (uint_t)dspp->numfromredactsnaps);
	if (dspp->numfromredactsnaps > 0) {
	kmem_free(dspp->fromredactsnaps,
	dspp->numfromredactsnaps * sizeof (uint64_t));
	dspp->fromredactsnaps = NULL;
	}
	}

	if (resuming \|\| book_resuming) {
	err = setup_resume_points(dspp, to_arg, from_arg,
	rlt_arg, smt_arg, resuming, os, redact_rl, nvl);
	if (err != 0)
	goto out;
	}

	if (featureflags & DMU_BACKUP_FEATURE_RAW) {
	uint64_t ivset_guid = ancestor_zb->zbm_ivset_guid;
	nvlist_t *keynvl = NULL;
	ASSERT(os->os_encrypted);

	err = dsl_crypto_populate_key_nvlist(os, ivset_guid,
	&keynvl);
	if (err != 0) {
	fnvlist_free(nvl);
	goto out;
	}

	fnvlist_add_nvlist(nvl, "crypt_keydata", keynvl);
	fnvlist_free(keynvl);
	}

	if (!nvlist_empty(nvl)) {
	payload = fnvlist_pack(nvl, &payload_len);
	drr->drr_payloadlen = payload_len;
	}

	fnvlist_free(nvl);
	err = dump_record(&dsc, payload, payload_len);
	fnvlist_pack_free(payload, payload_len);
	if (err != 0) {
	err = dsc.dsc_err;
	goto out;
	}

	setup_to_thread(to_arg, os, dssp, fromtxg, dspp->rawok);
	setup_from_thread(from_arg, from_rl, dssp);
	setup_redact_list_thread(rlt_arg, dspp, redact_rl, dssp);
	setup_merge_thread(smt_arg, dspp, from_arg, to_arg, rlt_arg, os);
	setup_reader_thread(srt_arg, dspp, smt_arg, featureflags);

	range = bqueue_dequeue(&srt_arg->q);
	while (err == 0 && !range->eos_marker) {
	err = do_dump(&dsc, range);
	range = get_next_range(&srt_arg->q, range);
	- if (issig(JUSTLOOKING) && issig(FORREAL))
	+ if (issig())
	err = SET_ERROR(EINTR);
	}

	/*
	* If we hit an error or are interrupted, cancel our worker threads and
	* clear the queue of any pending records. The threads will pass the
	* cancel up the tree of worker threads, and each one will clean up any
	* pending records before exiting.
	*/
	if (err != 0) {
	srt_arg->cancel = B_TRUE;
	while (!range->eos_marker) {
	range = get_next_range(&srt_arg->q, range);
	}
	}
	range_free(range);

	bqueue_destroy(&srt_arg->q);
	bqueue_destroy(&smt_arg->q);
	if (dspp->redactbook != NULL)
	bqueue_destroy(&rlt_arg->q);
	bqueue_destroy(&to_arg->q);
	bqueue_destroy(&from_arg->q);

	if (err == 0 && srt_arg->error != 0)
	err = srt_arg->error;

	if (err != 0)
	goto out;

	if (dsc.dsc_pending_op != PENDING_NONE)
	if (dump_record(&dsc, NULL, 0) != 0)
	err = SET_ERROR(EINTR);

	if (err != 0) {
	if (err == EINTR && dsc.dsc_err != 0)
	err = dsc.dsc_err;
	goto out;
	}

	/*
	* Send the DRR_END record if this is not a saved stream.
	* Otherwise, the omitted DRR_END record will signal to
	* the receive side that the stream is incomplete.
	*/
	if (!dspp->savedok) {
	memset(drr, 0, sizeof (dmu_replay_record_t));
	drr->drr_type = DRR_END;
	drr->drr_u.drr_end.drr_checksum = dsc.dsc_zc;
	drr->drr_u.drr_end.drr_toguid = dsc.dsc_toguid;

	if (dump_record(&dsc, NULL, 0) != 0)
	err = dsc.dsc_err;
	}
	out:
	mutex_enter(&to_ds->ds_sendstream_lock);
	list_remove(&to_ds->ds_sendstreams, dssp);
	mutex_exit(&to_ds->ds_sendstream_lock);

	VERIFY(err != 0 \|\| (dsc.dsc_sent_begin &&
	(dsc.dsc_sent_end \|\| dspp->savedok)));

	kmem_free(drr, sizeof (dmu_replay_record_t));
	kmem_free(dssp, sizeof (dmu_sendstatus_t));
	kmem_free(from_arg, sizeof (*from_arg));
	kmem_free(to_arg, sizeof (*to_arg));
	kmem_free(rlt_arg, sizeof (*rlt_arg));
	kmem_free(smt_arg, sizeof (*smt_arg));
	kmem_free(srt_arg, sizeof (*srt_arg));

	dsl_dataset_long_rele(to_ds, FTAG);
	if (from_rl != NULL) {
	dsl_redaction_list_long_rele(from_rl, FTAG);
	dsl_redaction_list_rele(from_rl, FTAG);
	}
	if (redact_rl != NULL) {
	dsl_redaction_list_long_rele(redact_rl, FTAG);
	dsl_redaction_list_rele(redact_rl, FTAG);
	}

	return (err);
	}

	int
	dmu_send_obj(const char *pool, uint64_t tosnap, uint64_t fromsnap,
	boolean_t embedok, boolean_t large_block_ok, boolean_t compressok,
	boolean_t rawok, boolean_t savedok, int outfd, offset_t *off,
	dmu_send_outparams_t *dsop)
	{
	int err;
	dsl_dataset_t *fromds;
	ds_hold_flags_t dsflags;
	struct dmu_send_params dspp = {0};
	dspp.embedok = embedok;
	dspp.large_block_ok = large_block_ok;
	dspp.compressok = compressok;
	dspp.outfd = outfd;
	dspp.off = off;
	dspp.dso = dsop;
	dspp.tag = FTAG;
	dspp.rawok = rawok;
	dspp.savedok = savedok;

	dsflags = (rawok) ? DS_HOLD_FLAG_NONE : DS_HOLD_FLAG_DECRYPT;
	err = dsl_pool_hold(pool, FTAG, &dspp.dp);
	if (err != 0)
	return (err);

	err = dsl_dataset_hold_obj_flags(dspp.dp, tosnap, dsflags, FTAG,
	&dspp.to_ds);
	if (err != 0) {
	dsl_pool_rele(dspp.dp, FTAG);
	return (err);
	}

	if (fromsnap != 0) {
	err = dsl_dataset_hold_obj_flags(dspp.dp, fromsnap, dsflags,
	FTAG, &fromds);
	if (err != 0) {
	dsl_dataset_rele_flags(dspp.to_ds, dsflags, FTAG);
	dsl_pool_rele(dspp.dp, FTAG);
	return (err);
	}
	dspp.ancestor_zb.zbm_guid = dsl_dataset_phys(fromds)->ds_guid;
	dspp.ancestor_zb.zbm_creation_txg =
	dsl_dataset_phys(fromds)->ds_creation_txg;
	dspp.ancestor_zb.zbm_creation_time =
	dsl_dataset_phys(fromds)->ds_creation_time;

	if (dsl_dataset_is_zapified(fromds)) {
	(void) zap_lookup(dspp.dp->dp_meta_objset,
	fromds->ds_object, DS_FIELD_IVSET_GUID, 8, 1,
	&dspp.ancestor_zb.zbm_ivset_guid);
	}

	/* See dmu_send for the reasons behind this. */
	uint64_t *fromredact;

	if (!dsl_dataset_get_uint64_array_feature(fromds,
	SPA_FEATURE_REDACTED_DATASETS,
	&dspp.numfromredactsnaps,
	&fromredact)) {
	dspp.numfromredactsnaps = NUM_SNAPS_NOT_REDACTED;
	} else if (dspp.numfromredactsnaps > 0) {
	uint64_t size = dspp.numfromredactsnaps *
	sizeof (uint64_t);
	dspp.fromredactsnaps = kmem_zalloc(size, KM_SLEEP);
	memcpy(dspp.fromredactsnaps, fromredact, size);
	}

	boolean_t is_before =
	dsl_dataset_is_before(dspp.to_ds, fromds, 0);
	dspp.is_clone = (dspp.to_ds->ds_dir !=
	fromds->ds_dir);
	dsl_dataset_rele(fromds, FTAG);
	if (!is_before) {
	dsl_pool_rele(dspp.dp, FTAG);
	err = SET_ERROR(EXDEV);
	} else {
	err = dmu_send_impl(&dspp);
	}
	} else {
	dspp.numfromredactsnaps = NUM_SNAPS_NOT_REDACTED;
	err = dmu_send_impl(&dspp);
	}
	if (dspp.fromredactsnaps)
	kmem_free(dspp.fromredactsnaps,
	dspp.numfromredactsnaps * sizeof (uint64_t));

	dsl_dataset_rele(dspp.to_ds, FTAG);
	return (err);
	}

	int
	dmu_send(const char tosnap, const char fromsnap, boolean_t embedok,
	boolean_t large_block_ok, boolean_t compressok, boolean_t rawok,
	boolean_t savedok, uint64_t resumeobj, uint64_t resumeoff,
	const char redactbook, int outfd, offset_t off,
	dmu_send_outparams_t *dsop)
	{
	int err = 0;
	ds_hold_flags_t dsflags;
	boolean_t owned = B_FALSE;
	dsl_dataset_t *fromds = NULL;
	zfs_bookmark_phys_t book = {0};
	struct dmu_send_params dspp = {0};

	dsflags = (rawok) ? DS_HOLD_FLAG_NONE : DS_HOLD_FLAG_DECRYPT;
	dspp.tosnap = tosnap;
	dspp.embedok = embedok;
	dspp.large_block_ok = large_block_ok;
	dspp.compressok = compressok;
	dspp.outfd = outfd;
	dspp.off = off;
	dspp.dso = dsop;
	dspp.tag = FTAG;
	dspp.resumeobj = resumeobj;
	dspp.resumeoff = resumeoff;
	dspp.rawok = rawok;
	dspp.savedok = savedok;

	if (fromsnap != NULL && strpbrk(fromsnap, "@#") == NULL)
	return (SET_ERROR(EINVAL));

	err = dsl_pool_hold(tosnap, FTAG, &dspp.dp);
	if (err != 0)
	return (err);

	if (strchr(tosnap, '@') == NULL && spa_writeable(dspp.dp->dp_spa)) {
	/*
	* We are sending a filesystem or volume. Ensure
	* that it doesn't change by owning the dataset.
	*/

	if (savedok) {
	/*
	* We are looking for the dataset that represents the
	* partially received send stream. If this stream was
	* received as a new snapshot of an existing dataset,
	* this will be saved in a hidden clone named
	* "<pool>/<dataset>/%recv". Otherwise, the stream
	* will be saved in the live dataset itself. In
	* either case we need to use dsl_dataset_own_force()
	* because the stream is marked as inconsistent,
	* which would normally make it unavailable to be
	* owned.
	*/
	char *name = kmem_asprintf("%s/%s", tosnap,
	recv_clone_name);
	err = dsl_dataset_own_force(dspp.dp, name, dsflags,
	FTAG, &dspp.to_ds);
	if (err == ENOENT) {
	err = dsl_dataset_own_force(dspp.dp, tosnap,
	dsflags, FTAG, &dspp.to_ds);
	}

	if (err == 0) {
	owned = B_TRUE;
	err = zap_lookup(dspp.dp->dp_meta_objset,
	dspp.to_ds->ds_object,
	DS_FIELD_RESUME_TOGUID, 8, 1,
	&dspp.saved_guid);
	}

	if (err == 0) {
	err = zap_lookup(dspp.dp->dp_meta_objset,
	dspp.to_ds->ds_object,
	DS_FIELD_RESUME_TONAME, 1,
	sizeof (dspp.saved_toname),
	dspp.saved_toname);
	}
	/* Only disown if there was an error in the lookups */
	if (owned && (err != 0))
	dsl_dataset_disown(dspp.to_ds, dsflags, FTAG);

	kmem_strfree(name);
	} else {
	err = dsl_dataset_own(dspp.dp, tosnap, dsflags,
	FTAG, &dspp.to_ds);
	if (err == 0)
	owned = B_TRUE;
	}
	} else {
	err = dsl_dataset_hold_flags(dspp.dp, tosnap, dsflags, FTAG,
	&dspp.to_ds);
	}

	if (err != 0) {
	/* Note: dsl dataset is not owned at this point */
	dsl_pool_rele(dspp.dp, FTAG);
	return (err);
	}

	if (redactbook != NULL) {
	char path[ZFS_MAX_DATASET_NAME_LEN];
	(void) strlcpy(path, tosnap, sizeof (path));
	char *at = strchr(path, '@');
	if (at == NULL) {
	err = EINVAL;
	} else {
	(void) snprintf(at, sizeof (path) - (at - path), "#%s",
	redactbook);
	err = dsl_bookmark_lookup(dspp.dp, path,
	NULL, &book);
	dspp.redactbook = &book;
	}
	}

	if (err != 0) {
	dsl_pool_rele(dspp.dp, FTAG);
	if (owned)
	dsl_dataset_disown(dspp.to_ds, dsflags, FTAG);
	else
	dsl_dataset_rele_flags(dspp.to_ds, dsflags, FTAG);
	return (err);
	}

	if (fromsnap != NULL) {
	zfs_bookmark_phys_t *zb = &dspp.ancestor_zb;
	int fsnamelen;
	if (strpbrk(tosnap, "@#") != NULL)
	fsnamelen = strpbrk(tosnap, "@#") - tosnap;
	else
	fsnamelen = strlen(tosnap);

	/*
	* If the fromsnap is in a different filesystem, then
	* mark the send stream as a clone.
	*/
	if (strncmp(tosnap, fromsnap, fsnamelen) != 0 \|\|
	(fromsnap[fsnamelen] != '@' &&
	fromsnap[fsnamelen] != '#')) {
	dspp.is_clone = B_TRUE;
	}

	if (strchr(fromsnap, '@') != NULL) {
	err = dsl_dataset_hold(dspp.dp, fromsnap, FTAG,
	&fromds);

	if (err != 0) {
	ASSERT3P(fromds, ==, NULL);
	} else {
	/*
	* We need to make a deep copy of the redact
	* snapshots of the from snapshot, because the
	* array will be freed when we evict from_ds.
	*/
	uint64_t *fromredact;
	if (!dsl_dataset_get_uint64_array_feature(
	fromds, SPA_FEATURE_REDACTED_DATASETS,
	&dspp.numfromredactsnaps,
	&fromredact)) {
	dspp.numfromredactsnaps =
	NUM_SNAPS_NOT_REDACTED;
	} else if (dspp.numfromredactsnaps > 0) {
	uint64_t size =
	dspp.numfromredactsnaps *
	sizeof (uint64_t);
	dspp.fromredactsnaps = kmem_zalloc(size,
	KM_SLEEP);
	memcpy(dspp.fromredactsnaps, fromredact,
	size);
	}
	if (!dsl_dataset_is_before(dspp.to_ds, fromds,
	0)) {
	err = SET_ERROR(EXDEV);
	} else {
	zb->zbm_creation_txg =
	dsl_dataset_phys(fromds)->
	ds_creation_txg;
	zb->zbm_creation_time =
	dsl_dataset_phys(fromds)->
	ds_creation_time;
	zb->zbm_guid =
	dsl_dataset_phys(fromds)->ds_guid;
	zb->zbm_redaction_obj = 0;

	if (dsl_dataset_is_zapified(fromds)) {
	(void) zap_lookup(
	dspp.dp->dp_meta_objset,
	fromds->ds_object,
	DS_FIELD_IVSET_GUID, 8, 1,
	&zb->zbm_ivset_guid);
	}
	}
	dsl_dataset_rele(fromds, FTAG);
	}
	} else {
	dspp.numfromredactsnaps = NUM_SNAPS_NOT_REDACTED;
	err = dsl_bookmark_lookup(dspp.dp, fromsnap, dspp.to_ds,
	zb);
	if (err == EXDEV && zb->zbm_redaction_obj != 0 &&
	zb->zbm_guid ==
	dsl_dataset_phys(dspp.to_ds)->ds_guid)
	err = 0;
	}

	if (err == 0) {
	/* dmu_send_impl will call dsl_pool_rele for us. */
	err = dmu_send_impl(&dspp);
	} else {
	if (dspp.fromredactsnaps)
	kmem_free(dspp.fromredactsnaps,
	dspp.numfromredactsnaps *
	sizeof (uint64_t));
	dsl_pool_rele(dspp.dp, FTAG);
	}
	} else {
	dspp.numfromredactsnaps = NUM_SNAPS_NOT_REDACTED;
	err = dmu_send_impl(&dspp);
	}
	if (owned)
	dsl_dataset_disown(dspp.to_ds, dsflags, FTAG);
	else
	dsl_dataset_rele_flags(dspp.to_ds, dsflags, FTAG);
	return (err);
	}

	static int
	dmu_adjust_send_estimate_for_indirects(dsl_dataset_t *ds, uint64_t uncompressed,
	uint64_t compressed, boolean_t stream_compressed, uint64_t *sizep)
	{
	int err = 0;
	uint64_t size;
	/*
	* Assume that space (both on-disk and in-stream) is dominated by
	* data. We will adjust for indirect blocks and the copies property,
	* but ignore per-object space used (eg, dnodes and DRR_OBJECT records).
	*/

	uint64_t recordsize;
	uint64_t record_count;
	objset_t *os;
	VERIFY0(dmu_objset_from_ds(ds, &os));

	/* Assume all (uncompressed) blocks are recordsize. */
	if (zfs_override_estimate_recordsize != 0) {
	recordsize = zfs_override_estimate_recordsize;
	} else if (os->os_phys->os_type == DMU_OST_ZVOL) {
	err = dsl_prop_get_int_ds(ds,
	zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), &recordsize);
	} else {
	err = dsl_prop_get_int_ds(ds,
	zfs_prop_to_name(ZFS_PROP_RECORDSIZE), &recordsize);
	}
	if (err != 0)
	return (err);
	record_count = uncompressed / recordsize;

	/*
	* If we're estimating a send size for a compressed stream, use the
	* compressed data size to estimate the stream size. Otherwise, use the
	* uncompressed data size.
	*/
	size = stream_compressed ? compressed : uncompressed;

	/*
	* Subtract out approximate space used by indirect blocks.
	* Assume most space is used by data blocks (non-indirect, non-dnode).
	* Assume no ditto blocks or internal fragmentation.
	*
	* Therefore, space used by indirect blocks is sizeof(blkptr_t) per
	* block.
	*/
	size -= record_count * sizeof (blkptr_t);

	/* Add in the space for the record associated with each block. */
	size += record_count * sizeof (dmu_replay_record_t);

	*sizep = size;

	return (0);
	}

	int
	dmu_send_estimate_fast(dsl_dataset_t origds, dsl_dataset_t fromds,
	zfs_bookmark_phys_t *frombook, boolean_t stream_compressed,
	boolean_t saved, uint64_t *sizep)
	{
	int err;
	dsl_dataset_t *ds = origds;
	uint64_t uncomp, comp;

	ASSERT(dsl_pool_config_held(origds->ds_dir->dd_pool));
	ASSERT(fromds == NULL \|\| frombook == NULL);

	/*
	* If this is a saved send we may actually be sending
	* from the %recv clone used for resuming.
	*/
	if (saved) {
	objset_t *mos = origds->ds_dir->dd_pool->dp_meta_objset;
	uint64_t guid;
	char dsname[ZFS_MAX_DATASET_NAME_LEN + 6];

	dsl_dataset_name(origds, dsname);
	(void) strcat(dsname, "/");
	(void) strlcat(dsname, recv_clone_name, sizeof (dsname));

	err = dsl_dataset_hold(origds->ds_dir->dd_pool,
	dsname, FTAG, &ds);
	if (err != ENOENT && err != 0) {
	return (err);
	} else if (err == ENOENT) {
	ds = origds;
	}

	/* check that this dataset has partially received data */
	err = zap_lookup(mos, ds->ds_object,
	DS_FIELD_RESUME_TOGUID, 8, 1, &guid);
	if (err != 0) {
	err = SET_ERROR(err == ENOENT ? EINVAL : err);
	goto out;
	}

	err = zap_lookup(mos, ds->ds_object,
	DS_FIELD_RESUME_TONAME, 1, sizeof (dsname), dsname);
	if (err != 0) {
	err = SET_ERROR(err == ENOENT ? EINVAL : err);
	goto out;
	}
	}

	/* tosnap must be a snapshot or the target of a saved send */
	if (!ds->ds_is_snapshot && ds == origds)
	return (SET_ERROR(EINVAL));

	if (fromds != NULL) {
	uint64_t used;
	if (!fromds->ds_is_snapshot) {
	err = SET_ERROR(EINVAL);
	goto out;
	}

	if (!dsl_dataset_is_before(ds, fromds, 0)) {
	err = SET_ERROR(EXDEV);
	goto out;
	}

	err = dsl_dataset_space_written(fromds, ds, &used, &comp,
	&uncomp);
	if (err != 0)
	goto out;
	} else if (frombook != NULL) {
	uint64_t used;
	err = dsl_dataset_space_written_bookmark(frombook, ds, &used,
	&comp, &uncomp);
	if (err != 0)
	goto out;
	} else {
	uncomp = dsl_dataset_phys(ds)->ds_uncompressed_bytes;
	comp = dsl_dataset_phys(ds)->ds_compressed_bytes;
	}

	err = dmu_adjust_send_estimate_for_indirects(ds, uncomp, comp,
	stream_compressed, sizep);
	/*
	* Add the size of the BEGIN and END records to the estimate.
	*/
	sizep += 2 sizeof (dmu_replay_record_t);

	out:
	if (ds != origds)
	dsl_dataset_rele(ds, FTAG);
	return (err);
	}

	ZFS_MODULE_PARAM(zfs_send, zfs_send_, corrupt_data, INT, ZMOD_RW,
	"Allow sending corrupt data");

	ZFS_MODULE_PARAM(zfs_send, zfs_send_, queue_length, UINT, ZMOD_RW,
	"Maximum send queue length");

	ZFS_MODULE_PARAM(zfs_send, zfs_send_, unmodified_spill_blocks, INT, ZMOD_RW,
	"Send unmodified spill blocks");

	ZFS_MODULE_PARAM(zfs_send, zfs_send_, no_prefetch_queue_length, UINT, ZMOD_RW,
	"Maximum send queue length for non-prefetch queues");

	ZFS_MODULE_PARAM(zfs_send, zfs_send_, queue_ff, UINT, ZMOD_RW,
	"Send queue fill fraction");

	ZFS_MODULE_PARAM(zfs_send, zfs_send_, no_prefetch_queue_ff, UINT, ZMOD_RW,
	"Send queue fill fraction for non-prefetch queues");

	ZFS_MODULE_PARAM(zfs_send, zfs_, override_estimate_recordsize, UINT, ZMOD_RW,
	"Override block size estimate with fixed size");
	diff --git a/sys/contrib/openzfs/module/zfs/metaslab.c b/sys/contrib/openzfs/module/zfs/metaslab.c
	index 9e762357b727..7170b5eefcea 100644
	--- a/sys/contrib/openzfs/module/zfs/metaslab.c
	+++ b/sys/contrib/openzfs/module/zfs/metaslab.c
	@@ -1,6284 +1,6287 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2019 by Delphix. All rights reserved.
	* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
	* Copyright (c) 2015, Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2017, Intel Corporation.
	*/

	#include <sys/zfs_context.h>
	#include <sys/dmu.h>
	#include <sys/dmu_tx.h>
	#include <sys/space_map.h>
	#include <sys/metaslab_impl.h>
	#include <sys/vdev_impl.h>
	#include <sys/vdev_draid.h>
	#include <sys/zio.h>
	#include <sys/spa_impl.h>
	#include <sys/zfeature.h>
	#include <sys/vdev_indirect_mapping.h>
	#include <sys/zap.h>
	#include <sys/btree.h>

	#define GANG_ALLOCATION(flags) \
	((flags) & (METASLAB_GANG_CHILD \| METASLAB_GANG_HEADER))

	/*
	* Metaslab granularity, in bytes. This is roughly similar to what would be
	* referred to as the "stripe size" in traditional RAID arrays. In normal
	* operation, we will try to write this amount of data to each disk before
	* moving on to the next top-level vdev.
	*/
	static uint64_t metaslab_aliquot = 1024 * 1024;

	/*
	* For testing, make some blocks above a certain size be gang blocks.
	*/
	uint64_t metaslab_force_ganging = SPA_MAXBLOCKSIZE + 1;

	/*
	* Of blocks of size >= metaslab_force_ganging, actually gang them this often.
	*/
	uint_t metaslab_force_ganging_pct = 3;

	/*
	* In pools where the log space map feature is not enabled we touch
	* multiple metaslabs (and their respective space maps) with each
	* transaction group. Thus, we benefit from having a small space map
	* block size since it allows us to issue more I/O operations scattered
	* around the disk. So a sane default for the space map block size
	* is 8~16K.
	*/
	int zfs_metaslab_sm_blksz_no_log = (1 << 14);

	/*
	* When the log space map feature is enabled, we accumulate a lot of
	* changes per metaslab that are flushed once in a while so we benefit
	* from a bigger block size like 128K for the metaslab space maps.
	*/
	int zfs_metaslab_sm_blksz_with_log = (1 << 17);

	/*
	* The in-core space map representation is more compact than its on-disk form.
	* The zfs_condense_pct determines how much more compact the in-core
	* space map representation must be before we compact it on-disk.
	* Values should be greater than or equal to 100.
	*/
	uint_t zfs_condense_pct = 200;

	/*
	* Condensing a metaslab is not guaranteed to actually reduce the amount of
	* space used on disk. In particular, a space map uses data in increments of
	* MAX(1 << ashift, space_map_blksz), so a metaslab might use the
	* same number of blocks after condensing. Since the goal of condensing is to
	* reduce the number of IOPs required to read the space map, we only want to
	* condense when we can be sure we will reduce the number of blocks used by the
	* space map. Unfortunately, we cannot precisely compute whether or not this is
	* the case in metaslab_should_condense since we are holding ms_lock. Instead,
	* we apply the following heuristic: do not condense a spacemap unless the
	* uncondensed size consumes greater than zfs_metaslab_condense_block_threshold
	* blocks.
	*/
	static const int zfs_metaslab_condense_block_threshold = 4;

	/*
	* The zfs_mg_noalloc_threshold defines which metaslab groups should
	* be eligible for allocation. The value is defined as a percentage of
	* free space. Metaslab groups that have more free space than
	* zfs_mg_noalloc_threshold are always eligible for allocations. Once
	* a metaslab group's free space is less than or equal to the
	* zfs_mg_noalloc_threshold the allocator will avoid allocating to that
	* group unless all groups in the pool have reached zfs_mg_noalloc_threshold.
	* Once all groups in the pool reach zfs_mg_noalloc_threshold then all
	* groups are allowed to accept allocations. Gang blocks are always
	* eligible to allocate on any metaslab group. The default value of 0 means
	* no metaslab group will be excluded based on this criterion.
	*/
	static uint_t zfs_mg_noalloc_threshold = 0;

	/*
	* Metaslab groups are considered eligible for allocations if their
	* fragmentation metric (measured as a percentage) is less than or
	* equal to zfs_mg_fragmentation_threshold. If a metaslab group
	* exceeds this threshold then it will be skipped unless all metaslab
	* groups within the metaslab class have also crossed this threshold.
	*
	* This tunable was introduced to avoid edge cases where we continue
	* allocating from very fragmented disks in our pool while other, less
	* fragmented disks, exists. On the other hand, if all disks in the
	* pool are uniformly approaching the threshold, the threshold can
	* be a speed bump in performance, where we keep switching the disks
	* that we allocate from (e.g. we allocate some segments from disk A
	* making it bypassing the threshold while freeing segments from disk
	* B getting its fragmentation below the threshold).
	*
	* Empirically, we've seen that our vdev selection for allocations is
	* good enough that fragmentation increases uniformly across all vdevs
	* the majority of the time. Thus we set the threshold percentage high
	* enough to avoid hitting the speed bump on pools that are being pushed
	* to the edge.
	*/
	static uint_t zfs_mg_fragmentation_threshold = 95;

	/*
	* Allow metaslabs to keep their active state as long as their fragmentation
	* percentage is less than or equal to zfs_metaslab_fragmentation_threshold. An
	* active metaslab that exceeds this threshold will no longer keep its active
	* status allowing better metaslabs to be selected.
	*/
	static uint_t zfs_metaslab_fragmentation_threshold = 70;

	/*
	* When set will load all metaslabs when pool is first opened.
	*/
	int metaslab_debug_load = B_FALSE;

	/*
	* When set will prevent metaslabs from being unloaded.
	*/
	static int metaslab_debug_unload = B_FALSE;

	/*
	* Minimum size which forces the dynamic allocator to change
	* it's allocation strategy. Once the space map cannot satisfy
	* an allocation of this size then it switches to using more
	* aggressive strategy (i.e search by size rather than offset).
	*/
	uint64_t metaslab_df_alloc_threshold = SPA_OLD_MAXBLOCKSIZE;

	/*
	* The minimum free space, in percent, which must be available
	* in a space map to continue allocations in a first-fit fashion.
	* Once the space map's free space drops below this level we dynamically
	* switch to using best-fit allocations.
	*/
	uint_t metaslab_df_free_pct = 4;

	/*
	* Maximum distance to search forward from the last offset. Without this
	* limit, fragmented pools can see >100,000 iterations and
	* metaslab_block_picker() becomes the performance limiting factor on
	* high-performance storage.
	*
	* With the default setting of 16MB, we typically see less than 500
	* iterations, even with very fragmented, ashift=9 pools. The maximum number
	* of iterations possible is:
	* metaslab_df_max_search / (2 * (1<<ashift))
	* With the default setting of 16MB this is 16*1024 (with ashift=9) or
	* 2048 (with ashift=12).
	*/
	static uint_t metaslab_df_max_search = 16 * 1024 * 1024;

	/*
	* Forces the metaslab_block_picker function to search for at least this many
	* segments forwards until giving up on finding a segment that the allocation
	* will fit into.
	*/
	static const uint32_t metaslab_min_search_count = 100;

	/*
	* If we are not searching forward (due to metaslab_df_max_search,
	* metaslab_df_free_pct, or metaslab_df_alloc_threshold), this tunable
	* controls what segment is used. If it is set, we will use the largest free
	* segment. If it is not set, we will use a segment of exactly the requested
	* size (or larger).
	*/
	static int metaslab_df_use_largest_segment = B_FALSE;

	/*
	* These tunables control how long a metaslab will remain loaded after the
	* last allocation from it. A metaslab can't be unloaded until at least
	* metaslab_unload_delay TXG's and metaslab_unload_delay_ms milliseconds
	* have elapsed. However, zfs_metaslab_mem_limit may cause it to be
	* unloaded sooner. These settings are intended to be generous -- to keep
	* metaslabs loaded for a long time, reducing the rate of metaslab loading.
	*/
	static uint_t metaslab_unload_delay = 32;
	static uint_t metaslab_unload_delay_ms = 10 * 60 * 1000; /* ten minutes */

	/*
	* Max number of metaslabs per group to preload.
	*/
	uint_t metaslab_preload_limit = 10;

	/*
	* Enable/disable preloading of metaslab.
	*/
	static int metaslab_preload_enabled = B_TRUE;

	/*
	* Enable/disable fragmentation weighting on metaslabs.
	*/
	static int metaslab_fragmentation_factor_enabled = B_TRUE;

	/*
	* Enable/disable lba weighting (i.e. outer tracks are given preference).
	*/
	static int metaslab_lba_weighting_enabled = B_TRUE;

	/*
	* Enable/disable metaslab group biasing.
	*/
	static int metaslab_bias_enabled = B_TRUE;

	/*
	* Enable/disable remapping of indirect DVAs to their concrete vdevs.
	*/
	static const boolean_t zfs_remap_blkptr_enable = B_TRUE;

	/*
	* Enable/disable segment-based metaslab selection.
	*/
	static int zfs_metaslab_segment_weight_enabled = B_TRUE;

	/*
	* When using segment-based metaslab selection, we will continue
	* allocating from the active metaslab until we have exhausted
	* zfs_metaslab_switch_threshold of its buckets.
	*/
	static int zfs_metaslab_switch_threshold = 2;

	/*
	* Internal switch to enable/disable the metaslab allocation tracing
	* facility.
	*/
	static const boolean_t metaslab_trace_enabled = B_FALSE;

	/*
	* Maximum entries that the metaslab allocation tracing facility will keep
	* in a given list when running in non-debug mode. We limit the number
	* of entries in non-debug mode to prevent us from using up too much memory.
	* The limit should be sufficiently large that we don't expect any allocation
	* to every exceed this value. In debug mode, the system will panic if this
	* limit is ever reached allowing for further investigation.
	*/
	static const uint64_t metaslab_trace_max_entries = 5000;

	/*
	* Maximum number of metaslabs per group that can be disabled
	* simultaneously.
	*/
	static const int max_disabled_ms = 3;

	/*
	* Time (in seconds) to respect ms_max_size when the metaslab is not loaded.
	* To avoid 64-bit overflow, don't set above UINT32_MAX.
	*/
	static uint64_t zfs_metaslab_max_size_cache_sec = 1 * 60 * 60; /* 1 hour */

	/*
	* Maximum percentage of memory to use on storing loaded metaslabs. If loading
	* a metaslab would take it over this percentage, the oldest selected metaslab
	* is automatically unloaded.
	*/
	static uint_t zfs_metaslab_mem_limit = 25;

	/*
	* Force the per-metaslab range trees to use 64-bit integers to store
	* segments. Used for debugging purposes.
	*/
	static const boolean_t zfs_metaslab_force_large_segs = B_FALSE;

	/*
	* By default we only store segments over a certain size in the size-sorted
	* metaslab trees (ms_allocatable_by_size and
	* ms_unflushed_frees_by_size). This dramatically reduces memory usage and
	* improves load and unload times at the cost of causing us to use slightly
	* larger segments than we would otherwise in some cases.
	*/
	static const uint32_t metaslab_by_size_min_shift = 14;

	/*
	* If not set, we will first try normal allocation. If that fails then
	* we will do a gang allocation. If that fails then we will do a "try hard"
	* gang allocation. If that fails then we will have a multi-layer gang
	* block.
	*
	* If set, we will first try normal allocation. If that fails then
	* we will do a "try hard" allocation. If that fails we will do a gang
	* allocation. If that fails we will do a "try hard" gang allocation. If
	* that fails then we will have a multi-layer gang block.
	*/
	static int zfs_metaslab_try_hard_before_gang = B_FALSE;

	/*
	* When not trying hard, we only consider the best zfs_metaslab_find_max_tries
	* metaslabs. This improves performance, especially when there are many
	* metaslabs per vdev and the allocation can't actually be satisfied (so we
	* would otherwise iterate all the metaslabs). If there is a metaslab with a
	* worse weight but it can actually satisfy the allocation, we won't find it
	* until trying hard. This may happen if the worse metaslab is not loaded
	* (and the true weight is better than we have calculated), or due to weight
	* bucketization. E.g. we are looking for a 60K segment, and the best
	* metaslabs all have free segments in the 32-63K bucket, but the best
	* zfs_metaslab_find_max_tries metaslabs have ms_max_size <60KB, and a
	* subsequent metaslab has ms_max_size >60KB (but fewer segments in this
	* bucket, and therefore a lower weight).
	*/
	static uint_t zfs_metaslab_find_max_tries = 100;

	static uint64_t metaslab_weight(metaslab_t *, boolean_t);
	static void metaslab_set_fragmentation(metaslab_t *, boolean_t);
	static void metaslab_free_impl(vdev_t *, uint64_t, uint64_t, boolean_t);
	static void metaslab_check_free_impl(vdev_t *, uint64_t, uint64_t);

	static void metaslab_passivate(metaslab_t *msp, uint64_t weight);
	static uint64_t metaslab_weight_from_range_tree(metaslab_t *msp);
	static void metaslab_flush_update(metaslab_t , dmu_tx_t );
	static unsigned int metaslab_idx_func(multilist_t , void );
	static void metaslab_evict(metaslab_t *, uint64_t);
	static void metaslab_rt_add(range_tree_t rt, range_seg_t rs, void *arg);
	kmem_cache_t *metaslab_alloc_trace_cache;

	typedef struct metaslab_stats {
	kstat_named_t metaslabstat_trace_over_limit;
	kstat_named_t metaslabstat_reload_tree;
	kstat_named_t metaslabstat_too_many_tries;
	kstat_named_t metaslabstat_try_hard;
	} metaslab_stats_t;

	static metaslab_stats_t metaslab_stats = {
	{ "trace_over_limit", KSTAT_DATA_UINT64 },
	{ "reload_tree", KSTAT_DATA_UINT64 },
	{ "too_many_tries", KSTAT_DATA_UINT64 },
	{ "try_hard", KSTAT_DATA_UINT64 },
	};

	#define METASLABSTAT_BUMP(stat) \
	atomic_inc_64(&metaslab_stats.stat.value.ui64);


	static kstat_t *metaslab_ksp;

	void
	metaslab_stat_init(void)
	{
	ASSERT(metaslab_alloc_trace_cache == NULL);
	metaslab_alloc_trace_cache = kmem_cache_create(
	"metaslab_alloc_trace_cache", sizeof (metaslab_alloc_trace_t),
	0, NULL, NULL, NULL, NULL, NULL, 0);
	metaslab_ksp = kstat_create("zfs", 0, "metaslab_stats",
	"misc", KSTAT_TYPE_NAMED, sizeof (metaslab_stats) /
	sizeof (kstat_named_t), KSTAT_FLAG_VIRTUAL);
	if (metaslab_ksp != NULL) {
	metaslab_ksp->ks_data = &metaslab_stats;
	kstat_install(metaslab_ksp);
	}
	}

	void
	metaslab_stat_fini(void)
	{
	if (metaslab_ksp != NULL) {
	kstat_delete(metaslab_ksp);
	metaslab_ksp = NULL;
	}

	kmem_cache_destroy(metaslab_alloc_trace_cache);
	metaslab_alloc_trace_cache = NULL;
	}

	/*
	* ==========================================================================
	* Metaslab classes
	* ==========================================================================
	*/
	metaslab_class_t *
	metaslab_class_create(spa_t spa, const metaslab_ops_t ops)
	{
	metaslab_class_t *mc;

	mc = kmem_zalloc(offsetof(metaslab_class_t,
	mc_allocator[spa->spa_alloc_count]), KM_SLEEP);

	mc->mc_spa = spa;
	mc->mc_ops = ops;
	mutex_init(&mc->mc_lock, NULL, MUTEX_DEFAULT, NULL);
	multilist_create(&mc->mc_metaslab_txg_list, sizeof (metaslab_t),
	offsetof(metaslab_t, ms_class_txg_node), metaslab_idx_func);
	for (int i = 0; i < spa->spa_alloc_count; i++) {
	metaslab_class_allocator_t *mca = &mc->mc_allocator[i];
	mca->mca_rotor = NULL;
	zfs_refcount_create_tracked(&mca->mca_alloc_slots);
	}

	return (mc);
	}

	void
	metaslab_class_destroy(metaslab_class_t *mc)
	{
	spa_t *spa = mc->mc_spa;

	ASSERT(mc->mc_alloc == 0);
	ASSERT(mc->mc_deferred == 0);
	ASSERT(mc->mc_space == 0);
	ASSERT(mc->mc_dspace == 0);

	for (int i = 0; i < spa->spa_alloc_count; i++) {
	metaslab_class_allocator_t *mca = &mc->mc_allocator[i];
	ASSERT(mca->mca_rotor == NULL);
	zfs_refcount_destroy(&mca->mca_alloc_slots);
	}
	mutex_destroy(&mc->mc_lock);
	multilist_destroy(&mc->mc_metaslab_txg_list);
	kmem_free(mc, offsetof(metaslab_class_t,
	mc_allocator[spa->spa_alloc_count]));
	}

	int
	metaslab_class_validate(metaslab_class_t *mc)
	{
	metaslab_group_t *mg;
	vdev_t *vd;

	/*
	* Must hold one of the spa_config locks.
	*/
	ASSERT(spa_config_held(mc->mc_spa, SCL_ALL, RW_READER) \|\|
	spa_config_held(mc->mc_spa, SCL_ALL, RW_WRITER));

	if ((mg = mc->mc_allocator[0].mca_rotor) == NULL)
	return (0);

	do {
	vd = mg->mg_vd;
	ASSERT(vd->vdev_mg != NULL);
	ASSERT3P(vd->vdev_top, ==, vd);
	ASSERT3P(mg->mg_class, ==, mc);
	ASSERT3P(vd->vdev_ops, !=, &vdev_hole_ops);
	} while ((mg = mg->mg_next) != mc->mc_allocator[0].mca_rotor);

	return (0);
	}

	static void
	metaslab_class_space_update(metaslab_class_t *mc, int64_t alloc_delta,
	int64_t defer_delta, int64_t space_delta, int64_t dspace_delta)
	{
	atomic_add_64(&mc->mc_alloc, alloc_delta);
	atomic_add_64(&mc->mc_deferred, defer_delta);
	atomic_add_64(&mc->mc_space, space_delta);
	atomic_add_64(&mc->mc_dspace, dspace_delta);
	}

	uint64_t
	metaslab_class_get_alloc(metaslab_class_t *mc)
	{
	return (mc->mc_alloc);
	}

	uint64_t
	metaslab_class_get_deferred(metaslab_class_t *mc)
	{
	return (mc->mc_deferred);
	}

	uint64_t
	metaslab_class_get_space(metaslab_class_t *mc)
	{
	return (mc->mc_space);
	}

	uint64_t
	metaslab_class_get_dspace(metaslab_class_t *mc)
	{
	return (spa_deflate(mc->mc_spa) ? mc->mc_dspace : mc->mc_space);
	}

	void
	metaslab_class_histogram_verify(metaslab_class_t *mc)
	{
	spa_t *spa = mc->mc_spa;
	vdev_t *rvd = spa->spa_root_vdev;
	uint64_t *mc_hist;
	int i;

	if ((zfs_flags & ZFS_DEBUG_HISTOGRAM_VERIFY) == 0)
	return;

	mc_hist = kmem_zalloc(sizeof (uint64_t) * RANGE_TREE_HISTOGRAM_SIZE,
	KM_SLEEP);

	mutex_enter(&mc->mc_lock);
	for (int c = 0; c < rvd->vdev_children; c++) {
	vdev_t *tvd = rvd->vdev_child[c];
	metaslab_group_t *mg = vdev_get_mg(tvd, mc);

	/*
	* Skip any holes, uninitialized top-levels, or
	* vdevs that are not in this metalab class.
	*/
	if (!vdev_is_concrete(tvd) \|\| tvd->vdev_ms_shift == 0 \|\|
	mg->mg_class != mc) {
	continue;
	}

	IMPLY(mg == mg->mg_vd->vdev_log_mg,
	mc == spa_embedded_log_class(mg->mg_vd->vdev_spa));

	for (i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i++)
	mc_hist[i] += mg->mg_histogram[i];
	}

	for (i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i++) {
	VERIFY3U(mc_hist[i], ==, mc->mc_histogram[i]);
	}

	mutex_exit(&mc->mc_lock);
	kmem_free(mc_hist, sizeof (uint64_t) * RANGE_TREE_HISTOGRAM_SIZE);
	}

	/*
	* Calculate the metaslab class's fragmentation metric. The metric
	* is weighted based on the space contribution of each metaslab group.
	* The return value will be a number between 0 and 100 (inclusive), or
	* ZFS_FRAG_INVALID if the metric has not been set. See comment above the
	* zfs_frag_table for more information about the metric.
	*/
	uint64_t
	metaslab_class_fragmentation(metaslab_class_t *mc)
	{
	vdev_t *rvd = mc->mc_spa->spa_root_vdev;
	uint64_t fragmentation = 0;

	spa_config_enter(mc->mc_spa, SCL_VDEV, FTAG, RW_READER);

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

	/*
	* Skip any holes, uninitialized top-levels,
	* or vdevs that are not in this metalab class.
	*/
	if (!vdev_is_concrete(tvd) \|\| tvd->vdev_ms_shift == 0 \|\|
	mg->mg_class != mc) {
	continue;
	}

	/*
	* If a metaslab group does not contain a fragmentation
	* metric then just bail out.
	*/
	if (mg->mg_fragmentation == ZFS_FRAG_INVALID) {
	spa_config_exit(mc->mc_spa, SCL_VDEV, FTAG);
	return (ZFS_FRAG_INVALID);
	}

	/*
	* Determine how much this metaslab_group is contributing
	* to the overall pool fragmentation metric.
	*/
	fragmentation += mg->mg_fragmentation *
	metaslab_group_get_space(mg);
	}
	fragmentation /= metaslab_class_get_space(mc);

	ASSERT3U(fragmentation, <=, 100);
	spa_config_exit(mc->mc_spa, SCL_VDEV, FTAG);
	return (fragmentation);
	}

	/*
	* Calculate the amount of expandable space that is available in
	* this metaslab class. If a device is expanded then its expandable
	* space will be the amount of allocatable space that is currently not
	* part of this metaslab class.
	*/
	uint64_t
	metaslab_class_expandable_space(metaslab_class_t *mc)
	{
	vdev_t *rvd = mc->mc_spa->spa_root_vdev;
	uint64_t space = 0;

	spa_config_enter(mc->mc_spa, SCL_VDEV, FTAG, RW_READER);
	for (int c = 0; c < rvd->vdev_children; c++) {
	vdev_t *tvd = rvd->vdev_child[c];
	metaslab_group_t *mg = tvd->vdev_mg;

	if (!vdev_is_concrete(tvd) \|\| tvd->vdev_ms_shift == 0 \|\|
	mg->mg_class != mc) {
	continue;
	}

	/*
	* Calculate if we have enough space to add additional
	* metaslabs. We report the expandable space in terms
	* of the metaslab size since that's the unit of expansion.
	*/
	- space += P2ALIGN(tvd->vdev_max_asize - tvd->vdev_asize,
	- 1ULL << tvd->vdev_ms_shift);
	+ space += P2ALIGN_TYPED(tvd->vdev_max_asize - tvd->vdev_asize,
	+ 1ULL << tvd->vdev_ms_shift, uint64_t);
	}
	spa_config_exit(mc->mc_spa, SCL_VDEV, FTAG);
	return (space);
	}

	void
	metaslab_class_evict_old(metaslab_class_t *mc, uint64_t txg)
	{
	multilist_t *ml = &mc->mc_metaslab_txg_list;
	+ hrtime_t now = gethrtime();
	for (int i = 0; i < multilist_get_num_sublists(ml); i++) {
	multilist_sublist_t *mls = multilist_sublist_lock_idx(ml, i);
	metaslab_t *msp = multilist_sublist_head(mls);
	multilist_sublist_unlock(mls);
	while (msp != NULL) {
	mutex_enter(&msp->ms_lock);

	/*
	* If the metaslab has been removed from the list
	* (which could happen if we were at the memory limit
	* and it was evicted during this loop), then we can't
	* proceed and we should restart the sublist.
	*/
	if (!multilist_link_active(&msp->ms_class_txg_node)) {
	mutex_exit(&msp->ms_lock);
	i--;
	break;
	}
	mls = multilist_sublist_lock_idx(ml, i);
	metaslab_t *next_msp = multilist_sublist_next(mls, msp);
	multilist_sublist_unlock(mls);
	if (txg >
	msp->ms_selected_txg + metaslab_unload_delay &&
	- gethrtime() > msp->ms_selected_time +
	- (uint64_t)MSEC2NSEC(metaslab_unload_delay_ms)) {
	+ now > msp->ms_selected_time +
	+ MSEC2NSEC(metaslab_unload_delay_ms) &&
	+ (msp->ms_allocator == -1 \|\|
	+ !metaslab_preload_enabled)) {
	metaslab_evict(msp, txg);
	} else {
	/*
	* Once we've hit a metaslab selected too
	* recently to evict, we're done evicting for
	* now.
	*/
	mutex_exit(&msp->ms_lock);
	break;
	}
	mutex_exit(&msp->ms_lock);
	msp = next_msp;
	}
	}
	}

	static int
	metaslab_compare(const void x1, const void x2)
	{
	const metaslab_t m1 = (const metaslab_t )x1;
	const metaslab_t m2 = (const metaslab_t )x2;

	int sort1 = 0;
	int sort2 = 0;
	if (m1->ms_allocator != -1 && m1->ms_primary)
	sort1 = 1;
	else if (m1->ms_allocator != -1 && !m1->ms_primary)
	sort1 = 2;
	if (m2->ms_allocator != -1 && m2->ms_primary)
	sort2 = 1;
	else if (m2->ms_allocator != -1 && !m2->ms_primary)
	sort2 = 2;

	/*
	* Sort inactive metaslabs first, then primaries, then secondaries. When
	* selecting a metaslab to allocate from, an allocator first tries its
	* primary, then secondary active metaslab. If it doesn't have active
	* metaslabs, or can't allocate from them, it searches for an inactive
	* metaslab to activate. If it can't find a suitable one, it will steal
	* a primary or secondary metaslab from another allocator.
	*/
	if (sort1 < sort2)
	return (-1);
	if (sort1 > sort2)
	return (1);

	int cmp = TREE_CMP(m2->ms_weight, m1->ms_weight);
	if (likely(cmp))
	return (cmp);

	IMPLY(TREE_CMP(m1->ms_start, m2->ms_start) == 0, m1 == m2);

	return (TREE_CMP(m1->ms_start, m2->ms_start));
	}

	/*
	* ==========================================================================
	* Metaslab groups
	* ==========================================================================
	*/
	/*
	* Update the allocatable flag and the metaslab group's capacity.
	* The allocatable flag is set to true if the capacity is below
	* the zfs_mg_noalloc_threshold or has a fragmentation value that is
	* greater than zfs_mg_fragmentation_threshold. If a metaslab group
	* transitions from allocatable to non-allocatable or vice versa then the
	* metaslab group's class is updated to reflect the transition.
	*/
	static void
	metaslab_group_alloc_update(metaslab_group_t *mg)
	{
	vdev_t *vd = mg->mg_vd;
	metaslab_class_t *mc = mg->mg_class;
	vdev_stat_t *vs = &vd->vdev_stat;
	boolean_t was_allocatable;
	boolean_t was_initialized;

	ASSERT(vd == vd->vdev_top);
	ASSERT3U(spa_config_held(mc->mc_spa, SCL_ALLOC, RW_READER), ==,
	SCL_ALLOC);

	mutex_enter(&mg->mg_lock);
	was_allocatable = mg->mg_allocatable;
	was_initialized = mg->mg_initialized;

	mg->mg_free_capacity = ((vs->vs_space - vs->vs_alloc) * 100) /
	(vs->vs_space + 1);

	mutex_enter(&mc->mc_lock);

	/*
	* If the metaslab group was just added then it won't
	* have any space until we finish syncing out this txg.
	* At that point we will consider it initialized and available
	* for allocations. We also don't consider non-activated
	* metaslab groups (e.g. vdevs that are in the middle of being removed)
	* to be initialized, because they can't be used for allocation.
	*/
	mg->mg_initialized = metaslab_group_initialized(mg);
	if (!was_initialized && mg->mg_initialized) {
	mc->mc_groups++;
	} else if (was_initialized && !mg->mg_initialized) {
	ASSERT3U(mc->mc_groups, >, 0);
	mc->mc_groups--;
	}
	if (mg->mg_initialized)
	mg->mg_no_free_space = B_FALSE;

	/*
	* A metaslab group is considered allocatable if it has plenty
	* of free space or is not heavily fragmented. We only take
	* fragmentation into account if the metaslab group has a valid
	* fragmentation metric (i.e. a value between 0 and 100).
	*/
	mg->mg_allocatable = (mg->mg_activation_count > 0 &&
	mg->mg_free_capacity > zfs_mg_noalloc_threshold &&
	(mg->mg_fragmentation == ZFS_FRAG_INVALID \|\|
	mg->mg_fragmentation <= zfs_mg_fragmentation_threshold));

	/*
	* The mc_alloc_groups maintains a count of the number of
	* groups in this metaslab class that are still above the
	* zfs_mg_noalloc_threshold. This is used by the allocating
	* threads to determine if they should avoid allocations to
	* a given group. The allocator will avoid allocations to a group
	* if that group has reached or is below the zfs_mg_noalloc_threshold
	* and there are still other groups that are above the threshold.
	* When a group transitions from allocatable to non-allocatable or
	* vice versa we update the metaslab class to reflect that change.
	* When the mc_alloc_groups value drops to 0 that means that all
	* groups have reached the zfs_mg_noalloc_threshold making all groups
	* eligible for allocations. This effectively means that all devices
	* are balanced again.
	*/
	if (was_allocatable && !mg->mg_allocatable)
	mc->mc_alloc_groups--;
	else if (!was_allocatable && mg->mg_allocatable)
	mc->mc_alloc_groups++;
	mutex_exit(&mc->mc_lock);

	mutex_exit(&mg->mg_lock);
	}

	int
	metaslab_sort_by_flushed(const void va, const void vb)
	{
	const metaslab_t *a = va;
	const metaslab_t *b = vb;

	int cmp = TREE_CMP(a->ms_unflushed_txg, b->ms_unflushed_txg);
	if (likely(cmp))
	return (cmp);

	uint64_t a_vdev_id = a->ms_group->mg_vd->vdev_id;
	uint64_t b_vdev_id = b->ms_group->mg_vd->vdev_id;
	cmp = TREE_CMP(a_vdev_id, b_vdev_id);
	if (cmp)
	return (cmp);

	return (TREE_CMP(a->ms_id, b->ms_id));
	}

	metaslab_group_t *
	metaslab_group_create(metaslab_class_t mc, vdev_t vd, int allocators)
	{
	metaslab_group_t *mg;

	mg = kmem_zalloc(offsetof(metaslab_group_t,
	mg_allocator[allocators]), KM_SLEEP);
	mutex_init(&mg->mg_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&mg->mg_ms_disabled_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&mg->mg_ms_disabled_cv, NULL, CV_DEFAULT, NULL);
	avl_create(&mg->mg_metaslab_tree, metaslab_compare,
	sizeof (metaslab_t), offsetof(metaslab_t, ms_group_node));
	mg->mg_vd = vd;
	mg->mg_class = mc;
	mg->mg_activation_count = 0;
	mg->mg_initialized = B_FALSE;
	mg->mg_no_free_space = B_TRUE;
	mg->mg_allocators = allocators;

	for (int i = 0; i < allocators; i++) {
	metaslab_group_allocator_t *mga = &mg->mg_allocator[i];
	zfs_refcount_create_tracked(&mga->mga_alloc_queue_depth);
	}

	return (mg);
	}

	void
	metaslab_group_destroy(metaslab_group_t *mg)
	{
	ASSERT(mg->mg_prev == NULL);
	ASSERT(mg->mg_next == NULL);
	/*
	* We may have gone below zero with the activation count
	* either because we never activated in the first place or
	* because we're done, and possibly removing the vdev.
	*/
	ASSERT(mg->mg_activation_count <= 0);

	avl_destroy(&mg->mg_metaslab_tree);
	mutex_destroy(&mg->mg_lock);
	mutex_destroy(&mg->mg_ms_disabled_lock);
	cv_destroy(&mg->mg_ms_disabled_cv);

	for (int i = 0; i < mg->mg_allocators; i++) {
	metaslab_group_allocator_t *mga = &mg->mg_allocator[i];
	zfs_refcount_destroy(&mga->mga_alloc_queue_depth);
	}
	kmem_free(mg, offsetof(metaslab_group_t,
	mg_allocator[mg->mg_allocators]));
	}

	void
	metaslab_group_activate(metaslab_group_t *mg)
	{
	metaslab_class_t *mc = mg->mg_class;
	spa_t *spa = mc->mc_spa;
	metaslab_group_t mgprev, mgnext;

	ASSERT3U(spa_config_held(spa, SCL_ALLOC, RW_WRITER), !=, 0);

	ASSERT(mg->mg_prev == NULL);
	ASSERT(mg->mg_next == NULL);
	ASSERT(mg->mg_activation_count <= 0);

	if (++mg->mg_activation_count <= 0)
	return;

	mg->mg_aliquot = metaslab_aliquot * MAX(1,
	vdev_get_ndisks(mg->mg_vd) - vdev_get_nparity(mg->mg_vd));
	metaslab_group_alloc_update(mg);

	if ((mgprev = mc->mc_allocator[0].mca_rotor) == NULL) {
	mg->mg_prev = mg;
	mg->mg_next = mg;
	} else {
	mgnext = mgprev->mg_next;
	mg->mg_prev = mgprev;
	mg->mg_next = mgnext;
	mgprev->mg_next = mg;
	mgnext->mg_prev = mg;
	}
	for (int i = 0; i < spa->spa_alloc_count; i++) {
	mc->mc_allocator[i].mca_rotor = mg;
	mg = mg->mg_next;
	}
	}

	/*
	* Passivate a metaslab group and remove it from the allocation rotor.
	* Callers must hold both the SCL_ALLOC and SCL_ZIO lock prior to passivating
	* a metaslab group. This function will momentarily drop spa_config_locks
	* that are lower than the SCL_ALLOC lock (see comment below).
	*/
	void
	metaslab_group_passivate(metaslab_group_t *mg)
	{
	metaslab_class_t *mc = mg->mg_class;
	spa_t *spa = mc->mc_spa;
	metaslab_group_t mgprev, mgnext;
	int locks = spa_config_held(spa, SCL_ALL, RW_WRITER);

	ASSERT3U(spa_config_held(spa, SCL_ALLOC \| SCL_ZIO, RW_WRITER), ==,
	(SCL_ALLOC \| SCL_ZIO));

	if (--mg->mg_activation_count != 0) {
	for (int i = 0; i < spa->spa_alloc_count; i++)
	ASSERT(mc->mc_allocator[i].mca_rotor != mg);
	ASSERT(mg->mg_prev == NULL);
	ASSERT(mg->mg_next == NULL);
	ASSERT(mg->mg_activation_count < 0);
	return;
	}

	/*
	* The spa_config_lock is an array of rwlocks, ordered as
	* follows (from highest to lowest):
	* SCL_CONFIG > SCL_STATE > SCL_L2ARC > SCL_ALLOC >
	* SCL_ZIO > SCL_FREE > SCL_VDEV
	* (For more information about the spa_config_lock see spa_misc.c)
	* The higher the lock, the broader its coverage. When we passivate
	* a metaslab group, we must hold both the SCL_ALLOC and the SCL_ZIO
	* config locks. However, the metaslab group's taskq might be trying
	* to preload metaslabs so we must drop the SCL_ZIO lock and any
	* lower locks to allow the I/O to complete. At a minimum,
	* we continue to hold the SCL_ALLOC lock, which prevents any future
	* allocations from taking place and any changes to the vdev tree.
	*/
	spa_config_exit(spa, locks & ~(SCL_ZIO - 1), spa);
	taskq_wait_outstanding(spa->spa_metaslab_taskq, 0);
	spa_config_enter(spa, locks & ~(SCL_ZIO - 1), spa, RW_WRITER);
	metaslab_group_alloc_update(mg);
	for (int i = 0; i < mg->mg_allocators; i++) {
	metaslab_group_allocator_t *mga = &mg->mg_allocator[i];
	metaslab_t *msp = mga->mga_primary;
	if (msp != NULL) {
	mutex_enter(&msp->ms_lock);
	metaslab_passivate(msp,
	metaslab_weight_from_range_tree(msp));
	mutex_exit(&msp->ms_lock);
	}
	msp = mga->mga_secondary;
	if (msp != NULL) {
	mutex_enter(&msp->ms_lock);
	metaslab_passivate(msp,
	metaslab_weight_from_range_tree(msp));
	mutex_exit(&msp->ms_lock);
	}
	}

	mgprev = mg->mg_prev;
	mgnext = mg->mg_next;

	if (mg == mgnext) {
	mgnext = NULL;
	} else {
	mgprev->mg_next = mgnext;
	mgnext->mg_prev = mgprev;
	}
	for (int i = 0; i < spa->spa_alloc_count; i++) {
	if (mc->mc_allocator[i].mca_rotor == mg)
	mc->mc_allocator[i].mca_rotor = mgnext;
	}

	mg->mg_prev = NULL;
	mg->mg_next = NULL;
	}

	boolean_t
	metaslab_group_initialized(metaslab_group_t *mg)
	{
	vdev_t *vd = mg->mg_vd;
	vdev_stat_t *vs = &vd->vdev_stat;

	return (vs->vs_space != 0 && mg->mg_activation_count > 0);
	}

	uint64_t
	metaslab_group_get_space(metaslab_group_t *mg)
	{
	/*
	* Note that the number of nodes in mg_metaslab_tree may be one less
	* than vdev_ms_count, due to the embedded log metaslab.
	*/
	mutex_enter(&mg->mg_lock);
	uint64_t ms_count = avl_numnodes(&mg->mg_metaslab_tree);
	mutex_exit(&mg->mg_lock);
	return ((1ULL << mg->mg_vd->vdev_ms_shift) * ms_count);
	}

	void
	metaslab_group_histogram_verify(metaslab_group_t *mg)
	{
	uint64_t *mg_hist;
	avl_tree_t *t = &mg->mg_metaslab_tree;
	uint64_t ashift = mg->mg_vd->vdev_ashift;

	if ((zfs_flags & ZFS_DEBUG_HISTOGRAM_VERIFY) == 0)
	return;

	mg_hist = kmem_zalloc(sizeof (uint64_t) * RANGE_TREE_HISTOGRAM_SIZE,
	KM_SLEEP);

	ASSERT3U(RANGE_TREE_HISTOGRAM_SIZE, >=,
	SPACE_MAP_HISTOGRAM_SIZE + ashift);

	mutex_enter(&mg->mg_lock);
	for (metaslab_t *msp = avl_first(t);
	msp != NULL; msp = AVL_NEXT(t, msp)) {
	VERIFY3P(msp->ms_group, ==, mg);
	/* skip if not active */
	if (msp->ms_sm == NULL)
	continue;

	for (int i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++) {
	mg_hist[i + ashift] +=
	msp->ms_sm->sm_phys->smp_histogram[i];
	}
	}

	for (int i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i ++)
	VERIFY3U(mg_hist[i], ==, mg->mg_histogram[i]);

	mutex_exit(&mg->mg_lock);

	kmem_free(mg_hist, sizeof (uint64_t) * RANGE_TREE_HISTOGRAM_SIZE);
	}

	static void
	metaslab_group_histogram_add(metaslab_group_t mg, metaslab_t msp)
	{
	metaslab_class_t *mc = mg->mg_class;
	uint64_t ashift = mg->mg_vd->vdev_ashift;

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	if (msp->ms_sm == NULL)
	return;

	mutex_enter(&mg->mg_lock);
	mutex_enter(&mc->mc_lock);
	for (int i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++) {
	IMPLY(mg == mg->mg_vd->vdev_log_mg,
	mc == spa_embedded_log_class(mg->mg_vd->vdev_spa));
	mg->mg_histogram[i + ashift] +=
	msp->ms_sm->sm_phys->smp_histogram[i];
	mc->mc_histogram[i + ashift] +=
	msp->ms_sm->sm_phys->smp_histogram[i];
	}
	mutex_exit(&mc->mc_lock);
	mutex_exit(&mg->mg_lock);
	}

	void
	metaslab_group_histogram_remove(metaslab_group_t mg, metaslab_t msp)
	{
	metaslab_class_t *mc = mg->mg_class;
	uint64_t ashift = mg->mg_vd->vdev_ashift;

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	if (msp->ms_sm == NULL)
	return;

	mutex_enter(&mg->mg_lock);
	mutex_enter(&mc->mc_lock);
	for (int i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++) {
	ASSERT3U(mg->mg_histogram[i + ashift], >=,
	msp->ms_sm->sm_phys->smp_histogram[i]);
	ASSERT3U(mc->mc_histogram[i + ashift], >=,
	msp->ms_sm->sm_phys->smp_histogram[i]);
	IMPLY(mg == mg->mg_vd->vdev_log_mg,
	mc == spa_embedded_log_class(mg->mg_vd->vdev_spa));

	mg->mg_histogram[i + ashift] -=
	msp->ms_sm->sm_phys->smp_histogram[i];
	mc->mc_histogram[i + ashift] -=
	msp->ms_sm->sm_phys->smp_histogram[i];
	}
	mutex_exit(&mc->mc_lock);
	mutex_exit(&mg->mg_lock);
	}

	static void
	metaslab_group_add(metaslab_group_t mg, metaslab_t msp)
	{
	ASSERT(msp->ms_group == NULL);
	mutex_enter(&mg->mg_lock);
	msp->ms_group = mg;
	msp->ms_weight = 0;
	avl_add(&mg->mg_metaslab_tree, msp);
	mutex_exit(&mg->mg_lock);

	mutex_enter(&msp->ms_lock);
	metaslab_group_histogram_add(mg, msp);
	mutex_exit(&msp->ms_lock);
	}

	static void
	metaslab_group_remove(metaslab_group_t mg, metaslab_t msp)
	{
	mutex_enter(&msp->ms_lock);
	metaslab_group_histogram_remove(mg, msp);
	mutex_exit(&msp->ms_lock);

	mutex_enter(&mg->mg_lock);
	ASSERT(msp->ms_group == mg);
	avl_remove(&mg->mg_metaslab_tree, msp);

	metaslab_class_t *mc = msp->ms_group->mg_class;
	multilist_sublist_t *mls =
	multilist_sublist_lock_obj(&mc->mc_metaslab_txg_list, msp);
	if (multilist_link_active(&msp->ms_class_txg_node))
	multilist_sublist_remove(mls, msp);
	multilist_sublist_unlock(mls);

	msp->ms_group = NULL;
	mutex_exit(&mg->mg_lock);
	}

	static void
	metaslab_group_sort_impl(metaslab_group_t mg, metaslab_t msp, uint64_t weight)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));
	ASSERT(MUTEX_HELD(&mg->mg_lock));
	ASSERT(msp->ms_group == mg);

	avl_remove(&mg->mg_metaslab_tree, msp);
	msp->ms_weight = weight;
	avl_add(&mg->mg_metaslab_tree, msp);

	}

	static void
	metaslab_group_sort(metaslab_group_t mg, metaslab_t msp, uint64_t weight)
	{
	/*
	* Although in principle the weight can be any value, in
	* practice we do not use values in the range [1, 511].
	*/
	ASSERT(weight >= SPA_MINBLOCKSIZE \|\| weight == 0);
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	mutex_enter(&mg->mg_lock);
	metaslab_group_sort_impl(mg, msp, weight);
	mutex_exit(&mg->mg_lock);
	}

	/*
	* Calculate the fragmentation for a given metaslab group. We can use
	* a simple average here since all metaslabs within the group must have
	* the same size. The return value will be a value between 0 and 100
	* (inclusive), or ZFS_FRAG_INVALID if less than half of the metaslab in this
	* group have a fragmentation metric.
	*/
	uint64_t
	metaslab_group_fragmentation(metaslab_group_t *mg)
	{
	vdev_t *vd = mg->mg_vd;
	uint64_t fragmentation = 0;
	uint64_t valid_ms = 0;

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

	if (msp->ms_fragmentation == ZFS_FRAG_INVALID)
	continue;
	if (msp->ms_group != mg)
	continue;

	valid_ms++;
	fragmentation += msp->ms_fragmentation;
	}

	if (valid_ms <= mg->mg_vd->vdev_ms_count / 2)
	return (ZFS_FRAG_INVALID);

	fragmentation /= valid_ms;
	ASSERT3U(fragmentation, <=, 100);
	return (fragmentation);
	}

	/*
	* Determine if a given metaslab group should skip allocations. A metaslab
	* group should avoid allocations if its free capacity is less than the
	* zfs_mg_noalloc_threshold or its fragmentation metric is greater than
	* zfs_mg_fragmentation_threshold and there is at least one metaslab group
	* that can still handle allocations. If the allocation throttle is enabled
	* then we skip allocations to devices that have reached their maximum
	* allocation queue depth unless the selected metaslab group is the only
	* eligible group remaining.
	*/
	static boolean_t
	metaslab_group_allocatable(metaslab_group_t mg, metaslab_group_t rotor,
	int flags, uint64_t psize, int allocator, int d)
	{
	spa_t *spa = mg->mg_vd->vdev_spa;
	metaslab_class_t *mc = mg->mg_class;

	/*
	* We can only consider skipping this metaslab group if it's
	* in the normal metaslab class and there are other metaslab
	* groups to select from. Otherwise, we always consider it eligible
	* for allocations.
	*/
	if ((mc != spa_normal_class(spa) &&
	mc != spa_special_class(spa) &&
	mc != spa_dedup_class(spa)) \|\|
	mc->mc_groups <= 1)
	return (B_TRUE);

	/*
	* If the metaslab group's mg_allocatable flag is set (see comments
	* in metaslab_group_alloc_update() for more information) and
	* the allocation throttle is disabled then allow allocations to this
	* device. However, if the allocation throttle is enabled then
	* check if we have reached our allocation limit (mga_alloc_queue_depth)
	* to determine if we should allow allocations to this metaslab group.
	* If all metaslab groups are no longer considered allocatable
	* (mc_alloc_groups == 0) or we're trying to allocate the smallest
	* gang block size then we allow allocations on this metaslab group
	* regardless of the mg_allocatable or throttle settings.
	*/
	if (mg->mg_allocatable) {
	metaslab_group_allocator_t *mga = &mg->mg_allocator[allocator];
	int64_t qdepth;
	uint64_t qmax = mga->mga_cur_max_alloc_queue_depth;

	if (!mc->mc_alloc_throttle_enabled)
	return (B_TRUE);

	/*
	* If this metaslab group does not have any free space, then
	* there is no point in looking further.
	*/
	if (mg->mg_no_free_space)
	return (B_FALSE);

	/*
	* Some allocations (e.g., those coming from device removal
	* where the * allocations are not even counted in the
	* metaslab * allocation queues) are allowed to bypass
	* the throttle.
	*/
	if (flags & METASLAB_DONT_THROTTLE)
	return (B_TRUE);

	/*
	* Relax allocation throttling for ditto blocks. Due to
	* random imbalances in allocation it tends to push copies
	* to one vdev, that looks a bit better at the moment.
	*/
	qmax = qmax * (4 + d) / 4;

	qdepth = zfs_refcount_count(&mga->mga_alloc_queue_depth);

	/*
	* If this metaslab group is below its qmax or it's
	* the only allocatable metaslab group, then attempt
	* to allocate from it.
	*/
	if (qdepth < qmax \|\| mc->mc_alloc_groups == 1)
	return (B_TRUE);
	ASSERT3U(mc->mc_alloc_groups, >, 1);

	/*
	* Since this metaslab group is at or over its qmax, we
	* need to determine if there are metaslab groups after this
	* one that might be able to handle this allocation. This is
	* racy since we can't hold the locks for all metaslab
	* groups at the same time when we make this check.
	*/
	for (metaslab_group_t *mgp = mg->mg_next;
	mgp != rotor; mgp = mgp->mg_next) {
	metaslab_group_allocator_t *mgap =
	&mgp->mg_allocator[allocator];
	qmax = mgap->mga_cur_max_alloc_queue_depth;
	qmax = qmax * (4 + d) / 4;
	qdepth =
	zfs_refcount_count(&mgap->mga_alloc_queue_depth);

	/*
	* If there is another metaslab group that
	* might be able to handle the allocation, then
	* we return false so that we skip this group.
	*/
	if (qdepth < qmax && !mgp->mg_no_free_space)
	return (B_FALSE);
	}

	/*
	* We didn't find another group to handle the allocation
	* so we can't skip this metaslab group even though
	* we are at or over our qmax.
	*/
	return (B_TRUE);

	} else if (mc->mc_alloc_groups == 0 \|\| psize == SPA_MINBLOCKSIZE) {
	return (B_TRUE);
	}
	return (B_FALSE);
	}

	/*
	* ==========================================================================
	* Range tree callbacks
	* ==========================================================================
	*/

	/*
	* Comparison function for the private size-ordered tree using 32-bit
	* ranges. Tree is sorted by size, larger sizes at the end of the tree.
	*/
	__attribute__((always_inline)) inline
	static int
	metaslab_rangesize32_compare(const void x1, const void x2)
	{
	const range_seg32_t *r1 = x1;
	const range_seg32_t *r2 = x2;

	uint64_t rs_size1 = r1->rs_end - r1->rs_start;
	uint64_t rs_size2 = r2->rs_end - r2->rs_start;

	int cmp = TREE_CMP(rs_size1, rs_size2);

	return (cmp + !cmp * TREE_CMP(r1->rs_start, r2->rs_start));
	}

	/*
	* Comparison function for the private size-ordered tree using 64-bit
	* ranges. Tree is sorted by size, larger sizes at the end of the tree.
	*/
	__attribute__((always_inline)) inline
	static int
	metaslab_rangesize64_compare(const void x1, const void x2)
	{
	const range_seg64_t *r1 = x1;
	const range_seg64_t *r2 = x2;

	uint64_t rs_size1 = r1->rs_end - r1->rs_start;
	uint64_t rs_size2 = r2->rs_end - r2->rs_start;

	int cmp = TREE_CMP(rs_size1, rs_size2);

	return (cmp + !cmp * TREE_CMP(r1->rs_start, r2->rs_start));
	}

	typedef struct metaslab_rt_arg {
	zfs_btree_t *mra_bt;
	uint32_t mra_floor_shift;
	} metaslab_rt_arg_t;

	struct mssa_arg {
	range_tree_t *rt;
	metaslab_rt_arg_t *mra;
	};

	static void
	metaslab_size_sorted_add(void *arg, uint64_t start, uint64_t size)
	{
	struct mssa_arg *mssap = arg;
	range_tree_t *rt = mssap->rt;
	metaslab_rt_arg_t *mrap = mssap->mra;
	range_seg_max_t seg = {0};
	rs_set_start(&seg, rt, start);
	rs_set_end(&seg, rt, start + size);
	metaslab_rt_add(rt, &seg, mrap);
	}

	static void
	metaslab_size_tree_full_load(range_tree_t *rt)
	{
	metaslab_rt_arg_t *mrap = rt->rt_arg;
	METASLABSTAT_BUMP(metaslabstat_reload_tree);
	ASSERT0(zfs_btree_numnodes(mrap->mra_bt));
	mrap->mra_floor_shift = 0;
	struct mssa_arg arg = {0};
	arg.rt = rt;
	arg.mra = mrap;
	range_tree_walk(rt, metaslab_size_sorted_add, &arg);
	}


	ZFS_BTREE_FIND_IN_BUF_FUNC(metaslab_rt_find_rangesize32_in_buf,
	range_seg32_t, metaslab_rangesize32_compare)

	ZFS_BTREE_FIND_IN_BUF_FUNC(metaslab_rt_find_rangesize64_in_buf,
	range_seg64_t, metaslab_rangesize64_compare)

	/*
	* Create any block allocator specific components. The current allocators
	* rely on using both a size-ordered range_tree_t and an array of uint64_t's.
	*/
	static void
	metaslab_rt_create(range_tree_t rt, void arg)
	{
	metaslab_rt_arg_t *mrap = arg;
	zfs_btree_t *size_tree = mrap->mra_bt;

	size_t size;
	int (compare) (const void , const void *);
	bt_find_in_buf_f bt_find;
	switch (rt->rt_type) {
	case RANGE_SEG32:
	size = sizeof (range_seg32_t);
	compare = metaslab_rangesize32_compare;
	bt_find = metaslab_rt_find_rangesize32_in_buf;
	break;
	case RANGE_SEG64:
	size = sizeof (range_seg64_t);
	compare = metaslab_rangesize64_compare;
	bt_find = metaslab_rt_find_rangesize64_in_buf;
	break;
	default:
	panic("Invalid range seg type %d", rt->rt_type);
	}
	zfs_btree_create(size_tree, compare, bt_find, size);
	mrap->mra_floor_shift = metaslab_by_size_min_shift;
	}

	static void
	metaslab_rt_destroy(range_tree_t rt, void arg)
	{
	(void) rt;
	metaslab_rt_arg_t *mrap = arg;
	zfs_btree_t *size_tree = mrap->mra_bt;

	zfs_btree_destroy(size_tree);
	kmem_free(mrap, sizeof (*mrap));
	}

	static void
	metaslab_rt_add(range_tree_t rt, range_seg_t rs, void *arg)
	{
	metaslab_rt_arg_t *mrap = arg;
	zfs_btree_t *size_tree = mrap->mra_bt;

	if (rs_get_end(rs, rt) - rs_get_start(rs, rt) <
	(1ULL << mrap->mra_floor_shift))
	return;

	zfs_btree_add(size_tree, rs);
	}

	static void
	metaslab_rt_remove(range_tree_t rt, range_seg_t rs, void *arg)
	{
	metaslab_rt_arg_t *mrap = arg;
	zfs_btree_t *size_tree = mrap->mra_bt;

	if (rs_get_end(rs, rt) - rs_get_start(rs, rt) < (1ULL <<
	mrap->mra_floor_shift))
	return;

	zfs_btree_remove(size_tree, rs);
	}

	static void
	metaslab_rt_vacate(range_tree_t rt, void arg)
	{
	metaslab_rt_arg_t *mrap = arg;
	zfs_btree_t *size_tree = mrap->mra_bt;
	zfs_btree_clear(size_tree);
	zfs_btree_destroy(size_tree);

	metaslab_rt_create(rt, arg);
	}

	static const range_tree_ops_t metaslab_rt_ops = {
	.rtop_create = metaslab_rt_create,
	.rtop_destroy = metaslab_rt_destroy,
	.rtop_add = metaslab_rt_add,
	.rtop_remove = metaslab_rt_remove,
	.rtop_vacate = metaslab_rt_vacate
	};

	/*
	* ==========================================================================
	* Common allocator routines
	* ==========================================================================
	*/

	/*
	* Return the maximum contiguous segment within the metaslab.
	*/
	uint64_t
	metaslab_largest_allocatable(metaslab_t *msp)
	{
	zfs_btree_t *t = &msp->ms_allocatable_by_size;
	range_seg_t *rs;

	if (t == NULL)
	return (0);
	if (zfs_btree_numnodes(t) == 0)
	metaslab_size_tree_full_load(msp->ms_allocatable);

	rs = zfs_btree_last(t, NULL);
	if (rs == NULL)
	return (0);

	return (rs_get_end(rs, msp->ms_allocatable) - rs_get_start(rs,
	msp->ms_allocatable));
	}

	/*
	* Return the maximum contiguous segment within the unflushed frees of this
	* metaslab.
	*/
	static uint64_t
	metaslab_largest_unflushed_free(metaslab_t *msp)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	if (msp->ms_unflushed_frees == NULL)
	return (0);

	if (zfs_btree_numnodes(&msp->ms_unflushed_frees_by_size) == 0)
	metaslab_size_tree_full_load(msp->ms_unflushed_frees);
	range_seg_t *rs = zfs_btree_last(&msp->ms_unflushed_frees_by_size,
	NULL);
	if (rs == NULL)
	return (0);

	/*
	* When a range is freed from the metaslab, that range is added to
	* both the unflushed frees and the deferred frees. While the block
	* will eventually be usable, if the metaslab were loaded the range
	* would not be added to the ms_allocatable tree until TXG_DEFER_SIZE
	* txgs had passed. As a result, when attempting to estimate an upper
	* bound for the largest currently-usable free segment in the
	* metaslab, we need to not consider any ranges currently in the defer
	* trees. This algorithm approximates the largest available chunk in
	* the largest range in the unflushed_frees tree by taking the first
	* chunk. While this may be a poor estimate, it should only remain so
	* briefly and should eventually self-correct as frees are no longer
	* deferred. Similar logic applies to the ms_freed tree. See
	* metaslab_load() for more details.
	*
	* There are two primary sources of inaccuracy in this estimate. Both
	* are tolerated for performance reasons. The first source is that we
	* only check the largest segment for overlaps. Smaller segments may
	* have more favorable overlaps with the other trees, resulting in
	* larger usable chunks. Second, we only look at the first chunk in
	* the largest segment; there may be other usable chunks in the
	* largest segment, but we ignore them.
	*/
	uint64_t rstart = rs_get_start(rs, msp->ms_unflushed_frees);
	uint64_t rsize = rs_get_end(rs, msp->ms_unflushed_frees) - rstart;
	for (int t = 0; t < TXG_DEFER_SIZE; t++) {
	uint64_t start = 0;
	uint64_t size = 0;
	boolean_t found = range_tree_find_in(msp->ms_defer[t], rstart,
	rsize, &start, &size);
	if (found) {
	if (rstart == start)
	return (0);
	rsize = start - rstart;
	}
	}

	uint64_t start = 0;
	uint64_t size = 0;
	boolean_t found = range_tree_find_in(msp->ms_freed, rstart,
	rsize, &start, &size);
	if (found)
	rsize = start - rstart;

	return (rsize);
	}

	static range_seg_t *
	metaslab_block_find(zfs_btree_t t, range_tree_t rt, uint64_t start,
	uint64_t size, zfs_btree_index_t *where)
	{
	range_seg_t *rs;
	range_seg_max_t rsearch;

	rs_set_start(&rsearch, rt, start);
	rs_set_end(&rsearch, rt, start + size);

	rs = zfs_btree_find(t, &rsearch, where);
	if (rs == NULL) {
	rs = zfs_btree_next(t, where, where);
	}

	return (rs);
	}

	/*
	* This is a helper function that can be used by the allocator to find a
	* suitable block to allocate. This will search the specified B-tree looking
	* for a block that matches the specified criteria.
	*/
	static uint64_t
	metaslab_block_picker(range_tree_t rt, uint64_t cursor, uint64_t size,
	uint64_t max_search)
	{
	if (*cursor == 0)
	*cursor = rt->rt_start;
	zfs_btree_t *bt = &rt->rt_root;
	zfs_btree_index_t where;
	range_seg_t rs = metaslab_block_find(bt, rt, cursor, size, &where);
	uint64_t first_found;
	int count_searched = 0;

	if (rs != NULL)
	first_found = rs_get_start(rs, rt);

	while (rs != NULL && (rs_get_start(rs, rt) - first_found <=
	max_search \|\| count_searched < metaslab_min_search_count)) {
	uint64_t offset = rs_get_start(rs, rt);
	if (offset + size <= rs_get_end(rs, rt)) {
	*cursor = offset + size;
	return (offset);
	}
	rs = zfs_btree_next(bt, &where, &where);
	count_searched++;
	}

	*cursor = 0;
	return (-1ULL);
	}

	static uint64_t metaslab_df_alloc(metaslab_t *msp, uint64_t size);
	static uint64_t metaslab_cf_alloc(metaslab_t *msp, uint64_t size);
	static uint64_t metaslab_ndf_alloc(metaslab_t *msp, uint64_t size);
	metaslab_ops_t metaslab_allocator(spa_t spa);

	static metaslab_ops_t metaslab_allocators[] = {
	{ "dynamic", metaslab_df_alloc },
	{ "cursor", metaslab_cf_alloc },
	{ "new-dynamic", metaslab_ndf_alloc },
	};

	static int
	spa_find_allocator_byname(const char *val)
	{
	int a = ARRAY_SIZE(metaslab_allocators) - 1;
	if (strcmp("new-dynamic", val) == 0)
	return (-1); /* remove when ndf is working */
	for (; a >= 0; a--) {
	if (strcmp(val, metaslab_allocators[a].msop_name) == 0)
	return (a);
	}
	return (-1);
	}

	void
	spa_set_allocator(spa_t spa, const char allocator)
	{
	int a = spa_find_allocator_byname(allocator);
	if (a < 0) a = 0;
	spa->spa_active_allocator = a;
	zfs_dbgmsg("spa allocator: %s\n", metaslab_allocators[a].msop_name);
	}

	int
	spa_get_allocator(spa_t *spa)
	{
	return (spa->spa_active_allocator);
	}

	#if defined(_KERNEL)
	int
	param_set_active_allocator_common(const char *val)
	{
	char *p;

	if (val == NULL)
	return (SET_ERROR(EINVAL));

	if ((p = strchr(val, '\n')) != NULL)
	*p = '\0';

	int a = spa_find_allocator_byname(val);
	if (a < 0)
	return (SET_ERROR(EINVAL));

	zfs_active_allocator = metaslab_allocators[a].msop_name;
	return (0);
	}
	#endif

	metaslab_ops_t *
	metaslab_allocator(spa_t *spa)
	{
	int allocator = spa_get_allocator(spa);
	return (&metaslab_allocators[allocator]);
	}

	/*
	* ==========================================================================
	* Dynamic Fit (df) block allocator
	*
	* Search for a free chunk of at least this size, starting from the last
	* offset (for this alignment of block) looking for up to
	* metaslab_df_max_search bytes (16MB). If a large enough free chunk is not
	* found within 16MB, then return a free chunk of exactly the requested size (or
	* larger).
	*
	* If it seems like searching from the last offset will be unproductive, skip
	* that and just return a free chunk of exactly the requested size (or larger).
	* This is based on metaslab_df_alloc_threshold and metaslab_df_free_pct. This
	* mechanism is probably not very useful and may be removed in the future.
	*
	* The behavior when not searching can be changed to return the largest free
	* chunk, instead of a free chunk of exactly the requested size, by setting
	* metaslab_df_use_largest_segment.
	* ==========================================================================
	*/
	static uint64_t
	metaslab_df_alloc(metaslab_t *msp, uint64_t size)
	{
	/*
	* Find the largest power of 2 block size that evenly divides the
	* requested size. This is used to try to allocate blocks with similar
	* alignment from the same area of the metaslab (i.e. same cursor
	* bucket) but it does not guarantee that other allocations sizes
	* may exist in the same region.
	*/
	uint64_t align = size & -size;
	uint64_t *cursor = &msp->ms_lbas[highbit64(align) - 1];
	range_tree_t *rt = msp->ms_allocatable;
	uint_t free_pct = range_tree_space(rt) * 100 / msp->ms_size;
	uint64_t offset;

	ASSERT(MUTEX_HELD(&msp->ms_lock));

	/*
	* If we're running low on space, find a segment based on size,
	* rather than iterating based on offset.
	*/
	if (metaslab_largest_allocatable(msp) < metaslab_df_alloc_threshold \|\|
	free_pct < metaslab_df_free_pct) {
	offset = -1;
	} else {
	offset = metaslab_block_picker(rt,
	cursor, size, metaslab_df_max_search);
	}

	if (offset == -1) {
	range_seg_t *rs;
	if (zfs_btree_numnodes(&msp->ms_allocatable_by_size) == 0)
	metaslab_size_tree_full_load(msp->ms_allocatable);

	if (metaslab_df_use_largest_segment) {
	/* use largest free segment */
	rs = zfs_btree_last(&msp->ms_allocatable_by_size, NULL);
	} else {
	zfs_btree_index_t where;
	/* use segment of this size, or next largest */
	rs = metaslab_block_find(&msp->ms_allocatable_by_size,
	rt, msp->ms_start, size, &where);
	}
	if (rs != NULL && rs_get_start(rs, rt) + size <= rs_get_end(rs,
	rt)) {
	offset = rs_get_start(rs, rt);
	*cursor = offset + size;
	}
	}

	return (offset);
	}

	/*
	* ==========================================================================
	* Cursor fit block allocator -
	* Select the largest region in the metaslab, set the cursor to the beginning
	* of the range and the cursor_end to the end of the range. As allocations
	* are made advance the cursor. Continue allocating from the cursor until
	* the range is exhausted and then find a new range.
	* ==========================================================================
	*/
	static uint64_t
	metaslab_cf_alloc(metaslab_t *msp, uint64_t size)
	{
	range_tree_t *rt = msp->ms_allocatable;
	zfs_btree_t *t = &msp->ms_allocatable_by_size;
	uint64_t *cursor = &msp->ms_lbas[0];
	uint64_t *cursor_end = &msp->ms_lbas[1];
	uint64_t offset = 0;

	ASSERT(MUTEX_HELD(&msp->ms_lock));

	ASSERT3U(cursor_end, >=, cursor);

	if ((cursor + size) > cursor_end) {
	range_seg_t *rs;

	if (zfs_btree_numnodes(t) == 0)
	metaslab_size_tree_full_load(msp->ms_allocatable);
	rs = zfs_btree_last(t, NULL);
	if (rs == NULL \|\| (rs_get_end(rs, rt) - rs_get_start(rs, rt)) <
	size)
	return (-1ULL);

	*cursor = rs_get_start(rs, rt);
	*cursor_end = rs_get_end(rs, rt);
	}

	offset = *cursor;
	*cursor += size;

	return (offset);
	}

	/*
	* ==========================================================================
	* New dynamic fit allocator -
	* Select a region that is large enough to allocate 2^metaslab_ndf_clump_shift
	* contiguous blocks. If no region is found then just use the largest segment
	* that remains.
	* ==========================================================================
	*/

	/*
	* Determines desired number of contiguous blocks (2^metaslab_ndf_clump_shift)
	* to request from the allocator.
	*/
	uint64_t metaslab_ndf_clump_shift = 4;

	static uint64_t
	metaslab_ndf_alloc(metaslab_t *msp, uint64_t size)
	{
	zfs_btree_t *t = &msp->ms_allocatable->rt_root;
	range_tree_t *rt = msp->ms_allocatable;
	zfs_btree_index_t where;
	range_seg_t *rs;
	range_seg_max_t rsearch;
	uint64_t hbit = highbit64(size);
	uint64_t *cursor = &msp->ms_lbas[hbit - 1];
	uint64_t max_size = metaslab_largest_allocatable(msp);

	ASSERT(MUTEX_HELD(&msp->ms_lock));

	if (max_size < size)
	return (-1ULL);

	rs_set_start(&rsearch, rt, *cursor);
	rs_set_end(&rsearch, rt, *cursor + size);

	rs = zfs_btree_find(t, &rsearch, &where);
	if (rs == NULL \|\| (rs_get_end(rs, rt) - rs_get_start(rs, rt)) < size) {
	t = &msp->ms_allocatable_by_size;

	rs_set_start(&rsearch, rt, 0);
	rs_set_end(&rsearch, rt, MIN(max_size, 1ULL << (hbit +
	metaslab_ndf_clump_shift)));

	rs = zfs_btree_find(t, &rsearch, &where);
	if (rs == NULL)
	rs = zfs_btree_next(t, &where, &where);
	ASSERT(rs != NULL);
	}

	if ((rs_get_end(rs, rt) - rs_get_start(rs, rt)) >= size) {
	*cursor = rs_get_start(rs, rt) + size;
	return (rs_get_start(rs, rt));
	}
	return (-1ULL);
	}

	/*
	* ==========================================================================
	* Metaslabs
	* ==========================================================================
	*/

	/*
	* Wait for any in-progress metaslab loads to complete.
	*/
	static void
	metaslab_load_wait(metaslab_t *msp)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	while (msp->ms_loading) {
	ASSERT(!msp->ms_loaded);
	cv_wait(&msp->ms_load_cv, &msp->ms_lock);
	}
	}

	/*
	* Wait for any in-progress flushing to complete.
	*/
	static void
	metaslab_flush_wait(metaslab_t *msp)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	while (msp->ms_flushing)
	cv_wait(&msp->ms_flush_cv, &msp->ms_lock);
	}

	static unsigned int
	metaslab_idx_func(multilist_t ml, void arg)
	{
	metaslab_t *msp = arg;

	/*
	* ms_id values are allocated sequentially, so full 64bit
	* division would be a waste of time, so limit it to 32 bits.
	*/
	return ((unsigned int)msp->ms_id % multilist_get_num_sublists(ml));
	}

	uint64_t
	metaslab_allocated_space(metaslab_t *msp)
	{
	return (msp->ms_allocated_space);
	}

	/*
	* Verify that the space accounting on disk matches the in-core range_trees.
	*/
	static void
	metaslab_verify_space(metaslab_t *msp, uint64_t txg)
	{
	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
	uint64_t allocating = 0;
	uint64_t sm_free_space, msp_free_space;

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	ASSERT(!msp->ms_condensing);

	if ((zfs_flags & ZFS_DEBUG_METASLAB_VERIFY) == 0)
	return;

	/*
	* We can only verify the metaslab space when we're called
	* from syncing context with a loaded metaslab that has an
	* allocated space map. Calling this in non-syncing context
	* does not provide a consistent view of the metaslab since
	* we're performing allocations in the future.
	*/
	if (txg != spa_syncing_txg(spa) \|\| msp->ms_sm == NULL \|\|
	!msp->ms_loaded)
	return;

	/*
	* Even though the smp_alloc field can get negative,
	* when it comes to a metaslab's space map, that should
	* never be the case.
	*/
	ASSERT3S(space_map_allocated(msp->ms_sm), >=, 0);

	ASSERT3U(space_map_allocated(msp->ms_sm), >=,
	range_tree_space(msp->ms_unflushed_frees));

	ASSERT3U(metaslab_allocated_space(msp), ==,
	space_map_allocated(msp->ms_sm) +
	range_tree_space(msp->ms_unflushed_allocs) -
	range_tree_space(msp->ms_unflushed_frees));

	sm_free_space = msp->ms_size - metaslab_allocated_space(msp);

	/*
	* Account for future allocations since we would have
	* already deducted that space from the ms_allocatable.
	*/
	for (int t = 0; t < TXG_CONCURRENT_STATES; t++) {
	allocating +=
	range_tree_space(msp->ms_allocating[(txg + t) & TXG_MASK]);
	}
	ASSERT3U(allocating + msp->ms_allocated_this_txg, ==,
	msp->ms_allocating_total);

	ASSERT3U(msp->ms_deferspace, ==,
	range_tree_space(msp->ms_defer[0]) +
	range_tree_space(msp->ms_defer[1]));

	msp_free_space = range_tree_space(msp->ms_allocatable) + allocating +
	msp->ms_deferspace + range_tree_space(msp->ms_freed);

	VERIFY3U(sm_free_space, ==, msp_free_space);
	}

	static void
	metaslab_aux_histograms_clear(metaslab_t *msp)
	{
	/*
	* Auxiliary histograms are only cleared when resetting them,
	* which can only happen while the metaslab is loaded.
	*/
	ASSERT(msp->ms_loaded);

	memset(msp->ms_synchist, 0, sizeof (msp->ms_synchist));
	for (int t = 0; t < TXG_DEFER_SIZE; t++)
	memset(msp->ms_deferhist[t], 0, sizeof (msp->ms_deferhist[t]));
	}

	static void
	metaslab_aux_histogram_add(uint64_t *histogram, uint64_t shift,
	range_tree_t *rt)
	{
	/*
	* This is modeled after space_map_histogram_add(), so refer to that
	* function for implementation details. We want this to work like
	* the space map histogram, and not the range tree histogram, as we
	* are essentially constructing a delta that will be later subtracted
	* from the space map histogram.
	*/
	int idx = 0;
	for (int i = shift; i < RANGE_TREE_HISTOGRAM_SIZE; i++) {
	ASSERT3U(i, >=, idx + shift);
	histogram[idx] += rt->rt_histogram[i] << (i - idx - shift);

	if (idx < SPACE_MAP_HISTOGRAM_SIZE - 1) {
	ASSERT3U(idx + shift, ==, i);
	idx++;
	ASSERT3U(idx, <, SPACE_MAP_HISTOGRAM_SIZE);
	}
	}
	}

	/*
	* Called at every sync pass that the metaslab gets synced.
	*
	* The reason is that we want our auxiliary histograms to be updated
	* wherever the metaslab's space map histogram is updated. This way
	* we stay consistent on which parts of the metaslab space map's
	* histogram are currently not available for allocations (e.g because
	* they are in the defer, freed, and freeing trees).
	*/
	static void
	metaslab_aux_histograms_update(metaslab_t *msp)
	{
	space_map_t *sm = msp->ms_sm;
	ASSERT(sm != NULL);

	/*
	* This is similar to the metaslab's space map histogram updates
	* that take place in metaslab_sync(). The only difference is that
	* we only care about segments that haven't made it into the
	* ms_allocatable tree yet.
	*/
	if (msp->ms_loaded) {
	metaslab_aux_histograms_clear(msp);

	metaslab_aux_histogram_add(msp->ms_synchist,
	sm->sm_shift, msp->ms_freed);

	for (int t = 0; t < TXG_DEFER_SIZE; t++) {
	metaslab_aux_histogram_add(msp->ms_deferhist[t],
	sm->sm_shift, msp->ms_defer[t]);
	}
	}

	metaslab_aux_histogram_add(msp->ms_synchist,
	sm->sm_shift, msp->ms_freeing);
	}

	/*
	* Called every time we are done syncing (writing to) the metaslab,
	* i.e. at the end of each sync pass.
	* [see the comment in metaslab_impl.h for ms_synchist, ms_deferhist]
	*/
	static void
	metaslab_aux_histograms_update_done(metaslab_t *msp, boolean_t defer_allowed)
	{
	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
	space_map_t *sm = msp->ms_sm;

	if (sm == NULL) {
	/*
	* We came here from metaslab_init() when creating/opening a
	* pool, looking at a metaslab that hasn't had any allocations
	* yet.
	*/
	return;
	}

	/*
	* This is similar to the actions that we take for the ms_freed
	* and ms_defer trees in metaslab_sync_done().
	*/
	uint64_t hist_index = spa_syncing_txg(spa) % TXG_DEFER_SIZE;
	if (defer_allowed) {
	memcpy(msp->ms_deferhist[hist_index], msp->ms_synchist,
	sizeof (msp->ms_synchist));
	} else {
	memset(msp->ms_deferhist[hist_index], 0,
	sizeof (msp->ms_deferhist[hist_index]));
	}
	memset(msp->ms_synchist, 0, sizeof (msp->ms_synchist));
	}

	/*
	* Ensure that the metaslab's weight and fragmentation are consistent
	* with the contents of the histogram (either the range tree's histogram
	* or the space map's depending whether the metaslab is loaded).
	*/
	static void
	metaslab_verify_weight_and_frag(metaslab_t *msp)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	if ((zfs_flags & ZFS_DEBUG_METASLAB_VERIFY) == 0)
	return;

	/*
	* We can end up here from vdev_remove_complete(), in which case we
	* cannot do these assertions because we hold spa config locks and
	* thus we are not allowed to read from the DMU.
	*
	* We check if the metaslab group has been removed and if that's
	* the case we return immediately as that would mean that we are
	* here from the aforementioned code path.
	*/
	if (msp->ms_group == NULL)
	return;

	/*
	* Devices being removed always return a weight of 0 and leave
	* fragmentation and ms_max_size as is - there is nothing for
	* us to verify here.
	*/
	vdev_t *vd = msp->ms_group->mg_vd;
	if (vd->vdev_removing)
	return;

	/*
	* If the metaslab is dirty it probably means that we've done
	* some allocations or frees that have changed our histograms
	* and thus the weight.
	*/
	for (int t = 0; t < TXG_SIZE; t++) {
	if (txg_list_member(&vd->vdev_ms_list, msp, t))
	return;
	}

	/*
	* This verification checks that our in-memory state is consistent
	* with what's on disk. If the pool is read-only then there aren't
	* any changes and we just have the initially-loaded state.
	*/
	if (!spa_writeable(msp->ms_group->mg_vd->vdev_spa))
	return;

	/* some extra verification for in-core tree if you can */
	if (msp->ms_loaded) {
	range_tree_stat_verify(msp->ms_allocatable);
	VERIFY(space_map_histogram_verify(msp->ms_sm,
	msp->ms_allocatable));
	}

	uint64_t weight = msp->ms_weight;
	uint64_t was_active = msp->ms_weight & METASLAB_ACTIVE_MASK;
	boolean_t space_based = WEIGHT_IS_SPACEBASED(msp->ms_weight);
	uint64_t frag = msp->ms_fragmentation;
	uint64_t max_segsize = msp->ms_max_size;

	msp->ms_weight = 0;
	msp->ms_fragmentation = 0;

	/*
	* This function is used for verification purposes and thus should
	* not introduce any side-effects/mutations on the system's state.
	*
	* Regardless of whether metaslab_weight() thinks this metaslab
	* should be active or not, we want to ensure that the actual weight
	* (and therefore the value of ms_weight) would be the same if it
	* was to be recalculated at this point.
	*
	* In addition we set the nodirty flag so metaslab_weight() does
	* not dirty the metaslab for future TXGs (e.g. when trying to
	* force condensing to upgrade the metaslab spacemaps).
	*/
	msp->ms_weight = metaslab_weight(msp, B_TRUE) \| was_active;

	VERIFY3U(max_segsize, ==, msp->ms_max_size);

	/*
	* If the weight type changed then there is no point in doing
	* verification. Revert fields to their original values.
	*/
	if ((space_based && !WEIGHT_IS_SPACEBASED(msp->ms_weight)) \|\|
	(!space_based && WEIGHT_IS_SPACEBASED(msp->ms_weight))) {
	msp->ms_fragmentation = frag;
	msp->ms_weight = weight;
	return;
	}

	VERIFY3U(msp->ms_fragmentation, ==, frag);
	VERIFY3U(msp->ms_weight, ==, weight);
	}

	/*
	* If we're over the zfs_metaslab_mem_limit, select the loaded metaslab from
	* this class that was used longest ago, and attempt to unload it. We don't
	* want to spend too much time in this loop to prevent performance
	* degradation, and we expect that most of the time this operation will
	* succeed. Between that and the normal unloading processing during txg sync,
	* we expect this to keep the metaslab memory usage under control.
	*/
	static void
	metaslab_potentially_evict(metaslab_class_t *mc)
	{
	#ifdef _KERNEL
	uint64_t allmem = arc_all_memory();
	uint64_t inuse = spl_kmem_cache_inuse(zfs_btree_leaf_cache);
	uint64_t size = spl_kmem_cache_entry_size(zfs_btree_leaf_cache);
	uint_t tries = 0;
	for (; allmem * zfs_metaslab_mem_limit / 100 < inuse * size &&
	tries < multilist_get_num_sublists(&mc->mc_metaslab_txg_list) * 2;
	tries++) {
	unsigned int idx = multilist_get_random_index(
	&mc->mc_metaslab_txg_list);
	multilist_sublist_t *mls =
	multilist_sublist_lock_idx(&mc->mc_metaslab_txg_list, idx);
	metaslab_t *msp = multilist_sublist_head(mls);
	multilist_sublist_unlock(mls);
	while (msp != NULL && allmem * zfs_metaslab_mem_limit / 100 <
	inuse * size) {
	VERIFY3P(mls, ==, multilist_sublist_lock_idx(
	&mc->mc_metaslab_txg_list, idx));
	ASSERT3U(idx, ==,
	metaslab_idx_func(&mc->mc_metaslab_txg_list, msp));

	if (!multilist_link_active(&msp->ms_class_txg_node)) {
	multilist_sublist_unlock(mls);
	break;
	}
	metaslab_t *next_msp = multilist_sublist_next(mls, msp);
	multilist_sublist_unlock(mls);
	/*
	* If the metaslab is currently loading there are two
	* cases. If it's the metaslab we're evicting, we
	* can't continue on or we'll panic when we attempt to
	* recursively lock the mutex. If it's another
	* metaslab that's loading, it can be safely skipped,
	* since we know it's very new and therefore not a
	* good eviction candidate. We check later once the
	* lock is held that the metaslab is fully loaded
	* before actually unloading it.
	*/
	if (msp->ms_loading) {
	msp = next_msp;
	inuse =
	spl_kmem_cache_inuse(zfs_btree_leaf_cache);
	continue;
	}
	/*
	* We can't unload metaslabs with no spacemap because
	* they're not ready to be unloaded yet. We can't
	* unload metaslabs with outstanding allocations
	* because doing so could cause the metaslab's weight
	* to decrease while it's unloaded, which violates an
	* invariant that we use to prevent unnecessary
	* loading. We also don't unload metaslabs that are
	* currently active because they are high-weight
	* metaslabs that are likely to be used in the near
	* future.
	*/
	mutex_enter(&msp->ms_lock);
	if (msp->ms_allocator == -1 && msp->ms_sm != NULL &&
	msp->ms_allocating_total == 0) {
	metaslab_unload(msp);
	}
	mutex_exit(&msp->ms_lock);
	msp = next_msp;
	inuse = spl_kmem_cache_inuse(zfs_btree_leaf_cache);
	}
	}
	#else
	(void) mc, (void) zfs_metaslab_mem_limit;
	#endif
	}

	static int
	metaslab_load_impl(metaslab_t *msp)
	{
	int error = 0;

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	ASSERT(msp->ms_loading);
	ASSERT(!msp->ms_condensing);

	/*
	* We temporarily drop the lock to unblock other operations while we
	* are reading the space map. Therefore, metaslab_sync() and
	* metaslab_sync_done() can run at the same time as we do.
	*
	* If we are using the log space maps, metaslab_sync() can't write to
	* the metaslab's space map while we are loading as we only write to
	* it when we are flushing the metaslab, and that can't happen while
	* we are loading it.
	*
	* If we are not using log space maps though, metaslab_sync() can
	* append to the space map while we are loading. Therefore we load
	* only entries that existed when we started the load. Additionally,
	* metaslab_sync_done() has to wait for the load to complete because
	* there are potential races like metaslab_load() loading parts of the
	* space map that are currently being appended by metaslab_sync(). If
	* we didn't, the ms_allocatable would have entries that
	* metaslab_sync_done() would try to re-add later.
	*
	* That's why before dropping the lock we remember the synced length
	* of the metaslab and read up to that point of the space map,
	* ignoring entries appended by metaslab_sync() that happen after we
	* drop the lock.
	*/
	uint64_t length = msp->ms_synced_length;
	mutex_exit(&msp->ms_lock);

	hrtime_t load_start = gethrtime();
	metaslab_rt_arg_t *mrap;
	if (msp->ms_allocatable->rt_arg == NULL) {
	mrap = kmem_zalloc(sizeof (*mrap), KM_SLEEP);
	} else {
	mrap = msp->ms_allocatable->rt_arg;
	msp->ms_allocatable->rt_ops = NULL;
	msp->ms_allocatable->rt_arg = NULL;
	}
	mrap->mra_bt = &msp->ms_allocatable_by_size;
	mrap->mra_floor_shift = metaslab_by_size_min_shift;

	if (msp->ms_sm != NULL) {
	error = space_map_load_length(msp->ms_sm, msp->ms_allocatable,
	SM_FREE, length);

	/* Now, populate the size-sorted tree. */
	metaslab_rt_create(msp->ms_allocatable, mrap);
	msp->ms_allocatable->rt_ops = &metaslab_rt_ops;
	msp->ms_allocatable->rt_arg = mrap;

	struct mssa_arg arg = {0};
	arg.rt = msp->ms_allocatable;
	arg.mra = mrap;
	range_tree_walk(msp->ms_allocatable, metaslab_size_sorted_add,
	&arg);
	} else {
	/*
	* Add the size-sorted tree first, since we don't need to load
	* the metaslab from the spacemap.
	*/
	metaslab_rt_create(msp->ms_allocatable, mrap);
	msp->ms_allocatable->rt_ops = &metaslab_rt_ops;
	msp->ms_allocatable->rt_arg = mrap;
	/*
	* The space map has not been allocated yet, so treat
	* all the space in the metaslab as free and add it to the
	* ms_allocatable tree.
	*/
	range_tree_add(msp->ms_allocatable,
	msp->ms_start, msp->ms_size);

	if (msp->ms_new) {
	/*
	* If the ms_sm doesn't exist, this means that this
	* metaslab hasn't gone through metaslab_sync() and
	* thus has never been dirtied. So we shouldn't
	* expect any unflushed allocs or frees from previous
	* TXGs.
	*/
	ASSERT(range_tree_is_empty(msp->ms_unflushed_allocs));
	ASSERT(range_tree_is_empty(msp->ms_unflushed_frees));
	}
	}

	/*
	* We need to grab the ms_sync_lock to prevent metaslab_sync() from
	* changing the ms_sm (or log_sm) and the metaslab's range trees
	* while we are about to use them and populate the ms_allocatable.
	* The ms_lock is insufficient for this because metaslab_sync() doesn't
	* hold the ms_lock while writing the ms_checkpointing tree to disk.
	*/
	mutex_enter(&msp->ms_sync_lock);
	mutex_enter(&msp->ms_lock);

	ASSERT(!msp->ms_condensing);
	ASSERT(!msp->ms_flushing);

	if (error != 0) {
	mutex_exit(&msp->ms_sync_lock);
	return (error);
	}

	ASSERT3P(msp->ms_group, !=, NULL);
	msp->ms_loaded = B_TRUE;

	/*
	* Apply all the unflushed changes to ms_allocatable right
	* away so any manipulations we do below have a clear view
	* of what is allocated and what is free.
	*/
	range_tree_walk(msp->ms_unflushed_allocs,
	range_tree_remove, msp->ms_allocatable);
	range_tree_walk(msp->ms_unflushed_frees,
	range_tree_add, msp->ms_allocatable);

	ASSERT3P(msp->ms_group, !=, NULL);
	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
	if (spa_syncing_log_sm(spa) != NULL) {
	ASSERT(spa_feature_is_enabled(spa,
	SPA_FEATURE_LOG_SPACEMAP));

	/*
	* If we use a log space map we add all the segments
	* that are in ms_unflushed_frees so they are available
	* for allocation.
	*
	* ms_allocatable needs to contain all free segments
	* that are ready for allocations (thus not segments
	* from ms_freeing, ms_freed, and the ms_defer trees).
	* But if we grab the lock in this code path at a sync
	* pass later that 1, then it also contains the
	* segments of ms_freed (they were added to it earlier
	* in this path through ms_unflushed_frees). So we
	* need to remove all the segments that exist in
	* ms_freed from ms_allocatable as they will be added
	* later in metaslab_sync_done().
	*
	* When there's no log space map, the ms_allocatable
	* correctly doesn't contain any segments that exist
	* in ms_freed [see ms_synced_length].
	*/
	range_tree_walk(msp->ms_freed,
	range_tree_remove, msp->ms_allocatable);
	}

	/*
	* If we are not using the log space map, ms_allocatable
	* contains the segments that exist in the ms_defer trees
	* [see ms_synced_length]. Thus we need to remove them
	* from ms_allocatable as they will be added again in
	* metaslab_sync_done().
	*
	* If we are using the log space map, ms_allocatable still
	* contains the segments that exist in the ms_defer trees.
	* Not because it read them through the ms_sm though. But
	* because these segments are part of ms_unflushed_frees
	* whose segments we add to ms_allocatable earlier in this
	* code path.
	*/
	for (int t = 0; t < TXG_DEFER_SIZE; t++) {
	range_tree_walk(msp->ms_defer[t],
	range_tree_remove, msp->ms_allocatable);
	}

	/*
	* Call metaslab_recalculate_weight_and_sort() now that the
	* metaslab is loaded so we get the metaslab's real weight.
	*
	* Unless this metaslab was created with older software and
	* has not yet been converted to use segment-based weight, we
	* expect the new weight to be better or equal to the weight
	* that the metaslab had while it was not loaded. This is
	* because the old weight does not take into account the
	* consolidation of adjacent segments between TXGs. [see
	* comment for ms_synchist and ms_deferhist[] for more info]
	*/
	uint64_t weight = msp->ms_weight;
	uint64_t max_size = msp->ms_max_size;
	metaslab_recalculate_weight_and_sort(msp);
	if (!WEIGHT_IS_SPACEBASED(weight))
	ASSERT3U(weight, <=, msp->ms_weight);
	msp->ms_max_size = metaslab_largest_allocatable(msp);
	ASSERT3U(max_size, <=, msp->ms_max_size);
	hrtime_t load_end = gethrtime();
	msp->ms_load_time = load_end;
	zfs_dbgmsg("metaslab_load: txg %llu, spa %s, vdev_id %llu, "
	"ms_id %llu, smp_length %llu, "
	"unflushed_allocs %llu, unflushed_frees %llu, "
	"freed %llu, defer %llu + %llu, unloaded time %llu ms, "
	"loading_time %lld ms, ms_max_size %llu, "
	"max size error %lld, "
	"old_weight %llx, new_weight %llx",
	(u_longlong_t)spa_syncing_txg(spa), spa_name(spa),
	(u_longlong_t)msp->ms_group->mg_vd->vdev_id,
	(u_longlong_t)msp->ms_id,
	(u_longlong_t)space_map_length(msp->ms_sm),
	(u_longlong_t)range_tree_space(msp->ms_unflushed_allocs),
	(u_longlong_t)range_tree_space(msp->ms_unflushed_frees),
	(u_longlong_t)range_tree_space(msp->ms_freed),
	(u_longlong_t)range_tree_space(msp->ms_defer[0]),
	(u_longlong_t)range_tree_space(msp->ms_defer[1]),
	(longlong_t)((load_start - msp->ms_unload_time) / 1000000),
	(longlong_t)((load_end - load_start) / 1000000),
	(u_longlong_t)msp->ms_max_size,
	(u_longlong_t)msp->ms_max_size - max_size,
	(u_longlong_t)weight, (u_longlong_t)msp->ms_weight);

	metaslab_verify_space(msp, spa_syncing_txg(spa));
	mutex_exit(&msp->ms_sync_lock);
	return (0);
	}

	int
	metaslab_load(metaslab_t *msp)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	/*
	* There may be another thread loading the same metaslab, if that's
	* the case just wait until the other thread is done and return.
	*/
	metaslab_load_wait(msp);
	if (msp->ms_loaded)
	return (0);
	VERIFY(!msp->ms_loading);
	ASSERT(!msp->ms_condensing);

	/*
	* We set the loading flag BEFORE potentially dropping the lock to
	* wait for an ongoing flush (see ms_flushing below). This way other
	* threads know that there is already a thread that is loading this
	* metaslab.
	*/
	msp->ms_loading = B_TRUE;

	/*
	* Wait for any in-progress flushing to finish as we drop the ms_lock
	* both here (during space_map_load()) and in metaslab_flush() (when
	* we flush our changes to the ms_sm).
	*/
	if (msp->ms_flushing)
	metaslab_flush_wait(msp);

	/*
	* In the possibility that we were waiting for the metaslab to be
	* flushed (where we temporarily dropped the ms_lock), ensure that
	* no one else loaded the metaslab somehow.
	*/
	ASSERT(!msp->ms_loaded);

	/*
	* If we're loading a metaslab in the normal class, consider evicting
	* another one to keep our memory usage under the limit defined by the
	* zfs_metaslab_mem_limit tunable.
	*/
	if (spa_normal_class(msp->ms_group->mg_class->mc_spa) ==
	msp->ms_group->mg_class) {
	metaslab_potentially_evict(msp->ms_group->mg_class);
	}

	int error = metaslab_load_impl(msp);

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	msp->ms_loading = B_FALSE;
	cv_broadcast(&msp->ms_load_cv);

	return (error);
	}

	void
	metaslab_unload(metaslab_t *msp)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	/*
	* This can happen if a metaslab is selected for eviction (in
	* metaslab_potentially_evict) and then unloaded during spa_sync (via
	* metaslab_class_evict_old).
	*/
	if (!msp->ms_loaded)
	return;

	range_tree_vacate(msp->ms_allocatable, NULL, NULL);
	msp->ms_loaded = B_FALSE;
	msp->ms_unload_time = gethrtime();

	msp->ms_activation_weight = 0;
	msp->ms_weight &= ~METASLAB_ACTIVE_MASK;

	if (msp->ms_group != NULL) {
	metaslab_class_t *mc = msp->ms_group->mg_class;
	multilist_sublist_t *mls =
	multilist_sublist_lock_obj(&mc->mc_metaslab_txg_list, msp);
	if (multilist_link_active(&msp->ms_class_txg_node))
	multilist_sublist_remove(mls, msp);
	multilist_sublist_unlock(mls);

	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
	zfs_dbgmsg("metaslab_unload: txg %llu, spa %s, vdev_id %llu, "
	"ms_id %llu, weight %llx, "
	"selected txg %llu (%llu ms ago), alloc_txg %llu, "
	"loaded %llu ms ago, max_size %llu",
	(u_longlong_t)spa_syncing_txg(spa), spa_name(spa),
	(u_longlong_t)msp->ms_group->mg_vd->vdev_id,
	(u_longlong_t)msp->ms_id,
	(u_longlong_t)msp->ms_weight,
	(u_longlong_t)msp->ms_selected_txg,
	(u_longlong_t)(msp->ms_unload_time -
	msp->ms_selected_time) / 1000 / 1000,
	(u_longlong_t)msp->ms_alloc_txg,
	(u_longlong_t)(msp->ms_unload_time -
	msp->ms_load_time) / 1000 / 1000,
	(u_longlong_t)msp->ms_max_size);
	}

	/*
	* We explicitly recalculate the metaslab's weight based on its space
	* map (as it is now not loaded). We want unload metaslabs to always
	* have their weights calculated from the space map histograms, while
	* loaded ones have it calculated from their in-core range tree
	* [see metaslab_load()]. This way, the weight reflects the information
	* available in-core, whether it is loaded or not.
	*
	* If ms_group == NULL means that we came here from metaslab_fini(),
	* at which point it doesn't make sense for us to do the recalculation
	* and the sorting.
	*/
	if (msp->ms_group != NULL)
	metaslab_recalculate_weight_and_sort(msp);
	}

	/*
	* We want to optimize the memory use of the per-metaslab range
	* trees. To do this, we store the segments in the range trees in
	* units of sectors, zero-indexing from the start of the metaslab. If
	* the vdev_ms_shift - the vdev_ashift is less than 32, we can store
	* the ranges using two uint32_ts, rather than two uint64_ts.
	*/
	range_seg_type_t
	metaslab_calculate_range_tree_type(vdev_t vdev, metaslab_t msp,
	uint64_t start, uint64_t shift)
	{
	if (vdev->vdev_ms_shift - vdev->vdev_ashift < 32 &&
	!zfs_metaslab_force_large_segs) {
	*shift = vdev->vdev_ashift;
	*start = msp->ms_start;
	return (RANGE_SEG32);
	} else {
	*shift = 0;
	*start = 0;
	return (RANGE_SEG64);
	}
	}

	void
	metaslab_set_selected_txg(metaslab_t *msp, uint64_t txg)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));
	metaslab_class_t *mc = msp->ms_group->mg_class;
	multilist_sublist_t *mls =
	multilist_sublist_lock_obj(&mc->mc_metaslab_txg_list, msp);
	if (multilist_link_active(&msp->ms_class_txg_node))
	multilist_sublist_remove(mls, msp);
	msp->ms_selected_txg = txg;
	msp->ms_selected_time = gethrtime();
	multilist_sublist_insert_tail(mls, msp);
	multilist_sublist_unlock(mls);
	}

	void
	metaslab_space_update(vdev_t vd, metaslab_class_t mc, int64_t alloc_delta,
	int64_t defer_delta, int64_t space_delta)
	{
	vdev_space_update(vd, alloc_delta, defer_delta, space_delta);

	ASSERT3P(vd->vdev_spa->spa_root_vdev, ==, vd->vdev_parent);
	ASSERT(vd->vdev_ms_count != 0);

	metaslab_class_space_update(mc, alloc_delta, defer_delta, space_delta,
	vdev_deflated_space(vd, space_delta));
	}

	int
	metaslab_init(metaslab_group_t *mg, uint64_t id, uint64_t object,
	uint64_t txg, metaslab_t **msp)
	{
	vdev_t *vd = mg->mg_vd;
	spa_t *spa = vd->vdev_spa;
	objset_t *mos = spa->spa_meta_objset;
	metaslab_t *ms;
	int error;

	ms = kmem_zalloc(sizeof (metaslab_t), KM_SLEEP);
	mutex_init(&ms->ms_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&ms->ms_sync_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&ms->ms_load_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&ms->ms_flush_cv, NULL, CV_DEFAULT, NULL);
	multilist_link_init(&ms->ms_class_txg_node);

	ms->ms_id = id;
	ms->ms_start = id << vd->vdev_ms_shift;
	ms->ms_size = 1ULL << vd->vdev_ms_shift;
	ms->ms_allocator = -1;
	ms->ms_new = B_TRUE;

	vdev_ops_t *ops = vd->vdev_ops;
	if (ops->vdev_op_metaslab_init != NULL)
	ops->vdev_op_metaslab_init(vd, &ms->ms_start, &ms->ms_size);

	/*
	* We only open space map objects that already exist. All others
	* will be opened when we finally allocate an object for it. For
	* readonly pools there is no need to open the space map object.
	*
	* Note:
	* When called from vdev_expand(), we can't call into the DMU as
	* we are holding the spa_config_lock as a writer and we would
	* deadlock [see relevant comment in vdev_metaslab_init()]. in
	* that case, the object parameter is zero though, so we won't
	* call into the DMU.
	*/
	if (object != 0 && !(spa->spa_mode == SPA_MODE_READ &&
	!spa->spa_read_spacemaps)) {
	error = space_map_open(&ms->ms_sm, mos, object, ms->ms_start,
	ms->ms_size, vd->vdev_ashift);

	if (error != 0) {
	kmem_free(ms, sizeof (metaslab_t));
	return (error);
	}

	ASSERT(ms->ms_sm != NULL);
	ms->ms_allocated_space = space_map_allocated(ms->ms_sm);
	}

	uint64_t shift, start;
	range_seg_type_t type =
	metaslab_calculate_range_tree_type(vd, ms, &start, &shift);

	ms->ms_allocatable = range_tree_create(NULL, type, NULL, start, shift);
	for (int t = 0; t < TXG_SIZE; t++) {
	ms->ms_allocating[t] = range_tree_create(NULL, type,
	NULL, start, shift);
	}
	ms->ms_freeing = range_tree_create(NULL, type, NULL, start, shift);
	ms->ms_freed = range_tree_create(NULL, type, NULL, start, shift);
	for (int t = 0; t < TXG_DEFER_SIZE; t++) {
	ms->ms_defer[t] = range_tree_create(NULL, type, NULL,
	start, shift);
	}
	ms->ms_checkpointing =
	range_tree_create(NULL, type, NULL, start, shift);
	ms->ms_unflushed_allocs =
	range_tree_create(NULL, type, NULL, start, shift);

	metaslab_rt_arg_t mrap = kmem_zalloc(sizeof (mrap), KM_SLEEP);
	mrap->mra_bt = &ms->ms_unflushed_frees_by_size;
	mrap->mra_floor_shift = metaslab_by_size_min_shift;
	ms->ms_unflushed_frees = range_tree_create(&metaslab_rt_ops,
	type, mrap, start, shift);

	ms->ms_trim = range_tree_create(NULL, type, NULL, start, shift);

	metaslab_group_add(mg, ms);
	metaslab_set_fragmentation(ms, B_FALSE);

	/*
	* If we're opening an existing pool (txg == 0) or creating
	* a new one (txg == TXG_INITIAL), all space is available now.
	* If we're adding space to an existing pool, the new space
	* does not become available until after this txg has synced.
	* The metaslab's weight will also be initialized when we sync
	* out this txg. This ensures that we don't attempt to allocate
	* from it before we have initialized it completely.
	*/
	if (txg <= TXG_INITIAL) {
	metaslab_sync_done(ms, 0);
	metaslab_space_update(vd, mg->mg_class,
	metaslab_allocated_space(ms), 0, 0);
	}

	if (txg != 0) {
	vdev_dirty(vd, 0, NULL, txg);
	vdev_dirty(vd, VDD_METASLAB, ms, txg);
	}

	*msp = ms;

	return (0);
	}

	static void
	metaslab_fini_flush_data(metaslab_t *msp)
	{
	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;

	if (metaslab_unflushed_txg(msp) == 0) {
	ASSERT3P(avl_find(&spa->spa_metaslabs_by_flushed, msp, NULL),
	==, NULL);
	return;
	}
	ASSERT(spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP));

	mutex_enter(&spa->spa_flushed_ms_lock);
	avl_remove(&spa->spa_metaslabs_by_flushed, msp);
	mutex_exit(&spa->spa_flushed_ms_lock);

	spa_log_sm_decrement_mscount(spa, metaslab_unflushed_txg(msp));
	spa_log_summary_decrement_mscount(spa, metaslab_unflushed_txg(msp),
	metaslab_unflushed_dirty(msp));
	}

	uint64_t
	metaslab_unflushed_changes_memused(metaslab_t *ms)
	{
	return ((range_tree_numsegs(ms->ms_unflushed_allocs) +
	range_tree_numsegs(ms->ms_unflushed_frees)) *
	ms->ms_unflushed_allocs->rt_root.bt_elem_size);
	}

	void
	metaslab_fini(metaslab_t *msp)
	{
	metaslab_group_t *mg = msp->ms_group;
	vdev_t *vd = mg->mg_vd;
	spa_t *spa = vd->vdev_spa;

	metaslab_fini_flush_data(msp);

	metaslab_group_remove(mg, msp);

	mutex_enter(&msp->ms_lock);
	VERIFY(msp->ms_group == NULL);

	/*
	* If this metaslab hasn't been through metaslab_sync_done() yet its
	* space hasn't been accounted for in its vdev and doesn't need to be
	* subtracted.
	*/
	if (!msp->ms_new) {
	metaslab_space_update(vd, mg->mg_class,
	-metaslab_allocated_space(msp), 0, -msp->ms_size);

	}
	space_map_close(msp->ms_sm);
	msp->ms_sm = NULL;

	metaslab_unload(msp);

	range_tree_destroy(msp->ms_allocatable);
	range_tree_destroy(msp->ms_freeing);
	range_tree_destroy(msp->ms_freed);

	ASSERT3U(spa->spa_unflushed_stats.sus_memused, >=,
	metaslab_unflushed_changes_memused(msp));
	spa->spa_unflushed_stats.sus_memused -=
	metaslab_unflushed_changes_memused(msp);
	range_tree_vacate(msp->ms_unflushed_allocs, NULL, NULL);
	range_tree_destroy(msp->ms_unflushed_allocs);
	range_tree_destroy(msp->ms_checkpointing);
	range_tree_vacate(msp->ms_unflushed_frees, NULL, NULL);
	range_tree_destroy(msp->ms_unflushed_frees);

	for (int t = 0; t < TXG_SIZE; t++) {
	range_tree_destroy(msp->ms_allocating[t]);
	}
	for (int t = 0; t < TXG_DEFER_SIZE; t++) {
	range_tree_destroy(msp->ms_defer[t]);
	}
	ASSERT0(msp->ms_deferspace);

	for (int t = 0; t < TXG_SIZE; t++)
	ASSERT(!txg_list_member(&vd->vdev_ms_list, msp, t));

	range_tree_vacate(msp->ms_trim, NULL, NULL);
	range_tree_destroy(msp->ms_trim);

	mutex_exit(&msp->ms_lock);
	cv_destroy(&msp->ms_load_cv);
	cv_destroy(&msp->ms_flush_cv);
	mutex_destroy(&msp->ms_lock);
	mutex_destroy(&msp->ms_sync_lock);
	ASSERT3U(msp->ms_allocator, ==, -1);

	kmem_free(msp, sizeof (metaslab_t));
	}

	#define FRAGMENTATION_TABLE_SIZE 17

	/*
	* This table defines a segment size based fragmentation metric that will
	* allow each metaslab to derive its own fragmentation value. This is done
	* by calculating the space in each bucket of the spacemap histogram and
	* multiplying that by the fragmentation metric in this table. Doing
	* this for all buckets and dividing it by the total amount of free
	* space in this metaslab (i.e. the total free space in all buckets) gives
	* us the fragmentation metric. This means that a high fragmentation metric
	* equates to most of the free space being comprised of small segments.
	* Conversely, if the metric is low, then most of the free space is in
	* large segments. A 10% change in fragmentation equates to approximately
	* double the number of segments.
	*
	* This table defines 0% fragmented space using 16MB segments. Testing has
	* shown that segments that are greater than or equal to 16MB do not suffer
	* from drastic performance problems. Using this value, we derive the rest
	* of the table. Since the fragmentation value is never stored on disk, it
	* is possible to change these calculations in the future.
	*/
	static const int zfs_frag_table[FRAGMENTATION_TABLE_SIZE] = {
	100, /* 512B */
	100, /* 1K */
	98, /* 2K */
	95, /* 4K */
	90, /* 8K */
	80, /* 16K */
	70, /* 32K */
	60, /* 64K */
	50, /* 128K */
	40, /* 256K */
	30, /* 512K */
	20, /* 1M */
	15, /* 2M */
	10, /* 4M */
	5, /* 8M */
	0 /* 16M */
	};

	/*
	* Calculate the metaslab's fragmentation metric and set ms_fragmentation.
	* Setting this value to ZFS_FRAG_INVALID means that the metaslab has not
	* been upgraded and does not support this metric. Otherwise, the return
	* value should be in the range [0, 100].
	*/
	static void
	metaslab_set_fragmentation(metaslab_t *msp, boolean_t nodirty)
	{
	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
	uint64_t fragmentation = 0;
	uint64_t total = 0;
	boolean_t feature_enabled = spa_feature_is_enabled(spa,
	SPA_FEATURE_SPACEMAP_HISTOGRAM);

	if (!feature_enabled) {
	msp->ms_fragmentation = ZFS_FRAG_INVALID;
	return;
	}

	/*
	* A null space map means that the entire metaslab is free
	* and thus is not fragmented.
	*/
	if (msp->ms_sm == NULL) {
	msp->ms_fragmentation = 0;
	return;
	}

	/*
	* If this metaslab's space map has not been upgraded, flag it
	* so that we upgrade next time we encounter it.
	*/
	if (msp->ms_sm->sm_dbuf->db_size != sizeof (space_map_phys_t)) {
	uint64_t txg = spa_syncing_txg(spa);
	vdev_t *vd = msp->ms_group->mg_vd;

	/*
	* If we've reached the final dirty txg, then we must
	* be shutting down the pool. We don't want to dirty
	* any data past this point so skip setting the condense
	* flag. We can retry this action the next time the pool
	* is imported. We also skip marking this metaslab for
	* condensing if the caller has explicitly set nodirty.
	*/
	if (!nodirty &&
	spa_writeable(spa) && txg < spa_final_dirty_txg(spa)) {
	msp->ms_condense_wanted = B_TRUE;
	vdev_dirty(vd, VDD_METASLAB, msp, txg + 1);
	zfs_dbgmsg("txg %llu, requesting force condense: "
	"ms_id %llu, vdev_id %llu", (u_longlong_t)txg,
	(u_longlong_t)msp->ms_id,
	(u_longlong_t)vd->vdev_id);
	}
	msp->ms_fragmentation = ZFS_FRAG_INVALID;
	return;
	}

	for (int i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++) {
	uint64_t space = 0;
	uint8_t shift = msp->ms_sm->sm_shift;

	int idx = MIN(shift - SPA_MINBLOCKSHIFT + i,
	FRAGMENTATION_TABLE_SIZE - 1);

	if (msp->ms_sm->sm_phys->smp_histogram[i] == 0)
	continue;

	space = msp->ms_sm->sm_phys->smp_histogram[i] << (i + shift);
	total += space;

	ASSERT3U(idx, <, FRAGMENTATION_TABLE_SIZE);
	fragmentation += space * zfs_frag_table[idx];
	}

	if (total > 0)
	fragmentation /= total;
	ASSERT3U(fragmentation, <=, 100);

	msp->ms_fragmentation = fragmentation;
	}

	/*
	* Compute a weight -- a selection preference value -- for the given metaslab.
	* This is based on the amount of free space, the level of fragmentation,
	* the LBA range, and whether the metaslab is loaded.
	*/
	static uint64_t
	metaslab_space_weight(metaslab_t *msp)
	{
	metaslab_group_t *mg = msp->ms_group;
	vdev_t *vd = mg->mg_vd;
	uint64_t weight, space;

	ASSERT(MUTEX_HELD(&msp->ms_lock));

	/*
	* The baseline weight is the metaslab's free space.
	*/
	space = msp->ms_size - metaslab_allocated_space(msp);

	if (metaslab_fragmentation_factor_enabled &&
	msp->ms_fragmentation != ZFS_FRAG_INVALID) {
	/*
	* Use the fragmentation information to inversely scale
	* down the baseline weight. We need to ensure that we
	* don't exclude this metaslab completely when it's 100%
	* fragmented. To avoid this we reduce the fragmented value
	* by 1.
	*/
	space = (space * (100 - (msp->ms_fragmentation - 1))) / 100;

	/*
	* If space < SPA_MINBLOCKSIZE, then we will not allocate from
	* this metaslab again. The fragmentation metric may have
	* decreased the space to something smaller than
	* SPA_MINBLOCKSIZE, so reset the space to SPA_MINBLOCKSIZE
	* so that we can consume any remaining space.
	*/
	if (space > 0 && space < SPA_MINBLOCKSIZE)
	space = SPA_MINBLOCKSIZE;
	}
	weight = space;

	/*
	* Modern disks have uniform bit density and constant angular velocity.
	* Therefore, the outer recording zones are faster (higher bandwidth)
	* than the inner zones by the ratio of outer to inner track diameter,
	* which is typically around 2:1. We account for this by assigning
	* higher weight to lower metaslabs (multiplier ranging from 2x to 1x).
	* In effect, this means that we'll select the metaslab with the most
	* free bandwidth rather than simply the one with the most free space.
	*/
	if (!vd->vdev_nonrot && metaslab_lba_weighting_enabled) {
	weight = 2 * weight - (msp->ms_id * weight) / vd->vdev_ms_count;
	ASSERT(weight >= space && weight <= 2 * space);
	}

	/*
	* If this metaslab is one we're actively using, adjust its
	* weight to make it preferable to any inactive metaslab so
	* we'll polish it off. If the fragmentation on this metaslab
	* has exceed our threshold, then don't mark it active.
	*/
	if (msp->ms_loaded && msp->ms_fragmentation != ZFS_FRAG_INVALID &&
	msp->ms_fragmentation <= zfs_metaslab_fragmentation_threshold) {
	weight \|= (msp->ms_weight & METASLAB_ACTIVE_MASK);
	}

	WEIGHT_SET_SPACEBASED(weight);
	return (weight);
	}

	/*
	* Return the weight of the specified metaslab, according to the segment-based
	* weighting algorithm. The metaslab must be loaded. This function can
	* be called within a sync pass since it relies only on the metaslab's
	* range tree which is always accurate when the metaslab is loaded.
	*/
	static uint64_t
	metaslab_weight_from_range_tree(metaslab_t *msp)
	{
	uint64_t weight = 0;
	uint32_t segments = 0;

	ASSERT(msp->ms_loaded);

	for (int i = RANGE_TREE_HISTOGRAM_SIZE - 1; i >= SPA_MINBLOCKSHIFT;
	i--) {
	uint8_t shift = msp->ms_group->mg_vd->vdev_ashift;
	int max_idx = SPACE_MAP_HISTOGRAM_SIZE + shift - 1;

	segments <<= 1;
	segments += msp->ms_allocatable->rt_histogram[i];

	/*
	* The range tree provides more precision than the space map
	* and must be downgraded so that all values fit within the
	* space map's histogram. This allows us to compare loaded
	* vs. unloaded metaslabs to determine which metaslab is
	* considered "best".
	*/
	if (i > max_idx)
	continue;

	if (segments != 0) {
	WEIGHT_SET_COUNT(weight, segments);
	WEIGHT_SET_INDEX(weight, i);
	WEIGHT_SET_ACTIVE(weight, 0);
	break;
	}
	}
	return (weight);
	}

	/*
	* Calculate the weight based on the on-disk histogram. Should be applied
	* only to unloaded metaslabs (i.e no incoming allocations) in-order to
	* give results consistent with the on-disk state
	*/
	static uint64_t
	metaslab_weight_from_spacemap(metaslab_t *msp)
	{
	space_map_t *sm = msp->ms_sm;
	ASSERT(!msp->ms_loaded);
	ASSERT(sm != NULL);
	ASSERT3U(space_map_object(sm), !=, 0);
	ASSERT3U(sm->sm_dbuf->db_size, ==, sizeof (space_map_phys_t));

	/*
	* Create a joint histogram from all the segments that have made
	* it to the metaslab's space map histogram, that are not yet
	* available for allocation because they are still in the freeing
	* pipeline (e.g. freeing, freed, and defer trees). Then subtract
	* these segments from the space map's histogram to get a more
	* accurate weight.
	*/
	uint64_t deferspace_histogram[SPACE_MAP_HISTOGRAM_SIZE] = {0};
	for (int i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++)
	deferspace_histogram[i] += msp->ms_synchist[i];
	for (int t = 0; t < TXG_DEFER_SIZE; t++) {
	for (int i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++) {
	deferspace_histogram[i] += msp->ms_deferhist[t][i];
	}
	}

	uint64_t weight = 0;
	for (int i = SPACE_MAP_HISTOGRAM_SIZE - 1; i >= 0; i--) {
	ASSERT3U(sm->sm_phys->smp_histogram[i], >=,
	deferspace_histogram[i]);
	uint64_t count =
	sm->sm_phys->smp_histogram[i] - deferspace_histogram[i];
	if (count != 0) {
	WEIGHT_SET_COUNT(weight, count);
	WEIGHT_SET_INDEX(weight, i + sm->sm_shift);
	WEIGHT_SET_ACTIVE(weight, 0);
	break;
	}
	}
	return (weight);
	}

	/*
	* Compute a segment-based weight for the specified metaslab. The weight
	* is determined by highest bucket in the histogram. The information
	* for the highest bucket is encoded into the weight value.
	*/
	static uint64_t
	metaslab_segment_weight(metaslab_t *msp)
	{
	metaslab_group_t *mg = msp->ms_group;
	uint64_t weight = 0;
	uint8_t shift = mg->mg_vd->vdev_ashift;

	ASSERT(MUTEX_HELD(&msp->ms_lock));

	/*
	* The metaslab is completely free.
	*/
	if (metaslab_allocated_space(msp) == 0) {
	int idx = highbit64(msp->ms_size) - 1;
	int max_idx = SPACE_MAP_HISTOGRAM_SIZE + shift - 1;

	if (idx < max_idx) {
	WEIGHT_SET_COUNT(weight, 1ULL);
	WEIGHT_SET_INDEX(weight, idx);
	} else {
	WEIGHT_SET_COUNT(weight, 1ULL << (idx - max_idx));
	WEIGHT_SET_INDEX(weight, max_idx);
	}
	WEIGHT_SET_ACTIVE(weight, 0);
	ASSERT(!WEIGHT_IS_SPACEBASED(weight));
	return (weight);
	}

	ASSERT3U(msp->ms_sm->sm_dbuf->db_size, ==, sizeof (space_map_phys_t));

	/*
	* If the metaslab is fully allocated then just make the weight 0.
	*/
	if (metaslab_allocated_space(msp) == msp->ms_size)
	return (0);
	/*
	* If the metaslab is already loaded, then use the range tree to
	* determine the weight. Otherwise, we rely on the space map information
	* to generate the weight.
	*/
	if (msp->ms_loaded) {
	weight = metaslab_weight_from_range_tree(msp);
	} else {
	weight = metaslab_weight_from_spacemap(msp);
	}

	/*
	* If the metaslab was active the last time we calculated its weight
	* then keep it active. We want to consume the entire region that
	* is associated with this weight.
	*/
	if (msp->ms_activation_weight != 0 && weight != 0)
	WEIGHT_SET_ACTIVE(weight, WEIGHT_GET_ACTIVE(msp->ms_weight));
	return (weight);
	}

	/*
	* Determine if we should attempt to allocate from this metaslab. If the
	* metaslab is loaded, then we can determine if the desired allocation
	* can be satisfied by looking at the size of the maximum free segment
	* on that metaslab. Otherwise, we make our decision based on the metaslab's
	* weight. For segment-based weighting we can determine the maximum
	* allocation based on the index encoded in its value. For space-based
	* weights we rely on the entire weight (excluding the weight-type bit).
	*/
	static boolean_t
	metaslab_should_allocate(metaslab_t *msp, uint64_t asize, boolean_t try_hard)
	{
	/*
	* This case will usually but not always get caught by the checks below;
	* metaslabs can be loaded by various means, including the trim and
	* initialize code. Once that happens, without this check they are
	* allocatable even before they finish their first txg sync.
	*/
	if (unlikely(msp->ms_new))
	return (B_FALSE);

	/*
	* If the metaslab is loaded, ms_max_size is definitive and we can use
	* the fast check. If it's not, the ms_max_size is a lower bound (once
	* set), and we should use the fast check as long as we're not in
	* try_hard and it's been less than zfs_metaslab_max_size_cache_sec
	* seconds since the metaslab was unloaded.
	*/
	if (msp->ms_loaded \|\|
	(msp->ms_max_size != 0 && !try_hard && gethrtime() <
	msp->ms_unload_time + SEC2NSEC(zfs_metaslab_max_size_cache_sec)))
	return (msp->ms_max_size >= asize);

	boolean_t should_allocate;
	if (!WEIGHT_IS_SPACEBASED(msp->ms_weight)) {
	/*
	* The metaslab segment weight indicates segments in the
	* range [2^i, 2^(i+1)), where i is the index in the weight.
	* Since the asize might be in the middle of the range, we
	* should attempt the allocation if asize < 2^(i+1).
	*/
	should_allocate = (asize <
	1ULL << (WEIGHT_GET_INDEX(msp->ms_weight) + 1));
	} else {
	should_allocate = (asize <=
	(msp->ms_weight & ~METASLAB_WEIGHT_TYPE));
	}

	return (should_allocate);
	}

	static uint64_t
	metaslab_weight(metaslab_t *msp, boolean_t nodirty)
	{
	vdev_t *vd = msp->ms_group->mg_vd;
	spa_t *spa = vd->vdev_spa;
	uint64_t weight;

	ASSERT(MUTEX_HELD(&msp->ms_lock));

	metaslab_set_fragmentation(msp, nodirty);

	/*
	* Update the maximum size. If the metaslab is loaded, this will
	* ensure that we get an accurate maximum size if newly freed space
	* has been added back into the free tree. If the metaslab is
	* unloaded, we check if there's a larger free segment in the
	* unflushed frees. This is a lower bound on the largest allocatable
	* segment size. Coalescing of adjacent entries may reveal larger
	* allocatable segments, but we aren't aware of those until loading
	* the space map into a range tree.
	*/
	if (msp->ms_loaded) {
	msp->ms_max_size = metaslab_largest_allocatable(msp);
	} else {
	msp->ms_max_size = MAX(msp->ms_max_size,
	metaslab_largest_unflushed_free(msp));
	}

	/*
	* Segment-based weighting requires space map histogram support.
	*/
	if (zfs_metaslab_segment_weight_enabled &&
	spa_feature_is_enabled(spa, SPA_FEATURE_SPACEMAP_HISTOGRAM) &&
	(msp->ms_sm == NULL \|\| msp->ms_sm->sm_dbuf->db_size ==
	sizeof (space_map_phys_t))) {
	weight = metaslab_segment_weight(msp);
	} else {
	weight = metaslab_space_weight(msp);
	}
	return (weight);
	}

	void
	metaslab_recalculate_weight_and_sort(metaslab_t *msp)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	/* note: we preserve the mask (e.g. indication of primary, etc..) */
	uint64_t was_active = msp->ms_weight & METASLAB_ACTIVE_MASK;
	metaslab_group_sort(msp->ms_group, msp,
	metaslab_weight(msp, B_FALSE) \| was_active);
	}

	static int
	metaslab_activate_allocator(metaslab_group_t mg, metaslab_t msp,
	int allocator, uint64_t activation_weight)
	{
	metaslab_group_allocator_t *mga = &mg->mg_allocator[allocator];
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	/*
	* If we're activating for the claim code, we don't want to actually
	* set the metaslab up for a specific allocator.
	*/
	if (activation_weight == METASLAB_WEIGHT_CLAIM) {
	ASSERT0(msp->ms_activation_weight);
	msp->ms_activation_weight = msp->ms_weight;
	metaslab_group_sort(mg, msp, msp->ms_weight \|
	activation_weight);
	return (0);
	}

	metaslab_t **mspp = (activation_weight == METASLAB_WEIGHT_PRIMARY ?
	&mga->mga_primary : &mga->mga_secondary);

	mutex_enter(&mg->mg_lock);
	if (*mspp != NULL) {
	mutex_exit(&mg->mg_lock);
	return (EEXIST);
	}

	*mspp = msp;
	ASSERT3S(msp->ms_allocator, ==, -1);
	msp->ms_allocator = allocator;
	msp->ms_primary = (activation_weight == METASLAB_WEIGHT_PRIMARY);

	ASSERT0(msp->ms_activation_weight);
	msp->ms_activation_weight = msp->ms_weight;
	metaslab_group_sort_impl(mg, msp,
	msp->ms_weight \| activation_weight);
	mutex_exit(&mg->mg_lock);

	return (0);
	}

	static int
	metaslab_activate(metaslab_t *msp, int allocator, uint64_t activation_weight)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	/*
	* The current metaslab is already activated for us so there
	* is nothing to do. Already activated though, doesn't mean
	* that this metaslab is activated for our allocator nor our
	* requested activation weight. The metaslab could have started
	* as an active one for our allocator but changed allocators
	* while we were waiting to grab its ms_lock or we stole it
	* [see find_valid_metaslab()]. This means that there is a
	* possibility of passivating a metaslab of another allocator
	* or from a different activation mask, from this thread.
	*/
	if ((msp->ms_weight & METASLAB_ACTIVE_MASK) != 0) {
	ASSERT(msp->ms_loaded);
	return (0);
	}

	int error = metaslab_load(msp);
	if (error != 0) {
	metaslab_group_sort(msp->ms_group, msp, 0);
	return (error);
	}

	/*
	* When entering metaslab_load() we may have dropped the
	* ms_lock because we were loading this metaslab, or we
	* were waiting for another thread to load it for us. In
	* that scenario, we recheck the weight of the metaslab
	* to see if it was activated by another thread.
	*
	* If the metaslab was activated for another allocator or
	* it was activated with a different activation weight (e.g.
	* we wanted to make it a primary but it was activated as
	* secondary) we return error (EBUSY).
	*
	* If the metaslab was activated for the same allocator
	* and requested activation mask, skip activating it.
	*/
	if ((msp->ms_weight & METASLAB_ACTIVE_MASK) != 0) {
	if (msp->ms_allocator != allocator)
	return (EBUSY);

	if ((msp->ms_weight & activation_weight) == 0)
	return (SET_ERROR(EBUSY));

	EQUIV((activation_weight == METASLAB_WEIGHT_PRIMARY),
	msp->ms_primary);
	return (0);
	}

	/*
	* If the metaslab has literally 0 space, it will have weight 0. In
	* that case, don't bother activating it. This can happen if the
	* metaslab had space during find_valid_metaslab, but another thread
	* loaded it and used all that space while we were waiting to grab the
	* lock.
	*/
	if (msp->ms_weight == 0) {
	ASSERT0(range_tree_space(msp->ms_allocatable));
	return (SET_ERROR(ENOSPC));
	}

	if ((error = metaslab_activate_allocator(msp->ms_group, msp,
	allocator, activation_weight)) != 0) {
	return (error);
	}

	ASSERT(msp->ms_loaded);
	ASSERT(msp->ms_weight & METASLAB_ACTIVE_MASK);

	return (0);
	}

	static void
	metaslab_passivate_allocator(metaslab_group_t mg, metaslab_t msp,
	uint64_t weight)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));
	ASSERT(msp->ms_loaded);

	if (msp->ms_weight & METASLAB_WEIGHT_CLAIM) {
	metaslab_group_sort(mg, msp, weight);
	return;
	}

	mutex_enter(&mg->mg_lock);
	ASSERT3P(msp->ms_group, ==, mg);
	ASSERT3S(0, <=, msp->ms_allocator);
	ASSERT3U(msp->ms_allocator, <, mg->mg_allocators);

	metaslab_group_allocator_t *mga = &mg->mg_allocator[msp->ms_allocator];
	if (msp->ms_primary) {
	ASSERT3P(mga->mga_primary, ==, msp);
	ASSERT(msp->ms_weight & METASLAB_WEIGHT_PRIMARY);
	mga->mga_primary = NULL;
	} else {
	ASSERT3P(mga->mga_secondary, ==, msp);
	ASSERT(msp->ms_weight & METASLAB_WEIGHT_SECONDARY);
	mga->mga_secondary = NULL;
	}
	msp->ms_allocator = -1;
	metaslab_group_sort_impl(mg, msp, weight);
	mutex_exit(&mg->mg_lock);
	}

	static void
	metaslab_passivate(metaslab_t *msp, uint64_t weight)
	{
	uint64_t size __maybe_unused = weight & ~METASLAB_WEIGHT_TYPE;

	/*
	* If size < SPA_MINBLOCKSIZE, then we will not allocate from
	* this metaslab again. In that case, it had better be empty,
	* or we would be leaving space on the table.
	*/
	ASSERT(!WEIGHT_IS_SPACEBASED(msp->ms_weight) \|\|
	size >= SPA_MINBLOCKSIZE \|\|
	range_tree_space(msp->ms_allocatable) == 0);
	ASSERT0(weight & METASLAB_ACTIVE_MASK);

	ASSERT(msp->ms_activation_weight != 0);
	msp->ms_activation_weight = 0;
	metaslab_passivate_allocator(msp->ms_group, msp, weight);
	ASSERT0(msp->ms_weight & METASLAB_ACTIVE_MASK);
	}

	/*
	* Segment-based metaslabs are activated once and remain active until
	* we either fail an allocation attempt (similar to space-based metaslabs)
	* or have exhausted the free space in zfs_metaslab_switch_threshold
	* buckets since the metaslab was activated. This function checks to see
	* if we've exhausted the zfs_metaslab_switch_threshold buckets in the
	* metaslab and passivates it proactively. This will allow us to select a
	* metaslab with a larger contiguous region, if any, remaining within this
	* metaslab group. If we're in sync pass > 1, then we continue using this
	* metaslab so that we don't dirty more block and cause more sync passes.
	*/
	static void
	metaslab_segment_may_passivate(metaslab_t *msp)
	{
	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;

	if (WEIGHT_IS_SPACEBASED(msp->ms_weight) \|\| spa_sync_pass(spa) > 1)
	return;

	/*
	* Since we are in the middle of a sync pass, the most accurate
	* information that is accessible to us is the in-core range tree
	* histogram; calculate the new weight based on that information.
	*/
	uint64_t weight = metaslab_weight_from_range_tree(msp);
	int activation_idx = WEIGHT_GET_INDEX(msp->ms_activation_weight);
	int current_idx = WEIGHT_GET_INDEX(weight);

	if (current_idx <= activation_idx - zfs_metaslab_switch_threshold)
	metaslab_passivate(msp, weight);
	}

	static void
	metaslab_preload(void *arg)
	{
	metaslab_t *msp = arg;
	metaslab_class_t *mc = msp->ms_group->mg_class;
	spa_t *spa = mc->mc_spa;
	fstrans_cookie_t cookie = spl_fstrans_mark();

	ASSERT(!MUTEX_HELD(&msp->ms_group->mg_lock));

	mutex_enter(&msp->ms_lock);
	(void) metaslab_load(msp);
	metaslab_set_selected_txg(msp, spa_syncing_txg(spa));
	mutex_exit(&msp->ms_lock);
	spl_fstrans_unmark(cookie);
	}

	static void
	metaslab_group_preload(metaslab_group_t *mg)
	{
	spa_t *spa = mg->mg_vd->vdev_spa;
	metaslab_t *msp;
	avl_tree_t *t = &mg->mg_metaslab_tree;
	int m = 0;

	if (spa_shutting_down(spa) \|\| !metaslab_preload_enabled)
	return;

	mutex_enter(&mg->mg_lock);

	/*
	* Load the next potential metaslabs
	*/
	for (msp = avl_first(t); msp != NULL; msp = AVL_NEXT(t, msp)) {
	ASSERT3P(msp->ms_group, ==, mg);

	/*
	* We preload only the maximum number of metaslabs specified
	* by metaslab_preload_limit. If a metaslab is being forced
	* to condense then we preload it too. This will ensure
	* that force condensing happens in the next txg.
	*/
	if (++m > metaslab_preload_limit && !msp->ms_condense_wanted) {
	continue;
	}

	VERIFY(taskq_dispatch(spa->spa_metaslab_taskq, metaslab_preload,
	msp, TQ_SLEEP \| (m <= mg->mg_allocators ? TQ_FRONT : 0))
	!= TASKQID_INVALID);
	}
	mutex_exit(&mg->mg_lock);
	}

	/*
	* Determine if the space map's on-disk footprint is past our tolerance for
	* inefficiency. We would like to use the following criteria to make our
	* decision:
	*
	* 1. Do not condense if the size of the space map object would dramatically
	* increase as a result of writing out the free space range tree.
	*
	* 2. Condense if the on on-disk space map representation is at least
	* zfs_condense_pct/100 times the size of the optimal representation
	* (i.e. zfs_condense_pct = 110 and in-core = 1MB, optimal = 1.1MB).
	*
	* 3. Do not condense if the on-disk size of the space map does not actually
	* decrease.
	*
	* Unfortunately, we cannot compute the on-disk size of the space map in this
	* context because we cannot accurately compute the effects of compression, etc.
	* Instead, we apply the heuristic described in the block comment for
	* zfs_metaslab_condense_block_threshold - we only condense if the space used
	* is greater than a threshold number of blocks.
	*/
	static boolean_t
	metaslab_should_condense(metaslab_t *msp)
	{
	space_map_t *sm = msp->ms_sm;
	vdev_t *vd = msp->ms_group->mg_vd;
	uint64_t vdev_blocksize = 1ULL << vd->vdev_ashift;

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	ASSERT(msp->ms_loaded);
	ASSERT(sm != NULL);
	ASSERT3U(spa_sync_pass(vd->vdev_spa), ==, 1);

	/*
	* We always condense metaslabs that are empty and metaslabs for
	* which a condense request has been made.
	*/
	if (range_tree_numsegs(msp->ms_allocatable) == 0 \|\|
	msp->ms_condense_wanted)
	return (B_TRUE);

	uint64_t record_size = MAX(sm->sm_blksz, vdev_blocksize);
	uint64_t object_size = space_map_length(sm);
	uint64_t optimal_size = space_map_estimate_optimal_size(sm,
	msp->ms_allocatable, SM_NO_VDEVID);

	return (object_size >= (optimal_size * zfs_condense_pct / 100) &&
	object_size > zfs_metaslab_condense_block_threshold * record_size);
	}

	/*
	* Condense the on-disk space map representation to its minimized form.
	* The minimized form consists of a small number of allocations followed
	* by the entries of the free range tree (ms_allocatable). The condensed
	* spacemap contains all the entries of previous TXGs (including those in
	* the pool-wide log spacemaps; thus this is effectively a superset of
	* metaslab_flush()), but this TXG's entries still need to be written.
	*/
	static void
	metaslab_condense(metaslab_t msp, dmu_tx_t tx)
	{
	range_tree_t *condense_tree;
	space_map_t *sm = msp->ms_sm;
	uint64_t txg = dmu_tx_get_txg(tx);
	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	ASSERT(msp->ms_loaded);
	ASSERT(msp->ms_sm != NULL);

	/*
	* In order to condense the space map, we need to change it so it
	* only describes which segments are currently allocated and free.
	*
	* All the current free space resides in the ms_allocatable, all
	* the ms_defer trees, and all the ms_allocating trees. We ignore
	* ms_freed because it is empty because we're in sync pass 1. We
	* ignore ms_freeing because these changes are not yet reflected
	* in the spacemap (they will be written later this txg).
	*
	* So to truncate the space map to represent all the entries of
	* previous TXGs we do the following:
	*
	* 1] We create a range tree (condense tree) that is 100% empty.
	* 2] We add to it all segments found in the ms_defer trees
	* as those segments are marked as free in the original space
	* map. We do the same with the ms_allocating trees for the same
	* reason. Adding these segments should be a relatively
	* inexpensive operation since we expect these trees to have a
	* small number of nodes.
	* 3] We vacate any unflushed allocs, since they are not frees we
	* need to add to the condense tree. Then we vacate any
	* unflushed frees as they should already be part of ms_allocatable.
	* 4] At this point, we would ideally like to add all segments
	* in the ms_allocatable tree from the condense tree. This way
	* we would write all the entries of the condense tree as the
	* condensed space map, which would only contain freed
	* segments with everything else assumed to be allocated.
	*
	* Doing so can be prohibitively expensive as ms_allocatable can
	* be large, and therefore computationally expensive to add to
	* the condense_tree. Instead we first sync out an entry marking
	* everything as allocated, then the condense_tree and then the
	* ms_allocatable, in the condensed space map. While this is not
	* optimal, it is typically close to optimal and more importantly
	* much cheaper to compute.
	*
	* 5] Finally, as both of the unflushed trees were written to our
	* new and condensed metaslab space map, we basically flushed
	* all the unflushed changes to disk, thus we call
	* metaslab_flush_update().
	*/
	ASSERT3U(spa_sync_pass(spa), ==, 1);
	ASSERT(range_tree_is_empty(msp->ms_freed)); /* since it is pass 1 */

	zfs_dbgmsg("condensing: txg %llu, msp[%llu] %px, vdev id %llu, "
	"spa %s, smp size %llu, segments %llu, forcing condense=%s",
	(u_longlong_t)txg, (u_longlong_t)msp->ms_id, msp,
	(u_longlong_t)msp->ms_group->mg_vd->vdev_id,
	spa->spa_name, (u_longlong_t)space_map_length(msp->ms_sm),
	(u_longlong_t)range_tree_numsegs(msp->ms_allocatable),
	msp->ms_condense_wanted ? "TRUE" : "FALSE");

	msp->ms_condense_wanted = B_FALSE;

	range_seg_type_t type;
	uint64_t shift, start;
	type = metaslab_calculate_range_tree_type(msp->ms_group->mg_vd, msp,
	&start, &shift);

	condense_tree = range_tree_create(NULL, type, NULL, start, shift);

	for (int t = 0; t < TXG_DEFER_SIZE; t++) {
	range_tree_walk(msp->ms_defer[t],
	range_tree_add, condense_tree);
	}

	for (int t = 0; t < TXG_CONCURRENT_STATES; t++) {
	range_tree_walk(msp->ms_allocating[(txg + t) & TXG_MASK],
	range_tree_add, condense_tree);
	}

	ASSERT3U(spa->spa_unflushed_stats.sus_memused, >=,
	metaslab_unflushed_changes_memused(msp));
	spa->spa_unflushed_stats.sus_memused -=
	metaslab_unflushed_changes_memused(msp);
	range_tree_vacate(msp->ms_unflushed_allocs, NULL, NULL);
	range_tree_vacate(msp->ms_unflushed_frees, NULL, NULL);

	/*
	* We're about to drop the metaslab's lock thus allowing other
	* consumers to change it's content. Set the metaslab's ms_condensing
	* flag to ensure that allocations on this metaslab do not occur
	* while we're in the middle of committing it to disk. This is only
	* critical for ms_allocatable as all other range trees use per TXG
	* views of their content.
	*/
	msp->ms_condensing = B_TRUE;

	mutex_exit(&msp->ms_lock);
	uint64_t object = space_map_object(msp->ms_sm);
	space_map_truncate(sm,
	spa_feature_is_enabled(spa, SPA_FEATURE_LOG_SPACEMAP) ?
	zfs_metaslab_sm_blksz_with_log : zfs_metaslab_sm_blksz_no_log, tx);

	/*
	* space_map_truncate() may have reallocated the spacemap object.
	* If so, update the vdev_ms_array.
	*/
	if (space_map_object(msp->ms_sm) != object) {
	object = space_map_object(msp->ms_sm);
	dmu_write(spa->spa_meta_objset,
	msp->ms_group->mg_vd->vdev_ms_array, sizeof (uint64_t) *
	msp->ms_id, sizeof (uint64_t), &object, tx);
	}

	/*
	* Note:
	* When the log space map feature is enabled, each space map will
	* always have ALLOCS followed by FREES for each sync pass. This is
	* typically true even when the log space map feature is disabled,
	* except from the case where a metaslab goes through metaslab_sync()
	* and gets condensed. In that case the metaslab's space map will have
	* ALLOCS followed by FREES (due to condensing) followed by ALLOCS
	* followed by FREES (due to space_map_write() in metaslab_sync()) for
	* sync pass 1.
	*/
	range_tree_t *tmp_tree = range_tree_create(NULL, type, NULL, start,
	shift);
	range_tree_add(tmp_tree, msp->ms_start, msp->ms_size);
	space_map_write(sm, tmp_tree, SM_ALLOC, SM_NO_VDEVID, tx);
	space_map_write(sm, msp->ms_allocatable, SM_FREE, SM_NO_VDEVID, tx);
	space_map_write(sm, condense_tree, SM_FREE, SM_NO_VDEVID, tx);

	range_tree_vacate(condense_tree, NULL, NULL);
	range_tree_destroy(condense_tree);
	range_tree_vacate(tmp_tree, NULL, NULL);
	range_tree_destroy(tmp_tree);
	mutex_enter(&msp->ms_lock);

	msp->ms_condensing = B_FALSE;
	metaslab_flush_update(msp, tx);
	}

	static void
	metaslab_unflushed_add(metaslab_t msp, dmu_tx_t tx)
	{
	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
	ASSERT(spa_syncing_log_sm(spa) != NULL);
	ASSERT(msp->ms_sm != NULL);
	ASSERT(range_tree_is_empty(msp->ms_unflushed_allocs));
	ASSERT(range_tree_is_empty(msp->ms_unflushed_frees));

	mutex_enter(&spa->spa_flushed_ms_lock);
	metaslab_set_unflushed_txg(msp, spa_syncing_txg(spa), tx);
	metaslab_set_unflushed_dirty(msp, B_TRUE);
	avl_add(&spa->spa_metaslabs_by_flushed, msp);
	mutex_exit(&spa->spa_flushed_ms_lock);

	spa_log_sm_increment_current_mscount(spa);
	spa_log_summary_add_flushed_metaslab(spa, B_TRUE);
	}

	void
	metaslab_unflushed_bump(metaslab_t msp, dmu_tx_t tx, boolean_t dirty)
	{
	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
	ASSERT(spa_syncing_log_sm(spa) != NULL);
	ASSERT(msp->ms_sm != NULL);
	ASSERT(metaslab_unflushed_txg(msp) != 0);
	ASSERT3P(avl_find(&spa->spa_metaslabs_by_flushed, msp, NULL), ==, msp);
	ASSERT(range_tree_is_empty(msp->ms_unflushed_allocs));
	ASSERT(range_tree_is_empty(msp->ms_unflushed_frees));

	VERIFY3U(tx->tx_txg, <=, spa_final_dirty_txg(spa));

	/* update metaslab's position in our flushing tree */
	uint64_t ms_prev_flushed_txg = metaslab_unflushed_txg(msp);
	boolean_t ms_prev_flushed_dirty = metaslab_unflushed_dirty(msp);
	mutex_enter(&spa->spa_flushed_ms_lock);
	avl_remove(&spa->spa_metaslabs_by_flushed, msp);
	metaslab_set_unflushed_txg(msp, spa_syncing_txg(spa), tx);
	metaslab_set_unflushed_dirty(msp, dirty);
	avl_add(&spa->spa_metaslabs_by_flushed, msp);
	mutex_exit(&spa->spa_flushed_ms_lock);

	/* update metaslab counts of spa_log_sm_t nodes */
	spa_log_sm_decrement_mscount(spa, ms_prev_flushed_txg);
	spa_log_sm_increment_current_mscount(spa);

	/* update log space map summary */
	spa_log_summary_decrement_mscount(spa, ms_prev_flushed_txg,
	ms_prev_flushed_dirty);
	spa_log_summary_add_flushed_metaslab(spa, dirty);

	/* cleanup obsolete logs if any */
	spa_cleanup_old_sm_logs(spa, tx);
	}

	/*
	* Called when the metaslab has been flushed (its own spacemap now reflects
	* all the contents of the pool-wide spacemap log). Updates the metaslab's
	* metadata and any pool-wide related log space map data (e.g. summary,
	* obsolete logs, etc..) to reflect that.
	*/
	static void
	metaslab_flush_update(metaslab_t msp, dmu_tx_t tx)
	{
	metaslab_group_t *mg = msp->ms_group;
	spa_t *spa = mg->mg_vd->vdev_spa;

	ASSERT(MUTEX_HELD(&msp->ms_lock));

	ASSERT3U(spa_sync_pass(spa), ==, 1);

	/*
	* Just because a metaslab got flushed, that doesn't mean that
	* it will pass through metaslab_sync_done(). Thus, make sure to
	* update ms_synced_length here in case it doesn't.
	*/
	msp->ms_synced_length = space_map_length(msp->ms_sm);

	/*
	* We may end up here from metaslab_condense() without the
	* feature being active. In that case this is a no-op.
	*/
	if (!spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP) \|\|
	metaslab_unflushed_txg(msp) == 0)
	return;

	metaslab_unflushed_bump(msp, tx, B_FALSE);
	}

	boolean_t
	metaslab_flush(metaslab_t msp, dmu_tx_t tx)
	{
	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	ASSERT3U(spa_sync_pass(spa), ==, 1);
	ASSERT(spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP));

	ASSERT(msp->ms_sm != NULL);
	ASSERT(metaslab_unflushed_txg(msp) != 0);
	ASSERT(avl_find(&spa->spa_metaslabs_by_flushed, msp, NULL) != NULL);

	/*
	* There is nothing wrong with flushing the same metaslab twice, as
	* this codepath should work on that case. However, the current
	* flushing scheme makes sure to avoid this situation as we would be
	* making all these calls without having anything meaningful to write
	* to disk. We assert this behavior here.
	*/
	ASSERT3U(metaslab_unflushed_txg(msp), <, dmu_tx_get_txg(tx));

	/*
	* We can not flush while loading, because then we would
	* not load the ms_unflushed_{allocs,frees}.
	*/
	if (msp->ms_loading)
	return (B_FALSE);

	metaslab_verify_space(msp, dmu_tx_get_txg(tx));
	metaslab_verify_weight_and_frag(msp);

	/*
	* Metaslab condensing is effectively flushing. Therefore if the
	* metaslab can be condensed we can just condense it instead of
	* flushing it.
	*
	* Note that metaslab_condense() does call metaslab_flush_update()
	* so we can just return immediately after condensing. We also
	* don't need to care about setting ms_flushing or broadcasting
	* ms_flush_cv, even if we temporarily drop the ms_lock in
	* metaslab_condense(), as the metaslab is already loaded.
	*/
	if (msp->ms_loaded && metaslab_should_condense(msp)) {
	metaslab_group_t *mg = msp->ms_group;

	/*
	* For all histogram operations below refer to the
	* comments of metaslab_sync() where we follow a
	* similar procedure.
	*/
	metaslab_group_histogram_verify(mg);
	metaslab_class_histogram_verify(mg->mg_class);
	metaslab_group_histogram_remove(mg, msp);

	metaslab_condense(msp, tx);

	space_map_histogram_clear(msp->ms_sm);
	space_map_histogram_add(msp->ms_sm, msp->ms_allocatable, tx);
	ASSERT(range_tree_is_empty(msp->ms_freed));
	for (int t = 0; t < TXG_DEFER_SIZE; t++) {
	space_map_histogram_add(msp->ms_sm,
	msp->ms_defer[t], tx);
	}
	metaslab_aux_histograms_update(msp);

	metaslab_group_histogram_add(mg, msp);
	metaslab_group_histogram_verify(mg);
	metaslab_class_histogram_verify(mg->mg_class);

	metaslab_verify_space(msp, dmu_tx_get_txg(tx));

	/*
	* Since we recreated the histogram (and potentially
	* the ms_sm too while condensing) ensure that the
	* weight is updated too because we are not guaranteed
	* that this metaslab is dirty and will go through
	* metaslab_sync_done().
	*/
	metaslab_recalculate_weight_and_sort(msp);
	return (B_TRUE);
	}

	msp->ms_flushing = B_TRUE;
	uint64_t sm_len_before = space_map_length(msp->ms_sm);

	mutex_exit(&msp->ms_lock);
	space_map_write(msp->ms_sm, msp->ms_unflushed_allocs, SM_ALLOC,
	SM_NO_VDEVID, tx);
	space_map_write(msp->ms_sm, msp->ms_unflushed_frees, SM_FREE,
	SM_NO_VDEVID, tx);
	mutex_enter(&msp->ms_lock);

	uint64_t sm_len_after = space_map_length(msp->ms_sm);
	if (zfs_flags & ZFS_DEBUG_LOG_SPACEMAP) {
	zfs_dbgmsg("flushing: txg %llu, spa %s, vdev_id %llu, "
	"ms_id %llu, unflushed_allocs %llu, unflushed_frees %llu, "
	"appended %llu bytes", (u_longlong_t)dmu_tx_get_txg(tx),
	spa_name(spa),
	(u_longlong_t)msp->ms_group->mg_vd->vdev_id,
	(u_longlong_t)msp->ms_id,
	(u_longlong_t)range_tree_space(msp->ms_unflushed_allocs),
	(u_longlong_t)range_tree_space(msp->ms_unflushed_frees),
	(u_longlong_t)(sm_len_after - sm_len_before));
	}

	ASSERT3U(spa->spa_unflushed_stats.sus_memused, >=,
	metaslab_unflushed_changes_memused(msp));
	spa->spa_unflushed_stats.sus_memused -=
	metaslab_unflushed_changes_memused(msp);
	range_tree_vacate(msp->ms_unflushed_allocs, NULL, NULL);
	range_tree_vacate(msp->ms_unflushed_frees, NULL, NULL);

	metaslab_verify_space(msp, dmu_tx_get_txg(tx));
	metaslab_verify_weight_and_frag(msp);

	metaslab_flush_update(msp, tx);

	metaslab_verify_space(msp, dmu_tx_get_txg(tx));
	metaslab_verify_weight_and_frag(msp);

	msp->ms_flushing = B_FALSE;
	cv_broadcast(&msp->ms_flush_cv);
	return (B_TRUE);
	}

	/*
	* Write a metaslab to disk in the context of the specified transaction group.
	*/
	void
	metaslab_sync(metaslab_t *msp, uint64_t txg)
	{
	metaslab_group_t *mg = msp->ms_group;
	vdev_t *vd = mg->mg_vd;
	spa_t *spa = vd->vdev_spa;
	objset_t *mos = spa_meta_objset(spa);
	range_tree_t *alloctree = msp->ms_allocating[txg & TXG_MASK];
	dmu_tx_t *tx;

	ASSERT(!vd->vdev_ishole);

	/*
	* This metaslab has just been added so there's no work to do now.
	*/
	if (msp->ms_new) {
	ASSERT0(range_tree_space(alloctree));
	ASSERT0(range_tree_space(msp->ms_freeing));
	ASSERT0(range_tree_space(msp->ms_freed));
	ASSERT0(range_tree_space(msp->ms_checkpointing));
	ASSERT0(range_tree_space(msp->ms_trim));
	return;
	}

	/*
	* Normally, we don't want to process a metaslab if there are no
	* allocations or frees to perform. However, if the metaslab is being
	* forced to condense, it's loaded and we're not beyond the final
	* dirty txg, we need to let it through. Not condensing beyond the
	* final dirty txg prevents an issue where metaslabs that need to be
	* condensed but were loaded for other reasons could cause a panic
	* here. By only checking the txg in that branch of the conditional,
	* we preserve the utility of the VERIFY statements in all other
	* cases.
	*/
	if (range_tree_is_empty(alloctree) &&
	range_tree_is_empty(msp->ms_freeing) &&
	range_tree_is_empty(msp->ms_checkpointing) &&
	!(msp->ms_loaded && msp->ms_condense_wanted &&
	txg <= spa_final_dirty_txg(spa)))
	return;


	VERIFY3U(txg, <=, spa_final_dirty_txg(spa));

	/*
	* The only state that can actually be changing concurrently
	* with metaslab_sync() is the metaslab's ms_allocatable. No
	* other thread can be modifying this txg's alloc, freeing,
	* freed, or space_map_phys_t. We drop ms_lock whenever we
	* could call into the DMU, because the DMU can call down to
	* us (e.g. via zio_free()) at any time.
	*
	* The spa_vdev_remove_thread() can be reading metaslab state
	* concurrently, and it is locked out by the ms_sync_lock.
	* Note that the ms_lock is insufficient for this, because it
	* is dropped by space_map_write().
	*/
	tx = dmu_tx_create_assigned(spa_get_dsl(spa), txg);

	/*
	* Generate a log space map if one doesn't exist already.
	*/
	spa_generate_syncing_log_sm(spa, tx);

	if (msp->ms_sm == NULL) {
	uint64_t new_object = space_map_alloc(mos,
	spa_feature_is_enabled(spa, SPA_FEATURE_LOG_SPACEMAP) ?
	zfs_metaslab_sm_blksz_with_log :
	zfs_metaslab_sm_blksz_no_log, tx);
	VERIFY3U(new_object, !=, 0);

	dmu_write(mos, vd->vdev_ms_array, sizeof (uint64_t) *
	msp->ms_id, sizeof (uint64_t), &new_object, tx);

	VERIFY0(space_map_open(&msp->ms_sm, mos, new_object,
	msp->ms_start, msp->ms_size, vd->vdev_ashift));
	ASSERT(msp->ms_sm != NULL);

	ASSERT(range_tree_is_empty(msp->ms_unflushed_allocs));
	ASSERT(range_tree_is_empty(msp->ms_unflushed_frees));
	ASSERT0(metaslab_allocated_space(msp));
	}

	if (!range_tree_is_empty(msp->ms_checkpointing) &&
	vd->vdev_checkpoint_sm == NULL) {
	ASSERT(spa_has_checkpoint(spa));

	uint64_t new_object = space_map_alloc(mos,
	zfs_vdev_standard_sm_blksz, tx);
	VERIFY3U(new_object, !=, 0);

	VERIFY0(space_map_open(&vd->vdev_checkpoint_sm,
	mos, new_object, 0, vd->vdev_asize, vd->vdev_ashift));
	ASSERT3P(vd->vdev_checkpoint_sm, !=, NULL);

	/*
	* We save the space map object as an entry in vdev_top_zap
	* so it can be retrieved when the pool is reopened after an
	* export or through zdb.
	*/
	VERIFY0(zap_add(vd->vdev_spa->spa_meta_objset,
	vd->vdev_top_zap, VDEV_TOP_ZAP_POOL_CHECKPOINT_SM,
	sizeof (new_object), 1, &new_object, tx));
	}

	mutex_enter(&msp->ms_sync_lock);
	mutex_enter(&msp->ms_lock);

	/*
	* Note: metaslab_condense() clears the space map's histogram.
	* Therefore we must verify and remove this histogram before
	* condensing.
	*/
	metaslab_group_histogram_verify(mg);
	metaslab_class_histogram_verify(mg->mg_class);
	metaslab_group_histogram_remove(mg, msp);

	if (spa->spa_sync_pass == 1 && msp->ms_loaded &&
	metaslab_should_condense(msp))
	metaslab_condense(msp, tx);

	/*
	* We'll be going to disk to sync our space accounting, thus we
	* drop the ms_lock during that time so allocations coming from
	* open-context (ZIL) for future TXGs do not block.
	*/
	mutex_exit(&msp->ms_lock);
	space_map_t *log_sm = spa_syncing_log_sm(spa);
	if (log_sm != NULL) {
	ASSERT(spa_feature_is_enabled(spa, SPA_FEATURE_LOG_SPACEMAP));
	if (metaslab_unflushed_txg(msp) == 0)
	metaslab_unflushed_add(msp, tx);
	else if (!metaslab_unflushed_dirty(msp))
	metaslab_unflushed_bump(msp, tx, B_TRUE);

	space_map_write(log_sm, alloctree, SM_ALLOC,
	vd->vdev_id, tx);
	space_map_write(log_sm, msp->ms_freeing, SM_FREE,
	vd->vdev_id, tx);
	mutex_enter(&msp->ms_lock);

	ASSERT3U(spa->spa_unflushed_stats.sus_memused, >=,
	metaslab_unflushed_changes_memused(msp));
	spa->spa_unflushed_stats.sus_memused -=
	metaslab_unflushed_changes_memused(msp);
	range_tree_remove_xor_add(alloctree,
	msp->ms_unflushed_frees, msp->ms_unflushed_allocs);
	range_tree_remove_xor_add(msp->ms_freeing,
	msp->ms_unflushed_allocs, msp->ms_unflushed_frees);
	spa->spa_unflushed_stats.sus_memused +=
	metaslab_unflushed_changes_memused(msp);
	} else {
	ASSERT(!spa_feature_is_enabled(spa, SPA_FEATURE_LOG_SPACEMAP));

	space_map_write(msp->ms_sm, alloctree, SM_ALLOC,
	SM_NO_VDEVID, tx);
	space_map_write(msp->ms_sm, msp->ms_freeing, SM_FREE,
	SM_NO_VDEVID, tx);
	mutex_enter(&msp->ms_lock);
	}

	msp->ms_allocated_space += range_tree_space(alloctree);
	ASSERT3U(msp->ms_allocated_space, >=,
	range_tree_space(msp->ms_freeing));
	msp->ms_allocated_space -= range_tree_space(msp->ms_freeing);

	if (!range_tree_is_empty(msp->ms_checkpointing)) {
	ASSERT(spa_has_checkpoint(spa));
	ASSERT3P(vd->vdev_checkpoint_sm, !=, NULL);

	/*
	* Since we are doing writes to disk and the ms_checkpointing
	* tree won't be changing during that time, we drop the
	* ms_lock while writing to the checkpoint space map, for the
	* same reason mentioned above.
	*/
	mutex_exit(&msp->ms_lock);
	space_map_write(vd->vdev_checkpoint_sm,
	msp->ms_checkpointing, SM_FREE, SM_NO_VDEVID, tx);
	mutex_enter(&msp->ms_lock);

	spa->spa_checkpoint_info.sci_dspace +=
	range_tree_space(msp->ms_checkpointing);
	vd->vdev_stat.vs_checkpoint_space +=
	range_tree_space(msp->ms_checkpointing);
	ASSERT3U(vd->vdev_stat.vs_checkpoint_space, ==,
	-space_map_allocated(vd->vdev_checkpoint_sm));

	range_tree_vacate(msp->ms_checkpointing, NULL, NULL);
	}

	if (msp->ms_loaded) {
	/*
	* When the space map is loaded, we have an accurate
	* histogram in the range tree. This gives us an opportunity
	* to bring the space map's histogram up-to-date so we clear
	* it first before updating it.
	*/
	space_map_histogram_clear(msp->ms_sm);
	space_map_histogram_add(msp->ms_sm, msp->ms_allocatable, tx);

	/*
	* Since we've cleared the histogram we need to add back
	* any free space that has already been processed, plus
	* any deferred space. This allows the on-disk histogram
	* to accurately reflect all free space even if some space
	* is not yet available for allocation (i.e. deferred).
	*/
	space_map_histogram_add(msp->ms_sm, msp->ms_freed, tx);

	/*
	* Add back any deferred free space that has not been
	* added back into the in-core free tree yet. This will
	* ensure that we don't end up with a space map histogram
	* that is completely empty unless the metaslab is fully
	* allocated.
	*/
	for (int t = 0; t < TXG_DEFER_SIZE; t++) {
	space_map_histogram_add(msp->ms_sm,
	msp->ms_defer[t], tx);
	}
	}

	/*
	* Always add the free space from this sync pass to the space
	* map histogram. We want to make sure that the on-disk histogram
	* accounts for all free space. If the space map is not loaded,
	* then we will lose some accuracy but will correct it the next
	* time we load the space map.
	*/
	space_map_histogram_add(msp->ms_sm, msp->ms_freeing, tx);
	metaslab_aux_histograms_update(msp);

	metaslab_group_histogram_add(mg, msp);
	metaslab_group_histogram_verify(mg);
	metaslab_class_histogram_verify(mg->mg_class);

	/*
	* For sync pass 1, we avoid traversing this txg's free range tree
	* and instead will just swap the pointers for freeing and freed.
	* We can safely do this since the freed_tree is guaranteed to be
	* empty on the initial pass.
	*
	* Keep in mind that even if we are currently using a log spacemap
	* we want current frees to end up in the ms_allocatable (but not
	* get appended to the ms_sm) so their ranges can be reused as usual.
	*/
	if (spa_sync_pass(spa) == 1) {
	range_tree_swap(&msp->ms_freeing, &msp->ms_freed);
	ASSERT0(msp->ms_allocated_this_txg);
	} else {
	range_tree_vacate(msp->ms_freeing,
	range_tree_add, msp->ms_freed);
	}
	msp->ms_allocated_this_txg += range_tree_space(alloctree);
	range_tree_vacate(alloctree, NULL, NULL);

	ASSERT0(range_tree_space(msp->ms_allocating[txg & TXG_MASK]));
	ASSERT0(range_tree_space(msp->ms_allocating[TXG_CLEAN(txg)
	& TXG_MASK]));
	ASSERT0(range_tree_space(msp->ms_freeing));
	ASSERT0(range_tree_space(msp->ms_checkpointing));

	mutex_exit(&msp->ms_lock);

	/*
	* Verify that the space map object ID has been recorded in the
	* vdev_ms_array.
	*/
	uint64_t object;
	VERIFY0(dmu_read(mos, vd->vdev_ms_array,
	msp->ms_id * sizeof (uint64_t), sizeof (uint64_t), &object, 0));
	VERIFY3U(object, ==, space_map_object(msp->ms_sm));

	mutex_exit(&msp->ms_sync_lock);
	dmu_tx_commit(tx);
	}

	static void
	metaslab_evict(metaslab_t *msp, uint64_t txg)
	{
	if (!msp->ms_loaded \|\| msp->ms_disabled != 0)
	return;

	for (int t = 1; t < TXG_CONCURRENT_STATES; t++) {
	VERIFY0(range_tree_space(
	msp->ms_allocating[(txg + t) & TXG_MASK]));
	}
	if (msp->ms_allocator != -1)
	metaslab_passivate(msp, msp->ms_weight & ~METASLAB_ACTIVE_MASK);

	if (!metaslab_debug_unload)
	metaslab_unload(msp);
	}

	/*
	* Called after a transaction group has completely synced to mark
	* all of the metaslab's free space as usable.
	*/
	void
	metaslab_sync_done(metaslab_t *msp, uint64_t txg)
	{
	metaslab_group_t *mg = msp->ms_group;
	vdev_t *vd = mg->mg_vd;
	spa_t *spa = vd->vdev_spa;
	range_tree_t **defer_tree;
	int64_t alloc_delta, defer_delta;
	boolean_t defer_allowed = B_TRUE;

	ASSERT(!vd->vdev_ishole);

	mutex_enter(&msp->ms_lock);

	if (msp->ms_new) {
	/* this is a new metaslab, add its capacity to the vdev */
	metaslab_space_update(vd, mg->mg_class, 0, 0, msp->ms_size);

	/* there should be no allocations nor frees at this point */
	VERIFY0(msp->ms_allocated_this_txg);
	VERIFY0(range_tree_space(msp->ms_freed));
	}

	ASSERT0(range_tree_space(msp->ms_freeing));
	ASSERT0(range_tree_space(msp->ms_checkpointing));

	defer_tree = &msp->ms_defer[txg % TXG_DEFER_SIZE];

	uint64_t free_space = metaslab_class_get_space(spa_normal_class(spa)) -
	metaslab_class_get_alloc(spa_normal_class(spa));
	if (free_space <= spa_get_slop_space(spa) \|\| vd->vdev_removing \|\|
	vd->vdev_rz_expanding) {
	defer_allowed = B_FALSE;
	}

	defer_delta = 0;
	alloc_delta = msp->ms_allocated_this_txg -
	range_tree_space(msp->ms_freed);

	if (defer_allowed) {
	defer_delta = range_tree_space(msp->ms_freed) -
	range_tree_space(*defer_tree);
	} else {
	defer_delta -= range_tree_space(*defer_tree);
	}
	metaslab_space_update(vd, mg->mg_class, alloc_delta + defer_delta,
	defer_delta, 0);

	if (spa_syncing_log_sm(spa) == NULL) {
	/*
	* If there's a metaslab_load() in progress and we don't have
	* a log space map, it means that we probably wrote to the
	* metaslab's space map. If this is the case, we need to
	* make sure that we wait for the load to complete so that we
	* have a consistent view at the in-core side of the metaslab.
	*/
	metaslab_load_wait(msp);
	} else {
	ASSERT(spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP));
	}

	/*
	* When auto-trimming is enabled, free ranges which are added to
	* ms_allocatable are also be added to ms_trim. The ms_trim tree is
	* periodically consumed by the vdev_autotrim_thread() which issues
	* trims for all ranges and then vacates the tree. The ms_trim tree
	* can be discarded at any time with the sole consequence of recent
	* frees not being trimmed.
	*/
	if (spa_get_autotrim(spa) == SPA_AUTOTRIM_ON) {
	range_tree_walk(*defer_tree, range_tree_add, msp->ms_trim);
	if (!defer_allowed) {
	range_tree_walk(msp->ms_freed, range_tree_add,
	msp->ms_trim);
	}
	} else {
	range_tree_vacate(msp->ms_trim, NULL, NULL);
	}

	/*
	* Move the frees from the defer_tree back to the free
	* range tree (if it's loaded). Swap the freed_tree and
	* the defer_tree -- this is safe to do because we've
	* just emptied out the defer_tree.
	*/
	range_tree_vacate(*defer_tree,
	msp->ms_loaded ? range_tree_add : NULL, msp->ms_allocatable);
	if (defer_allowed) {
	range_tree_swap(&msp->ms_freed, defer_tree);
	} else {
	range_tree_vacate(msp->ms_freed,
	msp->ms_loaded ? range_tree_add : NULL,
	msp->ms_allocatable);
	}

	msp->ms_synced_length = space_map_length(msp->ms_sm);

	msp->ms_deferspace += defer_delta;
	ASSERT3S(msp->ms_deferspace, >=, 0);
	ASSERT3S(msp->ms_deferspace, <=, msp->ms_size);
	if (msp->ms_deferspace != 0) {
	/*
	* Keep syncing this metaslab until all deferred frees
	* are back in circulation.
	*/
	vdev_dirty(vd, VDD_METASLAB, msp, txg + 1);
	}
	metaslab_aux_histograms_update_done(msp, defer_allowed);

	if (msp->ms_new) {
	msp->ms_new = B_FALSE;
	mutex_enter(&mg->mg_lock);
	mg->mg_ms_ready++;
	mutex_exit(&mg->mg_lock);
	}

	/*
	* Re-sort metaslab within its group now that we've adjusted
	* its allocatable space.
	*/
	metaslab_recalculate_weight_and_sort(msp);

	ASSERT0(range_tree_space(msp->ms_allocating[txg & TXG_MASK]));
	ASSERT0(range_tree_space(msp->ms_freeing));
	ASSERT0(range_tree_space(msp->ms_freed));
	ASSERT0(range_tree_space(msp->ms_checkpointing));
	msp->ms_allocating_total -= msp->ms_allocated_this_txg;
	msp->ms_allocated_this_txg = 0;
	mutex_exit(&msp->ms_lock);
	}

	void
	metaslab_sync_reassess(metaslab_group_t *mg)
	{
	spa_t *spa = mg->mg_class->mc_spa;

	spa_config_enter(spa, SCL_ALLOC, FTAG, RW_READER);
	metaslab_group_alloc_update(mg);
	mg->mg_fragmentation = metaslab_group_fragmentation(mg);

	/*
	* Preload the next potential metaslabs but only on active
	* metaslab groups. We can get into a state where the metaslab
	* is no longer active since we dirty metaslabs as we remove a
	* a device, thus potentially making the metaslab group eligible
	* for preloading.
	*/
	if (mg->mg_activation_count > 0) {
	metaslab_group_preload(mg);
	}
	spa_config_exit(spa, SCL_ALLOC, FTAG);
	}

	/*
	* When writing a ditto block (i.e. more than one DVA for a given BP) on
	* the same vdev as an existing DVA of this BP, then try to allocate it
	* on a different metaslab than existing DVAs (i.e. a unique metaslab).
	*/
	static boolean_t
	metaslab_is_unique(metaslab_t msp, dva_t dva)
	{
	uint64_t dva_ms_id;

	if (DVA_GET_ASIZE(dva) == 0)
	return (B_TRUE);

	if (msp->ms_group->mg_vd->vdev_id != DVA_GET_VDEV(dva))
	return (B_TRUE);

	dva_ms_id = DVA_GET_OFFSET(dva) >> msp->ms_group->mg_vd->vdev_ms_shift;

	return (msp->ms_id != dva_ms_id);
	}

	/*
	* ==========================================================================
	* Metaslab allocation tracing facility
	* ==========================================================================
	*/

	/*
	* Add an allocation trace element to the allocation tracing list.
	*/
	static void
	metaslab_trace_add(zio_alloc_list_t zal, metaslab_group_t mg,
	metaslab_t *msp, uint64_t psize, uint32_t dva_id, uint64_t offset,
	int allocator)
	{
	metaslab_alloc_trace_t *mat;

	if (!metaslab_trace_enabled)
	return;

	/*
	* When the tracing list reaches its maximum we remove
	* the second element in the list before adding a new one.
	* By removing the second element we preserve the original
	* entry as a clue to what allocations steps have already been
	* performed.
	*/
	if (zal->zal_size == metaslab_trace_max_entries) {
	metaslab_alloc_trace_t *mat_next;
	#ifdef ZFS_DEBUG
	panic("too many entries in allocation list");
	#endif
	METASLABSTAT_BUMP(metaslabstat_trace_over_limit);
	zal->zal_size--;
	mat_next = list_next(&zal->zal_list, list_head(&zal->zal_list));
	list_remove(&zal->zal_list, mat_next);
	kmem_cache_free(metaslab_alloc_trace_cache, mat_next);
	}

	mat = kmem_cache_alloc(metaslab_alloc_trace_cache, KM_SLEEP);
	list_link_init(&mat->mat_list_node);
	mat->mat_mg = mg;
	mat->mat_msp = msp;
	mat->mat_size = psize;
	mat->mat_dva_id = dva_id;
	mat->mat_offset = offset;
	mat->mat_weight = 0;
	mat->mat_allocator = allocator;

	if (msp != NULL)
	mat->mat_weight = msp->ms_weight;

	/*
	* The list is part of the zio so locking is not required. Only
	* a single thread will perform allocations for a given zio.
	*/
	list_insert_tail(&zal->zal_list, mat);
	zal->zal_size++;

	ASSERT3U(zal->zal_size, <=, metaslab_trace_max_entries);
	}

	void
	metaslab_trace_init(zio_alloc_list_t *zal)
	{
	list_create(&zal->zal_list, sizeof (metaslab_alloc_trace_t),
	offsetof(metaslab_alloc_trace_t, mat_list_node));
	zal->zal_size = 0;
	}

	void
	metaslab_trace_fini(zio_alloc_list_t *zal)
	{
	metaslab_alloc_trace_t *mat;

	while ((mat = list_remove_head(&zal->zal_list)) != NULL)
	kmem_cache_free(metaslab_alloc_trace_cache, mat);
	list_destroy(&zal->zal_list);
	zal->zal_size = 0;
	}

	/*
	* ==========================================================================
	* Metaslab block operations
	* ==========================================================================
	*/

	static void
	metaslab_group_alloc_increment(spa_t spa, uint64_t vdev, const void tag,
	int flags, int allocator)
	{
	if (!(flags & METASLAB_ASYNC_ALLOC) \|\|
	(flags & METASLAB_DONT_THROTTLE))
	return;

	metaslab_group_t *mg = vdev_lookup_top(spa, vdev)->vdev_mg;
	if (!mg->mg_class->mc_alloc_throttle_enabled)
	return;

	metaslab_group_allocator_t *mga = &mg->mg_allocator[allocator];
	(void) zfs_refcount_add(&mga->mga_alloc_queue_depth, tag);
	}

	static void
	metaslab_group_increment_qdepth(metaslab_group_t *mg, int allocator)
	{
	metaslab_group_allocator_t *mga = &mg->mg_allocator[allocator];
	metaslab_class_allocator_t *mca =
	&mg->mg_class->mc_allocator[allocator];
	uint64_t max = mg->mg_max_alloc_queue_depth;
	uint64_t cur = mga->mga_cur_max_alloc_queue_depth;
	while (cur < max) {
	if (atomic_cas_64(&mga->mga_cur_max_alloc_queue_depth,
	cur, cur + 1) == cur) {
	atomic_inc_64(&mca->mca_alloc_max_slots);
	return;
	}
	cur = mga->mga_cur_max_alloc_queue_depth;
	}
	}

	void
	metaslab_group_alloc_decrement(spa_t spa, uint64_t vdev, const void tag,
	int flags, int allocator, boolean_t io_complete)
	{
	if (!(flags & METASLAB_ASYNC_ALLOC) \|\|
	(flags & METASLAB_DONT_THROTTLE))
	return;

	metaslab_group_t *mg = vdev_lookup_top(spa, vdev)->vdev_mg;
	if (!mg->mg_class->mc_alloc_throttle_enabled)
	return;

	metaslab_group_allocator_t *mga = &mg->mg_allocator[allocator];
	(void) zfs_refcount_remove(&mga->mga_alloc_queue_depth, tag);
	if (io_complete)
	metaslab_group_increment_qdepth(mg, allocator);
	}

	void
	metaslab_group_alloc_verify(spa_t spa, const blkptr_t bp, const void *tag,
	int allocator)
	{
	#ifdef ZFS_DEBUG
	const dva_t *dva = bp->blk_dva;
	int ndvas = BP_GET_NDVAS(bp);

	for (int d = 0; d < ndvas; d++) {
	uint64_t vdev = DVA_GET_VDEV(&dva[d]);
	metaslab_group_t *mg = vdev_lookup_top(spa, vdev)->vdev_mg;
	metaslab_group_allocator_t *mga = &mg->mg_allocator[allocator];
	VERIFY(zfs_refcount_not_held(&mga->mga_alloc_queue_depth, tag));
	}
	#endif
	}

	static uint64_t
	metaslab_block_alloc(metaslab_t *msp, uint64_t size, uint64_t txg)
	{
	uint64_t start;
	range_tree_t *rt = msp->ms_allocatable;
	metaslab_class_t *mc = msp->ms_group->mg_class;

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	VERIFY(!msp->ms_condensing);
	VERIFY0(msp->ms_disabled);
	VERIFY0(msp->ms_new);

	start = mc->mc_ops->msop_alloc(msp, size);
	if (start != -1ULL) {
	metaslab_group_t *mg = msp->ms_group;
	vdev_t *vd = mg->mg_vd;

	VERIFY0(P2PHASE(start, 1ULL << vd->vdev_ashift));
	VERIFY0(P2PHASE(size, 1ULL << vd->vdev_ashift));
	VERIFY3U(range_tree_space(rt) - size, <=, msp->ms_size);
	range_tree_remove(rt, start, size);
	range_tree_clear(msp->ms_trim, start, size);

	if (range_tree_is_empty(msp->ms_allocating[txg & TXG_MASK]))
	vdev_dirty(mg->mg_vd, VDD_METASLAB, msp, txg);

	range_tree_add(msp->ms_allocating[txg & TXG_MASK], start, size);
	msp->ms_allocating_total += size;

	/* Track the last successful allocation */
	msp->ms_alloc_txg = txg;
	metaslab_verify_space(msp, txg);
	}

	/*
	* Now that we've attempted the allocation we need to update the
	* metaslab's maximum block size since it may have changed.
	*/
	msp->ms_max_size = metaslab_largest_allocatable(msp);
	return (start);
	}

	/*
	* Find the metaslab with the highest weight that is less than what we've
	* already tried. In the common case, this means that we will examine each
	* metaslab at most once. Note that concurrent callers could reorder metaslabs
	* by activation/passivation once we have dropped the mg_lock. If a metaslab is
	* activated by another thread, and we fail to allocate from the metaslab we
	* have selected, we may not try the newly-activated metaslab, and instead
	* activate another metaslab. This is not optimal, but generally does not cause
	* any problems (a possible exception being if every metaslab is completely full
	* except for the newly-activated metaslab which we fail to examine).
	*/
	static metaslab_t *
	find_valid_metaslab(metaslab_group_t *mg, uint64_t activation_weight,
	dva_t *dva, int d, boolean_t want_unique, uint64_t asize, int allocator,
	boolean_t try_hard, zio_alloc_list_t zal, metaslab_t search,
	boolean_t *was_active)
	{
	avl_index_t idx;
	avl_tree_t *t = &mg->mg_metaslab_tree;
	metaslab_t *msp = avl_find(t, search, &idx);
	if (msp == NULL)
	msp = avl_nearest(t, idx, AVL_AFTER);

	uint_t tries = 0;
	for (; msp != NULL; msp = AVL_NEXT(t, msp)) {
	int i;

	if (!try_hard && tries > zfs_metaslab_find_max_tries) {
	METASLABSTAT_BUMP(metaslabstat_too_many_tries);
	return (NULL);
	}
	tries++;

	if (!metaslab_should_allocate(msp, asize, try_hard)) {
	metaslab_trace_add(zal, mg, msp, asize, d,
	TRACE_TOO_SMALL, allocator);
	continue;
	}

	/*
	* If the selected metaslab is condensing or disabled, or
	* hasn't gone through a metaslab_sync_done(), then skip it.
	*/
	if (msp->ms_condensing \|\| msp->ms_disabled > 0 \|\| msp->ms_new)
	continue;

	*was_active = msp->ms_allocator != -1;
	/*
	* If we're activating as primary, this is our first allocation
	* from this disk, so we don't need to check how close we are.
	* If the metaslab under consideration was already active,
	* we're getting desperate enough to steal another allocator's
	* metaslab, so we still don't care about distances.
	*/
	if (activation_weight == METASLAB_WEIGHT_PRIMARY \|\| *was_active)
	break;

	for (i = 0; i < d; i++) {
	if (want_unique &&
	!metaslab_is_unique(msp, &dva[i]))
	break; /* try another metaslab */
	}
	if (i == d)
	break;
	}

	if (msp != NULL) {
	search->ms_weight = msp->ms_weight;
	search->ms_start = msp->ms_start + 1;
	search->ms_allocator = msp->ms_allocator;
	search->ms_primary = msp->ms_primary;
	}
	return (msp);
	}

	static void
	metaslab_active_mask_verify(metaslab_t *msp)
	{
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	if ((zfs_flags & ZFS_DEBUG_METASLAB_VERIFY) == 0)
	return;

	if ((msp->ms_weight & METASLAB_ACTIVE_MASK) == 0)
	return;

	if (msp->ms_weight & METASLAB_WEIGHT_PRIMARY) {
	VERIFY0(msp->ms_weight & METASLAB_WEIGHT_SECONDARY);
	VERIFY0(msp->ms_weight & METASLAB_WEIGHT_CLAIM);
	VERIFY3S(msp->ms_allocator, !=, -1);
	VERIFY(msp->ms_primary);
	return;
	}

	if (msp->ms_weight & METASLAB_WEIGHT_SECONDARY) {
	VERIFY0(msp->ms_weight & METASLAB_WEIGHT_PRIMARY);
	VERIFY0(msp->ms_weight & METASLAB_WEIGHT_CLAIM);
	VERIFY3S(msp->ms_allocator, !=, -1);
	VERIFY(!msp->ms_primary);
	return;
	}

	if (msp->ms_weight & METASLAB_WEIGHT_CLAIM) {
	VERIFY0(msp->ms_weight & METASLAB_WEIGHT_PRIMARY);
	VERIFY0(msp->ms_weight & METASLAB_WEIGHT_SECONDARY);
	VERIFY3S(msp->ms_allocator, ==, -1);
	return;
	}
	}

	static uint64_t
	metaslab_group_alloc_normal(metaslab_group_t mg, zio_alloc_list_t zal,
	uint64_t asize, uint64_t txg, boolean_t want_unique, dva_t *dva, int d,
	int allocator, boolean_t try_hard)
	{
	metaslab_t *msp = NULL;
	uint64_t offset = -1ULL;

	uint64_t activation_weight = METASLAB_WEIGHT_PRIMARY;
	for (int i = 0; i < d; i++) {
	if (activation_weight == METASLAB_WEIGHT_PRIMARY &&
	DVA_GET_VDEV(&dva[i]) == mg->mg_vd->vdev_id) {
	activation_weight = METASLAB_WEIGHT_SECONDARY;
	} else if (activation_weight == METASLAB_WEIGHT_SECONDARY &&
	DVA_GET_VDEV(&dva[i]) == mg->mg_vd->vdev_id) {
	activation_weight = METASLAB_WEIGHT_CLAIM;
	break;
	}
	}

	/*
	* If we don't have enough metaslabs active to fill the entire array, we
	* just use the 0th slot.
	*/
	if (mg->mg_ms_ready < mg->mg_allocators * 3)
	allocator = 0;
	metaslab_group_allocator_t *mga = &mg->mg_allocator[allocator];

	ASSERT3U(mg->mg_vd->vdev_ms_count, >=, 2);

	metaslab_t search = kmem_alloc(sizeof (search), KM_SLEEP);
	search->ms_weight = UINT64_MAX;
	search->ms_start = 0;
	/*
	* At the end of the metaslab tree are the already-active metaslabs,
	* first the primaries, then the secondaries. When we resume searching
	* through the tree, we need to consider ms_allocator and ms_primary so
	* we start in the location right after where we left off, and don't
	* accidentally loop forever considering the same metaslabs.
	*/
	search->ms_allocator = -1;
	search->ms_primary = B_TRUE;
	for (;;) {
	boolean_t was_active = B_FALSE;

	mutex_enter(&mg->mg_lock);

	if (activation_weight == METASLAB_WEIGHT_PRIMARY &&
	mga->mga_primary != NULL) {
	msp = mga->mga_primary;

	/*
	* Even though we don't hold the ms_lock for the
	* primary metaslab, those fields should not
	* change while we hold the mg_lock. Thus it is
	* safe to make assertions on them.
	*/
	ASSERT(msp->ms_primary);
	ASSERT3S(msp->ms_allocator, ==, allocator);
	ASSERT(msp->ms_loaded);

	was_active = B_TRUE;
	ASSERT(msp->ms_weight & METASLAB_ACTIVE_MASK);
	} else if (activation_weight == METASLAB_WEIGHT_SECONDARY &&
	mga->mga_secondary != NULL) {
	msp = mga->mga_secondary;

	/*
	* See comment above about the similar assertions
	* for the primary metaslab.
	*/
	ASSERT(!msp->ms_primary);
	ASSERT3S(msp->ms_allocator, ==, allocator);
	ASSERT(msp->ms_loaded);

	was_active = B_TRUE;
	ASSERT(msp->ms_weight & METASLAB_ACTIVE_MASK);
	} else {
	msp = find_valid_metaslab(mg, activation_weight, dva, d,
	want_unique, asize, allocator, try_hard, zal,
	search, &was_active);
	}

	mutex_exit(&mg->mg_lock);
	if (msp == NULL) {
	kmem_free(search, sizeof (*search));
	return (-1ULL);
	}
	mutex_enter(&msp->ms_lock);

	metaslab_active_mask_verify(msp);

	/*
	* This code is disabled out because of issues with
	* tracepoints in non-gpl kernel modules.
	*/
	#if 0
	DTRACE_PROBE3(ms__activation__attempt,
	metaslab_t *, msp, uint64_t, activation_weight,
	boolean_t, was_active);
	#endif

	/*
	* Ensure that the metaslab we have selected is still
	* capable of handling our request. It's possible that
	* another thread may have changed the weight while we
	* were blocked on the metaslab lock. We check the
	* active status first to see if we need to set_selected_txg
	* a new metaslab.
	*/
	if (was_active && !(msp->ms_weight & METASLAB_ACTIVE_MASK)) {
	ASSERT3S(msp->ms_allocator, ==, -1);
	mutex_exit(&msp->ms_lock);
	continue;
	}

	/*
	* If the metaslab was activated for another allocator
	* while we were waiting in the ms_lock above, or it's
	* a primary and we're seeking a secondary (or vice versa),
	* we go back and select a new metaslab.
	*/
	if (!was_active && (msp->ms_weight & METASLAB_ACTIVE_MASK) &&
	(msp->ms_allocator != -1) &&
	(msp->ms_allocator != allocator \|\| ((activation_weight ==
	METASLAB_WEIGHT_PRIMARY) != msp->ms_primary))) {
	ASSERT(msp->ms_loaded);
	ASSERT((msp->ms_weight & METASLAB_WEIGHT_CLAIM) \|\|
	msp->ms_allocator != -1);
	mutex_exit(&msp->ms_lock);
	continue;
	}

	/*
	* This metaslab was used for claiming regions allocated
	* by the ZIL during pool import. Once these regions are
	* claimed we don't need to keep the CLAIM bit set
	* anymore. Passivate this metaslab to zero its activation
	* mask.
	*/
	if (msp->ms_weight & METASLAB_WEIGHT_CLAIM &&
	activation_weight != METASLAB_WEIGHT_CLAIM) {
	ASSERT(msp->ms_loaded);
	ASSERT3S(msp->ms_allocator, ==, -1);
	metaslab_passivate(msp, msp->ms_weight &
	~METASLAB_WEIGHT_CLAIM);
	mutex_exit(&msp->ms_lock);
	continue;
	}

	metaslab_set_selected_txg(msp, txg);

	int activation_error =
	metaslab_activate(msp, allocator, activation_weight);
	metaslab_active_mask_verify(msp);

	/*
	* If the metaslab was activated by another thread for
	* another allocator or activation_weight (EBUSY), or it
	* failed because another metaslab was assigned as primary
	* for this allocator (EEXIST) we continue using this
	* metaslab for our allocation, rather than going on to a
	* worse metaslab (we waited for that metaslab to be loaded
	* after all).
	*
	* If the activation failed due to an I/O error or ENOSPC we
	* skip to the next metaslab.
	*/
	boolean_t activated;
	if (activation_error == 0) {
	activated = B_TRUE;
	} else if (activation_error == EBUSY \|\|
	activation_error == EEXIST) {
	activated = B_FALSE;
	} else {
	mutex_exit(&msp->ms_lock);
	continue;
	}
	ASSERT(msp->ms_loaded);

	/*
	* Now that we have the lock, recheck to see if we should
	* continue to use this metaslab for this allocation. The
	* the metaslab is now loaded so metaslab_should_allocate()
	* can accurately determine if the allocation attempt should
	* proceed.
	*/
	if (!metaslab_should_allocate(msp, asize, try_hard)) {
	/* Passivate this metaslab and select a new one. */
	metaslab_trace_add(zal, mg, msp, asize, d,
	TRACE_TOO_SMALL, allocator);
	goto next;
	}

	/*
	* If this metaslab is currently condensing then pick again
	* as we can't manipulate this metaslab until it's committed
	* to disk. If this metaslab is being initialized, we shouldn't
	* allocate from it since the allocated region might be
	* overwritten after allocation.
	*/
	if (msp->ms_condensing) {
	metaslab_trace_add(zal, mg, msp, asize, d,
	TRACE_CONDENSING, allocator);
	if (activated) {
	metaslab_passivate(msp, msp->ms_weight &
	~METASLAB_ACTIVE_MASK);
	}
	mutex_exit(&msp->ms_lock);
	continue;
	} else if (msp->ms_disabled > 0) {
	metaslab_trace_add(zal, mg, msp, asize, d,
	TRACE_DISABLED, allocator);
	if (activated) {
	metaslab_passivate(msp, msp->ms_weight &
	~METASLAB_ACTIVE_MASK);
	}
	mutex_exit(&msp->ms_lock);
	continue;
	}

	offset = metaslab_block_alloc(msp, asize, txg);
	metaslab_trace_add(zal, mg, msp, asize, d, offset, allocator);

	if (offset != -1ULL) {
	/* Proactively passivate the metaslab, if needed */
	if (activated)
	metaslab_segment_may_passivate(msp);
	break;
	}
	next:
	ASSERT(msp->ms_loaded);

	/*
	* This code is disabled out because of issues with
	* tracepoints in non-gpl kernel modules.
	*/
	#if 0
	DTRACE_PROBE2(ms__alloc__failure, metaslab_t *, msp,
	uint64_t, asize);
	#endif

	/*
	* We were unable to allocate from this metaslab so determine
	* a new weight for this metaslab. Now that we have loaded
	* the metaslab we can provide a better hint to the metaslab
	* selector.
	*
	* For space-based metaslabs, we use the maximum block size.
	* This information is only available when the metaslab
	* is loaded and is more accurate than the generic free
	* space weight that was calculated by metaslab_weight().
	* This information allows us to quickly compare the maximum
	* available allocation in the metaslab to the allocation
	* size being requested.
	*
	* For segment-based metaslabs, determine the new weight
	* based on the highest bucket in the range tree. We
	* explicitly use the loaded segment weight (i.e. the range
	* tree histogram) since it contains the space that is
	* currently available for allocation and is accurate
	* even within a sync pass.
	*/
	uint64_t weight;
	if (WEIGHT_IS_SPACEBASED(msp->ms_weight)) {
	weight = metaslab_largest_allocatable(msp);
	WEIGHT_SET_SPACEBASED(weight);
	} else {
	weight = metaslab_weight_from_range_tree(msp);
	}

	if (activated) {
	metaslab_passivate(msp, weight);
	} else {
	/*
	* For the case where we use the metaslab that is
	* active for another allocator we want to make
	* sure that we retain the activation mask.
	*
	* Note that we could attempt to use something like
	* metaslab_recalculate_weight_and_sort() that
	* retains the activation mask here. That function
	* uses metaslab_weight() to set the weight though
	* which is not as accurate as the calculations
	* above.
	*/
	weight \|= msp->ms_weight & METASLAB_ACTIVE_MASK;
	metaslab_group_sort(mg, msp, weight);
	}
	metaslab_active_mask_verify(msp);

	/*
	* We have just failed an allocation attempt, check
	* that metaslab_should_allocate() agrees. Otherwise,
	* we may end up in an infinite loop retrying the same
	* metaslab.
	*/
	ASSERT(!metaslab_should_allocate(msp, asize, try_hard));

	mutex_exit(&msp->ms_lock);
	}
	mutex_exit(&msp->ms_lock);
	kmem_free(search, sizeof (*search));
	return (offset);
	}

	static uint64_t
	metaslab_group_alloc(metaslab_group_t mg, zio_alloc_list_t zal,
	uint64_t asize, uint64_t txg, boolean_t want_unique, dva_t *dva, int d,
	int allocator, boolean_t try_hard)
	{
	uint64_t offset;

	offset = metaslab_group_alloc_normal(mg, zal, asize, txg, want_unique,
	dva, d, allocator, try_hard);

	mutex_enter(&mg->mg_lock);
	if (offset == -1ULL) {
	mg->mg_failed_allocations++;
	metaslab_trace_add(zal, mg, NULL, asize, d,
	TRACE_GROUP_FAILURE, allocator);
	if (asize == SPA_GANGBLOCKSIZE) {
	/*
	* This metaslab group was unable to allocate
	* the minimum gang block size so it must be out of
	* space. We must notify the allocation throttle
	* to start skipping allocation attempts to this
	* metaslab group until more space becomes available.
	* Note: this failure cannot be caused by the
	* allocation throttle since the allocation throttle
	* is only responsible for skipping devices and
	* not failing block allocations.
	*/
	mg->mg_no_free_space = B_TRUE;
	}
	}
	mg->mg_allocations++;
	mutex_exit(&mg->mg_lock);
	return (offset);
	}

	/*
	* Allocate a block for the specified i/o.
	*/
	int
	metaslab_alloc_dva(spa_t spa, metaslab_class_t mc, uint64_t psize,
	dva_t dva, int d, dva_t hintdva, uint64_t txg, int flags,
	zio_alloc_list_t *zal, int allocator)
	{
	metaslab_class_allocator_t *mca = &mc->mc_allocator[allocator];
	metaslab_group_t mg, rotor;
	vdev_t *vd;
	boolean_t try_hard = B_FALSE;

	ASSERT(!DVA_IS_VALID(&dva[d]));

	/*
	* For testing, make some blocks above a certain size be gang blocks.
	* This will result in more split blocks when using device removal,
	* and a large number of split blocks coupled with ztest-induced
	* damage can result in extremely long reconstruction times. This
	* will also test spilling from special to normal.
	*/
	if (psize >= metaslab_force_ganging &&
	metaslab_force_ganging_pct > 0 &&
	(random_in_range(100) < MIN(metaslab_force_ganging_pct, 100))) {
	metaslab_trace_add(zal, NULL, NULL, psize, d, TRACE_FORCE_GANG,
	allocator);
	return (SET_ERROR(ENOSPC));
	}

	/*
	* Start at the rotor and loop through all mgs until we find something.
	* Note that there's no locking on mca_rotor or mca_aliquot because
	* nothing actually breaks if we miss a few updates -- we just won't
	* allocate quite as evenly. It all balances out over time.
	*
	* If we are doing ditto or log blocks, try to spread them across
	* consecutive vdevs. If we're forced to reuse a vdev before we've
	* allocated all of our ditto blocks, then try and spread them out on
	* that vdev as much as possible. If it turns out to not be possible,
	* gradually lower our standards until anything becomes acceptable.
	* Also, allocating on consecutive vdevs (as opposed to random vdevs)
	* gives us hope of containing our fault domains to something we're
	* able to reason about. Otherwise, any two top-level vdev failures
	* will guarantee the loss of data. With consecutive allocation,
	* only two adjacent top-level vdev failures will result in data loss.
	*
	* If we are doing gang blocks (hintdva is non-NULL), try to keep
	* ourselves on the same vdev as our gang block header. That
	* way, we can hope for locality in vdev_cache, plus it makes our
	* fault domains something tractable.
	*/
	if (hintdva) {
	vd = vdev_lookup_top(spa, DVA_GET_VDEV(&hintdva[d]));

	/*
	* It's possible the vdev we're using as the hint no
	* longer exists or its mg has been closed (e.g. by
	* device removal). Consult the rotor when
	* all else fails.
	*/
	if (vd != NULL && vd->vdev_mg != NULL) {
	mg = vdev_get_mg(vd, mc);

	if (flags & METASLAB_HINTBP_AVOID)
	mg = mg->mg_next;
	} else {
	mg = mca->mca_rotor;
	}
	} else if (d != 0) {
	vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[d - 1]));
	mg = vd->vdev_mg->mg_next;
	} else {
	ASSERT(mca->mca_rotor != NULL);
	mg = mca->mca_rotor;
	}

	/*
	* If the hint put us into the wrong metaslab class, or into a
	* metaslab group that has been passivated, just follow the rotor.
	*/
	if (mg->mg_class != mc \|\| mg->mg_activation_count <= 0)
	mg = mca->mca_rotor;

	rotor = mg;
	top:
	do {
	boolean_t allocatable;

	ASSERT(mg->mg_activation_count == 1);
	vd = mg->mg_vd;

	/*
	* Don't allocate from faulted devices.
	*/
	if (try_hard) {
	spa_config_enter(spa, SCL_ZIO, FTAG, RW_READER);
	allocatable = vdev_allocatable(vd);
	spa_config_exit(spa, SCL_ZIO, FTAG);
	} else {
	allocatable = vdev_allocatable(vd);
	}

	/*
	* Determine if the selected metaslab group is eligible
	* for allocations. If we're ganging then don't allow
	* this metaslab group to skip allocations since that would
	* inadvertently return ENOSPC and suspend the pool
	* even though space is still available.
	*/
	if (allocatable && !GANG_ALLOCATION(flags) && !try_hard) {
	allocatable = metaslab_group_allocatable(mg, rotor,
	flags, psize, allocator, d);
	}

	if (!allocatable) {
	metaslab_trace_add(zal, mg, NULL, psize, d,
	TRACE_NOT_ALLOCATABLE, allocator);
	goto next;
	}

	/*
	* Avoid writing single-copy data to an unhealthy,
	* non-redundant vdev, unless we've already tried all
	* other vdevs.
	*/
	if (vd->vdev_state < VDEV_STATE_HEALTHY &&
	d == 0 && !try_hard && vd->vdev_children == 0) {
	metaslab_trace_add(zal, mg, NULL, psize, d,
	TRACE_VDEV_ERROR, allocator);
	goto next;
	}

	ASSERT(mg->mg_class == mc);

	uint64_t asize = vdev_psize_to_asize_txg(vd, psize, txg);
	ASSERT(P2PHASE(asize, 1ULL << vd->vdev_ashift) == 0);

	/*
	* If we don't need to try hard, then require that the
	* block be on a different metaslab from any other DVAs
	* in this BP (unique=true). If we are trying hard, then
	* allow any metaslab to be used (unique=false).
	*/
	uint64_t offset = metaslab_group_alloc(mg, zal, asize, txg,
	!try_hard, dva, d, allocator, try_hard);

	if (offset != -1ULL) {
	/*
	* If we've just selected this metaslab group,
	* figure out whether the corresponding vdev is
	* over- or under-used relative to the pool,
	* and set an allocation bias to even it out.
	*
	* Bias is also used to compensate for unequally
	* sized vdevs so that space is allocated fairly.
	*/
	if (mca->mca_aliquot == 0 && metaslab_bias_enabled) {
	vdev_stat_t *vs = &vd->vdev_stat;
	int64_t vs_free = vs->vs_space - vs->vs_alloc;
	int64_t mc_free = mc->mc_space - mc->mc_alloc;
	int64_t ratio;

	/*
	* Calculate how much more or less we should
	* try to allocate from this device during
	* this iteration around the rotor.
	*
	* This basically introduces a zero-centered
	* bias towards the devices with the most
	* free space, while compensating for vdev
	* size differences.
	*
	* Examples:
	* vdev V1 = 16M/128M
	* vdev V2 = 16M/128M
	* ratio(V1) = 100% ratio(V2) = 100%
	*
	* vdev V1 = 16M/128M
	* vdev V2 = 64M/128M
	* ratio(V1) = 127% ratio(V2) = 72%
	*
	* vdev V1 = 16M/128M
	* vdev V2 = 64M/512M
	* ratio(V1) = 40% ratio(V2) = 160%
	*/
	ratio = (vs_free * mc->mc_alloc_groups * 100) /
	(mc_free + 1);
	mg->mg_bias = ((ratio - 100) *
	(int64_t)mg->mg_aliquot) / 100;
	} else if (!metaslab_bias_enabled) {
	mg->mg_bias = 0;
	}

	if ((flags & METASLAB_ZIL) \|\|
	atomic_add_64_nv(&mca->mca_aliquot, asize) >=
	mg->mg_aliquot + mg->mg_bias) {
	mca->mca_rotor = mg->mg_next;
	mca->mca_aliquot = 0;
	}

	DVA_SET_VDEV(&dva[d], vd->vdev_id);
	DVA_SET_OFFSET(&dva[d], offset);
	DVA_SET_GANG(&dva[d],
	((flags & METASLAB_GANG_HEADER) ? 1 : 0));
	DVA_SET_ASIZE(&dva[d], asize);

	return (0);
	}
	next:
	mca->mca_rotor = mg->mg_next;
	mca->mca_aliquot = 0;
	} while ((mg = mg->mg_next) != rotor);

	/*
	* If we haven't tried hard, perhaps do so now.
	*/
	if (!try_hard && (zfs_metaslab_try_hard_before_gang \|\|
	GANG_ALLOCATION(flags) \|\| (flags & METASLAB_ZIL) != 0 \|\|
	psize <= 1 << spa->spa_min_ashift)) {
	METASLABSTAT_BUMP(metaslabstat_try_hard);
	try_hard = B_TRUE;
	goto top;
	}

	memset(&dva[d], 0, sizeof (dva_t));

	metaslab_trace_add(zal, rotor, NULL, psize, d, TRACE_ENOSPC, allocator);
	return (SET_ERROR(ENOSPC));
	}

	void
	metaslab_free_concrete(vdev_t *vd, uint64_t offset, uint64_t asize,
	boolean_t checkpoint)
	{
	metaslab_t *msp;
	spa_t *spa = vd->vdev_spa;

	ASSERT(vdev_is_concrete(vd));
	ASSERT3U(spa_config_held(spa, SCL_ALL, RW_READER), !=, 0);
	ASSERT3U(offset >> vd->vdev_ms_shift, <, vd->vdev_ms_count);

	msp = vd->vdev_ms[offset >> vd->vdev_ms_shift];

	VERIFY(!msp->ms_condensing);
	VERIFY3U(offset, >=, msp->ms_start);
	VERIFY3U(offset + asize, <=, msp->ms_start + msp->ms_size);
	VERIFY0(P2PHASE(offset, 1ULL << vd->vdev_ashift));
	VERIFY0(P2PHASE(asize, 1ULL << vd->vdev_ashift));

	metaslab_check_free_impl(vd, offset, asize);

	mutex_enter(&msp->ms_lock);
	if (range_tree_is_empty(msp->ms_freeing) &&
	range_tree_is_empty(msp->ms_checkpointing)) {
	vdev_dirty(vd, VDD_METASLAB, msp, spa_syncing_txg(spa));
	}

	if (checkpoint) {
	ASSERT(spa_has_checkpoint(spa));
	range_tree_add(msp->ms_checkpointing, offset, asize);
	} else {
	range_tree_add(msp->ms_freeing, offset, asize);
	}
	mutex_exit(&msp->ms_lock);
	}

	void
	metaslab_free_impl_cb(uint64_t inner_offset, vdev_t *vd, uint64_t offset,
	uint64_t size, void *arg)
	{
	(void) inner_offset;
	boolean_t *checkpoint = arg;

	ASSERT3P(checkpoint, !=, NULL);

	if (vd->vdev_ops->vdev_op_remap != NULL)
	vdev_indirect_mark_obsolete(vd, offset, size);
	else
	metaslab_free_impl(vd, offset, size, *checkpoint);
	}

	static void
	metaslab_free_impl(vdev_t *vd, uint64_t offset, uint64_t size,
	boolean_t checkpoint)
	{
	spa_t *spa = vd->vdev_spa;

	ASSERT3U(spa_config_held(spa, SCL_ALL, RW_READER), !=, 0);

	if (spa_syncing_txg(spa) > spa_freeze_txg(spa))
	return;

	if (spa->spa_vdev_removal != NULL &&
	spa->spa_vdev_removal->svr_vdev_id == vd->vdev_id &&
	vdev_is_concrete(vd)) {
	/*
	* Note: we check if the vdev is concrete because when
	* we complete the removal, we first change the vdev to be
	* an indirect vdev (in open context), and then (in syncing
	* context) clear spa_vdev_removal.
	*/
	free_from_removing_vdev(vd, offset, size);
	} else if (vd->vdev_ops->vdev_op_remap != NULL) {
	vdev_indirect_mark_obsolete(vd, offset, size);
	vd->vdev_ops->vdev_op_remap(vd, offset, size,
	metaslab_free_impl_cb, &checkpoint);
	} else {
	metaslab_free_concrete(vd, offset, size, checkpoint);
	}
	}

	typedef struct remap_blkptr_cb_arg {
	blkptr_t *rbca_bp;
	spa_remap_cb_t rbca_cb;
	vdev_t *rbca_remap_vd;
	uint64_t rbca_remap_offset;
	void *rbca_cb_arg;
	} remap_blkptr_cb_arg_t;

	static void
	remap_blkptr_cb(uint64_t inner_offset, vdev_t *vd, uint64_t offset,
	uint64_t size, void *arg)
	{
	remap_blkptr_cb_arg_t *rbca = arg;
	blkptr_t *bp = rbca->rbca_bp;

	/* We can not remap split blocks. */
	if (size != DVA_GET_ASIZE(&bp->blk_dva[0]))
	return;
	ASSERT0(inner_offset);

	if (rbca->rbca_cb != NULL) {
	/*
	* At this point we know that we are not handling split
	* blocks and we invoke the callback on the previous
	* vdev which must be indirect.
	*/
	ASSERT3P(rbca->rbca_remap_vd->vdev_ops, ==, &vdev_indirect_ops);

	rbca->rbca_cb(rbca->rbca_remap_vd->vdev_id,
	rbca->rbca_remap_offset, size, rbca->rbca_cb_arg);

	/* set up remap_blkptr_cb_arg for the next call */
	rbca->rbca_remap_vd = vd;
	rbca->rbca_remap_offset = offset;
	}

	/*
	* The phys birth time is that of dva[0]. This ensures that we know
	* when each dva was written, so that resilver can determine which
	* blocks need to be scrubbed (i.e. those written during the time
	* the vdev was offline). It also ensures that the key used in
	* the ARC hash table is unique (i.e. dva[0] + phys_birth). If
	* we didn't change the phys_birth, a lookup in the ARC for a
	* remapped BP could find the data that was previously stored at
	* this vdev + offset.
	*/
	vdev_t *oldvd = vdev_lookup_top(vd->vdev_spa,
	DVA_GET_VDEV(&bp->blk_dva[0]));
	vdev_indirect_births_t *vib = oldvd->vdev_indirect_births;
	uint64_t physical_birth = vdev_indirect_births_physbirth(vib,
	DVA_GET_OFFSET(&bp->blk_dva[0]), DVA_GET_ASIZE(&bp->blk_dva[0]));
	BP_SET_PHYSICAL_BIRTH(bp, physical_birth);

	DVA_SET_VDEV(&bp->blk_dva[0], vd->vdev_id);
	DVA_SET_OFFSET(&bp->blk_dva[0], offset);
	}

	/*
	* If the block pointer contains any indirect DVAs, modify them to refer to
	* concrete DVAs. Note that this will sometimes not be possible, leaving
	* the indirect DVA in place. This happens if the indirect DVA spans multiple
	* segments in the mapping (i.e. it is a "split block").
	*
	* If the BP was remapped, calls the callback on the original dva (note the
	* callback can be called multiple times if the original indirect DVA refers
	* to another indirect DVA, etc).
	*
	* Returns TRUE if the BP was remapped.
	*/
	boolean_t
	spa_remap_blkptr(spa_t spa, blkptr_t bp, spa_remap_cb_t callback, void *arg)
	{
	remap_blkptr_cb_arg_t rbca;

	if (!zfs_remap_blkptr_enable)
	return (B_FALSE);

	if (!spa_feature_is_enabled(spa, SPA_FEATURE_OBSOLETE_COUNTS))
	return (B_FALSE);

	/*
	* Dedup BP's can not be remapped, because ddt_phys_select() depends
	* on DVA[0] being the same in the BP as in the DDT (dedup table).
	*/
	if (BP_GET_DEDUP(bp))
	return (B_FALSE);

	/*
	* Gang blocks can not be remapped, because
	* zio_checksum_gang_verifier() depends on the DVA[0] that's in
	* the BP used to read the gang block header (GBH) being the same
	* as the DVA[0] that we allocated for the GBH.
	*/
	if (BP_IS_GANG(bp))
	return (B_FALSE);

	/*
	* Embedded BP's have no DVA to remap.
	*/
	if (BP_GET_NDVAS(bp) < 1)
	return (B_FALSE);

	/*
	* Note: we only remap dva[0]. If we remapped other dvas, we
	* would no longer know what their phys birth txg is.
	*/
	dva_t *dva = &bp->blk_dva[0];

	uint64_t offset = DVA_GET_OFFSET(dva);
	uint64_t size = DVA_GET_ASIZE(dva);
	vdev_t *vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva));

	if (vd->vdev_ops->vdev_op_remap == NULL)
	return (B_FALSE);

	rbca.rbca_bp = bp;
	rbca.rbca_cb = callback;
	rbca.rbca_remap_vd = vd;
	rbca.rbca_remap_offset = offset;
	rbca.rbca_cb_arg = arg;

	/*
	* remap_blkptr_cb() will be called in order for each level of
	* indirection, until a concrete vdev is reached or a split block is
	* encountered. old_vd and old_offset are updated within the callback
	* as we go from the one indirect vdev to the next one (either concrete
	* or indirect again) in that order.
	*/
	vd->vdev_ops->vdev_op_remap(vd, offset, size, remap_blkptr_cb, &rbca);

	/* Check if the DVA wasn't remapped because it is a split block */
	if (DVA_GET_VDEV(&rbca.rbca_bp->blk_dva[0]) == vd->vdev_id)
	return (B_FALSE);

	return (B_TRUE);
	}

	/*
	* Undo the allocation of a DVA which happened in the given transaction group.
	*/
	void
	metaslab_unalloc_dva(spa_t spa, const dva_t dva, uint64_t txg)
	{
	metaslab_t *msp;
	vdev_t *vd;
	uint64_t vdev = DVA_GET_VDEV(dva);
	uint64_t offset = DVA_GET_OFFSET(dva);
	uint64_t size = DVA_GET_ASIZE(dva);

	ASSERT(DVA_IS_VALID(dva));
	ASSERT3U(spa_config_held(spa, SCL_ALL, RW_READER), !=, 0);

	if (txg > spa_freeze_txg(spa))
	return;

	if ((vd = vdev_lookup_top(spa, vdev)) == NULL \|\| !DVA_IS_VALID(dva) \|\|
	(offset >> vd->vdev_ms_shift) >= vd->vdev_ms_count) {
	zfs_panic_recover("metaslab_free_dva(): bad DVA %llu:%llu:%llu",
	(u_longlong_t)vdev, (u_longlong_t)offset,
	(u_longlong_t)size);
	return;
	}

	ASSERT(!vd->vdev_removing);
	ASSERT(vdev_is_concrete(vd));
	ASSERT0(vd->vdev_indirect_config.vic_mapping_object);
	ASSERT3P(vd->vdev_indirect_mapping, ==, NULL);

	if (DVA_GET_GANG(dva))
	size = vdev_gang_header_asize(vd);

	msp = vd->vdev_ms[offset >> vd->vdev_ms_shift];

	mutex_enter(&msp->ms_lock);
	range_tree_remove(msp->ms_allocating[txg & TXG_MASK],
	offset, size);
	msp->ms_allocating_total -= size;

	VERIFY(!msp->ms_condensing);
	VERIFY3U(offset, >=, msp->ms_start);
	VERIFY3U(offset + size, <=, msp->ms_start + msp->ms_size);
	VERIFY3U(range_tree_space(msp->ms_allocatable) + size, <=,
	msp->ms_size);
	VERIFY0(P2PHASE(offset, 1ULL << vd->vdev_ashift));
	VERIFY0(P2PHASE(size, 1ULL << vd->vdev_ashift));
	range_tree_add(msp->ms_allocatable, offset, size);
	mutex_exit(&msp->ms_lock);
	}

	/*
	* Free the block represented by the given DVA.
	*/
	void
	metaslab_free_dva(spa_t spa, const dva_t dva, boolean_t checkpoint)
	{
	uint64_t vdev = DVA_GET_VDEV(dva);
	uint64_t offset = DVA_GET_OFFSET(dva);
	uint64_t size = DVA_GET_ASIZE(dva);
	vdev_t *vd = vdev_lookup_top(spa, vdev);

	ASSERT(DVA_IS_VALID(dva));
	ASSERT3U(spa_config_held(spa, SCL_ALL, RW_READER), !=, 0);

	if (DVA_GET_GANG(dva)) {
	size = vdev_gang_header_asize(vd);
	}

	metaslab_free_impl(vd, offset, size, checkpoint);
	}

	/*
	* Reserve some allocation slots. The reservation system must be called
	* before we call into the allocator. If there aren't any available slots
	* then the I/O will be throttled until an I/O completes and its slots are
	* freed up. The function returns true if it was successful in placing
	* the reservation.
	*/
	boolean_t
	metaslab_class_throttle_reserve(metaslab_class_t *mc, int slots, int allocator,
	zio_t *zio, int flags)
	{
	metaslab_class_allocator_t *mca = &mc->mc_allocator[allocator];
	uint64_t max = mca->mca_alloc_max_slots;

	ASSERT(mc->mc_alloc_throttle_enabled);
	if (GANG_ALLOCATION(flags) \|\| (flags & METASLAB_MUST_RESERVE) \|\|
	zfs_refcount_count(&mca->mca_alloc_slots) + slots <= max) {
	/*
	* The potential race between _count() and _add() is covered
	* by the allocator lock in most cases, or irrelevant due to
	* GANG_ALLOCATION() or METASLAB_MUST_RESERVE set in others.
	* But even if we assume some other non-existing scenario, the
	* worst that can happen is few more I/Os get to allocation
	* earlier, that is not a problem.
	*
	* We reserve the slots individually so that we can unreserve
	* them individually when an I/O completes.
	*/
	zfs_refcount_add_few(&mca->mca_alloc_slots, slots, zio);
	zio->io_flags \|= ZIO_FLAG_IO_ALLOCATING;
	return (B_TRUE);
	}
	return (B_FALSE);
	}

	void
	metaslab_class_throttle_unreserve(metaslab_class_t *mc, int slots,
	int allocator, zio_t *zio)
	{
	metaslab_class_allocator_t *mca = &mc->mc_allocator[allocator];

	ASSERT(mc->mc_alloc_throttle_enabled);
	zfs_refcount_remove_few(&mca->mca_alloc_slots, slots, zio);
	}

	static int
	metaslab_claim_concrete(vdev_t *vd, uint64_t offset, uint64_t size,
	uint64_t txg)
	{
	metaslab_t *msp;
	spa_t *spa = vd->vdev_spa;
	int error = 0;

	if (offset >> vd->vdev_ms_shift >= vd->vdev_ms_count)
	return (SET_ERROR(ENXIO));

	ASSERT3P(vd->vdev_ms, !=, NULL);
	msp = vd->vdev_ms[offset >> vd->vdev_ms_shift];

	mutex_enter(&msp->ms_lock);

	if ((txg != 0 && spa_writeable(spa)) \|\| !msp->ms_loaded) {
	error = metaslab_activate(msp, 0, METASLAB_WEIGHT_CLAIM);
	if (error == EBUSY) {
	ASSERT(msp->ms_loaded);
	ASSERT(msp->ms_weight & METASLAB_ACTIVE_MASK);
	error = 0;
	}
	}

	if (error == 0 &&
	!range_tree_contains(msp->ms_allocatable, offset, size))
	error = SET_ERROR(ENOENT);

	if (error \|\| txg == 0) { /* txg == 0 indicates dry run */
	mutex_exit(&msp->ms_lock);
	return (error);
	}

	VERIFY(!msp->ms_condensing);
	VERIFY0(P2PHASE(offset, 1ULL << vd->vdev_ashift));
	VERIFY0(P2PHASE(size, 1ULL << vd->vdev_ashift));
	VERIFY3U(range_tree_space(msp->ms_allocatable) - size, <=,
	msp->ms_size);
	range_tree_remove(msp->ms_allocatable, offset, size);
	range_tree_clear(msp->ms_trim, offset, size);

	if (spa_writeable(spa)) { /* don't dirty if we're zdb(8) */
	metaslab_class_t *mc = msp->ms_group->mg_class;
	multilist_sublist_t *mls =
	multilist_sublist_lock_obj(&mc->mc_metaslab_txg_list, msp);
	if (!multilist_link_active(&msp->ms_class_txg_node)) {
	msp->ms_selected_txg = txg;
	multilist_sublist_insert_head(mls, msp);
	}
	multilist_sublist_unlock(mls);

	if (range_tree_is_empty(msp->ms_allocating[txg & TXG_MASK]))
	vdev_dirty(vd, VDD_METASLAB, msp, txg);
	range_tree_add(msp->ms_allocating[txg & TXG_MASK],
	offset, size);
	msp->ms_allocating_total += size;
	}

	mutex_exit(&msp->ms_lock);

	return (0);
	}

	typedef struct metaslab_claim_cb_arg_t {
	uint64_t mcca_txg;
	int mcca_error;
	} metaslab_claim_cb_arg_t;

	static void
	metaslab_claim_impl_cb(uint64_t inner_offset, vdev_t *vd, uint64_t offset,
	uint64_t size, void *arg)
	{
	(void) inner_offset;
	metaslab_claim_cb_arg_t *mcca_arg = arg;

	if (mcca_arg->mcca_error == 0) {
	mcca_arg->mcca_error = metaslab_claim_concrete(vd, offset,
	size, mcca_arg->mcca_txg);
	}
	}

	int
	metaslab_claim_impl(vdev_t *vd, uint64_t offset, uint64_t size, uint64_t txg)
	{
	if (vd->vdev_ops->vdev_op_remap != NULL) {
	metaslab_claim_cb_arg_t arg;

	/*
	* Only zdb(8) can claim on indirect vdevs. This is used
	* to detect leaks of mapped space (that are not accounted
	* for in the obsolete counts, spacemap, or bpobj).
	*/
	ASSERT(!spa_writeable(vd->vdev_spa));
	arg.mcca_error = 0;
	arg.mcca_txg = txg;

	vd->vdev_ops->vdev_op_remap(vd, offset, size,
	metaslab_claim_impl_cb, &arg);

	if (arg.mcca_error == 0) {
	arg.mcca_error = metaslab_claim_concrete(vd,
	offset, size, txg);
	}
	return (arg.mcca_error);
	} else {
	return (metaslab_claim_concrete(vd, offset, size, txg));
	}
	}

	/*
	* Intent log support: upon opening the pool after a crash, notify the SPA
	* of blocks that the intent log has allocated for immediate write, but
	* which are still considered free by the SPA because the last transaction
	* group didn't commit yet.
	*/
	static int
	metaslab_claim_dva(spa_t spa, const dva_t dva, uint64_t txg)
	{
	uint64_t vdev = DVA_GET_VDEV(dva);
	uint64_t offset = DVA_GET_OFFSET(dva);
	uint64_t size = DVA_GET_ASIZE(dva);
	vdev_t *vd;

	if ((vd = vdev_lookup_top(spa, vdev)) == NULL) {
	return (SET_ERROR(ENXIO));
	}

	ASSERT(DVA_IS_VALID(dva));

	if (DVA_GET_GANG(dva))
	size = vdev_gang_header_asize(vd);

	return (metaslab_claim_impl(vd, offset, size, txg));
	}

	int
	metaslab_alloc(spa_t spa, metaslab_class_t mc, uint64_t psize, blkptr_t *bp,
	int ndvas, uint64_t txg, blkptr_t *hintbp, int flags,
	zio_alloc_list_t zal, zio_t zio, int allocator)
	{
	dva_t *dva = bp->blk_dva;
	dva_t *hintdva = (hintbp != NULL) ? hintbp->blk_dva : NULL;
	int error = 0;

	ASSERT0(BP_GET_LOGICAL_BIRTH(bp));
	ASSERT0(BP_GET_PHYSICAL_BIRTH(bp));

	spa_config_enter(spa, SCL_ALLOC, FTAG, RW_READER);

	if (mc->mc_allocator[allocator].mca_rotor == NULL) {
	/* no vdevs in this class */
	spa_config_exit(spa, SCL_ALLOC, FTAG);
	return (SET_ERROR(ENOSPC));
	}

	ASSERT(ndvas > 0 && ndvas <= spa_max_replication(spa));
	ASSERT(BP_GET_NDVAS(bp) == 0);
	ASSERT(hintbp == NULL \|\| ndvas <= BP_GET_NDVAS(hintbp));
	ASSERT3P(zal, !=, NULL);

	for (int d = 0; d < ndvas; d++) {
	error = metaslab_alloc_dva(spa, mc, psize, dva, d, hintdva,
	txg, flags, zal, allocator);
	if (error != 0) {
	for (d--; d >= 0; d--) {
	metaslab_unalloc_dva(spa, &dva[d], txg);
	metaslab_group_alloc_decrement(spa,
	DVA_GET_VDEV(&dva[d]), zio, flags,
	allocator, B_FALSE);
	memset(&dva[d], 0, sizeof (dva_t));
	}
	spa_config_exit(spa, SCL_ALLOC, FTAG);
	return (error);
	} else {
	/*
	* Update the metaslab group's queue depth
	* based on the newly allocated dva.
	*/
	metaslab_group_alloc_increment(spa,
	DVA_GET_VDEV(&dva[d]), zio, flags, allocator);
	}
	}
	ASSERT(error == 0);
	ASSERT(BP_GET_NDVAS(bp) == ndvas);

	spa_config_exit(spa, SCL_ALLOC, FTAG);

	BP_SET_BIRTH(bp, txg, 0);

	return (0);
	}

	void
	metaslab_free(spa_t spa, const blkptr_t bp, uint64_t txg, boolean_t now)
	{
	const dva_t *dva = bp->blk_dva;
	int ndvas = BP_GET_NDVAS(bp);

	ASSERT(!BP_IS_HOLE(bp));
	ASSERT(!now \|\| BP_GET_LOGICAL_BIRTH(bp) >= spa_syncing_txg(spa));

	/*
	* If we have a checkpoint for the pool we need to make sure that
	* the blocks that we free that are part of the checkpoint won't be
	* reused until the checkpoint is discarded or we revert to it.
	*
	* The checkpoint flag is passed down the metaslab_free code path
	* and is set whenever we want to add a block to the checkpoint's
	* accounting. That is, we "checkpoint" blocks that existed at the
	* time the checkpoint was created and are therefore referenced by
	* the checkpointed uberblock.
	*
	* Note that, we don't checkpoint any blocks if the current
	* syncing txg <= spa_checkpoint_txg. We want these frees to sync
	* normally as they will be referenced by the checkpointed uberblock.
	*/
	boolean_t checkpoint = B_FALSE;
	if (BP_GET_LOGICAL_BIRTH(bp) <= spa->spa_checkpoint_txg &&
	spa_syncing_txg(spa) > spa->spa_checkpoint_txg) {
	/*
	* At this point, if the block is part of the checkpoint
	* there is no way it was created in the current txg.
	*/
	ASSERT(!now);
	ASSERT3U(spa_syncing_txg(spa), ==, txg);
	checkpoint = B_TRUE;
	}

	spa_config_enter(spa, SCL_FREE, FTAG, RW_READER);

	for (int d = 0; d < ndvas; d++) {
	if (now) {
	metaslab_unalloc_dva(spa, &dva[d], txg);
	} else {
	ASSERT3U(txg, ==, spa_syncing_txg(spa));
	metaslab_free_dva(spa, &dva[d], checkpoint);
	}
	}

	spa_config_exit(spa, SCL_FREE, FTAG);
	}

	int
	metaslab_claim(spa_t spa, const blkptr_t bp, uint64_t txg)
	{
	const dva_t *dva = bp->blk_dva;
	int ndvas = BP_GET_NDVAS(bp);
	int error = 0;

	ASSERT(!BP_IS_HOLE(bp));

	if (txg != 0) {
	/*
	* First do a dry run to make sure all DVAs are claimable,
	* so we don't have to unwind from partial failures below.
	*/
	if ((error = metaslab_claim(spa, bp, 0)) != 0)
	return (error);
	}

	spa_config_enter(spa, SCL_ALLOC, FTAG, RW_READER);

	for (int d = 0; d < ndvas; d++) {
	error = metaslab_claim_dva(spa, &dva[d], txg);
	if (error != 0)
	break;
	}

	spa_config_exit(spa, SCL_ALLOC, FTAG);

	ASSERT(error == 0 \|\| txg == 0);

	return (error);
	}

	static void
	metaslab_check_free_impl_cb(uint64_t inner, vdev_t *vd, uint64_t offset,
	uint64_t size, void *arg)
	{
	(void) inner, (void) arg;

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

	metaslab_check_free_impl(vd, offset, size);
	}

	static void
	metaslab_check_free_impl(vdev_t *vd, uint64_t offset, uint64_t size)
	{
	metaslab_t *msp;
	spa_t *spa __maybe_unused = vd->vdev_spa;

	if ((zfs_flags & ZFS_DEBUG_ZIO_FREE) == 0)
	return;

	if (vd->vdev_ops->vdev_op_remap != NULL) {
	vd->vdev_ops->vdev_op_remap(vd, offset, size,
	metaslab_check_free_impl_cb, NULL);
	return;
	}

	ASSERT(vdev_is_concrete(vd));
	ASSERT3U(offset >> vd->vdev_ms_shift, <, vd->vdev_ms_count);
	ASSERT3U(spa_config_held(spa, SCL_ALL, RW_READER), !=, 0);

	msp = vd->vdev_ms[offset >> vd->vdev_ms_shift];

	mutex_enter(&msp->ms_lock);
	if (msp->ms_loaded) {
	range_tree_verify_not_present(msp->ms_allocatable,
	offset, size);
	}

	/*
	* Check all segments that currently exist in the freeing pipeline.
	*
	* It would intuitively make sense to also check the current allocating
	* tree since metaslab_unalloc_dva() exists for extents that are
	* allocated and freed in the same sync pass within the same txg.
	* Unfortunately there are places (e.g. the ZIL) where we allocate a
	* segment but then we free part of it within the same txg
	* [see zil_sync()]. Thus, we don't call range_tree_verify() in the
	* current allocating tree.
	*/
	range_tree_verify_not_present(msp->ms_freeing, offset, size);
	range_tree_verify_not_present(msp->ms_checkpointing, offset, size);
	range_tree_verify_not_present(msp->ms_freed, offset, size);
	for (int j = 0; j < TXG_DEFER_SIZE; j++)
	range_tree_verify_not_present(msp->ms_defer[j], offset, size);
	range_tree_verify_not_present(msp->ms_trim, offset, size);
	mutex_exit(&msp->ms_lock);
	}

	void
	metaslab_check_free(spa_t spa, const blkptr_t bp)
	{
	if ((zfs_flags & ZFS_DEBUG_ZIO_FREE) == 0)
	return;

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
	for (int i = 0; i < BP_GET_NDVAS(bp); i++) {
	uint64_t vdev = DVA_GET_VDEV(&bp->blk_dva[i]);
	vdev_t *vd = vdev_lookup_top(spa, vdev);
	uint64_t offset = DVA_GET_OFFSET(&bp->blk_dva[i]);
	uint64_t size = DVA_GET_ASIZE(&bp->blk_dva[i]);

	if (DVA_GET_GANG(&bp->blk_dva[i]))
	size = vdev_gang_header_asize(vd);

	ASSERT3P(vd, !=, NULL);

	metaslab_check_free_impl(vd, offset, size);
	}
	spa_config_exit(spa, SCL_VDEV, FTAG);
	}

	static void
	metaslab_group_disable_wait(metaslab_group_t *mg)
	{
	ASSERT(MUTEX_HELD(&mg->mg_ms_disabled_lock));
	while (mg->mg_disabled_updating) {
	cv_wait(&mg->mg_ms_disabled_cv, &mg->mg_ms_disabled_lock);
	}
	}

	static void
	metaslab_group_disabled_increment(metaslab_group_t *mg)
	{
	ASSERT(MUTEX_HELD(&mg->mg_ms_disabled_lock));
	ASSERT(mg->mg_disabled_updating);

	while (mg->mg_ms_disabled >= max_disabled_ms) {
	cv_wait(&mg->mg_ms_disabled_cv, &mg->mg_ms_disabled_lock);
	}
	mg->mg_ms_disabled++;
	ASSERT3U(mg->mg_ms_disabled, <=, max_disabled_ms);
	}

	/*
	* Mark the metaslab as disabled to prevent any allocations on this metaslab.
	* We must also track how many metaslabs are currently disabled within a
	* metaslab group and limit them to prevent allocation failures from
	* occurring because all metaslabs are disabled.
	*/
	void
	metaslab_disable(metaslab_t *msp)
	{
	ASSERT(!MUTEX_HELD(&msp->ms_lock));
	metaslab_group_t *mg = msp->ms_group;

	mutex_enter(&mg->mg_ms_disabled_lock);

	/*
	* To keep an accurate count of how many threads have disabled
	* a specific metaslab group, we only allow one thread to mark
	* the metaslab group at a time. This ensures that the value of
	* ms_disabled will be accurate when we decide to mark a metaslab
	* group as disabled. To do this we force all other threads
	* to wait till the metaslab's mg_disabled_updating flag is no
	* longer set.
	*/
	metaslab_group_disable_wait(mg);
	mg->mg_disabled_updating = B_TRUE;
	if (msp->ms_disabled == 0) {
	metaslab_group_disabled_increment(mg);
	}
	mutex_enter(&msp->ms_lock);
	msp->ms_disabled++;
	mutex_exit(&msp->ms_lock);

	mg->mg_disabled_updating = B_FALSE;
	cv_broadcast(&mg->mg_ms_disabled_cv);
	mutex_exit(&mg->mg_ms_disabled_lock);
	}

	void
	metaslab_enable(metaslab_t *msp, boolean_t sync, boolean_t unload)
	{
	metaslab_group_t *mg = msp->ms_group;
	spa_t *spa = mg->mg_vd->vdev_spa;

	/*
	* Wait for the outstanding IO to be synced to prevent newly
	* allocated blocks from being overwritten. This used by
	* initialize and TRIM which are modifying unallocated space.
	*/
	if (sync)
	txg_wait_synced(spa_get_dsl(spa), 0);

	mutex_enter(&mg->mg_ms_disabled_lock);
	mutex_enter(&msp->ms_lock);
	if (--msp->ms_disabled == 0) {
	mg->mg_ms_disabled--;
	cv_broadcast(&mg->mg_ms_disabled_cv);
	if (unload)
	metaslab_unload(msp);
	}
	mutex_exit(&msp->ms_lock);
	mutex_exit(&mg->mg_ms_disabled_lock);
	}

	void
	metaslab_set_unflushed_dirty(metaslab_t *ms, boolean_t dirty)
	{
	ms->ms_unflushed_dirty = dirty;
	}

	static void
	metaslab_update_ondisk_flush_data(metaslab_t ms, dmu_tx_t tx)
	{
	vdev_t *vd = ms->ms_group->mg_vd;
	spa_t *spa = vd->vdev_spa;
	objset_t *mos = spa_meta_objset(spa);

	ASSERT(spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP));

	metaslab_unflushed_phys_t entry = {
	.msp_unflushed_txg = metaslab_unflushed_txg(ms),
	};
	uint64_t entry_size = sizeof (entry);
	uint64_t entry_offset = ms->ms_id * entry_size;

	uint64_t object = 0;
	int err = zap_lookup(mos, vd->vdev_top_zap,
	VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS, sizeof (uint64_t), 1,
	&object);
	if (err == ENOENT) {
	object = dmu_object_alloc(mos, DMU_OTN_UINT64_METADATA,
	SPA_OLD_MAXBLOCKSIZE, DMU_OT_NONE, 0, tx);
	VERIFY0(zap_add(mos, vd->vdev_top_zap,
	VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS, sizeof (uint64_t), 1,
	&object, tx));
	} else {
	VERIFY0(err);
	}

	dmu_write(spa_meta_objset(spa), object, entry_offset, entry_size,
	&entry, tx);
	}

	void
	metaslab_set_unflushed_txg(metaslab_t ms, uint64_t txg, dmu_tx_t tx)
	{
	ms->ms_unflushed_txg = txg;
	metaslab_update_ondisk_flush_data(ms, tx);
	}

	boolean_t
	metaslab_unflushed_dirty(metaslab_t *ms)
	{
	return (ms->ms_unflushed_dirty);
	}

	uint64_t
	metaslab_unflushed_txg(metaslab_t *ms)
	{
	return (ms->ms_unflushed_txg);
	}

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, aliquot, U64, ZMOD_RW,
	"Allocation granularity (a.k.a. stripe size)");

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, debug_load, INT, ZMOD_RW,
	"Load all metaslabs when pool is first opened");

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, debug_unload, INT, ZMOD_RW,
	"Prevent metaslabs from being unloaded");

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, preload_enabled, INT, ZMOD_RW,
	"Preload potential metaslabs during reassessment");

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, preload_limit, UINT, ZMOD_RW,
	"Max number of metaslabs per group to preload");

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, unload_delay, UINT, ZMOD_RW,
	"Delay in txgs after metaslab was last used before unloading");

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, unload_delay_ms, UINT, ZMOD_RW,
	"Delay in milliseconds after metaslab was last used before unloading");

	/* BEGIN CSTYLED */
	ZFS_MODULE_PARAM(zfs_mg, zfs_mg_, noalloc_threshold, UINT, ZMOD_RW,
	"Percentage of metaslab group size that should be free to make it "
	"eligible for allocation");

	ZFS_MODULE_PARAM(zfs_mg, zfs_mg_, fragmentation_threshold, UINT, ZMOD_RW,
	"Percentage of metaslab group size that should be considered eligible "
	"for allocations unless all metaslab groups within the metaslab class "
	"have also crossed this threshold");

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, fragmentation_factor_enabled, INT,
	ZMOD_RW,
	"Use the fragmentation metric to prefer less fragmented metaslabs");
	/* END CSTYLED */

	ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, fragmentation_threshold, UINT,
	ZMOD_RW, "Fragmentation for metaslab to allow allocation");

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, lba_weighting_enabled, INT, ZMOD_RW,
	"Prefer metaslabs with lower LBAs");

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, bias_enabled, INT, ZMOD_RW,
	"Enable metaslab group biasing");

	ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, segment_weight_enabled, INT,
	ZMOD_RW, "Enable segment-based metaslab selection");

	ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, switch_threshold, INT, ZMOD_RW,
	"Segment-based metaslab selection maximum buckets before switching");

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, force_ganging, U64, ZMOD_RW,
	"Blocks larger than this size are sometimes forced to be gang blocks");

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, force_ganging_pct, UINT, ZMOD_RW,
	"Percentage of large blocks that will be forced to be gang blocks");

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, df_max_search, UINT, ZMOD_RW,
	"Max distance (bytes) to search forward before using size tree");

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, df_use_largest_segment, INT, ZMOD_RW,
	"When looking in size tree, use largest segment instead of exact fit");

	ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, max_size_cache_sec, U64,
	ZMOD_RW, "How long to trust the cached max chunk size of a metaslab");

	ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, mem_limit, UINT, ZMOD_RW,
	"Percentage of memory that can be used to store metaslab range trees");

	ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, try_hard_before_gang, INT,
	ZMOD_RW, "Try hard to allocate before ganging");

	ZFS_MODULE_PARAM(zfs_metaslab, zfs_metaslab_, find_max_tries, UINT, ZMOD_RW,
	"Normally only consider this many of the best metaslabs in each vdev");

	/* BEGIN CSTYLED */
	ZFS_MODULE_PARAM_CALL(zfs, zfs_, active_allocator,
	param_set_active_allocator, param_get_charp, ZMOD_RW,
	"SPA active allocator");
	/* END CSTYLED */
	diff --git a/sys/contrib/openzfs/module/zfs/spa.c b/sys/contrib/openzfs/module/zfs/spa.c
	index c3800e018c73..638572996c3a 100644
	--- a/sys/contrib/openzfs/module/zfs/spa.c
	+++ b/sys/contrib/openzfs/module/zfs/spa.c
	@@ -1,10946 +1,10972 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2024 by Delphix. All rights reserved.
	* Copyright (c) 2018, Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
	* Copyright 2013 Saso Kiselkov. All rights reserved.
	* Copyright (c) 2014 Integros [integros.com]
	* Copyright 2016 Toomas Soome <tsoome@me.com>
	* Copyright (c) 2016 Actifio, Inc. All rights reserved.
	* Copyright 2018 Joyent, Inc.
	* Copyright (c) 2017, 2019, Datto Inc. All rights reserved.
	* Copyright 2017 Joyent, Inc.
	* Copyright (c) 2017, Intel Corporation.
	* Copyright (c) 2021, Colm Buckley <colm@tuatha.org>
	* Copyright (c) 2023 Hewlett Packard Enterprise Development LP.
	+ * Copyright (c) 2024, Klara Inc.
	*/

	/*
	* SPA: Storage Pool Allocator
	*
	* This file contains all the routines used when modifying on-disk SPA state.
	* This includes opening, importing, destroying, exporting a pool, and syncing a
	* pool.
	*/

	#include <sys/zfs_context.h>
	#include <sys/fm/fs/zfs.h>
	#include <sys/spa_impl.h>
	#include <sys/zio.h>
	#include <sys/zio_checksum.h>
	#include <sys/dmu.h>
	#include <sys/dmu_tx.h>
	#include <sys/zap.h>
	#include <sys/zil.h>
	#include <sys/brt.h>
	#include <sys/ddt.h>
	#include <sys/vdev_impl.h>
	#include <sys/vdev_removal.h>
	#include <sys/vdev_indirect_mapping.h>
	#include <sys/vdev_indirect_births.h>
	#include <sys/vdev_initialize.h>
	#include <sys/vdev_rebuild.h>
	#include <sys/vdev_trim.h>
	#include <sys/vdev_disk.h>
	#include <sys/vdev_raidz.h>
	#include <sys/vdev_draid.h>
	#include <sys/metaslab.h>
	#include <sys/metaslab_impl.h>
	#include <sys/mmp.h>
	#include <sys/uberblock_impl.h>
	#include <sys/txg.h>
	#include <sys/avl.h>
	#include <sys/bpobj.h>
	#include <sys/dmu_traverse.h>
	#include <sys/dmu_objset.h>
	#include <sys/unique.h>
	#include <sys/dsl_pool.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_dir.h>
	#include <sys/dsl_prop.h>
	#include <sys/dsl_synctask.h>
	#include <sys/fs/zfs.h>
	#include <sys/arc.h>
	#include <sys/callb.h>
	#include <sys/systeminfo.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/dsl_scan.h>
	#include <sys/zfeature.h>
	#include <sys/dsl_destroy.h>
	#include <sys/zvol.h>

	#ifdef _KERNEL
	#include <sys/fm/protocol.h>
	#include <sys/fm/util.h>
	#include <sys/callb.h>
	#include <sys/zone.h>
	#include <sys/vmsystm.h>
	#endif /* _KERNEL */

	#include "zfs_prop.h"
	#include "zfs_comutil.h"
	#include <cityhash.h>

	/*
	* spa_thread() existed on Illumos as a parent thread for the various worker
	* threads that actually run the pool, as a way to both reference the entire
	* pool work as a single object, and to share properties like scheduling
	* options. It has not yet been adapted to Linux or FreeBSD. This define is
	* used to mark related parts of the code to make things easier for the reader,
	* and to compile this code out. It can be removed when someone implements it,
	* moves it to some Illumos-specific place, or removes it entirely.
	*/
	#undef HAVE_SPA_THREAD

	/*
	* The "System Duty Cycle" scheduling class is an Illumos feature to help
	* prevent CPU-intensive kernel threads from affecting latency on interactive
	* threads. It doesn't exist on Linux or FreeBSD, so the supporting code is
	* gated behind a define. On Illumos SDC depends on spa_thread(), but
	* spa_thread() also has other uses, so this is a separate define.
	*/
	#undef HAVE_SYSDC

	/*
	* The interval, in seconds, at which failed configuration cache file writes
	* should be retried.
	*/
	int zfs_ccw_retry_interval = 300;

	typedef enum zti_modes {
	ZTI_MODE_FIXED, /* value is # of threads (min 1) */
	ZTI_MODE_SCALE, /* Taskqs scale with CPUs. */
	ZTI_MODE_SYNC, /* sync thread assigned */
	ZTI_MODE_NULL, /* don't create a taskq */
	ZTI_NMODES
	} zti_modes_t;

	#define ZTI_P(n, q) { ZTI_MODE_FIXED, (n), (q) }
	#define ZTI_PCT(n) { ZTI_MODE_ONLINE_PERCENT, (n), 1 }
	#define ZTI_SCALE { ZTI_MODE_SCALE, 0, 1 }
	#define ZTI_SYNC { ZTI_MODE_SYNC, 0, 1 }
	#define ZTI_NULL { ZTI_MODE_NULL, 0, 0 }

	#define ZTI_N(n) ZTI_P(n, 1)
	#define ZTI_ONE ZTI_N(1)

	typedef struct zio_taskq_info {
	zti_modes_t zti_mode;
	uint_t zti_value;
	uint_t zti_count;
	} zio_taskq_info_t;

	static const char *const zio_taskq_types[ZIO_TASKQ_TYPES] = {
	"iss", "iss_h", "int", "int_h"
	};

	/*
	* This table defines the taskq settings for each ZFS I/O type. When
	* initializing a pool, we use this table to create an appropriately sized
	* taskq. Some operations are low volume and therefore have a small, static
	* number of threads assigned to their taskqs using the ZTI_N(#) or ZTI_ONE
	* macros. Other operations process a large amount of data; the ZTI_SCALE
	* macro causes us to create a taskq oriented for throughput. Some operations
	* are so high frequency and short-lived that the taskq itself can become a
	* point of lock contention. The ZTI_P(#, #) macro indicates that we need an
	* additional degree of parallelism specified by the number of threads per-
	* taskq and the number of taskqs; when dispatching an event in this case, the
	* particular taskq is chosen at random. ZTI_SCALE uses a number of taskqs
	* that scales with the number of CPUs.
	*
	* The different taskq priorities are to handle the different contexts (issue
	- * and interrupt) and then to reserve threads for ZIO_PRIORITY_NOW I/Os that
	- * need to be handled with minimum delay.
	+ * and interrupt) and then to reserve threads for high priority I/Os that
	+ * need to be handled with minimum delay. Illumos taskq has unfair TQ_FRONT
	+ * implementation, so separate high priority threads are used there.
	*/
	static zio_taskq_info_t zio_taskqs[ZIO_TYPES][ZIO_TASKQ_TYPES] = {
	/* ISSUE ISSUE_HIGH INTR INTR_HIGH */
	{ ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* NULL */
	{ ZTI_N(8), ZTI_NULL, ZTI_SCALE, ZTI_NULL }, /* READ */
	+#ifdef illumos
	{ ZTI_SYNC, ZTI_N(5), ZTI_SCALE, ZTI_N(5) }, /* WRITE */
	+#else
	+ { ZTI_SYNC, ZTI_NULL, ZTI_SCALE, ZTI_NULL }, /* WRITE */
	+#endif
	{ ZTI_SCALE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* FREE */
	{ ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* CLAIM */
	{ ZTI_ONE, ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* FLUSH */
	{ ZTI_N(4), ZTI_NULL, ZTI_ONE, ZTI_NULL }, /* TRIM */
	};

	static void spa_sync_version(void arg, dmu_tx_t tx);
	static void spa_sync_props(void arg, dmu_tx_t tx);
	static boolean_t spa_has_active_shared_spare(spa_t *spa);
	static int spa_load_impl(spa_t *spa, spa_import_type_t type,
	const char **ereport);
	static void spa_vdev_resilver_done(spa_t *spa);

	/*
	* Percentage of all CPUs that can be used by the metaslab preload taskq.
	*/
	static uint_t metaslab_preload_pct = 50;

	static uint_t zio_taskq_batch_pct = 80; /* 1 thread per cpu in pset */
	static uint_t zio_taskq_batch_tpq; /* threads per taskq */

	#ifdef HAVE_SYSDC
	static const boolean_t zio_taskq_sysdc = B_TRUE; /* use SDC scheduling class */
	static const uint_t zio_taskq_basedc = 80; /* base duty cycle */
	#endif

	#ifdef HAVE_SPA_THREAD
	static const boolean_t spa_create_process = B_TRUE; /* no process => no sysdc */
	#endif

	static uint_t zio_taskq_write_tpq = 16;

	/*
	* Report any spa_load_verify errors found, but do not fail spa_load.
	* This is used by zdb to analyze non-idle pools.
	*/
	boolean_t spa_load_verify_dryrun = B_FALSE;

	/*
	* Allow read spacemaps in case of readonly import (spa_mode == SPA_MODE_READ).
	* This is used by zdb for spacemaps verification.
	*/
	boolean_t spa_mode_readable_spacemaps = B_FALSE;

	/*
	* This (illegal) pool name is used when temporarily importing a spa_t in order
	* to get the vdev stats associated with the imported devices.
	*/
	#define TRYIMPORT_NAME "$import"

	/*
	* For debugging purposes: print out vdev tree during pool import.
	*/
	static int spa_load_print_vdev_tree = B_FALSE;

	/*
	* A non-zero value for zfs_max_missing_tvds means that we allow importing
	* pools with missing top-level vdevs. This is strictly intended for advanced
	* pool recovery cases since missing data is almost inevitable. Pools with
	* missing devices can only be imported read-only for safety reasons, and their
	* fail-mode will be automatically set to "continue".
	*
	* With 1 missing vdev we should be able to import the pool and mount all
	* datasets. User data that was not modified after the missing device has been
	* added should be recoverable. This means that snapshots created prior to the
	* addition of that device should be completely intact.
	*
	* With 2 missing vdevs, some datasets may fail to mount since there are
	* dataset statistics that are stored as regular metadata. Some data might be
	* recoverable if those vdevs were added recently.
	*
	* With 3 or more missing vdevs, the pool is severely damaged and MOS entries
	* may be missing entirely. Chances of data recovery are very low. Note that
	* there are also risks of performing an inadvertent rewind as we might be
	* missing all the vdevs with the latest uberblocks.
	*/
	uint64_t zfs_max_missing_tvds = 0;

	/*
	* The parameters below are similar to zfs_max_missing_tvds but are only
	* intended for a preliminary open of the pool with an untrusted config which
	* might be incomplete or out-dated.
	*
	* We are more tolerant for pools opened from a cachefile since we could have
	* an out-dated cachefile where a device removal was not registered.
	* We could have set the limit arbitrarily high but in the case where devices
	* are really missing we would want to return the proper error codes; we chose
	* SPA_DVAS_PER_BP - 1 so that some copies of the MOS would still be available
	* and we get a chance to retrieve the trusted config.
	*/
	uint64_t zfs_max_missing_tvds_cachefile = SPA_DVAS_PER_BP - 1;

	/*
	* In the case where config was assembled by scanning device paths (/dev/dsks
	* by default) we are less tolerant since all the existing devices should have
	* been detected and we want spa_load to return the right error codes.
	*/
	uint64_t zfs_max_missing_tvds_scan = 0;

	/*
	* Debugging aid that pauses spa_sync() towards the end.
	*/
	static const boolean_t zfs_pause_spa_sync = B_FALSE;

	/*
	* Variables to indicate the livelist condense zthr func should wait at certain
	* points for the livelist to be removed - used to test condense/destroy races
	*/
	static int zfs_livelist_condense_zthr_pause = 0;
	static int zfs_livelist_condense_sync_pause = 0;

	/*
	* Variables to track whether or not condense cancellation has been
	* triggered in testing.
	*/
	static int zfs_livelist_condense_sync_cancel = 0;
	static int zfs_livelist_condense_zthr_cancel = 0;

	/*
	* Variable to track whether or not extra ALLOC blkptrs were added to a
	* livelist entry while it was being condensed (caused by the way we track
	* remapped blkptrs in dbuf_remap_impl)
	*/
	static int zfs_livelist_condense_new_alloc = 0;

	/*
	* ==========================================================================
	* SPA properties routines
	* ==========================================================================
	*/

	/*
	* Add a (source=src, propname=propval) list to an nvlist.
	*/
	static void
	spa_prop_add_list(nvlist_t nvl, zpool_prop_t prop, const char strval,
	uint64_t intval, zprop_source_t src)
	{
	const char *propname = zpool_prop_to_name(prop);
	nvlist_t *propval;

	propval = fnvlist_alloc();
	fnvlist_add_uint64(propval, ZPROP_SOURCE, src);

	if (strval != NULL)
	fnvlist_add_string(propval, ZPROP_VALUE, strval);
	else
	fnvlist_add_uint64(propval, ZPROP_VALUE, intval);

	fnvlist_add_nvlist(nvl, propname, propval);
	nvlist_free(propval);
	}

	/*
	* Add a user property (source=src, propname=propval) to an nvlist.
	*/
	static void
	spa_prop_add_user(nvlist_t nvl, const char propname, char *strval,
	zprop_source_t src)
	{
	nvlist_t *propval;

	VERIFY(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_SLEEP) == 0);
	VERIFY(nvlist_add_uint64(propval, ZPROP_SOURCE, src) == 0);
	VERIFY(nvlist_add_string(propval, ZPROP_VALUE, strval) == 0);
	VERIFY(nvlist_add_nvlist(nvl, propname, propval) == 0);
	nvlist_free(propval);
	}

	/*
	* Get property values from the spa configuration.
	*/
	static void
	spa_prop_get_config(spa_t spa, nvlist_t *nvp)
	{
	vdev_t *rvd = spa->spa_root_vdev;
	dsl_pool_t *pool = spa->spa_dsl_pool;
	uint64_t size, alloc, cap, version;
	const zprop_source_t src = ZPROP_SRC_NONE;
	spa_config_dirent_t *dp;
	metaslab_class_t *mc = spa_normal_class(spa);

	ASSERT(MUTEX_HELD(&spa->spa_props_lock));

	if (rvd != NULL) {
	alloc = metaslab_class_get_alloc(mc);
	alloc += metaslab_class_get_alloc(spa_special_class(spa));
	alloc += metaslab_class_get_alloc(spa_dedup_class(spa));
	alloc += metaslab_class_get_alloc(spa_embedded_log_class(spa));

	size = metaslab_class_get_space(mc);
	size += metaslab_class_get_space(spa_special_class(spa));
	size += metaslab_class_get_space(spa_dedup_class(spa));
	size += metaslab_class_get_space(spa_embedded_log_class(spa));

	spa_prop_add_list(*nvp, ZPOOL_PROP_NAME, spa_name(spa), 0, src);
	spa_prop_add_list(*nvp, ZPOOL_PROP_SIZE, NULL, size, src);
	spa_prop_add_list(*nvp, ZPOOL_PROP_ALLOCATED, NULL, alloc, src);
	spa_prop_add_list(*nvp, ZPOOL_PROP_FREE, NULL,
	size - alloc, src);
	spa_prop_add_list(*nvp, ZPOOL_PROP_CHECKPOINT, NULL,
	spa->spa_checkpoint_info.sci_dspace, src);

	spa_prop_add_list(*nvp, ZPOOL_PROP_FRAGMENTATION, NULL,
	metaslab_class_fragmentation(mc), src);
	spa_prop_add_list(*nvp, ZPOOL_PROP_EXPANDSZ, NULL,
	metaslab_class_expandable_space(mc), src);
	spa_prop_add_list(*nvp, ZPOOL_PROP_READONLY, NULL,
	(spa_mode(spa) == SPA_MODE_READ), src);

	cap = (size == 0) ? 0 : (alloc * 100 / size);
	spa_prop_add_list(*nvp, ZPOOL_PROP_CAPACITY, NULL, cap, src);

	spa_prop_add_list(*nvp, ZPOOL_PROP_DEDUPRATIO, NULL,
	ddt_get_pool_dedup_ratio(spa), src);
	spa_prop_add_list(*nvp, ZPOOL_PROP_BCLONEUSED, NULL,
	brt_get_used(spa), src);
	spa_prop_add_list(*nvp, ZPOOL_PROP_BCLONESAVED, NULL,
	brt_get_saved(spa), src);
	spa_prop_add_list(*nvp, ZPOOL_PROP_BCLONERATIO, NULL,
	brt_get_ratio(spa), src);

	spa_prop_add_list(*nvp, ZPOOL_PROP_HEALTH, NULL,
	rvd->vdev_state, src);

	version = spa_version(spa);
	if (version == zpool_prop_default_numeric(ZPOOL_PROP_VERSION)) {
	spa_prop_add_list(*nvp, ZPOOL_PROP_VERSION, NULL,
	version, ZPROP_SRC_DEFAULT);
	} else {
	spa_prop_add_list(*nvp, ZPOOL_PROP_VERSION, NULL,
	version, ZPROP_SRC_LOCAL);
	}
	spa_prop_add_list(*nvp, ZPOOL_PROP_LOAD_GUID,
	NULL, spa_load_guid(spa), src);
	}

	if (pool != NULL) {
	/*
	* The $FREE directory was introduced in SPA_VERSION_DEADLISTS,
	* when opening pools before this version freedir will be NULL.
	*/
	if (pool->dp_free_dir != NULL) {
	spa_prop_add_list(*nvp, ZPOOL_PROP_FREEING, NULL,
	dsl_dir_phys(pool->dp_free_dir)->dd_used_bytes,
	src);
	} else {
	spa_prop_add_list(*nvp, ZPOOL_PROP_FREEING,
	NULL, 0, src);
	}

	if (pool->dp_leak_dir != NULL) {
	spa_prop_add_list(*nvp, ZPOOL_PROP_LEAKED, NULL,
	dsl_dir_phys(pool->dp_leak_dir)->dd_used_bytes,
	src);
	} else {
	spa_prop_add_list(*nvp, ZPOOL_PROP_LEAKED,
	NULL, 0, src);
	}
	}

	spa_prop_add_list(*nvp, ZPOOL_PROP_GUID, NULL, spa_guid(spa), src);

	if (spa->spa_comment != NULL) {
	spa_prop_add_list(*nvp, ZPOOL_PROP_COMMENT, spa->spa_comment,
	0, ZPROP_SRC_LOCAL);
	}

	if (spa->spa_compatibility != NULL) {
	spa_prop_add_list(*nvp, ZPOOL_PROP_COMPATIBILITY,
	spa->spa_compatibility, 0, ZPROP_SRC_LOCAL);
	}

	if (spa->spa_root != NULL)
	spa_prop_add_list(*nvp, ZPOOL_PROP_ALTROOT, spa->spa_root,
	0, ZPROP_SRC_LOCAL);

	if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS)) {
	spa_prop_add_list(*nvp, ZPOOL_PROP_MAXBLOCKSIZE, NULL,
	MIN(zfs_max_recordsize, SPA_MAXBLOCKSIZE), ZPROP_SRC_NONE);
	} else {
	spa_prop_add_list(*nvp, ZPOOL_PROP_MAXBLOCKSIZE, NULL,
	SPA_OLD_MAXBLOCKSIZE, ZPROP_SRC_NONE);
	}

	if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_DNODE)) {
	spa_prop_add_list(*nvp, ZPOOL_PROP_MAXDNODESIZE, NULL,
	DNODE_MAX_SIZE, ZPROP_SRC_NONE);
	} else {
	spa_prop_add_list(*nvp, ZPOOL_PROP_MAXDNODESIZE, NULL,
	DNODE_MIN_SIZE, ZPROP_SRC_NONE);
	}

	if ((dp = list_head(&spa->spa_config_list)) != NULL) {
	if (dp->scd_path == NULL) {
	spa_prop_add_list(*nvp, ZPOOL_PROP_CACHEFILE,
	"none", 0, ZPROP_SRC_LOCAL);
	} else if (strcmp(dp->scd_path, spa_config_path) != 0) {
	spa_prop_add_list(*nvp, ZPOOL_PROP_CACHEFILE,
	dp->scd_path, 0, ZPROP_SRC_LOCAL);
	}
	}
	}

	/*
	* Get zpool property values.
	*/
	int
	spa_prop_get(spa_t spa, nvlist_t *nvp)
	{
	objset_t *mos = spa->spa_meta_objset;
	zap_cursor_t zc;
	zap_attribute_t za;
	dsl_pool_t *dp;
	int err;

	err = nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_SLEEP);
	if (err)
	return (err);

	dp = spa_get_dsl(spa);
	dsl_pool_config_enter(dp, FTAG);
	mutex_enter(&spa->spa_props_lock);

	/*
	* Get properties from the spa config.
	*/
	spa_prop_get_config(spa, nvp);

	/* If no pool property object, no more prop to get. */
	if (mos == NULL \|\| spa->spa_pool_props_object == 0)
	goto out;

	/*
	* Get properties from the MOS pool property object.
	*/
	for (zap_cursor_init(&zc, mos, spa->spa_pool_props_object);
	(err = zap_cursor_retrieve(&zc, &za)) == 0;
	zap_cursor_advance(&zc)) {
	uint64_t intval = 0;
	char *strval = NULL;
	zprop_source_t src = ZPROP_SRC_DEFAULT;
	zpool_prop_t prop;

	if ((prop = zpool_name_to_prop(za.za_name)) ==
	ZPOOL_PROP_INVAL && !zfs_prop_user(za.za_name))
	continue;

	switch (za.za_integer_length) {
	case 8:
	/* integer property */
	if (za.za_first_integer !=
	zpool_prop_default_numeric(prop))
	src = ZPROP_SRC_LOCAL;

	if (prop == ZPOOL_PROP_BOOTFS) {
	dsl_dataset_t *ds = NULL;

	err = dsl_dataset_hold_obj(dp,
	za.za_first_integer, FTAG, &ds);
	if (err != 0)
	break;

	strval = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN,
	KM_SLEEP);
	dsl_dataset_name(ds, strval);
	dsl_dataset_rele(ds, FTAG);
	} else {
	strval = NULL;
	intval = za.za_first_integer;
	}

	spa_prop_add_list(*nvp, prop, strval, intval, src);

	if (strval != NULL)
	kmem_free(strval, ZFS_MAX_DATASET_NAME_LEN);

	break;

	case 1:
	/* string property */
	strval = kmem_alloc(za.za_num_integers, KM_SLEEP);
	err = zap_lookup(mos, spa->spa_pool_props_object,
	za.za_name, 1, za.za_num_integers, strval);
	if (err) {
	kmem_free(strval, za.za_num_integers);
	break;
	}
	if (prop != ZPOOL_PROP_INVAL) {
	spa_prop_add_list(*nvp, prop, strval, 0, src);
	} else {
	src = ZPROP_SRC_LOCAL;
	spa_prop_add_user(*nvp, za.za_name, strval,
	src);
	}
	kmem_free(strval, za.za_num_integers);
	break;

	default:
	break;
	}
	}
	zap_cursor_fini(&zc);
	out:
	mutex_exit(&spa->spa_props_lock);
	dsl_pool_config_exit(dp, FTAG);
	if (err && err != ENOENT) {
	nvlist_free(*nvp);
	*nvp = NULL;
	return (err);
	}

	return (0);
	}

	/*
	* Validate the given pool properties nvlist and modify the list
	* for the property values to be set.
	*/
	static int
	spa_prop_validate(spa_t spa, nvlist_t props)
	{
	nvpair_t *elem;
	int error = 0, reset_bootfs = 0;
	uint64_t objnum = 0;
	boolean_t has_feature = B_FALSE;

	elem = NULL;
	while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
	uint64_t intval;
	const char strval, slash, check, fname;
	const char *propname = nvpair_name(elem);
	zpool_prop_t prop = zpool_name_to_prop(propname);

	switch (prop) {
	case ZPOOL_PROP_INVAL:
	/*
	* Sanitize the input.
	*/
	if (zfs_prop_user(propname)) {
	if (strlen(propname) >= ZAP_MAXNAMELEN) {
	error = SET_ERROR(ENAMETOOLONG);
	break;
	}

	if (strlen(fnvpair_value_string(elem)) >=
	ZAP_MAXVALUELEN) {
	error = SET_ERROR(E2BIG);
	break;
	}
	} else if (zpool_prop_feature(propname)) {
	if (nvpair_type(elem) != DATA_TYPE_UINT64) {
	error = SET_ERROR(EINVAL);
	break;
	}

	if (nvpair_value_uint64(elem, &intval) != 0) {
	error = SET_ERROR(EINVAL);
	break;
	}

	if (intval != 0) {
	error = SET_ERROR(EINVAL);
	break;
	}

	fname = strchr(propname, '@') + 1;
	if (zfeature_lookup_name(fname, NULL) != 0) {
	error = SET_ERROR(EINVAL);
	break;
	}

	has_feature = B_TRUE;
	} else {
	error = SET_ERROR(EINVAL);
	break;
	}
	break;

	case ZPOOL_PROP_VERSION:
	error = nvpair_value_uint64(elem, &intval);
	if (!error &&
	(intval < spa_version(spa) \|\|
	intval > SPA_VERSION_BEFORE_FEATURES \|\|
	has_feature))
	error = SET_ERROR(EINVAL);
	break;

	case ZPOOL_PROP_DELEGATION:
	case ZPOOL_PROP_AUTOREPLACE:
	case ZPOOL_PROP_LISTSNAPS:
	case ZPOOL_PROP_AUTOEXPAND:
	case ZPOOL_PROP_AUTOTRIM:
	error = nvpair_value_uint64(elem, &intval);
	if (!error && intval > 1)
	error = SET_ERROR(EINVAL);
	break;

	case ZPOOL_PROP_MULTIHOST:
	error = nvpair_value_uint64(elem, &intval);
	if (!error && intval > 1)
	error = SET_ERROR(EINVAL);

	if (!error) {
	uint32_t hostid = zone_get_hostid(NULL);
	if (hostid)
	spa->spa_hostid = hostid;
	else
	error = SET_ERROR(ENOTSUP);
	}

	break;

	case ZPOOL_PROP_BOOTFS:
	/*
	* If the pool version is less than SPA_VERSION_BOOTFS,
	* or the pool is still being created (version == 0),
	* the bootfs property cannot be set.
	*/
	if (spa_version(spa) < SPA_VERSION_BOOTFS) {
	error = SET_ERROR(ENOTSUP);
	break;
	}

	/*
	* Make sure the vdev config is bootable
	*/
	if (!vdev_is_bootable(spa->spa_root_vdev)) {
	error = SET_ERROR(ENOTSUP);
	break;
	}

	reset_bootfs = 1;

	error = nvpair_value_string(elem, &strval);

	if (!error) {
	objset_t *os;

	if (strval == NULL \|\| strval[0] == '\0') {
	objnum = zpool_prop_default_numeric(
	ZPOOL_PROP_BOOTFS);
	break;
	}

	error = dmu_objset_hold(strval, FTAG, &os);
	if (error != 0)
	break;

	/* Must be ZPL. */
	if (dmu_objset_type(os) != DMU_OST_ZFS) {
	error = SET_ERROR(ENOTSUP);
	} else {
	objnum = dmu_objset_id(os);
	}
	dmu_objset_rele(os, FTAG);
	}
	break;

	case ZPOOL_PROP_FAILUREMODE:
	error = nvpair_value_uint64(elem, &intval);
	if (!error && intval > ZIO_FAILURE_MODE_PANIC)
	error = SET_ERROR(EINVAL);

	/*
	* This is a special case which only occurs when
	* the pool has completely failed. This allows
	* the user to change the in-core failmode property
	* without syncing it out to disk (I/Os might
	* currently be blocked). We do this by returning
	* EIO to the caller (spa_prop_set) to trick it
	* into thinking we encountered a property validation
	* error.
	*/
	if (!error && spa_suspended(spa)) {
	spa->spa_failmode = intval;
	error = SET_ERROR(EIO);
	}
	break;

	case ZPOOL_PROP_CACHEFILE:
	if ((error = nvpair_value_string(elem, &strval)) != 0)
	break;

	if (strval[0] == '\0')
	break;

	if (strcmp(strval, "none") == 0)
	break;

	if (strval[0] != '/') {
	error = SET_ERROR(EINVAL);
	break;
	}

	slash = strrchr(strval, '/');
	ASSERT(slash != NULL);

	if (slash[1] == '\0' \|\| strcmp(slash, "/.") == 0 \|\|
	strcmp(slash, "/..") == 0)
	error = SET_ERROR(EINVAL);
	break;

	case ZPOOL_PROP_COMMENT:
	if ((error = nvpair_value_string(elem, &strval)) != 0)
	break;
	for (check = strval; *check != '\0'; check++) {
	if (!isprint(*check)) {
	error = SET_ERROR(EINVAL);
	break;
	}
	}
	if (strlen(strval) > ZPROP_MAX_COMMENT)
	error = SET_ERROR(E2BIG);
	break;

	default:
	break;
	}

	if (error)
	break;
	}

	(void) nvlist_remove_all(props,
	zpool_prop_to_name(ZPOOL_PROP_DEDUPDITTO));

	if (!error && reset_bootfs) {
	error = nvlist_remove(props,
	zpool_prop_to_name(ZPOOL_PROP_BOOTFS), DATA_TYPE_STRING);

	if (!error) {
	error = nvlist_add_uint64(props,
	zpool_prop_to_name(ZPOOL_PROP_BOOTFS), objnum);
	}
	}

	return (error);
	}

	void
	spa_configfile_set(spa_t spa, nvlist_t nvp, boolean_t need_sync)
	{
	const char *cachefile;
	spa_config_dirent_t *dp;

	if (nvlist_lookup_string(nvp, zpool_prop_to_name(ZPOOL_PROP_CACHEFILE),
	&cachefile) != 0)
	return;

	dp = kmem_alloc(sizeof (spa_config_dirent_t),
	KM_SLEEP);

	if (cachefile[0] == '\0')
	dp->scd_path = spa_strdup(spa_config_path);
	else if (strcmp(cachefile, "none") == 0)
	dp->scd_path = NULL;
	else
	dp->scd_path = spa_strdup(cachefile);

	list_insert_head(&spa->spa_config_list, dp);
	if (need_sync)
	spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE);
	}

	int
	spa_prop_set(spa_t spa, nvlist_t nvp)
	{
	int error;
	nvpair_t *elem = NULL;
	boolean_t need_sync = B_FALSE;

	if ((error = spa_prop_validate(spa, nvp)) != 0)
	return (error);

	while ((elem = nvlist_next_nvpair(nvp, elem)) != NULL) {
	zpool_prop_t prop = zpool_name_to_prop(nvpair_name(elem));

	if (prop == ZPOOL_PROP_CACHEFILE \|\|
	prop == ZPOOL_PROP_ALTROOT \|\|
	prop == ZPOOL_PROP_READONLY)
	continue;

	if (prop == ZPOOL_PROP_INVAL &&
	zfs_prop_user(nvpair_name(elem))) {
	need_sync = B_TRUE;
	break;
	}

	if (prop == ZPOOL_PROP_VERSION \|\| prop == ZPOOL_PROP_INVAL) {
	uint64_t ver = 0;

	if (prop == ZPOOL_PROP_VERSION) {
	VERIFY(nvpair_value_uint64(elem, &ver) == 0);
	} else {
	ASSERT(zpool_prop_feature(nvpair_name(elem)));
	ver = SPA_VERSION_FEATURES;
	need_sync = B_TRUE;
	}

	/* Save time if the version is already set. */
	if (ver == spa_version(spa))
	continue;

	/*
	* In addition to the pool directory object, we might
	* create the pool properties object, the features for
	* read object, the features for write object, or the
	* feature descriptions object.
	*/
	error = dsl_sync_task(spa->spa_name, NULL,
	spa_sync_version, &ver,
	6, ZFS_SPACE_CHECK_RESERVED);
	if (error)
	return (error);
	continue;
	}

	need_sync = B_TRUE;
	break;
	}

	if (need_sync) {
	return (dsl_sync_task(spa->spa_name, NULL, spa_sync_props,
	nvp, 6, ZFS_SPACE_CHECK_RESERVED));
	}

	return (0);
	}

	/*
	* If the bootfs property value is dsobj, clear it.
	*/
	void
	spa_prop_clear_bootfs(spa_t spa, uint64_t dsobj, dmu_tx_t tx)
	{
	if (spa->spa_bootfs == dsobj && spa->spa_pool_props_object != 0) {
	VERIFY(zap_remove(spa->spa_meta_objset,
	spa->spa_pool_props_object,
	zpool_prop_to_name(ZPOOL_PROP_BOOTFS), tx) == 0);
	spa->spa_bootfs = 0;
	}
	}

	static int
	spa_change_guid_check(void arg, dmu_tx_t tx)
	{
	uint64_t *newguid __maybe_unused = arg;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	vdev_t *rvd = spa->spa_root_vdev;
	uint64_t vdev_state;

	if (spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT)) {
	int error = (spa_has_checkpoint(spa)) ?
	ZFS_ERR_CHECKPOINT_EXISTS : ZFS_ERR_DISCARDING_CHECKPOINT;
	return (SET_ERROR(error));
	}

	spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
	vdev_state = rvd->vdev_state;
	spa_config_exit(spa, SCL_STATE, FTAG);

	if (vdev_state != VDEV_STATE_HEALTHY)
	return (SET_ERROR(ENXIO));

	ASSERT3U(spa_guid(spa), !=, *newguid);

	return (0);
	}

	static void
	spa_change_guid_sync(void arg, dmu_tx_t tx)
	{
	uint64_t *newguid = arg;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	uint64_t oldguid;
	vdev_t *rvd = spa->spa_root_vdev;

	oldguid = spa_guid(spa);

	spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
	rvd->vdev_guid = *newguid;
	rvd->vdev_guid_sum += (*newguid - oldguid);
	vdev_config_dirty(rvd);
	spa_config_exit(spa, SCL_STATE, FTAG);

	spa_history_log_internal(spa, "guid change", tx, "old=%llu new=%llu",
	(u_longlong_t)oldguid, (u_longlong_t)*newguid);
	}

	/*
	* Change the GUID for the pool. This is done so that we can later
	* re-import a pool built from a clone of our own vdevs. We will modify
	* the root vdev's guid, our own pool guid, and then mark all of our
	* vdevs dirty. Note that we must make sure that all our vdevs are
	* online when we do this, or else any vdevs that weren't present
	* would be orphaned from our pool. We are also going to issue a
	* sysevent to update any watchers.
	*/
	int
	spa_change_guid(spa_t *spa)
	{
	int error;
	uint64_t guid;

	mutex_enter(&spa->spa_vdev_top_lock);
	mutex_enter(&spa_namespace_lock);
	guid = spa_generate_guid(NULL);

	error = dsl_sync_task(spa->spa_name, spa_change_guid_check,
	spa_change_guid_sync, &guid, 5, ZFS_SPACE_CHECK_RESERVED);

	if (error == 0) {
	/*
	* Clear the kobj flag from all the vdevs to allow
	* vdev_cache_process_kobj_evt() to post events to all the
	* vdevs since GUID is updated.
	*/
	vdev_clear_kobj_evt(spa->spa_root_vdev);
	for (int i = 0; i < spa->spa_l2cache.sav_count; i++)
	vdev_clear_kobj_evt(spa->spa_l2cache.sav_vdevs[i]);

	spa_write_cachefile(spa, B_FALSE, B_TRUE, B_TRUE);
	spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_REGUID);
	}

	mutex_exit(&spa_namespace_lock);
	mutex_exit(&spa->spa_vdev_top_lock);

	return (error);
	}

	/*
	* ==========================================================================
	* SPA state manipulation (open/create/destroy/import/export)
	* ==========================================================================
	*/

	static int
	spa_error_entry_compare(const void a, const void b)
	{
	const spa_error_entry_t sa = (const spa_error_entry_t )a;
	const spa_error_entry_t sb = (const spa_error_entry_t )b;
	int ret;

	ret = memcmp(&sa->se_bookmark, &sb->se_bookmark,
	sizeof (zbookmark_phys_t));

	return (TREE_ISIGN(ret));
	}

	/*
	* Utility function which retrieves copies of the current logs and
	* re-initializes them in the process.
	*/
	void
	spa_get_errlists(spa_t spa, avl_tree_t last, avl_tree_t *scrub)
	{
	ASSERT(MUTEX_HELD(&spa->spa_errlist_lock));

	memcpy(last, &spa->spa_errlist_last, sizeof (avl_tree_t));
	memcpy(scrub, &spa->spa_errlist_scrub, sizeof (avl_tree_t));

	avl_create(&spa->spa_errlist_scrub,
	spa_error_entry_compare, sizeof (spa_error_entry_t),
	offsetof(spa_error_entry_t, se_avl));
	avl_create(&spa->spa_errlist_last,
	spa_error_entry_compare, sizeof (spa_error_entry_t),
	offsetof(spa_error_entry_t, se_avl));
	}

	static void
	spa_taskqs_init(spa_t *spa, zio_type_t t, zio_taskq_type_t q)
	{
	const zio_taskq_info_t *ztip = &zio_taskqs[t][q];
	enum zti_modes mode = ztip->zti_mode;
	uint_t value = ztip->zti_value;
	uint_t count = ztip->zti_count;
	spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];
	uint_t cpus, flags = TASKQ_DYNAMIC;

	switch (mode) {
	case ZTI_MODE_FIXED:
	ASSERT3U(value, >, 0);
	break;

	case ZTI_MODE_SYNC:

	/*
	* Create one wr_iss taskq for every 'zio_taskq_write_tpq' CPUs,
	* not to exceed the number of spa allocators, and align to it.
	*/
	cpus = MAX(1, boot_ncpus * zio_taskq_batch_pct / 100);
	count = MAX(1, cpus / MAX(1, zio_taskq_write_tpq));
	count = MAX(count, (zio_taskq_batch_pct + 99) / 100);
	count = MIN(count, spa->spa_alloc_count);
	while (spa->spa_alloc_count % count != 0 &&
	spa->spa_alloc_count < count * 2)
	count--;

	/*
	* zio_taskq_batch_pct is unbounded and may exceed 100%, but no
	* single taskq may have more threads than 100% of online cpus.
	*/
	value = (zio_taskq_batch_pct + count / 2) / count;
	value = MIN(value, 100);
	flags \|= TASKQ_THREADS_CPU_PCT;
	break;

	case ZTI_MODE_SCALE:
	flags \|= TASKQ_THREADS_CPU_PCT;
	/*
	* We want more taskqs to reduce lock contention, but we want
	* less for better request ordering and CPU utilization.
	*/
	cpus = MAX(1, boot_ncpus * zio_taskq_batch_pct / 100);
	if (zio_taskq_batch_tpq > 0) {
	count = MAX(1, (cpus + zio_taskq_batch_tpq / 2) /
	zio_taskq_batch_tpq);
	} else {
	/*
	* Prefer 6 threads per taskq, but no more taskqs
	* than threads in them on large systems. For 80%:
	*
	* taskq taskq total
	* cpus taskqs percent threads threads
	* ------- ------- ------- ------- -------
	* 1 1 80% 1 1
	* 2 1 80% 1 1
	* 4 1 80% 3 3
	* 8 2 40% 3 6
	* 16 3 27% 4 12
	* 32 5 16% 5 25
	* 64 7 11% 7 49
	* 128 10 8% 10 100
	* 256 14 6% 15 210
	*/
	count = 1 + cpus / 6;
	while (count * count > cpus)
	count--;
	}
	/* Limit each taskq within 100% to not trigger assertion. */
	count = MAX(count, (zio_taskq_batch_pct + 99) / 100);
	value = (zio_taskq_batch_pct + count / 2) / count;
	break;

	case ZTI_MODE_NULL:
	tqs->stqs_count = 0;
	tqs->stqs_taskq = NULL;
	return;

	default:
	panic("unrecognized mode for %s_%s taskq (%u:%u) in "
	"spa_taskqs_init()",
	zio_type_name[t], zio_taskq_types[q], mode, value);
	break;
	}

	ASSERT3U(count, >, 0);
	tqs->stqs_count = count;
	tqs->stqs_taskq = kmem_alloc(count * sizeof (taskq_t *), KM_SLEEP);

	for (uint_t i = 0; i < count; i++) {
	taskq_t *tq;
	char name[32];

	if (count > 1)
	(void) snprintf(name, sizeof (name), "%s_%s_%u",
	zio_type_name[t], zio_taskq_types[q], i);
	else
	(void) snprintf(name, sizeof (name), "%s_%s",
	zio_type_name[t], zio_taskq_types[q]);

	#ifdef HAVE_SYSDC
	if (zio_taskq_sysdc && spa->spa_proc != &p0) {
	(void) zio_taskq_basedc;
	tq = taskq_create_sysdc(name, value, 50, INT_MAX,
	spa->spa_proc, zio_taskq_basedc, flags);
	} else {
	#endif
	pri_t pri = maxclsyspri;
	/*
	* The write issue taskq can be extremely CPU
	* intensive. Run it at slightly less important
	* priority than the other taskqs.
	*
	* Under Linux and FreeBSD this means incrementing
	* the priority value as opposed to platforms like
	* illumos where it should be decremented.
	*
	* On FreeBSD, if priorities divided by four (RQ_PPQ)
	* are equal then a difference between them is
	* insignificant.
	*/
	if (t == ZIO_TYPE_WRITE && q == ZIO_TASKQ_ISSUE) {
	#if defined(__linux__)
	pri++;
	#elif defined(__FreeBSD__)
	pri += 4;
	#else
	#error "unknown OS"
	#endif
	}
	tq = taskq_create_proc(name, value, pri, 50,
	INT_MAX, spa->spa_proc, flags);
	#ifdef HAVE_SYSDC
	}
	#endif

	tqs->stqs_taskq[i] = tq;
	}
	}

	static void
	spa_taskqs_fini(spa_t *spa, zio_type_t t, zio_taskq_type_t q)
	{
	spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];

	if (tqs->stqs_taskq == NULL) {
	ASSERT3U(tqs->stqs_count, ==, 0);
	return;
	}

	for (uint_t i = 0; i < tqs->stqs_count; i++) {
	ASSERT3P(tqs->stqs_taskq[i], !=, NULL);
	taskq_destroy(tqs->stqs_taskq[i]);
	}

	kmem_free(tqs->stqs_taskq, tqs->stqs_count * sizeof (taskq_t *));
	tqs->stqs_taskq = NULL;
	}

	#ifdef _KERNEL
	/*
	* The READ and WRITE rows of zio_taskqs are configurable at module load time
	* by setting zio_taskq_read or zio_taskq_write.
	*
	* Example (the defaults for READ and WRITE)
	* zio_taskq_read='fixed,1,8 null scale null'
	- * zio_taskq_write='sync fixed,1,5 scale fixed,1,5'
	+ * zio_taskq_write='sync null scale null'
	*
	* Each sets the entire row at a time.
	*
	* 'fixed' is parameterised: fixed,Q,T where Q is number of taskqs, T is number
	* of threads per taskq.
	*
	* 'null' can only be set on the high-priority queues (queue selection for
	* high-priority queues will fall back to the regular queue if the high-pri
	* is NULL.
	*/
	static const char *const modes[ZTI_NMODES] = {
	"fixed", "scale", "sync", "null"
	};

	/* Parse the incoming config string. Modifies cfg */
	static int
	spa_taskq_param_set(zio_type_t t, char *cfg)
	{
	int err = 0;

	zio_taskq_info_t row[ZIO_TASKQ_TYPES] = {{0}};

	char next = cfg, tok, *c;

	/*
	* Parse out each element from the string and fill `row`. The entire
	* row has to be set at once, so any errors are flagged by just
	* breaking out of this loop early.
	*/
	uint_t q;
	for (q = 0; q < ZIO_TASKQ_TYPES; q++) {
	/* `next` is the start of the config */
	if (next == NULL)
	break;

	/* Eat up leading space */
	while (isspace(*next))
	next++;
	if (*next == '\0')
	break;

	/* Mode ends at space or end of string */
	tok = next;
	next = strchr(tok, ' ');
	if (next != NULL) *next++ = '\0';

	/* Parameters start after a comma */
	c = strchr(tok, ',');
	if (c != NULL) *c++ = '\0';

	/* Match mode string */
	uint_t mode;
	for (mode = 0; mode < ZTI_NMODES; mode++)
	if (strcmp(tok, modes[mode]) == 0)
	break;
	if (mode == ZTI_NMODES)
	break;

	/* Invalid canary */
	row[q].zti_mode = ZTI_NMODES;

	/* Per-mode setup */
	switch (mode) {

	/*
	* FIXED is parameterised: number of queues, and number of
	* threads per queue.
	*/
	case ZTI_MODE_FIXED: {
	/* No parameters? */
	if (c == NULL \|\| *c == '\0')
	break;

	/* Find next parameter */
	tok = c;
	c = strchr(tok, ',');
	if (c == NULL)
	break;

	/* Take digits and convert */
	unsigned long long nq;
	if (!(isdigit(*tok)))
	break;
	err = ddi_strtoull(tok, &tok, 10, &nq);
	/* Must succeed and also end at the next param sep */
	if (err != 0 \|\| tok != c)
	break;

	/* Move past the comma */
	tok++;
	/* Need another number */
	if (!(isdigit(*tok)))
	break;
	/* Remember start to make sure we moved */
	c = tok;

	/* Take digits */
	unsigned long long ntpq;
	err = ddi_strtoull(tok, &tok, 10, &ntpq);
	/* Must succeed, and moved forward */
	if (err != 0 \|\| tok == c \|\| *tok != '\0')
	break;

	/*
	* sanity; zero queues/threads make no sense, and
	* 16K is almost certainly more than anyone will ever
	* need and avoids silly numbers like UINT32_MAX
	*/
	if (nq == 0 \|\| nq >= 16384 \|\|
	ntpq == 0 \|\| ntpq >= 16384)
	break;

	const zio_taskq_info_t zti = ZTI_P(ntpq, nq);
	row[q] = zti;
	break;
	}

	case ZTI_MODE_SCALE: {
	const zio_taskq_info_t zti = ZTI_SCALE;
	row[q] = zti;
	break;
	}

	case ZTI_MODE_SYNC: {
	const zio_taskq_info_t zti = ZTI_SYNC;
	row[q] = zti;
	break;
	}

	case ZTI_MODE_NULL: {
	/*
	* Can only null the high-priority queues; the general-
	* purpose ones have to exist.
	*/
	if (q != ZIO_TASKQ_ISSUE_HIGH &&
	q != ZIO_TASKQ_INTERRUPT_HIGH)
	break;

	const zio_taskq_info_t zti = ZTI_NULL;
	row[q] = zti;
	break;
	}

	default:
	break;
	}

	/* Ensure we set a mode */
	if (row[q].zti_mode == ZTI_NMODES)
	break;
	}

	/* Didn't get a full row, fail */
	if (q < ZIO_TASKQ_TYPES)
	return (SET_ERROR(EINVAL));

	/* Eat trailing space */
	if (next != NULL)
	while (isspace(*next))
	next++;

	/* If there's anything left over then fail */
	if (next != NULL && *next != '\0')
	return (SET_ERROR(EINVAL));

	/* Success! Copy it into the real config */
	for (q = 0; q < ZIO_TASKQ_TYPES; q++)
	zio_taskqs[t][q] = row[q];

	return (0);
	}

	static int
	spa_taskq_param_get(zio_type_t t, char *buf, boolean_t add_newline)
	{
	int pos = 0;

	/* Build paramater string from live config */
	const char *sep = "";
	for (uint_t q = 0; q < ZIO_TASKQ_TYPES; q++) {
	const zio_taskq_info_t *zti = &zio_taskqs[t][q];
	if (zti->zti_mode == ZTI_MODE_FIXED)
	pos += sprintf(&buf[pos], "%s%s,%u,%u", sep,
	modes[zti->zti_mode], zti->zti_count,
	zti->zti_value);
	else
	pos += sprintf(&buf[pos], "%s%s", sep,
	modes[zti->zti_mode]);
	sep = " ";
	}

	if (add_newline)
	buf[pos++] = '\n';
	buf[pos] = '\0';

	return (pos);
	}

	#ifdef __linux__
	static int
	spa_taskq_read_param_set(const char val, zfs_kernel_param_t kp)
	{
	char *cfg = kmem_strdup(val);
	int err = spa_taskq_param_set(ZIO_TYPE_READ, cfg);
	kmem_free(cfg, strlen(val)+1);
	return (-err);
	}
	static int
	spa_taskq_read_param_get(char buf, zfs_kernel_param_t kp)
	{
	return (spa_taskq_param_get(ZIO_TYPE_READ, buf, TRUE));
	}

	static int
	spa_taskq_write_param_set(const char val, zfs_kernel_param_t kp)
	{
	char *cfg = kmem_strdup(val);
	int err = spa_taskq_param_set(ZIO_TYPE_WRITE, cfg);
	kmem_free(cfg, strlen(val)+1);
	return (-err);
	}
	static int
	spa_taskq_write_param_get(char buf, zfs_kernel_param_t kp)
	{
	return (spa_taskq_param_get(ZIO_TYPE_WRITE, buf, TRUE));
	}
	#else
	/*
	* On FreeBSD load-time parameters can be set up before malloc() is available,
	* so we have to do all the parsing work on the stack.
	*/
	#define SPA_TASKQ_PARAM_MAX (128)

	static int
	spa_taskq_read_param(ZFS_MODULE_PARAM_ARGS)
	{
	char buf[SPA_TASKQ_PARAM_MAX];
	int err;

	(void) spa_taskq_param_get(ZIO_TYPE_READ, buf, FALSE);
	err = sysctl_handle_string(oidp, buf, sizeof (buf), req);
	if (err \|\| req->newptr == NULL)
	return (err);
	return (spa_taskq_param_set(ZIO_TYPE_READ, buf));
	}

	static int
	spa_taskq_write_param(ZFS_MODULE_PARAM_ARGS)
	{
	char buf[SPA_TASKQ_PARAM_MAX];
	int err;

	(void) spa_taskq_param_get(ZIO_TYPE_WRITE, buf, FALSE);
	err = sysctl_handle_string(oidp, buf, sizeof (buf), req);
	if (err \|\| req->newptr == NULL)
	return (err);
	return (spa_taskq_param_set(ZIO_TYPE_WRITE, buf));
	}
	#endif
	#endif /* _KERNEL */

	/*
	* Dispatch a task to the appropriate taskq for the ZFS I/O type and priority.
	* Note that a type may have multiple discrete taskqs to avoid lock contention
	* on the taskq itself.
	*/
	-static taskq_t *
	-spa_taskq_dispatch_select(spa_t *spa, zio_type_t t, zio_taskq_type_t q,
	- zio_t *zio)
	+void
	+spa_taskq_dispatch(spa_t *spa, zio_type_t t, zio_taskq_type_t q,
	+ task_func_t func, zio_t zio, boolean_t cutinline)
	{
	spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];
	taskq_t *tq;

	ASSERT3P(tqs->stqs_taskq, !=, NULL);
	ASSERT3U(tqs->stqs_count, !=, 0);

	+ /*
	+ * NB: We are assuming that the zio can only be dispatched
	+ * to a single taskq at a time. It would be a grievous error
	+ * to dispatch the zio to another taskq at the same time.
	+ */
	+ ASSERT(zio);
	+ ASSERT(taskq_empty_ent(&zio->io_tqent));
	+
	if (tqs->stqs_count == 1) {
	tq = tqs->stqs_taskq[0];
	} else if ((t == ZIO_TYPE_WRITE) && (q == ZIO_TASKQ_ISSUE) &&
	- (zio != NULL) && ZIO_HAS_ALLOCATOR(zio)) {
	+ ZIO_HAS_ALLOCATOR(zio)) {
	tq = tqs->stqs_taskq[zio->io_allocator % tqs->stqs_count];
	} else {
	tq = tqs->stqs_taskq[((uint64_t)gethrtime()) % tqs->stqs_count];
	}
	- return (tq);
	-}
	-
	-void
	-spa_taskq_dispatch_ent(spa_t *spa, zio_type_t t, zio_taskq_type_t q,
	- task_func_t func, void arg, uint_t flags, taskq_ent_t *ent,
	- zio_t *zio)
	-{
	- taskq_t *tq = spa_taskq_dispatch_select(spa, t, q, zio);
	- taskq_dispatch_ent(tq, func, arg, flags, ent);
	-}

	-/*
	- * Same as spa_taskq_dispatch_ent() but block on the task until completion.
	- */
	-void
	-spa_taskq_dispatch_sync(spa_t *spa, zio_type_t t, zio_taskq_type_t q,
	- task_func_t func, void arg, uint_t flags)
	-{
	- taskq_t *tq = spa_taskq_dispatch_select(spa, t, q, NULL);
	- taskqid_t id = taskq_dispatch(tq, func, arg, flags);
	- if (id)
	- taskq_wait_id(tq, id);
	+ taskq_dispatch_ent(tq, func, zio, cutinline ? TQ_FRONT : 0,
	+ &zio->io_tqent);
	}

	static void
	spa_create_zio_taskqs(spa_t *spa)
	{
	for (int t = 0; t < ZIO_TYPES; t++) {
	for (int q = 0; q < ZIO_TASKQ_TYPES; q++) {
	spa_taskqs_init(spa, t, q);
	}
	}
	}

	#if defined(_KERNEL) && defined(HAVE_SPA_THREAD)
	static void
	spa_thread(void *arg)
	{
	psetid_t zio_taskq_psrset_bind = PS_NONE;
	callb_cpr_t cprinfo;

	spa_t *spa = arg;
	user_t *pu = PTOU(curproc);

	CALLB_CPR_INIT(&cprinfo, &spa->spa_proc_lock, callb_generic_cpr,
	spa->spa_name);

	ASSERT(curproc != &p0);
	(void) snprintf(pu->u_psargs, sizeof (pu->u_psargs),
	"zpool-%s", spa->spa_name);
	(void) strlcpy(pu->u_comm, pu->u_psargs, sizeof (pu->u_comm));

	/* bind this thread to the requested psrset */
	if (zio_taskq_psrset_bind != PS_NONE) {
	pool_lock();
	mutex_enter(&cpu_lock);
	mutex_enter(&pidlock);
	mutex_enter(&curproc->p_lock);

	if (cpupart_bind_thread(curthread, zio_taskq_psrset_bind,
	0, NULL, NULL) == 0) {
	curthread->t_bind_pset = zio_taskq_psrset_bind;
	} else {
	cmn_err(CE_WARN,
	"Couldn't bind process for zfs pool \"%s\" to "
	"pset %d\n", spa->spa_name, zio_taskq_psrset_bind);
	}

	mutex_exit(&curproc->p_lock);
	mutex_exit(&pidlock);
	mutex_exit(&cpu_lock);
	pool_unlock();
	}

	#ifdef HAVE_SYSDC
	if (zio_taskq_sysdc) {
	sysdc_thread_enter(curthread, 100, 0);
	}
	#endif

	spa->spa_proc = curproc;
	spa->spa_did = curthread->t_did;

	spa_create_zio_taskqs(spa);

	mutex_enter(&spa->spa_proc_lock);
	ASSERT(spa->spa_proc_state == SPA_PROC_CREATED);

	spa->spa_proc_state = SPA_PROC_ACTIVE;
	cv_broadcast(&spa->spa_proc_cv);

	CALLB_CPR_SAFE_BEGIN(&cprinfo);
	while (spa->spa_proc_state == SPA_PROC_ACTIVE)
	cv_wait(&spa->spa_proc_cv, &spa->spa_proc_lock);
	CALLB_CPR_SAFE_END(&cprinfo, &spa->spa_proc_lock);

	ASSERT(spa->spa_proc_state == SPA_PROC_DEACTIVATE);
	spa->spa_proc_state = SPA_PROC_GONE;
	spa->spa_proc = &p0;
	cv_broadcast(&spa->spa_proc_cv);
	CALLB_CPR_EXIT(&cprinfo); /* drops spa_proc_lock */

	mutex_enter(&curproc->p_lock);
	lwp_exit();
	}
	#endif

	extern metaslab_ops_t metaslab_allocator(spa_t spa);

	/*
	* Activate an uninitialized pool.
	*/
	static void
	spa_activate(spa_t *spa, spa_mode_t mode)
	{
	metaslab_ops_t *msp = metaslab_allocator(spa);
	ASSERT(spa->spa_state == POOL_STATE_UNINITIALIZED);

	spa->spa_state = POOL_STATE_ACTIVE;
	spa->spa_mode = mode;
	spa->spa_read_spacemaps = spa_mode_readable_spacemaps;

	spa->spa_normal_class = metaslab_class_create(spa, msp);
	spa->spa_log_class = metaslab_class_create(spa, msp);
	spa->spa_embedded_log_class = metaslab_class_create(spa, msp);
	spa->spa_special_class = metaslab_class_create(spa, msp);
	spa->spa_dedup_class = metaslab_class_create(spa, msp);

	/* Try to create a covering process */
	mutex_enter(&spa->spa_proc_lock);
	ASSERT(spa->spa_proc_state == SPA_PROC_NONE);
	ASSERT(spa->spa_proc == &p0);
	spa->spa_did = 0;

	#ifdef HAVE_SPA_THREAD
	/* Only create a process if we're going to be around a while. */
	if (spa_create_process && strcmp(spa->spa_name, TRYIMPORT_NAME) != 0) {
	if (newproc(spa_thread, (caddr_t)spa, syscid, maxclsyspri,
	NULL, 0) == 0) {
	spa->spa_proc_state = SPA_PROC_CREATED;
	while (spa->spa_proc_state == SPA_PROC_CREATED) {
	cv_wait(&spa->spa_proc_cv,
	&spa->spa_proc_lock);
	}
	ASSERT(spa->spa_proc_state == SPA_PROC_ACTIVE);
	ASSERT(spa->spa_proc != &p0);
	ASSERT(spa->spa_did != 0);
	} else {
	#ifdef _KERNEL
	cmn_err(CE_WARN,
	"Couldn't create process for zfs pool \"%s\"\n",
	spa->spa_name);
	#endif
	}
	}
	#endif /* HAVE_SPA_THREAD */
	mutex_exit(&spa->spa_proc_lock);

	/* If we didn't create a process, we need to create our taskqs. */
	if (spa->spa_proc == &p0) {
	spa_create_zio_taskqs(spa);
	}

	for (size_t i = 0; i < TXG_SIZE; i++) {
	spa->spa_txg_zio[i] = zio_root(spa, NULL, NULL,
	ZIO_FLAG_CANFAIL);
	}

	list_create(&spa->spa_config_dirty_list, sizeof (vdev_t),
	offsetof(vdev_t, vdev_config_dirty_node));
	list_create(&spa->spa_evicting_os_list, sizeof (objset_t),
	offsetof(objset_t, os_evicting_node));
	list_create(&spa->spa_state_dirty_list, sizeof (vdev_t),
	offsetof(vdev_t, vdev_state_dirty_node));

	txg_list_create(&spa->spa_vdev_txg_list, spa,
	offsetof(struct vdev, vdev_txg_node));

	avl_create(&spa->spa_errlist_scrub,
	spa_error_entry_compare, sizeof (spa_error_entry_t),
	offsetof(spa_error_entry_t, se_avl));
	avl_create(&spa->spa_errlist_last,
	spa_error_entry_compare, sizeof (spa_error_entry_t),
	offsetof(spa_error_entry_t, se_avl));
	avl_create(&spa->spa_errlist_healed,
	spa_error_entry_compare, sizeof (spa_error_entry_t),
	offsetof(spa_error_entry_t, se_avl));

	spa_activate_os(spa);

	spa_keystore_init(&spa->spa_keystore);

	/*
	* This taskq is used to perform zvol-minor-related tasks
	* asynchronously. This has several advantages, including easy
	* resolution of various deadlocks.
	*
	* The taskq must be single threaded to ensure tasks are always
	* processed in the order in which they were dispatched.
	*
	* A taskq per pool allows one to keep the pools independent.
	* This way if one pool is suspended, it will not impact another.
	*
	* The preferred location to dispatch a zvol minor task is a sync
	* task. In this context, there is easy access to the spa_t and minimal
	* error handling is required because the sync task must succeed.
	*/
	spa->spa_zvol_taskq = taskq_create("z_zvol", 1, defclsyspri,
	1, INT_MAX, 0);

	/*
	* The taskq to preload metaslabs.
	*/
	spa->spa_metaslab_taskq = taskq_create("z_metaslab",
	metaslab_preload_pct, maxclsyspri, 1, INT_MAX,
	TASKQ_DYNAMIC \| TASKQ_THREADS_CPU_PCT);

	/*
	* Taskq dedicated to prefetcher threads: this is used to prevent the
	* pool traverse code from monopolizing the global (and limited)
	* system_taskq by inappropriately scheduling long running tasks on it.
	*/
	spa->spa_prefetch_taskq = taskq_create("z_prefetch", 100,
	defclsyspri, 1, INT_MAX, TASKQ_DYNAMIC \| TASKQ_THREADS_CPU_PCT);

	/*
	* The taskq to upgrade datasets in this pool. Currently used by
	* feature SPA_FEATURE_USEROBJ_ACCOUNTING/SPA_FEATURE_PROJECT_QUOTA.
	*/
	spa->spa_upgrade_taskq = taskq_create("z_upgrade", 100,
	defclsyspri, 1, INT_MAX, TASKQ_DYNAMIC \| TASKQ_THREADS_CPU_PCT);
	}

	/*
	* Opposite of spa_activate().
	*/
	static void
	spa_deactivate(spa_t *spa)
	{
	ASSERT(spa->spa_sync_on == B_FALSE);
	ASSERT(spa->spa_dsl_pool == NULL);
	ASSERT(spa->spa_root_vdev == NULL);
	ASSERT(spa->spa_async_zio_root == NULL);
	ASSERT(spa->spa_state != POOL_STATE_UNINITIALIZED);

	spa_evicting_os_wait(spa);

	if (spa->spa_zvol_taskq) {
	taskq_destroy(spa->spa_zvol_taskq);
	spa->spa_zvol_taskq = NULL;
	}

	if (spa->spa_metaslab_taskq) {
	taskq_destroy(spa->spa_metaslab_taskq);
	spa->spa_metaslab_taskq = NULL;
	}

	if (spa->spa_prefetch_taskq) {
	taskq_destroy(spa->spa_prefetch_taskq);
	spa->spa_prefetch_taskq = NULL;
	}

	if (spa->spa_upgrade_taskq) {
	taskq_destroy(spa->spa_upgrade_taskq);
	spa->spa_upgrade_taskq = NULL;
	}

	txg_list_destroy(&spa->spa_vdev_txg_list);

	list_destroy(&spa->spa_config_dirty_list);
	list_destroy(&spa->spa_evicting_os_list);
	list_destroy(&spa->spa_state_dirty_list);

	taskq_cancel_id(system_delay_taskq, spa->spa_deadman_tqid);

	for (int t = 0; t < ZIO_TYPES; t++) {
	for (int q = 0; q < ZIO_TASKQ_TYPES; q++) {
	spa_taskqs_fini(spa, t, q);
	}
	}

	for (size_t i = 0; i < TXG_SIZE; i++) {
	ASSERT3P(spa->spa_txg_zio[i], !=, NULL);
	VERIFY0(zio_wait(spa->spa_txg_zio[i]));
	spa->spa_txg_zio[i] = NULL;
	}

	metaslab_class_destroy(spa->spa_normal_class);
	spa->spa_normal_class = NULL;

	metaslab_class_destroy(spa->spa_log_class);
	spa->spa_log_class = NULL;

	metaslab_class_destroy(spa->spa_embedded_log_class);
	spa->spa_embedded_log_class = NULL;

	metaslab_class_destroy(spa->spa_special_class);
	spa->spa_special_class = NULL;

	metaslab_class_destroy(spa->spa_dedup_class);
	spa->spa_dedup_class = NULL;

	/*
	* If this was part of an import or the open otherwise failed, we may
	* still have errors left in the queues. Empty them just in case.
	*/
	spa_errlog_drain(spa);
	avl_destroy(&spa->spa_errlist_scrub);
	avl_destroy(&spa->spa_errlist_last);
	avl_destroy(&spa->spa_errlist_healed);

	spa_keystore_fini(&spa->spa_keystore);

	spa->spa_state = POOL_STATE_UNINITIALIZED;

	mutex_enter(&spa->spa_proc_lock);
	if (spa->spa_proc_state != SPA_PROC_NONE) {
	ASSERT(spa->spa_proc_state == SPA_PROC_ACTIVE);
	spa->spa_proc_state = SPA_PROC_DEACTIVATE;
	cv_broadcast(&spa->spa_proc_cv);
	while (spa->spa_proc_state == SPA_PROC_DEACTIVATE) {
	ASSERT(spa->spa_proc != &p0);
	cv_wait(&spa->spa_proc_cv, &spa->spa_proc_lock);
	}
	ASSERT(spa->spa_proc_state == SPA_PROC_GONE);
	spa->spa_proc_state = SPA_PROC_NONE;
	}
	ASSERT(spa->spa_proc == &p0);
	mutex_exit(&spa->spa_proc_lock);

	/*
	* We want to make sure spa_thread() has actually exited the ZFS
	* module, so that the module can't be unloaded out from underneath
	* it.
	*/
	if (spa->spa_did != 0) {
	thread_join(spa->spa_did);
	spa->spa_did = 0;
	}

	spa_deactivate_os(spa);

	}

	/*
	* Verify a pool configuration, and construct the vdev tree appropriately. This
	* will create all the necessary vdevs in the appropriate layout, with each vdev
	* in the CLOSED state. This will prep the pool before open/creation/import.
	* All vdev validation is done by the vdev_alloc() routine.
	*/
	int
	spa_config_parse(spa_t spa, vdev_t vdp, nvlist_t nv, vdev_t *parent,
	uint_t id, int atype)
	{
	nvlist_t **child;
	uint_t children;
	int error;

	if ((error = vdev_alloc(spa, vdp, nv, parent, id, atype)) != 0)
	return (error);

	if ((*vdp)->vdev_ops->vdev_op_leaf)
	return (0);

	error = nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	&child, &children);

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

	if (error) {
	vdev_free(*vdp);
	*vdp = NULL;
	return (SET_ERROR(EINVAL));
	}

	for (int c = 0; c < children; c++) {
	vdev_t *vd;
	if ((error = spa_config_parse(spa, &vd, child[c], *vdp, c,
	atype)) != 0) {
	vdev_free(*vdp);
	*vdp = NULL;
	return (error);
	}
	}

	ASSERT(*vdp != NULL);

	return (0);
	}

	static boolean_t
	spa_should_flush_logs_on_unload(spa_t *spa)
	{
	if (!spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP))
	return (B_FALSE);

	if (!spa_writeable(spa))
	return (B_FALSE);

	if (!spa->spa_sync_on)
	return (B_FALSE);

	if (spa_state(spa) != POOL_STATE_EXPORTED)
	return (B_FALSE);

	if (zfs_keep_log_spacemaps_at_export)
	return (B_FALSE);

	return (B_TRUE);
	}

	/*
	* Opens a transaction that will set the flag that will instruct
	* spa_sync to attempt to flush all the metaslabs for that txg.
	*/
	static void
	spa_unload_log_sm_flush_all(spa_t *spa)
	{
	dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
	VERIFY0(dmu_tx_assign(tx, TXG_WAIT));

	ASSERT3U(spa->spa_log_flushall_txg, ==, 0);
	spa->spa_log_flushall_txg = dmu_tx_get_txg(tx);

	dmu_tx_commit(tx);
	txg_wait_synced(spa_get_dsl(spa), spa->spa_log_flushall_txg);
	}

	static void
	spa_unload_log_sm_metadata(spa_t *spa)
	{
	void *cookie = NULL;
	spa_log_sm_t *sls;
	log_summary_entry_t *e;

	while ((sls = avl_destroy_nodes(&spa->spa_sm_logs_by_txg,
	&cookie)) != NULL) {
	VERIFY0(sls->sls_mscount);
	kmem_free(sls, sizeof (spa_log_sm_t));
	}

	while ((e = list_remove_head(&spa->spa_log_summary)) != NULL) {
	VERIFY0(e->lse_mscount);
	kmem_free(e, sizeof (log_summary_entry_t));
	}

	spa->spa_unflushed_stats.sus_nblocks = 0;
	spa->spa_unflushed_stats.sus_memused = 0;
	spa->spa_unflushed_stats.sus_blocklimit = 0;
	}

	static void
	spa_destroy_aux_threads(spa_t *spa)
	{
	if (spa->spa_condense_zthr != NULL) {
	zthr_destroy(spa->spa_condense_zthr);
	spa->spa_condense_zthr = NULL;
	}
	if (spa->spa_checkpoint_discard_zthr != NULL) {
	zthr_destroy(spa->spa_checkpoint_discard_zthr);
	spa->spa_checkpoint_discard_zthr = NULL;
	}
	if (spa->spa_livelist_delete_zthr != NULL) {
	zthr_destroy(spa->spa_livelist_delete_zthr);
	spa->spa_livelist_delete_zthr = NULL;
	}
	if (spa->spa_livelist_condense_zthr != NULL) {
	zthr_destroy(spa->spa_livelist_condense_zthr);
	spa->spa_livelist_condense_zthr = NULL;
	}
	if (spa->spa_raidz_expand_zthr != NULL) {
	zthr_destroy(spa->spa_raidz_expand_zthr);
	spa->spa_raidz_expand_zthr = NULL;
	}
	}

	/*
	* Opposite of spa_load().
	*/
	static void
	spa_unload(spa_t *spa)
	{
	- ASSERT(MUTEX_HELD(&spa_namespace_lock));
	+ ASSERT(MUTEX_HELD(&spa_namespace_lock) \|\|
	+ spa->spa_export_thread == curthread);
	ASSERT(spa_state(spa) != POOL_STATE_UNINITIALIZED);

	spa_import_progress_remove(spa_guid(spa));
	spa_load_note(spa, "UNLOADING");

	spa_wake_waiters(spa);

	/*
	* If we have set the spa_final_txg, we have already performed the
	* tasks below in spa_export_common(). We should not redo it here since
	* we delay the final TXGs beyond what spa_final_txg is set at.
	*/
	if (spa->spa_final_txg == UINT64_MAX) {
	/*
	* If the log space map feature is enabled and the pool is
	* getting exported (but not destroyed), we want to spend some
	* time flushing as many metaslabs as we can in an attempt to
	* destroy log space maps and save import time.
	*/
	if (spa_should_flush_logs_on_unload(spa))
	spa_unload_log_sm_flush_all(spa);

	/*
	* Stop async tasks.
	*/
	spa_async_suspend(spa);

	if (spa->spa_root_vdev) {
	vdev_t *root_vdev = spa->spa_root_vdev;
	vdev_initialize_stop_all(root_vdev,
	VDEV_INITIALIZE_ACTIVE);
	vdev_trim_stop_all(root_vdev, VDEV_TRIM_ACTIVE);
	vdev_autotrim_stop_all(spa);
	vdev_rebuild_stop_all(spa);
	}
	}

	/*
	* Stop syncing.
	*/
	if (spa->spa_sync_on) {
	txg_sync_stop(spa->spa_dsl_pool);
	spa->spa_sync_on = B_FALSE;
	}

	/*
	* This ensures that there is no async metaslab prefetching
	* while we attempt to unload the spa.
	*/
	taskq_wait(spa->spa_metaslab_taskq);

	if (spa->spa_mmp.mmp_thread)
	mmp_thread_stop(spa);

	/*
	* Wait for any outstanding async I/O to complete.
	*/
	if (spa->spa_async_zio_root != NULL) {
	for (int i = 0; i < max_ncpus; i++)
	(void) zio_wait(spa->spa_async_zio_root[i]);
	kmem_free(spa->spa_async_zio_root, max_ncpus * sizeof (void *));
	spa->spa_async_zio_root = NULL;
	}

	if (spa->spa_vdev_removal != NULL) {
	spa_vdev_removal_destroy(spa->spa_vdev_removal);
	spa->spa_vdev_removal = NULL;
	}

	spa_destroy_aux_threads(spa);

	spa_condense_fini(spa);

	bpobj_close(&spa->spa_deferred_bpobj);

	spa_config_enter(spa, SCL_ALL, spa, RW_WRITER);

	/*
	* Close all vdevs.
	*/
	if (spa->spa_root_vdev)
	vdev_free(spa->spa_root_vdev);
	ASSERT(spa->spa_root_vdev == NULL);

	/*
	* Close the dsl pool.
	*/
	if (spa->spa_dsl_pool) {
	dsl_pool_close(spa->spa_dsl_pool);
	spa->spa_dsl_pool = NULL;
	spa->spa_meta_objset = NULL;
	}

	ddt_unload(spa);
	brt_unload(spa);
	spa_unload_log_sm_metadata(spa);

	/*
	* Drop and purge level 2 cache
	*/
	spa_l2cache_drop(spa);

	if (spa->spa_spares.sav_vdevs) {
	for (int i = 0; i < spa->spa_spares.sav_count; i++)
	vdev_free(spa->spa_spares.sav_vdevs[i]);
	kmem_free(spa->spa_spares.sav_vdevs,
	spa->spa_spares.sav_count * sizeof (void *));
	spa->spa_spares.sav_vdevs = NULL;
	}
	if (spa->spa_spares.sav_config) {
	nvlist_free(spa->spa_spares.sav_config);
	spa->spa_spares.sav_config = NULL;
	}
	spa->spa_spares.sav_count = 0;

	if (spa->spa_l2cache.sav_vdevs) {
	for (int i = 0; i < spa->spa_l2cache.sav_count; i++) {
	vdev_clear_stats(spa->spa_l2cache.sav_vdevs[i]);
	vdev_free(spa->spa_l2cache.sav_vdevs[i]);
	}
	kmem_free(spa->spa_l2cache.sav_vdevs,
	spa->spa_l2cache.sav_count * sizeof (void *));
	spa->spa_l2cache.sav_vdevs = NULL;
	}
	if (spa->spa_l2cache.sav_config) {
	nvlist_free(spa->spa_l2cache.sav_config);
	spa->spa_l2cache.sav_config = NULL;
	}
	spa->spa_l2cache.sav_count = 0;

	spa->spa_async_suspended = 0;

	spa->spa_indirect_vdevs_loaded = B_FALSE;

	if (spa->spa_comment != NULL) {
	spa_strfree(spa->spa_comment);
	spa->spa_comment = NULL;
	}
	if (spa->spa_compatibility != NULL) {
	spa_strfree(spa->spa_compatibility);
	spa->spa_compatibility = NULL;
	}

	spa->spa_raidz_expand = NULL;

	spa_config_exit(spa, SCL_ALL, spa);
	}

	/*
	* Load (or re-load) the current list of vdevs describing the active spares for
	* this pool. When this is called, we have some form of basic information in
	* 'spa_spares.sav_config'. We parse this into vdevs, try to open them, and
	* then re-generate a more complete list including status information.
	*/
	void
	spa_load_spares(spa_t *spa)
	{
	nvlist_t **spares;
	uint_t nspares;
	int i;
	vdev_t vd, tvd;

	#ifndef _KERNEL
	/*
	* zdb opens both the current state of the pool and the
	* checkpointed state (if present), with a different spa_t.
	*
	* As spare vdevs are shared among open pools, we skip loading
	* them when we load the checkpointed state of the pool.
	*/
	if (!spa_writeable(spa))
	return;
	#endif

	ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);

	/*
	* First, close and free any existing spare vdevs.
	*/
	if (spa->spa_spares.sav_vdevs) {
	for (i = 0; i < spa->spa_spares.sav_count; i++) {
	vd = spa->spa_spares.sav_vdevs[i];

	/* Undo the call to spa_activate() below */
	if ((tvd = spa_lookup_by_guid(spa, vd->vdev_guid,
	B_FALSE)) != NULL && tvd->vdev_isspare)
	spa_spare_remove(tvd);
	vdev_close(vd);
	vdev_free(vd);
	}

	kmem_free(spa->spa_spares.sav_vdevs,
	spa->spa_spares.sav_count * sizeof (void *));
	}

	if (spa->spa_spares.sav_config == NULL)
	nspares = 0;
	else
	VERIFY0(nvlist_lookup_nvlist_array(spa->spa_spares.sav_config,
	ZPOOL_CONFIG_SPARES, &spares, &nspares));

	spa->spa_spares.sav_count = (int)nspares;
	spa->spa_spares.sav_vdevs = NULL;

	if (nspares == 0)
	return;

	/*
	* Construct the array of vdevs, opening them to get status in the
	* process. For each spare, there is potentially two different vdev_t
	* structures associated with it: one in the list of spares (used only
	* for basic validation purposes) and one in the active vdev
	* configuration (if it's spared in). During this phase we open and
	* validate each vdev on the spare list. If the vdev also exists in the
	* active configuration, then we also mark this vdev as an active spare.
	*/
	spa->spa_spares.sav_vdevs = kmem_zalloc(nspares * sizeof (void *),
	KM_SLEEP);
	for (i = 0; i < spa->spa_spares.sav_count; i++) {
	VERIFY(spa_config_parse(spa, &vd, spares[i], NULL, 0,
	VDEV_ALLOC_SPARE) == 0);
	ASSERT(vd != NULL);

	spa->spa_spares.sav_vdevs[i] = vd;

	if ((tvd = spa_lookup_by_guid(spa, vd->vdev_guid,
	B_FALSE)) != NULL) {
	if (!tvd->vdev_isspare)
	spa_spare_add(tvd);

	/*
	* We only mark the spare active if we were successfully
	* able to load the vdev. Otherwise, importing a pool
	* with a bad active spare would result in strange
	* behavior, because multiple pool would think the spare
	* is actively in use.
	*
	* There is a vulnerability here to an equally bizarre
	* circumstance, where a dead active spare is later
	* brought back to life (onlined or otherwise). Given
	* the rarity of this scenario, and the extra complexity
	* it adds, we ignore the possibility.
	*/
	if (!vdev_is_dead(tvd))
	spa_spare_activate(tvd);
	}

	vd->vdev_top = vd;
	vd->vdev_aux = &spa->spa_spares;

	if (vdev_open(vd) != 0)
	continue;

	if (vdev_validate_aux(vd) == 0)
	spa_spare_add(vd);
	}

	/*
	* Recompute the stashed list of spares, with status information
	* this time.
	*/
	fnvlist_remove(spa->spa_spares.sav_config, ZPOOL_CONFIG_SPARES);

	spares = kmem_alloc(spa->spa_spares.sav_count * sizeof (void *),
	KM_SLEEP);
	for (i = 0; i < spa->spa_spares.sav_count; i++)
	spares[i] = vdev_config_generate(spa,
	spa->spa_spares.sav_vdevs[i], B_TRUE, VDEV_CONFIG_SPARE);
	fnvlist_add_nvlist_array(spa->spa_spares.sav_config,
	ZPOOL_CONFIG_SPARES, (const nvlist_t * const *)spares,
	spa->spa_spares.sav_count);
	for (i = 0; i < spa->spa_spares.sav_count; i++)
	nvlist_free(spares[i]);
	kmem_free(spares, spa->spa_spares.sav_count * sizeof (void *));
	}

	/*
	* Load (or re-load) the current list of vdevs describing the active l2cache for
	* this pool. When this is called, we have some form of basic information in
	* 'spa_l2cache.sav_config'. We parse this into vdevs, try to open them, and
	* then re-generate a more complete list including status information.
	* Devices which are already active have their details maintained, and are
	* not re-opened.
	*/
	void
	spa_load_l2cache(spa_t *spa)
	{
	nvlist_t **l2cache = NULL;
	uint_t nl2cache;
	int i, j, oldnvdevs;
	uint64_t guid;
	vdev_t vd, oldvdevs, *newvdevs;
	spa_aux_vdev_t *sav = &spa->spa_l2cache;

	#ifndef _KERNEL
	/*
	* zdb opens both the current state of the pool and the
	* checkpointed state (if present), with a different spa_t.
	*
	* As L2 caches are part of the ARC which is shared among open
	* pools, we skip loading them when we load the checkpointed
	* state of the pool.
	*/
	if (!spa_writeable(spa))
	return;
	#endif

	ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);

	oldvdevs = sav->sav_vdevs;
	oldnvdevs = sav->sav_count;
	sav->sav_vdevs = NULL;
	sav->sav_count = 0;

	if (sav->sav_config == NULL) {
	nl2cache = 0;
	newvdevs = NULL;
	goto out;
	}

	VERIFY0(nvlist_lookup_nvlist_array(sav->sav_config,
	ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache));
	newvdevs = kmem_alloc(nl2cache * sizeof (void *), KM_SLEEP);

	/*
	* Process new nvlist of vdevs.
	*/
	for (i = 0; i < nl2cache; i++) {
	guid = fnvlist_lookup_uint64(l2cache[i], ZPOOL_CONFIG_GUID);

	newvdevs[i] = NULL;
	for (j = 0; j < oldnvdevs; j++) {
	vd = oldvdevs[j];
	if (vd != NULL && guid == vd->vdev_guid) {
	/*
	* Retain previous vdev for add/remove ops.
	*/
	newvdevs[i] = vd;
	oldvdevs[j] = NULL;
	break;
	}
	}

	if (newvdevs[i] == NULL) {
	/*
	* Create new vdev
	*/
	VERIFY(spa_config_parse(spa, &vd, l2cache[i], NULL, 0,
	VDEV_ALLOC_L2CACHE) == 0);
	ASSERT(vd != NULL);
	newvdevs[i] = vd;

	/*
	* Commit this vdev as an l2cache device,
	* even if it fails to open.
	*/
	spa_l2cache_add(vd);

	vd->vdev_top = vd;
	vd->vdev_aux = sav;

	spa_l2cache_activate(vd);

	if (vdev_open(vd) != 0)
	continue;

	(void) vdev_validate_aux(vd);

	if (!vdev_is_dead(vd))
	l2arc_add_vdev(spa, vd);

	/*
	* Upon cache device addition to a pool or pool
	* creation with a cache device or if the header
	* of the device is invalid we issue an async
	* TRIM command for the whole device which will
	* execute if l2arc_trim_ahead > 0.
	*/
	spa_async_request(spa, SPA_ASYNC_L2CACHE_TRIM);
	}
	}

	sav->sav_vdevs = newvdevs;
	sav->sav_count = (int)nl2cache;

	/*
	* Recompute the stashed list of l2cache devices, with status
	* information this time.
	*/
	fnvlist_remove(sav->sav_config, ZPOOL_CONFIG_L2CACHE);

	if (sav->sav_count > 0)
	l2cache = kmem_alloc(sav->sav_count * sizeof (void *),
	KM_SLEEP);
	for (i = 0; i < sav->sav_count; i++)
	l2cache[i] = vdev_config_generate(spa,
	sav->sav_vdevs[i], B_TRUE, VDEV_CONFIG_L2CACHE);
	fnvlist_add_nvlist_array(sav->sav_config, ZPOOL_CONFIG_L2CACHE,
	(const nvlist_t * const *)l2cache, sav->sav_count);

	out:
	/*
	* Purge vdevs that were dropped
	*/
	if (oldvdevs) {
	for (i = 0; i < oldnvdevs; i++) {
	uint64_t pool;

	vd = oldvdevs[i];
	if (vd != NULL) {
	ASSERT(vd->vdev_isl2cache);

	if (spa_l2cache_exists(vd->vdev_guid, &pool) &&
	pool != 0ULL && l2arc_vdev_present(vd))
	l2arc_remove_vdev(vd);
	vdev_clear_stats(vd);
	vdev_free(vd);
	}
	}

	kmem_free(oldvdevs, oldnvdevs * sizeof (void *));
	}

	for (i = 0; i < sav->sav_count; i++)
	nvlist_free(l2cache[i]);
	if (sav->sav_count)
	kmem_free(l2cache, sav->sav_count * sizeof (void *));
	}

	static int
	load_nvlist(spa_t spa, uint64_t obj, nvlist_t *value)
	{
	dmu_buf_t *db;
	char *packed = NULL;
	size_t nvsize = 0;
	int error;
	*value = NULL;

	error = dmu_bonus_hold(spa->spa_meta_objset, obj, FTAG, &db);
	if (error)
	return (error);

	nvsize = (uint64_t )db->db_data;
	dmu_buf_rele(db, FTAG);

	packed = vmem_alloc(nvsize, KM_SLEEP);
	error = dmu_read(spa->spa_meta_objset, obj, 0, nvsize, packed,
	DMU_READ_PREFETCH);
	if (error == 0)
	error = nvlist_unpack(packed, nvsize, value, 0);
	vmem_free(packed, nvsize);

	return (error);
	}

	/*
	* Concrete top-level vdevs that are not missing and are not logs. At every
	* spa_sync we write new uberblocks to at least SPA_SYNC_MIN_VDEVS core tvds.
	*/
	static uint64_t
	spa_healthy_core_tvds(spa_t *spa)
	{
	vdev_t *rvd = spa->spa_root_vdev;
	uint64_t tvds = 0;

	for (uint64_t i = 0; i < rvd->vdev_children; i++) {
	vdev_t *vd = rvd->vdev_child[i];
	if (vd->vdev_islog)
	continue;
	if (vdev_is_concrete(vd) && !vdev_is_dead(vd))
	tvds++;
	}

	return (tvds);
	}

	/*
	* Checks to see if the given vdev could not be opened, in which case we post a
	* sysevent to notify the autoreplace code that the device has been removed.
	*/
	static void
	spa_check_removed(vdev_t *vd)
	{
	for (uint64_t c = 0; c < vd->vdev_children; c++)
	spa_check_removed(vd->vdev_child[c]);

	if (vd->vdev_ops->vdev_op_leaf && vdev_is_dead(vd) &&
	vdev_is_concrete(vd)) {
	zfs_post_autoreplace(vd->vdev_spa, vd);
	spa_event_notify(vd->vdev_spa, vd, NULL, ESC_ZFS_VDEV_CHECK);
	}
	}

	static int
	spa_check_for_missing_logs(spa_t *spa)
	{
	vdev_t *rvd = spa->spa_root_vdev;

	/*
	* If we're doing a normal import, then build up any additional
	* diagnostic information about missing log devices.
	* We'll pass this up to the user for further processing.
	*/
	if (!(spa->spa_import_flags & ZFS_IMPORT_MISSING_LOG)) {
	nvlist_t *child, nv;
	uint64_t idx = 0;

	child = kmem_alloc(rvd->vdev_children * sizeof (nvlist_t *),
	KM_SLEEP);
	nv = fnvlist_alloc();

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

	/*
	* We consider a device as missing only if it failed
	* to open (i.e. offline or faulted is not considered
	* as missing).
	*/
	if (tvd->vdev_islog &&
	tvd->vdev_state == VDEV_STATE_CANT_OPEN) {
	child[idx++] = vdev_config_generate(spa, tvd,
	B_FALSE, VDEV_CONFIG_MISSING);
	}
	}

	if (idx > 0) {
	fnvlist_add_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
	(const nvlist_t * const *)child, idx);
	fnvlist_add_nvlist(spa->spa_load_info,
	ZPOOL_CONFIG_MISSING_DEVICES, nv);

	for (uint64_t i = 0; i < idx; i++)
	nvlist_free(child[i]);
	}
	nvlist_free(nv);
	kmem_free(child, rvd->vdev_children * sizeof (char **));

	if (idx > 0) {
	spa_load_failed(spa, "some log devices are missing");
	vdev_dbgmsg_print_tree(rvd, 2);
	return (SET_ERROR(ENXIO));
	}
	} else {
	for (uint64_t c = 0; c < rvd->vdev_children; c++) {
	vdev_t *tvd = rvd->vdev_child[c];

	if (tvd->vdev_islog &&
	tvd->vdev_state == VDEV_STATE_CANT_OPEN) {
	spa_set_log_state(spa, SPA_LOG_CLEAR);
	spa_load_note(spa, "some log devices are "
	"missing, ZIL is dropped.");
	vdev_dbgmsg_print_tree(rvd, 2);
	break;
	}
	}
	}

	return (0);
	}

	/*
	* Check for missing log devices
	*/
	static boolean_t
	spa_check_logs(spa_t *spa)
	{
	boolean_t rv = B_FALSE;
	dsl_pool_t *dp = spa_get_dsl(spa);

	switch (spa->spa_log_state) {
	default:
	break;
	case SPA_LOG_MISSING:
	/* need to recheck in case slog has been restored */
	case SPA_LOG_UNKNOWN:
	rv = (dmu_objset_find_dp(dp, dp->dp_root_dir_obj,
	zil_check_log_chain, NULL, DS_FIND_CHILDREN) != 0);
	if (rv)
	spa_set_log_state(spa, SPA_LOG_MISSING);
	break;
	}
	return (rv);
	}

	/*
	* Passivate any log vdevs (note, does not apply to embedded log metaslabs).
	*/
	static boolean_t
	spa_passivate_log(spa_t *spa)
	{
	vdev_t *rvd = spa->spa_root_vdev;
	boolean_t slog_found = B_FALSE;

	ASSERT(spa_config_held(spa, SCL_ALLOC, RW_WRITER));

	for (int c = 0; c < rvd->vdev_children; c++) {
	vdev_t *tvd = rvd->vdev_child[c];

	if (tvd->vdev_islog) {
	ASSERT3P(tvd->vdev_log_mg, ==, NULL);
	metaslab_group_passivate(tvd->vdev_mg);
	slog_found = B_TRUE;
	}
	}

	return (slog_found);
	}

	/*
	* Activate any log vdevs (note, does not apply to embedded log metaslabs).
	*/
	static void
	spa_activate_log(spa_t *spa)
	{
	vdev_t *rvd = spa->spa_root_vdev;

	ASSERT(spa_config_held(spa, SCL_ALLOC, RW_WRITER));

	for (int c = 0; c < rvd->vdev_children; c++) {
	vdev_t *tvd = rvd->vdev_child[c];

	if (tvd->vdev_islog) {
	ASSERT3P(tvd->vdev_log_mg, ==, NULL);
	metaslab_group_activate(tvd->vdev_mg);
	}
	}
	}

	int
	spa_reset_logs(spa_t *spa)
	{
	int error;

	error = dmu_objset_find(spa_name(spa), zil_reset,
	NULL, DS_FIND_CHILDREN);
	if (error == 0) {
	/*
	* We successfully offlined the log device, sync out the
	* current txg so that the "stubby" block can be removed
	* by zil_sync().
	*/
	txg_wait_synced(spa->spa_dsl_pool, 0);
	}
	return (error);
	}

	static void
	spa_aux_check_removed(spa_aux_vdev_t *sav)
	{
	for (int i = 0; i < sav->sav_count; i++)
	spa_check_removed(sav->sav_vdevs[i]);
	}

	void
	spa_claim_notify(zio_t *zio)
	{
	spa_t *spa = zio->io_spa;

	if (zio->io_error)
	return;

	mutex_enter(&spa->spa_props_lock); /* any mutex will do */
	if (spa->spa_claim_max_txg < BP_GET_LOGICAL_BIRTH(zio->io_bp))
	spa->spa_claim_max_txg = BP_GET_LOGICAL_BIRTH(zio->io_bp);
	mutex_exit(&spa->spa_props_lock);
	}

	typedef struct spa_load_error {
	boolean_t sle_verify_data;
	uint64_t sle_meta_count;
	uint64_t sle_data_count;
	} spa_load_error_t;

	static void
	spa_load_verify_done(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;
	spa_load_error_t *sle = zio->io_private;
	dmu_object_type_t type = BP_GET_TYPE(bp);
	int error = zio->io_error;
	spa_t *spa = zio->io_spa;

	abd_free(zio->io_abd);
	if (error) {
	if ((BP_GET_LEVEL(bp) != 0 \|\| DMU_OT_IS_METADATA(type)) &&
	type != DMU_OT_INTENT_LOG)
	atomic_inc_64(&sle->sle_meta_count);
	else
	atomic_inc_64(&sle->sle_data_count);
	}

	mutex_enter(&spa->spa_scrub_lock);
	spa->spa_load_verify_bytes -= BP_GET_PSIZE(bp);
	cv_broadcast(&spa->spa_scrub_io_cv);
	mutex_exit(&spa->spa_scrub_lock);
	}

	/*
	* Maximum number of inflight bytes is the log2 fraction of the arc size.
	* By default, we set it to 1/16th of the arc.
	*/
	static uint_t spa_load_verify_shift = 4;
	static int spa_load_verify_metadata = B_TRUE;
	static int spa_load_verify_data = B_TRUE;

	static int
	spa_load_verify_cb(spa_t spa, zilog_t zilog, const blkptr_t *bp,
	const zbookmark_phys_t zb, const dnode_phys_t dnp, void *arg)
	{
	zio_t *rio = arg;
	spa_load_error_t *sle = rio->io_private;

	(void) zilog, (void) dnp;

	/*
	* Note: normally this routine will not be called if
	* spa_load_verify_metadata is not set. However, it may be useful
	* to manually set the flag after the traversal has begun.
	*/
	if (!spa_load_verify_metadata)
	return (0);

	/*
	* Sanity check the block pointer in order to detect obvious damage
	* before using the contents in subsequent checks or in zio_read().
	* When damaged consider it to be a metadata error since we cannot
	* trust the BP_GET_TYPE and BP_GET_LEVEL values.
	*/
	if (!zfs_blkptr_verify(spa, bp, BLK_CONFIG_NEEDED, BLK_VERIFY_LOG)) {
	atomic_inc_64(&sle->sle_meta_count);
	return (0);
	}

	if (zb->zb_level == ZB_DNODE_LEVEL \|\| BP_IS_HOLE(bp) \|\|
	BP_IS_EMBEDDED(bp) \|\| BP_IS_REDACTED(bp))
	return (0);

	if (!BP_IS_METADATA(bp) &&
	(!spa_load_verify_data \|\| !sle->sle_verify_data))
	return (0);

	uint64_t maxinflight_bytes =
	arc_target_bytes() >> spa_load_verify_shift;
	size_t size = BP_GET_PSIZE(bp);

	mutex_enter(&spa->spa_scrub_lock);
	while (spa->spa_load_verify_bytes >= maxinflight_bytes)
	cv_wait(&spa->spa_scrub_io_cv, &spa->spa_scrub_lock);
	spa->spa_load_verify_bytes += size;
	mutex_exit(&spa->spa_scrub_lock);

	zio_nowait(zio_read(rio, spa, bp, abd_alloc_for_io(size, B_FALSE), size,
	spa_load_verify_done, rio->io_private, ZIO_PRIORITY_SCRUB,
	ZIO_FLAG_SPECULATIVE \| ZIO_FLAG_CANFAIL \|
	ZIO_FLAG_SCRUB \| ZIO_FLAG_RAW, zb));
	return (0);
	}

	static int
	verify_dataset_name_len(dsl_pool_t dp, dsl_dataset_t ds, void *arg)
	{
	(void) dp, (void) arg;

	if (dsl_dataset_namelen(ds) >= ZFS_MAX_DATASET_NAME_LEN)
	return (SET_ERROR(ENAMETOOLONG));

	return (0);
	}

	static int
	spa_load_verify(spa_t *spa)
	{
	zio_t *rio;
	spa_load_error_t sle = { 0 };
	zpool_load_policy_t policy;
	boolean_t verify_ok = B_FALSE;
	int error = 0;

	zpool_get_load_policy(spa->spa_config, &policy);

	if (policy.zlp_rewind & ZPOOL_NEVER_REWIND \|\|
	policy.zlp_maxmeta == UINT64_MAX)
	return (0);

	dsl_pool_config_enter(spa->spa_dsl_pool, FTAG);
	error = dmu_objset_find_dp(spa->spa_dsl_pool,
	spa->spa_dsl_pool->dp_root_dir_obj, verify_dataset_name_len, NULL,
	DS_FIND_CHILDREN);
	dsl_pool_config_exit(spa->spa_dsl_pool, FTAG);
	if (error != 0)
	return (error);

	/*
	* Verify data only if we are rewinding or error limit was set.
	* Otherwise nothing except dbgmsg care about it to waste time.
	*/
	sle.sle_verify_data = (policy.zlp_rewind & ZPOOL_REWIND_MASK) \|\|
	(policy.zlp_maxdata < UINT64_MAX);

	rio = zio_root(spa, NULL, &sle,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_SPECULATIVE);

	if (spa_load_verify_metadata) {
	if (spa->spa_extreme_rewind) {
	spa_load_note(spa, "performing a complete scan of the "
	"pool since extreme rewind is on. This may take "
	"a very long time.\n (spa_load_verify_data=%u, "
	"spa_load_verify_metadata=%u)",
	spa_load_verify_data, spa_load_verify_metadata);
	}

	error = traverse_pool(spa, spa->spa_verify_min_txg,
	TRAVERSE_PRE \| TRAVERSE_PREFETCH_METADATA \|
	TRAVERSE_NO_DECRYPT, spa_load_verify_cb, rio);
	}

	(void) zio_wait(rio);
	ASSERT0(spa->spa_load_verify_bytes);

	spa->spa_load_meta_errors = sle.sle_meta_count;
	spa->spa_load_data_errors = sle.sle_data_count;

	if (sle.sle_meta_count != 0 \|\| sle.sle_data_count != 0) {
	spa_load_note(spa, "spa_load_verify found %llu metadata errors "
	"and %llu data errors", (u_longlong_t)sle.sle_meta_count,
	(u_longlong_t)sle.sle_data_count);
	}

	if (spa_load_verify_dryrun \|\|
	(!error && sle.sle_meta_count <= policy.zlp_maxmeta &&
	sle.sle_data_count <= policy.zlp_maxdata)) {
	int64_t loss = 0;

	verify_ok = B_TRUE;
	spa->spa_load_txg = spa->spa_uberblock.ub_txg;
	spa->spa_load_txg_ts = spa->spa_uberblock.ub_timestamp;

	loss = spa->spa_last_ubsync_txg_ts - spa->spa_load_txg_ts;
	fnvlist_add_uint64(spa->spa_load_info, ZPOOL_CONFIG_LOAD_TIME,
	spa->spa_load_txg_ts);
	fnvlist_add_int64(spa->spa_load_info, ZPOOL_CONFIG_REWIND_TIME,
	loss);
	fnvlist_add_uint64(spa->spa_load_info,
	ZPOOL_CONFIG_LOAD_META_ERRORS, sle.sle_meta_count);
	fnvlist_add_uint64(spa->spa_load_info,
	ZPOOL_CONFIG_LOAD_DATA_ERRORS, sle.sle_data_count);
	} else {
	spa->spa_load_max_txg = spa->spa_uberblock.ub_txg;
	}

	if (spa_load_verify_dryrun)
	return (0);

	if (error) {
	if (error != ENXIO && error != EIO)
	error = SET_ERROR(EIO);
	return (error);
	}

	return (verify_ok ? 0 : EIO);
	}

	/*
	* Find a value in the pool props object.
	*/
	static void
	spa_prop_find(spa_t spa, zpool_prop_t prop, uint64_t val)
	{
	(void) zap_lookup(spa->spa_meta_objset, spa->spa_pool_props_object,
	zpool_prop_to_name(prop), sizeof (uint64_t), 1, val);
	}

	/*
	* Find a value in the pool directory object.
	*/
	static int
	spa_dir_prop(spa_t spa, const char name, uint64_t *val, boolean_t log_enoent)
	{
	int error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	name, sizeof (uint64_t), 1, val);

	if (error != 0 && (error != ENOENT \|\| log_enoent)) {
	spa_load_failed(spa, "couldn't get '%s' value in MOS directory "
	"[error=%d]", name, error);
	}

	return (error);
	}

	static int
	spa_vdev_err(vdev_t *vdev, vdev_aux_t aux, int err)
	{
	vdev_set_state(vdev, B_TRUE, VDEV_STATE_CANT_OPEN, aux);
	return (SET_ERROR(err));
	}

	boolean_t
	spa_livelist_delete_check(spa_t *spa)
	{
	return (spa->spa_livelists_to_delete != 0);
	}

	static boolean_t
	spa_livelist_delete_cb_check(void arg, zthr_t z)
	{
	(void) z;
	spa_t *spa = arg;
	return (spa_livelist_delete_check(spa));
	}

	static int
	delete_blkptr_cb(void arg, const blkptr_t bp, dmu_tx_t *tx)
	{
	spa_t *spa = arg;
	zio_free(spa, tx->tx_txg, bp);
	dsl_dir_diduse_space(tx->tx_pool->dp_free_dir, DD_USED_HEAD,
	-bp_get_dsize_sync(spa, bp),
	-BP_GET_PSIZE(bp), -BP_GET_UCSIZE(bp), tx);
	return (0);
	}

	static int
	dsl_get_next_livelist_obj(objset_t os, uint64_t zap_obj, uint64_t llp)
	{
	int err;
	zap_cursor_t zc;
	zap_attribute_t za;
	zap_cursor_init(&zc, os, zap_obj);
	err = zap_cursor_retrieve(&zc, &za);
	zap_cursor_fini(&zc);
	if (err == 0)
	*llp = za.za_first_integer;
	return (err);
	}

	/*
	* Components of livelist deletion that must be performed in syncing
	* context: freeing block pointers and updating the pool-wide data
	* structures to indicate how much work is left to do
	*/
	typedef struct sublist_delete_arg {
	spa_t *spa;
	dsl_deadlist_t *ll;
	uint64_t key;
	bplist_t *to_free;
	} sublist_delete_arg_t;

	static void
	sublist_delete_sync(void arg, dmu_tx_t tx)
	{
	sublist_delete_arg_t *sda = arg;
	spa_t *spa = sda->spa;
	dsl_deadlist_t *ll = sda->ll;
	uint64_t key = sda->key;
	bplist_t *to_free = sda->to_free;

	bplist_iterate(to_free, delete_blkptr_cb, spa, tx);
	dsl_deadlist_remove_entry(ll, key, tx);
	}

	typedef struct livelist_delete_arg {
	spa_t *spa;
	uint64_t ll_obj;
	uint64_t zap_obj;
	} livelist_delete_arg_t;

	static void
	livelist_delete_sync(void arg, dmu_tx_t tx)
	{
	livelist_delete_arg_t *lda = arg;
	spa_t *spa = lda->spa;
	uint64_t ll_obj = lda->ll_obj;
	uint64_t zap_obj = lda->zap_obj;
	objset_t *mos = spa->spa_meta_objset;
	uint64_t count;

	/* free the livelist and decrement the feature count */
	VERIFY0(zap_remove_int(mos, zap_obj, ll_obj, tx));
	dsl_deadlist_free(mos, ll_obj, tx);
	spa_feature_decr(spa, SPA_FEATURE_LIVELIST, tx);
	VERIFY0(zap_count(mos, zap_obj, &count));
	if (count == 0) {
	/* no more livelists to delete */
	VERIFY0(zap_remove(mos, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_DELETED_CLONES, tx));
	VERIFY0(zap_destroy(mos, zap_obj, tx));
	spa->spa_livelists_to_delete = 0;
	spa_notify_waiters(spa);
	}
	}

	/*
	* Load in the value for the livelist to be removed and open it. Then,
	* load its first sublist and determine which block pointers should actually
	* be freed. Then, call a synctask which performs the actual frees and updates
	* the pool-wide livelist data.
	*/
	static void
	spa_livelist_delete_cb(void arg, zthr_t z)
	{
	spa_t *spa = arg;
	uint64_t ll_obj = 0, count;
	objset_t *mos = spa->spa_meta_objset;
	uint64_t zap_obj = spa->spa_livelists_to_delete;
	/*
	* Determine the next livelist to delete. This function should only
	* be called if there is at least one deleted clone.
	*/
	VERIFY0(dsl_get_next_livelist_obj(mos, zap_obj, &ll_obj));
	VERIFY0(zap_count(mos, ll_obj, &count));
	if (count > 0) {
	dsl_deadlist_t *ll;
	dsl_deadlist_entry_t *dle;
	bplist_t to_free;
	ll = kmem_zalloc(sizeof (dsl_deadlist_t), KM_SLEEP);
	dsl_deadlist_open(ll, mos, ll_obj);
	dle = dsl_deadlist_first(ll);
	ASSERT3P(dle, !=, NULL);
	bplist_create(&to_free);
	int err = dsl_process_sub_livelist(&dle->dle_bpobj, &to_free,
	z, NULL);
	if (err == 0) {
	sublist_delete_arg_t sync_arg = {
	.spa = spa,
	.ll = ll,
	.key = dle->dle_mintxg,
	.to_free = &to_free
	};
	zfs_dbgmsg("deleting sublist (id %llu) from"
	" livelist %llu, %lld remaining",
	(u_longlong_t)dle->dle_bpobj.bpo_object,
	(u_longlong_t)ll_obj, (longlong_t)count - 1);
	VERIFY0(dsl_sync_task(spa_name(spa), NULL,
	sublist_delete_sync, &sync_arg, 0,
	ZFS_SPACE_CHECK_DESTROY));
	} else {
	VERIFY3U(err, ==, EINTR);
	}
	bplist_clear(&to_free);
	bplist_destroy(&to_free);
	dsl_deadlist_close(ll);
	kmem_free(ll, sizeof (dsl_deadlist_t));
	} else {
	livelist_delete_arg_t sync_arg = {
	.spa = spa,
	.ll_obj = ll_obj,
	.zap_obj = zap_obj
	};
	zfs_dbgmsg("deletion of livelist %llu completed",
	(u_longlong_t)ll_obj);
	VERIFY0(dsl_sync_task(spa_name(spa), NULL, livelist_delete_sync,
	&sync_arg, 0, ZFS_SPACE_CHECK_DESTROY));
	}
	}

	static void
	spa_start_livelist_destroy_thread(spa_t *spa)
	{
	ASSERT3P(spa->spa_livelist_delete_zthr, ==, NULL);
	spa->spa_livelist_delete_zthr =
	zthr_create("z_livelist_destroy",
	spa_livelist_delete_cb_check, spa_livelist_delete_cb, spa,
	minclsyspri);
	}

	typedef struct livelist_new_arg {
	bplist_t *allocs;
	bplist_t *frees;
	} livelist_new_arg_t;

	static int
	livelist_track_new_cb(void arg, const blkptr_t bp, boolean_t bp_freed,
	dmu_tx_t *tx)
	{
	ASSERT(tx == NULL);
	livelist_new_arg_t *lna = arg;
	if (bp_freed) {
	bplist_append(lna->frees, bp);
	} else {
	bplist_append(lna->allocs, bp);
	zfs_livelist_condense_new_alloc++;
	}
	return (0);
	}

	typedef struct livelist_condense_arg {
	spa_t *spa;
	bplist_t to_keep;
	uint64_t first_size;
	uint64_t next_size;
	} livelist_condense_arg_t;

	static void
	spa_livelist_condense_sync(void arg, dmu_tx_t tx)
	{
	livelist_condense_arg_t *lca = arg;
	spa_t *spa = lca->spa;
	bplist_t new_frees;
	dsl_dataset_t *ds = spa->spa_to_condense.ds;

	/* Have we been cancelled? */
	if (spa->spa_to_condense.cancelled) {
	zfs_livelist_condense_sync_cancel++;
	goto out;
	}

	dsl_deadlist_entry_t *first = spa->spa_to_condense.first;
	dsl_deadlist_entry_t *next = spa->spa_to_condense.next;
	dsl_deadlist_t *ll = &ds->ds_dir->dd_livelist;

	/*
	* It's possible that the livelist was changed while the zthr was
	* running. Therefore, we need to check for new blkptrs in the two
	* entries being condensed and continue to track them in the livelist.
	* Because of the way we handle remapped blkptrs (see dbuf_remap_impl),
	* it's possible that the newly added blkptrs are FREEs or ALLOCs so
	* we need to sort them into two different bplists.
	*/
	uint64_t first_obj = first->dle_bpobj.bpo_object;
	uint64_t next_obj = next->dle_bpobj.bpo_object;
	uint64_t cur_first_size = first->dle_bpobj.bpo_phys->bpo_num_blkptrs;
	uint64_t cur_next_size = next->dle_bpobj.bpo_phys->bpo_num_blkptrs;

	bplist_create(&new_frees);
	livelist_new_arg_t new_bps = {
	.allocs = &lca->to_keep,
	.frees = &new_frees,
	};

	if (cur_first_size > lca->first_size) {
	VERIFY0(livelist_bpobj_iterate_from_nofree(&first->dle_bpobj,
	livelist_track_new_cb, &new_bps, lca->first_size));
	}
	if (cur_next_size > lca->next_size) {
	VERIFY0(livelist_bpobj_iterate_from_nofree(&next->dle_bpobj,
	livelist_track_new_cb, &new_bps, lca->next_size));
	}

	dsl_deadlist_clear_entry(first, ll, tx);
	ASSERT(bpobj_is_empty(&first->dle_bpobj));
	dsl_deadlist_remove_entry(ll, next->dle_mintxg, tx);

	bplist_iterate(&lca->to_keep, dsl_deadlist_insert_alloc_cb, ll, tx);
	bplist_iterate(&new_frees, dsl_deadlist_insert_free_cb, ll, tx);
	bplist_destroy(&new_frees);

	char dsname[ZFS_MAX_DATASET_NAME_LEN];
	dsl_dataset_name(ds, dsname);
	zfs_dbgmsg("txg %llu condensing livelist of %s (id %llu), bpobj %llu "
	"(%llu blkptrs) and bpobj %llu (%llu blkptrs) -> bpobj %llu "
	"(%llu blkptrs)", (u_longlong_t)tx->tx_txg, dsname,
	(u_longlong_t)ds->ds_object, (u_longlong_t)first_obj,
	(u_longlong_t)cur_first_size, (u_longlong_t)next_obj,
	(u_longlong_t)cur_next_size,
	(u_longlong_t)first->dle_bpobj.bpo_object,
	(u_longlong_t)first->dle_bpobj.bpo_phys->bpo_num_blkptrs);
	out:
	dmu_buf_rele(ds->ds_dbuf, spa);
	spa->spa_to_condense.ds = NULL;
	bplist_clear(&lca->to_keep);
	bplist_destroy(&lca->to_keep);
	kmem_free(lca, sizeof (livelist_condense_arg_t));
	spa->spa_to_condense.syncing = B_FALSE;
	}

	static void
	spa_livelist_condense_cb(void arg, zthr_t t)
	{
	while (zfs_livelist_condense_zthr_pause &&
	!(zthr_has_waiters(t) \|\| zthr_iscancelled(t)))
	delay(1);

	spa_t *spa = arg;
	dsl_deadlist_entry_t *first = spa->spa_to_condense.first;
	dsl_deadlist_entry_t *next = spa->spa_to_condense.next;
	uint64_t first_size, next_size;

	livelist_condense_arg_t *lca =
	kmem_alloc(sizeof (livelist_condense_arg_t), KM_SLEEP);
	bplist_create(&lca->to_keep);

	/*
	* Process the livelists (matching FREEs and ALLOCs) in open context
	* so we have minimal work in syncing context to condense.
	*
	* We save bpobj sizes (first_size and next_size) to use later in
	* syncing context to determine if entries were added to these sublists
	* while in open context. This is possible because the clone is still
	* active and open for normal writes and we want to make sure the new,
	* unprocessed blockpointers are inserted into the livelist normally.
	*
	* Note that dsl_process_sub_livelist() both stores the size number of
	* blockpointers and iterates over them while the bpobj's lock held, so
	* the sizes returned to us are consistent which what was actually
	* processed.
	*/
	int err = dsl_process_sub_livelist(&first->dle_bpobj, &lca->to_keep, t,
	&first_size);
	if (err == 0)
	err = dsl_process_sub_livelist(&next->dle_bpobj, &lca->to_keep,
	t, &next_size);

	if (err == 0) {
	while (zfs_livelist_condense_sync_pause &&
	!(zthr_has_waiters(t) \|\| zthr_iscancelled(t)))
	delay(1);

	dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
	dmu_tx_mark_netfree(tx);
	dmu_tx_hold_space(tx, 1);
	err = dmu_tx_assign(tx, TXG_NOWAIT \| TXG_NOTHROTTLE);
	if (err == 0) {
	/*
	* Prevent the condense zthr restarting before
	* the synctask completes.
	*/
	spa->spa_to_condense.syncing = B_TRUE;
	lca->spa = spa;
	lca->first_size = first_size;
	lca->next_size = next_size;
	dsl_sync_task_nowait(spa_get_dsl(spa),
	spa_livelist_condense_sync, lca, tx);
	dmu_tx_commit(tx);
	return;
	}
	}
	/*
	* Condensing can not continue: either it was externally stopped or
	* we were unable to assign to a tx because the pool has run out of
	* space. In the second case, we'll just end up trying to condense
	* again in a later txg.
	*/
	ASSERT(err != 0);
	bplist_clear(&lca->to_keep);
	bplist_destroy(&lca->to_keep);
	kmem_free(lca, sizeof (livelist_condense_arg_t));
	dmu_buf_rele(spa->spa_to_condense.ds->ds_dbuf, spa);
	spa->spa_to_condense.ds = NULL;
	if (err == EINTR)
	zfs_livelist_condense_zthr_cancel++;
	}

	/*
	* Check that there is something to condense but that a condense is not
	* already in progress and that condensing has not been cancelled.
	*/
	static boolean_t
	spa_livelist_condense_cb_check(void arg, zthr_t z)
	{
	(void) z;
	spa_t *spa = arg;
	if ((spa->spa_to_condense.ds != NULL) &&
	(spa->spa_to_condense.syncing == B_FALSE) &&
	(spa->spa_to_condense.cancelled == B_FALSE)) {
	return (B_TRUE);
	}
	return (B_FALSE);
	}

	static void
	spa_start_livelist_condensing_thread(spa_t *spa)
	{
	spa->spa_to_condense.ds = NULL;
	spa->spa_to_condense.first = NULL;
	spa->spa_to_condense.next = NULL;
	spa->spa_to_condense.syncing = B_FALSE;
	spa->spa_to_condense.cancelled = B_FALSE;

	ASSERT3P(spa->spa_livelist_condense_zthr, ==, NULL);
	spa->spa_livelist_condense_zthr =
	zthr_create("z_livelist_condense",
	spa_livelist_condense_cb_check,
	spa_livelist_condense_cb, spa, minclsyspri);
	}

	static void
	spa_spawn_aux_threads(spa_t *spa)
	{
	ASSERT(spa_writeable(spa));

	spa_start_raidz_expansion_thread(spa);
	spa_start_indirect_condensing_thread(spa);
	spa_start_livelist_destroy_thread(spa);
	spa_start_livelist_condensing_thread(spa);

	ASSERT3P(spa->spa_checkpoint_discard_zthr, ==, NULL);
	spa->spa_checkpoint_discard_zthr =
	zthr_create("z_checkpoint_discard",
	spa_checkpoint_discard_thread_check,
	spa_checkpoint_discard_thread, spa, minclsyspri);
	}

	/*
	* Fix up config after a partly-completed split. This is done with the
	* ZPOOL_CONFIG_SPLIT nvlist. Both the splitting pool and the split-off
	* pool have that entry in their config, but only the splitting one contains
	* a list of all the guids of the vdevs that are being split off.
	*
	* This function determines what to do with that list: either rejoin
	* all the disks to the pool, or complete the splitting process. To attempt
	* the rejoin, each disk that is offlined is marked online again, and
	* we do a reopen() call. If the vdev label for every disk that was
	* marked online indicates it was successfully split off (VDEV_AUX_SPLIT_POOL)
	* then we call vdev_split() on each disk, and complete the split.
	*
	* Otherwise we leave the config alone, with all the vdevs in place in
	* the original pool.
	*/
	static void
	spa_try_repair(spa_t spa, nvlist_t config)
	{
	uint_t extracted;
	uint64_t *glist;
	uint_t i, gcount;
	nvlist_t *nvl;
	vdev_t **vd;
	boolean_t attempt_reopen;

	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_SPLIT, &nvl) != 0)
	return;

	/* check that the config is complete */
	if (nvlist_lookup_uint64_array(nvl, ZPOOL_CONFIG_SPLIT_LIST,
	&glist, &gcount) != 0)
	return;

	vd = kmem_zalloc(gcount * sizeof (vdev_t *), KM_SLEEP);

	/* attempt to online all the vdevs & validate */
	attempt_reopen = B_TRUE;
	for (i = 0; i < gcount; i++) {
	if (glist[i] == 0) /* vdev is hole */
	continue;

	vd[i] = spa_lookup_by_guid(spa, glist[i], B_FALSE);
	if (vd[i] == NULL) {
	/*
	* Don't bother attempting to reopen the disks;
	* just do the split.
	*/
	attempt_reopen = B_FALSE;
	} else {
	/* attempt to re-online it */
	vd[i]->vdev_offline = B_FALSE;
	}
	}

	if (attempt_reopen) {
	vdev_reopen(spa->spa_root_vdev);

	/* check each device to see what state it's in */
	for (extracted = 0, i = 0; i < gcount; i++) {
	if (vd[i] != NULL &&
	vd[i]->vdev_stat.vs_aux != VDEV_AUX_SPLIT_POOL)
	break;
	++extracted;
	}
	}

	/*
	* If every disk has been moved to the new pool, or if we never
	* even attempted to look at them, then we split them off for
	* good.
	*/
	if (!attempt_reopen \|\| gcount == extracted) {
	for (i = 0; i < gcount; i++)
	if (vd[i] != NULL)
	vdev_split(vd[i]);
	vdev_reopen(spa->spa_root_vdev);
	}

	kmem_free(vd, gcount * sizeof (vdev_t *));
	}

	static int
	spa_load(spa_t *spa, spa_load_state_t state, spa_import_type_t type)
	{
	const char *ereport = FM_EREPORT_ZFS_POOL;
	int error;

	spa->spa_load_state = state;
	(void) spa_import_progress_set_state(spa_guid(spa),
	spa_load_state(spa));
	spa_import_progress_set_notes(spa, "spa_load()");

	gethrestime(&spa->spa_loaded_ts);
	error = spa_load_impl(spa, type, &ereport);

	/*
	* Don't count references from objsets that are already closed
	* and are making their way through the eviction process.
	*/
	spa_evicting_os_wait(spa);
	spa->spa_minref = zfs_refcount_count(&spa->spa_refcount);
	if (error) {
	if (error != EEXIST) {
	spa->spa_loaded_ts.tv_sec = 0;
	spa->spa_loaded_ts.tv_nsec = 0;
	}
	if (error != EBADF) {
	(void) zfs_ereport_post(ereport, spa,
	NULL, NULL, NULL, 0);
	}
	}
	spa->spa_load_state = error ? SPA_LOAD_ERROR : SPA_LOAD_NONE;
	spa->spa_ena = 0;

	(void) spa_import_progress_set_state(spa_guid(spa),
	spa_load_state(spa));

	return (error);
	}

	#ifdef ZFS_DEBUG
	/*
	* Count the number of per-vdev ZAPs associated with all of the vdevs in the
	* vdev tree rooted in the given vd, and ensure that each ZAP is present in the
	* spa's per-vdev ZAP list.
	*/
	static uint64_t
	vdev_count_verify_zaps(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;
	uint64_t total = 0;

	if (spa_feature_is_active(vd->vdev_spa, SPA_FEATURE_AVZ_V2) &&
	vd->vdev_root_zap != 0) {
	total++;
	ASSERT0(zap_lookup_int(spa->spa_meta_objset,
	spa->spa_all_vdev_zaps, vd->vdev_root_zap));
	}
	if (vd->vdev_top_zap != 0) {
	total++;
	ASSERT0(zap_lookup_int(spa->spa_meta_objset,
	spa->spa_all_vdev_zaps, vd->vdev_top_zap));
	}
	if (vd->vdev_leaf_zap != 0) {
	total++;
	ASSERT0(zap_lookup_int(spa->spa_meta_objset,
	spa->spa_all_vdev_zaps, vd->vdev_leaf_zap));
	}

	for (uint64_t i = 0; i < vd->vdev_children; i++) {
	total += vdev_count_verify_zaps(vd->vdev_child[i]);
	}

	return (total);
	}
	#else
	#define vdev_count_verify_zaps(vd) ((void) sizeof (vd), 0)
	#endif

	/*
	* Determine whether the activity check is required.
	*/
	static boolean_t
	spa_activity_check_required(spa_t spa, uberblock_t ub, nvlist_t *label,
	nvlist_t *config)
	{
	uint64_t state = 0;
	uint64_t hostid = 0;
	uint64_t tryconfig_txg = 0;
	uint64_t tryconfig_timestamp = 0;
	uint16_t tryconfig_mmp_seq = 0;
	nvlist_t *nvinfo;

	if (nvlist_exists(config, ZPOOL_CONFIG_LOAD_INFO)) {
	nvinfo = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO);
	(void) nvlist_lookup_uint64(nvinfo, ZPOOL_CONFIG_MMP_TXG,
	&tryconfig_txg);
	(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_TIMESTAMP,
	&tryconfig_timestamp);
	(void) nvlist_lookup_uint16(nvinfo, ZPOOL_CONFIG_MMP_SEQ,
	&tryconfig_mmp_seq);
	}

	(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE, &state);

	/*
	* Disable the MMP activity check - This is used by zdb which
	* is intended to be used on potentially active pools.
	*/
	if (spa->spa_import_flags & ZFS_IMPORT_SKIP_MMP)
	return (B_FALSE);

	/*
	* Skip the activity check when the MMP feature is disabled.
	*/
	if (ub->ub_mmp_magic == MMP_MAGIC && ub->ub_mmp_delay == 0)
	return (B_FALSE);

	/*
	* If the tryconfig_ values are nonzero, they are the results of an
	* earlier tryimport. If they all match the uberblock we just found,
	* then the pool has not changed and we return false so we do not test
	* a second time.
	*/
	if (tryconfig_txg && tryconfig_txg == ub->ub_txg &&
	tryconfig_timestamp && tryconfig_timestamp == ub->ub_timestamp &&
	tryconfig_mmp_seq && tryconfig_mmp_seq ==
	(MMP_SEQ_VALID(ub) ? MMP_SEQ(ub) : 0))
	return (B_FALSE);

	/*
	* Allow the activity check to be skipped when importing the pool
	* on the same host which last imported it. Since the hostid from
	* configuration may be stale use the one read from the label.
	*/
	if (nvlist_exists(label, ZPOOL_CONFIG_HOSTID))
	hostid = fnvlist_lookup_uint64(label, ZPOOL_CONFIG_HOSTID);

	if (hostid == spa_get_hostid(spa))
	return (B_FALSE);

	/*
	* Skip the activity test when the pool was cleanly exported.
	*/
	if (state != POOL_STATE_ACTIVE)
	return (B_FALSE);

	return (B_TRUE);
	}

	/*
	* Nanoseconds the activity check must watch for changes on-disk.
	*/
	static uint64_t
	spa_activity_check_duration(spa_t spa, uberblock_t ub)
	{
	uint64_t import_intervals = MAX(zfs_multihost_import_intervals, 1);
	uint64_t multihost_interval = MSEC2NSEC(
	MMP_INTERVAL_OK(zfs_multihost_interval));
	uint64_t import_delay = MAX(NANOSEC, import_intervals *
	multihost_interval);

	/*
	* Local tunables determine a minimum duration except for the case
	* where we know when the remote host will suspend the pool if MMP
	* writes do not land.
	*
	* See Big Theory comment at the top of mmp.c for the reasoning behind
	* these cases and times.
	*/

	ASSERT(MMP_IMPORT_SAFETY_FACTOR >= 100);

	if (MMP_INTERVAL_VALID(ub) && MMP_FAIL_INT_VALID(ub) &&
	MMP_FAIL_INT(ub) > 0) {

	/* MMP on remote host will suspend pool after failed writes */
	import_delay = MMP_FAIL_INT(ub) * MSEC2NSEC(MMP_INTERVAL(ub)) *
	MMP_IMPORT_SAFETY_FACTOR / 100;

	zfs_dbgmsg("fail_intvals>0 import_delay=%llu ub_mmp "
	"mmp_fails=%llu ub_mmp mmp_interval=%llu "
	"import_intervals=%llu", (u_longlong_t)import_delay,
	(u_longlong_t)MMP_FAIL_INT(ub),
	(u_longlong_t)MMP_INTERVAL(ub),
	(u_longlong_t)import_intervals);

	} else if (MMP_INTERVAL_VALID(ub) && MMP_FAIL_INT_VALID(ub) &&
	MMP_FAIL_INT(ub) == 0) {

	/* MMP on remote host will never suspend pool */
	import_delay = MAX(import_delay, (MSEC2NSEC(MMP_INTERVAL(ub)) +
	ub->ub_mmp_delay) * import_intervals);

	zfs_dbgmsg("fail_intvals=0 import_delay=%llu ub_mmp "
	"mmp_interval=%llu ub_mmp_delay=%llu "
	"import_intervals=%llu", (u_longlong_t)import_delay,
	(u_longlong_t)MMP_INTERVAL(ub),
	(u_longlong_t)ub->ub_mmp_delay,
	(u_longlong_t)import_intervals);

	} else if (MMP_VALID(ub)) {
	/*
	* zfs-0.7 compatibility case
	*/

	import_delay = MAX(import_delay, (multihost_interval +
	ub->ub_mmp_delay) * import_intervals);

	zfs_dbgmsg("import_delay=%llu ub_mmp_delay=%llu "
	"import_intervals=%llu leaves=%u",
	(u_longlong_t)import_delay,
	(u_longlong_t)ub->ub_mmp_delay,
	(u_longlong_t)import_intervals,
	vdev_count_leaves(spa));
	} else {
	/* Using local tunings is the only reasonable option */
	zfs_dbgmsg("pool last imported on non-MMP aware "
	"host using import_delay=%llu multihost_interval=%llu "
	"import_intervals=%llu", (u_longlong_t)import_delay,
	(u_longlong_t)multihost_interval,
	(u_longlong_t)import_intervals);
	}

	return (import_delay);
	}

	/*
	* Remote host activity check.
	*
	* error results:
	* 0 - no activity detected
	* EREMOTEIO - remote activity detected
	* EINTR - user canceled the operation
	*/
	static int
	spa_activity_check(spa_t spa, uberblock_t ub, nvlist_t *config,
	boolean_t importing)
	{
	uint64_t txg = ub->ub_txg;
	uint64_t timestamp = ub->ub_timestamp;
	uint64_t mmp_config = ub->ub_mmp_config;
	uint16_t mmp_seq = MMP_SEQ_VALID(ub) ? MMP_SEQ(ub) : 0;
	uint64_t import_delay;
	hrtime_t import_expire, now;
	nvlist_t *mmp_label = NULL;
	vdev_t *rvd = spa->spa_root_vdev;
	kcondvar_t cv;
	kmutex_t mtx;
	int error = 0;

	cv_init(&cv, NULL, CV_DEFAULT, NULL);
	mutex_init(&mtx, NULL, MUTEX_DEFAULT, NULL);
	mutex_enter(&mtx);

	/*
	* If ZPOOL_CONFIG_MMP_TXG is present an activity check was performed
	* during the earlier tryimport. If the txg recorded there is 0 then
	* the pool is known to be active on another host.
	*
	* Otherwise, the pool might be in use on another host. Check for
	* changes in the uberblocks on disk if necessary.
	*/
	if (nvlist_exists(config, ZPOOL_CONFIG_LOAD_INFO)) {
	nvlist_t *nvinfo = fnvlist_lookup_nvlist(config,
	ZPOOL_CONFIG_LOAD_INFO);

	if (nvlist_exists(nvinfo, ZPOOL_CONFIG_MMP_TXG) &&
	fnvlist_lookup_uint64(nvinfo, ZPOOL_CONFIG_MMP_TXG) == 0) {
	vdev_uberblock_load(rvd, ub, &mmp_label);
	error = SET_ERROR(EREMOTEIO);
	goto out;
	}
	}

	import_delay = spa_activity_check_duration(spa, ub);

	/* Add a small random factor in case of simultaneous imports (0-25%) */
	import_delay += import_delay * random_in_range(250) / 1000;

	import_expire = gethrtime() + import_delay;

	if (importing) {
	spa_import_progress_set_notes(spa, "Checking MMP activity, "
	"waiting %llu ms", (u_longlong_t)NSEC2MSEC(import_delay));
	}

	int iterations = 0;
	while ((now = gethrtime()) < import_expire) {
	if (importing && iterations++ % 30 == 0) {
	spa_import_progress_set_notes(spa, "Checking MMP "
	"activity, %llu ms remaining",
	(u_longlong_t)NSEC2MSEC(import_expire - now));
	}

	if (importing) {
	(void) spa_import_progress_set_mmp_check(spa_guid(spa),
	NSEC2SEC(import_expire - gethrtime()));
	}

	vdev_uberblock_load(rvd, ub, &mmp_label);

	if (txg != ub->ub_txg \|\| timestamp != ub->ub_timestamp \|\|
	mmp_seq != (MMP_SEQ_VALID(ub) ? MMP_SEQ(ub) : 0)) {
	zfs_dbgmsg("multihost activity detected "
	"txg %llu ub_txg %llu "
	"timestamp %llu ub_timestamp %llu "
	"mmp_config %#llx ub_mmp_config %#llx",
	(u_longlong_t)txg, (u_longlong_t)ub->ub_txg,
	(u_longlong_t)timestamp,
	(u_longlong_t)ub->ub_timestamp,
	(u_longlong_t)mmp_config,
	(u_longlong_t)ub->ub_mmp_config);

	error = SET_ERROR(EREMOTEIO);
	break;
	}

	if (mmp_label) {
	nvlist_free(mmp_label);
	mmp_label = NULL;
	}

	error = cv_timedwait_sig(&cv, &mtx, ddi_get_lbolt() + hz);
	if (error != -1) {
	error = SET_ERROR(EINTR);
	break;
	}
	error = 0;
	}

	out:
	mutex_exit(&mtx);
	mutex_destroy(&mtx);
	cv_destroy(&cv);

	/*
	* If the pool is determined to be active store the status in the
	* spa->spa_load_info nvlist. If the remote hostname or hostid are
	* available from configuration read from disk store them as well.
	* This allows 'zpool import' to generate a more useful message.
	*
	* ZPOOL_CONFIG_MMP_STATE - observed pool status (mandatory)
	* ZPOOL_CONFIG_MMP_HOSTNAME - hostname from the active pool
	* ZPOOL_CONFIG_MMP_HOSTID - hostid from the active pool
	*/
	if (error == EREMOTEIO) {
	const char *hostname = "<unknown>";
	uint64_t hostid = 0;

	if (mmp_label) {
	if (nvlist_exists(mmp_label, ZPOOL_CONFIG_HOSTNAME)) {
	hostname = fnvlist_lookup_string(mmp_label,
	ZPOOL_CONFIG_HOSTNAME);
	fnvlist_add_string(spa->spa_load_info,
	ZPOOL_CONFIG_MMP_HOSTNAME, hostname);
	}

	if (nvlist_exists(mmp_label, ZPOOL_CONFIG_HOSTID)) {
	hostid = fnvlist_lookup_uint64(mmp_label,
	ZPOOL_CONFIG_HOSTID);
	fnvlist_add_uint64(spa->spa_load_info,
	ZPOOL_CONFIG_MMP_HOSTID, hostid);
	}
	}

	fnvlist_add_uint64(spa->spa_load_info,
	ZPOOL_CONFIG_MMP_STATE, MMP_STATE_ACTIVE);
	fnvlist_add_uint64(spa->spa_load_info,
	ZPOOL_CONFIG_MMP_TXG, 0);

	error = spa_vdev_err(rvd, VDEV_AUX_ACTIVE, EREMOTEIO);
	}

	if (mmp_label)
	nvlist_free(mmp_label);

	return (error);
	}

	/*
	* Called from zfs_ioc_clear for a pool that was suspended
	* after failing mmp write checks.
	*/
	boolean_t
	spa_mmp_remote_host_activity(spa_t *spa)
	{
	ASSERT(spa_multihost(spa) && spa_suspended(spa));

	nvlist_t *best_label;
	uberblock_t best_ub;

	/*
	* Locate the best uberblock on disk
	*/
	vdev_uberblock_load(spa->spa_root_vdev, &best_ub, &best_label);
	if (best_label) {
	/*
	* confirm that the best hostid matches our hostid
	*/
	if (nvlist_exists(best_label, ZPOOL_CONFIG_HOSTID) &&
	spa_get_hostid(spa) !=
	fnvlist_lookup_uint64(best_label, ZPOOL_CONFIG_HOSTID)) {
	nvlist_free(best_label);
	return (B_TRUE);
	}
	nvlist_free(best_label);
	} else {
	return (B_TRUE);
	}

	if (!MMP_VALID(&best_ub) \|\|
	!MMP_FAIL_INT_VALID(&best_ub) \|\|
	MMP_FAIL_INT(&best_ub) == 0) {
	return (B_TRUE);
	}

	if (best_ub.ub_txg != spa->spa_uberblock.ub_txg \|\|
	best_ub.ub_timestamp != spa->spa_uberblock.ub_timestamp) {
	zfs_dbgmsg("txg mismatch detected during pool clear "
	"txg %llu ub_txg %llu timestamp %llu ub_timestamp %llu",
	(u_longlong_t)spa->spa_uberblock.ub_txg,
	(u_longlong_t)best_ub.ub_txg,
	(u_longlong_t)spa->spa_uberblock.ub_timestamp,
	(u_longlong_t)best_ub.ub_timestamp);
	return (B_TRUE);
	}

	/*
	* Perform an activity check looking for any remote writer
	*/
	return (spa_activity_check(spa, &spa->spa_uberblock, spa->spa_config,
	B_FALSE) != 0);
	}

	static int
	spa_verify_host(spa_t spa, nvlist_t mos_config)
	{
	uint64_t hostid;
	const char *hostname;
	uint64_t myhostid = 0;

	if (!spa_is_root(spa) && nvlist_lookup_uint64(mos_config,
	ZPOOL_CONFIG_HOSTID, &hostid) == 0) {
	hostname = fnvlist_lookup_string(mos_config,
	ZPOOL_CONFIG_HOSTNAME);

	myhostid = zone_get_hostid(NULL);

	if (hostid != 0 && myhostid != 0 && hostid != myhostid) {
	cmn_err(CE_WARN, "pool '%s' could not be "
	"loaded as it was last accessed by "
	"another system (host: %s hostid: 0x%llx). "
	"See: https://openzfs.github.io/openzfs-docs/msg/"
	"ZFS-8000-EY",
	spa_name(spa), hostname, (u_longlong_t)hostid);
	spa_load_failed(spa, "hostid verification failed: pool "
	"last accessed by host: %s (hostid: 0x%llx)",
	hostname, (u_longlong_t)hostid);
	return (SET_ERROR(EBADF));
	}
	}

	return (0);
	}

	static int
	spa_ld_parse_config(spa_t *spa, spa_import_type_t type)
	{
	int error = 0;
	nvlist_t nvtree, nvl, *config = spa->spa_config;
	int parse;
	vdev_t *rvd;
	uint64_t pool_guid;
	const char *comment;
	const char *compatibility;

	/*
	* Versioning wasn't explicitly added to the label until later, so if
	* it's not present treat it as the initial version.
	*/
	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
	&spa->spa_ubsync.ub_version) != 0)
	spa->spa_ubsync.ub_version = SPA_VERSION_INITIAL;

	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &pool_guid)) {
	spa_load_failed(spa, "invalid config provided: '%s' missing",
	ZPOOL_CONFIG_POOL_GUID);
	return (SET_ERROR(EINVAL));
	}

	/*
	* If we are doing an import, ensure that the pool is not already
	* imported by checking if its pool guid already exists in the
	* spa namespace.
	*
	* The only case that we allow an already imported pool to be
	* imported again, is when the pool is checkpointed and we want to
	* look at its checkpointed state from userland tools like zdb.
	*/
	#ifdef _KERNEL
	if ((spa->spa_load_state == SPA_LOAD_IMPORT \|\|
	spa->spa_load_state == SPA_LOAD_TRYIMPORT) &&
	spa_guid_exists(pool_guid, 0)) {
	#else
	if ((spa->spa_load_state == SPA_LOAD_IMPORT \|\|
	spa->spa_load_state == SPA_LOAD_TRYIMPORT) &&
	spa_guid_exists(pool_guid, 0) &&
	!spa_importing_readonly_checkpoint(spa)) {
	#endif
	spa_load_failed(spa, "a pool with guid %llu is already open",
	(u_longlong_t)pool_guid);
	return (SET_ERROR(EEXIST));
	}

	spa->spa_config_guid = pool_guid;

	nvlist_free(spa->spa_load_info);
	spa->spa_load_info = fnvlist_alloc();

	ASSERT(spa->spa_comment == NULL);
	if (nvlist_lookup_string(config, ZPOOL_CONFIG_COMMENT, &comment) == 0)
	spa->spa_comment = spa_strdup(comment);

	ASSERT(spa->spa_compatibility == NULL);
	if (nvlist_lookup_string(config, ZPOOL_CONFIG_COMPATIBILITY,
	&compatibility) == 0)
	spa->spa_compatibility = spa_strdup(compatibility);

	(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG,
	&spa->spa_config_txg);

	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_SPLIT, &nvl) == 0)
	spa->spa_config_splitting = fnvlist_dup(nvl);

	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvtree)) {
	spa_load_failed(spa, "invalid config provided: '%s' missing",
	ZPOOL_CONFIG_VDEV_TREE);
	return (SET_ERROR(EINVAL));
	}

	/*
	* Create "The Godfather" zio to hold all async IOs
	*/
	spa->spa_async_zio_root = kmem_alloc(max_ncpus * sizeof (void *),
	KM_SLEEP);
	for (int i = 0; i < max_ncpus; i++) {
	spa->spa_async_zio_root[i] = zio_root(spa, NULL, NULL,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_SPECULATIVE \|
	ZIO_FLAG_GODFATHER);
	}

	/*
	* Parse the configuration into a vdev tree. We explicitly set the
	* value that will be returned by spa_version() since parsing the
	* configuration requires knowing the version number.
	*/
	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	parse = (type == SPA_IMPORT_EXISTING ?
	VDEV_ALLOC_LOAD : VDEV_ALLOC_SPLIT);
	error = spa_config_parse(spa, &rvd, nvtree, NULL, 0, parse);
	spa_config_exit(spa, SCL_ALL, FTAG);

	if (error != 0) {
	spa_load_failed(spa, "unable to parse config [error=%d]",
	error);
	return (error);
	}

	ASSERT(spa->spa_root_vdev == rvd);
	ASSERT3U(spa->spa_min_ashift, >=, SPA_MINBLOCKSHIFT);
	ASSERT3U(spa->spa_max_ashift, <=, SPA_MAXBLOCKSHIFT);

	if (type != SPA_IMPORT_ASSEMBLE) {
	ASSERT(spa_guid(spa) == pool_guid);
	}

	return (0);
	}

	/*
	* Recursively open all vdevs in the vdev tree. This function is called twice:
	* first with the untrusted config, then with the trusted config.
	*/
	static int
	spa_ld_open_vdevs(spa_t *spa)
	{
	int error = 0;

	/*
	* spa_missing_tvds_allowed defines how many top-level vdevs can be
	* missing/unopenable for the root vdev to be still considered openable.
	*/
	if (spa->spa_trust_config) {
	spa->spa_missing_tvds_allowed = zfs_max_missing_tvds;
	} else if (spa->spa_config_source == SPA_CONFIG_SRC_CACHEFILE) {
	spa->spa_missing_tvds_allowed = zfs_max_missing_tvds_cachefile;
	} else if (spa->spa_config_source == SPA_CONFIG_SRC_SCAN) {
	spa->spa_missing_tvds_allowed = zfs_max_missing_tvds_scan;
	} else {
	spa->spa_missing_tvds_allowed = 0;
	}

	spa->spa_missing_tvds_allowed =
	MAX(zfs_max_missing_tvds, spa->spa_missing_tvds_allowed);

	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	error = vdev_open(spa->spa_root_vdev);
	spa_config_exit(spa, SCL_ALL, FTAG);

	if (spa->spa_missing_tvds != 0) {
	spa_load_note(spa, "vdev tree has %lld missing top-level "
	"vdevs.", (u_longlong_t)spa->spa_missing_tvds);
	if (spa->spa_trust_config && (spa->spa_mode & SPA_MODE_WRITE)) {
	/*
	* Although theoretically we could allow users to open
	* incomplete pools in RW mode, we'd need to add a lot
	* of extra logic (e.g. adjust pool space to account
	* for missing vdevs).
	* This limitation also prevents users from accidentally
	* opening the pool in RW mode during data recovery and
	* damaging it further.
	*/
	spa_load_note(spa, "pools with missing top-level "
	"vdevs can only be opened in read-only mode.");
	error = SET_ERROR(ENXIO);
	} else {
	spa_load_note(spa, "current settings allow for maximum "
	"%lld missing top-level vdevs at this stage.",
	(u_longlong_t)spa->spa_missing_tvds_allowed);
	}
	}
	if (error != 0) {
	spa_load_failed(spa, "unable to open vdev tree [error=%d]",
	error);
	}
	if (spa->spa_missing_tvds != 0 \|\| error != 0)
	vdev_dbgmsg_print_tree(spa->spa_root_vdev, 2);

	return (error);
	}

	/*
	* We need to validate the vdev labels against the configuration that
	* we have in hand. This function is called twice: first with an untrusted
	* config, then with a trusted config. The validation is more strict when the
	* config is trusted.
	*/
	static int
	spa_ld_validate_vdevs(spa_t *spa)
	{
	int error = 0;
	vdev_t *rvd = spa->spa_root_vdev;

	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	error = vdev_validate(rvd);
	spa_config_exit(spa, SCL_ALL, FTAG);

	if (error != 0) {
	spa_load_failed(spa, "vdev_validate failed [error=%d]", error);
	return (error);
	}

	if (rvd->vdev_state <= VDEV_STATE_CANT_OPEN) {
	spa_load_failed(spa, "cannot open vdev tree after invalidating "
	"some vdevs");
	vdev_dbgmsg_print_tree(rvd, 2);
	return (SET_ERROR(ENXIO));
	}

	return (0);
	}

	static void
	spa_ld_select_uberblock_done(spa_t spa, uberblock_t ub)
	{
	spa->spa_state = POOL_STATE_ACTIVE;
	spa->spa_ubsync = spa->spa_uberblock;
	spa->spa_verify_min_txg = spa->spa_extreme_rewind ?
	TXG_INITIAL - 1 : spa_last_synced_txg(spa) - TXG_DEFER_SIZE - 1;
	spa->spa_first_txg = spa->spa_last_ubsync_txg ?
	spa->spa_last_ubsync_txg : spa_last_synced_txg(spa) + 1;
	spa->spa_claim_max_txg = spa->spa_first_txg;
	spa->spa_prev_software_version = ub->ub_software_version;
	}

	static int
	spa_ld_select_uberblock(spa_t *spa, spa_import_type_t type)
	{
	vdev_t *rvd = spa->spa_root_vdev;
	nvlist_t *label;
	uberblock_t *ub = &spa->spa_uberblock;
	boolean_t activity_check = B_FALSE;

	/*
	* If we are opening the checkpointed state of the pool by
	* rewinding to it, at this point we will have written the
	* checkpointed uberblock to the vdev labels, so searching
	* the labels will find the right uberblock. However, if
	* we are opening the checkpointed state read-only, we have
	* not modified the labels. Therefore, we must ignore the
	* labels and continue using the spa_uberblock that was set
	* by spa_ld_checkpoint_rewind.
	*
	* Note that it would be fine to ignore the labels when
	* rewinding (opening writeable) as well. However, if we
	* crash just after writing the labels, we will end up
	* searching the labels. Doing so in the common case means
	* that this code path gets exercised normally, rather than
	* just in the edge case.
	*/
	if (ub->ub_checkpoint_txg != 0 &&
	spa_importing_readonly_checkpoint(spa)) {
	spa_ld_select_uberblock_done(spa, ub);
	return (0);
	}

	/*
	* Find the best uberblock.
	*/
	vdev_uberblock_load(rvd, ub, &label);

	/*
	* If we weren't able to find a single valid uberblock, return failure.
	*/
	if (ub->ub_txg == 0) {
	nvlist_free(label);
	spa_load_failed(spa, "no valid uberblock found");
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, ENXIO));
	}

	if (spa->spa_load_max_txg != UINT64_MAX) {
	(void) spa_import_progress_set_max_txg(spa_guid(spa),
	(u_longlong_t)spa->spa_load_max_txg);
	}
	spa_load_note(spa, "using uberblock with txg=%llu",
	(u_longlong_t)ub->ub_txg);
	if (ub->ub_raidz_reflow_info != 0) {
	spa_load_note(spa, "uberblock raidz_reflow_info: "
	"state=%u offset=%llu",
	(int)RRSS_GET_STATE(ub),
	(u_longlong_t)RRSS_GET_OFFSET(ub));
	}


	/*
	* For pools which have the multihost property on determine if the
	* pool is truly inactive and can be safely imported. Prevent
	* hosts which don't have a hostid set from importing the pool.
	*/
	activity_check = spa_activity_check_required(spa, ub, label,
	spa->spa_config);
	if (activity_check) {
	if (ub->ub_mmp_magic == MMP_MAGIC && ub->ub_mmp_delay &&
	spa_get_hostid(spa) == 0) {
	nvlist_free(label);
	fnvlist_add_uint64(spa->spa_load_info,
	ZPOOL_CONFIG_MMP_STATE, MMP_STATE_NO_HOSTID);
	return (spa_vdev_err(rvd, VDEV_AUX_ACTIVE, EREMOTEIO));
	}

	int error =
	spa_activity_check(spa, ub, spa->spa_config, B_TRUE);
	if (error) {
	nvlist_free(label);
	return (error);
	}

	fnvlist_add_uint64(spa->spa_load_info,
	ZPOOL_CONFIG_MMP_STATE, MMP_STATE_INACTIVE);
	fnvlist_add_uint64(spa->spa_load_info,
	ZPOOL_CONFIG_MMP_TXG, ub->ub_txg);
	fnvlist_add_uint16(spa->spa_load_info,
	ZPOOL_CONFIG_MMP_SEQ,
	(MMP_SEQ_VALID(ub) ? MMP_SEQ(ub) : 0));
	}

	/*
	* If the pool has an unsupported version we can't open it.
	*/
	if (!SPA_VERSION_IS_SUPPORTED(ub->ub_version)) {
	nvlist_free(label);
	spa_load_failed(spa, "version %llu is not supported",
	(u_longlong_t)ub->ub_version);
	return (spa_vdev_err(rvd, VDEV_AUX_VERSION_NEWER, ENOTSUP));
	}

	if (ub->ub_version >= SPA_VERSION_FEATURES) {
	nvlist_t *features;

	/*
	* If we weren't able to find what's necessary for reading the
	* MOS in the label, return failure.
	*/
	if (label == NULL) {
	spa_load_failed(spa, "label config unavailable");
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA,
	ENXIO));
	}

	if (nvlist_lookup_nvlist(label, ZPOOL_CONFIG_FEATURES_FOR_READ,
	&features) != 0) {
	nvlist_free(label);
	spa_load_failed(spa, "invalid label: '%s' missing",
	ZPOOL_CONFIG_FEATURES_FOR_READ);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA,
	ENXIO));
	}

	/*
	* Update our in-core representation with the definitive values
	* from the label.
	*/
	nvlist_free(spa->spa_label_features);
	spa->spa_label_features = fnvlist_dup(features);
	}

	nvlist_free(label);

	/*
	* Look through entries in the label nvlist's features_for_read. If
	* there is a feature listed there which we don't understand then we
	* cannot open a pool.
	*/
	if (ub->ub_version >= SPA_VERSION_FEATURES) {
	nvlist_t *unsup_feat;

	unsup_feat = fnvlist_alloc();

	for (nvpair_t *nvp = nvlist_next_nvpair(spa->spa_label_features,
	NULL); nvp != NULL;
	nvp = nvlist_next_nvpair(spa->spa_label_features, nvp)) {
	if (!zfeature_is_supported(nvpair_name(nvp))) {
	fnvlist_add_string(unsup_feat,
	nvpair_name(nvp), "");
	}
	}

	if (!nvlist_empty(unsup_feat)) {
	fnvlist_add_nvlist(spa->spa_load_info,
	ZPOOL_CONFIG_UNSUP_FEAT, unsup_feat);
	nvlist_free(unsup_feat);
	spa_load_failed(spa, "some features are unsupported");
	return (spa_vdev_err(rvd, VDEV_AUX_UNSUP_FEAT,
	ENOTSUP));
	}

	nvlist_free(unsup_feat);
	}

	if (type != SPA_IMPORT_ASSEMBLE && spa->spa_config_splitting) {
	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	spa_try_repair(spa, spa->spa_config);
	spa_config_exit(spa, SCL_ALL, FTAG);
	nvlist_free(spa->spa_config_splitting);
	spa->spa_config_splitting = NULL;
	}

	/*
	* Initialize internal SPA structures.
	*/
	spa_ld_select_uberblock_done(spa, ub);

	return (0);
	}

	static int
	spa_ld_open_rootbp(spa_t *spa)
	{
	int error = 0;
	vdev_t *rvd = spa->spa_root_vdev;

	error = dsl_pool_init(spa, spa->spa_first_txg, &spa->spa_dsl_pool);
	if (error != 0) {
	spa_load_failed(spa, "unable to open rootbp in dsl_pool_init "
	"[error=%d]", error);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}
	spa->spa_meta_objset = spa->spa_dsl_pool->dp_meta_objset;

	return (0);
	}

	static int
	spa_ld_trusted_config(spa_t *spa, spa_import_type_t type,
	boolean_t reloading)
	{
	vdev_t mrvd, rvd = spa->spa_root_vdev;
	nvlist_t nv, mos_config, *policy;
	int error = 0, copy_error;
	uint64_t healthy_tvds, healthy_tvds_mos;
	uint64_t mos_config_txg;

	if (spa_dir_prop(spa, DMU_POOL_CONFIG, &spa->spa_config_object, B_TRUE)
	!= 0)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));

	/*
	* If we're assembling a pool from a split, the config provided is
	* already trusted so there is nothing to do.
	*/
	if (type == SPA_IMPORT_ASSEMBLE)
	return (0);

	healthy_tvds = spa_healthy_core_tvds(spa);

	if (load_nvlist(spa, spa->spa_config_object, &mos_config)
	!= 0) {
	spa_load_failed(spa, "unable to retrieve MOS config");
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	/*
	* If we are doing an open, pool owner wasn't verified yet, thus do
	* the verification here.
	*/
	if (spa->spa_load_state == SPA_LOAD_OPEN) {
	error = spa_verify_host(spa, mos_config);
	if (error != 0) {
	nvlist_free(mos_config);
	return (error);
	}
	}

	nv = fnvlist_lookup_nvlist(mos_config, ZPOOL_CONFIG_VDEV_TREE);

	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);

	/*
	* Build a new vdev tree from the trusted config
	*/
	error = spa_config_parse(spa, &mrvd, nv, NULL, 0, VDEV_ALLOC_LOAD);
	if (error != 0) {
	nvlist_free(mos_config);
	spa_config_exit(spa, SCL_ALL, FTAG);
	spa_load_failed(spa, "spa_config_parse failed [error=%d]",
	error);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, error));
	}

	/*
	* Vdev paths in the MOS may be obsolete. If the untrusted config was
	* obtained by scanning /dev/dsk, then it will have the right vdev
	* paths. We update the trusted MOS config with this information.
	* We first try to copy the paths with vdev_copy_path_strict, which
	* succeeds only when both configs have exactly the same vdev tree.
	* If that fails, we fall back to a more flexible method that has a
	* best effort policy.
	*/
	copy_error = vdev_copy_path_strict(rvd, mrvd);
	if (copy_error != 0 \|\| spa_load_print_vdev_tree) {
	spa_load_note(spa, "provided vdev tree:");
	vdev_dbgmsg_print_tree(rvd, 2);
	spa_load_note(spa, "MOS vdev tree:");
	vdev_dbgmsg_print_tree(mrvd, 2);
	}
	if (copy_error != 0) {
	spa_load_note(spa, "vdev_copy_path_strict failed, falling "
	"back to vdev_copy_path_relaxed");
	vdev_copy_path_relaxed(rvd, mrvd);
	}

	vdev_close(rvd);
	vdev_free(rvd);
	spa->spa_root_vdev = mrvd;
	rvd = mrvd;
	spa_config_exit(spa, SCL_ALL, FTAG);

	/*
	* If 'zpool import' used a cached config, then the on-disk hostid and
	* hostname may be different to the cached config in ways that should
	* prevent import. Userspace can't discover this without a scan, but
	* we know, so we add these values to LOAD_INFO so the caller can know
	* the difference.
	*
	* Note that we have to do this before the config is regenerated,
	* because the new config will have the hostid and hostname for this
	* host, in readiness for import.
	*/
	if (nvlist_exists(mos_config, ZPOOL_CONFIG_HOSTID))
	fnvlist_add_uint64(spa->spa_load_info, ZPOOL_CONFIG_HOSTID,
	fnvlist_lookup_uint64(mos_config, ZPOOL_CONFIG_HOSTID));
	if (nvlist_exists(mos_config, ZPOOL_CONFIG_HOSTNAME))
	fnvlist_add_string(spa->spa_load_info, ZPOOL_CONFIG_HOSTNAME,
	fnvlist_lookup_string(mos_config, ZPOOL_CONFIG_HOSTNAME));

	/*
	* We will use spa_config if we decide to reload the spa or if spa_load
	* fails and we rewind. We must thus regenerate the config using the
	* MOS information with the updated paths. ZPOOL_LOAD_POLICY is used to
	* pass settings on how to load the pool and is not stored in the MOS.
	* We copy it over to our new, trusted config.
	*/
	mos_config_txg = fnvlist_lookup_uint64(mos_config,
	ZPOOL_CONFIG_POOL_TXG);
	nvlist_free(mos_config);
	mos_config = spa_config_generate(spa, NULL, mos_config_txg, B_FALSE);
	if (nvlist_lookup_nvlist(spa->spa_config, ZPOOL_LOAD_POLICY,
	&policy) == 0)
	fnvlist_add_nvlist(mos_config, ZPOOL_LOAD_POLICY, policy);
	spa_config_set(spa, mos_config);
	spa->spa_config_source = SPA_CONFIG_SRC_MOS;

	/*
	* Now that we got the config from the MOS, we should be more strict
	* in checking blkptrs and can make assumptions about the consistency
	* of the vdev tree. spa_trust_config must be set to true before opening
	* vdevs in order for them to be writeable.
	*/
	spa->spa_trust_config = B_TRUE;

	/*
	* Open and validate the new vdev tree
	*/
	error = spa_ld_open_vdevs(spa);
	if (error != 0)
	return (error);

	error = spa_ld_validate_vdevs(spa);
	if (error != 0)
	return (error);

	if (copy_error != 0 \|\| spa_load_print_vdev_tree) {
	spa_load_note(spa, "final vdev tree:");
	vdev_dbgmsg_print_tree(rvd, 2);
	}

	if (spa->spa_load_state != SPA_LOAD_TRYIMPORT &&
	!spa->spa_extreme_rewind && zfs_max_missing_tvds == 0) {
	/*
	* Sanity check to make sure that we are indeed loading the
	* latest uberblock. If we missed SPA_SYNC_MIN_VDEVS tvds
	* in the config provided and they happened to be the only ones
	* to have the latest uberblock, we could involuntarily perform
	* an extreme rewind.
	*/
	healthy_tvds_mos = spa_healthy_core_tvds(spa);
	if (healthy_tvds_mos - healthy_tvds >=
	SPA_SYNC_MIN_VDEVS) {
	spa_load_note(spa, "config provided misses too many "
	"top-level vdevs compared to MOS (%lld vs %lld). ",
	(u_longlong_t)healthy_tvds,
	(u_longlong_t)healthy_tvds_mos);
	spa_load_note(spa, "vdev tree:");
	vdev_dbgmsg_print_tree(rvd, 2);
	if (reloading) {
	spa_load_failed(spa, "config was already "
	"provided from MOS. Aborting.");
	return (spa_vdev_err(rvd,
	VDEV_AUX_CORRUPT_DATA, EIO));
	}
	spa_load_note(spa, "spa must be reloaded using MOS "
	"config");
	return (SET_ERROR(EAGAIN));
	}
	}

	error = spa_check_for_missing_logs(spa);
	if (error != 0)
	return (spa_vdev_err(rvd, VDEV_AUX_BAD_GUID_SUM, ENXIO));

	if (rvd->vdev_guid_sum != spa->spa_uberblock.ub_guid_sum) {
	spa_load_failed(spa, "uberblock guid sum doesn't match MOS "
	"guid sum (%llu != %llu)",
	(u_longlong_t)spa->spa_uberblock.ub_guid_sum,
	(u_longlong_t)rvd->vdev_guid_sum);
	return (spa_vdev_err(rvd, VDEV_AUX_BAD_GUID_SUM,
	ENXIO));
	}

	return (0);
	}

	static int
	spa_ld_open_indirect_vdev_metadata(spa_t *spa)
	{
	int error = 0;
	vdev_t *rvd = spa->spa_root_vdev;

	/*
	* Everything that we read before spa_remove_init() must be stored
	* on concreted vdevs. Therefore we do this as early as possible.
	*/
	error = spa_remove_init(spa);
	if (error != 0) {
	spa_load_failed(spa, "spa_remove_init failed [error=%d]",
	error);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	/*
	* Retrieve information needed to condense indirect vdev mappings.
	*/
	error = spa_condense_init(spa);
	if (error != 0) {
	spa_load_failed(spa, "spa_condense_init failed [error=%d]",
	error);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, error));
	}

	return (0);
	}

	static int
	spa_ld_check_features(spa_t spa, boolean_t missing_feat_writep)
	{
	int error = 0;
	vdev_t *rvd = spa->spa_root_vdev;

	if (spa_version(spa) >= SPA_VERSION_FEATURES) {
	boolean_t missing_feat_read = B_FALSE;
	nvlist_t unsup_feat, enabled_feat;

	if (spa_dir_prop(spa, DMU_POOL_FEATURES_FOR_READ,
	&spa->spa_feat_for_read_obj, B_TRUE) != 0) {
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	if (spa_dir_prop(spa, DMU_POOL_FEATURES_FOR_WRITE,
	&spa->spa_feat_for_write_obj, B_TRUE) != 0) {
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	if (spa_dir_prop(spa, DMU_POOL_FEATURE_DESCRIPTIONS,
	&spa->spa_feat_desc_obj, B_TRUE) != 0) {
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	enabled_feat = fnvlist_alloc();
	unsup_feat = fnvlist_alloc();

	if (!spa_features_check(spa, B_FALSE,
	unsup_feat, enabled_feat))
	missing_feat_read = B_TRUE;

	if (spa_writeable(spa) \|\|
	spa->spa_load_state == SPA_LOAD_TRYIMPORT) {
	if (!spa_features_check(spa, B_TRUE,
	unsup_feat, enabled_feat)) {
	*missing_feat_writep = B_TRUE;
	}
	}

	fnvlist_add_nvlist(spa->spa_load_info,
	ZPOOL_CONFIG_ENABLED_FEAT, enabled_feat);

	if (!nvlist_empty(unsup_feat)) {
	fnvlist_add_nvlist(spa->spa_load_info,
	ZPOOL_CONFIG_UNSUP_FEAT, unsup_feat);
	}

	fnvlist_free(enabled_feat);
	fnvlist_free(unsup_feat);

	if (!missing_feat_read) {
	fnvlist_add_boolean(spa->spa_load_info,
	ZPOOL_CONFIG_CAN_RDONLY);
	}

	/*
	* If the state is SPA_LOAD_TRYIMPORT, our objective is
	* twofold: to determine whether the pool is available for
	* import in read-write mode and (if it is not) whether the
	* pool is available for import in read-only mode. If the pool
	* is available for import in read-write mode, it is displayed
	* as available in userland; if it is not available for import
	* in read-only mode, it is displayed as unavailable in
	* userland. If the pool is available for import in read-only
	* mode but not read-write mode, it is displayed as unavailable
	* in userland with a special note that the pool is actually
	* available for open in read-only mode.
	*
	* As a result, if the state is SPA_LOAD_TRYIMPORT and we are
	* missing a feature for write, we must first determine whether
	* the pool can be opened read-only before returning to
	* userland in order to know whether to display the
	* abovementioned note.
	*/
	if (missing_feat_read \|\| (*missing_feat_writep &&
	spa_writeable(spa))) {
	spa_load_failed(spa, "pool uses unsupported features");
	return (spa_vdev_err(rvd, VDEV_AUX_UNSUP_FEAT,
	ENOTSUP));
	}

	/*
	* Load refcounts for ZFS features from disk into an in-memory
	* cache during SPA initialization.
	*/
	for (spa_feature_t i = 0; i < SPA_FEATURES; i++) {
	uint64_t refcount;

	error = feature_get_refcount_from_disk(spa,
	&spa_feature_table[i], &refcount);
	if (error == 0) {
	spa->spa_feat_refcount_cache[i] = refcount;
	} else if (error == ENOTSUP) {
	spa->spa_feat_refcount_cache[i] =
	SPA_FEATURE_DISABLED;
	} else {
	spa_load_failed(spa, "error getting refcount "
	"for feature %s [error=%d]",
	spa_feature_table[i].fi_guid, error);
	return (spa_vdev_err(rvd,
	VDEV_AUX_CORRUPT_DATA, EIO));
	}
	}
	}

	if (spa_feature_is_active(spa, SPA_FEATURE_ENABLED_TXG)) {
	if (spa_dir_prop(spa, DMU_POOL_FEATURE_ENABLED_TXG,
	&spa->spa_feat_enabled_txg_obj, B_TRUE) != 0)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	/*
	* Encryption was added before bookmark_v2, even though bookmark_v2
	* is now a dependency. If this pool has encryption enabled without
	* bookmark_v2, trigger an errata message.
	*/
	if (spa_feature_is_enabled(spa, SPA_FEATURE_ENCRYPTION) &&
	!spa_feature_is_enabled(spa, SPA_FEATURE_BOOKMARK_V2)) {
	spa->spa_errata = ZPOOL_ERRATA_ZOL_8308_ENCRYPTION;
	}

	return (0);
	}

	static int
	spa_ld_load_special_directories(spa_t *spa)
	{
	int error = 0;
	vdev_t *rvd = spa->spa_root_vdev;

	spa->spa_is_initializing = B_TRUE;
	error = dsl_pool_open(spa->spa_dsl_pool);
	spa->spa_is_initializing = B_FALSE;
	if (error != 0) {
	spa_load_failed(spa, "dsl_pool_open failed [error=%d]", error);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	return (0);
	}

	static int
	spa_ld_get_props(spa_t *spa)
	{
	int error = 0;
	uint64_t obj;
	vdev_t *rvd = spa->spa_root_vdev;

	/* Grab the checksum salt from the MOS. */
	error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_CHECKSUM_SALT, 1,
	sizeof (spa->spa_cksum_salt.zcs_bytes),
	spa->spa_cksum_salt.zcs_bytes);
	if (error == ENOENT) {
	/* Generate a new salt for subsequent use */
	(void) random_get_pseudo_bytes(spa->spa_cksum_salt.zcs_bytes,
	sizeof (spa->spa_cksum_salt.zcs_bytes));
	} else if (error != 0) {
	spa_load_failed(spa, "unable to retrieve checksum salt from "
	"MOS [error=%d]", error);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	if (spa_dir_prop(spa, DMU_POOL_SYNC_BPOBJ, &obj, B_TRUE) != 0)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	error = bpobj_open(&spa->spa_deferred_bpobj, spa->spa_meta_objset, obj);
	if (error != 0) {
	spa_load_failed(spa, "error opening deferred-frees bpobj "
	"[error=%d]", error);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	/*
	* Load the bit that tells us to use the new accounting function
	* (raid-z deflation). If we have an older pool, this will not
	* be present.
	*/
	error = spa_dir_prop(spa, DMU_POOL_DEFLATE, &spa->spa_deflate, B_FALSE);
	if (error != 0 && error != ENOENT)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));

	error = spa_dir_prop(spa, DMU_POOL_CREATION_VERSION,
	&spa->spa_creation_version, B_FALSE);
	if (error != 0 && error != ENOENT)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));

	/*
	* Load the persistent error log. If we have an older pool, this will
	* not be present.
	*/
	error = spa_dir_prop(spa, DMU_POOL_ERRLOG_LAST, &spa->spa_errlog_last,
	B_FALSE);
	if (error != 0 && error != ENOENT)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));

	error = spa_dir_prop(spa, DMU_POOL_ERRLOG_SCRUB,
	&spa->spa_errlog_scrub, B_FALSE);
	if (error != 0 && error != ENOENT)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));

	/*
	* Load the livelist deletion field. If a livelist is queued for
	* deletion, indicate that in the spa
	*/
	error = spa_dir_prop(spa, DMU_POOL_DELETED_CLONES,
	&spa->spa_livelists_to_delete, B_FALSE);
	if (error != 0 && error != ENOENT)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));

	/*
	* Load the history object. If we have an older pool, this
	* will not be present.
	*/
	error = spa_dir_prop(spa, DMU_POOL_HISTORY, &spa->spa_history, B_FALSE);
	if (error != 0 && error != ENOENT)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));

	/*
	* Load the per-vdev ZAP map. If we have an older pool, this will not
	* be present; in this case, defer its creation to a later time to
	* avoid dirtying the MOS this early / out of sync context. See
	* spa_sync_config_object.
	*/

	/* The sentinel is only available in the MOS config. */
	nvlist_t *mos_config;
	if (load_nvlist(spa, spa->spa_config_object, &mos_config) != 0) {
	spa_load_failed(spa, "unable to retrieve MOS config");
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	error = spa_dir_prop(spa, DMU_POOL_VDEV_ZAP_MAP,
	&spa->spa_all_vdev_zaps, B_FALSE);

	if (error == ENOENT) {
	VERIFY(!nvlist_exists(mos_config,
	ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS));
	spa->spa_avz_action = AVZ_ACTION_INITIALIZE;
	ASSERT0(vdev_count_verify_zaps(spa->spa_root_vdev));
	} else if (error != 0) {
	nvlist_free(mos_config);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	} else if (!nvlist_exists(mos_config, ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS)) {
	/*
	* An older version of ZFS overwrote the sentinel value, so
	* we have orphaned per-vdev ZAPs in the MOS. Defer their
	* destruction to later; see spa_sync_config_object.
	*/
	spa->spa_avz_action = AVZ_ACTION_DESTROY;
	/*
	* We're assuming that no vdevs have had their ZAPs created
	* before this. Better be sure of it.
	*/
	ASSERT0(vdev_count_verify_zaps(spa->spa_root_vdev));
	}
	nvlist_free(mos_config);

	spa->spa_delegation = zpool_prop_default_numeric(ZPOOL_PROP_DELEGATION);

	error = spa_dir_prop(spa, DMU_POOL_PROPS, &spa->spa_pool_props_object,
	B_FALSE);
	if (error && error != ENOENT)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));

	if (error == 0) {
	uint64_t autoreplace = 0;

	spa_prop_find(spa, ZPOOL_PROP_BOOTFS, &spa->spa_bootfs);
	spa_prop_find(spa, ZPOOL_PROP_AUTOREPLACE, &autoreplace);
	spa_prop_find(spa, ZPOOL_PROP_DELEGATION, &spa->spa_delegation);
	spa_prop_find(spa, ZPOOL_PROP_FAILUREMODE, &spa->spa_failmode);
	spa_prop_find(spa, ZPOOL_PROP_AUTOEXPAND, &spa->spa_autoexpand);
	spa_prop_find(spa, ZPOOL_PROP_MULTIHOST, &spa->spa_multihost);
	spa_prop_find(spa, ZPOOL_PROP_AUTOTRIM, &spa->spa_autotrim);
	spa->spa_autoreplace = (autoreplace != 0);
	}

	/*
	* If we are importing a pool with missing top-level vdevs,
	* we enforce that the pool doesn't panic or get suspended on
	* error since the likelihood of missing data is extremely high.
	*/
	if (spa->spa_missing_tvds > 0 &&
	spa->spa_failmode != ZIO_FAILURE_MODE_CONTINUE &&
	spa->spa_load_state != SPA_LOAD_TRYIMPORT) {
	spa_load_note(spa, "forcing failmode to 'continue' "
	"as some top level vdevs are missing");
	spa->spa_failmode = ZIO_FAILURE_MODE_CONTINUE;
	}

	return (0);
	}

	static int
	spa_ld_open_aux_vdevs(spa_t *spa, spa_import_type_t type)
	{
	int error = 0;
	vdev_t *rvd = spa->spa_root_vdev;

	/*
	* If we're assembling the pool from the split-off vdevs of
	* an existing pool, we don't want to attach the spares & cache
	* devices.
	*/

	/*
	* Load any hot spares for this pool.
	*/
	error = spa_dir_prop(spa, DMU_POOL_SPARES, &spa->spa_spares.sav_object,
	B_FALSE);
	if (error != 0 && error != ENOENT)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	if (error == 0 && type != SPA_IMPORT_ASSEMBLE) {
	ASSERT(spa_version(spa) >= SPA_VERSION_SPARES);
	if (load_nvlist(spa, spa->spa_spares.sav_object,
	&spa->spa_spares.sav_config) != 0) {
	spa_load_failed(spa, "error loading spares nvlist");
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	spa_load_spares(spa);
	spa_config_exit(spa, SCL_ALL, FTAG);
	} else if (error == 0) {
	spa->spa_spares.sav_sync = B_TRUE;
	}

	/*
	* Load any level 2 ARC devices for this pool.
	*/
	error = spa_dir_prop(spa, DMU_POOL_L2CACHE,
	&spa->spa_l2cache.sav_object, B_FALSE);
	if (error != 0 && error != ENOENT)
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	if (error == 0 && type != SPA_IMPORT_ASSEMBLE) {
	ASSERT(spa_version(spa) >= SPA_VERSION_L2CACHE);
	if (load_nvlist(spa, spa->spa_l2cache.sav_object,
	&spa->spa_l2cache.sav_config) != 0) {
	spa_load_failed(spa, "error loading l2cache nvlist");
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	spa_load_l2cache(spa);
	spa_config_exit(spa, SCL_ALL, FTAG);
	} else if (error == 0) {
	spa->spa_l2cache.sav_sync = B_TRUE;
	}

	return (0);
	}

	static int
	spa_ld_load_vdev_metadata(spa_t *spa)
	{
	int error = 0;
	vdev_t *rvd = spa->spa_root_vdev;

	/*
	* If the 'multihost' property is set, then never allow a pool to
	* be imported when the system hostid is zero. The exception to
	* this rule is zdb which is always allowed to access pools.
	*/
	if (spa_multihost(spa) && spa_get_hostid(spa) == 0 &&
	(spa->spa_import_flags & ZFS_IMPORT_SKIP_MMP) == 0) {
	fnvlist_add_uint64(spa->spa_load_info,
	ZPOOL_CONFIG_MMP_STATE, MMP_STATE_NO_HOSTID);
	return (spa_vdev_err(rvd, VDEV_AUX_ACTIVE, EREMOTEIO));
	}

	/*
	* If the 'autoreplace' property is set, then post a resource notifying
	* the ZFS DE that it should not issue any faults for unopenable
	* devices. We also iterate over the vdevs, and post a sysevent for any
	* unopenable vdevs so that the normal autoreplace handler can take
	* over.
	*/
	if (spa->spa_autoreplace && spa->spa_load_state != SPA_LOAD_TRYIMPORT) {
	spa_check_removed(spa->spa_root_vdev);
	/*
	* For the import case, this is done in spa_import(), because
	* at this point we're using the spare definitions from
	* the MOS config, not necessarily from the userland config.
	*/
	if (spa->spa_load_state != SPA_LOAD_IMPORT) {
	spa_aux_check_removed(&spa->spa_spares);
	spa_aux_check_removed(&spa->spa_l2cache);
	}
	}

	/*
	* Load the vdev metadata such as metaslabs, DTLs, spacemap object, etc.
	*/
	error = vdev_load(rvd);
	if (error != 0) {
	spa_load_failed(spa, "vdev_load failed [error=%d]", error);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, error));
	}

	error = spa_ld_log_spacemaps(spa);
	if (error != 0) {
	spa_load_failed(spa, "spa_ld_log_spacemaps failed [error=%d]",
	error);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, error));
	}

	/*
	* Propagate the leaf DTLs we just loaded all the way up the vdev tree.
	*/
	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	vdev_dtl_reassess(rvd, 0, 0, B_FALSE, B_FALSE);
	spa_config_exit(spa, SCL_ALL, FTAG);

	return (0);
	}

	static int
	spa_ld_load_dedup_tables(spa_t *spa)
	{
	int error = 0;
	vdev_t *rvd = spa->spa_root_vdev;

	error = ddt_load(spa);
	if (error != 0) {
	spa_load_failed(spa, "ddt_load failed [error=%d]", error);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	return (0);
	}

	static int
	spa_ld_load_brt(spa_t *spa)
	{
	int error = 0;
	vdev_t *rvd = spa->spa_root_vdev;

	error = brt_load(spa);
	if (error != 0) {
	spa_load_failed(spa, "brt_load failed [error=%d]", error);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA, EIO));
	}

	return (0);
	}

	static int
	spa_ld_verify_logs(spa_t spa, spa_import_type_t type, const char *ereport)
	{
	vdev_t *rvd = spa->spa_root_vdev;

	if (type != SPA_IMPORT_ASSEMBLE && spa_writeable(spa)) {
	boolean_t missing = spa_check_logs(spa);
	if (missing) {
	if (spa->spa_missing_tvds != 0) {
	spa_load_note(spa, "spa_check_logs failed "
	"so dropping the logs");
	} else {
	*ereport = FM_EREPORT_ZFS_LOG_REPLAY;
	spa_load_failed(spa, "spa_check_logs failed");
	return (spa_vdev_err(rvd, VDEV_AUX_BAD_LOG,
	ENXIO));
	}
	}
	}

	return (0);
	}

	static int
	spa_ld_verify_pool_data(spa_t *spa)
	{
	int error = 0;
	vdev_t *rvd = spa->spa_root_vdev;

	/*
	* We've successfully opened the pool, verify that we're ready
	* to start pushing transactions.
	*/
	if (spa->spa_load_state != SPA_LOAD_TRYIMPORT) {
	error = spa_load_verify(spa);
	if (error != 0) {
	spa_load_failed(spa, "spa_load_verify failed "
	"[error=%d]", error);
	return (spa_vdev_err(rvd, VDEV_AUX_CORRUPT_DATA,
	error));
	}
	}

	return (0);
	}

	static void
	spa_ld_claim_log_blocks(spa_t *spa)
	{
	dmu_tx_t *tx;
	dsl_pool_t *dp = spa_get_dsl(spa);

	/*
	* Claim log blocks that haven't been committed yet.
	* This must all happen in a single txg.
	* Note: spa_claim_max_txg is updated by spa_claim_notify(),
	* invoked from zil_claim_log_block()'s i/o done callback.
	* Price of rollback is that we abandon the log.
	*/
	spa->spa_claiming = B_TRUE;

	tx = dmu_tx_create_assigned(dp, spa_first_txg(spa));
	(void) dmu_objset_find_dp(dp, dp->dp_root_dir_obj,
	zil_claim, tx, DS_FIND_CHILDREN);
	dmu_tx_commit(tx);

	spa->spa_claiming = B_FALSE;

	spa_set_log_state(spa, SPA_LOG_GOOD);
	}

	static void
	spa_ld_check_for_config_update(spa_t *spa, uint64_t config_cache_txg,
	boolean_t update_config_cache)
	{
	vdev_t *rvd = spa->spa_root_vdev;
	int need_update = B_FALSE;

	/*
	* If the config cache is stale, or we have uninitialized
	* metaslabs (see spa_vdev_add()), then update the config.
	*
	* If this is a verbatim import, trust the current
	* in-core spa_config and update the disk labels.
	*/
	if (update_config_cache \|\| config_cache_txg != spa->spa_config_txg \|\|
	spa->spa_load_state == SPA_LOAD_IMPORT \|\|
	spa->spa_load_state == SPA_LOAD_RECOVER \|\|
	(spa->spa_import_flags & ZFS_IMPORT_VERBATIM))
	need_update = B_TRUE;

	for (int c = 0; c < rvd->vdev_children; c++)
	if (rvd->vdev_child[c]->vdev_ms_array == 0)
	need_update = B_TRUE;

	/*
	* Update the config cache asynchronously in case we're the
	* root pool, in which case the config cache isn't writable yet.
	*/
	if (need_update)
	spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE);
	}

	static void
	spa_ld_prepare_for_reload(spa_t *spa)
	{
	spa_mode_t mode = spa->spa_mode;
	int async_suspended = spa->spa_async_suspended;

	spa_unload(spa);
	spa_deactivate(spa);
	spa_activate(spa, mode);

	/*
	* We save the value of spa_async_suspended as it gets reset to 0 by
	* spa_unload(). We want to restore it back to the original value before
	* returning as we might be calling spa_async_resume() later.
	*/
	spa->spa_async_suspended = async_suspended;
	}

	static int
	spa_ld_read_checkpoint_txg(spa_t *spa)
	{
	uberblock_t checkpoint;
	int error = 0;

	ASSERT0(spa->spa_checkpoint_txg);
	ASSERT(MUTEX_HELD(&spa_namespace_lock) \|\|
	spa->spa_load_thread == curthread);

	error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_ZPOOL_CHECKPOINT, sizeof (uint64_t),
	sizeof (uberblock_t) / sizeof (uint64_t), &checkpoint);

	if (error == ENOENT)
	return (0);

	if (error != 0)
	return (error);

	ASSERT3U(checkpoint.ub_txg, !=, 0);
	ASSERT3U(checkpoint.ub_checkpoint_txg, !=, 0);
	ASSERT3U(checkpoint.ub_timestamp, !=, 0);
	spa->spa_checkpoint_txg = checkpoint.ub_txg;
	spa->spa_checkpoint_info.sci_timestamp = checkpoint.ub_timestamp;

	return (0);
	}

	static int
	spa_ld_mos_init(spa_t *spa, spa_import_type_t type)
	{
	int error = 0;

	ASSERT(MUTEX_HELD(&spa_namespace_lock));
	ASSERT(spa->spa_config_source != SPA_CONFIG_SRC_NONE);

	/*
	* Never trust the config that is provided unless we are assembling
	* a pool following a split.
	* This means don't trust blkptrs and the vdev tree in general. This
	* also effectively puts the spa in read-only mode since
	* spa_writeable() checks for spa_trust_config to be true.
	* We will later load a trusted config from the MOS.
	*/
	if (type != SPA_IMPORT_ASSEMBLE)
	spa->spa_trust_config = B_FALSE;

	/*
	* Parse the config provided to create a vdev tree.
	*/
	error = spa_ld_parse_config(spa, type);
	if (error != 0)
	return (error);

	spa_import_progress_add(spa);

	/*
	* Now that we have the vdev tree, try to open each vdev. This involves
	* opening the underlying physical device, retrieving its geometry and
	* probing the vdev with a dummy I/O. The state of each vdev will be set
	* based on the success of those operations. After this we'll be ready
	* to read from the vdevs.
	*/
	error = spa_ld_open_vdevs(spa);
	if (error != 0)
	return (error);

	/*
	* Read the label of each vdev and make sure that the GUIDs stored
	* there match the GUIDs in the config provided.
	* If we're assembling a new pool that's been split off from an
	* existing pool, the labels haven't yet been updated so we skip
	* validation for now.
	*/
	if (type != SPA_IMPORT_ASSEMBLE) {
	error = spa_ld_validate_vdevs(spa);
	if (error != 0)
	return (error);
	}

	/*
	* Read all vdev labels to find the best uberblock (i.e. latest,
	* unless spa_load_max_txg is set) and store it in spa_uberblock. We
	* get the list of features required to read blkptrs in the MOS from
	* the vdev label with the best uberblock and verify that our version
	* of zfs supports them all.
	*/
	error = spa_ld_select_uberblock(spa, type);
	if (error != 0)
	return (error);

	/*
	* Pass that uberblock to the dsl_pool layer which will open the root
	* blkptr. This blkptr points to the latest version of the MOS and will
	* allow us to read its contents.
	*/
	error = spa_ld_open_rootbp(spa);
	if (error != 0)
	return (error);

	return (0);
	}

	static int
	spa_ld_checkpoint_rewind(spa_t *spa)
	{
	uberblock_t checkpoint;
	int error = 0;

	ASSERT(MUTEX_HELD(&spa_namespace_lock));
	ASSERT(spa->spa_import_flags & ZFS_IMPORT_CHECKPOINT);

	error = zap_lookup(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_ZPOOL_CHECKPOINT, sizeof (uint64_t),
	sizeof (uberblock_t) / sizeof (uint64_t), &checkpoint);

	if (error != 0) {
	spa_load_failed(spa, "unable to retrieve checkpointed "
	"uberblock from the MOS config [error=%d]", error);

	if (error == ENOENT)
	error = ZFS_ERR_NO_CHECKPOINT;

	return (error);
	}

	ASSERT3U(checkpoint.ub_txg, <, spa->spa_uberblock.ub_txg);
	ASSERT3U(checkpoint.ub_txg, ==, checkpoint.ub_checkpoint_txg);

	/*
	* We need to update the txg and timestamp of the checkpointed
	* uberblock to be higher than the latest one. This ensures that
	* the checkpointed uberblock is selected if we were to close and
	* reopen the pool right after we've written it in the vdev labels.
	* (also see block comment in vdev_uberblock_compare)
	*/
	checkpoint.ub_txg = spa->spa_uberblock.ub_txg + 1;
	checkpoint.ub_timestamp = gethrestime_sec();

	/*
	* Set current uberblock to be the checkpointed uberblock.
	*/
	spa->spa_uberblock = checkpoint;

	/*
	* If we are doing a normal rewind, then the pool is open for
	* writing and we sync the "updated" checkpointed uberblock to
	* disk. Once this is done, we've basically rewound the whole
	* pool and there is no way back.
	*
	* There are cases when we don't want to attempt and sync the
	* checkpointed uberblock to disk because we are opening a
	* pool as read-only. Specifically, verifying the checkpointed
	* state with zdb, and importing the checkpointed state to get
	* a "preview" of its content.
	*/
	if (spa_writeable(spa)) {
	vdev_t *rvd = spa->spa_root_vdev;

	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	vdev_t *svd[SPA_SYNC_MIN_VDEVS] = { NULL };
	int svdcount = 0;
	int children = rvd->vdev_children;
	int c0 = random_in_range(children);

	for (int c = 0; c < children; c++) {
	vdev_t *vd = rvd->vdev_child[(c0 + c) % children];

	/* Stop when revisiting the first vdev */
	if (c > 0 && svd[0] == vd)
	break;

	if (vd->vdev_ms_array == 0 \|\| vd->vdev_islog \|\|
	!vdev_is_concrete(vd))
	continue;

	svd[svdcount++] = vd;
	if (svdcount == SPA_SYNC_MIN_VDEVS)
	break;
	}
	error = vdev_config_sync(svd, svdcount, spa->spa_first_txg);
	if (error == 0)
	spa->spa_last_synced_guid = rvd->vdev_guid;
	spa_config_exit(spa, SCL_ALL, FTAG);

	if (error != 0) {
	spa_load_failed(spa, "failed to write checkpointed "
	"uberblock to the vdev labels [error=%d]", error);
	return (error);
	}
	}

	return (0);
	}

	static int
	spa_ld_mos_with_trusted_config(spa_t *spa, spa_import_type_t type,
	boolean_t *update_config_cache)
	{
	int error;

	/*
	* Parse the config for pool, open and validate vdevs,
	* select an uberblock, and use that uberblock to open
	* the MOS.
	*/
	error = spa_ld_mos_init(spa, type);
	if (error != 0)
	return (error);

	/*
	* Retrieve the trusted config stored in the MOS and use it to create
	* a new, exact version of the vdev tree, then reopen all vdevs.
	*/
	error = spa_ld_trusted_config(spa, type, B_FALSE);
	if (error == EAGAIN) {
	if (update_config_cache != NULL)
	*update_config_cache = B_TRUE;

	/*
	* Redo the loading process with the trusted config if it is
	* too different from the untrusted config.
	*/
	spa_ld_prepare_for_reload(spa);
	spa_load_note(spa, "RELOADING");
	error = spa_ld_mos_init(spa, type);
	if (error != 0)
	return (error);

	error = spa_ld_trusted_config(spa, type, B_TRUE);
	if (error != 0)
	return (error);

	} else if (error != 0) {
	return (error);
	}

	return (0);
	}

	/*
	* Load an existing storage pool, using the config provided. This config
	* describes which vdevs are part of the pool and is later validated against
	* partial configs present in each vdev's label and an entire copy of the
	* config stored in the MOS.
	*/
	static int
	spa_load_impl(spa_t spa, spa_import_type_t type, const char *ereport)
	{
	int error = 0;
	boolean_t missing_feat_write = B_FALSE;
	boolean_t checkpoint_rewind =
	(spa->spa_import_flags & ZFS_IMPORT_CHECKPOINT);
	boolean_t update_config_cache = B_FALSE;
	hrtime_t load_start = gethrtime();

	ASSERT(MUTEX_HELD(&spa_namespace_lock));
	ASSERT(spa->spa_config_source != SPA_CONFIG_SRC_NONE);

	spa_load_note(spa, "LOADING");

	error = spa_ld_mos_with_trusted_config(spa, type, &update_config_cache);
	if (error != 0)
	return (error);

	/*
	* If we are rewinding to the checkpoint then we need to repeat
	* everything we've done so far in this function but this time
	* selecting the checkpointed uberblock and using that to open
	* the MOS.
	*/
	if (checkpoint_rewind) {
	/*
	* If we are rewinding to the checkpoint update config cache
	* anyway.
	*/
	update_config_cache = B_TRUE;

	/*
	* Extract the checkpointed uberblock from the current MOS
	* and use this as the pool's uberblock from now on. If the
	* pool is imported as writeable we also write the checkpoint
	* uberblock to the labels, making the rewind permanent.
	*/
	error = spa_ld_checkpoint_rewind(spa);
	if (error != 0)
	return (error);

	/*
	* Redo the loading process again with the
	* checkpointed uberblock.
	*/
	spa_ld_prepare_for_reload(spa);
	spa_load_note(spa, "LOADING checkpointed uberblock");
	error = spa_ld_mos_with_trusted_config(spa, type, NULL);
	if (error != 0)
	return (error);
	}

	/*
	* Drop the namespace lock for the rest of the function.
	*/
	spa->spa_load_thread = curthread;
	mutex_exit(&spa_namespace_lock);

	/*
	* Retrieve the checkpoint txg if the pool has a checkpoint.
	*/
	spa_import_progress_set_notes(spa, "Loading checkpoint txg");
	error = spa_ld_read_checkpoint_txg(spa);
	if (error != 0)
	goto fail;

	/*
	* Retrieve the mapping of indirect vdevs. Those vdevs were removed
	* from the pool and their contents were re-mapped to other vdevs. Note
	* that everything that we read before this step must have been
	* rewritten on concrete vdevs after the last device removal was
	* initiated. Otherwise we could be reading from indirect vdevs before
	* we have loaded their mappings.
	*/
	spa_import_progress_set_notes(spa, "Loading indirect vdev metadata");
	error = spa_ld_open_indirect_vdev_metadata(spa);
	if (error != 0)
	goto fail;

	/*
	* Retrieve the full list of active features from the MOS and check if
	* they are all supported.
	*/
	spa_import_progress_set_notes(spa, "Checking feature flags");
	error = spa_ld_check_features(spa, &missing_feat_write);
	if (error != 0)
	goto fail;

	/*
	* Load several special directories from the MOS needed by the dsl_pool
	* layer.
	*/
	spa_import_progress_set_notes(spa, "Loading special MOS directories");
	error = spa_ld_load_special_directories(spa);
	if (error != 0)
	goto fail;

	/*
	* Retrieve pool properties from the MOS.
	*/
	spa_import_progress_set_notes(spa, "Loading properties");
	error = spa_ld_get_props(spa);
	if (error != 0)
	goto fail;

	/*
	* Retrieve the list of auxiliary devices - cache devices and spares -
	* and open them.
	*/
	spa_import_progress_set_notes(spa, "Loading AUX vdevs");
	error = spa_ld_open_aux_vdevs(spa, type);
	if (error != 0)
	goto fail;

	/*
	* Load the metadata for all vdevs. Also check if unopenable devices
	* should be autoreplaced.
	*/
	spa_import_progress_set_notes(spa, "Loading vdev metadata");
	error = spa_ld_load_vdev_metadata(spa);
	if (error != 0)
	goto fail;

	spa_import_progress_set_notes(spa, "Loading dedup tables");
	error = spa_ld_load_dedup_tables(spa);
	if (error != 0)
	goto fail;

	spa_import_progress_set_notes(spa, "Loading BRT");
	error = spa_ld_load_brt(spa);
	if (error != 0)
	goto fail;

	/*
	* Verify the logs now to make sure we don't have any unexpected errors
	* when we claim log blocks later.
	*/
	spa_import_progress_set_notes(spa, "Verifying Log Devices");
	error = spa_ld_verify_logs(spa, type, ereport);
	if (error != 0)
	goto fail;

	if (missing_feat_write) {
	ASSERT(spa->spa_load_state == SPA_LOAD_TRYIMPORT);

	/*
	* At this point, we know that we can open the pool in
	* read-only mode but not read-write mode. We now have enough
	* information and can return to userland.
	*/
	error = spa_vdev_err(spa->spa_root_vdev, VDEV_AUX_UNSUP_FEAT,
	ENOTSUP);
	goto fail;
	}

	/*
	* Traverse the last txgs to make sure the pool was left off in a safe
	* state. When performing an extreme rewind, we verify the whole pool,
	* which can take a very long time.
	*/
	spa_import_progress_set_notes(spa, "Verifying pool data");
	error = spa_ld_verify_pool_data(spa);
	if (error != 0)
	goto fail;

	/*
	* Calculate the deflated space for the pool. This must be done before
	* we write anything to the pool because we'd need to update the space
	* accounting using the deflated sizes.
	*/
	spa_import_progress_set_notes(spa, "Calculating deflated space");
	spa_update_dspace(spa);

	/*
	* We have now retrieved all the information we needed to open the
	* pool. If we are importing the pool in read-write mode, a few
	* additional steps must be performed to finish the import.
	*/
	spa_import_progress_set_notes(spa, "Starting import");
	if (spa_writeable(spa) && (spa->spa_load_state == SPA_LOAD_RECOVER \|\|
	spa->spa_load_max_txg == UINT64_MAX)) {
	uint64_t config_cache_txg = spa->spa_config_txg;

	ASSERT(spa->spa_load_state != SPA_LOAD_TRYIMPORT);

	/*
	* Before we do any zio_write's, complete the raidz expansion
	* scratch space copying, if necessary.
	*/
	if (RRSS_GET_STATE(&spa->spa_uberblock) == RRSS_SCRATCH_VALID)
	vdev_raidz_reflow_copy_scratch(spa);

	/*
	* In case of a checkpoint rewind, log the original txg
	* of the checkpointed uberblock.
	*/
	if (checkpoint_rewind) {
	spa_history_log_internal(spa, "checkpoint rewind",
	NULL, "rewound state to txg=%llu",
	(u_longlong_t)spa->spa_uberblock.ub_checkpoint_txg);
	}

	spa_import_progress_set_notes(spa, "Claiming ZIL blocks");
	/*
	* Traverse the ZIL and claim all blocks.
	*/
	spa_ld_claim_log_blocks(spa);

	/*
	* Kick-off the syncing thread.
	*/
	spa->spa_sync_on = B_TRUE;
	txg_sync_start(spa->spa_dsl_pool);
	mmp_thread_start(spa);

	/*
	* Wait for all claims to sync. We sync up to the highest
	* claimed log block birth time so that claimed log blocks
	* don't appear to be from the future. spa_claim_max_txg
	* will have been set for us by ZIL traversal operations
	* performed above.
	*/
	spa_import_progress_set_notes(spa, "Syncing ZIL claims");
	txg_wait_synced(spa->spa_dsl_pool, spa->spa_claim_max_txg);

	/*
	* Check if we need to request an update of the config. On the
	* next sync, we would update the config stored in vdev labels
	* and the cachefile (by default /etc/zfs/zpool.cache).
	*/
	spa_import_progress_set_notes(spa, "Updating configs");
	spa_ld_check_for_config_update(spa, config_cache_txg,
	update_config_cache);

	/*
	* Check if a rebuild was in progress and if so resume it.
	* Then check all DTLs to see if anything needs resilvering.
	* The resilver will be deferred if a rebuild was started.
	*/
	spa_import_progress_set_notes(spa, "Starting resilvers");
	if (vdev_rebuild_active(spa->spa_root_vdev)) {
	vdev_rebuild_restart(spa);
	} else if (!dsl_scan_resilvering(spa->spa_dsl_pool) &&
	vdev_resilver_needed(spa->spa_root_vdev, NULL, NULL)) {
	spa_async_request(spa, SPA_ASYNC_RESILVER);
	}

	/*
	* Log the fact that we booted up (so that we can detect if
	* we rebooted in the middle of an operation).
	*/
	spa_history_log_version(spa, "open", NULL);

	spa_import_progress_set_notes(spa,
	"Restarting device removals");
	spa_restart_removal(spa);
	spa_spawn_aux_threads(spa);

	/*
	* Delete any inconsistent datasets.
	*
	* Note:
	* Since we may be issuing deletes for clones here,
	* we make sure to do so after we've spawned all the
	* auxiliary threads above (from which the livelist
	* deletion zthr is part of).
	*/
	spa_import_progress_set_notes(spa,
	"Cleaning up inconsistent objsets");
	(void) dmu_objset_find(spa_name(spa),
	dsl_destroy_inconsistent, NULL, DS_FIND_CHILDREN);

	/*
	* Clean up any stale temporary dataset userrefs.
	*/
	spa_import_progress_set_notes(spa,
	"Cleaning up temporary userrefs");
	dsl_pool_clean_tmp_userrefs(spa->spa_dsl_pool);

	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	spa_import_progress_set_notes(spa, "Restarting initialize");
	vdev_initialize_restart(spa->spa_root_vdev);
	spa_import_progress_set_notes(spa, "Restarting TRIM");
	vdev_trim_restart(spa->spa_root_vdev);
	vdev_autotrim_restart(spa);
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	spa_import_progress_set_notes(spa, "Finished importing");
	}
	zio_handle_import_delay(spa, gethrtime() - load_start);

	spa_import_progress_remove(spa_guid(spa));
	spa_async_request(spa, SPA_ASYNC_L2CACHE_REBUILD);

	spa_load_note(spa, "LOADED");
	fail:
	mutex_enter(&spa_namespace_lock);
	spa->spa_load_thread = NULL;
	cv_broadcast(&spa_namespace_cv);

	return (error);

	}

	static int
	spa_load_retry(spa_t *spa, spa_load_state_t state)
	{
	spa_mode_t mode = spa->spa_mode;

	spa_unload(spa);
	spa_deactivate(spa);

	spa->spa_load_max_txg = spa->spa_uberblock.ub_txg - 1;

	spa_activate(spa, mode);
	spa_async_suspend(spa);

	spa_load_note(spa, "spa_load_retry: rewind, max txg: %llu",
	(u_longlong_t)spa->spa_load_max_txg);

	return (spa_load(spa, state, SPA_IMPORT_EXISTING));
	}

	/*
	* If spa_load() fails this function will try loading prior txg's. If
	* 'state' is SPA_LOAD_RECOVER and one of these loads succeeds the pool
	* will be rewound to that txg. If 'state' is not SPA_LOAD_RECOVER this
	* function will not rewind the pool and will return the same error as
	* spa_load().
	*/
	static int
	spa_load_best(spa_t *spa, spa_load_state_t state, uint64_t max_request,
	int rewind_flags)
	{
	nvlist_t *loadinfo = NULL;
	nvlist_t *config = NULL;
	int load_error, rewind_error;
	uint64_t safe_rewind_txg;
	uint64_t min_txg;

	if (spa->spa_load_txg && state == SPA_LOAD_RECOVER) {
	spa->spa_load_max_txg = spa->spa_load_txg;
	spa_set_log_state(spa, SPA_LOG_CLEAR);
	} else {
	spa->spa_load_max_txg = max_request;
	if (max_request != UINT64_MAX)
	spa->spa_extreme_rewind = B_TRUE;
	}

	load_error = rewind_error = spa_load(spa, state, SPA_IMPORT_EXISTING);
	if (load_error == 0)
	return (0);
	if (load_error == ZFS_ERR_NO_CHECKPOINT) {
	/*
	* When attempting checkpoint-rewind on a pool with no
	* checkpoint, we should not attempt to load uberblocks
	* from previous txgs when spa_load fails.
	*/
	ASSERT(spa->spa_import_flags & ZFS_IMPORT_CHECKPOINT);
	spa_import_progress_remove(spa_guid(spa));
	return (load_error);
	}

	if (spa->spa_root_vdev != NULL)
	config = spa_config_generate(spa, NULL, -1ULL, B_TRUE);

	spa->spa_last_ubsync_txg = spa->spa_uberblock.ub_txg;
	spa->spa_last_ubsync_txg_ts = spa->spa_uberblock.ub_timestamp;

	if (rewind_flags & ZPOOL_NEVER_REWIND) {
	nvlist_free(config);
	spa_import_progress_remove(spa_guid(spa));
	return (load_error);
	}

	if (state == SPA_LOAD_RECOVER) {
	/* Price of rolling back is discarding txgs, including log */
	spa_set_log_state(spa, SPA_LOG_CLEAR);
	} else {
	/*
	* If we aren't rolling back save the load info from our first
	* import attempt so that we can restore it after attempting
	* to rewind.
	*/
	loadinfo = spa->spa_load_info;
	spa->spa_load_info = fnvlist_alloc();
	}

	spa->spa_load_max_txg = spa->spa_last_ubsync_txg;
	safe_rewind_txg = spa->spa_last_ubsync_txg - TXG_DEFER_SIZE;
	min_txg = (rewind_flags & ZPOOL_EXTREME_REWIND) ?
	TXG_INITIAL : safe_rewind_txg;

	/*
	* Continue as long as we're finding errors, we're still within
	* the acceptable rewind range, and we're still finding uberblocks
	*/
	while (rewind_error && spa->spa_uberblock.ub_txg >= min_txg &&
	spa->spa_uberblock.ub_txg <= spa->spa_load_max_txg) {
	if (spa->spa_load_max_txg < safe_rewind_txg)
	spa->spa_extreme_rewind = B_TRUE;
	rewind_error = spa_load_retry(spa, state);
	}

	spa->spa_extreme_rewind = B_FALSE;
	spa->spa_load_max_txg = UINT64_MAX;

	if (config && (rewind_error \|\| state != SPA_LOAD_RECOVER))
	spa_config_set(spa, config);
	else
	nvlist_free(config);

	if (state == SPA_LOAD_RECOVER) {
	ASSERT3P(loadinfo, ==, NULL);
	spa_import_progress_remove(spa_guid(spa));
	return (rewind_error);
	} else {
	/* Store the rewind info as part of the initial load info */
	fnvlist_add_nvlist(loadinfo, ZPOOL_CONFIG_REWIND_INFO,
	spa->spa_load_info);

	/* Restore the initial load info */
	fnvlist_free(spa->spa_load_info);
	spa->spa_load_info = loadinfo;

	spa_import_progress_remove(spa_guid(spa));
	return (load_error);
	}
	}

	/*
	* Pool Open/Import
	*
	* The import case is identical to an open except that the configuration is sent
	* down from userland, instead of grabbed from the configuration cache. For the
	* case of an open, the pool configuration will exist in the
	* POOL_STATE_UNINITIALIZED state.
	*
	* The stats information (gen/count/ustats) is used to gather vdev statistics at
	* the same time open the pool, without having to keep around the spa_t in some
	* ambiguous state.
	*/
	static int
	spa_open_common(const char pool, spa_t spapp, const void tag,
	nvlist_t nvpolicy, nvlist_t *config)
	{
	spa_t *spa;
	spa_load_state_t state = SPA_LOAD_OPEN;
	int error;
	int locked = B_FALSE;
	int firstopen = B_FALSE;

	*spapp = NULL;

	/*
	* As disgusting as this is, we need to support recursive calls to this
	* function because dsl_dir_open() is called during spa_load(), and ends
	* up calling spa_open() again. The real fix is to figure out how to
	* avoid dsl_dir_open() calling this in the first place.
	*/
	if (MUTEX_NOT_HELD(&spa_namespace_lock)) {
	mutex_enter(&spa_namespace_lock);
	locked = B_TRUE;
	}

	if ((spa = spa_lookup(pool)) == NULL) {
	if (locked)
	mutex_exit(&spa_namespace_lock);
	return (SET_ERROR(ENOENT));
	}

	if (spa->spa_state == POOL_STATE_UNINITIALIZED) {
	zpool_load_policy_t policy;

	firstopen = B_TRUE;

	zpool_get_load_policy(nvpolicy ? nvpolicy : spa->spa_config,
	&policy);
	if (policy.zlp_rewind & ZPOOL_DO_REWIND)
	state = SPA_LOAD_RECOVER;

	spa_activate(spa, spa_mode_global);

	if (state != SPA_LOAD_RECOVER)
	spa->spa_last_ubsync_txg = spa->spa_load_txg = 0;
	spa->spa_config_source = SPA_CONFIG_SRC_CACHEFILE;

	zfs_dbgmsg("spa_open_common: opening %s", pool);
	error = spa_load_best(spa, state, policy.zlp_txg,
	policy.zlp_rewind);

	if (error == EBADF) {
	/*
	* If vdev_validate() returns failure (indicated by
	* EBADF), it indicates that one of the vdevs indicates
	* that the pool has been exported or destroyed. If
	* this is the case, the config cache is out of sync and
	* we should remove the pool from the namespace.
	*/
	spa_unload(spa);
	spa_deactivate(spa);
	spa_write_cachefile(spa, B_TRUE, B_TRUE, B_FALSE);
	spa_remove(spa);
	if (locked)
	mutex_exit(&spa_namespace_lock);
	return (SET_ERROR(ENOENT));
	}

	if (error) {
	/*
	* We can't open the pool, but we still have useful
	* information: the state of each vdev after the
	* attempted vdev_open(). Return this to the user.
	*/
	if (config != NULL && spa->spa_config) {
	*config = fnvlist_dup(spa->spa_config);
	fnvlist_add_nvlist(*config,
	ZPOOL_CONFIG_LOAD_INFO,
	spa->spa_load_info);
	}
	spa_unload(spa);
	spa_deactivate(spa);
	spa->spa_last_open_failed = error;
	if (locked)
	mutex_exit(&spa_namespace_lock);
	*spapp = NULL;
	return (error);
	}
	}

	spa_open_ref(spa, tag);

	if (config != NULL)
	*config = spa_config_generate(spa, NULL, -1ULL, B_TRUE);

	/*
	* If we've recovered the pool, pass back any information we
	* gathered while doing the load.
	*/
	if (state == SPA_LOAD_RECOVER && config != NULL) {
	fnvlist_add_nvlist(*config, ZPOOL_CONFIG_LOAD_INFO,
	spa->spa_load_info);
	}

	if (locked) {
	spa->spa_last_open_failed = 0;
	spa->spa_last_ubsync_txg = 0;
	spa->spa_load_txg = 0;
	mutex_exit(&spa_namespace_lock);
	}

	if (firstopen)
	zvol_create_minors_recursive(spa_name(spa));

	*spapp = spa;

	return (0);
	}

	int
	spa_open_rewind(const char name, spa_t spapp, const void tag,
	nvlist_t policy, nvlist_t *config)
	{
	return (spa_open_common(name, spapp, tag, policy, config));
	}

	int
	spa_open(const char name, spa_t spapp, const void tag)
	{
	return (spa_open_common(name, spapp, tag, NULL, NULL));
	}

	/*
	* Lookup the given spa_t, incrementing the inject count in the process,
	* preventing it from being exported or destroyed.
	*/
	spa_t *
	spa_inject_addref(char *name)
	{
	spa_t *spa;

	mutex_enter(&spa_namespace_lock);
	if ((spa = spa_lookup(name)) == NULL) {
	mutex_exit(&spa_namespace_lock);
	return (NULL);
	}
	spa->spa_inject_ref++;
	mutex_exit(&spa_namespace_lock);

	return (spa);
	}

	void
	spa_inject_delref(spa_t *spa)
	{
	mutex_enter(&spa_namespace_lock);
	spa->spa_inject_ref--;
	mutex_exit(&spa_namespace_lock);
	}

	/*
	* Add spares device information to the nvlist.
	*/
	static void
	spa_add_spares(spa_t spa, nvlist_t config)
	{
	nvlist_t **spares;
	uint_t i, nspares;
	nvlist_t *nvroot;
	uint64_t guid;
	vdev_stat_t *vs;
	uint_t vsc;
	uint64_t pool;

	ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER));

	if (spa->spa_spares.sav_count == 0)
	return;

	nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);
	VERIFY0(nvlist_lookup_nvlist_array(spa->spa_spares.sav_config,
	ZPOOL_CONFIG_SPARES, &spares, &nspares));
	if (nspares != 0) {
	fnvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
	(const nvlist_t * const *)spares, nspares);
	VERIFY0(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
	&spares, &nspares));

	/*
	* Go through and find any spares which have since been
	* repurposed as an active spare. If this is the case, update
	* their status appropriately.
	*/
	for (i = 0; i < nspares; i++) {
	guid = fnvlist_lookup_uint64(spares[i],
	ZPOOL_CONFIG_GUID);
	VERIFY0(nvlist_lookup_uint64_array(spares[i],
	ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc));
	if (spa_spare_exists(guid, &pool, NULL) &&
	pool != 0ULL) {
	vs->vs_state = VDEV_STATE_CANT_OPEN;
	vs->vs_aux = VDEV_AUX_SPARED;
	} else {
	vs->vs_state =
	spa->spa_spares.sav_vdevs[i]->vdev_state;
	}
	}
	}
	}

	/*
	* Add l2cache device information to the nvlist, including vdev stats.
	*/
	static void
	spa_add_l2cache(spa_t spa, nvlist_t config)
	{
	nvlist_t **l2cache;
	uint_t i, j, nl2cache;
	nvlist_t *nvroot;
	uint64_t guid;
	vdev_t *vd;
	vdev_stat_t *vs;
	uint_t vsc;

	ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER));

	if (spa->spa_l2cache.sav_count == 0)
	return;

	nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);
	VERIFY0(nvlist_lookup_nvlist_array(spa->spa_l2cache.sav_config,
	ZPOOL_CONFIG_L2CACHE, &l2cache, &nl2cache));
	if (nl2cache != 0) {
	fnvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
	(const nvlist_t * const *)l2cache, nl2cache);
	VERIFY0(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
	&l2cache, &nl2cache));

	/*
	* Update level 2 cache device stats.
	*/

	for (i = 0; i < nl2cache; i++) {
	guid = fnvlist_lookup_uint64(l2cache[i],
	ZPOOL_CONFIG_GUID);

	vd = NULL;
	for (j = 0; j < spa->spa_l2cache.sav_count; j++) {
	if (guid ==
	spa->spa_l2cache.sav_vdevs[j]->vdev_guid) {
	vd = spa->spa_l2cache.sav_vdevs[j];
	break;
	}
	}
	ASSERT(vd != NULL);

	VERIFY0(nvlist_lookup_uint64_array(l2cache[i],
	ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc));
	vdev_get_stats(vd, vs);
	vdev_config_generate_stats(vd, l2cache[i]);

	}
	}
	}

	static void
	spa_feature_stats_from_disk(spa_t spa, nvlist_t features)
	{
	zap_cursor_t zc;
	zap_attribute_t za;

	if (spa->spa_feat_for_read_obj != 0) {
	for (zap_cursor_init(&zc, spa->spa_meta_objset,
	spa->spa_feat_for_read_obj);
	zap_cursor_retrieve(&zc, &za) == 0;
	zap_cursor_advance(&zc)) {
	ASSERT(za.za_integer_length == sizeof (uint64_t) &&
	za.za_num_integers == 1);
	VERIFY0(nvlist_add_uint64(features, za.za_name,
	za.za_first_integer));
	}
	zap_cursor_fini(&zc);
	}

	if (spa->spa_feat_for_write_obj != 0) {
	for (zap_cursor_init(&zc, spa->spa_meta_objset,
	spa->spa_feat_for_write_obj);
	zap_cursor_retrieve(&zc, &za) == 0;
	zap_cursor_advance(&zc)) {
	ASSERT(za.za_integer_length == sizeof (uint64_t) &&
	za.za_num_integers == 1);
	VERIFY0(nvlist_add_uint64(features, za.za_name,
	za.za_first_integer));
	}
	zap_cursor_fini(&zc);
	}
	}

	static void
	spa_feature_stats_from_cache(spa_t spa, nvlist_t features)
	{
	int i;

	for (i = 0; i < SPA_FEATURES; i++) {
	zfeature_info_t feature = spa_feature_table[i];
	uint64_t refcount;

	if (feature_get_refcount(spa, &feature, &refcount) != 0)
	continue;

	VERIFY0(nvlist_add_uint64(features, feature.fi_guid, refcount));
	}
	}

	/*
	* Store a list of pool features and their reference counts in the
	* config.
	*
	* The first time this is called on a spa, allocate a new nvlist, fetch
	* the pool features and reference counts from disk, then save the list
	* in the spa. In subsequent calls on the same spa use the saved nvlist
	* and refresh its values from the cached reference counts. This
	* ensures we don't block here on I/O on a suspended pool so 'zpool
	* clear' can resume the pool.
	*/
	static void
	spa_add_feature_stats(spa_t spa, nvlist_t config)
	{
	nvlist_t *features;

	ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER));

	mutex_enter(&spa->spa_feat_stats_lock);
	features = spa->spa_feat_stats;

	if (features != NULL) {
	spa_feature_stats_from_cache(spa, features);
	} else {
	VERIFY0(nvlist_alloc(&features, NV_UNIQUE_NAME, KM_SLEEP));
	spa->spa_feat_stats = features;
	spa_feature_stats_from_disk(spa, features);
	}

	VERIFY0(nvlist_add_nvlist(config, ZPOOL_CONFIG_FEATURE_STATS,
	features));

	mutex_exit(&spa->spa_feat_stats_lock);
	}

	int
	spa_get_stats(const char name, nvlist_t *config,
	char *altroot, size_t buflen)
	{
	int error;
	spa_t *spa;

	*config = NULL;
	error = spa_open_common(name, &spa, FTAG, NULL, config);

	if (spa != NULL) {
	/*
	* This still leaves a window of inconsistency where the spares
	* or l2cache devices could change and the config would be
	* self-inconsistent.
	*/
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);

	if (*config != NULL) {
	uint64_t loadtimes[2];

	loadtimes[0] = spa->spa_loaded_ts.tv_sec;
	loadtimes[1] = spa->spa_loaded_ts.tv_nsec;
	fnvlist_add_uint64_array(*config,
	ZPOOL_CONFIG_LOADED_TIME, loadtimes, 2);

	fnvlist_add_uint64(*config,
	ZPOOL_CONFIG_ERRCOUNT,
	spa_approx_errlog_size(spa));

	if (spa_suspended(spa)) {
	fnvlist_add_uint64(*config,
	ZPOOL_CONFIG_SUSPENDED,
	spa->spa_failmode);
	fnvlist_add_uint64(*config,
	ZPOOL_CONFIG_SUSPENDED_REASON,
	spa->spa_suspended);
	}

	spa_add_spares(spa, *config);
	spa_add_l2cache(spa, *config);
	spa_add_feature_stats(spa, *config);
	}
	}

	/*
	* We want to get the alternate root even for faulted pools, so we cheat
	* and call spa_lookup() directly.
	*/
	if (altroot) {
	if (spa == NULL) {
	mutex_enter(&spa_namespace_lock);
	spa = spa_lookup(name);
	if (spa)
	spa_altroot(spa, altroot, buflen);
	else
	altroot[0] = '\0';
	spa = NULL;
	mutex_exit(&spa_namespace_lock);
	} else {
	spa_altroot(spa, altroot, buflen);
	}
	}

	if (spa != NULL) {
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	spa_close(spa, FTAG);
	}

	return (error);
	}

	/*
	* Validate that the auxiliary device array is well formed. We must have an
	* array of nvlists, each which describes a valid leaf vdev. If this is an
	* import (mode is VDEV_ALLOC_SPARE), then we allow corrupted spares to be
	* specified, as long as they are well-formed.
	*/
	static int
	spa_validate_aux_devs(spa_t spa, nvlist_t nvroot, uint64_t crtxg, int mode,
	spa_aux_vdev_t sav, const char config, uint64_t version,
	vdev_labeltype_t label)
	{
	nvlist_t **dev;
	uint_t i, ndev;
	vdev_t *vd;
	int error;

	ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);

	/*
	* It's acceptable to have no devs specified.
	*/
	if (nvlist_lookup_nvlist_array(nvroot, config, &dev, &ndev) != 0)
	return (0);

	if (ndev == 0)
	return (SET_ERROR(EINVAL));

	/*
	* Make sure the pool is formatted with a version that supports this
	* device type.
	*/
	if (spa_version(spa) < version)
	return (SET_ERROR(ENOTSUP));

	/*
	* Set the pending device list so we correctly handle device in-use
	* checking.
	*/
	sav->sav_pending = dev;
	sav->sav_npending = ndev;

	for (i = 0; i < ndev; i++) {
	if ((error = spa_config_parse(spa, &vd, dev[i], NULL, 0,
	mode)) != 0)
	goto out;

	if (!vd->vdev_ops->vdev_op_leaf) {
	vdev_free(vd);
	error = SET_ERROR(EINVAL);
	goto out;
	}

	vd->vdev_top = vd;

	if ((error = vdev_open(vd)) == 0 &&
	(error = vdev_label_init(vd, crtxg, label)) == 0) {
	fnvlist_add_uint64(dev[i], ZPOOL_CONFIG_GUID,
	vd->vdev_guid);
	}

	vdev_free(vd);

	if (error &&
	(mode != VDEV_ALLOC_SPARE && mode != VDEV_ALLOC_L2CACHE))
	goto out;
	else
	error = 0;
	}

	out:
	sav->sav_pending = NULL;
	sav->sav_npending = 0;
	return (error);
	}

	static int
	spa_validate_aux(spa_t spa, nvlist_t nvroot, uint64_t crtxg, int mode)
	{
	int error;

	ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);

	if ((error = spa_validate_aux_devs(spa, nvroot, crtxg, mode,
	&spa->spa_spares, ZPOOL_CONFIG_SPARES, SPA_VERSION_SPARES,
	VDEV_LABEL_SPARE)) != 0) {
	return (error);
	}

	return (spa_validate_aux_devs(spa, nvroot, crtxg, mode,
	&spa->spa_l2cache, ZPOOL_CONFIG_L2CACHE, SPA_VERSION_L2CACHE,
	VDEV_LABEL_L2CACHE));
	}

	static void
	spa_set_aux_vdevs(spa_aux_vdev_t sav, nvlist_t *devs, int ndevs,
	const char *config)
	{
	int i;

	if (sav->sav_config != NULL) {
	nvlist_t **olddevs;
	uint_t oldndevs;
	nvlist_t **newdevs;

	/*
	* Generate new dev list by concatenating with the
	* current dev list.
	*/
	VERIFY0(nvlist_lookup_nvlist_array(sav->sav_config, config,
	&olddevs, &oldndevs));

	newdevs = kmem_alloc(sizeof (void )
	(ndevs + oldndevs), KM_SLEEP);
	for (i = 0; i < oldndevs; i++)
	newdevs[i] = fnvlist_dup(olddevs[i]);
	for (i = 0; i < ndevs; i++)
	newdevs[i + oldndevs] = fnvlist_dup(devs[i]);

	fnvlist_remove(sav->sav_config, config);

	fnvlist_add_nvlist_array(sav->sav_config, config,
	(const nvlist_t * const *)newdevs, ndevs + oldndevs);
	for (i = 0; i < oldndevs + ndevs; i++)
	nvlist_free(newdevs[i]);
	kmem_free(newdevs, (oldndevs + ndevs) * sizeof (void *));
	} else {
	/*
	* Generate a new dev list.
	*/
	sav->sav_config = fnvlist_alloc();
	fnvlist_add_nvlist_array(sav->sav_config, config,
	(const nvlist_t * const *)devs, ndevs);
	}
	}

	/*
	* Stop and drop level 2 ARC devices
	*/
	void
	spa_l2cache_drop(spa_t *spa)
	{
	vdev_t *vd;
	int i;
	spa_aux_vdev_t *sav = &spa->spa_l2cache;

	for (i = 0; i < sav->sav_count; i++) {
	uint64_t pool;

	vd = sav->sav_vdevs[i];
	ASSERT(vd != NULL);

	if (spa_l2cache_exists(vd->vdev_guid, &pool) &&
	pool != 0ULL && l2arc_vdev_present(vd))
	l2arc_remove_vdev(vd);
	}
	}

	/*
	* Verify encryption parameters for spa creation. If we are encrypting, we must
	* have the encryption feature flag enabled.
	*/
	static int
	spa_create_check_encryption_params(dsl_crypto_params_t *dcp,
	boolean_t has_encryption)
	{
	if (dcp->cp_crypt != ZIO_CRYPT_OFF &&
	dcp->cp_crypt != ZIO_CRYPT_INHERIT &&
	!has_encryption)
	return (SET_ERROR(ENOTSUP));

	return (dmu_objset_create_crypt_check(NULL, dcp, NULL));
	}

	/*
	* Pool Creation
	*/
	int
	spa_create(const char pool, nvlist_t nvroot, nvlist_t *props,
	nvlist_t zplprops, dsl_crypto_params_t dcp)
	{
	spa_t *spa;
	const char *altroot = NULL;
	vdev_t *rvd;
	dsl_pool_t *dp;
	dmu_tx_t *tx;
	int error = 0;
	uint64_t txg = TXG_INITIAL;
	nvlist_t spares, l2cache;
	uint_t nspares, nl2cache;
	uint64_t version, obj, ndraid = 0;
	boolean_t has_features;
	boolean_t has_encryption;
	boolean_t has_allocclass;
	spa_feature_t feat;
	const char *feat_name;
	const char *poolname;
	nvlist_t *nvl;

	if (props == NULL \|\|
	nvlist_lookup_string(props,
	zpool_prop_to_name(ZPOOL_PROP_TNAME), &poolname) != 0)
	poolname = (char *)pool;

	/*
	* If this pool already exists, return failure.
	*/
	mutex_enter(&spa_namespace_lock);
	if (spa_lookup(poolname) != NULL) {
	mutex_exit(&spa_namespace_lock);
	return (SET_ERROR(EEXIST));
	}

	/*
	* Allocate a new spa_t structure.
	*/
	nvl = fnvlist_alloc();
	fnvlist_add_string(nvl, ZPOOL_CONFIG_POOL_NAME, pool);
	(void) nvlist_lookup_string(props,
	zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot);
	spa = spa_add(poolname, nvl, altroot);
	fnvlist_free(nvl);
	spa_activate(spa, spa_mode_global);

	if (props && (error = spa_prop_validate(spa, props))) {
	spa_deactivate(spa);
	spa_remove(spa);
	mutex_exit(&spa_namespace_lock);
	return (error);
	}

	/*
	* Temporary pool names should never be written to disk.
	*/
	if (poolname != pool)
	spa->spa_import_flags \|= ZFS_IMPORT_TEMP_NAME;

	has_features = B_FALSE;
	has_encryption = B_FALSE;
	has_allocclass = B_FALSE;
	for (nvpair_t *elem = nvlist_next_nvpair(props, NULL);
	elem != NULL; elem = nvlist_next_nvpair(props, elem)) {
	if (zpool_prop_feature(nvpair_name(elem))) {
	has_features = B_TRUE;

	feat_name = strchr(nvpair_name(elem), '@') + 1;
	VERIFY0(zfeature_lookup_name(feat_name, &feat));
	if (feat == SPA_FEATURE_ENCRYPTION)
	has_encryption = B_TRUE;
	if (feat == SPA_FEATURE_ALLOCATION_CLASSES)
	has_allocclass = B_TRUE;
	}
	}

	/* verify encryption params, if they were provided */
	if (dcp != NULL) {
	error = spa_create_check_encryption_params(dcp, has_encryption);
	if (error != 0) {
	spa_deactivate(spa);
	spa_remove(spa);
	mutex_exit(&spa_namespace_lock);
	return (error);
	}
	}
	if (!has_allocclass && zfs_special_devs(nvroot, NULL)) {
	spa_deactivate(spa);
	spa_remove(spa);
	mutex_exit(&spa_namespace_lock);
	return (ENOTSUP);
	}

	if (has_features \|\| nvlist_lookup_uint64(props,
	zpool_prop_to_name(ZPOOL_PROP_VERSION), &version) != 0) {
	version = SPA_VERSION;
	}
	ASSERT(SPA_VERSION_IS_SUPPORTED(version));

	spa->spa_first_txg = txg;
	spa->spa_uberblock.ub_txg = txg - 1;
	spa->spa_uberblock.ub_version = version;
	spa->spa_ubsync = spa->spa_uberblock;
	spa->spa_load_state = SPA_LOAD_CREATE;
	spa->spa_removing_phys.sr_state = DSS_NONE;
	spa->spa_removing_phys.sr_removing_vdev = -1;
	spa->spa_removing_phys.sr_prev_indirect_vdev = -1;
	spa->spa_indirect_vdevs_loaded = B_TRUE;

	/*
	* Create "The Godfather" zio to hold all async IOs
	*/
	spa->spa_async_zio_root = kmem_alloc(max_ncpus * sizeof (void *),
	KM_SLEEP);
	for (int i = 0; i < max_ncpus; i++) {
	spa->spa_async_zio_root[i] = zio_root(spa, NULL, NULL,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_SPECULATIVE \|
	ZIO_FLAG_GODFATHER);
	}

	/*
	* Create the root vdev.
	*/
	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);

	error = spa_config_parse(spa, &rvd, nvroot, NULL, 0, VDEV_ALLOC_ADD);

	ASSERT(error != 0 \|\| rvd != NULL);
	ASSERT(error != 0 \|\| spa->spa_root_vdev == rvd);

	if (error == 0 && !zfs_allocatable_devs(nvroot))
	error = SET_ERROR(EINVAL);

	if (error == 0 &&
	(error = vdev_create(rvd, txg, B_FALSE)) == 0 &&
	(error = vdev_draid_spare_create(nvroot, rvd, &ndraid, 0)) == 0 &&
	(error = spa_validate_aux(spa, nvroot, txg, VDEV_ALLOC_ADD)) == 0) {
	/*
	* instantiate the metaslab groups (this will dirty the vdevs)
	* we can no longer error exit past this point
	*/
	for (int c = 0; error == 0 && c < rvd->vdev_children; c++) {
	vdev_t *vd = rvd->vdev_child[c];

	vdev_metaslab_set_size(vd);
	vdev_expand(vd, txg);
	}
	}

	spa_config_exit(spa, SCL_ALL, FTAG);

	if (error != 0) {
	spa_unload(spa);
	spa_deactivate(spa);
	spa_remove(spa);
	mutex_exit(&spa_namespace_lock);
	return (error);
	}

	/*
	* Get the list of spares, if specified.
	*/
	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
	&spares, &nspares) == 0) {
	spa->spa_spares.sav_config = fnvlist_alloc();
	fnvlist_add_nvlist_array(spa->spa_spares.sav_config,
	ZPOOL_CONFIG_SPARES, (const nvlist_t * const *)spares,
	nspares);
	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	spa_load_spares(spa);
	spa_config_exit(spa, SCL_ALL, FTAG);
	spa->spa_spares.sav_sync = B_TRUE;
	}

	/*
	* Get the list of level 2 cache devices, if specified.
	*/
	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
	&l2cache, &nl2cache) == 0) {
	VERIFY0(nvlist_alloc(&spa->spa_l2cache.sav_config,
	NV_UNIQUE_NAME, KM_SLEEP));
	fnvlist_add_nvlist_array(spa->spa_l2cache.sav_config,
	ZPOOL_CONFIG_L2CACHE, (const nvlist_t * const *)l2cache,
	nl2cache);
	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	spa_load_l2cache(spa);
	spa_config_exit(spa, SCL_ALL, FTAG);
	spa->spa_l2cache.sav_sync = B_TRUE;
	}

	spa->spa_is_initializing = B_TRUE;
	spa->spa_dsl_pool = dp = dsl_pool_create(spa, zplprops, dcp, txg);
	spa->spa_is_initializing = B_FALSE;

	/*
	* Create DDTs (dedup tables).
	*/
	ddt_create(spa);
	/*
	* Create BRT table and BRT table object.
	*/
	brt_create(spa);

	spa_update_dspace(spa);

	tx = dmu_tx_create_assigned(dp, txg);

	/*
	* Create the pool's history object.
	*/
	if (version >= SPA_VERSION_ZPOOL_HISTORY && !spa->spa_history)
	spa_history_create_obj(spa, tx);

	spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_CREATE);
	spa_history_log_version(spa, "create", tx);

	/*
	* Create the pool config object.
	*/
	spa->spa_config_object = dmu_object_alloc(spa->spa_meta_objset,
	DMU_OT_PACKED_NVLIST, SPA_CONFIG_BLOCKSIZE,
	DMU_OT_PACKED_NVLIST_SIZE, sizeof (uint64_t), tx);

	if (zap_add(spa->spa_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CONFIG,
	sizeof (uint64_t), 1, &spa->spa_config_object, tx) != 0) {
	cmn_err(CE_PANIC, "failed to add pool config");
	}

	if (zap_add(spa->spa_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CREATION_VERSION,
	sizeof (uint64_t), 1, &version, tx) != 0) {
	cmn_err(CE_PANIC, "failed to add pool version");
	}

	/* Newly created pools with the right version are always deflated. */
	if (version >= SPA_VERSION_RAIDZ_DEFLATE) {
	spa->spa_deflate = TRUE;
	if (zap_add(spa->spa_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DEFLATE,
	sizeof (uint64_t), 1, &spa->spa_deflate, tx) != 0) {
	cmn_err(CE_PANIC, "failed to add deflate");
	}
	}

	/*
	* Create the deferred-free bpobj. Turn off compression
	* because sync-to-convergence takes longer if the blocksize
	* keeps changing.
	*/
	obj = bpobj_alloc(spa->spa_meta_objset, 1 << 14, tx);
	dmu_object_set_compress(spa->spa_meta_objset, obj,
	ZIO_COMPRESS_OFF, tx);
	if (zap_add(spa->spa_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_SYNC_BPOBJ,
	sizeof (uint64_t), 1, &obj, tx) != 0) {
	cmn_err(CE_PANIC, "failed to add bpobj");
	}
	VERIFY3U(0, ==, bpobj_open(&spa->spa_deferred_bpobj,
	spa->spa_meta_objset, obj));

	/*
	* Generate some random noise for salted checksums to operate on.
	*/
	(void) random_get_pseudo_bytes(spa->spa_cksum_salt.zcs_bytes,
	sizeof (spa->spa_cksum_salt.zcs_bytes));

	/*
	* Set pool properties.
	*/
	spa->spa_bootfs = zpool_prop_default_numeric(ZPOOL_PROP_BOOTFS);
	spa->spa_delegation = zpool_prop_default_numeric(ZPOOL_PROP_DELEGATION);
	spa->spa_failmode = zpool_prop_default_numeric(ZPOOL_PROP_FAILUREMODE);
	spa->spa_autoexpand = zpool_prop_default_numeric(ZPOOL_PROP_AUTOEXPAND);
	spa->spa_multihost = zpool_prop_default_numeric(ZPOOL_PROP_MULTIHOST);
	spa->spa_autotrim = zpool_prop_default_numeric(ZPOOL_PROP_AUTOTRIM);

	if (props != NULL) {
	spa_configfile_set(spa, props, B_FALSE);
	spa_sync_props(props, tx);
	}

	for (int i = 0; i < ndraid; i++)
	spa_feature_incr(spa, SPA_FEATURE_DRAID, tx);

	dmu_tx_commit(tx);

	spa->spa_sync_on = B_TRUE;
	txg_sync_start(dp);
	mmp_thread_start(spa);
	txg_wait_synced(dp, txg);

	spa_spawn_aux_threads(spa);

	spa_write_cachefile(spa, B_FALSE, B_TRUE, B_TRUE);

	/*
	* Don't count references from objsets that are already closed
	* and are making their way through the eviction process.
	*/
	spa_evicting_os_wait(spa);
	spa->spa_minref = zfs_refcount_count(&spa->spa_refcount);
	spa->spa_load_state = SPA_LOAD_NONE;

	spa_import_os(spa);

	mutex_exit(&spa_namespace_lock);

	return (0);
	}

	/*
	* Import a non-root pool into the system.
	*/
	int
	spa_import(char pool, nvlist_t config, nvlist_t *props, uint64_t flags)
	{
	spa_t *spa;
	const char *altroot = NULL;
	spa_load_state_t state = SPA_LOAD_IMPORT;
	zpool_load_policy_t policy;
	spa_mode_t mode = spa_mode_global;
	uint64_t readonly = B_FALSE;
	int error;
	nvlist_t *nvroot;
	nvlist_t spares, l2cache;
	uint_t nspares, nl2cache;

	/*
	* If a pool with this name exists, return failure.
	*/
	mutex_enter(&spa_namespace_lock);
	if (spa_lookup(pool) != NULL) {
	mutex_exit(&spa_namespace_lock);
	return (SET_ERROR(EEXIST));
	}

	/*
	* Create and initialize the spa structure.
	*/
	(void) nvlist_lookup_string(props,
	zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot);
	(void) nvlist_lookup_uint64(props,
	zpool_prop_to_name(ZPOOL_PROP_READONLY), &readonly);
	if (readonly)
	mode = SPA_MODE_READ;
	spa = spa_add(pool, config, altroot);
	spa->spa_import_flags = flags;

	/*
	* Verbatim import - Take a pool and insert it into the namespace
	* as if it had been loaded at boot.
	*/
	if (spa->spa_import_flags & ZFS_IMPORT_VERBATIM) {
	if (props != NULL)
	spa_configfile_set(spa, props, B_FALSE);

	spa_write_cachefile(spa, B_FALSE, B_TRUE, B_FALSE);
	spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_IMPORT);
	zfs_dbgmsg("spa_import: verbatim import of %s", pool);
	mutex_exit(&spa_namespace_lock);
	return (0);
	}

	spa_activate(spa, mode);

	/*
	* Don't start async tasks until we know everything is healthy.
	*/
	spa_async_suspend(spa);

	zpool_get_load_policy(config, &policy);
	if (policy.zlp_rewind & ZPOOL_DO_REWIND)
	state = SPA_LOAD_RECOVER;

	spa->spa_config_source = SPA_CONFIG_SRC_TRYIMPORT;

	if (state != SPA_LOAD_RECOVER) {
	spa->spa_last_ubsync_txg = spa->spa_load_txg = 0;
	zfs_dbgmsg("spa_import: importing %s", pool);
	} else {
	zfs_dbgmsg("spa_import: importing %s, max_txg=%lld "
	"(RECOVERY MODE)", pool, (longlong_t)policy.zlp_txg);
	}
	error = spa_load_best(spa, state, policy.zlp_txg, policy.zlp_rewind);

	/*
	* Propagate anything learned while loading the pool and pass it
	* back to caller (i.e. rewind info, missing devices, etc).
	*/
	fnvlist_add_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, spa->spa_load_info);

	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	/*
	* Toss any existing sparelist, as it doesn't have any validity
	* anymore, and conflicts with spa_has_spare().
	*/
	if (spa->spa_spares.sav_config) {
	nvlist_free(spa->spa_spares.sav_config);
	spa->spa_spares.sav_config = NULL;
	spa_load_spares(spa);
	}
	if (spa->spa_l2cache.sav_config) {
	nvlist_free(spa->spa_l2cache.sav_config);
	spa->spa_l2cache.sav_config = NULL;
	spa_load_l2cache(spa);
	}

	nvroot = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);
	spa_config_exit(spa, SCL_ALL, FTAG);

	if (props != NULL)
	spa_configfile_set(spa, props, B_FALSE);

	if (error != 0 \|\| (props && spa_writeable(spa) &&
	(error = spa_prop_set(spa, props)))) {
	spa_unload(spa);
	spa_deactivate(spa);
	spa_remove(spa);
	mutex_exit(&spa_namespace_lock);
	return (error);
	}

	spa_async_resume(spa);

	/*
	* Override any spares and level 2 cache devices as specified by
	* the user, as these may have correct device names/devids, etc.
	*/
	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
	&spares, &nspares) == 0) {
	if (spa->spa_spares.sav_config)
	fnvlist_remove(spa->spa_spares.sav_config,
	ZPOOL_CONFIG_SPARES);
	else
	spa->spa_spares.sav_config = fnvlist_alloc();
	fnvlist_add_nvlist_array(spa->spa_spares.sav_config,
	ZPOOL_CONFIG_SPARES, (const nvlist_t * const *)spares,
	nspares);
	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	spa_load_spares(spa);
	spa_config_exit(spa, SCL_ALL, FTAG);
	spa->spa_spares.sav_sync = B_TRUE;
	}
	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
	&l2cache, &nl2cache) == 0) {
	if (spa->spa_l2cache.sav_config)
	fnvlist_remove(spa->spa_l2cache.sav_config,
	ZPOOL_CONFIG_L2CACHE);
	else
	spa->spa_l2cache.sav_config = fnvlist_alloc();
	fnvlist_add_nvlist_array(spa->spa_l2cache.sav_config,
	ZPOOL_CONFIG_L2CACHE, (const nvlist_t * const *)l2cache,
	nl2cache);
	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	spa_load_l2cache(spa);
	spa_config_exit(spa, SCL_ALL, FTAG);
	spa->spa_l2cache.sav_sync = B_TRUE;
	}

	/*
	* Check for any removed devices.
	*/
	if (spa->spa_autoreplace) {
	spa_aux_check_removed(&spa->spa_spares);
	spa_aux_check_removed(&spa->spa_l2cache);
	}

	if (spa_writeable(spa)) {
	/*
	* Update the config cache to include the newly-imported pool.
	*/
	spa_config_update(spa, SPA_CONFIG_UPDATE_POOL);
	}

	/*
	* It's possible that the pool was expanded while it was exported.
	* We kick off an async task to handle this for us.
	*/
	spa_async_request(spa, SPA_ASYNC_AUTOEXPAND);

	spa_history_log_version(spa, "import", NULL);

	spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_IMPORT);

	mutex_exit(&spa_namespace_lock);

	zvol_create_minors_recursive(pool);

	spa_import_os(spa);

	return (0);
	}

	nvlist_t *
	spa_tryimport(nvlist_t *tryconfig)
	{
	nvlist_t *config = NULL;
	const char poolname, cachefile;
	spa_t *spa;
	uint64_t state;
	int error;
	zpool_load_policy_t policy;

	if (nvlist_lookup_string(tryconfig, ZPOOL_CONFIG_POOL_NAME, &poolname))
	return (NULL);

	if (nvlist_lookup_uint64(tryconfig, ZPOOL_CONFIG_POOL_STATE, &state))
	return (NULL);

	/*
	* Create and initialize the spa structure.
	*/
	char *name = kmem_alloc(MAXPATHLEN, KM_SLEEP);
	(void) snprintf(name, MAXPATHLEN, "%s-%llx-%s",
	TRYIMPORT_NAME, (u_longlong_t)(uintptr_t)curthread, poolname);

	mutex_enter(&spa_namespace_lock);
	spa = spa_add(name, tryconfig, NULL);
	spa_activate(spa, SPA_MODE_READ);
	kmem_free(name, MAXPATHLEN);

	/*
	* Rewind pool if a max txg was provided.
	*/
	zpool_get_load_policy(spa->spa_config, &policy);
	if (policy.zlp_txg != UINT64_MAX) {
	spa->spa_load_max_txg = policy.zlp_txg;
	spa->spa_extreme_rewind = B_TRUE;
	zfs_dbgmsg("spa_tryimport: importing %s, max_txg=%lld",
	poolname, (longlong_t)policy.zlp_txg);
	} else {
	zfs_dbgmsg("spa_tryimport: importing %s", poolname);
	}

	if (nvlist_lookup_string(tryconfig, ZPOOL_CONFIG_CACHEFILE, &cachefile)
	== 0) {
	zfs_dbgmsg("spa_tryimport: using cachefile '%s'", cachefile);
	spa->spa_config_source = SPA_CONFIG_SRC_CACHEFILE;
	} else {
	spa->spa_config_source = SPA_CONFIG_SRC_SCAN;
	}

	/*
	* spa_import() relies on a pool config fetched by spa_try_import()
	* for spare/cache devices. Import flags are not passed to
	* spa_tryimport(), which makes it return early due to a missing log
	* device and missing retrieving the cache device and spare eventually.
	* Passing ZFS_IMPORT_MISSING_LOG to spa_tryimport() makes it fetch
	* the correct configuration regardless of the missing log device.
	*/
	spa->spa_import_flags \|= ZFS_IMPORT_MISSING_LOG;

	error = spa_load(spa, SPA_LOAD_TRYIMPORT, SPA_IMPORT_EXISTING);

	/*
	* If 'tryconfig' was at least parsable, return the current config.
	*/
	if (spa->spa_root_vdev != NULL) {
	config = spa_config_generate(spa, NULL, -1ULL, B_TRUE);
	fnvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME, poolname);
	fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE, state);
	fnvlist_add_uint64(config, ZPOOL_CONFIG_TIMESTAMP,
	spa->spa_uberblock.ub_timestamp);
	fnvlist_add_nvlist(config, ZPOOL_CONFIG_LOAD_INFO,
	spa->spa_load_info);
	fnvlist_add_uint64(config, ZPOOL_CONFIG_ERRATA,
	spa->spa_errata);

	/*
	* If the bootfs property exists on this pool then we
	* copy it out so that external consumers can tell which
	* pools are bootable.
	*/
	if ((!error \|\| error == EEXIST) && spa->spa_bootfs) {
	char *tmpname = kmem_alloc(MAXPATHLEN, KM_SLEEP);

	/*
	* We have to play games with the name since the
	* pool was opened as TRYIMPORT_NAME.
	*/
	if (dsl_dsobj_to_dsname(spa_name(spa),
	spa->spa_bootfs, tmpname) == 0) {
	char *cp;
	char *dsname;

	dsname = kmem_alloc(MAXPATHLEN, KM_SLEEP);

	cp = strchr(tmpname, '/');
	if (cp == NULL) {
	(void) strlcpy(dsname, tmpname,
	MAXPATHLEN);
	} else {
	(void) snprintf(dsname, MAXPATHLEN,
	"%s/%s", poolname, ++cp);
	}
	fnvlist_add_string(config, ZPOOL_CONFIG_BOOTFS,
	dsname);
	kmem_free(dsname, MAXPATHLEN);
	}
	kmem_free(tmpname, MAXPATHLEN);
	}

	/*
	* Add the list of hot spares and level 2 cache devices.
	*/
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	spa_add_spares(spa, config);
	spa_add_l2cache(spa, config);
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	}

	spa_unload(spa);
	spa_deactivate(spa);
	spa_remove(spa);
	mutex_exit(&spa_namespace_lock);

	return (config);
	}

	/*
	* Pool export/destroy
	*
	* The act of destroying or exporting a pool is very simple. We make sure there
	* is no more pending I/O and any references to the pool are gone. Then, we
	* update the pool state and sync all the labels to disk, removing the
	* configuration from the cache afterwards. If the 'hardforce' flag is set, then
	* we don't sync the labels or remove the configuration cache.
	*/
	static int
	spa_export_common(const char pool, int new_state, nvlist_t *oldconfig,
	boolean_t force, boolean_t hardforce)
	{
	- int error;
	+ int error = 0;
	spa_t *spa;
	hrtime_t export_start = gethrtime();

	if (oldconfig)
	*oldconfig = NULL;

	if (!(spa_mode_global & SPA_MODE_WRITE))
	return (SET_ERROR(EROFS));

	mutex_enter(&spa_namespace_lock);
	if ((spa = spa_lookup(pool)) == NULL) {
	mutex_exit(&spa_namespace_lock);
	return (SET_ERROR(ENOENT));
	}

	if (spa->spa_is_exporting) {
	/* the pool is being exported by another thread */
	mutex_exit(&spa_namespace_lock);
	return (SET_ERROR(ZFS_ERR_EXPORT_IN_PROGRESS));
	}
	spa->spa_is_exporting = B_TRUE;

	/*
	- * Put a hold on the pool, drop the namespace lock, stop async tasks,
	- * reacquire the namespace lock, and see if we can export.
	+ * Put a hold on the pool, drop the namespace lock, stop async tasks
	+ * and see if we can export.
	*/
	spa_open_ref(spa, FTAG);
	mutex_exit(&spa_namespace_lock);
	spa_async_suspend(spa);
	if (spa->spa_zvol_taskq) {
	zvol_remove_minors(spa, spa_name(spa), B_TRUE);
	taskq_wait(spa->spa_zvol_taskq);
	}
	mutex_enter(&spa_namespace_lock);
	+ spa->spa_export_thread = curthread;
	spa_close(spa, FTAG);

	- if (spa->spa_state == POOL_STATE_UNINITIALIZED)
	+ if (spa->spa_state == POOL_STATE_UNINITIALIZED) {
	+ mutex_exit(&spa_namespace_lock);
	goto export_spa;
	+ }
	+
	/*
	* The pool will be in core if it's openable, in which case we can
	* modify its state. Objsets may be open only because they're dirty,
	* so we have to force it to sync before checking spa_refcnt.
	*/
	if (spa->spa_sync_on) {
	txg_wait_synced(spa->spa_dsl_pool, 0);
	spa_evicting_os_wait(spa);
	}

	/*
	* A pool cannot be exported or destroyed if there are active
	* references. If we are resetting a pool, allow references by
	* fault injection handlers.
	*/
	if (!spa_refcount_zero(spa) \|\| (spa->spa_inject_ref != 0)) {
	error = SET_ERROR(EBUSY);
	goto fail;
	}

	+ mutex_exit(&spa_namespace_lock);
	+ /*
	+ * At this point we no longer hold the spa_namespace_lock and
	+ * there were no references on the spa. Future spa_lookups will
	+ * notice the spa->spa_export_thread and wait until we signal
	+ * that we are finshed.
	+ */
	+
	if (spa->spa_sync_on) {
	vdev_t *rvd = spa->spa_root_vdev;
	/*
	* A pool cannot be exported if it has an active shared spare.
	* This is to prevent other pools stealing the active spare
	* from an exported pool. At user's own will, such pool can
	* be forcedly exported.
	*/
	if (!force && new_state == POOL_STATE_EXPORTED &&
	spa_has_active_shared_spare(spa)) {
	error = SET_ERROR(EXDEV);
	+ mutex_enter(&spa_namespace_lock);
	goto fail;
	}

	/*
	* We're about to export or destroy this pool. Make sure
	* we stop all initialization and trim activity here before
	* we set the spa_final_txg. This will ensure that all
	* dirty data resulting from the initialization is
	* committed to disk before we unload the pool.
	*/
	vdev_initialize_stop_all(rvd, VDEV_INITIALIZE_ACTIVE);
	vdev_trim_stop_all(rvd, VDEV_TRIM_ACTIVE);
	vdev_autotrim_stop_all(spa);
	vdev_rebuild_stop_all(spa);

	/*
	* We want this to be reflected on every label,
	* so mark them all dirty. spa_unload() will do the
	* final sync that pushes these changes out.
	*/
	if (new_state != POOL_STATE_UNINITIALIZED && !hardforce) {
	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	spa->spa_state = new_state;
	vdev_config_dirty(rvd);
	spa_config_exit(spa, SCL_ALL, FTAG);
	}

	/*
	* If the log space map feature is enabled and the pool is
	* getting exported (but not destroyed), we want to spend some
	* time flushing as many metaslabs as we can in an attempt to
	* destroy log space maps and save import time. This has to be
	* done before we set the spa_final_txg, otherwise
	* spa_sync() -> spa_flush_metaslabs() may dirty the final TXGs.
	* spa_should_flush_logs_on_unload() should be called after
	* spa_state has been set to the new_state.
	*/
	if (spa_should_flush_logs_on_unload(spa))
	spa_unload_log_sm_flush_all(spa);

	if (new_state != POOL_STATE_UNINITIALIZED && !hardforce) {
	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	spa->spa_final_txg = spa_last_synced_txg(spa) +
	TXG_DEFER_SIZE + 1;
	spa_config_exit(spa, SCL_ALL, FTAG);
	}
	}

	export_spa:
	spa_export_os(spa);

	if (new_state == POOL_STATE_DESTROYED)
	spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_DESTROY);
	else if (new_state == POOL_STATE_EXPORTED)
	spa_event_notify(spa, NULL, NULL, ESC_ZFS_POOL_EXPORT);

	if (spa->spa_state != POOL_STATE_UNINITIALIZED) {
	spa_unload(spa);
	spa_deactivate(spa);
	}

	if (oldconfig && spa->spa_config)
	*oldconfig = fnvlist_dup(spa->spa_config);

	+ if (new_state == POOL_STATE_EXPORTED)
	+ zio_handle_export_delay(spa, gethrtime() - export_start);
	+
	+ /*
	+ * Take the namespace lock for the actual spa_t removal
	+ */
	+ mutex_enter(&spa_namespace_lock);
	if (new_state != POOL_STATE_UNINITIALIZED) {
	if (!hardforce)
	spa_write_cachefile(spa, B_TRUE, B_TRUE, B_FALSE);
	spa_remove(spa);
	} else {
	/*
	* If spa_remove() is not called for this spa_t and
	* there is any possibility that it can be reused,
	* we make sure to reset the exporting flag.
	*/
	spa->spa_is_exporting = B_FALSE;
	+ spa->spa_export_thread = NULL;
	}

	- if (new_state == POOL_STATE_EXPORTED)
	- zio_handle_export_delay(spa, gethrtime() - export_start);
	-
	+ /*
	+ * Wake up any waiters in spa_lookup()
	+ */
	+ cv_broadcast(&spa_namespace_cv);
	mutex_exit(&spa_namespace_lock);
	return (0);

	fail:
	spa->spa_is_exporting = B_FALSE;
	+ spa->spa_export_thread = NULL;
	+
	spa_async_resume(spa);
	+ /*
	+ * Wake up any waiters in spa_lookup()
	+ */
	+ cv_broadcast(&spa_namespace_cv);
	mutex_exit(&spa_namespace_lock);
	return (error);
	}

	/*
	* Destroy a storage pool.
	*/
	int
	spa_destroy(const char *pool)
	{
	return (spa_export_common(pool, POOL_STATE_DESTROYED, NULL,
	B_FALSE, B_FALSE));
	}

	/*
	* Export a storage pool.
	*/
	int
	spa_export(const char pool, nvlist_t *oldconfig, boolean_t force,
	boolean_t hardforce)
	{
	return (spa_export_common(pool, POOL_STATE_EXPORTED, oldconfig,
	force, hardforce));
	}

	/*
	* Similar to spa_export(), this unloads the spa_t without actually removing it
	* from the namespace in any way.
	*/
	int
	spa_reset(const char *pool)
	{
	return (spa_export_common(pool, POOL_STATE_UNINITIALIZED, NULL,
	B_FALSE, B_FALSE));
	}

	/*
	* ==========================================================================
	* Device manipulation
	* ==========================================================================
	*/

	/*
	* This is called as a synctask to increment the draid feature flag
	*/
	static void
	spa_draid_feature_incr(void arg, dmu_tx_t tx)
	{
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	int draid = (int)(uintptr_t)arg;

	for (int c = 0; c < draid; c++)
	spa_feature_incr(spa, SPA_FEATURE_DRAID, tx);
	}

	/*
	* Add a device to a storage pool.
	*/
	int
	spa_vdev_add(spa_t spa, nvlist_t nvroot, boolean_t check_ashift)
	{
	uint64_t txg, ndraid = 0;
	int error;
	vdev_t *rvd = spa->spa_root_vdev;
	vdev_t vd, tvd;
	nvlist_t spares, l2cache;
	uint_t nspares, nl2cache;

	ASSERT(spa_writeable(spa));

	txg = spa_vdev_enter(spa);

	if ((error = spa_config_parse(spa, &vd, nvroot, NULL, 0,
	VDEV_ALLOC_ADD)) != 0)
	return (spa_vdev_exit(spa, NULL, txg, error));

	spa->spa_pending_vdev = vd; /* spa_vdev_exit() will clear this */

	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, &spares,
	&nspares) != 0)
	nspares = 0;

	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, &l2cache,
	&nl2cache) != 0)
	nl2cache = 0;

	if (vd->vdev_children == 0 && nspares == 0 && nl2cache == 0)
	return (spa_vdev_exit(spa, vd, txg, EINVAL));

	if (vd->vdev_children != 0 &&
	(error = vdev_create(vd, txg, B_FALSE)) != 0) {
	return (spa_vdev_exit(spa, vd, txg, error));
	}

	/*
	* The virtual dRAID spares must be added after vdev tree is created
	* and the vdev guids are generated. The guid of their associated
	* dRAID is stored in the config and used when opening the spare.
	*/
	if ((error = vdev_draid_spare_create(nvroot, vd, &ndraid,
	rvd->vdev_children)) == 0) {
	if (ndraid > 0 && nvlist_lookup_nvlist_array(nvroot,
	ZPOOL_CONFIG_SPARES, &spares, &nspares) != 0)
	nspares = 0;
	} else {
	return (spa_vdev_exit(spa, vd, txg, error));
	}

	/*
	* We must validate the spares and l2cache devices after checking the
	* children. Otherwise, vdev_inuse() will blindly overwrite the spare.
	*/
	if ((error = spa_validate_aux(spa, nvroot, txg, VDEV_ALLOC_ADD)) != 0)
	return (spa_vdev_exit(spa, vd, txg, error));

	/*
	* If we are in the middle of a device removal, we can only add
	* devices which match the existing devices in the pool.
	* If we are in the middle of a removal, or have some indirect
	* vdevs, we can not add raidz or dRAID top levels.
	*/
	if (spa->spa_vdev_removal != NULL \|\|
	spa->spa_removing_phys.sr_prev_indirect_vdev != -1) {
	for (int c = 0; c < vd->vdev_children; c++) {
	tvd = vd->vdev_child[c];
	if (spa->spa_vdev_removal != NULL &&
	tvd->vdev_ashift != spa->spa_max_ashift) {
	return (spa_vdev_exit(spa, vd, txg, EINVAL));
	}
	/* Fail if top level vdev is raidz or a dRAID */
	if (vdev_get_nparity(tvd) != 0)
	return (spa_vdev_exit(spa, vd, txg, EINVAL));

	/*
	* Need the top level mirror to be
	* a mirror of leaf vdevs only
	*/
	if (tvd->vdev_ops == &vdev_mirror_ops) {
	for (uint64_t cid = 0;
	cid < tvd->vdev_children; cid++) {
	vdev_t *cvd = tvd->vdev_child[cid];
	if (!cvd->vdev_ops->vdev_op_leaf) {
	return (spa_vdev_exit(spa, vd,
	txg, EINVAL));
	}
	}
	}
	}
	}

	if (check_ashift && spa->spa_max_ashift == spa->spa_min_ashift) {
	for (int c = 0; c < vd->vdev_children; c++) {
	tvd = vd->vdev_child[c];
	if (tvd->vdev_ashift != spa->spa_max_ashift) {
	return (spa_vdev_exit(spa, vd, txg,
	ZFS_ERR_ASHIFT_MISMATCH));
	}
	}
	}

	for (int c = 0; c < vd->vdev_children; c++) {
	tvd = vd->vdev_child[c];
	vdev_remove_child(vd, tvd);
	tvd->vdev_id = rvd->vdev_children;
	vdev_add_child(rvd, tvd);
	vdev_config_dirty(tvd);
	}

	if (nspares != 0) {
	spa_set_aux_vdevs(&spa->spa_spares, spares, nspares,
	ZPOOL_CONFIG_SPARES);
	spa_load_spares(spa);
	spa->spa_spares.sav_sync = B_TRUE;
	}

	if (nl2cache != 0) {
	spa_set_aux_vdevs(&spa->spa_l2cache, l2cache, nl2cache,
	ZPOOL_CONFIG_L2CACHE);
	spa_load_l2cache(spa);
	spa->spa_l2cache.sav_sync = B_TRUE;
	}

	/*
	* We can't increment a feature while holding spa_vdev so we
	* have to do it in a synctask.
	*/
	if (ndraid != 0) {
	dmu_tx_t *tx;

	tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);
	dsl_sync_task_nowait(spa->spa_dsl_pool, spa_draid_feature_incr,
	(void *)(uintptr_t)ndraid, tx);
	dmu_tx_commit(tx);
	}

	/*
	* We have to be careful when adding new vdevs to an existing pool.
	* If other threads start allocating from these vdevs before we
	* sync the config cache, and we lose power, then upon reboot we may
	* fail to open the pool because there are DVAs that the config cache
	* can't translate. Therefore, we first add the vdevs without
	* initializing metaslabs; sync the config cache (via spa_vdev_exit());
	* and then let spa_config_update() initialize the new metaslabs.
	*
	* spa_load() checks for added-but-not-initialized vdevs, so that
	* if we lose power at any point in this sequence, the remaining
	* steps will be completed the next time we load the pool.
	*/
	(void) spa_vdev_exit(spa, vd, txg, 0);

	mutex_enter(&spa_namespace_lock);
	spa_config_update(spa, SPA_CONFIG_UPDATE_POOL);
	spa_event_notify(spa, NULL, NULL, ESC_ZFS_VDEV_ADD);
	mutex_exit(&spa_namespace_lock);

	return (0);
	}

	/*
	* Attach a device to a vdev specified by its guid. The vdev type can be
	* a mirror, a raidz, or a leaf device that is also a top-level (e.g. a
	* single device). When the vdev is a single device, a mirror vdev will be
	* automatically inserted.
	*
	* If 'replacing' is specified, the new device is intended to replace the
	* existing device; in this case the two devices are made into their own
	* mirror using the 'replacing' vdev, which is functionally identical to
	* the mirror vdev (it actually reuses all the same ops) but has a few
	* extra rules: you can't attach to it after it's been created, and upon
	* completion of resilvering, the first disk (the one being replaced)
	* is automatically detached.
	*
	* If 'rebuild' is specified, then sequential reconstruction (a.ka. rebuild)
	* should be performed instead of traditional healing reconstruction. From
	* an administrators perspective these are both resilver operations.
	*/
	int
	spa_vdev_attach(spa_t spa, uint64_t guid, nvlist_t nvroot, int replacing,
	int rebuild)
	{
	uint64_t txg, dtl_max_txg;
	vdev_t *rvd = spa->spa_root_vdev;
	vdev_t oldvd, newvd, newrootvd, pvd, *tvd;
	vdev_ops_t *pvops;
	char oldvdpath, newvdpath;
	int newvd_isspare = B_FALSE;
	int error;

	ASSERT(spa_writeable(spa));

	txg = spa_vdev_enter(spa);

	oldvd = spa_lookup_by_guid(spa, guid, B_FALSE);

	ASSERT(MUTEX_HELD(&spa_namespace_lock));
	if (spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT)) {
	error = (spa_has_checkpoint(spa)) ?
	ZFS_ERR_CHECKPOINT_EXISTS : ZFS_ERR_DISCARDING_CHECKPOINT;
	return (spa_vdev_exit(spa, NULL, txg, error));
	}

	if (rebuild) {
	if (!spa_feature_is_enabled(spa, SPA_FEATURE_DEVICE_REBUILD))
	return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));

	if (dsl_scan_resilvering(spa_get_dsl(spa)) \|\|
	dsl_scan_resilver_scheduled(spa_get_dsl(spa))) {
	return (spa_vdev_exit(spa, NULL, txg,
	ZFS_ERR_RESILVER_IN_PROGRESS));
	}
	} else {
	if (vdev_rebuild_active(rvd))
	return (spa_vdev_exit(spa, NULL, txg,
	ZFS_ERR_REBUILD_IN_PROGRESS));
	}

	if (spa->spa_vdev_removal != NULL) {
	return (spa_vdev_exit(spa, NULL, txg,
	ZFS_ERR_DEVRM_IN_PROGRESS));
	}

	if (oldvd == NULL)
	return (spa_vdev_exit(spa, NULL, txg, ENODEV));

	boolean_t raidz = oldvd->vdev_ops == &vdev_raidz_ops;

	if (raidz) {
	if (!spa_feature_is_enabled(spa, SPA_FEATURE_RAIDZ_EXPANSION))
	return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));

	/*
	* Can't expand a raidz while prior expand is in progress.
	*/
	if (spa->spa_raidz_expand != NULL) {
	return (spa_vdev_exit(spa, NULL, txg,
	ZFS_ERR_RAIDZ_EXPAND_IN_PROGRESS));
	}
	} else if (!oldvd->vdev_ops->vdev_op_leaf) {
	return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));
	}

	if (raidz)
	pvd = oldvd;
	else
	pvd = oldvd->vdev_parent;

	if (spa_config_parse(spa, &newrootvd, nvroot, NULL, 0,
	VDEV_ALLOC_ATTACH) != 0)
	return (spa_vdev_exit(spa, NULL, txg, EINVAL));

	if (newrootvd->vdev_children != 1)
	return (spa_vdev_exit(spa, newrootvd, txg, EINVAL));

	newvd = newrootvd->vdev_child[0];

	if (!newvd->vdev_ops->vdev_op_leaf)
	return (spa_vdev_exit(spa, newrootvd, txg, EINVAL));

	if ((error = vdev_create(newrootvd, txg, replacing)) != 0)
	return (spa_vdev_exit(spa, newrootvd, txg, error));

	/*
	* log, dedup and special vdevs should not be replaced by spares.
	*/
	if ((oldvd->vdev_top->vdev_alloc_bias != VDEV_BIAS_NONE \|\|
	oldvd->vdev_top->vdev_islog) && newvd->vdev_isspare) {
	return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
	}

	/*
	* A dRAID spare can only replace a child of its parent dRAID vdev.
	*/
	if (newvd->vdev_ops == &vdev_draid_spare_ops &&
	oldvd->vdev_top != vdev_draid_spare_get_parent(newvd)) {
	return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
	}

	if (rebuild) {
	/*
	* For rebuilds, the top vdev must support reconstruction
	* using only space maps. This means the only allowable
	* vdevs types are the root vdev, a mirror, or dRAID.
	*/
	tvd = pvd;
	if (pvd->vdev_top != NULL)
	tvd = pvd->vdev_top;

	if (tvd->vdev_ops != &vdev_mirror_ops &&
	tvd->vdev_ops != &vdev_root_ops &&
	tvd->vdev_ops != &vdev_draid_ops) {
	return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
	}
	}

	if (!replacing) {
	/*
	* For attach, the only allowable parent is a mirror or
	* the root vdev. A raidz vdev can be attached to, but
	* you cannot attach to a raidz child.
	*/
	if (pvd->vdev_ops != &vdev_mirror_ops &&
	pvd->vdev_ops != &vdev_root_ops &&
	!raidz)
	return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));

	pvops = &vdev_mirror_ops;
	} else {
	/*
	* Active hot spares can only be replaced by inactive hot
	* spares.
	*/
	if (pvd->vdev_ops == &vdev_spare_ops &&
	oldvd->vdev_isspare &&
	!spa_has_spare(spa, newvd->vdev_guid))
	return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));

	/*
	* If the source is a hot spare, and the parent isn't already a
	* spare, then we want to create a new hot spare. Otherwise, we
	* want to create a replacing vdev. The user is not allowed to
	* attach to a spared vdev child unless the 'isspare' state is
	* the same (spare replaces spare, non-spare replaces
	* non-spare).
	*/
	if (pvd->vdev_ops == &vdev_replacing_ops &&
	spa_version(spa) < SPA_VERSION_MULTI_REPLACE) {
	return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
	} else if (pvd->vdev_ops == &vdev_spare_ops &&
	newvd->vdev_isspare != oldvd->vdev_isspare) {
	return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));
	}

	if (newvd->vdev_isspare)
	pvops = &vdev_spare_ops;
	else
	pvops = &vdev_replacing_ops;
	}

	/*
	* Make sure the new device is big enough.
	*/
	vdev_t *min_vdev = raidz ? oldvd->vdev_child[0] : oldvd;
	if (newvd->vdev_asize < vdev_get_min_asize(min_vdev))
	return (spa_vdev_exit(spa, newrootvd, txg, EOVERFLOW));

	/*
	* The new device cannot have a higher alignment requirement
	* than the top-level vdev.
	*/
	if (newvd->vdev_ashift > oldvd->vdev_top->vdev_ashift)
	return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));

	/*
	* RAIDZ-expansion-specific checks.
	*/
	if (raidz) {
	if (vdev_raidz_attach_check(newvd) != 0)
	return (spa_vdev_exit(spa, newrootvd, txg, ENOTSUP));

	/*
	* Fail early if a child is not healthy or being replaced
	*/
	for (int i = 0; i < oldvd->vdev_children; i++) {
	if (vdev_is_dead(oldvd->vdev_child[i]) \|\|
	!oldvd->vdev_child[i]->vdev_ops->vdev_op_leaf) {
	return (spa_vdev_exit(spa, newrootvd, txg,
	ENXIO));
	}
	/* Also fail if reserved boot area is in-use */
	if (vdev_check_boot_reserve(spa, oldvd->vdev_child[i])
	!= 0) {
	return (spa_vdev_exit(spa, newrootvd, txg,
	EADDRINUSE));
	}
	}
	}

	if (raidz) {
	/*
	* Note: oldvdpath is freed by spa_strfree(), but
	* kmem_asprintf() is freed by kmem_strfree(), so we have to
	* move it to a spa_strdup-ed string.
	*/
	char *tmp = kmem_asprintf("raidz%u-%u",
	(uint_t)vdev_get_nparity(oldvd), (uint_t)oldvd->vdev_id);
	oldvdpath = spa_strdup(tmp);
	kmem_strfree(tmp);
	} else {
	oldvdpath = spa_strdup(oldvd->vdev_path);
	}
	newvdpath = spa_strdup(newvd->vdev_path);

	/*
	* If this is an in-place replacement, update oldvd's path and devid
	* to make it distinguishable from newvd, and unopenable from now on.
	*/
	if (strcmp(oldvdpath, newvdpath) == 0) {
	spa_strfree(oldvd->vdev_path);
	oldvd->vdev_path = kmem_alloc(strlen(newvdpath) + 5,
	KM_SLEEP);
	(void) sprintf(oldvd->vdev_path, "%s/old",
	newvdpath);
	if (oldvd->vdev_devid != NULL) {
	spa_strfree(oldvd->vdev_devid);
	oldvd->vdev_devid = NULL;
	}
	spa_strfree(oldvdpath);
	oldvdpath = spa_strdup(oldvd->vdev_path);
	}

	/*
	* If the parent is not a mirror, or if we're replacing, insert the new
	* mirror/replacing/spare vdev above oldvd.
	*/
	if (!raidz && pvd->vdev_ops != pvops) {
	pvd = vdev_add_parent(oldvd, pvops);
	ASSERT(pvd->vdev_ops == pvops);
	ASSERT(oldvd->vdev_parent == pvd);
	}

	ASSERT(pvd->vdev_top->vdev_parent == rvd);

	/*
	* Extract the new device from its root and add it to pvd.
	*/
	vdev_remove_child(newrootvd, newvd);
	newvd->vdev_id = pvd->vdev_children;
	newvd->vdev_crtxg = oldvd->vdev_crtxg;
	vdev_add_child(pvd, newvd);

	/*
	* Reevaluate the parent vdev state.
	*/
	vdev_propagate_state(pvd);

	tvd = newvd->vdev_top;
	ASSERT(pvd->vdev_top == tvd);
	ASSERT(tvd->vdev_parent == rvd);

	vdev_config_dirty(tvd);

	/*
	* Set newvd's DTL to [TXG_INITIAL, dtl_max_txg) so that we account
	* for any dmu_sync-ed blocks. It will propagate upward when
	* spa_vdev_exit() calls vdev_dtl_reassess().
	*/
	dtl_max_txg = txg + TXG_CONCURRENT_STATES;

	if (raidz) {
	/*
	* Wait for the youngest allocations and frees to sync,
	* and then wait for the deferral of those frees to finish.
	*/
	spa_vdev_config_exit(spa, NULL,
	txg + TXG_CONCURRENT_STATES + TXG_DEFER_SIZE, 0, FTAG);

	vdev_initialize_stop_all(tvd, VDEV_INITIALIZE_ACTIVE);
	vdev_trim_stop_all(tvd, VDEV_TRIM_ACTIVE);
	vdev_autotrim_stop_wait(tvd);

	dtl_max_txg = spa_vdev_config_enter(spa);

	tvd->vdev_rz_expanding = B_TRUE;

	vdev_dirty_leaves(tvd, VDD_DTL, dtl_max_txg);
	vdev_config_dirty(tvd);

	dmu_tx_t *tx = dmu_tx_create_assigned(spa->spa_dsl_pool,
	dtl_max_txg);
	dsl_sync_task_nowait(spa->spa_dsl_pool, vdev_raidz_attach_sync,
	newvd, tx);
	dmu_tx_commit(tx);
	} else {
	vdev_dtl_dirty(newvd, DTL_MISSING, TXG_INITIAL,
	dtl_max_txg - TXG_INITIAL);

	if (newvd->vdev_isspare) {
	spa_spare_activate(newvd);
	spa_event_notify(spa, newvd, NULL, ESC_ZFS_VDEV_SPARE);
	}

	newvd_isspare = newvd->vdev_isspare;

	/*
	* Mark newvd's DTL dirty in this txg.
	*/
	vdev_dirty(tvd, VDD_DTL, newvd, txg);

	/*
	* Schedule the resilver or rebuild to restart in the future.
	* We do this to ensure that dmu_sync-ed blocks have been
	* stitched into the respective datasets.
	*/
	if (rebuild) {
	newvd->vdev_rebuild_txg = txg;

	vdev_rebuild(tvd);
	} else {
	newvd->vdev_resilver_txg = txg;

	if (dsl_scan_resilvering(spa_get_dsl(spa)) &&
	spa_feature_is_enabled(spa,
	SPA_FEATURE_RESILVER_DEFER)) {
	vdev_defer_resilver(newvd);
	} else {
	dsl_scan_restart_resilver(spa->spa_dsl_pool,
	dtl_max_txg);
	}
	}
	}

	if (spa->spa_bootfs)
	spa_event_notify(spa, newvd, NULL, ESC_ZFS_BOOTFS_VDEV_ATTACH);

	spa_event_notify(spa, newvd, NULL, ESC_ZFS_VDEV_ATTACH);

	/*
	* Commit the config
	*/
	(void) spa_vdev_exit(spa, newrootvd, dtl_max_txg, 0);

	spa_history_log_internal(spa, "vdev attach", NULL,
	"%s vdev=%s %s vdev=%s",
	replacing && newvd_isspare ? "spare in" :
	replacing ? "replace" : "attach", newvdpath,
	replacing ? "for" : "to", oldvdpath);

	spa_strfree(oldvdpath);
	spa_strfree(newvdpath);

	return (0);
	}

	/*
	* Detach a device from a mirror or replacing vdev.
	*
	* If 'replace_done' is specified, only detach if the parent
	* is a replacing or a spare vdev.
	*/
	int
	spa_vdev_detach(spa_t *spa, uint64_t guid, uint64_t pguid, int replace_done)
	{
	uint64_t txg;
	int error;
	vdev_t *rvd __maybe_unused = spa->spa_root_vdev;
	vdev_t vd, pvd, cvd, tvd;
	boolean_t unspare = B_FALSE;
	uint64_t unspare_guid = 0;
	char *vdpath;

	ASSERT(spa_writeable(spa));

	txg = spa_vdev_detach_enter(spa, guid);

	vd = spa_lookup_by_guid(spa, guid, B_FALSE);

	/*
	* Besides being called directly from the userland through the
	* ioctl interface, spa_vdev_detach() can be potentially called
	* at the end of spa_vdev_resilver_done().
	*
	* In the regular case, when we have a checkpoint this shouldn't
	* happen as we never empty the DTLs of a vdev during the scrub
	* [see comment in dsl_scan_done()]. Thus spa_vdev_resilvering_done()
	* should never get here when we have a checkpoint.
	*
	* That said, even in a case when we checkpoint the pool exactly
	* as spa_vdev_resilver_done() calls this function everything
	* should be fine as the resilver will return right away.
	*/
	ASSERT(MUTEX_HELD(&spa_namespace_lock));
	if (spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT)) {
	error = (spa_has_checkpoint(spa)) ?
	ZFS_ERR_CHECKPOINT_EXISTS : ZFS_ERR_DISCARDING_CHECKPOINT;
	return (spa_vdev_exit(spa, NULL, txg, error));
	}

	if (vd == NULL)
	return (spa_vdev_exit(spa, NULL, txg, ENODEV));

	if (!vd->vdev_ops->vdev_op_leaf)
	return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));

	pvd = vd->vdev_parent;

	/*
	* If the parent/child relationship is not as expected, don't do it.
	* Consider M(A,R(B,C)) -- that is, a mirror of A with a replacing
	* vdev that's replacing B with C. The user's intent in replacing
	* is to go from M(A,B) to M(A,C). If the user decides to cancel
	* the replace by detaching C, the expected behavior is to end up
	* M(A,B). But suppose that right after deciding to detach C,
	* the replacement of B completes. We would have M(A,C), and then
	* ask to detach C, which would leave us with just A -- not what
	* the user wanted. To prevent this, we make sure that the
	* parent/child relationship hasn't changed -- in this example,
	* that C's parent is still the replacing vdev R.
	*/
	if (pvd->vdev_guid != pguid && pguid != 0)
	return (spa_vdev_exit(spa, NULL, txg, EBUSY));

	/*
	* Only 'replacing' or 'spare' vdevs can be replaced.
	*/
	if (replace_done && pvd->vdev_ops != &vdev_replacing_ops &&
	pvd->vdev_ops != &vdev_spare_ops)
	return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));

	ASSERT(pvd->vdev_ops != &vdev_spare_ops \|\|
	spa_version(spa) >= SPA_VERSION_SPARES);

	/*
	* Only mirror, replacing, and spare vdevs support detach.
	*/
	if (pvd->vdev_ops != &vdev_replacing_ops &&
	pvd->vdev_ops != &vdev_mirror_ops &&
	pvd->vdev_ops != &vdev_spare_ops)
	return (spa_vdev_exit(spa, NULL, txg, ENOTSUP));

	/*
	* If this device has the only valid copy of some data,
	* we cannot safely detach it.
	*/
	if (vdev_dtl_required(vd))
	return (spa_vdev_exit(spa, NULL, txg, EBUSY));

	ASSERT(pvd->vdev_children >= 2);

	/*
	* If we are detaching the second disk from a replacing vdev, then
	* check to see if we changed the original vdev's path to have "/old"
	* at the end in spa_vdev_attach(). If so, undo that change now.
	*/
	if (pvd->vdev_ops == &vdev_replacing_ops && vd->vdev_id > 0 &&
	vd->vdev_path != NULL) {
	size_t len = strlen(vd->vdev_path);

	for (int c = 0; c < pvd->vdev_children; c++) {
	cvd = pvd->vdev_child[c];

	if (cvd == vd \|\| cvd->vdev_path == NULL)
	continue;

	if (strncmp(cvd->vdev_path, vd->vdev_path, len) == 0 &&
	strcmp(cvd->vdev_path + len, "/old") == 0) {
	spa_strfree(cvd->vdev_path);
	cvd->vdev_path = spa_strdup(vd->vdev_path);
	break;
	}
	}
	}

	/*
	* If we are detaching the original disk from a normal spare, then it
	* implies that the spare should become a real disk, and be removed
	* from the active spare list for the pool. dRAID spares on the
	* other hand are coupled to the pool and thus should never be removed
	* from the spares list.
	*/
	if (pvd->vdev_ops == &vdev_spare_ops && vd->vdev_id == 0) {
	vdev_t *last_cvd = pvd->vdev_child[pvd->vdev_children - 1];

	if (last_cvd->vdev_isspare &&
	last_cvd->vdev_ops != &vdev_draid_spare_ops) {
	unspare = B_TRUE;
	}
	}

	/*
	* Erase the disk labels so the disk can be used for other things.
	* This must be done after all other error cases are handled,
	* but before we disembowel vd (so we can still do I/O to it).
	* But if we can't do it, don't treat the error as fatal --
	* it may be that the unwritability of the disk is the reason
	* it's being detached!
	*/
	(void) vdev_label_init(vd, 0, VDEV_LABEL_REMOVE);

	/*
	* Remove vd from its parent and compact the parent's children.
	*/
	vdev_remove_child(pvd, vd);
	vdev_compact_children(pvd);

	/*
	* Remember one of the remaining children so we can get tvd below.
	*/
	cvd = pvd->vdev_child[pvd->vdev_children - 1];

	/*
	* If we need to remove the remaining child from the list of hot spares,
	* do it now, marking the vdev as no longer a spare in the process.
	* We must do this before vdev_remove_parent(), because that can
	* change the GUID if it creates a new toplevel GUID. For a similar
	* reason, we must remove the spare now, in the same txg as the detach;
	* otherwise someone could attach a new sibling, change the GUID, and
	* the subsequent attempt to spa_vdev_remove(unspare_guid) would fail.
	*/
	if (unspare) {
	ASSERT(cvd->vdev_isspare);
	spa_spare_remove(cvd);
	unspare_guid = cvd->vdev_guid;
	(void) spa_vdev_remove(spa, unspare_guid, B_TRUE);
	cvd->vdev_unspare = B_TRUE;
	}

	/*
	* If the parent mirror/replacing vdev only has one child,
	* the parent is no longer needed. Remove it from the tree.
	*/
	if (pvd->vdev_children == 1) {
	if (pvd->vdev_ops == &vdev_spare_ops)
	cvd->vdev_unspare = B_FALSE;
	vdev_remove_parent(cvd);
	}

	/*
	* We don't set tvd until now because the parent we just removed
	* may have been the previous top-level vdev.
	*/
	tvd = cvd->vdev_top;
	ASSERT(tvd->vdev_parent == rvd);

	/*
	* Reevaluate the parent vdev state.
	*/
	vdev_propagate_state(cvd);

	/*
	* If the 'autoexpand' property is set on the pool then automatically
	* try to expand the size of the pool. For example if the device we
	* just detached was smaller than the others, it may be possible to
	* add metaslabs (i.e. grow the pool). We need to reopen the vdev
	* first so that we can obtain the updated sizes of the leaf vdevs.
	*/
	if (spa->spa_autoexpand) {
	vdev_reopen(tvd);
	vdev_expand(tvd, txg);
	}

	vdev_config_dirty(tvd);

	/*
	* Mark vd's DTL as dirty in this txg. vdev_dtl_sync() will see that
	* vd->vdev_detached is set and free vd's DTL object in syncing context.
	* But first make sure we're not on any other txg's DTL list, to
	* prevent vd from being accessed after it's freed.
	*/
	vdpath = spa_strdup(vd->vdev_path ? vd->vdev_path : "none");
	for (int t = 0; t < TXG_SIZE; t++)
	(void) txg_list_remove_this(&tvd->vdev_dtl_list, vd, t);
	vd->vdev_detached = B_TRUE;
	vdev_dirty(tvd, VDD_DTL, vd, txg);

	spa_event_notify(spa, vd, NULL, ESC_ZFS_VDEV_REMOVE);
	spa_notify_waiters(spa);

	/* hang on to the spa before we release the lock */
	spa_open_ref(spa, FTAG);

	error = spa_vdev_exit(spa, vd, txg, 0);

	spa_history_log_internal(spa, "detach", NULL,
	"vdev=%s", vdpath);
	spa_strfree(vdpath);

	/*
	* If this was the removal of the original device in a hot spare vdev,
	* then we want to go through and remove the device from the hot spare
	* list of every other pool.
	*/
	if (unspare) {
	spa_t *altspa = NULL;

	mutex_enter(&spa_namespace_lock);
	while ((altspa = spa_next(altspa)) != NULL) {
	if (altspa->spa_state != POOL_STATE_ACTIVE \|\|
	altspa == spa)
	continue;

	spa_open_ref(altspa, FTAG);
	mutex_exit(&spa_namespace_lock);
	(void) spa_vdev_remove(altspa, unspare_guid, B_TRUE);
	mutex_enter(&spa_namespace_lock);
	spa_close(altspa, FTAG);
	}
	mutex_exit(&spa_namespace_lock);

	/* search the rest of the vdevs for spares to remove */
	spa_vdev_resilver_done(spa);
	}

	/* all done with the spa; OK to release */
	mutex_enter(&spa_namespace_lock);
	spa_close(spa, FTAG);
	mutex_exit(&spa_namespace_lock);

	return (error);
	}

	static int
	spa_vdev_initialize_impl(spa_t *spa, uint64_t guid, uint64_t cmd_type,
	list_t *vd_list)
	{
	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	spa_config_enter(spa, SCL_CONFIG \| SCL_STATE, FTAG, RW_READER);

	/* Look up vdev and ensure it's a leaf. */
	vdev_t *vd = spa_lookup_by_guid(spa, guid, B_FALSE);
	if (vd == NULL \|\| vd->vdev_detached) {
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	return (SET_ERROR(ENODEV));
	} else if (!vd->vdev_ops->vdev_op_leaf \|\| !vdev_is_concrete(vd)) {
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	return (SET_ERROR(EINVAL));
	} else if (!vdev_writeable(vd)) {
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	return (SET_ERROR(EROFS));
	}
	mutex_enter(&vd->vdev_initialize_lock);
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);

	/*
	* When we activate an initialize action we check to see
	* if the vdev_initialize_thread is NULL. We do this instead
	* of using the vdev_initialize_state since there might be
	* a previous initialization process which has completed but
	* the thread is not exited.
	*/
	if (cmd_type == POOL_INITIALIZE_START &&
	(vd->vdev_initialize_thread != NULL \|\|
	vd->vdev_top->vdev_removing \|\| vd->vdev_top->vdev_rz_expanding)) {
	mutex_exit(&vd->vdev_initialize_lock);
	return (SET_ERROR(EBUSY));
	} else if (cmd_type == POOL_INITIALIZE_CANCEL &&
	(vd->vdev_initialize_state != VDEV_INITIALIZE_ACTIVE &&
	vd->vdev_initialize_state != VDEV_INITIALIZE_SUSPENDED)) {
	mutex_exit(&vd->vdev_initialize_lock);
	return (SET_ERROR(ESRCH));
	} else if (cmd_type == POOL_INITIALIZE_SUSPEND &&
	vd->vdev_initialize_state != VDEV_INITIALIZE_ACTIVE) {
	mutex_exit(&vd->vdev_initialize_lock);
	return (SET_ERROR(ESRCH));
	} else if (cmd_type == POOL_INITIALIZE_UNINIT &&
	vd->vdev_initialize_thread != NULL) {
	mutex_exit(&vd->vdev_initialize_lock);
	return (SET_ERROR(EBUSY));
	}

	switch (cmd_type) {
	case POOL_INITIALIZE_START:
	vdev_initialize(vd);
	break;
	case POOL_INITIALIZE_CANCEL:
	vdev_initialize_stop(vd, VDEV_INITIALIZE_CANCELED, vd_list);
	break;
	case POOL_INITIALIZE_SUSPEND:
	vdev_initialize_stop(vd, VDEV_INITIALIZE_SUSPENDED, vd_list);
	break;
	case POOL_INITIALIZE_UNINIT:
	vdev_uninitialize(vd);
	break;
	default:
	panic("invalid cmd_type %llu", (unsigned long long)cmd_type);
	}
	mutex_exit(&vd->vdev_initialize_lock);

	return (0);
	}

	int
	spa_vdev_initialize(spa_t spa, nvlist_t nv, uint64_t cmd_type,
	nvlist_t *vdev_errlist)
	{
	int total_errors = 0;
	list_t vd_list;

	list_create(&vd_list, sizeof (vdev_t),
	offsetof(vdev_t, vdev_initialize_node));

	/*
	* We hold the namespace lock through the whole function
	* to prevent any changes to the pool while we're starting or
	* stopping initialization. The config and state locks are held so that
	* we can properly assess the vdev state before we commit to
	* the initializing operation.
	*/
	mutex_enter(&spa_namespace_lock);

	for (nvpair_t *pair = nvlist_next_nvpair(nv, NULL);
	pair != NULL; pair = nvlist_next_nvpair(nv, pair)) {
	uint64_t vdev_guid = fnvpair_value_uint64(pair);

	int error = spa_vdev_initialize_impl(spa, vdev_guid, cmd_type,
	&vd_list);
	if (error != 0) {
	char guid_as_str[MAXNAMELEN];

	(void) snprintf(guid_as_str, sizeof (guid_as_str),
	"%llu", (unsigned long long)vdev_guid);
	fnvlist_add_int64(vdev_errlist, guid_as_str, error);
	total_errors++;
	}
	}

	/* Wait for all initialize threads to stop. */
	vdev_initialize_stop_wait(spa, &vd_list);

	/* Sync out the initializing state */
	txg_wait_synced(spa->spa_dsl_pool, 0);
	mutex_exit(&spa_namespace_lock);

	list_destroy(&vd_list);

	return (total_errors);
	}

	static int
	spa_vdev_trim_impl(spa_t *spa, uint64_t guid, uint64_t cmd_type,
	uint64_t rate, boolean_t partial, boolean_t secure, list_t *vd_list)
	{
	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	spa_config_enter(spa, SCL_CONFIG \| SCL_STATE, FTAG, RW_READER);

	/* Look up vdev and ensure it's a leaf. */
	vdev_t *vd = spa_lookup_by_guid(spa, guid, B_FALSE);
	if (vd == NULL \|\| vd->vdev_detached) {
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	return (SET_ERROR(ENODEV));
	} else if (!vd->vdev_ops->vdev_op_leaf \|\| !vdev_is_concrete(vd)) {
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	return (SET_ERROR(EINVAL));
	} else if (!vdev_writeable(vd)) {
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	return (SET_ERROR(EROFS));
	} else if (!vd->vdev_has_trim) {
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	return (SET_ERROR(EOPNOTSUPP));
	} else if (secure && !vd->vdev_has_securetrim) {
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	return (SET_ERROR(EOPNOTSUPP));
	}
	mutex_enter(&vd->vdev_trim_lock);
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);

	/*
	* When we activate a TRIM action we check to see if the
	* vdev_trim_thread is NULL. We do this instead of using the
	* vdev_trim_state since there might be a previous TRIM process
	* which has completed but the thread is not exited.
	*/
	if (cmd_type == POOL_TRIM_START &&
	(vd->vdev_trim_thread != NULL \|\| vd->vdev_top->vdev_removing \|\|
	vd->vdev_top->vdev_rz_expanding)) {
	mutex_exit(&vd->vdev_trim_lock);
	return (SET_ERROR(EBUSY));
	} else if (cmd_type == POOL_TRIM_CANCEL &&
	(vd->vdev_trim_state != VDEV_TRIM_ACTIVE &&
	vd->vdev_trim_state != VDEV_TRIM_SUSPENDED)) {
	mutex_exit(&vd->vdev_trim_lock);
	return (SET_ERROR(ESRCH));
	} else if (cmd_type == POOL_TRIM_SUSPEND &&
	vd->vdev_trim_state != VDEV_TRIM_ACTIVE) {
	mutex_exit(&vd->vdev_trim_lock);
	return (SET_ERROR(ESRCH));
	}

	switch (cmd_type) {
	case POOL_TRIM_START:
	vdev_trim(vd, rate, partial, secure);
	break;
	case POOL_TRIM_CANCEL:
	vdev_trim_stop(vd, VDEV_TRIM_CANCELED, vd_list);
	break;
	case POOL_TRIM_SUSPEND:
	vdev_trim_stop(vd, VDEV_TRIM_SUSPENDED, vd_list);
	break;
	default:
	panic("invalid cmd_type %llu", (unsigned long long)cmd_type);
	}
	mutex_exit(&vd->vdev_trim_lock);

	return (0);
	}

	/*
	* Initiates a manual TRIM for the requested vdevs. This kicks off individual
	* TRIM threads for each child vdev. These threads pass over all of the free
	* space in the vdev's metaslabs and issues TRIM commands for that space.
	*/
	int
	spa_vdev_trim(spa_t spa, nvlist_t nv, uint64_t cmd_type, uint64_t rate,
	boolean_t partial, boolean_t secure, nvlist_t *vdev_errlist)
	{
	int total_errors = 0;
	list_t vd_list;

	list_create(&vd_list, sizeof (vdev_t),
	offsetof(vdev_t, vdev_trim_node));

	/*
	* We hold the namespace lock through the whole function
	* to prevent any changes to the pool while we're starting or
	* stopping TRIM. The config and state locks are held so that
	* we can properly assess the vdev state before we commit to
	* the TRIM operation.
	*/
	mutex_enter(&spa_namespace_lock);

	for (nvpair_t *pair = nvlist_next_nvpair(nv, NULL);
	pair != NULL; pair = nvlist_next_nvpair(nv, pair)) {
	uint64_t vdev_guid = fnvpair_value_uint64(pair);

	int error = spa_vdev_trim_impl(spa, vdev_guid, cmd_type,
	rate, partial, secure, &vd_list);
	if (error != 0) {
	char guid_as_str[MAXNAMELEN];

	(void) snprintf(guid_as_str, sizeof (guid_as_str),
	"%llu", (unsigned long long)vdev_guid);
	fnvlist_add_int64(vdev_errlist, guid_as_str, error);
	total_errors++;
	}
	}

	/* Wait for all TRIM threads to stop. */
	vdev_trim_stop_wait(spa, &vd_list);

	/* Sync out the TRIM state */
	txg_wait_synced(spa->spa_dsl_pool, 0);
	mutex_exit(&spa_namespace_lock);

	list_destroy(&vd_list);

	return (total_errors);
	}

	/*
	* Split a set of devices from their mirrors, and create a new pool from them.
	*/
	int
	spa_vdev_split_mirror(spa_t spa, const char newname, nvlist_t *config,
	nvlist_t *props, boolean_t exp)
	{
	int error = 0;
	uint64_t txg, *glist;
	spa_t *newspa;
	uint_t c, children, lastlog;
	nvlist_t *child, nvl, *tmp;
	dmu_tx_t *tx;
	const char *altroot = NULL;
	vdev_t rvd, vml = NULL; / vdev modify list */
	boolean_t activate_slog;

	ASSERT(spa_writeable(spa));

	txg = spa_vdev_enter(spa);

	ASSERT(MUTEX_HELD(&spa_namespace_lock));
	if (spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT)) {
	error = (spa_has_checkpoint(spa)) ?
	ZFS_ERR_CHECKPOINT_EXISTS : ZFS_ERR_DISCARDING_CHECKPOINT;
	return (spa_vdev_exit(spa, NULL, txg, error));
	}

	/* clear the log and flush everything up to now */
	activate_slog = spa_passivate_log(spa);
	(void) spa_vdev_config_exit(spa, NULL, txg, 0, FTAG);
	error = spa_reset_logs(spa);
	txg = spa_vdev_config_enter(spa);

	if (activate_slog)
	spa_activate_log(spa);

	if (error != 0)
	return (spa_vdev_exit(spa, NULL, txg, error));

	/* check new spa name before going any further */
	if (spa_lookup(newname) != NULL)
	return (spa_vdev_exit(spa, NULL, txg, EEXIST));

	/*
	* scan through all the children to ensure they're all mirrors
	*/
	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &nvl) != 0 \|\|
	nvlist_lookup_nvlist_array(nvl, ZPOOL_CONFIG_CHILDREN, &child,
	&children) != 0)
	return (spa_vdev_exit(spa, NULL, txg, EINVAL));

	/* first, check to ensure we've got the right child count */
	rvd = spa->spa_root_vdev;
	lastlog = 0;
	for (c = 0; c < rvd->vdev_children; c++) {
	vdev_t *vd = rvd->vdev_child[c];

	/* don't count the holes & logs as children */
	if (vd->vdev_islog \|\| (vd->vdev_ops != &vdev_indirect_ops &&
	!vdev_is_concrete(vd))) {
	if (lastlog == 0)
	lastlog = c;
	continue;
	}

	lastlog = 0;
	}
	if (children != (lastlog != 0 ? lastlog : rvd->vdev_children))
	return (spa_vdev_exit(spa, NULL, txg, EINVAL));

	/* next, ensure no spare or cache devices are part of the split */
	if (nvlist_lookup_nvlist(nvl, ZPOOL_CONFIG_SPARES, &tmp) == 0 \|\|
	nvlist_lookup_nvlist(nvl, ZPOOL_CONFIG_L2CACHE, &tmp) == 0)
	return (spa_vdev_exit(spa, NULL, txg, EINVAL));

	vml = kmem_zalloc(children * sizeof (vdev_t *), KM_SLEEP);
	glist = kmem_zalloc(children * sizeof (uint64_t), KM_SLEEP);

	/* then, loop over each vdev and validate it */
	for (c = 0; c < children; c++) {
	uint64_t is_hole = 0;

	(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
	&is_hole);

	if (is_hole != 0) {
	if (spa->spa_root_vdev->vdev_child[c]->vdev_ishole \|\|
	spa->spa_root_vdev->vdev_child[c]->vdev_islog) {
	continue;
	} else {
	error = SET_ERROR(EINVAL);
	break;
	}
	}

	/* deal with indirect vdevs */
	if (spa->spa_root_vdev->vdev_child[c]->vdev_ops ==
	&vdev_indirect_ops)
	continue;

	/* which disk is going to be split? */
	if (nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_GUID,
	&glist[c]) != 0) {
	error = SET_ERROR(EINVAL);
	break;
	}

	/* look it up in the spa */
	vml[c] = spa_lookup_by_guid(spa, glist[c], B_FALSE);
	if (vml[c] == NULL) {
	error = SET_ERROR(ENODEV);
	break;
	}

	/* make sure there's nothing stopping the split */
	if (vml[c]->vdev_parent->vdev_ops != &vdev_mirror_ops \|\|
	vml[c]->vdev_islog \|\|
	!vdev_is_concrete(vml[c]) \|\|
	vml[c]->vdev_isspare \|\|
	vml[c]->vdev_isl2cache \|\|
	!vdev_writeable(vml[c]) \|\|
	vml[c]->vdev_children != 0 \|\|
	vml[c]->vdev_state != VDEV_STATE_HEALTHY \|\|
	c != spa->spa_root_vdev->vdev_child[c]->vdev_id) {
	error = SET_ERROR(EINVAL);
	break;
	}

	if (vdev_dtl_required(vml[c]) \|\|
	vdev_resilver_needed(vml[c], NULL, NULL)) {
	error = SET_ERROR(EBUSY);
	break;
	}

	/* we need certain info from the top level */
	fnvlist_add_uint64(child[c], ZPOOL_CONFIG_METASLAB_ARRAY,
	vml[c]->vdev_top->vdev_ms_array);
	fnvlist_add_uint64(child[c], ZPOOL_CONFIG_METASLAB_SHIFT,
	vml[c]->vdev_top->vdev_ms_shift);
	fnvlist_add_uint64(child[c], ZPOOL_CONFIG_ASIZE,
	vml[c]->vdev_top->vdev_asize);
	fnvlist_add_uint64(child[c], ZPOOL_CONFIG_ASHIFT,
	vml[c]->vdev_top->vdev_ashift);

	/* transfer per-vdev ZAPs */
	ASSERT3U(vml[c]->vdev_leaf_zap, !=, 0);
	VERIFY0(nvlist_add_uint64(child[c],
	ZPOOL_CONFIG_VDEV_LEAF_ZAP, vml[c]->vdev_leaf_zap));

	ASSERT3U(vml[c]->vdev_top->vdev_top_zap, !=, 0);
	VERIFY0(nvlist_add_uint64(child[c],
	ZPOOL_CONFIG_VDEV_TOP_ZAP,
	vml[c]->vdev_parent->vdev_top_zap));
	}

	if (error != 0) {
	kmem_free(vml, children * sizeof (vdev_t *));
	kmem_free(glist, children * sizeof (uint64_t));
	return (spa_vdev_exit(spa, NULL, txg, error));
	}

	/* stop writers from using the disks */
	for (c = 0; c < children; c++) {
	if (vml[c] != NULL)
	vml[c]->vdev_offline = B_TRUE;
	}
	vdev_reopen(spa->spa_root_vdev);

	/*
	* Temporarily record the splitting vdevs in the spa config. This
	* will disappear once the config is regenerated.
	*/
	nvl = fnvlist_alloc();
	fnvlist_add_uint64_array(nvl, ZPOOL_CONFIG_SPLIT_LIST, glist, children);
	kmem_free(glist, children * sizeof (uint64_t));

	mutex_enter(&spa->spa_props_lock);
	fnvlist_add_nvlist(spa->spa_config, ZPOOL_CONFIG_SPLIT, nvl);
	mutex_exit(&spa->spa_props_lock);
	spa->spa_config_splitting = nvl;
	vdev_config_dirty(spa->spa_root_vdev);

	/* configure and create the new pool */
	fnvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME, newname);
	fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE,
	exp ? POOL_STATE_EXPORTED : POOL_STATE_ACTIVE);
	fnvlist_add_uint64(config, ZPOOL_CONFIG_VERSION, spa_version(spa));
	fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG, spa->spa_config_txg);
	fnvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID,
	spa_generate_guid(NULL));
	VERIFY0(nvlist_add_boolean(config, ZPOOL_CONFIG_HAS_PER_VDEV_ZAPS));
	(void) nvlist_lookup_string(props,
	zpool_prop_to_name(ZPOOL_PROP_ALTROOT), &altroot);

	/* add the new pool to the namespace */
	newspa = spa_add(newname, config, altroot);
	newspa->spa_avz_action = AVZ_ACTION_REBUILD;
	newspa->spa_config_txg = spa->spa_config_txg;
	spa_set_log_state(newspa, SPA_LOG_CLEAR);

	/* release the spa config lock, retaining the namespace lock */
	spa_vdev_config_exit(spa, NULL, txg, 0, FTAG);

	if (zio_injection_enabled)
	zio_handle_panic_injection(spa, FTAG, 1);

	spa_activate(newspa, spa_mode_global);
	spa_async_suspend(newspa);

	/*
	* Temporarily stop the initializing and TRIM activity. We set the
	* state to ACTIVE so that we know to resume initializing or TRIM
	* once the split has completed.
	*/
	list_t vd_initialize_list;
	list_create(&vd_initialize_list, sizeof (vdev_t),
	offsetof(vdev_t, vdev_initialize_node));

	list_t vd_trim_list;
	list_create(&vd_trim_list, sizeof (vdev_t),
	offsetof(vdev_t, vdev_trim_node));

	for (c = 0; c < children; c++) {
	if (vml[c] != NULL && vml[c]->vdev_ops != &vdev_indirect_ops) {
	mutex_enter(&vml[c]->vdev_initialize_lock);
	vdev_initialize_stop(vml[c],
	VDEV_INITIALIZE_ACTIVE, &vd_initialize_list);
	mutex_exit(&vml[c]->vdev_initialize_lock);

	mutex_enter(&vml[c]->vdev_trim_lock);
	vdev_trim_stop(vml[c], VDEV_TRIM_ACTIVE, &vd_trim_list);
	mutex_exit(&vml[c]->vdev_trim_lock);
	}
	}

	vdev_initialize_stop_wait(spa, &vd_initialize_list);
	vdev_trim_stop_wait(spa, &vd_trim_list);

	list_destroy(&vd_initialize_list);
	list_destroy(&vd_trim_list);

	newspa->spa_config_source = SPA_CONFIG_SRC_SPLIT;
	newspa->spa_is_splitting = B_TRUE;

	/* create the new pool from the disks of the original pool */
	error = spa_load(newspa, SPA_LOAD_IMPORT, SPA_IMPORT_ASSEMBLE);
	if (error)
	goto out;

	/* if that worked, generate a real config for the new pool */
	if (newspa->spa_root_vdev != NULL) {
	newspa->spa_config_splitting = fnvlist_alloc();
	fnvlist_add_uint64(newspa->spa_config_splitting,
	ZPOOL_CONFIG_SPLIT_GUID, spa_guid(spa));
	spa_config_set(newspa, spa_config_generate(newspa, NULL, -1ULL,
	B_TRUE));
	}

	/* set the props */
	if (props != NULL) {
	spa_configfile_set(newspa, props, B_FALSE);
	error = spa_prop_set(newspa, props);
	if (error)
	goto out;
	}

	/* flush everything */
	txg = spa_vdev_config_enter(newspa);
	vdev_config_dirty(newspa->spa_root_vdev);
	(void) spa_vdev_config_exit(newspa, NULL, txg, 0, FTAG);

	if (zio_injection_enabled)
	zio_handle_panic_injection(spa, FTAG, 2);

	spa_async_resume(newspa);

	/* finally, update the original pool's config */
	txg = spa_vdev_config_enter(spa);
	tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error != 0)
	dmu_tx_abort(tx);
	for (c = 0; c < children; c++) {
	if (vml[c] != NULL && vml[c]->vdev_ops != &vdev_indirect_ops) {
	vdev_t *tvd = vml[c]->vdev_top;

	/*
	* Need to be sure the detachable VDEV is not
	* on any other txg's DTL list to prevent it
	* from being accessed after it's freed.
	*/
	for (int t = 0; t < TXG_SIZE; t++) {
	(void) txg_list_remove_this(
	&tvd->vdev_dtl_list, vml[c], t);
	}

	vdev_split(vml[c]);
	if (error == 0)
	spa_history_log_internal(spa, "detach", tx,
	"vdev=%s", vml[c]->vdev_path);

	vdev_free(vml[c]);
	}
	}
	spa->spa_avz_action = AVZ_ACTION_REBUILD;
	vdev_config_dirty(spa->spa_root_vdev);
	spa->spa_config_splitting = NULL;
	nvlist_free(nvl);
	if (error == 0)
	dmu_tx_commit(tx);
	(void) spa_vdev_exit(spa, NULL, txg, 0);

	if (zio_injection_enabled)
	zio_handle_panic_injection(spa, FTAG, 3);

	/* split is complete; log a history record */
	spa_history_log_internal(newspa, "split", NULL,
	"from pool %s", spa_name(spa));

	newspa->spa_is_splitting = B_FALSE;
	kmem_free(vml, children * sizeof (vdev_t *));

	/* if we're not going to mount the filesystems in userland, export */
	if (exp)
	error = spa_export_common(newname, POOL_STATE_EXPORTED, NULL,
	B_FALSE, B_FALSE);

	return (error);

	out:
	spa_unload(newspa);
	spa_deactivate(newspa);
	spa_remove(newspa);

	txg = spa_vdev_config_enter(spa);

	/* re-online all offlined disks */
	for (c = 0; c < children; c++) {
	if (vml[c] != NULL)
	vml[c]->vdev_offline = B_FALSE;
	}

	/* restart initializing or trimming disks as necessary */
	spa_async_request(spa, SPA_ASYNC_INITIALIZE_RESTART);
	spa_async_request(spa, SPA_ASYNC_TRIM_RESTART);
	spa_async_request(spa, SPA_ASYNC_AUTOTRIM_RESTART);

	vdev_reopen(spa->spa_root_vdev);

	nvlist_free(spa->spa_config_splitting);
	spa->spa_config_splitting = NULL;
	(void) spa_vdev_exit(spa, NULL, txg, error);

	kmem_free(vml, children * sizeof (vdev_t *));
	return (error);
	}

	/*
	* Find any device that's done replacing, or a vdev marked 'unspare' that's
	* currently spared, so we can detach it.
	*/
	static vdev_t *
	spa_vdev_resilver_done_hunt(vdev_t *vd)
	{
	vdev_t newvd, oldvd;

	for (int c = 0; c < vd->vdev_children; c++) {
	oldvd = spa_vdev_resilver_done_hunt(vd->vdev_child[c]);
	if (oldvd != NULL)
	return (oldvd);
	}

	/*
	* Check for a completed replacement. We always consider the first
	* vdev in the list to be the oldest vdev, and the last one to be
	* the newest (see spa_vdev_attach() for how that works). In
	* the case where the newest vdev is faulted, we will not automatically
	* remove it after a resilver completes. This is OK as it will require
	* user intervention to determine which disk the admin wishes to keep.
	*/
	if (vd->vdev_ops == &vdev_replacing_ops) {
	ASSERT(vd->vdev_children > 1);

	newvd = vd->vdev_child[vd->vdev_children - 1];
	oldvd = vd->vdev_child[0];

	if (vdev_dtl_empty(newvd, DTL_MISSING) &&
	vdev_dtl_empty(newvd, DTL_OUTAGE) &&
	!vdev_dtl_required(oldvd))
	return (oldvd);
	}

	/*
	* Check for a completed resilver with the 'unspare' flag set.
	* Also potentially update faulted state.
	*/
	if (vd->vdev_ops == &vdev_spare_ops) {
	vdev_t *first = vd->vdev_child[0];
	vdev_t *last = vd->vdev_child[vd->vdev_children - 1];

	if (last->vdev_unspare) {
	oldvd = first;
	newvd = last;
	} else if (first->vdev_unspare) {
	oldvd = last;
	newvd = first;
	} else {
	oldvd = NULL;
	}

	if (oldvd != NULL &&
	vdev_dtl_empty(newvd, DTL_MISSING) &&
	vdev_dtl_empty(newvd, DTL_OUTAGE) &&
	!vdev_dtl_required(oldvd))
	return (oldvd);

	vdev_propagate_state(vd);

	/*
	* If there are more than two spares attached to a disk,
	* and those spares are not required, then we want to
	* attempt to free them up now so that they can be used
	* by other pools. Once we're back down to a single
	* disk+spare, we stop removing them.
	*/
	if (vd->vdev_children > 2) {
	newvd = vd->vdev_child[1];

	if (newvd->vdev_isspare && last->vdev_isspare &&
	vdev_dtl_empty(last, DTL_MISSING) &&
	vdev_dtl_empty(last, DTL_OUTAGE) &&
	!vdev_dtl_required(newvd))
	return (newvd);
	}
	}

	return (NULL);
	}

	static void
	spa_vdev_resilver_done(spa_t *spa)
	{
	vdev_t vd, pvd, *ppvd;
	uint64_t guid, sguid, pguid, ppguid;

	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);

	while ((vd = spa_vdev_resilver_done_hunt(spa->spa_root_vdev)) != NULL) {
	pvd = vd->vdev_parent;
	ppvd = pvd->vdev_parent;
	guid = vd->vdev_guid;
	pguid = pvd->vdev_guid;
	ppguid = ppvd->vdev_guid;
	sguid = 0;
	/*
	* If we have just finished replacing a hot spared device, then
	* we need to detach the parent's first child (the original hot
	* spare) as well.
	*/
	if (ppvd->vdev_ops == &vdev_spare_ops && pvd->vdev_id == 0 &&
	ppvd->vdev_children == 2) {
	ASSERT(pvd->vdev_ops == &vdev_replacing_ops);
	sguid = ppvd->vdev_child[1]->vdev_guid;
	}
	ASSERT(vd->vdev_resilver_txg == 0 \|\| !vdev_dtl_required(vd));

	spa_config_exit(spa, SCL_ALL, FTAG);
	if (spa_vdev_detach(spa, guid, pguid, B_TRUE) != 0)
	return;
	if (sguid && spa_vdev_detach(spa, sguid, ppguid, B_TRUE) != 0)
	return;
	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	}

	spa_config_exit(spa, SCL_ALL, FTAG);

	/*
	* If a detach was not performed above replace waiters will not have
	* been notified. In which case we must do so now.
	*/
	spa_notify_waiters(spa);
	}

	/*
	* Update the stored path or FRU for this vdev.
	*/
	static int
	spa_vdev_set_common(spa_t spa, uint64_t guid, const char value,
	boolean_t ispath)
	{
	vdev_t *vd;
	boolean_t sync = B_FALSE;

	ASSERT(spa_writeable(spa));

	spa_vdev_state_enter(spa, SCL_ALL);

	if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL)
	return (spa_vdev_state_exit(spa, NULL, ENOENT));

	if (!vd->vdev_ops->vdev_op_leaf)
	return (spa_vdev_state_exit(spa, NULL, ENOTSUP));

	if (ispath) {
	if (strcmp(value, vd->vdev_path) != 0) {
	spa_strfree(vd->vdev_path);
	vd->vdev_path = spa_strdup(value);
	sync = B_TRUE;
	}
	} else {
	if (vd->vdev_fru == NULL) {
	vd->vdev_fru = spa_strdup(value);
	sync = B_TRUE;
	} else if (strcmp(value, vd->vdev_fru) != 0) {
	spa_strfree(vd->vdev_fru);
	vd->vdev_fru = spa_strdup(value);
	sync = B_TRUE;
	}
	}

	return (spa_vdev_state_exit(spa, sync ? vd : NULL, 0));
	}

	int
	spa_vdev_setpath(spa_t spa, uint64_t guid, const char newpath)
	{
	return (spa_vdev_set_common(spa, guid, newpath, B_TRUE));
	}

	int
	spa_vdev_setfru(spa_t spa, uint64_t guid, const char newfru)
	{
	return (spa_vdev_set_common(spa, guid, newfru, B_FALSE));
	}

	/*
	* ==========================================================================
	* SPA Scanning
	* ==========================================================================
	*/
	int
	spa_scrub_pause_resume(spa_t *spa, pool_scrub_cmd_t cmd)
	{
	ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0);

	if (dsl_scan_resilvering(spa->spa_dsl_pool))
	return (SET_ERROR(EBUSY));

	return (dsl_scrub_set_pause_resume(spa->spa_dsl_pool, cmd));
	}

	int
	spa_scan_stop(spa_t *spa)
	{
	ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0);
	if (dsl_scan_resilvering(spa->spa_dsl_pool))
	return (SET_ERROR(EBUSY));

	return (dsl_scan_cancel(spa->spa_dsl_pool));
	}

	int
	spa_scan(spa_t *spa, pool_scan_func_t func)
	{
	ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == 0);

	if (func >= POOL_SCAN_FUNCS \|\| func == POOL_SCAN_NONE)
	return (SET_ERROR(ENOTSUP));

	if (func == POOL_SCAN_RESILVER &&
	!spa_feature_is_enabled(spa, SPA_FEATURE_RESILVER_DEFER))
	return (SET_ERROR(ENOTSUP));

	/*
	* If a resilver was requested, but there is no DTL on a
	* writeable leaf device, we have nothing to do.
	*/
	if (func == POOL_SCAN_RESILVER &&
	!vdev_resilver_needed(spa->spa_root_vdev, NULL, NULL)) {
	spa_async_request(spa, SPA_ASYNC_RESILVER_DONE);
	return (0);
	}

	if (func == POOL_SCAN_ERRORSCRUB &&
	!spa_feature_is_enabled(spa, SPA_FEATURE_HEAD_ERRLOG))
	return (SET_ERROR(ENOTSUP));

	return (dsl_scan(spa->spa_dsl_pool, func));
	}

	/*
	* ==========================================================================
	* SPA async task processing
	* ==========================================================================
	*/

	static void
	spa_async_remove(spa_t spa, vdev_t vd)
	{
	if (vd->vdev_remove_wanted) {
	vd->vdev_remove_wanted = B_FALSE;
	vd->vdev_delayed_close = B_FALSE;
	vdev_set_state(vd, B_FALSE, VDEV_STATE_REMOVED, VDEV_AUX_NONE);

	/*
	* We want to clear the stats, but we don't want to do a full
	* vdev_clear() as that will cause us to throw away
	* degraded/faulted state as well as attempt to reopen the
	* device, all of which is a waste.
	*/
	vd->vdev_stat.vs_read_errors = 0;
	vd->vdev_stat.vs_write_errors = 0;
	vd->vdev_stat.vs_checksum_errors = 0;

	vdev_state_dirty(vd->vdev_top);

	/* Tell userspace that the vdev is gone. */
	zfs_post_remove(spa, vd);
	}

	for (int c = 0; c < vd->vdev_children; c++)
	spa_async_remove(spa, vd->vdev_child[c]);
	}

	static void
	spa_async_fault_vdev(spa_t spa, vdev_t vd)
	{
	if (vd->vdev_fault_wanted) {
	vd->vdev_fault_wanted = B_FALSE;
	vdev_set_state(vd, B_TRUE, VDEV_STATE_FAULTED,
	VDEV_AUX_ERR_EXCEEDED);
	}

	for (int c = 0; c < vd->vdev_children; c++)
	spa_async_fault_vdev(spa, vd->vdev_child[c]);
	}

	static void
	spa_async_autoexpand(spa_t spa, vdev_t vd)
	{
	if (!spa->spa_autoexpand)
	return;

	for (int c = 0; c < vd->vdev_children; c++) {
	vdev_t *cvd = vd->vdev_child[c];
	spa_async_autoexpand(spa, cvd);
	}

	if (!vd->vdev_ops->vdev_op_leaf \|\| vd->vdev_physpath == NULL)
	return;

	spa_event_notify(vd->vdev_spa, vd, NULL, ESC_ZFS_VDEV_AUTOEXPAND);
	}

	static __attribute__((noreturn)) void
	spa_async_thread(void *arg)
	{
	spa_t spa = (spa_t )arg;
	dsl_pool_t *dp = spa->spa_dsl_pool;
	int tasks;

	ASSERT(spa->spa_sync_on);

	mutex_enter(&spa->spa_async_lock);
	tasks = spa->spa_async_tasks;
	spa->spa_async_tasks = 0;
	mutex_exit(&spa->spa_async_lock);

	/*
	* See if the config needs to be updated.
	*/
	if (tasks & SPA_ASYNC_CONFIG_UPDATE) {
	uint64_t old_space, new_space;

	mutex_enter(&spa_namespace_lock);
	old_space = metaslab_class_get_space(spa_normal_class(spa));
	old_space += metaslab_class_get_space(spa_special_class(spa));
	old_space += metaslab_class_get_space(spa_dedup_class(spa));
	old_space += metaslab_class_get_space(
	spa_embedded_log_class(spa));

	spa_config_update(spa, SPA_CONFIG_UPDATE_POOL);

	new_space = metaslab_class_get_space(spa_normal_class(spa));
	new_space += metaslab_class_get_space(spa_special_class(spa));
	new_space += metaslab_class_get_space(spa_dedup_class(spa));
	new_space += metaslab_class_get_space(
	spa_embedded_log_class(spa));
	mutex_exit(&spa_namespace_lock);

	/*
	* If the pool grew as a result of the config update,
	* then log an internal history event.
	*/
	if (new_space != old_space) {
	spa_history_log_internal(spa, "vdev online", NULL,
	"pool '%s' size: %llu(+%llu)",
	spa_name(spa), (u_longlong_t)new_space,
	(u_longlong_t)(new_space - old_space));
	}
	}

	/*
	* See if any devices need to be marked REMOVED.
	*/
	if (tasks & SPA_ASYNC_REMOVE) {
	spa_vdev_state_enter(spa, SCL_NONE);
	spa_async_remove(spa, spa->spa_root_vdev);
	for (int i = 0; i < spa->spa_l2cache.sav_count; i++)
	spa_async_remove(spa, spa->spa_l2cache.sav_vdevs[i]);
	for (int i = 0; i < spa->spa_spares.sav_count; i++)
	spa_async_remove(spa, spa->spa_spares.sav_vdevs[i]);
	(void) spa_vdev_state_exit(spa, NULL, 0);
	}

	if ((tasks & SPA_ASYNC_AUTOEXPAND) && !spa_suspended(spa)) {
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	spa_async_autoexpand(spa, spa->spa_root_vdev);
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	}

	/*
	* See if any devices need to be marked faulted.
	*/
	if (tasks & SPA_ASYNC_FAULT_VDEV) {
	spa_vdev_state_enter(spa, SCL_NONE);
	spa_async_fault_vdev(spa, spa->spa_root_vdev);
	(void) spa_vdev_state_exit(spa, NULL, 0);
	}

	/*
	* If any devices are done replacing, detach them.
	*/
	if (tasks & SPA_ASYNC_RESILVER_DONE \|\|
	tasks & SPA_ASYNC_REBUILD_DONE \|\|
	tasks & SPA_ASYNC_DETACH_SPARE) {
	spa_vdev_resilver_done(spa);
	}

	/*
	* Kick off a resilver.
	*/
	if (tasks & SPA_ASYNC_RESILVER &&
	!vdev_rebuild_active(spa->spa_root_vdev) &&
	(!dsl_scan_resilvering(dp) \|\|
	!spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_RESILVER_DEFER)))
	dsl_scan_restart_resilver(dp, 0);

	if (tasks & SPA_ASYNC_INITIALIZE_RESTART) {
	mutex_enter(&spa_namespace_lock);
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	vdev_initialize_restart(spa->spa_root_vdev);
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	mutex_exit(&spa_namespace_lock);
	}

	if (tasks & SPA_ASYNC_TRIM_RESTART) {
	mutex_enter(&spa_namespace_lock);
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	vdev_trim_restart(spa->spa_root_vdev);
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	mutex_exit(&spa_namespace_lock);
	}

	if (tasks & SPA_ASYNC_AUTOTRIM_RESTART) {
	mutex_enter(&spa_namespace_lock);
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	vdev_autotrim_restart(spa);
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	mutex_exit(&spa_namespace_lock);
	}

	/*
	* Kick off L2 cache whole device TRIM.
	*/
	if (tasks & SPA_ASYNC_L2CACHE_TRIM) {
	mutex_enter(&spa_namespace_lock);
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	vdev_trim_l2arc(spa);
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	mutex_exit(&spa_namespace_lock);
	}

	/*
	* Kick off L2 cache rebuilding.
	*/
	if (tasks & SPA_ASYNC_L2CACHE_REBUILD) {
	mutex_enter(&spa_namespace_lock);
	spa_config_enter(spa, SCL_L2ARC, FTAG, RW_READER);
	l2arc_spa_rebuild_start(spa);
	spa_config_exit(spa, SCL_L2ARC, FTAG);
	mutex_exit(&spa_namespace_lock);
	}

	/*
	* Let the world know that we're done.
	*/
	mutex_enter(&spa->spa_async_lock);
	spa->spa_async_thread = NULL;
	cv_broadcast(&spa->spa_async_cv);
	mutex_exit(&spa->spa_async_lock);
	thread_exit();
	}

	void
	spa_async_suspend(spa_t *spa)
	{
	mutex_enter(&spa->spa_async_lock);
	spa->spa_async_suspended++;
	while (spa->spa_async_thread != NULL)
	cv_wait(&spa->spa_async_cv, &spa->spa_async_lock);
	mutex_exit(&spa->spa_async_lock);

	spa_vdev_remove_suspend(spa);

	zthr_t *condense_thread = spa->spa_condense_zthr;
	if (condense_thread != NULL)
	zthr_cancel(condense_thread);

	zthr_t *raidz_expand_thread = spa->spa_raidz_expand_zthr;
	if (raidz_expand_thread != NULL)
	zthr_cancel(raidz_expand_thread);

	zthr_t *discard_thread = spa->spa_checkpoint_discard_zthr;
	if (discard_thread != NULL)
	zthr_cancel(discard_thread);

	zthr_t *ll_delete_thread = spa->spa_livelist_delete_zthr;
	if (ll_delete_thread != NULL)
	zthr_cancel(ll_delete_thread);

	zthr_t *ll_condense_thread = spa->spa_livelist_condense_zthr;
	if (ll_condense_thread != NULL)
	zthr_cancel(ll_condense_thread);
	}

	void
	spa_async_resume(spa_t *spa)
	{
	mutex_enter(&spa->spa_async_lock);
	ASSERT(spa->spa_async_suspended != 0);
	spa->spa_async_suspended--;
	mutex_exit(&spa->spa_async_lock);
	spa_restart_removal(spa);

	zthr_t *condense_thread = spa->spa_condense_zthr;
	if (condense_thread != NULL)
	zthr_resume(condense_thread);

	zthr_t *raidz_expand_thread = spa->spa_raidz_expand_zthr;
	if (raidz_expand_thread != NULL)
	zthr_resume(raidz_expand_thread);

	zthr_t *discard_thread = spa->spa_checkpoint_discard_zthr;
	if (discard_thread != NULL)
	zthr_resume(discard_thread);

	zthr_t *ll_delete_thread = spa->spa_livelist_delete_zthr;
	if (ll_delete_thread != NULL)
	zthr_resume(ll_delete_thread);

	zthr_t *ll_condense_thread = spa->spa_livelist_condense_zthr;
	if (ll_condense_thread != NULL)
	zthr_resume(ll_condense_thread);
	}

	static boolean_t
	spa_async_tasks_pending(spa_t *spa)
	{
	uint_t non_config_tasks;
	uint_t config_task;
	boolean_t config_task_suspended;

	non_config_tasks = spa->spa_async_tasks & ~SPA_ASYNC_CONFIG_UPDATE;
	config_task = spa->spa_async_tasks & SPA_ASYNC_CONFIG_UPDATE;
	if (spa->spa_ccw_fail_time == 0) {
	config_task_suspended = B_FALSE;
	} else {
	config_task_suspended =
	(gethrtime() - spa->spa_ccw_fail_time) <
	((hrtime_t)zfs_ccw_retry_interval * NANOSEC);
	}

	return (non_config_tasks \|\| (config_task && !config_task_suspended));
	}

	static void
	spa_async_dispatch(spa_t *spa)
	{
	mutex_enter(&spa->spa_async_lock);
	if (spa_async_tasks_pending(spa) &&
	!spa->spa_async_suspended &&
	spa->spa_async_thread == NULL)
	spa->spa_async_thread = thread_create(NULL, 0,
	spa_async_thread, spa, 0, &p0, TS_RUN, maxclsyspri);
	mutex_exit(&spa->spa_async_lock);
	}

	void
	spa_async_request(spa_t *spa, int task)
	{
	zfs_dbgmsg("spa=%s async request task=%u", spa->spa_name, task);
	mutex_enter(&spa->spa_async_lock);
	spa->spa_async_tasks \|= task;
	mutex_exit(&spa->spa_async_lock);
	}

	int
	spa_async_tasks(spa_t *spa)
	{
	return (spa->spa_async_tasks);
	}

	/*
	* ==========================================================================
	* SPA syncing routines
	* ==========================================================================
	*/


	static int
	bpobj_enqueue_cb(void arg, const blkptr_t bp, boolean_t bp_freed,
	dmu_tx_t *tx)
	{
	bpobj_t *bpo = arg;
	bpobj_enqueue(bpo, bp, bp_freed, tx);
	return (0);
	}

	int
	bpobj_enqueue_alloc_cb(void arg, const blkptr_t bp, dmu_tx_t *tx)
	{
	return (bpobj_enqueue_cb(arg, bp, B_FALSE, tx));
	}

	int
	bpobj_enqueue_free_cb(void arg, const blkptr_t bp, dmu_tx_t *tx)
	{
	return (bpobj_enqueue_cb(arg, bp, B_TRUE, tx));
	}

	static int
	spa_free_sync_cb(void arg, const blkptr_t bp, dmu_tx_t *tx)
	{
	zio_t *pio = arg;

	zio_nowait(zio_free_sync(pio, pio->io_spa, dmu_tx_get_txg(tx), bp,
	pio->io_flags));
	return (0);
	}

	static int
	bpobj_spa_free_sync_cb(void arg, const blkptr_t bp, boolean_t bp_freed,
	dmu_tx_t *tx)
	{
	ASSERT(!bp_freed);
	return (spa_free_sync_cb(arg, bp, tx));
	}

	/*
	* Note: this simple function is not inlined to make it easier to dtrace the
	* amount of time spent syncing frees.
	*/
	static void
	spa_sync_frees(spa_t spa, bplist_t bpl, dmu_tx_t *tx)
	{
	zio_t *zio = zio_root(spa, NULL, NULL, 0);
	bplist_iterate(bpl, spa_free_sync_cb, zio, tx);
	VERIFY(zio_wait(zio) == 0);
	}

	/*
	* Note: this simple function is not inlined to make it easier to dtrace the
	* amount of time spent syncing deferred frees.
	*/
	static void
	spa_sync_deferred_frees(spa_t spa, dmu_tx_t tx)
	{
	if (spa_sync_pass(spa) != 1)
	return;

	/*
	* Note:
	* If the log space map feature is active, we stop deferring
	* frees to the next TXG and therefore running this function
	* would be considered a no-op as spa_deferred_bpobj should
	* not have any entries.
	*
	* That said we run this function anyway (instead of returning
	* immediately) for the edge-case scenario where we just
	* activated the log space map feature in this TXG but we have
	* deferred frees from the previous TXG.
	*/
	zio_t *zio = zio_root(spa, NULL, NULL, 0);
	VERIFY3U(bpobj_iterate(&spa->spa_deferred_bpobj,
	bpobj_spa_free_sync_cb, zio, tx), ==, 0);
	VERIFY0(zio_wait(zio));
	}

	static void
	spa_sync_nvlist(spa_t spa, uint64_t obj, nvlist_t nv, dmu_tx_t *tx)
	{
	char *packed = NULL;
	size_t bufsize;
	size_t nvsize = 0;
	dmu_buf_t *db;

	VERIFY(nvlist_size(nv, &nvsize, NV_ENCODE_XDR) == 0);

	/*
	* Write full (SPA_CONFIG_BLOCKSIZE) blocks of configuration
	* information. This avoids the dmu_buf_will_dirty() path and
	* saves us a pre-read to get data we don't actually care about.
	*/
	bufsize = P2ROUNDUP((uint64_t)nvsize, SPA_CONFIG_BLOCKSIZE);
	packed = vmem_alloc(bufsize, KM_SLEEP);

	VERIFY(nvlist_pack(nv, &packed, &nvsize, NV_ENCODE_XDR,
	KM_SLEEP) == 0);
	memset(packed + nvsize, 0, bufsize - nvsize);

	dmu_write(spa->spa_meta_objset, obj, 0, bufsize, packed, tx);

	vmem_free(packed, bufsize);

	VERIFY(0 == dmu_bonus_hold(spa->spa_meta_objset, obj, FTAG, &db));
	dmu_buf_will_dirty(db, tx);
	(uint64_t )db->db_data = nvsize;
	dmu_buf_rele(db, FTAG);
	}

	static void
	spa_sync_aux_dev(spa_t spa, spa_aux_vdev_t sav, dmu_tx_t *tx,
	const char config, const char entry)
	{
	nvlist_t *nvroot;
	nvlist_t **list;
	int i;

	if (!sav->sav_sync)
	return;

	/*
	* Update the MOS nvlist describing the list of available devices.
	* spa_validate_aux() will have already made sure this nvlist is
	* valid and the vdevs are labeled appropriately.
	*/
	if (sav->sav_object == 0) {
	sav->sav_object = dmu_object_alloc(spa->spa_meta_objset,
	DMU_OT_PACKED_NVLIST, 1 << 14, DMU_OT_PACKED_NVLIST_SIZE,
	sizeof (uint64_t), tx);
	VERIFY(zap_update(spa->spa_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT, entry, sizeof (uint64_t), 1,
	&sav->sav_object, tx) == 0);
	}

	nvroot = fnvlist_alloc();
	if (sav->sav_count == 0) {
	fnvlist_add_nvlist_array(nvroot, config,
	(const nvlist_t * const *)NULL, 0);
	} else {
	list = kmem_alloc(sav->sav_countsizeof (void ), KM_SLEEP);
	for (i = 0; i < sav->sav_count; i++)
	list[i] = vdev_config_generate(spa, sav->sav_vdevs[i],
	B_FALSE, VDEV_CONFIG_L2CACHE);
	fnvlist_add_nvlist_array(nvroot, config,
	(const nvlist_t * const *)list, sav->sav_count);
	for (i = 0; i < sav->sav_count; i++)
	nvlist_free(list[i]);
	kmem_free(list, sav->sav_count * sizeof (void *));
	}

	spa_sync_nvlist(spa, sav->sav_object, nvroot, tx);
	nvlist_free(nvroot);

	sav->sav_sync = B_FALSE;
	}

	/*
	* Rebuild spa's all-vdev ZAP from the vdev ZAPs indicated in each vdev_t.
	* The all-vdev ZAP must be empty.
	*/
	static void
	spa_avz_build(vdev_t vd, uint64_t avz, dmu_tx_t tx)
	{
	spa_t *spa = vd->vdev_spa;

	if (vd->vdev_root_zap != 0 &&
	spa_feature_is_active(spa, SPA_FEATURE_AVZ_V2)) {
	VERIFY0(zap_add_int(spa->spa_meta_objset, avz,
	vd->vdev_root_zap, tx));
	}
	if (vd->vdev_top_zap != 0) {
	VERIFY0(zap_add_int(spa->spa_meta_objset, avz,
	vd->vdev_top_zap, tx));
	}
	if (vd->vdev_leaf_zap != 0) {
	VERIFY0(zap_add_int(spa->spa_meta_objset, avz,
	vd->vdev_leaf_zap, tx));
	}
	for (uint64_t i = 0; i < vd->vdev_children; i++) {
	spa_avz_build(vd->vdev_child[i], avz, tx);
	}
	}

	static void
	spa_sync_config_object(spa_t spa, dmu_tx_t tx)
	{
	nvlist_t *config;

	/*
	* If the pool is being imported from a pre-per-vdev-ZAP version of ZFS,
	* its config may not be dirty but we still need to build per-vdev ZAPs.
	* Similarly, if the pool is being assembled (e.g. after a split), we
	* need to rebuild the AVZ although the config may not be dirty.
	*/
	if (list_is_empty(&spa->spa_config_dirty_list) &&
	spa->spa_avz_action == AVZ_ACTION_NONE)
	return;

	spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);

	ASSERT(spa->spa_avz_action == AVZ_ACTION_NONE \|\|
	spa->spa_avz_action == AVZ_ACTION_INITIALIZE \|\|
	spa->spa_all_vdev_zaps != 0);

	if (spa->spa_avz_action == AVZ_ACTION_REBUILD) {
	/* Make and build the new AVZ */
	uint64_t new_avz = zap_create(spa->spa_meta_objset,
	DMU_OTN_ZAP_METADATA, DMU_OT_NONE, 0, tx);
	spa_avz_build(spa->spa_root_vdev, new_avz, tx);

	/* Diff old AVZ with new one */
	zap_cursor_t zc;
	zap_attribute_t za;

	for (zap_cursor_init(&zc, spa->spa_meta_objset,
	spa->spa_all_vdev_zaps);
	zap_cursor_retrieve(&zc, &za) == 0;
	zap_cursor_advance(&zc)) {
	uint64_t vdzap = za.za_first_integer;
	if (zap_lookup_int(spa->spa_meta_objset, new_avz,
	vdzap) == ENOENT) {
	/*
	* ZAP is listed in old AVZ but not in new one;
	* destroy it
	*/
	VERIFY0(zap_destroy(spa->spa_meta_objset, vdzap,
	tx));
	}
	}

	zap_cursor_fini(&zc);

	/* Destroy the old AVZ */
	VERIFY0(zap_destroy(spa->spa_meta_objset,
	spa->spa_all_vdev_zaps, tx));

	/* Replace the old AVZ in the dir obj with the new one */
	VERIFY0(zap_update(spa->spa_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_VDEV_ZAP_MAP,
	sizeof (new_avz), 1, &new_avz, tx));

	spa->spa_all_vdev_zaps = new_avz;
	} else if (spa->spa_avz_action == AVZ_ACTION_DESTROY) {
	zap_cursor_t zc;
	zap_attribute_t za;

	/* Walk through the AVZ and destroy all listed ZAPs */
	for (zap_cursor_init(&zc, spa->spa_meta_objset,
	spa->spa_all_vdev_zaps);
	zap_cursor_retrieve(&zc, &za) == 0;
	zap_cursor_advance(&zc)) {
	uint64_t zap = za.za_first_integer;
	VERIFY0(zap_destroy(spa->spa_meta_objset, zap, tx));
	}

	zap_cursor_fini(&zc);

	/* Destroy and unlink the AVZ itself */
	VERIFY0(zap_destroy(spa->spa_meta_objset,
	spa->spa_all_vdev_zaps, tx));
	VERIFY0(zap_remove(spa->spa_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_VDEV_ZAP_MAP, tx));
	spa->spa_all_vdev_zaps = 0;
	}

	if (spa->spa_all_vdev_zaps == 0) {
	spa->spa_all_vdev_zaps = zap_create_link(spa->spa_meta_objset,
	DMU_OTN_ZAP_METADATA, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_VDEV_ZAP_MAP, tx);
	}
	spa->spa_avz_action = AVZ_ACTION_NONE;

	/* Create ZAPs for vdevs that don't have them. */
	vdev_construct_zaps(spa->spa_root_vdev, tx);

	config = spa_config_generate(spa, spa->spa_root_vdev,
	dmu_tx_get_txg(tx), B_FALSE);

	/*
	* If we're upgrading the spa version then make sure that
	* the config object gets updated with the correct version.
	*/
	if (spa->spa_ubsync.ub_version < spa->spa_uberblock.ub_version)
	fnvlist_add_uint64(config, ZPOOL_CONFIG_VERSION,
	spa->spa_uberblock.ub_version);

	spa_config_exit(spa, SCL_STATE, FTAG);

	nvlist_free(spa->spa_config_syncing);
	spa->spa_config_syncing = config;

	spa_sync_nvlist(spa, spa->spa_config_object, config, tx);
	}

	static void
	spa_sync_version(void arg, dmu_tx_t tx)
	{
	uint64_t *versionp = arg;
	uint64_t version = *versionp;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;

	/*
	* Setting the version is special cased when first creating the pool.
	*/
	ASSERT(tx->tx_txg != TXG_INITIAL);

	ASSERT(SPA_VERSION_IS_SUPPORTED(version));
	ASSERT(version >= spa_version(spa));

	spa->spa_uberblock.ub_version = version;
	vdev_config_dirty(spa->spa_root_vdev);
	spa_history_log_internal(spa, "set", tx, "version=%lld",
	(longlong_t)version);
	}

	/*
	* Set zpool properties.
	*/
	static void
	spa_sync_props(void arg, dmu_tx_t tx)
	{
	nvlist_t *nvp = arg;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	objset_t *mos = spa->spa_meta_objset;
	nvpair_t *elem = NULL;

	mutex_enter(&spa->spa_props_lock);

	while ((elem = nvlist_next_nvpair(nvp, elem))) {
	uint64_t intval;
	const char strval, fname;
	zpool_prop_t prop;
	const char *propname;
	const char *elemname = nvpair_name(elem);
	zprop_type_t proptype;
	spa_feature_t fid;

	switch (prop = zpool_name_to_prop(elemname)) {
	case ZPOOL_PROP_VERSION:
	intval = fnvpair_value_uint64(elem);
	/*
	* The version is synced separately before other
	* properties and should be correct by now.
	*/
	ASSERT3U(spa_version(spa), >=, intval);
	break;

	case ZPOOL_PROP_ALTROOT:
	/*
	* 'altroot' is a non-persistent property. It should
	* have been set temporarily at creation or import time.
	*/
	ASSERT(spa->spa_root != NULL);
	break;

	case ZPOOL_PROP_READONLY:
	case ZPOOL_PROP_CACHEFILE:
	/*
	* 'readonly' and 'cachefile' are also non-persistent
	* properties.
	*/
	break;
	case ZPOOL_PROP_COMMENT:
	strval = fnvpair_value_string(elem);
	if (spa->spa_comment != NULL)
	spa_strfree(spa->spa_comment);
	spa->spa_comment = spa_strdup(strval);
	/*
	* We need to dirty the configuration on all the vdevs
	* so that their labels get updated. We also need to
	* update the cache file to keep it in sync with the
	* MOS version. It's unnecessary to do this for pool
	* creation since the vdev's configuration has already
	* been dirtied.
	*/
	if (tx->tx_txg != TXG_INITIAL) {
	vdev_config_dirty(spa->spa_root_vdev);
	spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE);
	}
	spa_history_log_internal(spa, "set", tx,
	"%s=%s", elemname, strval);
	break;
	case ZPOOL_PROP_COMPATIBILITY:
	strval = fnvpair_value_string(elem);
	if (spa->spa_compatibility != NULL)
	spa_strfree(spa->spa_compatibility);
	spa->spa_compatibility = spa_strdup(strval);
	/*
	* Dirty the configuration on vdevs as above.
	*/
	if (tx->tx_txg != TXG_INITIAL) {
	vdev_config_dirty(spa->spa_root_vdev);
	spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE);
	}

	spa_history_log_internal(spa, "set", tx,
	"%s=%s", nvpair_name(elem), strval);
	break;

	case ZPOOL_PROP_INVAL:
	if (zpool_prop_feature(elemname)) {
	fname = strchr(elemname, '@') + 1;
	VERIFY0(zfeature_lookup_name(fname, &fid));

	spa_feature_enable(spa, fid, tx);
	spa_history_log_internal(spa, "set", tx,
	"%s=enabled", elemname);
	break;
	} else if (!zfs_prop_user(elemname)) {
	ASSERT(zpool_prop_feature(elemname));
	break;
	}
	zfs_fallthrough;
	default:
	/*
	* Set pool property values in the poolprops mos object.
	*/
	if (spa->spa_pool_props_object == 0) {
	spa->spa_pool_props_object =
	zap_create_link(mos, DMU_OT_POOL_PROPS,
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_PROPS,
	tx);
	}

	/* normalize the property name */
	if (prop == ZPOOL_PROP_INVAL) {
	propname = elemname;
	proptype = PROP_TYPE_STRING;
	} else {
	propname = zpool_prop_to_name(prop);
	proptype = zpool_prop_get_type(prop);
	}

	if (nvpair_type(elem) == DATA_TYPE_STRING) {
	ASSERT(proptype == PROP_TYPE_STRING);
	strval = fnvpair_value_string(elem);
	VERIFY0(zap_update(mos,
	spa->spa_pool_props_object, propname,
	1, strlen(strval) + 1, strval, tx));
	spa_history_log_internal(spa, "set", tx,
	"%s=%s", elemname, strval);
	} else if (nvpair_type(elem) == DATA_TYPE_UINT64) {
	intval = fnvpair_value_uint64(elem);

	if (proptype == PROP_TYPE_INDEX) {
	const char *unused;
	VERIFY0(zpool_prop_index_to_string(
	prop, intval, &unused));
	}
	VERIFY0(zap_update(mos,
	spa->spa_pool_props_object, propname,
	8, 1, &intval, tx));
	spa_history_log_internal(spa, "set", tx,
	"%s=%lld", elemname,
	(longlong_t)intval);

	switch (prop) {
	case ZPOOL_PROP_DELEGATION:
	spa->spa_delegation = intval;
	break;
	case ZPOOL_PROP_BOOTFS:
	spa->spa_bootfs = intval;
	break;
	case ZPOOL_PROP_FAILUREMODE:
	spa->spa_failmode = intval;
	break;
	case ZPOOL_PROP_AUTOTRIM:
	spa->spa_autotrim = intval;
	spa_async_request(spa,
	SPA_ASYNC_AUTOTRIM_RESTART);
	break;
	case ZPOOL_PROP_AUTOEXPAND:
	spa->spa_autoexpand = intval;
	if (tx->tx_txg != TXG_INITIAL)
	spa_async_request(spa,
	SPA_ASYNC_AUTOEXPAND);
	break;
	case ZPOOL_PROP_MULTIHOST:
	spa->spa_multihost = intval;
	break;
	default:
	break;
	}
	} else {
	ASSERT(0); /* not allowed */
	}
	}

	}

	mutex_exit(&spa->spa_props_lock);
	}

	/*
	* Perform one-time upgrade on-disk changes. spa_version() does not
	* reflect the new version this txg, so there must be no changes this
	* txg to anything that the upgrade code depends on after it executes.
	* Therefore this must be called after dsl_pool_sync() does the sync
	* tasks.
	*/
	static void
	spa_sync_upgrades(spa_t spa, dmu_tx_t tx)
	{
	if (spa_sync_pass(spa) != 1)
	return;

	dsl_pool_t *dp = spa->spa_dsl_pool;
	rrw_enter(&dp->dp_config_rwlock, RW_WRITER, FTAG);

	if (spa->spa_ubsync.ub_version < SPA_VERSION_ORIGIN &&
	spa->spa_uberblock.ub_version >= SPA_VERSION_ORIGIN) {
	dsl_pool_create_origin(dp, tx);

	/* Keeping the origin open increases spa_minref */
	spa->spa_minref += 3;
	}

	if (spa->spa_ubsync.ub_version < SPA_VERSION_NEXT_CLONES &&
	spa->spa_uberblock.ub_version >= SPA_VERSION_NEXT_CLONES) {
	dsl_pool_upgrade_clones(dp, tx);
	}

	if (spa->spa_ubsync.ub_version < SPA_VERSION_DIR_CLONES &&
	spa->spa_uberblock.ub_version >= SPA_VERSION_DIR_CLONES) {
	dsl_pool_upgrade_dir_clones(dp, tx);

	/* Keeping the freedir open increases spa_minref */
	spa->spa_minref += 3;
	}

	if (spa->spa_ubsync.ub_version < SPA_VERSION_FEATURES &&
	spa->spa_uberblock.ub_version >= SPA_VERSION_FEATURES) {
	spa_feature_create_zap_objects(spa, tx);
	}

	/*
	* LZ4_COMPRESS feature's behaviour was changed to activate_on_enable
	* when possibility to use lz4 compression for metadata was added
	* Old pools that have this feature enabled must be upgraded to have
	* this feature active
	*/
	if (spa->spa_uberblock.ub_version >= SPA_VERSION_FEATURES) {
	boolean_t lz4_en = spa_feature_is_enabled(spa,
	SPA_FEATURE_LZ4_COMPRESS);
	boolean_t lz4_ac = spa_feature_is_active(spa,
	SPA_FEATURE_LZ4_COMPRESS);

	if (lz4_en && !lz4_ac)
	spa_feature_incr(spa, SPA_FEATURE_LZ4_COMPRESS, tx);
	}

	/*
	* If we haven't written the salt, do so now. Note that the
	* feature may not be activated yet, but that's fine since
	* the presence of this ZAP entry is backwards compatible.
	*/
	if (zap_contains(spa->spa_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
	DMU_POOL_CHECKSUM_SALT) == ENOENT) {
	VERIFY0(zap_add(spa->spa_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_CHECKSUM_SALT, 1,
	sizeof (spa->spa_cksum_salt.zcs_bytes),
	spa->spa_cksum_salt.zcs_bytes, tx));
	}

	rrw_exit(&dp->dp_config_rwlock, FTAG);
	}

	static void
	vdev_indirect_state_sync_verify(vdev_t *vd)
	{
	vdev_indirect_mapping_t *vim __maybe_unused = vd->vdev_indirect_mapping;
	vdev_indirect_births_t *vib __maybe_unused = vd->vdev_indirect_births;

	if (vd->vdev_ops == &vdev_indirect_ops) {
	ASSERT(vim != NULL);
	ASSERT(vib != NULL);
	}

	uint64_t obsolete_sm_object = 0;
	ASSERT0(vdev_obsolete_sm_object(vd, &obsolete_sm_object));
	if (obsolete_sm_object != 0) {
	ASSERT(vd->vdev_obsolete_sm != NULL);
	ASSERT(vd->vdev_removing \|\|
	vd->vdev_ops == &vdev_indirect_ops);
	ASSERT(vdev_indirect_mapping_num_entries(vim) > 0);
	ASSERT(vdev_indirect_mapping_bytes_mapped(vim) > 0);
	ASSERT3U(obsolete_sm_object, ==,
	space_map_object(vd->vdev_obsolete_sm));
	ASSERT3U(vdev_indirect_mapping_bytes_mapped(vim), >=,
	space_map_allocated(vd->vdev_obsolete_sm));
	}
	ASSERT(vd->vdev_obsolete_segments != NULL);

	/*
	* Since frees / remaps to an indirect vdev can only
	* happen in syncing context, the obsolete segments
	* tree must be empty when we start syncing.
	*/
	ASSERT0(range_tree_space(vd->vdev_obsolete_segments));
	}

	/*
	* Set the top-level vdev's max queue depth. Evaluate each top-level's
	* async write queue depth in case it changed. The max queue depth will
	* not change in the middle of syncing out this txg.
	*/
	static void
	spa_sync_adjust_vdev_max_queue_depth(spa_t *spa)
	{
	ASSERT(spa_writeable(spa));

	vdev_t *rvd = spa->spa_root_vdev;
	uint32_t max_queue_depth = zfs_vdev_async_write_max_active *
	zfs_vdev_queue_depth_pct / 100;
	metaslab_class_t *normal = spa_normal_class(spa);
	metaslab_class_t *special = spa_special_class(spa);
	metaslab_class_t *dedup = spa_dedup_class(spa);

	uint64_t slots_per_allocator = 0;
	for (int c = 0; c < rvd->vdev_children; c++) {
	vdev_t *tvd = rvd->vdev_child[c];

	metaslab_group_t *mg = tvd->vdev_mg;
	if (mg == NULL \|\| !metaslab_group_initialized(mg))
	continue;

	metaslab_class_t *mc = mg->mg_class;
	if (mc != normal && mc != special && mc != dedup)
	continue;

	/*
	* It is safe to do a lock-free check here because only async
	* allocations look at mg_max_alloc_queue_depth, and async
	* allocations all happen from spa_sync().
	*/
	for (int i = 0; i < mg->mg_allocators; i++) {
	ASSERT0(zfs_refcount_count(
	&(mg->mg_allocator[i].mga_alloc_queue_depth)));
	}
	mg->mg_max_alloc_queue_depth = max_queue_depth;

	for (int i = 0; i < mg->mg_allocators; i++) {
	mg->mg_allocator[i].mga_cur_max_alloc_queue_depth =
	zfs_vdev_def_queue_depth;
	}
	slots_per_allocator += zfs_vdev_def_queue_depth;
	}

	for (int i = 0; i < spa->spa_alloc_count; i++) {
	ASSERT0(zfs_refcount_count(&normal->mc_allocator[i].
	mca_alloc_slots));
	ASSERT0(zfs_refcount_count(&special->mc_allocator[i].
	mca_alloc_slots));
	ASSERT0(zfs_refcount_count(&dedup->mc_allocator[i].
	mca_alloc_slots));
	normal->mc_allocator[i].mca_alloc_max_slots =
	slots_per_allocator;
	special->mc_allocator[i].mca_alloc_max_slots =
	slots_per_allocator;
	dedup->mc_allocator[i].mca_alloc_max_slots =
	slots_per_allocator;
	}
	normal->mc_alloc_throttle_enabled = zio_dva_throttle_enabled;
	special->mc_alloc_throttle_enabled = zio_dva_throttle_enabled;
	dedup->mc_alloc_throttle_enabled = zio_dva_throttle_enabled;
	}

	static void
	spa_sync_condense_indirect(spa_t spa, dmu_tx_t tx)
	{
	ASSERT(spa_writeable(spa));

	vdev_t *rvd = spa->spa_root_vdev;
	for (int c = 0; c < rvd->vdev_children; c++) {
	vdev_t *vd = rvd->vdev_child[c];
	vdev_indirect_state_sync_verify(vd);

	if (vdev_indirect_should_condense(vd)) {
	spa_condense_indirect_start_sync(vd, tx);
	break;
	}
	}
	}

	static void
	spa_sync_iterate_to_convergence(spa_t spa, dmu_tx_t tx)
	{
	objset_t *mos = spa->spa_meta_objset;
	dsl_pool_t *dp = spa->spa_dsl_pool;
	uint64_t txg = tx->tx_txg;
	bplist_t *free_bpl = &spa->spa_free_bplist[txg & TXG_MASK];

	do {
	int pass = ++spa->spa_sync_pass;

	spa_sync_config_object(spa, tx);
	spa_sync_aux_dev(spa, &spa->spa_spares, tx,
	ZPOOL_CONFIG_SPARES, DMU_POOL_SPARES);
	spa_sync_aux_dev(spa, &spa->spa_l2cache, tx,
	ZPOOL_CONFIG_L2CACHE, DMU_POOL_L2CACHE);
	spa_errlog_sync(spa, txg);
	dsl_pool_sync(dp, txg);

	if (pass < zfs_sync_pass_deferred_free \|\|
	spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP)) {
	/*
	* If the log space map feature is active we don't
	* care about deferred frees and the deferred bpobj
	* as the log space map should effectively have the
	* same results (i.e. appending only to one object).
	*/
	spa_sync_frees(spa, free_bpl, tx);
	} else {
	/*
	* We can not defer frees in pass 1, because
	* we sync the deferred frees later in pass 1.
	*/
	ASSERT3U(pass, >, 1);
	bplist_iterate(free_bpl, bpobj_enqueue_alloc_cb,
	&spa->spa_deferred_bpobj, tx);
	}

	brt_sync(spa, txg);
	ddt_sync(spa, txg);
	dsl_scan_sync(dp, tx);
	dsl_errorscrub_sync(dp, tx);
	svr_sync(spa, tx);
	spa_sync_upgrades(spa, tx);

	spa_flush_metaslabs(spa, tx);

	vdev_t *vd = NULL;
	while ((vd = txg_list_remove(&spa->spa_vdev_txg_list, txg))
	!= NULL)
	vdev_sync(vd, txg);

	if (pass == 1) {
	/*
	* dsl_pool_sync() -> dp_sync_tasks may have dirtied
	* the config. If that happens, this txg should not
	* be a no-op. So we must sync the config to the MOS
	* before checking for no-op.
	*
	* Note that when the config is dirty, it will
	* be written to the MOS (i.e. the MOS will be
	* dirtied) every time we call spa_sync_config_object()
	* in this txg. Therefore we can't call this after
	* dsl_pool_sync() every pass, because it would
	* prevent us from converging, since we'd dirty
	* the MOS every pass.
	*
	* Sync tasks can only be processed in pass 1, so
	* there's no need to do this in later passes.
	*/
	spa_sync_config_object(spa, tx);
	}

	/*
	* Note: We need to check if the MOS is dirty because we could
	* have marked the MOS dirty without updating the uberblock
	* (e.g. if we have sync tasks but no dirty user data). We need
	* to check the uberblock's rootbp because it is updated if we
	* have synced out dirty data (though in this case the MOS will
	* most likely also be dirty due to second order effects, we
	* don't want to rely on that here).
	*/
	if (pass == 1 &&
	BP_GET_LOGICAL_BIRTH(&spa->spa_uberblock.ub_rootbp) < txg &&
	!dmu_objset_is_dirty(mos, txg)) {
	/*
	* Nothing changed on the first pass, therefore this
	* TXG is a no-op. Avoid syncing deferred frees, so
	* that we can keep this TXG as a no-op.
	*/
	ASSERT(txg_list_empty(&dp->dp_dirty_datasets, txg));
	ASSERT(txg_list_empty(&dp->dp_dirty_dirs, txg));
	ASSERT(txg_list_empty(&dp->dp_sync_tasks, txg));
	ASSERT(txg_list_empty(&dp->dp_early_sync_tasks, txg));
	break;
	}

	spa_sync_deferred_frees(spa, tx);
	} while (dmu_objset_is_dirty(mos, txg));
	}

	/*
	* Rewrite the vdev configuration (which includes the uberblock) to
	* commit the transaction group.
	*
	* If there are no dirty vdevs, we sync the uberblock to a few random
	* top-level vdevs that are known to be visible in the config cache
	* (see spa_vdev_add() for a complete description). If there are dirty
	* vdevs, sync the uberblock to all vdevs.
	*/
	static void
	spa_sync_rewrite_vdev_config(spa_t spa, dmu_tx_t tx)
	{
	vdev_t *rvd = spa->spa_root_vdev;
	uint64_t txg = tx->tx_txg;

	for (;;) {
	int error = 0;

	/*
	* We hold SCL_STATE to prevent vdev open/close/etc.
	* while we're attempting to write the vdev labels.
	*/
	spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);

	if (list_is_empty(&spa->spa_config_dirty_list)) {
	vdev_t *svd[SPA_SYNC_MIN_VDEVS] = { NULL };
	int svdcount = 0;
	int children = rvd->vdev_children;
	int c0 = random_in_range(children);

	for (int c = 0; c < children; c++) {
	vdev_t *vd =
	rvd->vdev_child[(c0 + c) % children];

	/* Stop when revisiting the first vdev */
	if (c > 0 && svd[0] == vd)
	break;

	if (vd->vdev_ms_array == 0 \|\|
	vd->vdev_islog \|\|
	!vdev_is_concrete(vd))
	continue;

	svd[svdcount++] = vd;
	if (svdcount == SPA_SYNC_MIN_VDEVS)
	break;
	}
	error = vdev_config_sync(svd, svdcount, txg);
	} else {
	error = vdev_config_sync(rvd->vdev_child,
	rvd->vdev_children, txg);
	}

	if (error == 0)
	spa->spa_last_synced_guid = rvd->vdev_guid;

	spa_config_exit(spa, SCL_STATE, FTAG);

	if (error == 0)
	break;
	zio_suspend(spa, NULL, ZIO_SUSPEND_IOERR);
	zio_resume_wait(spa);
	}
	}

	/*
	* Sync the specified transaction group. New blocks may be dirtied as
	* part of the process, so we iterate until it converges.
	*/
	void
	spa_sync(spa_t *spa, uint64_t txg)
	{
	vdev_t *vd = NULL;

	VERIFY(spa_writeable(spa));

	/*
	* Wait for i/os issued in open context that need to complete
	* before this txg syncs.
	*/
	(void) zio_wait(spa->spa_txg_zio[txg & TXG_MASK]);
	spa->spa_txg_zio[txg & TXG_MASK] = zio_root(spa, NULL, NULL,
	ZIO_FLAG_CANFAIL);

	/*
	* Now that there can be no more cloning in this transaction group,
	* but we are still before issuing frees, we can process pending BRT
	* updates.
	*/
	brt_pending_apply(spa, txg);

	/*
	* Lock out configuration changes.
	*/
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);

	spa->spa_syncing_txg = txg;
	spa->spa_sync_pass = 0;

	for (int i = 0; i < spa->spa_alloc_count; i++) {
	mutex_enter(&spa->spa_allocs[i].spaa_lock);
	VERIFY0(avl_numnodes(&spa->spa_allocs[i].spaa_tree));
	mutex_exit(&spa->spa_allocs[i].spaa_lock);
	}

	/*
	* If there are any pending vdev state changes, convert them
	* into config changes that go out with this transaction group.
	*/
	spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
	while ((vd = list_head(&spa->spa_state_dirty_list)) != NULL) {
	/* Avoid holding the write lock unless actually necessary */
	if (vd->vdev_aux == NULL) {
	vdev_state_clean(vd);
	vdev_config_dirty(vd);
	continue;
	}
	/*
	* We need the write lock here because, for aux vdevs,
	* calling vdev_config_dirty() modifies sav_config.
	* This is ugly and will become unnecessary when we
	* eliminate the aux vdev wart by integrating all vdevs
	* into the root vdev tree.
	*/
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	spa_config_enter(spa, SCL_CONFIG \| SCL_STATE, FTAG, RW_WRITER);
	while ((vd = list_head(&spa->spa_state_dirty_list)) != NULL) {
	vdev_state_clean(vd);
	vdev_config_dirty(vd);
	}
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	spa_config_enter(spa, SCL_CONFIG \| SCL_STATE, FTAG, RW_READER);
	}
	spa_config_exit(spa, SCL_STATE, FTAG);

	dsl_pool_t *dp = spa->spa_dsl_pool;
	dmu_tx_t *tx = dmu_tx_create_assigned(dp, txg);

	spa->spa_sync_starttime = gethrtime();
	taskq_cancel_id(system_delay_taskq, spa->spa_deadman_tqid);
	spa->spa_deadman_tqid = taskq_dispatch_delay(system_delay_taskq,
	spa_deadman, spa, TQ_SLEEP, ddi_get_lbolt() +
	NSEC_TO_TICK(spa->spa_deadman_synctime));

	/*
	* If we are upgrading to SPA_VERSION_RAIDZ_DEFLATE this txg,
	* set spa_deflate if we have no raid-z vdevs.
	*/
	if (spa->spa_ubsync.ub_version < SPA_VERSION_RAIDZ_DEFLATE &&
	spa->spa_uberblock.ub_version >= SPA_VERSION_RAIDZ_DEFLATE) {
	vdev_t *rvd = spa->spa_root_vdev;

	int i;
	for (i = 0; i < rvd->vdev_children; i++) {
	vd = rvd->vdev_child[i];
	if (vd->vdev_deflate_ratio != SPA_MINBLOCKSIZE)
	break;
	}
	if (i == rvd->vdev_children) {
	spa->spa_deflate = TRUE;
	VERIFY0(zap_add(spa->spa_meta_objset,
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_DEFLATE,
	sizeof (uint64_t), 1, &spa->spa_deflate, tx));
	}
	}

	spa_sync_adjust_vdev_max_queue_depth(spa);

	spa_sync_condense_indirect(spa, tx);

	spa_sync_iterate_to_convergence(spa, tx);

	#ifdef ZFS_DEBUG
	if (!list_is_empty(&spa->spa_config_dirty_list)) {
	/*
	* Make sure that the number of ZAPs for all the vdevs matches
	* the number of ZAPs in the per-vdev ZAP list. This only gets
	* called if the config is dirty; otherwise there may be
	* outstanding AVZ operations that weren't completed in
	* spa_sync_config_object.
	*/
	uint64_t all_vdev_zap_entry_count;
	ASSERT0(zap_count(spa->spa_meta_objset,
	spa->spa_all_vdev_zaps, &all_vdev_zap_entry_count));
	ASSERT3U(vdev_count_verify_zaps(spa->spa_root_vdev), ==,
	all_vdev_zap_entry_count);
	}
	#endif

	if (spa->spa_vdev_removal != NULL) {
	ASSERT0(spa->spa_vdev_removal->svr_bytes_done[txg & TXG_MASK]);
	}

	spa_sync_rewrite_vdev_config(spa, tx);
	dmu_tx_commit(tx);

	taskq_cancel_id(system_delay_taskq, spa->spa_deadman_tqid);
	spa->spa_deadman_tqid = 0;

	/*
	* Clear the dirty config list.
	*/
	while ((vd = list_head(&spa->spa_config_dirty_list)) != NULL)
	vdev_config_clean(vd);

	/*
	* Now that the new config has synced transactionally,
	* let it become visible to the config cache.
	*/
	if (spa->spa_config_syncing != NULL) {
	spa_config_set(spa, spa->spa_config_syncing);
	spa->spa_config_txg = txg;
	spa->spa_config_syncing = NULL;
	}

	dsl_pool_sync_done(dp, txg);

	for (int i = 0; i < spa->spa_alloc_count; i++) {
	mutex_enter(&spa->spa_allocs[i].spaa_lock);
	VERIFY0(avl_numnodes(&spa->spa_allocs[i].spaa_tree));
	mutex_exit(&spa->spa_allocs[i].spaa_lock);
	}

	/*
	* Update usable space statistics.
	*/
	while ((vd = txg_list_remove(&spa->spa_vdev_txg_list, TXG_CLEAN(txg)))
	!= NULL)
	vdev_sync_done(vd, txg);

	metaslab_class_evict_old(spa->spa_normal_class, txg);
	metaslab_class_evict_old(spa->spa_log_class, txg);
	+ /* spa_embedded_log_class has only one metaslab per vdev. */
	+ metaslab_class_evict_old(spa->spa_special_class, txg);
	+ metaslab_class_evict_old(spa->spa_dedup_class, txg);

	spa_sync_close_syncing_log_sm(spa);

	spa_update_dspace(spa);

	if (spa_get_autotrim(spa) == SPA_AUTOTRIM_ON)
	vdev_autotrim_kick(spa);

	/*
	* It had better be the case that we didn't dirty anything
	* since vdev_config_sync().
	*/
	ASSERT(txg_list_empty(&dp->dp_dirty_datasets, txg));
	ASSERT(txg_list_empty(&dp->dp_dirty_dirs, txg));
	ASSERT(txg_list_empty(&spa->spa_vdev_txg_list, txg));

	while (zfs_pause_spa_sync)
	delay(1);

	spa->spa_sync_pass = 0;

	/*
	* Update the last synced uberblock here. We want to do this at
	* the end of spa_sync() so that consumers of spa_last_synced_txg()
	* will be guaranteed that all the processing associated with
	* that txg has been completed.
	*/
	spa->spa_ubsync = spa->spa_uberblock;
	spa_config_exit(spa, SCL_CONFIG, FTAG);

	spa_handle_ignored_writes(spa);

	/*
	* If any async tasks have been requested, kick them off.
	*/
	spa_async_dispatch(spa);
	}

	/*
	* Sync all pools. We don't want to hold the namespace lock across these
	* operations, so we take a reference on the spa_t and drop the lock during the
	* sync.
	*/
	void
	spa_sync_allpools(void)
	{
	spa_t *spa = NULL;
	mutex_enter(&spa_namespace_lock);
	while ((spa = spa_next(spa)) != NULL) {
	if (spa_state(spa) != POOL_STATE_ACTIVE \|\|
	!spa_writeable(spa) \|\| spa_suspended(spa))
	continue;
	spa_open_ref(spa, FTAG);
	mutex_exit(&spa_namespace_lock);
	txg_wait_synced(spa_get_dsl(spa), 0);
	mutex_enter(&spa_namespace_lock);
	spa_close(spa, FTAG);
	}
	mutex_exit(&spa_namespace_lock);
	}

	taskq_t *
	spa_sync_tq_create(spa_t spa, const char name)
	{
	kthread_t **kthreads;

	ASSERT(spa->spa_sync_tq == NULL);
	ASSERT3S(spa->spa_alloc_count, <=, boot_ncpus);

	/*
	* - do not allow more allocators than cpus.
	* - there may be more cpus than allocators.
	* - do not allow more sync taskq threads than allocators or cpus.
	*/
	int nthreads = spa->spa_alloc_count;
	spa->spa_syncthreads = kmem_zalloc(sizeof (spa_syncthread_info_t) *
	nthreads, KM_SLEEP);

	spa->spa_sync_tq = taskq_create_synced(name, nthreads, minclsyspri,
	nthreads, INT_MAX, TASKQ_PREPOPULATE, &kthreads);
	VERIFY(spa->spa_sync_tq != NULL);
	VERIFY(kthreads != NULL);

	spa_syncthread_info_t *ti = spa->spa_syncthreads;
	for (int i = 0; i < nthreads; i++, ti++) {
	ti->sti_thread = kthreads[i];
	ti->sti_allocator = i;
	}

	kmem_free(kthreads, sizeof (kthreads) nthreads);
	return (spa->spa_sync_tq);
	}

	void
	spa_sync_tq_destroy(spa_t *spa)
	{
	ASSERT(spa->spa_sync_tq != NULL);

	taskq_wait(spa->spa_sync_tq);
	taskq_destroy(spa->spa_sync_tq);
	kmem_free(spa->spa_syncthreads,
	sizeof (spa_syncthread_info_t) * spa->spa_alloc_count);
	spa->spa_sync_tq = NULL;
	}

	uint_t
	spa_acq_allocator(spa_t *spa)
	{
	int i;

	if (spa->spa_alloc_count == 1)
	return (0);

	mutex_enter(&spa->spa_allocs_use->sau_lock);
	uint_t r = spa->spa_allocs_use->sau_rotor;
	do {
	if (++r == spa->spa_alloc_count)
	r = 0;
	} while (spa->spa_allocs_use->sau_inuse[r]);
	spa->spa_allocs_use->sau_inuse[r] = B_TRUE;
	spa->spa_allocs_use->sau_rotor = r;
	mutex_exit(&spa->spa_allocs_use->sau_lock);

	spa_syncthread_info_t *ti = spa->spa_syncthreads;
	for (i = 0; i < spa->spa_alloc_count; i++, ti++) {
	if (ti->sti_thread == curthread) {
	ti->sti_allocator = r;
	break;
	}
	}
	ASSERT3S(i, <, spa->spa_alloc_count);
	return (r);
	}

	void
	spa_rel_allocator(spa_t *spa, uint_t allocator)
	{
	if (spa->spa_alloc_count > 1)
	spa->spa_allocs_use->sau_inuse[allocator] = B_FALSE;
	}

	void
	spa_select_allocator(zio_t *zio)
	{
	zbookmark_phys_t *bm = &zio->io_bookmark;
	spa_t *spa = zio->io_spa;

	ASSERT(zio->io_type == ZIO_TYPE_WRITE);

	/*
	* A gang block (for example) may have inherited its parent's
	* allocator, in which case there is nothing further to do here.
	*/
	if (ZIO_HAS_ALLOCATOR(zio))
	return;

	ASSERT(spa != NULL);
	ASSERT(bm != NULL);

	/*
	* First try to use an allocator assigned to the syncthread, and set
	* the corresponding write issue taskq for the allocator.
	* Note, we must have an open pool to do this.
	*/
	if (spa->spa_sync_tq != NULL) {
	spa_syncthread_info_t *ti = spa->spa_syncthreads;
	for (int i = 0; i < spa->spa_alloc_count; i++, ti++) {
	if (ti->sti_thread == curthread) {
	zio->io_allocator = ti->sti_allocator;
	return;
	}
	}
	}

	/*
	* We want to try to use as many allocators as possible to help improve
	* performance, but we also want logically adjacent IOs to be physically
	* adjacent to improve sequential read performance. We chunk each object
	* into 2^20 block regions, and then hash based on the objset, object,
	* level, and region to accomplish both of these goals.
	*/
	uint64_t hv = cityhash4(bm->zb_objset, bm->zb_object, bm->zb_level,
	bm->zb_blkid >> 20);

	zio->io_allocator = (uint_t)hv % spa->spa_alloc_count;
	}

	/*
	* ==========================================================================
	* Miscellaneous routines
	* ==========================================================================
	*/

	/*
	* Remove all pools in the system.
	*/
	void
	spa_evict_all(void)
	{
	spa_t *spa;

	/*
	* Remove all cached state. All pools should be closed now,
	* so every spa in the AVL tree should be unreferenced.
	*/
	mutex_enter(&spa_namespace_lock);
	while ((spa = spa_next(NULL)) != NULL) {
	/*
	* Stop async tasks. The async thread may need to detach
	* a device that's been replaced, which requires grabbing
	* spa_namespace_lock, so we must drop it here.
	*/
	spa_open_ref(spa, FTAG);
	mutex_exit(&spa_namespace_lock);
	spa_async_suspend(spa);
	mutex_enter(&spa_namespace_lock);
	spa_close(spa, FTAG);

	if (spa->spa_state != POOL_STATE_UNINITIALIZED) {
	spa_unload(spa);
	spa_deactivate(spa);
	}
	spa_remove(spa);
	}
	mutex_exit(&spa_namespace_lock);
	}

	vdev_t *
	spa_lookup_by_guid(spa_t *spa, uint64_t guid, boolean_t aux)
	{
	vdev_t *vd;
	int i;

	if ((vd = vdev_lookup_by_guid(spa->spa_root_vdev, guid)) != NULL)
	return (vd);

	if (aux) {
	for (i = 0; i < spa->spa_l2cache.sav_count; i++) {
	vd = spa->spa_l2cache.sav_vdevs[i];
	if (vd->vdev_guid == guid)
	return (vd);
	}

	for (i = 0; i < spa->spa_spares.sav_count; i++) {
	vd = spa->spa_spares.sav_vdevs[i];
	if (vd->vdev_guid == guid)
	return (vd);
	}
	}

	return (NULL);
	}

	void
	spa_upgrade(spa_t *spa, uint64_t version)
	{
	ASSERT(spa_writeable(spa));

	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);

	/*
	* This should only be called for a non-faulted pool, and since a
	* future version would result in an unopenable pool, this shouldn't be
	* possible.
	*/
	ASSERT(SPA_VERSION_IS_SUPPORTED(spa->spa_uberblock.ub_version));
	ASSERT3U(version, >=, spa->spa_uberblock.ub_version);

	spa->spa_uberblock.ub_version = version;
	vdev_config_dirty(spa->spa_root_vdev);

	spa_config_exit(spa, SCL_ALL, FTAG);

	txg_wait_synced(spa_get_dsl(spa), 0);
	}

	static boolean_t
	spa_has_aux_vdev(spa_t spa, uint64_t guid, spa_aux_vdev_t sav)
	{
	(void) spa;
	int i;
	uint64_t vdev_guid;

	for (i = 0; i < sav->sav_count; i++)
	if (sav->sav_vdevs[i]->vdev_guid == guid)
	return (B_TRUE);

	for (i = 0; i < sav->sav_npending; i++) {
	if (nvlist_lookup_uint64(sav->sav_pending[i], ZPOOL_CONFIG_GUID,
	&vdev_guid) == 0 && vdev_guid == guid)
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	boolean_t
	spa_has_l2cache(spa_t *spa, uint64_t guid)
	{
	return (spa_has_aux_vdev(spa, guid, &spa->spa_l2cache));
	}

	boolean_t
	spa_has_spare(spa_t *spa, uint64_t guid)
	{
	return (spa_has_aux_vdev(spa, guid, &spa->spa_spares));
	}

	/*
	* Check if a pool has an active shared spare device.
	* Note: reference count of an active spare is 2, as a spare and as a replace
	*/
	static boolean_t
	spa_has_active_shared_spare(spa_t *spa)
	{
	int i, refcnt;
	uint64_t pool;
	spa_aux_vdev_t *sav = &spa->spa_spares;

	for (i = 0; i < sav->sav_count; i++) {
	if (spa_spare_exists(sav->sav_vdevs[i]->vdev_guid, &pool,
	&refcnt) && pool != 0ULL && pool == spa_guid(spa) &&
	refcnt > 2)
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	uint64_t
	spa_total_metaslabs(spa_t *spa)
	{
	vdev_t *rvd = spa->spa_root_vdev;

	uint64_t m = 0;
	for (uint64_t c = 0; c < rvd->vdev_children; c++) {
	vdev_t *vd = rvd->vdev_child[c];
	if (!vdev_is_concrete(vd))
	continue;
	m += vd->vdev_ms_count;
	}
	return (m);
	}

	/*
	* Notify any waiting threads that some activity has switched from being in-
	* progress to not-in-progress so that the thread can wake up and determine
	* whether it is finished waiting.
	*/
	void
	spa_notify_waiters(spa_t *spa)
	{
	/*
	* Acquiring spa_activities_lock here prevents the cv_broadcast from
	* happening between the waiting thread's check and cv_wait.
	*/
	mutex_enter(&spa->spa_activities_lock);
	cv_broadcast(&spa->spa_activities_cv);
	mutex_exit(&spa->spa_activities_lock);
	}

	/*
	* Notify any waiting threads that the pool is exporting, and then block until
	* they are finished using the spa_t.
	*/
	void
	spa_wake_waiters(spa_t *spa)
	{
	mutex_enter(&spa->spa_activities_lock);
	spa->spa_waiters_cancel = B_TRUE;
	cv_broadcast(&spa->spa_activities_cv);
	while (spa->spa_waiters != 0)
	cv_wait(&spa->spa_waiters_cv, &spa->spa_activities_lock);
	spa->spa_waiters_cancel = B_FALSE;
	mutex_exit(&spa->spa_activities_lock);
	}

	/* Whether the vdev or any of its descendants are being initialized/trimmed. */
	static boolean_t
	spa_vdev_activity_in_progress_impl(vdev_t *vd, zpool_wait_activity_t activity)
	{
	spa_t *spa = vd->vdev_spa;

	ASSERT(spa_config_held(spa, SCL_CONFIG \| SCL_STATE, RW_READER));
	ASSERT(MUTEX_HELD(&spa->spa_activities_lock));
	ASSERT(activity == ZPOOL_WAIT_INITIALIZE \|\|
	activity == ZPOOL_WAIT_TRIM);

	kmutex_t *lock = activity == ZPOOL_WAIT_INITIALIZE ?
	&vd->vdev_initialize_lock : &vd->vdev_trim_lock;

	mutex_exit(&spa->spa_activities_lock);
	mutex_enter(lock);
	mutex_enter(&spa->spa_activities_lock);

	boolean_t in_progress = (activity == ZPOOL_WAIT_INITIALIZE) ?
	(vd->vdev_initialize_state == VDEV_INITIALIZE_ACTIVE) :
	(vd->vdev_trim_state == VDEV_TRIM_ACTIVE);
	mutex_exit(lock);

	if (in_progress)
	return (B_TRUE);

	for (int i = 0; i < vd->vdev_children; i++) {
	if (spa_vdev_activity_in_progress_impl(vd->vdev_child[i],
	activity))
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	/*
	* If use_guid is true, this checks whether the vdev specified by guid is
	* being initialized/trimmed. Otherwise, it checks whether any vdev in the pool
	* is being initialized/trimmed. The caller must hold the config lock and
	* spa_activities_lock.
	*/
	static int
	spa_vdev_activity_in_progress(spa_t *spa, boolean_t use_guid, uint64_t guid,
	zpool_wait_activity_t activity, boolean_t *in_progress)
	{
	mutex_exit(&spa->spa_activities_lock);
	spa_config_enter(spa, SCL_CONFIG \| SCL_STATE, FTAG, RW_READER);
	mutex_enter(&spa->spa_activities_lock);

	vdev_t *vd;
	if (use_guid) {
	vd = spa_lookup_by_guid(spa, guid, B_FALSE);
	if (vd == NULL \|\| !vd->vdev_ops->vdev_op_leaf) {
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	return (EINVAL);
	}
	} else {
	vd = spa->spa_root_vdev;
	}

	*in_progress = spa_vdev_activity_in_progress_impl(vd, activity);

	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	return (0);
	}

	/*
	* Locking for waiting threads
	* ---------------------------
	*
	* Waiting threads need a way to check whether a given activity is in progress,
	* and then, if it is, wait for it to complete. Each activity will have some
	* in-memory representation of the relevant on-disk state which can be used to
	* determine whether or not the activity is in progress. The in-memory state and
	* the locking used to protect it will be different for each activity, and may
	* not be suitable for use with a cvar (e.g., some state is protected by the
	* config lock). To allow waiting threads to wait without any races, another
	* lock, spa_activities_lock, is used.
	*
	* When the state is checked, both the activity-specific lock (if there is one)
	* and spa_activities_lock are held. In some cases, the activity-specific lock
	* is acquired explicitly (e.g. the config lock). In others, the locking is
	* internal to some check (e.g. bpobj_is_empty). After checking, the waiting
	* thread releases the activity-specific lock and, if the activity is in
	* progress, then cv_waits using spa_activities_lock.
	*
	* The waiting thread is woken when another thread, one completing some
	* activity, updates the state of the activity and then calls
	* spa_notify_waiters, which will cv_broadcast. This 'completing' thread only
	* needs to hold its activity-specific lock when updating the state, and this
	* lock can (but doesn't have to) be dropped before calling spa_notify_waiters.
	*
	* Because spa_notify_waiters acquires spa_activities_lock before broadcasting,
	* and because it is held when the waiting thread checks the state of the
	* activity, it can never be the case that the completing thread both updates
	* the activity state and cv_broadcasts in between the waiting thread's check
	* and cv_wait. Thus, a waiting thread can never miss a wakeup.
	*
	* In order to prevent deadlock, when the waiting thread does its check, in some
	* cases it will temporarily drop spa_activities_lock in order to acquire the
	* activity-specific lock. The order in which spa_activities_lock and the
	* activity specific lock are acquired in the waiting thread is determined by
	* the order in which they are acquired in the completing thread; if the
	* completing thread calls spa_notify_waiters with the activity-specific lock
	* held, then the waiting thread must also acquire the activity-specific lock
	* first.
	*/

	static int
	spa_activity_in_progress(spa_t *spa, zpool_wait_activity_t activity,
	boolean_t use_tag, uint64_t tag, boolean_t *in_progress)
	{
	int error = 0;

	ASSERT(MUTEX_HELD(&spa->spa_activities_lock));

	switch (activity) {
	case ZPOOL_WAIT_CKPT_DISCARD:
	*in_progress =
	(spa_feature_is_active(spa, SPA_FEATURE_POOL_CHECKPOINT) &&
	zap_contains(spa_meta_objset(spa),
	DMU_POOL_DIRECTORY_OBJECT, DMU_POOL_ZPOOL_CHECKPOINT) ==
	ENOENT);
	break;
	case ZPOOL_WAIT_FREE:
	*in_progress = ((spa_version(spa) >= SPA_VERSION_DEADLISTS &&
	!bpobj_is_empty(&spa->spa_dsl_pool->dp_free_bpobj)) \|\|
	spa_feature_is_active(spa, SPA_FEATURE_ASYNC_DESTROY) \|\|
	spa_livelist_delete_check(spa));
	break;
	case ZPOOL_WAIT_INITIALIZE:
	case ZPOOL_WAIT_TRIM:
	error = spa_vdev_activity_in_progress(spa, use_tag, tag,
	activity, in_progress);
	break;
	case ZPOOL_WAIT_REPLACE:
	mutex_exit(&spa->spa_activities_lock);
	spa_config_enter(spa, SCL_CONFIG \| SCL_STATE, FTAG, RW_READER);
	mutex_enter(&spa->spa_activities_lock);

	*in_progress = vdev_replace_in_progress(spa->spa_root_vdev);
	spa_config_exit(spa, SCL_CONFIG \| SCL_STATE, FTAG);
	break;
	case ZPOOL_WAIT_REMOVE:
	*in_progress = (spa->spa_removing_phys.sr_state ==
	DSS_SCANNING);
	break;
	case ZPOOL_WAIT_RESILVER:
	*in_progress = vdev_rebuild_active(spa->spa_root_vdev);
	if (*in_progress)
	break;
	zfs_fallthrough;
	case ZPOOL_WAIT_SCRUB:
	{
	boolean_t scanning, paused, is_scrub;
	dsl_scan_t *scn = spa->spa_dsl_pool->dp_scan;

	is_scrub = (scn->scn_phys.scn_func == POOL_SCAN_SCRUB);
	scanning = (scn->scn_phys.scn_state == DSS_SCANNING);
	paused = dsl_scan_is_paused_scrub(scn);
	*in_progress = (scanning && !paused &&
	is_scrub == (activity == ZPOOL_WAIT_SCRUB));
	break;
	}
	case ZPOOL_WAIT_RAIDZ_EXPAND:
	{
	vdev_raidz_expand_t *vre = spa->spa_raidz_expand;
	*in_progress = (vre != NULL && vre->vre_state == DSS_SCANNING);
	break;
	}
	default:
	panic("unrecognized value for activity %d", activity);
	}

	return (error);
	}

	static int
	spa_wait_common(const char *pool, zpool_wait_activity_t activity,
	boolean_t use_tag, uint64_t tag, boolean_t *waited)
	{
	/*
	* The tag is used to distinguish between instances of an activity.
	* 'initialize' and 'trim' are the only activities that we use this for.
	* The other activities can only have a single instance in progress in a
	* pool at one time, making the tag unnecessary.
	*
	* There can be multiple devices being replaced at once, but since they
	* all finish once resilvering finishes, we don't bother keeping track
	* of them individually, we just wait for them all to finish.
	*/
	if (use_tag && activity != ZPOOL_WAIT_INITIALIZE &&
	activity != ZPOOL_WAIT_TRIM)
	return (EINVAL);

	if (activity < 0 \|\| activity >= ZPOOL_WAIT_NUM_ACTIVITIES)
	return (EINVAL);

	spa_t *spa;
	int error = spa_open(pool, &spa, FTAG);
	if (error != 0)
	return (error);

	/*
	* Increment the spa's waiter count so that we can call spa_close and
	* still ensure that the spa_t doesn't get freed before this thread is
	* finished with it when the pool is exported. We want to call spa_close
	* before we start waiting because otherwise the additional ref would
	* prevent the pool from being exported or destroyed throughout the
	* potentially long wait.
	*/
	mutex_enter(&spa->spa_activities_lock);
	spa->spa_waiters++;
	spa_close(spa, FTAG);

	*waited = B_FALSE;
	for (;;) {
	boolean_t in_progress;
	error = spa_activity_in_progress(spa, activity, use_tag, tag,
	&in_progress);

	if (error \|\| !in_progress \|\| spa->spa_waiters_cancel)
	break;

	*waited = B_TRUE;

	if (cv_wait_sig(&spa->spa_activities_cv,
	&spa->spa_activities_lock) == 0) {
	error = EINTR;
	break;
	}
	}

	spa->spa_waiters--;
	cv_signal(&spa->spa_waiters_cv);
	mutex_exit(&spa->spa_activities_lock);

	return (error);
	}

	/*
	* Wait for a particular instance of the specified activity to complete, where
	* the instance is identified by 'tag'
	*/
	int
	spa_wait_tag(const char *pool, zpool_wait_activity_t activity, uint64_t tag,
	boolean_t *waited)
	{
	return (spa_wait_common(pool, activity, B_TRUE, tag, waited));
	}

	/*
	* Wait for all instances of the specified activity complete
	*/
	int
	spa_wait(const char pool, zpool_wait_activity_t activity, boolean_t waited)
	{

	return (spa_wait_common(pool, activity, B_FALSE, 0, waited));
	}

	sysevent_t *
	spa_event_create(spa_t spa, vdev_t vd, nvlist_t hist_nvl, const char name)
	{
	sysevent_t *ev = NULL;
	#ifdef _KERNEL
	nvlist_t *resource;

	resource = zfs_event_create(spa, vd, FM_SYSEVENT_CLASS, name, hist_nvl);
	if (resource) {
	ev = kmem_alloc(sizeof (sysevent_t), KM_SLEEP);
	ev->resource = resource;
	}
	#else
	(void) spa, (void) vd, (void) hist_nvl, (void) name;
	#endif
	return (ev);
	}

	void
	spa_event_post(sysevent_t *ev)
	{
	#ifdef _KERNEL
	if (ev) {
	zfs_zevent_post(ev->resource, NULL, zfs_zevent_post_cb);
	kmem_free(ev, sizeof (*ev));
	}
	#else
	(void) ev;
	#endif
	}

	/*
	* Post a zevent corresponding to the given sysevent. The 'name' must be one
	* of the event definitions in sys/sysevent/eventdefs.h. The payload will be
	* filled in from the spa and (optionally) the vdev. This doesn't do anything
	* in the userland libzpool, as we don't want consumers to misinterpret ztest
	* or zdb as real changes.
	*/
	void
	spa_event_notify(spa_t spa, vdev_t vd, nvlist_t hist_nvl, const char name)
	{
	spa_event_post(spa_event_create(spa, vd, hist_nvl, name));
	}

	/* state manipulation functions */
	EXPORT_SYMBOL(spa_open);
	EXPORT_SYMBOL(spa_open_rewind);
	EXPORT_SYMBOL(spa_get_stats);
	EXPORT_SYMBOL(spa_create);
	EXPORT_SYMBOL(spa_import);
	EXPORT_SYMBOL(spa_tryimport);
	EXPORT_SYMBOL(spa_destroy);
	EXPORT_SYMBOL(spa_export);
	EXPORT_SYMBOL(spa_reset);
	EXPORT_SYMBOL(spa_async_request);
	EXPORT_SYMBOL(spa_async_suspend);
	EXPORT_SYMBOL(spa_async_resume);
	EXPORT_SYMBOL(spa_inject_addref);
	EXPORT_SYMBOL(spa_inject_delref);
	EXPORT_SYMBOL(spa_scan_stat_init);
	EXPORT_SYMBOL(spa_scan_get_stats);

	/* device manipulation */
	EXPORT_SYMBOL(spa_vdev_add);
	EXPORT_SYMBOL(spa_vdev_attach);
	EXPORT_SYMBOL(spa_vdev_detach);
	EXPORT_SYMBOL(spa_vdev_setpath);
	EXPORT_SYMBOL(spa_vdev_setfru);
	EXPORT_SYMBOL(spa_vdev_split_mirror);

	/* spare statech is global across all pools) */
	EXPORT_SYMBOL(spa_spare_add);
	EXPORT_SYMBOL(spa_spare_remove);
	EXPORT_SYMBOL(spa_spare_exists);
	EXPORT_SYMBOL(spa_spare_activate);

	/* L2ARC statech is global across all pools) */
	EXPORT_SYMBOL(spa_l2cache_add);
	EXPORT_SYMBOL(spa_l2cache_remove);
	EXPORT_SYMBOL(spa_l2cache_exists);
	EXPORT_SYMBOL(spa_l2cache_activate);
	EXPORT_SYMBOL(spa_l2cache_drop);

	/* scanning */
	EXPORT_SYMBOL(spa_scan);
	EXPORT_SYMBOL(spa_scan_stop);

	/* spa syncing */
	EXPORT_SYMBOL(spa_sync); /* only for DMU use */
	EXPORT_SYMBOL(spa_sync_allpools);

	/* properties */
	EXPORT_SYMBOL(spa_prop_set);
	EXPORT_SYMBOL(spa_prop_get);
	EXPORT_SYMBOL(spa_prop_clear_bootfs);

	/* asynchronous event notification */
	EXPORT_SYMBOL(spa_event_notify);

	ZFS_MODULE_PARAM(zfs_metaslab, metaslab_, preload_pct, UINT, ZMOD_RW,
	"Percentage of CPUs to run a metaslab preload taskq");

	/* BEGIN CSTYLED */
	ZFS_MODULE_PARAM(zfs_spa, spa_, load_verify_shift, UINT, ZMOD_RW,
	"log2 fraction of arc that can be used by inflight I/Os when "
	"verifying pool during import");
	/* END CSTYLED */

	ZFS_MODULE_PARAM(zfs_spa, spa_, load_verify_metadata, INT, ZMOD_RW,
	"Set to traverse metadata on pool import");

	ZFS_MODULE_PARAM(zfs_spa, spa_, load_verify_data, INT, ZMOD_RW,
	"Set to traverse data on pool import");

	ZFS_MODULE_PARAM(zfs_spa, spa_, load_print_vdev_tree, INT, ZMOD_RW,
	"Print vdev tree to zfs_dbgmsg during pool import");

	ZFS_MODULE_PARAM(zfs_zio, zio_, taskq_batch_pct, UINT, ZMOD_RW,
	"Percentage of CPUs to run an IO worker thread");

	ZFS_MODULE_PARAM(zfs_zio, zio_, taskq_batch_tpq, UINT, ZMOD_RW,
	"Number of threads per IO worker taskqueue");

	/* BEGIN CSTYLED */
	ZFS_MODULE_PARAM(zfs, zfs_, max_missing_tvds, U64, ZMOD_RW,
	"Allow importing pool with up to this number of missing top-level "
	"vdevs (in read-only mode)");
	/* END CSTYLED */

	ZFS_MODULE_PARAM(zfs_livelist_condense, zfs_livelist_condense_, zthr_pause, INT,
	ZMOD_RW, "Set the livelist condense zthr to pause");

	ZFS_MODULE_PARAM(zfs_livelist_condense, zfs_livelist_condense_, sync_pause, INT,
	ZMOD_RW, "Set the livelist condense synctask to pause");

	/* BEGIN CSTYLED */
	ZFS_MODULE_PARAM(zfs_livelist_condense, zfs_livelist_condense_, sync_cancel,
	INT, ZMOD_RW,
	"Whether livelist condensing was canceled in the synctask");

	ZFS_MODULE_PARAM(zfs_livelist_condense, zfs_livelist_condense_, zthr_cancel,
	INT, ZMOD_RW,
	"Whether livelist condensing was canceled in the zthr function");

	ZFS_MODULE_PARAM(zfs_livelist_condense, zfs_livelist_condense_, new_alloc, INT,
	ZMOD_RW,
	"Whether extra ALLOC blkptrs were added to a livelist entry while it "
	"was being condensed");

	#ifdef _KERNEL
	ZFS_MODULE_VIRTUAL_PARAM_CALL(zfs_zio, zio_, taskq_read,
	spa_taskq_read_param_set, spa_taskq_read_param_get, ZMOD_RW,
	"Configure IO queues for read IO");
	ZFS_MODULE_VIRTUAL_PARAM_CALL(zfs_zio, zio_, taskq_write,
	spa_taskq_write_param_set, spa_taskq_write_param_get, ZMOD_RW,
	"Configure IO queues for write IO");
	#endif
	/* END CSTYLED */

	ZFS_MODULE_PARAM(zfs_zio, zio_, taskq_write_tpq, UINT, ZMOD_RW,
	"Number of CPUs per write issue taskq");
	diff --git a/sys/contrib/openzfs/module/zfs/spa_misc.c b/sys/contrib/openzfs/module/zfs/spa_misc.c
	index e6d4a9bdb29c..d1d41bbe7214 100644
	--- a/sys/contrib/openzfs/module/zfs/spa_misc.c
	+++ b/sys/contrib/openzfs/module/zfs/spa_misc.c
	@@ -1,3116 +1,3129 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2024 by Delphix. All rights reserved.
	* Copyright 2015 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
	* Copyright 2013 Saso Kiselkov. All rights reserved.
	* Copyright (c) 2017 Datto Inc.
	* Copyright (c) 2017, Intel Corporation.
	* Copyright (c) 2019, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
	- * Copyright (c) 2023, Klara Inc.
	+ * Copyright (c) 2023, 2024, Klara Inc.
	*/

	#include <sys/zfs_context.h>
	#include <sys/zfs_chksum.h>
	#include <sys/spa_impl.h>
	#include <sys/zio.h>
	#include <sys/zio_checksum.h>
	#include <sys/zio_compress.h>
	#include <sys/dmu.h>
	#include <sys/dmu_tx.h>
	#include <sys/zap.h>
	#include <sys/zil.h>
	#include <sys/vdev_impl.h>
	#include <sys/vdev_initialize.h>
	#include <sys/vdev_trim.h>
	#include <sys/vdev_file.h>
	#include <sys/vdev_raidz.h>
	#include <sys/metaslab.h>
	#include <sys/uberblock_impl.h>
	#include <sys/txg.h>
	#include <sys/avl.h>
	#include <sys/unique.h>
	#include <sys/dsl_pool.h>
	#include <sys/dsl_dir.h>
	#include <sys/dsl_prop.h>
	#include <sys/fm/util.h>
	#include <sys/dsl_scan.h>
	#include <sys/fs/zfs.h>
	#include <sys/metaslab_impl.h>
	#include <sys/arc.h>
	#include <sys/brt.h>
	#include <sys/ddt.h>
	#include <sys/kstat.h>
	#include "zfs_prop.h"
	#include <sys/btree.h>
	#include <sys/zfeature.h>
	#include <sys/qat.h>
	#include <sys/zstd/zstd.h>

	/*
	* SPA locking
	*
	* There are three basic locks for managing spa_t structures:
	*
	* spa_namespace_lock (global mutex)
	*
	* This lock must be acquired to do any of the following:
	*
	* - Lookup a spa_t by name
	* - Add or remove a spa_t from the namespace
	* - Increase spa_refcount from non-zero
	* - Check if spa_refcount is zero
	* - Rename a spa_t
	* - add/remove/attach/detach devices
	- * - Held for the duration of create/destroy/export
	- * - Held at the start and end of import
	+ * - Held for the duration of create/destroy
	+ * - Held at the start and end of import and export
	*
	* It does not need to handle recursion. A create or destroy may
	* reference objects (files or zvols) in other pools, but by
	* definition they must have an existing reference, and will never need
	* to lookup a spa_t by name.
	*
	* spa_refcount (per-spa zfs_refcount_t protected by mutex)
	*
	* This reference count keep track of any active users of the spa_t. The
	* spa_t cannot be destroyed or freed while this is non-zero. Internally,
	* the refcount is never really 'zero' - opening a pool implicitly keeps
	* some references in the DMU. Internally we check against spa_minref, but
	* present the image of a zero/non-zero value to consumers.
	*
	* spa_config_lock[] (per-spa array of rwlocks)
	*
	* This protects the spa_t from config changes, and must be held in
	* the following circumstances:
	*
	* - RW_READER to perform I/O to the spa
	* - RW_WRITER to change the vdev config
	*
	* The locking order is fairly straightforward:
	*
	* spa_namespace_lock -> spa_refcount
	*
	* The namespace lock must be acquired to increase the refcount from 0
	* or to check if it is zero.
	*
	* spa_refcount -> spa_config_lock[]
	*
	* There must be at least one valid reference on the spa_t to acquire
	* the config lock.
	*
	* spa_namespace_lock -> spa_config_lock[]
	*
	* The namespace lock must always be taken before the config lock.
	*
	*
	* The spa_namespace_lock can be acquired directly and is globally visible.
	*
	* The namespace is manipulated using the following functions, all of which
	* require the spa_namespace_lock to be held.
	*
	* spa_lookup() Lookup a spa_t by name.
	*
	* spa_add() Create a new spa_t in the namespace.
	*
	* spa_remove() Remove a spa_t from the namespace. This also
	* frees up any memory associated with the spa_t.
	*
	* spa_next() Returns the next spa_t in the system, or the
	* first if NULL is passed.
	*
	* spa_evict_all() Shutdown and remove all spa_t structures in
	* the system.
	*
	* spa_guid_exists() Determine whether a pool/device guid exists.
	*
	* The spa_refcount is manipulated using the following functions:
	*
	* spa_open_ref() Adds a reference to the given spa_t. Must be
	* called with spa_namespace_lock held if the
	* refcount is currently zero.
	*
	* spa_close() Remove a reference from the spa_t. This will
	* not free the spa_t or remove it from the
	* namespace. No locking is required.
	*
	* spa_refcount_zero() Returns true if the refcount is currently
	* zero. Must be called with spa_namespace_lock
	* held.
	*
	* The spa_config_lock[] is an array of rwlocks, ordered as follows:
	* SCL_CONFIG > SCL_STATE > SCL_ALLOC > SCL_ZIO > SCL_FREE > SCL_VDEV.
	* spa_config_lock[] is manipulated with spa_config_{enter,exit,held}().
	*
	* To read the configuration, it suffices to hold one of these locks as reader.
	* To modify the configuration, you must hold all locks as writer. To modify
	* vdev state without altering the vdev tree's topology (e.g. online/offline),
	* you must hold SCL_STATE and SCL_ZIO as writer.
	*
	* We use these distinct config locks to avoid recursive lock entry.
	* For example, spa_sync() (which holds SCL_CONFIG as reader) induces
	* block allocations (SCL_ALLOC), which may require reading space maps
	* from disk (dmu_read() -> zio_read() -> SCL_ZIO).
	*
	* The spa config locks cannot be normal rwlocks because we need the
	* ability to hand off ownership. For example, SCL_ZIO is acquired
	* by the issuing thread and later released by an interrupt thread.
	* They do, however, obey the usual write-wanted semantics to prevent
	* writer (i.e. system administrator) starvation.
	*
	* The lock acquisition rules are as follows:
	*
	* SCL_CONFIG
	* Protects changes to the vdev tree topology, such as vdev
	* add/remove/attach/detach. Protects the dirty config list
	* (spa_config_dirty_list) and the set of spares and l2arc devices.
	*
	* SCL_STATE
	* Protects changes to pool state and vdev state, such as vdev
	* online/offline/fault/degrade/clear. Protects the dirty state list
	* (spa_state_dirty_list) and global pool state (spa_state).
	*
	* SCL_ALLOC
	* Protects changes to metaslab groups and classes.
	* Held as reader by metaslab_alloc() and metaslab_claim().
	*
	* SCL_ZIO
	* Held by bp-level zios (those which have no io_vd upon entry)
	* to prevent changes to the vdev tree. The bp-level zio implicitly
	* protects all of its vdev child zios, which do not hold SCL_ZIO.
	*
	* SCL_FREE
	* Protects changes to metaslab groups and classes.
	* Held as reader by metaslab_free(). SCL_FREE is distinct from
	* SCL_ALLOC, and lower than SCL_ZIO, so that we can safely free
	* blocks in zio_done() while another i/o that holds either
	* SCL_ALLOC or SCL_ZIO is waiting for this i/o to complete.
	*
	* SCL_VDEV
	* Held as reader to prevent changes to the vdev tree during trivial
	* inquiries such as bp_get_dsize(). SCL_VDEV is distinct from the
	* other locks, and lower than all of them, to ensure that it's safe
	* to acquire regardless of caller context.
	*
	* In addition, the following rules apply:
	*
	* (a) spa_props_lock protects pool properties, spa_config and spa_config_list.
	* The lock ordering is SCL_CONFIG > spa_props_lock.
	*
	* (b) I/O operations on leaf vdevs. For any zio operation that takes
	* an explicit vdev_t argument -- such as zio_ioctl(), zio_read_phys(),
	* or zio_write_phys() -- the caller must ensure that the config cannot
	* cannot change in the interim, and that the vdev cannot be reopened.
	* SCL_STATE as reader suffices for both.
	*
	* The vdev configuration is protected by spa_vdev_enter() / spa_vdev_exit().
	*
	* spa_vdev_enter() Acquire the namespace lock and the config lock
	* for writing.
	*
	* spa_vdev_exit() Release the config lock, wait for all I/O
	* to complete, sync the updated configs to the
	* cache, and release the namespace lock.
	*
	* vdev state is protected by spa_vdev_state_enter() / spa_vdev_state_exit().
	* Like spa_vdev_enter/exit, these are convenience wrappers -- the actual
	* locking is, always, based on spa_namespace_lock and spa_config_lock[].
	*/

	avl_tree_t spa_namespace_avl;
	kmutex_t spa_namespace_lock;
	kcondvar_t spa_namespace_cv;
	static const int spa_max_replication_override = SPA_DVAS_PER_BP;

	static kmutex_t spa_spare_lock;
	static avl_tree_t spa_spare_avl;
	static kmutex_t spa_l2cache_lock;
	static avl_tree_t spa_l2cache_avl;

	spa_mode_t spa_mode_global = SPA_MODE_UNINIT;

	#ifdef ZFS_DEBUG
	/*
	* Everything except dprintf, set_error, spa, and indirect_remap is on
	* by default in debug builds.
	*/
	int zfs_flags = ~(ZFS_DEBUG_DPRINTF \| ZFS_DEBUG_SET_ERROR \|
	ZFS_DEBUG_INDIRECT_REMAP);
	#else
	int zfs_flags = 0;
	#endif

	/*
	* zfs_recover can be set to nonzero to attempt to recover from
	* otherwise-fatal errors, typically caused by on-disk corruption. When
	* set, calls to zfs_panic_recover() will turn into warning messages.
	* This should only be used as a last resort, as it typically results
	* in leaked space, or worse.
	*/
	int zfs_recover = B_FALSE;

	/*
	* If destroy encounters an EIO while reading metadata (e.g. indirect
	* blocks), space referenced by the missing metadata can not be freed.
	* Normally this causes the background destroy to become "stalled", as
	* it is unable to make forward progress. While in this stalled state,
	* all remaining space to free from the error-encountering filesystem is
	* "temporarily leaked". Set this flag to cause it to ignore the EIO,
	* permanently leak the space from indirect blocks that can not be read,
	* and continue to free everything else that it can.
	*
	* The default, "stalling" behavior is useful if the storage partially
	* fails (i.e. some but not all i/os fail), and then later recovers. In
	* this case, we will be able to continue pool operations while it is
	* partially failed, and when it recovers, we can continue to free the
	* space, with no leaks. However, note that this case is actually
	* fairly rare.
	*
	* Typically pools either (a) fail completely (but perhaps temporarily,
	* e.g. a top-level vdev going offline), or (b) have localized,
	* permanent errors (e.g. disk returns the wrong data due to bit flip or
	* firmware bug). In case (a), this setting does not matter because the
	* pool will be suspended and the sync thread will not be able to make
	* forward progress regardless. In case (b), because the error is
	* permanent, the best we can do is leak the minimum amount of space,
	* which is what setting this flag will do. Therefore, it is reasonable
	* for this flag to normally be set, but we chose the more conservative
	* approach of not setting it, so that there is no possibility of
	* leaking space in the "partial temporary" failure case.
	*/
	int zfs_free_leak_on_eio = B_FALSE;

	/*
	* Expiration time in milliseconds. This value has two meanings. First it is
	* used to determine when the spa_deadman() logic should fire. By default the
	* spa_deadman() will fire if spa_sync() has not completed in 600 seconds.
	* Secondly, the value determines if an I/O is considered "hung". Any I/O that
	* has not completed in zfs_deadman_synctime_ms is considered "hung" resulting
	* in one of three behaviors controlled by zfs_deadman_failmode.
	*/
	uint64_t zfs_deadman_synctime_ms = 600000UL; /* 10 min. */

	/*
	* This value controls the maximum amount of time zio_wait() will block for an
	* outstanding IO. By default this is 300 seconds at which point the "hung"
	* behavior will be applied as described for zfs_deadman_synctime_ms.
	*/
	uint64_t zfs_deadman_ziotime_ms = 300000UL; /* 5 min. */

	/*
	* Check time in milliseconds. This defines the frequency at which we check
	* for hung I/O.
	*/
	uint64_t zfs_deadman_checktime_ms = 60000UL; /* 1 min. */

	/*
	* By default the deadman is enabled.
	*/
	int zfs_deadman_enabled = B_TRUE;

	/*
	* Controls the behavior of the deadman when it detects a "hung" I/O.
	* Valid values are zfs_deadman_failmode=<wait\|continue\|panic>.
	*
	* wait - Wait for the "hung" I/O (default)
	* continue - Attempt to recover from a "hung" I/O
	* panic - Panic the system
	*/
	const char *zfs_deadman_failmode = "wait";

	/*
	* The worst case is single-sector max-parity RAID-Z blocks, in which
	* case the space requirement is exactly (VDEV_RAIDZ_MAXPARITY + 1)
	* times the size; so just assume that. Add to this the fact that
	* we can have up to 3 DVAs per bp, and one more factor of 2 because
	* the block may be dittoed with up to 3 DVAs by ddt_sync(). All together,
	* the worst case is:
	* (VDEV_RAIDZ_MAXPARITY + 1) * SPA_DVAS_PER_BP * 2 == 24
	*/
	uint_t spa_asize_inflation = 24;

	/*
	* Normally, we don't allow the last 3.2% (1/(2^spa_slop_shift)) of space in
	* the pool to be consumed (bounded by spa_max_slop). This ensures that we
	* don't run the pool completely out of space, due to unaccounted changes (e.g.
	* to the MOS). It also limits the worst-case time to allocate space. If we
	* have less than this amount of free space, most ZPL operations (e.g. write,
	* create) will return ENOSPC. The ZIL metaslabs (spa_embedded_log_class) are
	* also part of this 3.2% of space which can't be consumed by normal writes;
	* the slop space "proper" (spa_get_slop_space()) is decreased by the embedded
	* log space.
	*
	* Certain operations (e.g. file removal, most administrative actions) can
	* use half the slop space. They will only return ENOSPC if less than half
	* the slop space is free. Typically, once the pool has less than the slop
	* space free, the user will use these operations to free up space in the pool.
	* These are the operations that call dsl_pool_adjustedsize() with the netfree
	* argument set to TRUE.
	*
	* Operations that are almost guaranteed to free up space in the absence of
	* a pool checkpoint can use up to three quarters of the slop space
	* (e.g zfs destroy).
	*
	* A very restricted set of operations are always permitted, regardless of
	* the amount of free space. These are the operations that call
	* dsl_sync_task(ZFS_SPACE_CHECK_NONE). If these operations result in a net
	* increase in the amount of space used, it is possible to run the pool
	* completely out of space, causing it to be permanently read-only.
	*
	* Note that on very small pools, the slop space will be larger than
	* 3.2%, in an effort to have it be at least spa_min_slop (128MB),
	* but we never allow it to be more than half the pool size.
	*
	* Further, on very large pools, the slop space will be smaller than
	* 3.2%, to avoid reserving much more space than we actually need; bounded
	* by spa_max_slop (128GB).
	*
	* See also the comments in zfs_space_check_t.
	*/
	uint_t spa_slop_shift = 5;
	static const uint64_t spa_min_slop = 128ULL * 1024 * 1024;
	static const uint64_t spa_max_slop = 128ULL * 1024 * 1024 * 1024;

	/*
	* Number of allocators to use, per spa instance
	*/
	static int spa_num_allocators = 4;
	static int spa_cpus_per_allocator = 4;

	/*
	* Spa active allocator.
	* Valid values are zfs_active_allocator=<dynamic\|cursor\|new-dynamic>.
	*/
	const char *zfs_active_allocator = "dynamic";

	void
	spa_load_failed(spa_t spa, const char fmt, ...)
	{
	va_list adx;
	char buf[256];

	va_start(adx, fmt);
	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
	va_end(adx);

	zfs_dbgmsg("spa_load(%s, config %s): FAILED: %s", spa->spa_name,
	spa->spa_trust_config ? "trusted" : "untrusted", buf);
	}

	void
	spa_load_note(spa_t spa, const char fmt, ...)
	{
	va_list adx;
	char buf[256];

	va_start(adx, fmt);
	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
	va_end(adx);

	zfs_dbgmsg("spa_load(%s, config %s): %s", spa->spa_name,
	spa->spa_trust_config ? "trusted" : "untrusted", buf);

	spa_import_progress_set_notes_nolog(spa, "%s", buf);
	}

	/*
	* By default dedup and user data indirects land in the special class
	*/
	static int zfs_ddt_data_is_special = B_TRUE;
	static int zfs_user_indirect_is_special = B_TRUE;

	/*
	* The percentage of special class final space reserved for metadata only.
	* Once we allocate 100 - zfs_special_class_metadata_reserve_pct we only
	* let metadata into the class.
	*/
	static uint_t zfs_special_class_metadata_reserve_pct = 25;

	/*
	* ==========================================================================
	* SPA config locking
	* ==========================================================================
	*/
	static void
	spa_config_lock_init(spa_t *spa)
	{
	for (int i = 0; i < SCL_LOCKS; i++) {
	spa_config_lock_t *scl = &spa->spa_config_lock[i];
	mutex_init(&scl->scl_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&scl->scl_cv, NULL, CV_DEFAULT, NULL);
	scl->scl_writer = NULL;
	scl->scl_write_wanted = 0;
	scl->scl_count = 0;
	}
	}

	static void
	spa_config_lock_destroy(spa_t *spa)
	{
	for (int i = 0; i < SCL_LOCKS; i++) {
	spa_config_lock_t *scl = &spa->spa_config_lock[i];
	mutex_destroy(&scl->scl_lock);
	cv_destroy(&scl->scl_cv);
	ASSERT(scl->scl_writer == NULL);
	ASSERT(scl->scl_write_wanted == 0);
	ASSERT(scl->scl_count == 0);
	}
	}

	int
	spa_config_tryenter(spa_t spa, int locks, const void tag, krw_t rw)
	{
	for (int i = 0; i < SCL_LOCKS; i++) {
	spa_config_lock_t *scl = &spa->spa_config_lock[i];
	if (!(locks & (1 << i)))
	continue;
	mutex_enter(&scl->scl_lock);
	if (rw == RW_READER) {
	if (scl->scl_writer \|\| scl->scl_write_wanted) {
	mutex_exit(&scl->scl_lock);
	spa_config_exit(spa, locks & ((1 << i) - 1),
	tag);
	return (0);
	}
	} else {
	ASSERT(scl->scl_writer != curthread);
	if (scl->scl_count != 0) {
	mutex_exit(&scl->scl_lock);
	spa_config_exit(spa, locks & ((1 << i) - 1),
	tag);
	return (0);
	}
	scl->scl_writer = curthread;
	}
	scl->scl_count++;
	mutex_exit(&scl->scl_lock);
	}
	return (1);
	}

	static void
	spa_config_enter_impl(spa_t spa, int locks, const void tag, krw_t rw,
	int mmp_flag)
	{
	(void) tag;
	int wlocks_held = 0;

	ASSERT3U(SCL_LOCKS, <, sizeof (wlocks_held) * NBBY);

	for (int i = 0; i < SCL_LOCKS; i++) {
	spa_config_lock_t *scl = &spa->spa_config_lock[i];
	if (scl->scl_writer == curthread)
	wlocks_held \|= (1 << i);
	if (!(locks & (1 << i)))
	continue;
	mutex_enter(&scl->scl_lock);
	if (rw == RW_READER) {
	while (scl->scl_writer \|\|
	(!mmp_flag && scl->scl_write_wanted)) {
	cv_wait(&scl->scl_cv, &scl->scl_lock);
	}
	} else {
	ASSERT(scl->scl_writer != curthread);
	while (scl->scl_count != 0) {
	scl->scl_write_wanted++;
	cv_wait(&scl->scl_cv, &scl->scl_lock);
	scl->scl_write_wanted--;
	}
	scl->scl_writer = curthread;
	}
	scl->scl_count++;
	mutex_exit(&scl->scl_lock);
	}
	ASSERT3U(wlocks_held, <=, locks);
	}

	void
	spa_config_enter(spa_t spa, int locks, const void tag, krw_t rw)
	{
	spa_config_enter_impl(spa, locks, tag, rw, 0);
	}

	/*
	* The spa_config_enter_mmp() allows the mmp thread to cut in front of
	* outstanding write lock requests. This is needed since the mmp updates are
	* time sensitive and failure to service them promptly will result in a
	* suspended pool. This pool suspension has been seen in practice when there is
	* a single disk in a pool that is responding slowly and presumably about to
	* fail.
	*/

	void
	spa_config_enter_mmp(spa_t spa, int locks, const void tag, krw_t rw)
	{
	spa_config_enter_impl(spa, locks, tag, rw, 1);
	}

	void
	spa_config_exit(spa_t spa, int locks, const void tag)
	{
	(void) tag;
	for (int i = SCL_LOCKS - 1; i >= 0; i--) {
	spa_config_lock_t *scl = &spa->spa_config_lock[i];
	if (!(locks & (1 << i)))
	continue;
	mutex_enter(&scl->scl_lock);
	ASSERT(scl->scl_count > 0);
	if (--scl->scl_count == 0) {
	ASSERT(scl->scl_writer == NULL \|\|
	scl->scl_writer == curthread);
	scl->scl_writer = NULL; /* OK in either case */
	cv_broadcast(&scl->scl_cv);
	}
	mutex_exit(&scl->scl_lock);
	}
	}

	int
	spa_config_held(spa_t *spa, int locks, krw_t rw)
	{
	int locks_held = 0;

	for (int i = 0; i < SCL_LOCKS; i++) {
	spa_config_lock_t *scl = &spa->spa_config_lock[i];
	if (!(locks & (1 << i)))
	continue;
	if ((rw == RW_READER && scl->scl_count != 0) \|\|
	(rw == RW_WRITER && scl->scl_writer == curthread))
	locks_held \|= 1 << i;
	}

	return (locks_held);
	}

	/*
	* ==========================================================================
	* SPA namespace functions
	* ==========================================================================
	*/

	/*
	* Lookup the named spa_t in the AVL tree. The spa_namespace_lock must be held.
	* Returns NULL if no matching spa_t is found.
	*/
	spa_t *
	spa_lookup(const char *name)
	{
	static spa_t search; /* spa_t is large; don't allocate on stack */
	spa_t *spa;
	avl_index_t where;
	char *cp;

	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	retry:
	(void) strlcpy(search.spa_name, name, sizeof (search.spa_name));

	/*
	* If it's a full dataset name, figure out the pool name and
	* just use that.
	*/
	cp = strpbrk(search.spa_name, "/@#");
	if (cp != NULL)
	*cp = '\0';

	spa = avl_find(&spa_namespace_avl, &search, &where);
	if (spa == NULL)
	return (NULL);

	- if (spa->spa_load_thread != NULL &&
	- spa->spa_load_thread != curthread) {
	+ /*
	+ * Avoid racing with import/export, which don't hold the namespace
	+ * lock for their entire duration.
	+ */
	+ if ((spa->spa_load_thread != NULL &&
	+ spa->spa_load_thread != curthread) \|\|
	+ (spa->spa_export_thread != NULL &&
	+ spa->spa_export_thread != curthread)) {
	cv_wait(&spa_namespace_cv, &spa_namespace_lock);
	goto retry;
	}

	return (spa);
	}

	/*
	* Fires when spa_sync has not completed within zfs_deadman_synctime_ms.
	* If the zfs_deadman_enabled flag is set then it inspects all vdev queues
	* looking for potentially hung I/Os.
	*/
	void
	spa_deadman(void *arg)
	{
	spa_t *spa = arg;

	/* Disable the deadman if the pool is suspended. */
	if (spa_suspended(spa))
	return;

	zfs_dbgmsg("slow spa_sync: started %llu seconds ago, calls %llu",
	(gethrtime() - spa->spa_sync_starttime) / NANOSEC,
	(u_longlong_t)++spa->spa_deadman_calls);
	if (zfs_deadman_enabled)
	vdev_deadman(spa->spa_root_vdev, FTAG);

	spa->spa_deadman_tqid = taskq_dispatch_delay(system_delay_taskq,
	spa_deadman, spa, TQ_SLEEP, ddi_get_lbolt() +
	MSEC_TO_TICK(zfs_deadman_checktime_ms));
	}

	static int
	spa_log_sm_sort_by_txg(const void va, const void vb)
	{
	const spa_log_sm_t *a = va;
	const spa_log_sm_t *b = vb;

	return (TREE_CMP(a->sls_txg, b->sls_txg));
	}

	/*
	* Create an uninitialized spa_t with the given name. Requires
	* spa_namespace_lock. The caller must ensure that the spa_t doesn't already
	* exist by calling spa_lookup() first.
	*/
	spa_t *
	spa_add(const char name, nvlist_t config, const char *altroot)
	{
	spa_t *spa;
	spa_config_dirent_t *dp;

	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	spa = kmem_zalloc(sizeof (spa_t), KM_SLEEP);

	mutex_init(&spa->spa_async_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_errlist_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_errlog_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_evicting_os_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_history_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_proc_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_props_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_cksum_tmpls_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_scrub_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_suspend_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_vdev_top_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_feat_stats_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_flushed_ms_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa->spa_activities_lock, NULL, MUTEX_DEFAULT, NULL);

	cv_init(&spa->spa_async_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&spa->spa_evicting_os_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&spa->spa_proc_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&spa->spa_scrub_io_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&spa->spa_suspend_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&spa->spa_activities_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&spa->spa_waiters_cv, NULL, CV_DEFAULT, NULL);

	for (int t = 0; t < TXG_SIZE; t++)
	bplist_create(&spa->spa_free_bplist[t]);

	(void) strlcpy(spa->spa_name, name, sizeof (spa->spa_name));
	spa->spa_state = POOL_STATE_UNINITIALIZED;
	spa->spa_freeze_txg = UINT64_MAX;
	spa->spa_final_txg = UINT64_MAX;
	spa->spa_load_max_txg = UINT64_MAX;
	spa->spa_proc = &p0;
	spa->spa_proc_state = SPA_PROC_NONE;
	spa->spa_trust_config = B_TRUE;
	spa->spa_hostid = zone_get_hostid(NULL);

	spa->spa_deadman_synctime = MSEC2NSEC(zfs_deadman_synctime_ms);
	spa->spa_deadman_ziotime = MSEC2NSEC(zfs_deadman_ziotime_ms);
	spa_set_deadman_failmode(spa, zfs_deadman_failmode);
	spa_set_allocator(spa, zfs_active_allocator);

	zfs_refcount_create(&spa->spa_refcount);
	spa_config_lock_init(spa);
	spa_stats_init(spa);

	ASSERT(MUTEX_HELD(&spa_namespace_lock));
	avl_add(&spa_namespace_avl, spa);

	/*
	* Set the alternate root, if there is one.
	*/
	if (altroot)
	spa->spa_root = spa_strdup(altroot);

	/* Do not allow more allocators than fraction of CPUs. */
	spa->spa_alloc_count = MAX(MIN(spa_num_allocators,
	boot_ncpus / MAX(spa_cpus_per_allocator, 1)), 1);

	spa->spa_allocs = kmem_zalloc(spa->spa_alloc_count *
	sizeof (spa_alloc_t), KM_SLEEP);
	for (int i = 0; i < spa->spa_alloc_count; i++) {
	mutex_init(&spa->spa_allocs[i].spaa_lock, NULL, MUTEX_DEFAULT,
	NULL);
	avl_create(&spa->spa_allocs[i].spaa_tree, zio_bookmark_compare,
	sizeof (zio_t), offsetof(zio_t, io_queue_node.a));
	}
	if (spa->spa_alloc_count > 1) {
	spa->spa_allocs_use = kmem_zalloc(offsetof(spa_allocs_use_t,
	sau_inuse[spa->spa_alloc_count]), KM_SLEEP);
	mutex_init(&spa->spa_allocs_use->sau_lock, NULL, MUTEX_DEFAULT,
	NULL);
	}

	avl_create(&spa->spa_metaslabs_by_flushed, metaslab_sort_by_flushed,
	sizeof (metaslab_t), offsetof(metaslab_t, ms_spa_txg_node));
	avl_create(&spa->spa_sm_logs_by_txg, spa_log_sm_sort_by_txg,
	sizeof (spa_log_sm_t), offsetof(spa_log_sm_t, sls_node));
	list_create(&spa->spa_log_summary, sizeof (log_summary_entry_t),
	offsetof(log_summary_entry_t, lse_node));

	/*
	* Every pool starts with the default cachefile
	*/
	list_create(&spa->spa_config_list, sizeof (spa_config_dirent_t),
	offsetof(spa_config_dirent_t, scd_link));

	dp = kmem_zalloc(sizeof (spa_config_dirent_t), KM_SLEEP);
	dp->scd_path = altroot ? NULL : spa_strdup(spa_config_path);
	list_insert_head(&spa->spa_config_list, dp);

	VERIFY(nvlist_alloc(&spa->spa_load_info, NV_UNIQUE_NAME,
	KM_SLEEP) == 0);

	if (config != NULL) {
	nvlist_t *features;

	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_FEATURES_FOR_READ,
	&features) == 0) {
	VERIFY(nvlist_dup(features, &spa->spa_label_features,
	0) == 0);
	}

	VERIFY(nvlist_dup(config, &spa->spa_config, 0) == 0);
	}

	if (spa->spa_label_features == NULL) {
	VERIFY(nvlist_alloc(&spa->spa_label_features, NV_UNIQUE_NAME,
	KM_SLEEP) == 0);
	}

	spa->spa_min_ashift = INT_MAX;
	spa->spa_max_ashift = 0;
	spa->spa_min_alloc = INT_MAX;
	spa->spa_gcd_alloc = INT_MAX;

	/* Reset cached value */
	spa->spa_dedup_dspace = ~0ULL;

	/*
	* As a pool is being created, treat all features as disabled by
	* setting SPA_FEATURE_DISABLED for all entries in the feature
	* refcount cache.
	*/
	for (int i = 0; i < SPA_FEATURES; i++) {
	spa->spa_feat_refcount_cache[i] = SPA_FEATURE_DISABLED;
	}

	list_create(&spa->spa_leaf_list, sizeof (vdev_t),
	offsetof(vdev_t, vdev_leaf_node));

	return (spa);
	}

	/*
	* Removes a spa_t from the namespace, freeing up any memory used. Requires
	* spa_namespace_lock. This is called only after the spa_t has been closed and
	* deactivated.
	*/
	void
	spa_remove(spa_t *spa)
	{
	spa_config_dirent_t *dp;

	ASSERT(MUTEX_HELD(&spa_namespace_lock));
	ASSERT(spa_state(spa) == POOL_STATE_UNINITIALIZED);
	ASSERT3U(zfs_refcount_count(&spa->spa_refcount), ==, 0);
	ASSERT0(spa->spa_waiters);

	nvlist_free(spa->spa_config_splitting);

	avl_remove(&spa_namespace_avl, spa);

	if (spa->spa_root)
	spa_strfree(spa->spa_root);

	while ((dp = list_remove_head(&spa->spa_config_list)) != NULL) {
	if (dp->scd_path != NULL)
	spa_strfree(dp->scd_path);
	kmem_free(dp, sizeof (spa_config_dirent_t));
	}

	for (int i = 0; i < spa->spa_alloc_count; i++) {
	avl_destroy(&spa->spa_allocs[i].spaa_tree);
	mutex_destroy(&spa->spa_allocs[i].spaa_lock);
	}
	kmem_free(spa->spa_allocs, spa->spa_alloc_count *
	sizeof (spa_alloc_t));
	if (spa->spa_alloc_count > 1) {
	mutex_destroy(&spa->spa_allocs_use->sau_lock);
	kmem_free(spa->spa_allocs_use, offsetof(spa_allocs_use_t,
	sau_inuse[spa->spa_alloc_count]));
	}

	avl_destroy(&spa->spa_metaslabs_by_flushed);
	avl_destroy(&spa->spa_sm_logs_by_txg);
	list_destroy(&spa->spa_log_summary);
	list_destroy(&spa->spa_config_list);
	list_destroy(&spa->spa_leaf_list);

	nvlist_free(spa->spa_label_features);
	nvlist_free(spa->spa_load_info);
	nvlist_free(spa->spa_feat_stats);
	spa_config_set(spa, NULL);

	zfs_refcount_destroy(&spa->spa_refcount);

	spa_stats_destroy(spa);
	spa_config_lock_destroy(spa);

	for (int t = 0; t < TXG_SIZE; t++)
	bplist_destroy(&spa->spa_free_bplist[t]);

	zio_checksum_templates_free(spa);

	cv_destroy(&spa->spa_async_cv);
	cv_destroy(&spa->spa_evicting_os_cv);
	cv_destroy(&spa->spa_proc_cv);
	cv_destroy(&spa->spa_scrub_io_cv);
	cv_destroy(&spa->spa_suspend_cv);
	cv_destroy(&spa->spa_activities_cv);
	cv_destroy(&spa->spa_waiters_cv);

	mutex_destroy(&spa->spa_flushed_ms_lock);
	mutex_destroy(&spa->spa_async_lock);
	mutex_destroy(&spa->spa_errlist_lock);
	mutex_destroy(&spa->spa_errlog_lock);
	mutex_destroy(&spa->spa_evicting_os_lock);
	mutex_destroy(&spa->spa_history_lock);
	mutex_destroy(&spa->spa_proc_lock);
	mutex_destroy(&spa->spa_props_lock);
	mutex_destroy(&spa->spa_cksum_tmpls_lock);
	mutex_destroy(&spa->spa_scrub_lock);
	mutex_destroy(&spa->spa_suspend_lock);
	mutex_destroy(&spa->spa_vdev_top_lock);
	mutex_destroy(&spa->spa_feat_stats_lock);
	mutex_destroy(&spa->spa_activities_lock);

	kmem_free(spa, sizeof (spa_t));
	}

	/*
	* Given a pool, return the next pool in the namespace, or NULL if there is
	* none. If 'prev' is NULL, return the first pool.
	*/
	spa_t *
	spa_next(spa_t *prev)
	{
	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	if (prev)
	return (AVL_NEXT(&spa_namespace_avl, prev));
	else
	return (avl_first(&spa_namespace_avl));
	}

	/*
	* ==========================================================================
	* SPA refcount functions
	* ==========================================================================
	*/

	/*
	* Add a reference to the given spa_t. Must have at least one reference, or
	* have the namespace lock held.
	*/
	void
	spa_open_ref(spa_t spa, const void tag)
	{
	ASSERT(zfs_refcount_count(&spa->spa_refcount) >= spa->spa_minref \|\|
	MUTEX_HELD(&spa_namespace_lock) \|\|
	spa->spa_load_thread == curthread);
	(void) zfs_refcount_add(&spa->spa_refcount, tag);
	}

	/*
	* Remove a reference to the given spa_t. Must have at least one reference, or
	- * have the namespace lock held.
	+ * have the namespace lock held or be part of a pool import/export.
	*/
	void
	spa_close(spa_t spa, const void tag)
	{
	ASSERT(zfs_refcount_count(&spa->spa_refcount) > spa->spa_minref \|\|
	MUTEX_HELD(&spa_namespace_lock) \|\|
	- spa->spa_load_thread == curthread);
	+ spa->spa_load_thread == curthread \|\|
	+ spa->spa_export_thread == curthread);
	(void) zfs_refcount_remove(&spa->spa_refcount, tag);
	}

	/*
	* Remove a reference to the given spa_t held by a dsl dir that is
	* being asynchronously released. Async releases occur from a taskq
	* performing eviction of dsl datasets and dirs. The namespace lock
	* isn't held and the hold by the object being evicted may contribute to
	* spa_minref (e.g. dataset or directory released during pool export),
	* so the asserts in spa_close() do not apply.
	*/
	void
	spa_async_close(spa_t spa, const void tag)
	{
	(void) zfs_refcount_remove(&spa->spa_refcount, tag);
	}

	/*
	* Check to see if the spa refcount is zero. Must be called with
	- * spa_namespace_lock held. We really compare against spa_minref, which is the
	- * number of references acquired when opening a pool
	+ * spa_namespace_lock held or be the spa export thread. We really
	+ * compare against spa_minref, which is the number of references
	+ * acquired when opening a pool
	*/
	boolean_t
	spa_refcount_zero(spa_t *spa)
	{
	- ASSERT(MUTEX_HELD(&spa_namespace_lock));
	+ ASSERT(MUTEX_HELD(&spa_namespace_lock) \|\|
	+ spa->spa_export_thread == curthread);

	return (zfs_refcount_count(&spa->spa_refcount) == spa->spa_minref);
	}

	/*
	* ==========================================================================
	* SPA spare and l2cache tracking
	* ==========================================================================
	*/

	/*
	* Hot spares and cache devices are tracked using the same code below,
	* for 'auxiliary' devices.
	*/

	typedef struct spa_aux {
	uint64_t aux_guid;
	uint64_t aux_pool;
	avl_node_t aux_avl;
	int aux_count;
	} spa_aux_t;

	static inline int
	spa_aux_compare(const void a, const void b)
	{
	const spa_aux_t sa = (const spa_aux_t )a;
	const spa_aux_t sb = (const spa_aux_t )b;

	return (TREE_CMP(sa->aux_guid, sb->aux_guid));
	}

	static void
	spa_aux_add(vdev_t vd, avl_tree_t avl)
	{
	avl_index_t where;
	spa_aux_t search;
	spa_aux_t *aux;

	search.aux_guid = vd->vdev_guid;
	if ((aux = avl_find(avl, &search, &where)) != NULL) {
	aux->aux_count++;
	} else {
	aux = kmem_zalloc(sizeof (spa_aux_t), KM_SLEEP);
	aux->aux_guid = vd->vdev_guid;
	aux->aux_count = 1;
	avl_insert(avl, aux, where);
	}
	}

	static void
	spa_aux_remove(vdev_t vd, avl_tree_t avl)
	{
	spa_aux_t search;
	spa_aux_t *aux;
	avl_index_t where;

	search.aux_guid = vd->vdev_guid;
	aux = avl_find(avl, &search, &where);

	ASSERT(aux != NULL);

	if (--aux->aux_count == 0) {
	avl_remove(avl, aux);
	kmem_free(aux, sizeof (spa_aux_t));
	} else if (aux->aux_pool == spa_guid(vd->vdev_spa)) {
	aux->aux_pool = 0ULL;
	}
	}

	static boolean_t
	spa_aux_exists(uint64_t guid, uint64_t pool, int refcnt, avl_tree_t *avl)
	{
	spa_aux_t search, *found;

	search.aux_guid = guid;
	found = avl_find(avl, &search, NULL);

	if (pool) {
	if (found)
	*pool = found->aux_pool;
	else
	*pool = 0ULL;
	}

	if (refcnt) {
	if (found)
	*refcnt = found->aux_count;
	else
	*refcnt = 0;
	}

	return (found != NULL);
	}

	static void
	spa_aux_activate(vdev_t vd, avl_tree_t avl)
	{
	spa_aux_t search, *found;
	avl_index_t where;

	search.aux_guid = vd->vdev_guid;
	found = avl_find(avl, &search, &where);
	ASSERT(found != NULL);
	ASSERT(found->aux_pool == 0ULL);

	found->aux_pool = spa_guid(vd->vdev_spa);
	}

	/*
	* Spares are tracked globally due to the following constraints:
	*
	* - A spare may be part of multiple pools.
	* - A spare may be added to a pool even if it's actively in use within
	* another pool.
	* - A spare in use in any pool can only be the source of a replacement if
	* the target is a spare in the same pool.
	*
	* We keep track of all spares on the system through the use of a reference
	* counted AVL tree. When a vdev is added as a spare, or used as a replacement
	* spare, then we bump the reference count in the AVL tree. In addition, we set
	* the 'vdev_isspare' member to indicate that the device is a spare (active or
	* inactive). When a spare is made active (used to replace a device in the
	* pool), we also keep track of which pool its been made a part of.
	*
	* The 'spa_spare_lock' protects the AVL tree. These functions are normally
	* called under the spa_namespace lock as part of vdev reconfiguration. The
	* separate spare lock exists for the status query path, which does not need to
	* be completely consistent with respect to other vdev configuration changes.
	*/

	static int
	spa_spare_compare(const void a, const void b)
	{
	return (spa_aux_compare(a, b));
	}

	void
	spa_spare_add(vdev_t *vd)
	{
	mutex_enter(&spa_spare_lock);
	ASSERT(!vd->vdev_isspare);
	spa_aux_add(vd, &spa_spare_avl);
	vd->vdev_isspare = B_TRUE;
	mutex_exit(&spa_spare_lock);
	}

	void
	spa_spare_remove(vdev_t *vd)
	{
	mutex_enter(&spa_spare_lock);
	ASSERT(vd->vdev_isspare);
	spa_aux_remove(vd, &spa_spare_avl);
	vd->vdev_isspare = B_FALSE;
	mutex_exit(&spa_spare_lock);
	}

	boolean_t
	spa_spare_exists(uint64_t guid, uint64_t pool, int refcnt)
	{
	boolean_t found;

	mutex_enter(&spa_spare_lock);
	found = spa_aux_exists(guid, pool, refcnt, &spa_spare_avl);
	mutex_exit(&spa_spare_lock);

	return (found);
	}

	void
	spa_spare_activate(vdev_t *vd)
	{
	mutex_enter(&spa_spare_lock);
	ASSERT(vd->vdev_isspare);
	spa_aux_activate(vd, &spa_spare_avl);
	mutex_exit(&spa_spare_lock);
	}

	/*
	* Level 2 ARC devices are tracked globally for the same reasons as spares.
	* Cache devices currently only support one pool per cache device, and so
	* for these devices the aux reference count is currently unused beyond 1.
	*/

	static int
	spa_l2cache_compare(const void a, const void b)
	{
	return (spa_aux_compare(a, b));
	}

	void
	spa_l2cache_add(vdev_t *vd)
	{
	mutex_enter(&spa_l2cache_lock);
	ASSERT(!vd->vdev_isl2cache);
	spa_aux_add(vd, &spa_l2cache_avl);
	vd->vdev_isl2cache = B_TRUE;
	mutex_exit(&spa_l2cache_lock);
	}

	void
	spa_l2cache_remove(vdev_t *vd)
	{
	mutex_enter(&spa_l2cache_lock);
	ASSERT(vd->vdev_isl2cache);
	spa_aux_remove(vd, &spa_l2cache_avl);
	vd->vdev_isl2cache = B_FALSE;
	mutex_exit(&spa_l2cache_lock);
	}

	boolean_t
	spa_l2cache_exists(uint64_t guid, uint64_t *pool)
	{
	boolean_t found;

	mutex_enter(&spa_l2cache_lock);
	found = spa_aux_exists(guid, pool, NULL, &spa_l2cache_avl);
	mutex_exit(&spa_l2cache_lock);

	return (found);
	}

	void
	spa_l2cache_activate(vdev_t *vd)
	{
	mutex_enter(&spa_l2cache_lock);
	ASSERT(vd->vdev_isl2cache);
	spa_aux_activate(vd, &spa_l2cache_avl);
	mutex_exit(&spa_l2cache_lock);
	}

	/*
	* ==========================================================================
	* SPA vdev locking
	* ==========================================================================
	*/

	/*
	* Lock the given spa_t for the purpose of adding or removing a vdev.
	* Grabs the global spa_namespace_lock plus the spa config lock for writing.
	* It returns the next transaction group for the spa_t.
	*/
	uint64_t
	spa_vdev_enter(spa_t *spa)
	{
	mutex_enter(&spa->spa_vdev_top_lock);
	mutex_enter(&spa_namespace_lock);

	+ ASSERT0(spa->spa_export_thread);
	+
	vdev_autotrim_stop_all(spa);

	return (spa_vdev_config_enter(spa));
	}

	/*
	* The same as spa_vdev_enter() above but additionally takes the guid of
	* the vdev being detached. When there is a rebuild in process it will be
	* suspended while the vdev tree is modified then resumed by spa_vdev_exit().
	* The rebuild is canceled if only a single child remains after the detach.
	*/
	uint64_t
	spa_vdev_detach_enter(spa_t *spa, uint64_t guid)
	{
	mutex_enter(&spa->spa_vdev_top_lock);
	mutex_enter(&spa_namespace_lock);

	+ ASSERT0(spa->spa_export_thread);
	+
	vdev_autotrim_stop_all(spa);

	if (guid != 0) {
	vdev_t *vd = spa_lookup_by_guid(spa, guid, B_FALSE);
	if (vd) {
	vdev_rebuild_stop_wait(vd->vdev_top);
	}
	}

	return (spa_vdev_config_enter(spa));
	}

	/*
	* Internal implementation for spa_vdev_enter(). Used when a vdev
	* operation requires multiple syncs (i.e. removing a device) while
	* keeping the spa_namespace_lock held.
	*/
	uint64_t
	spa_vdev_config_enter(spa_t *spa)
	{
	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	spa_config_enter(spa, SCL_ALL, spa, RW_WRITER);

	return (spa_last_synced_txg(spa) + 1);
	}

	/*
	* Used in combination with spa_vdev_config_enter() to allow the syncing
	* of multiple transactions without releasing the spa_namespace_lock.
	*/
	void
	spa_vdev_config_exit(spa_t spa, vdev_t vd, uint64_t txg, int error,
	const char *tag)
	{
	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	int config_changed = B_FALSE;

	ASSERT(txg > spa_last_synced_txg(spa));

	spa->spa_pending_vdev = NULL;

	/*
	* Reassess the DTLs.
	*/
	vdev_dtl_reassess(spa->spa_root_vdev, 0, 0, B_FALSE, B_FALSE);

	if (error == 0 && !list_is_empty(&spa->spa_config_dirty_list)) {
	config_changed = B_TRUE;
	spa->spa_config_generation++;
	}

	/*
	* Verify the metaslab classes.
	*/
	ASSERT(metaslab_class_validate(spa_normal_class(spa)) == 0);
	ASSERT(metaslab_class_validate(spa_log_class(spa)) == 0);
	ASSERT(metaslab_class_validate(spa_embedded_log_class(spa)) == 0);
	ASSERT(metaslab_class_validate(spa_special_class(spa)) == 0);
	ASSERT(metaslab_class_validate(spa_dedup_class(spa)) == 0);

	spa_config_exit(spa, SCL_ALL, spa);

	/*
	* Panic the system if the specified tag requires it. This
	* is useful for ensuring that configurations are updated
	* transactionally.
	*/
	if (zio_injection_enabled)
	zio_handle_panic_injection(spa, tag, 0);

	/*
	* Note: this txg_wait_synced() is important because it ensures
	* that there won't be more than one config change per txg.
	* This allows us to use the txg as the generation number.
	*/
	if (error == 0)
	txg_wait_synced(spa->spa_dsl_pool, txg);

	if (vd != NULL) {
	ASSERT(!vd->vdev_detached \|\| vd->vdev_dtl_sm == NULL);
	if (vd->vdev_ops->vdev_op_leaf) {
	mutex_enter(&vd->vdev_initialize_lock);
	vdev_initialize_stop(vd, VDEV_INITIALIZE_CANCELED,
	NULL);
	mutex_exit(&vd->vdev_initialize_lock);

	mutex_enter(&vd->vdev_trim_lock);
	vdev_trim_stop(vd, VDEV_TRIM_CANCELED, NULL);
	mutex_exit(&vd->vdev_trim_lock);
	}

	/*
	* The vdev may be both a leaf and top-level device.
	*/
	vdev_autotrim_stop_wait(vd);

	spa_config_enter(spa, SCL_STATE_ALL, spa, RW_WRITER);
	vdev_free(vd);
	spa_config_exit(spa, SCL_STATE_ALL, spa);
	}

	/*
	* If the config changed, update the config cache.
	*/
	if (config_changed)
	spa_write_cachefile(spa, B_FALSE, B_TRUE, B_TRUE);
	}

	/*
	* Unlock the spa_t after adding or removing a vdev. Besides undoing the
	* locking of spa_vdev_enter(), we also want make sure the transactions have
	* synced to disk, and then update the global configuration cache with the new
	* information.
	*/
	int
	spa_vdev_exit(spa_t spa, vdev_t vd, uint64_t txg, int error)
	{
	vdev_autotrim_restart(spa);
	vdev_rebuild_restart(spa);

	spa_vdev_config_exit(spa, vd, txg, error, FTAG);
	mutex_exit(&spa_namespace_lock);
	mutex_exit(&spa->spa_vdev_top_lock);

	return (error);
	}

	/*
	* Lock the given spa_t for the purpose of changing vdev state.
	*/
	void
	spa_vdev_state_enter(spa_t *spa, int oplocks)
	{
	int locks = SCL_STATE_ALL \| oplocks;

	/*
	* Root pools may need to read of the underlying devfs filesystem
	* when opening up a vdev. Unfortunately if we're holding the
	* SCL_ZIO lock it will result in a deadlock when we try to issue
	* the read from the root filesystem. Instead we "prefetch"
	* the associated vnodes that we need prior to opening the
	* underlying devices and cache them so that we can prevent
	* any I/O when we are doing the actual open.
	*/
	if (spa_is_root(spa)) {
	int low = locks & ~(SCL_ZIO - 1);
	int high = locks & ~low;

	spa_config_enter(spa, high, spa, RW_WRITER);
	vdev_hold(spa->spa_root_vdev);
	spa_config_enter(spa, low, spa, RW_WRITER);
	} else {
	spa_config_enter(spa, locks, spa, RW_WRITER);
	}
	spa->spa_vdev_locks = locks;
	}

	int
	spa_vdev_state_exit(spa_t spa, vdev_t vd, int error)
	{
	boolean_t config_changed = B_FALSE;
	vdev_t *vdev_top;

	if (vd == NULL \|\| vd == spa->spa_root_vdev) {
	vdev_top = spa->spa_root_vdev;
	} else {
	vdev_top = vd->vdev_top;
	}

	if (vd != NULL \|\| error == 0)
	vdev_dtl_reassess(vdev_top, 0, 0, B_FALSE, B_FALSE);

	if (vd != NULL) {
	if (vd != spa->spa_root_vdev)
	vdev_state_dirty(vdev_top);

	config_changed = B_TRUE;
	spa->spa_config_generation++;
	}

	if (spa_is_root(spa))
	vdev_rele(spa->spa_root_vdev);

	ASSERT3U(spa->spa_vdev_locks, >=, SCL_STATE_ALL);
	spa_config_exit(spa, spa->spa_vdev_locks, spa);

	/*
	* If anything changed, wait for it to sync. This ensures that,
	* from the system administrator's perspective, zpool(8) commands
	* are synchronous. This is important for things like zpool offline:
	* when the command completes, you expect no further I/O from ZFS.
	*/
	if (vd != NULL)
	txg_wait_synced(spa->spa_dsl_pool, 0);

	/*
	* If the config changed, update the config cache.
	*/
	if (config_changed) {
	mutex_enter(&spa_namespace_lock);
	spa_write_cachefile(spa, B_FALSE, B_TRUE, B_FALSE);
	mutex_exit(&spa_namespace_lock);
	}

	return (error);
	}

	/*
	* ==========================================================================
	* Miscellaneous functions
	* ==========================================================================
	*/

	void
	spa_activate_mos_feature(spa_t spa, const char feature, dmu_tx_t *tx)
	{
	if (!nvlist_exists(spa->spa_label_features, feature)) {
	fnvlist_add_boolean(spa->spa_label_features, feature);
	/*
	* When we are creating the pool (tx_txg==TXG_INITIAL), we can't
	* dirty the vdev config because lock SCL_CONFIG is not held.
	* Thankfully, in this case we don't need to dirty the config
	* because it will be written out anyway when we finish
	* creating the pool.
	*/
	if (tx->tx_txg != TXG_INITIAL)
	vdev_config_dirty(spa->spa_root_vdev);
	}
	}

	void
	spa_deactivate_mos_feature(spa_t spa, const char feature)
	{
	if (nvlist_remove_all(spa->spa_label_features, feature) == 0)
	vdev_config_dirty(spa->spa_root_vdev);
	}

	/*
	* Return the spa_t associated with given pool_guid, if it exists. If
	* device_guid is non-zero, determine whether the pool exists and contains
	* a device with the specified device_guid.
	*/
	spa_t *
	spa_by_guid(uint64_t pool_guid, uint64_t device_guid)
	{
	spa_t *spa;
	avl_tree_t *t = &spa_namespace_avl;

	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	for (spa = avl_first(t); spa != NULL; spa = AVL_NEXT(t, spa)) {
	if (spa->spa_state == POOL_STATE_UNINITIALIZED)
	continue;
	if (spa->spa_root_vdev == NULL)
	continue;
	if (spa_guid(spa) == pool_guid) {
	if (device_guid == 0)
	break;

	if (vdev_lookup_by_guid(spa->spa_root_vdev,
	device_guid) != NULL)
	break;

	/*
	* Check any devices we may be in the process of adding.
	*/
	if (spa->spa_pending_vdev) {
	if (vdev_lookup_by_guid(spa->spa_pending_vdev,
	device_guid) != NULL)
	break;
	}
	}
	}

	return (spa);
	}

	/*
	* Determine whether a pool with the given pool_guid exists.
	*/
	boolean_t
	spa_guid_exists(uint64_t pool_guid, uint64_t device_guid)
	{
	return (spa_by_guid(pool_guid, device_guid) != NULL);
	}

	char *
	spa_strdup(const char *s)
	{
	size_t len;
	char *new;

	len = strlen(s);
	new = kmem_alloc(len + 1, KM_SLEEP);
	memcpy(new, s, len + 1);

	return (new);
	}

	void
	spa_strfree(char *s)
	{
	kmem_free(s, strlen(s) + 1);
	}

	uint64_t
	spa_generate_guid(spa_t *spa)
	{
	uint64_t guid;

	if (spa != NULL) {
	do {
	(void) random_get_pseudo_bytes((void *)&guid,
	sizeof (guid));
	} while (guid == 0 \|\| spa_guid_exists(spa_guid(spa), guid));
	} else {
	do {
	(void) random_get_pseudo_bytes((void *)&guid,
	sizeof (guid));
	} while (guid == 0 \|\| spa_guid_exists(guid, 0));
	}

	return (guid);
	}

	void
	snprintf_blkptr(char buf, size_t buflen, const blkptr_t bp)
	{
	char type[256];
	const char *checksum = NULL;
	const char *compress = NULL;

	if (bp != NULL) {
	if (BP_GET_TYPE(bp) & DMU_OT_NEWTYPE) {
	dmu_object_byteswap_t bswap =
	DMU_OT_BYTESWAP(BP_GET_TYPE(bp));
	(void) snprintf(type, sizeof (type), "bswap %s %s",
	DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) ?
	"metadata" : "data",
	dmu_ot_byteswap[bswap].ob_name);
	} else {
	(void) strlcpy(type, dmu_ot[BP_GET_TYPE(bp)].ot_name,
	sizeof (type));
	}
	if (!BP_IS_EMBEDDED(bp)) {
	checksum =
	zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_name;
	}
	compress = zio_compress_table[BP_GET_COMPRESS(bp)].ci_name;
	}

	SNPRINTF_BLKPTR(kmem_scnprintf, ' ', buf, buflen, bp, type, checksum,
	compress);
	}

	void
	spa_freeze(spa_t *spa)
	{
	uint64_t freeze_txg = 0;

	spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER);
	if (spa->spa_freeze_txg == UINT64_MAX) {
	freeze_txg = spa_last_synced_txg(spa) + TXG_SIZE;
	spa->spa_freeze_txg = freeze_txg;
	}
	spa_config_exit(spa, SCL_ALL, FTAG);
	if (freeze_txg != 0)
	txg_wait_synced(spa_get_dsl(spa), freeze_txg);
	}

	void
	zfs_panic_recover(const char *fmt, ...)
	{
	va_list adx;

	va_start(adx, fmt);
	vcmn_err(zfs_recover ? CE_WARN : CE_PANIC, fmt, adx);
	va_end(adx);
	}

	/*
	* This is a stripped-down version of strtoull, suitable only for converting
	* lowercase hexadecimal numbers that don't overflow.
	*/
	uint64_t
	zfs_strtonum(const char str, char *nptr)
	{
	uint64_t val = 0;
	char c;
	int digit;

	while ((c = *str) != '\0') {
	if (c >= '0' && c <= '9')
	digit = c - '0';
	else if (c >= 'a' && c <= 'f')
	digit = 10 + c - 'a';
	else
	break;

	val *= 16;
	val += digit;

	str++;
	}

	if (nptr)
	nptr = (char )str;

	return (val);
	}

	void
	spa_activate_allocation_classes(spa_t spa, dmu_tx_t tx)
	{
	/*
	* We bump the feature refcount for each special vdev added to the pool
	*/
	ASSERT(spa_feature_is_enabled(spa, SPA_FEATURE_ALLOCATION_CLASSES));
	spa_feature_incr(spa, SPA_FEATURE_ALLOCATION_CLASSES, tx);
	}

	/*
	* ==========================================================================
	* Accessor functions
	* ==========================================================================
	*/

	boolean_t
	spa_shutting_down(spa_t *spa)
	{
	return (spa->spa_async_suspended);
	}

	dsl_pool_t *
	spa_get_dsl(spa_t *spa)
	{
	return (spa->spa_dsl_pool);
	}

	boolean_t
	spa_is_initializing(spa_t *spa)
	{
	return (spa->spa_is_initializing);
	}

	boolean_t
	spa_indirect_vdevs_loaded(spa_t *spa)
	{
	return (spa->spa_indirect_vdevs_loaded);
	}

	blkptr_t *
	spa_get_rootblkptr(spa_t *spa)
	{
	return (&spa->spa_ubsync.ub_rootbp);
	}

	void
	spa_set_rootblkptr(spa_t spa, const blkptr_t bp)
	{
	spa->spa_uberblock.ub_rootbp = *bp;
	}

	void
	spa_altroot(spa_t spa, char buf, size_t buflen)
	{
	if (spa->spa_root == NULL)
	buf[0] = '\0';
	else
	(void) strlcpy(buf, spa->spa_root, buflen);
	}

	uint32_t
	spa_sync_pass(spa_t *spa)
	{
	return (spa->spa_sync_pass);
	}

	char *
	spa_name(spa_t *spa)
	{
	return (spa->spa_name);
	}

	uint64_t
	spa_guid(spa_t *spa)
	{
	dsl_pool_t *dp = spa_get_dsl(spa);
	uint64_t guid;

	/*
	* If we fail to parse the config during spa_load(), we can go through
	* the error path (which posts an ereport) and end up here with no root
	* vdev. We stash the original pool guid in 'spa_config_guid' to handle
	* this case.
	*/
	if (spa->spa_root_vdev == NULL)
	return (spa->spa_config_guid);

	guid = spa->spa_last_synced_guid != 0 ?
	spa->spa_last_synced_guid : spa->spa_root_vdev->vdev_guid;

	/*
	* Return the most recently synced out guid unless we're
	* in syncing context.
	*/
	if (dp && dsl_pool_sync_context(dp))
	return (spa->spa_root_vdev->vdev_guid);
	else
	return (guid);
	}

	uint64_t
	spa_load_guid(spa_t *spa)
	{
	/*
	* This is a GUID that exists solely as a reference for the
	* purposes of the arc. It is generated at load time, and
	* is never written to persistent storage.
	*/
	return (spa->spa_load_guid);
	}

	uint64_t
	spa_last_synced_txg(spa_t *spa)
	{
	return (spa->spa_ubsync.ub_txg);
	}

	uint64_t
	spa_first_txg(spa_t *spa)
	{
	return (spa->spa_first_txg);
	}

	uint64_t
	spa_syncing_txg(spa_t *spa)
	{
	return (spa->spa_syncing_txg);
	}

	/*
	* Return the last txg where data can be dirtied. The final txgs
	* will be used to just clear out any deferred frees that remain.
	*/
	uint64_t
	spa_final_dirty_txg(spa_t *spa)
	{
	return (spa->spa_final_txg - TXG_DEFER_SIZE);
	}

	pool_state_t
	spa_state(spa_t *spa)
	{
	return (spa->spa_state);
	}

	spa_load_state_t
	spa_load_state(spa_t *spa)
	{
	return (spa->spa_load_state);
	}

	uint64_t
	spa_freeze_txg(spa_t *spa)
	{
	return (spa->spa_freeze_txg);
	}

	/*
	* Return the inflated asize for a logical write in bytes. This is used by the
	* DMU to calculate the space a logical write will require on disk.
	* If lsize is smaller than the largest physical block size allocatable on this
	* pool we use its value instead, since the write will end up using the whole
	* block anyway.
	*/
	uint64_t
	spa_get_worst_case_asize(spa_t *spa, uint64_t lsize)
	{
	if (lsize == 0)
	return (0); /* No inflation needed */
	return (MAX(lsize, 1 << spa->spa_max_ashift) * spa_asize_inflation);
	}

	/*
	* Return the amount of slop space in bytes. It is typically 1/32 of the pool
	* (3.2%), minus the embedded log space. On very small pools, it may be
	* slightly larger than this. On very large pools, it will be capped to
	* the value of spa_max_slop. The embedded log space is not included in
	* spa_dspace. By subtracting it, the usable space (per "zfs list") is a
	* constant 97% of the total space, regardless of metaslab size (assuming the
	* default spa_slop_shift=5 and a non-tiny pool).
	*
	* See the comment above spa_slop_shift for more details.
	*/
	uint64_t
	spa_get_slop_space(spa_t *spa)
	{
	uint64_t space = 0;
	uint64_t slop = 0;

	/*
	* Make sure spa_dedup_dspace has been set.
	*/
	if (spa->spa_dedup_dspace == ~0ULL)
	spa_update_dspace(spa);

	/*
	* spa_get_dspace() includes the space only logically "used" by
	* deduplicated data, so since it's not useful to reserve more
	* space with more deduplicated data, we subtract that out here.
	*/
	space =
	spa_get_dspace(spa) - spa->spa_dedup_dspace - brt_get_dspace(spa);
	slop = MIN(space >> spa_slop_shift, spa_max_slop);

	/*
	* Subtract the embedded log space, but no more than half the (3.2%)
	* unusable space. Note, the "no more than half" is only relevant if
	* zfs_embedded_slog_min_ms >> spa_slop_shift < 2, which is not true by
	* default.
	*/
	uint64_t embedded_log =
	metaslab_class_get_dspace(spa_embedded_log_class(spa));
	slop -= MIN(embedded_log, slop >> 1);

	/*
	* Slop space should be at least spa_min_slop, but no more than half
	* the entire pool.
	*/
	slop = MAX(slop, MIN(space >> 1, spa_min_slop));
	return (slop);
	}

	uint64_t
	spa_get_dspace(spa_t *spa)
	{
	return (spa->spa_dspace);
	}

	uint64_t
	spa_get_checkpoint_space(spa_t *spa)
	{
	return (spa->spa_checkpoint_info.sci_dspace);
	}

	void
	spa_update_dspace(spa_t *spa)
	{
	spa->spa_dspace = metaslab_class_get_dspace(spa_normal_class(spa)) +
	ddt_get_dedup_dspace(spa) + brt_get_dspace(spa);
	if (spa->spa_nonallocating_dspace > 0) {
	/*
	* Subtract the space provided by all non-allocating vdevs that
	* contribute to dspace. If a file is overwritten, its old
	* blocks are freed and new blocks are allocated. If there are
	* no snapshots of the file, the available space should remain
	* the same. The old blocks could be freed from the
	* non-allocating vdev, but the new blocks must be allocated on
	* other (allocating) vdevs. By reserving the entire size of
	* the non-allocating vdevs (including allocated space), we
	* ensure that there will be enough space on the allocating
	* vdevs for this file overwrite to succeed.
	*
	* Note that the DMU/DSL doesn't actually know or care
	* how much space is allocated (it does its own tracking
	* of how much space has been logically used). So it
	* doesn't matter that the data we are moving may be
	* allocated twice (on the old device and the new device).
	*/
	ASSERT3U(spa->spa_dspace, >=, spa->spa_nonallocating_dspace);
	spa->spa_dspace -= spa->spa_nonallocating_dspace;
	}
	}

	/*
	* Return the failure mode that has been set to this pool. The default
	* behavior will be to block all I/Os when a complete failure occurs.
	*/
	uint64_t
	spa_get_failmode(spa_t *spa)
	{
	return (spa->spa_failmode);
	}

	boolean_t
	spa_suspended(spa_t *spa)
	{
	return (spa->spa_suspended != ZIO_SUSPEND_NONE);
	}

	uint64_t
	spa_version(spa_t *spa)
	{
	return (spa->spa_ubsync.ub_version);
	}

	boolean_t
	spa_deflate(spa_t *spa)
	{
	return (spa->spa_deflate);
	}

	metaslab_class_t *
	spa_normal_class(spa_t *spa)
	{
	return (spa->spa_normal_class);
	}

	metaslab_class_t *
	spa_log_class(spa_t *spa)
	{
	return (spa->spa_log_class);
	}

	metaslab_class_t *
	spa_embedded_log_class(spa_t *spa)
	{
	return (spa->spa_embedded_log_class);
	}

	metaslab_class_t *
	spa_special_class(spa_t *spa)
	{
	return (spa->spa_special_class);
	}

	metaslab_class_t *
	spa_dedup_class(spa_t *spa)
	{
	return (spa->spa_dedup_class);
	}

	/*
	* Locate an appropriate allocation class
	*/
	metaslab_class_t *
	spa_preferred_class(spa_t *spa, uint64_t size, dmu_object_type_t objtype,
	uint_t level, uint_t special_smallblk)
	{
	/*
	* ZIL allocations determine their class in zio_alloc_zil().
	*/
	ASSERT(objtype != DMU_OT_INTENT_LOG);

	boolean_t has_special_class = spa->spa_special_class->mc_groups != 0;

	if (DMU_OT_IS_DDT(objtype)) {
	if (spa->spa_dedup_class->mc_groups != 0)
	return (spa_dedup_class(spa));
	else if (has_special_class && zfs_ddt_data_is_special)
	return (spa_special_class(spa));
	else
	return (spa_normal_class(spa));
	}

	/* Indirect blocks for user data can land in special if allowed */
	if (level > 0 && (DMU_OT_IS_FILE(objtype) \|\| objtype == DMU_OT_ZVOL)) {
	if (has_special_class && zfs_user_indirect_is_special)
	return (spa_special_class(spa));
	else
	return (spa_normal_class(spa));
	}

	if (DMU_OT_IS_METADATA(objtype) \|\| level > 0) {
	if (has_special_class)
	return (spa_special_class(spa));
	else
	return (spa_normal_class(spa));
	}

	/*
	* Allow small file blocks in special class in some cases (like
	* for the dRAID vdev feature). But always leave a reserve of
	* zfs_special_class_metadata_reserve_pct exclusively for metadata.
	*/
	if (DMU_OT_IS_FILE(objtype) &&
	has_special_class && size <= special_smallblk) {
	metaslab_class_t *special = spa_special_class(spa);
	uint64_t alloc = metaslab_class_get_alloc(special);
	uint64_t space = metaslab_class_get_space(special);
	uint64_t limit =
	(space * (100 - zfs_special_class_metadata_reserve_pct))
	/ 100;

	if (alloc < limit)
	return (special);
	}

	return (spa_normal_class(spa));
	}

	void
	spa_evicting_os_register(spa_t spa, objset_t os)
	{
	mutex_enter(&spa->spa_evicting_os_lock);
	list_insert_head(&spa->spa_evicting_os_list, os);
	mutex_exit(&spa->spa_evicting_os_lock);
	}

	void
	spa_evicting_os_deregister(spa_t spa, objset_t os)
	{
	mutex_enter(&spa->spa_evicting_os_lock);
	list_remove(&spa->spa_evicting_os_list, os);
	cv_broadcast(&spa->spa_evicting_os_cv);
	mutex_exit(&spa->spa_evicting_os_lock);
	}

	void
	spa_evicting_os_wait(spa_t *spa)
	{
	mutex_enter(&spa->spa_evicting_os_lock);
	while (!list_is_empty(&spa->spa_evicting_os_list))
	cv_wait(&spa->spa_evicting_os_cv, &spa->spa_evicting_os_lock);
	mutex_exit(&spa->spa_evicting_os_lock);

	dmu_buf_user_evict_wait();
	}

	int
	spa_max_replication(spa_t *spa)
	{
	/*
	* As of SPA_VERSION == SPA_VERSION_DITTO_BLOCKS, we are able to
	* handle BPs with more than one DVA allocated. Set our max
	* replication level accordingly.
	*/
	if (spa_version(spa) < SPA_VERSION_DITTO_BLOCKS)
	return (1);
	return (MIN(SPA_DVAS_PER_BP, spa_max_replication_override));
	}

	int
	spa_prev_software_version(spa_t *spa)
	{
	return (spa->spa_prev_software_version);
	}

	uint64_t
	spa_deadman_synctime(spa_t *spa)
	{
	return (spa->spa_deadman_synctime);
	}

	spa_autotrim_t
	spa_get_autotrim(spa_t *spa)
	{
	return (spa->spa_autotrim);
	}

	uint64_t
	spa_deadman_ziotime(spa_t *spa)
	{
	return (spa->spa_deadman_ziotime);
	}

	uint64_t
	spa_get_deadman_failmode(spa_t *spa)
	{
	return (spa->spa_deadman_failmode);
	}

	void
	spa_set_deadman_failmode(spa_t spa, const char failmode)
	{
	if (strcmp(failmode, "wait") == 0)
	spa->spa_deadman_failmode = ZIO_FAILURE_MODE_WAIT;
	else if (strcmp(failmode, "continue") == 0)
	spa->spa_deadman_failmode = ZIO_FAILURE_MODE_CONTINUE;
	else if (strcmp(failmode, "panic") == 0)
	spa->spa_deadman_failmode = ZIO_FAILURE_MODE_PANIC;
	else
	spa->spa_deadman_failmode = ZIO_FAILURE_MODE_WAIT;
	}

	void
	spa_set_deadman_ziotime(hrtime_t ns)
	{
	spa_t *spa = NULL;

	if (spa_mode_global != SPA_MODE_UNINIT) {
	mutex_enter(&spa_namespace_lock);
	while ((spa = spa_next(spa)) != NULL)
	spa->spa_deadman_ziotime = ns;
	mutex_exit(&spa_namespace_lock);
	}
	}

	void
	spa_set_deadman_synctime(hrtime_t ns)
	{
	spa_t *spa = NULL;

	if (spa_mode_global != SPA_MODE_UNINIT) {
	mutex_enter(&spa_namespace_lock);
	while ((spa = spa_next(spa)) != NULL)
	spa->spa_deadman_synctime = ns;
	mutex_exit(&spa_namespace_lock);
	}
	}

	uint64_t
	dva_get_dsize_sync(spa_t spa, const dva_t dva)
	{
	uint64_t asize = DVA_GET_ASIZE(dva);
	uint64_t dsize = asize;

	ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0);

	if (asize != 0 && spa->spa_deflate) {
	vdev_t *vd = vdev_lookup_top(spa, DVA_GET_VDEV(dva));
	if (vd != NULL)
	dsize = (asize >> SPA_MINBLOCKSHIFT) *
	vd->vdev_deflate_ratio;
	}

	return (dsize);
	}

	uint64_t
	bp_get_dsize_sync(spa_t spa, const blkptr_t bp)
	{
	uint64_t dsize = 0;

	for (int d = 0; d < BP_GET_NDVAS(bp); d++)
	dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]);

	return (dsize);
	}

	uint64_t
	bp_get_dsize(spa_t spa, const blkptr_t bp)
	{
	uint64_t dsize = 0;

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);

	for (int d = 0; d < BP_GET_NDVAS(bp); d++)
	dsize += dva_get_dsize_sync(spa, &bp->blk_dva[d]);

	spa_config_exit(spa, SCL_VDEV, FTAG);

	return (dsize);
	}

	uint64_t
	spa_dirty_data(spa_t *spa)
	{
	return (spa->spa_dsl_pool->dp_dirty_total);
	}

	/*
	* ==========================================================================
	* SPA Import Progress Routines
	* ==========================================================================
	*/

	typedef struct spa_import_progress {
	uint64_t pool_guid; /* unique id for updates */
	char *pool_name;
	spa_load_state_t spa_load_state;
	char *spa_load_notes;
	uint64_t mmp_sec_remaining; /* MMP activity check */
	uint64_t spa_load_max_txg; /* rewind txg */
	procfs_list_node_t smh_node;
	} spa_import_progress_t;

	spa_history_list_t *spa_import_progress_list = NULL;

	static int
	spa_import_progress_show_header(struct seq_file *f)
	{
	seq_printf(f, "%-20s %-14s %-14s %-12s %-16s %s\n", "pool_guid",
	"load_state", "multihost_secs", "max_txg",
	"pool_name", "notes");
	return (0);
	}

	static int
	spa_import_progress_show(struct seq_file f, void data)
	{
	spa_import_progress_t sip = (spa_import_progress_t )data;

	seq_printf(f, "%-20llu %-14llu %-14llu %-12llu %-16s %s\n",
	(u_longlong_t)sip->pool_guid, (u_longlong_t)sip->spa_load_state,
	(u_longlong_t)sip->mmp_sec_remaining,
	(u_longlong_t)sip->spa_load_max_txg,
	(sip->pool_name ? sip->pool_name : "-"),
	(sip->spa_load_notes ? sip->spa_load_notes : "-"));

	return (0);
	}

	/* Remove oldest elements from list until there are no more than 'size' left */
	static void
	spa_import_progress_truncate(spa_history_list_t *shl, unsigned int size)
	{
	spa_import_progress_t *sip;
	while (shl->size > size) {
	sip = list_remove_head(&shl->procfs_list.pl_list);
	if (sip->pool_name)
	spa_strfree(sip->pool_name);
	if (sip->spa_load_notes)
	kmem_strfree(sip->spa_load_notes);
	kmem_free(sip, sizeof (spa_import_progress_t));
	shl->size--;
	}

	IMPLY(size == 0, list_is_empty(&shl->procfs_list.pl_list));
	}

	static void
	spa_import_progress_init(void)
	{
	spa_import_progress_list = kmem_zalloc(sizeof (spa_history_list_t),
	KM_SLEEP);

	spa_import_progress_list->size = 0;

	spa_import_progress_list->procfs_list.pl_private =
	spa_import_progress_list;

	procfs_list_install("zfs",
	NULL,
	"import_progress",
	0644,
	&spa_import_progress_list->procfs_list,
	spa_import_progress_show,
	spa_import_progress_show_header,
	NULL,
	offsetof(spa_import_progress_t, smh_node));
	}

	static void
	spa_import_progress_destroy(void)
	{
	spa_history_list_t *shl = spa_import_progress_list;
	procfs_list_uninstall(&shl->procfs_list);
	spa_import_progress_truncate(shl, 0);
	procfs_list_destroy(&shl->procfs_list);
	kmem_free(shl, sizeof (spa_history_list_t));
	}

	int
	spa_import_progress_set_state(uint64_t pool_guid,
	spa_load_state_t load_state)
	{
	spa_history_list_t *shl = spa_import_progress_list;
	spa_import_progress_t *sip;
	int error = ENOENT;

	if (shl->size == 0)
	return (0);

	mutex_enter(&shl->procfs_list.pl_lock);
	for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL;
	sip = list_prev(&shl->procfs_list.pl_list, sip)) {
	if (sip->pool_guid == pool_guid) {
	sip->spa_load_state = load_state;
	if (sip->spa_load_notes != NULL) {
	kmem_strfree(sip->spa_load_notes);
	sip->spa_load_notes = NULL;
	}
	error = 0;
	break;
	}
	}
	mutex_exit(&shl->procfs_list.pl_lock);

	return (error);
	}

	static void
	spa_import_progress_set_notes_impl(spa_t *spa, boolean_t log_dbgmsg,
	const char *fmt, va_list adx)
	{
	spa_history_list_t *shl = spa_import_progress_list;
	spa_import_progress_t *sip;
	uint64_t pool_guid = spa_guid(spa);

	if (shl->size == 0)
	return;

	char *notes = kmem_vasprintf(fmt, adx);

	mutex_enter(&shl->procfs_list.pl_lock);
	for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL;
	sip = list_prev(&shl->procfs_list.pl_list, sip)) {
	if (sip->pool_guid == pool_guid) {
	if (sip->spa_load_notes != NULL) {
	kmem_strfree(sip->spa_load_notes);
	sip->spa_load_notes = NULL;
	}
	sip->spa_load_notes = notes;
	if (log_dbgmsg)
	zfs_dbgmsg("'%s' %s", sip->pool_name, notes);
	notes = NULL;
	break;
	}
	}
	mutex_exit(&shl->procfs_list.pl_lock);
	if (notes != NULL)
	kmem_strfree(notes);
	}

	void
	spa_import_progress_set_notes(spa_t spa, const char fmt, ...)
	{
	va_list adx;

	va_start(adx, fmt);
	spa_import_progress_set_notes_impl(spa, B_TRUE, fmt, adx);
	va_end(adx);
	}

	void
	spa_import_progress_set_notes_nolog(spa_t spa, const char fmt, ...)
	{
	va_list adx;

	va_start(adx, fmt);
	spa_import_progress_set_notes_impl(spa, B_FALSE, fmt, adx);
	va_end(adx);
	}

	int
	spa_import_progress_set_max_txg(uint64_t pool_guid, uint64_t load_max_txg)
	{
	spa_history_list_t *shl = spa_import_progress_list;
	spa_import_progress_t *sip;
	int error = ENOENT;

	if (shl->size == 0)
	return (0);

	mutex_enter(&shl->procfs_list.pl_lock);
	for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL;
	sip = list_prev(&shl->procfs_list.pl_list, sip)) {
	if (sip->pool_guid == pool_guid) {
	sip->spa_load_max_txg = load_max_txg;
	error = 0;
	break;
	}
	}
	mutex_exit(&shl->procfs_list.pl_lock);

	return (error);
	}

	int
	spa_import_progress_set_mmp_check(uint64_t pool_guid,
	uint64_t mmp_sec_remaining)
	{
	spa_history_list_t *shl = spa_import_progress_list;
	spa_import_progress_t *sip;
	int error = ENOENT;

	if (shl->size == 0)
	return (0);

	mutex_enter(&shl->procfs_list.pl_lock);
	for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL;
	sip = list_prev(&shl->procfs_list.pl_list, sip)) {
	if (sip->pool_guid == pool_guid) {
	sip->mmp_sec_remaining = mmp_sec_remaining;
	error = 0;
	break;
	}
	}
	mutex_exit(&shl->procfs_list.pl_lock);

	return (error);
	}

	/*
	* A new import is in progress, add an entry.
	*/
	void
	spa_import_progress_add(spa_t *spa)
	{
	spa_history_list_t *shl = spa_import_progress_list;
	spa_import_progress_t *sip;
	const char *poolname = NULL;

	sip = kmem_zalloc(sizeof (spa_import_progress_t), KM_SLEEP);
	sip->pool_guid = spa_guid(spa);

	(void) nvlist_lookup_string(spa->spa_config, ZPOOL_CONFIG_POOL_NAME,
	&poolname);
	if (poolname == NULL)
	poolname = spa_name(spa);
	sip->pool_name = spa_strdup(poolname);
	sip->spa_load_state = spa_load_state(spa);
	sip->spa_load_notes = NULL;

	mutex_enter(&shl->procfs_list.pl_lock);
	procfs_list_add(&shl->procfs_list, sip);
	shl->size++;
	mutex_exit(&shl->procfs_list.pl_lock);
	}

	void
	spa_import_progress_remove(uint64_t pool_guid)
	{
	spa_history_list_t *shl = spa_import_progress_list;
	spa_import_progress_t *sip;

	mutex_enter(&shl->procfs_list.pl_lock);
	for (sip = list_tail(&shl->procfs_list.pl_list); sip != NULL;
	sip = list_prev(&shl->procfs_list.pl_list, sip)) {
	if (sip->pool_guid == pool_guid) {
	if (sip->pool_name)
	spa_strfree(sip->pool_name);
	if (sip->spa_load_notes)
	spa_strfree(sip->spa_load_notes);
	list_remove(&shl->procfs_list.pl_list, sip);
	shl->size--;
	kmem_free(sip, sizeof (spa_import_progress_t));
	break;
	}
	}
	mutex_exit(&shl->procfs_list.pl_lock);
	}

	/*
	* ==========================================================================
	* Initialization and Termination
	* ==========================================================================
	*/

	static int
	spa_name_compare(const void a1, const void a2)
	{
	const spa_t *s1 = a1;
	const spa_t *s2 = a2;
	int s;

	s = strcmp(s1->spa_name, s2->spa_name);

	return (TREE_ISIGN(s));
	}

	void
	spa_boot_init(void)
	{
	spa_config_load();
	}

	void
	spa_init(spa_mode_t mode)
	{
	mutex_init(&spa_namespace_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa_spare_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&spa_l2cache_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&spa_namespace_cv, NULL, CV_DEFAULT, NULL);

	avl_create(&spa_namespace_avl, spa_name_compare, sizeof (spa_t),
	offsetof(spa_t, spa_avl));

	avl_create(&spa_spare_avl, spa_spare_compare, sizeof (spa_aux_t),
	offsetof(spa_aux_t, aux_avl));

	avl_create(&spa_l2cache_avl, spa_l2cache_compare, sizeof (spa_aux_t),
	offsetof(spa_aux_t, aux_avl));

	spa_mode_global = mode;

	#ifndef _KERNEL
	if (spa_mode_global != SPA_MODE_READ && dprintf_find_string("watch")) {
	struct sigaction sa;

	sa.sa_flags = SA_SIGINFO;
	sigemptyset(&sa.sa_mask);
	sa.sa_sigaction = arc_buf_sigsegv;

	if (sigaction(SIGSEGV, &sa, NULL) == -1) {
	perror("could not enable watchpoints: "
	"sigaction(SIGSEGV, ...) = ");
	} else {
	arc_watch = B_TRUE;
	}
	}
	#endif

	fm_init();
	zfs_refcount_init();
	unique_init();
	zfs_btree_init();
	metaslab_stat_init();
	brt_init();
	ddt_init();
	zio_init();
	dmu_init();
	zil_init();
	vdev_mirror_stat_init();
	vdev_raidz_math_init();
	vdev_file_init();
	zfs_prop_init();
	chksum_init();
	zpool_prop_init();
	zpool_feature_init();
	spa_config_load();
	vdev_prop_init();
	l2arc_start();
	scan_init();
	qat_init();
	spa_import_progress_init();
	}

	void
	spa_fini(void)
	{
	l2arc_stop();

	spa_evict_all();

	vdev_file_fini();
	vdev_mirror_stat_fini();
	vdev_raidz_math_fini();
	chksum_fini();
	zil_fini();
	dmu_fini();
	zio_fini();
	ddt_fini();
	brt_fini();
	metaslab_stat_fini();
	zfs_btree_fini();
	unique_fini();
	zfs_refcount_fini();
	fm_fini();
	scan_fini();
	qat_fini();
	spa_import_progress_destroy();

	avl_destroy(&spa_namespace_avl);
	avl_destroy(&spa_spare_avl);
	avl_destroy(&spa_l2cache_avl);

	cv_destroy(&spa_namespace_cv);
	mutex_destroy(&spa_namespace_lock);
	mutex_destroy(&spa_spare_lock);
	mutex_destroy(&spa_l2cache_lock);
	}

	/*
	* Return whether this pool has a dedicated slog device. No locking needed.
	* It's not a problem if the wrong answer is returned as it's only for
	* performance and not correctness.
	*/
	boolean_t
	spa_has_slogs(spa_t *spa)
	{
	return (spa->spa_log_class->mc_groups != 0);
	}

	spa_log_state_t
	spa_get_log_state(spa_t *spa)
	{
	return (spa->spa_log_state);
	}

	void
	spa_set_log_state(spa_t *spa, spa_log_state_t state)
	{
	spa->spa_log_state = state;
	}

	boolean_t
	spa_is_root(spa_t *spa)
	{
	return (spa->spa_is_root);
	}

	boolean_t
	spa_writeable(spa_t *spa)
	{
	return (!!(spa->spa_mode & SPA_MODE_WRITE) && spa->spa_trust_config);
	}

	/*
	* Returns true if there is a pending sync task in any of the current
	* syncing txg, the current quiescing txg, or the current open txg.
	*/
	boolean_t
	spa_has_pending_synctask(spa_t *spa)
	{
	return (!txg_all_lists_empty(&spa->spa_dsl_pool->dp_sync_tasks) \|\|
	!txg_all_lists_empty(&spa->spa_dsl_pool->dp_early_sync_tasks));
	}

	spa_mode_t
	spa_mode(spa_t *spa)
	{
	return (spa->spa_mode);
	}

	uint64_t
	spa_bootfs(spa_t *spa)
	{
	return (spa->spa_bootfs);
	}

	uint64_t
	spa_delegation(spa_t *spa)
	{
	return (spa->spa_delegation);
	}

	objset_t *
	spa_meta_objset(spa_t *spa)
	{
	return (spa->spa_meta_objset);
	}

	enum zio_checksum
	spa_dedup_checksum(spa_t *spa)
	{
	return (spa->spa_dedup_checksum);
	}

	/*
	* Reset pool scan stat per scan pass (or reboot).
	*/
	void
	spa_scan_stat_init(spa_t *spa)
	{
	/* data not stored on disk */
	spa->spa_scan_pass_start = gethrestime_sec();
	if (dsl_scan_is_paused_scrub(spa->spa_dsl_pool->dp_scan))
	spa->spa_scan_pass_scrub_pause = spa->spa_scan_pass_start;
	else
	spa->spa_scan_pass_scrub_pause = 0;

	if (dsl_errorscrub_is_paused(spa->spa_dsl_pool->dp_scan))
	spa->spa_scan_pass_errorscrub_pause = spa->spa_scan_pass_start;
	else
	spa->spa_scan_pass_errorscrub_pause = 0;

	spa->spa_scan_pass_scrub_spent_paused = 0;
	spa->spa_scan_pass_exam = 0;
	spa->spa_scan_pass_issued = 0;

	// error scrub stats
	spa->spa_scan_pass_errorscrub_spent_paused = 0;
	}

	/*
	* Get scan stats for zpool status reports
	*/
	int
	spa_scan_get_stats(spa_t spa, pool_scan_stat_t ps)
	{
	dsl_scan_t *scn = spa->spa_dsl_pool ? spa->spa_dsl_pool->dp_scan : NULL;

	if (scn == NULL \|\| (scn->scn_phys.scn_func == POOL_SCAN_NONE &&
	scn->errorscrub_phys.dep_func == POOL_SCAN_NONE))
	return (SET_ERROR(ENOENT));

	memset(ps, 0, sizeof (pool_scan_stat_t));

	/* data stored on disk */
	ps->pss_func = scn->scn_phys.scn_func;
	ps->pss_state = scn->scn_phys.scn_state;
	ps->pss_start_time = scn->scn_phys.scn_start_time;
	ps->pss_end_time = scn->scn_phys.scn_end_time;
	ps->pss_to_examine = scn->scn_phys.scn_to_examine;
	ps->pss_examined = scn->scn_phys.scn_examined;
	ps->pss_skipped = scn->scn_phys.scn_skipped;
	ps->pss_processed = scn->scn_phys.scn_processed;
	ps->pss_errors = scn->scn_phys.scn_errors;

	/* data not stored on disk */
	ps->pss_pass_exam = spa->spa_scan_pass_exam;
	ps->pss_pass_start = spa->spa_scan_pass_start;
	ps->pss_pass_scrub_pause = spa->spa_scan_pass_scrub_pause;
	ps->pss_pass_scrub_spent_paused = spa->spa_scan_pass_scrub_spent_paused;
	ps->pss_pass_issued = spa->spa_scan_pass_issued;
	ps->pss_issued =
	scn->scn_issued_before_pass + spa->spa_scan_pass_issued;

	/* error scrub data stored on disk */
	ps->pss_error_scrub_func = scn->errorscrub_phys.dep_func;
	ps->pss_error_scrub_state = scn->errorscrub_phys.dep_state;
	ps->pss_error_scrub_start = scn->errorscrub_phys.dep_start_time;
	ps->pss_error_scrub_end = scn->errorscrub_phys.dep_end_time;
	ps->pss_error_scrub_examined = scn->errorscrub_phys.dep_examined;
	ps->pss_error_scrub_to_be_examined =
	scn->errorscrub_phys.dep_to_examine;

	/* error scrub data not stored on disk */
	ps->pss_pass_error_scrub_pause = spa->spa_scan_pass_errorscrub_pause;

	return (0);
	}

	int
	spa_maxblocksize(spa_t *spa)
	{
	if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS))
	return (SPA_MAXBLOCKSIZE);
	else
	return (SPA_OLD_MAXBLOCKSIZE);
	}


	/*
	* Returns the txg that the last device removal completed. No indirect mappings
	* have been added since this txg.
	*/
	uint64_t
	spa_get_last_removal_txg(spa_t *spa)
	{
	uint64_t vdevid;
	uint64_t ret = -1ULL;

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
	/*
	* sr_prev_indirect_vdev is only modified while holding all the
	* config locks, so it is sufficient to hold SCL_VDEV as reader when
	* examining it.
	*/
	vdevid = spa->spa_removing_phys.sr_prev_indirect_vdev;

	while (vdevid != -1ULL) {
	vdev_t *vd = vdev_lookup_top(spa, vdevid);
	vdev_indirect_births_t *vib = vd->vdev_indirect_births;

	ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);

	/*
	* If the removal did not remap any data, we don't care.
	*/
	if (vdev_indirect_births_count(vib) != 0) {
	ret = vdev_indirect_births_last_entry_txg(vib);
	break;
	}

	vdevid = vd->vdev_indirect_config.vic_prev_indirect_vdev;
	}
	spa_config_exit(spa, SCL_VDEV, FTAG);

	IMPLY(ret != -1ULL,
	spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REMOVAL));

	return (ret);
	}

	int
	spa_maxdnodesize(spa_t *spa)
	{
	if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_DNODE))
	return (DNODE_MAX_SIZE);
	else
	return (DNODE_MIN_SIZE);
	}

	boolean_t
	spa_multihost(spa_t *spa)
	{
	return (spa->spa_multihost ? B_TRUE : B_FALSE);
	}

	uint32_t
	spa_get_hostid(spa_t *spa)
	{
	return (spa->spa_hostid);
	}

	boolean_t
	spa_trust_config(spa_t *spa)
	{
	return (spa->spa_trust_config);
	}

	uint64_t
	spa_missing_tvds_allowed(spa_t *spa)
	{
	return (spa->spa_missing_tvds_allowed);
	}

	space_map_t *
	spa_syncing_log_sm(spa_t *spa)
	{
	return (spa->spa_syncing_log_sm);
	}

	void
	spa_set_missing_tvds(spa_t *spa, uint64_t missing)
	{
	spa->spa_missing_tvds = missing;
	}

	/*
	* Return the pool state string ("ONLINE", "DEGRADED", "SUSPENDED", etc).
	*/
	const char *
	spa_state_to_name(spa_t *spa)
	{
	ASSERT3P(spa, !=, NULL);

	/*
	* it is possible for the spa to exist, without root vdev
	* as the spa transitions during import/export
	*/
	vdev_t *rvd = spa->spa_root_vdev;
	if (rvd == NULL) {
	return ("TRANSITIONING");
	}
	vdev_state_t state = rvd->vdev_state;
	vdev_aux_t aux = rvd->vdev_stat.vs_aux;

	if (spa_suspended(spa))
	return ("SUSPENDED");

	switch (state) {
	case VDEV_STATE_CLOSED:
	case VDEV_STATE_OFFLINE:
	return ("OFFLINE");
	case VDEV_STATE_REMOVED:
	return ("REMOVED");
	case VDEV_STATE_CANT_OPEN:
	if (aux == VDEV_AUX_CORRUPT_DATA \|\| aux == VDEV_AUX_BAD_LOG)
	return ("FAULTED");
	else if (aux == VDEV_AUX_SPLIT_POOL)
	return ("SPLIT");
	else
	return ("UNAVAIL");
	case VDEV_STATE_FAULTED:
	return ("FAULTED");
	case VDEV_STATE_DEGRADED:
	return ("DEGRADED");
	case VDEV_STATE_HEALTHY:
	return ("ONLINE");
	default:
	break;
	}

	return ("UNKNOWN");
	}

	boolean_t
	spa_top_vdevs_spacemap_addressable(spa_t *spa)
	{
	vdev_t *rvd = spa->spa_root_vdev;
	for (uint64_t c = 0; c < rvd->vdev_children; c++) {
	if (!vdev_is_spacemap_addressable(rvd->vdev_child[c]))
	return (B_FALSE);
	}
	return (B_TRUE);
	}

	boolean_t
	spa_has_checkpoint(spa_t *spa)
	{
	return (spa->spa_checkpoint_txg != 0);
	}

	boolean_t
	spa_importing_readonly_checkpoint(spa_t *spa)
	{
	return ((spa->spa_import_flags & ZFS_IMPORT_CHECKPOINT) &&
	spa->spa_mode == SPA_MODE_READ);
	}

	uint64_t
	spa_min_claim_txg(spa_t *spa)
	{
	uint64_t checkpoint_txg = spa->spa_uberblock.ub_checkpoint_txg;

	if (checkpoint_txg != 0)
	return (checkpoint_txg + 1);

	return (spa->spa_first_txg);
	}

	/*
	* If there is a checkpoint, async destroys may consume more space from
	* the pool instead of freeing it. In an attempt to save the pool from
	* getting suspended when it is about to run out of space, we stop
	* processing async destroys.
	*/
	boolean_t
	spa_suspend_async_destroy(spa_t *spa)
	{
	dsl_pool_t *dp = spa_get_dsl(spa);

	uint64_t unreserved = dsl_pool_unreserved_space(dp,
	ZFS_SPACE_CHECK_EXTRA_RESERVED);
	uint64_t used = dsl_dir_phys(dp->dp_root_dir)->dd_used_bytes;
	uint64_t avail = (unreserved > used) ? (unreserved - used) : 0;

	if (spa_has_checkpoint(spa) && avail == 0)
	return (B_TRUE);

	return (B_FALSE);
	}

	#if defined(_KERNEL)

	int
	param_set_deadman_failmode_common(const char *val)
	{
	spa_t *spa = NULL;
	char *p;

	if (val == NULL)
	return (SET_ERROR(EINVAL));

	if ((p = strchr(val, '\n')) != NULL)
	*p = '\0';

	if (strcmp(val, "wait") != 0 && strcmp(val, "continue") != 0 &&
	strcmp(val, "panic"))
	return (SET_ERROR(EINVAL));

	if (spa_mode_global != SPA_MODE_UNINIT) {
	mutex_enter(&spa_namespace_lock);
	while ((spa = spa_next(spa)) != NULL)
	spa_set_deadman_failmode(spa, val);
	mutex_exit(&spa_namespace_lock);
	}

	return (0);
	}
	#endif

	/* Namespace manipulation */
	EXPORT_SYMBOL(spa_lookup);
	EXPORT_SYMBOL(spa_add);
	EXPORT_SYMBOL(spa_remove);
	EXPORT_SYMBOL(spa_next);

	/* Refcount functions */
	EXPORT_SYMBOL(spa_open_ref);
	EXPORT_SYMBOL(spa_close);
	EXPORT_SYMBOL(spa_refcount_zero);

	/* Pool configuration lock */
	EXPORT_SYMBOL(spa_config_tryenter);
	EXPORT_SYMBOL(spa_config_enter);
	EXPORT_SYMBOL(spa_config_exit);
	EXPORT_SYMBOL(spa_config_held);

	/* Pool vdev add/remove lock */
	EXPORT_SYMBOL(spa_vdev_enter);
	EXPORT_SYMBOL(spa_vdev_exit);

	/* Pool vdev state change lock */
	EXPORT_SYMBOL(spa_vdev_state_enter);
	EXPORT_SYMBOL(spa_vdev_state_exit);

	/* Accessor functions */
	EXPORT_SYMBOL(spa_shutting_down);
	EXPORT_SYMBOL(spa_get_dsl);
	EXPORT_SYMBOL(spa_get_rootblkptr);
	EXPORT_SYMBOL(spa_set_rootblkptr);
	EXPORT_SYMBOL(spa_altroot);
	EXPORT_SYMBOL(spa_sync_pass);
	EXPORT_SYMBOL(spa_name);
	EXPORT_SYMBOL(spa_guid);
	EXPORT_SYMBOL(spa_last_synced_txg);
	EXPORT_SYMBOL(spa_first_txg);
	EXPORT_SYMBOL(spa_syncing_txg);
	EXPORT_SYMBOL(spa_version);
	EXPORT_SYMBOL(spa_state);
	EXPORT_SYMBOL(spa_load_state);
	EXPORT_SYMBOL(spa_freeze_txg);
	EXPORT_SYMBOL(spa_get_dspace);
	EXPORT_SYMBOL(spa_update_dspace);
	EXPORT_SYMBOL(spa_deflate);
	EXPORT_SYMBOL(spa_normal_class);
	EXPORT_SYMBOL(spa_log_class);
	EXPORT_SYMBOL(spa_special_class);
	EXPORT_SYMBOL(spa_preferred_class);
	EXPORT_SYMBOL(spa_max_replication);
	EXPORT_SYMBOL(spa_prev_software_version);
	EXPORT_SYMBOL(spa_get_failmode);
	EXPORT_SYMBOL(spa_suspended);
	EXPORT_SYMBOL(spa_bootfs);
	EXPORT_SYMBOL(spa_delegation);
	EXPORT_SYMBOL(spa_meta_objset);
	EXPORT_SYMBOL(spa_maxblocksize);
	EXPORT_SYMBOL(spa_maxdnodesize);

	/* Miscellaneous support routines */
	EXPORT_SYMBOL(spa_guid_exists);
	EXPORT_SYMBOL(spa_strdup);
	EXPORT_SYMBOL(spa_strfree);
	EXPORT_SYMBOL(spa_generate_guid);
	EXPORT_SYMBOL(snprintf_blkptr);
	EXPORT_SYMBOL(spa_freeze);
	EXPORT_SYMBOL(spa_upgrade);
	EXPORT_SYMBOL(spa_evict_all);
	EXPORT_SYMBOL(spa_lookup_by_guid);
	EXPORT_SYMBOL(spa_has_spare);
	EXPORT_SYMBOL(dva_get_dsize_sync);
	EXPORT_SYMBOL(bp_get_dsize_sync);
	EXPORT_SYMBOL(bp_get_dsize);
	EXPORT_SYMBOL(spa_has_slogs);
	EXPORT_SYMBOL(spa_is_root);
	EXPORT_SYMBOL(spa_writeable);
	EXPORT_SYMBOL(spa_mode);
	EXPORT_SYMBOL(spa_namespace_lock);
	EXPORT_SYMBOL(spa_trust_config);
	EXPORT_SYMBOL(spa_missing_tvds_allowed);
	EXPORT_SYMBOL(spa_set_missing_tvds);
	EXPORT_SYMBOL(spa_state_to_name);
	EXPORT_SYMBOL(spa_importing_readonly_checkpoint);
	EXPORT_SYMBOL(spa_min_claim_txg);
	EXPORT_SYMBOL(spa_suspend_async_destroy);
	EXPORT_SYMBOL(spa_has_checkpoint);
	EXPORT_SYMBOL(spa_top_vdevs_spacemap_addressable);

	ZFS_MODULE_PARAM(zfs, zfs_, flags, UINT, ZMOD_RW,
	"Set additional debugging flags");

	ZFS_MODULE_PARAM(zfs, zfs_, recover, INT, ZMOD_RW,
	"Set to attempt to recover from fatal errors");

	ZFS_MODULE_PARAM(zfs, zfs_, free_leak_on_eio, INT, ZMOD_RW,
	"Set to ignore IO errors during free and permanently leak the space");

	ZFS_MODULE_PARAM(zfs_deadman, zfs_deadman_, checktime_ms, U64, ZMOD_RW,
	"Dead I/O check interval in milliseconds");

	ZFS_MODULE_PARAM(zfs_deadman, zfs_deadman_, enabled, INT, ZMOD_RW,
	"Enable deadman timer");

	ZFS_MODULE_PARAM(zfs_spa, spa_, asize_inflation, UINT, ZMOD_RW,
	"SPA size estimate multiplication factor");

	ZFS_MODULE_PARAM(zfs, zfs_, ddt_data_is_special, INT, ZMOD_RW,
	"Place DDT data into the special class");

	ZFS_MODULE_PARAM(zfs, zfs_, user_indirect_is_special, INT, ZMOD_RW,
	"Place user data indirect blocks into the special class");

	/* BEGIN CSTYLED */
	ZFS_MODULE_PARAM_CALL(zfs_deadman, zfs_deadman_, failmode,
	param_set_deadman_failmode, param_get_charp, ZMOD_RW,
	"Failmode for deadman timer");

	ZFS_MODULE_PARAM_CALL(zfs_deadman, zfs_deadman_, synctime_ms,
	param_set_deadman_synctime, spl_param_get_u64, ZMOD_RW,
	"Pool sync expiration time in milliseconds");

	ZFS_MODULE_PARAM_CALL(zfs_deadman, zfs_deadman_, ziotime_ms,
	param_set_deadman_ziotime, spl_param_get_u64, ZMOD_RW,
	"IO expiration time in milliseconds");

	ZFS_MODULE_PARAM(zfs, zfs_, special_class_metadata_reserve_pct, UINT, ZMOD_RW,
	"Small file blocks in special vdevs depends on this much "
	"free space available");
	/* END CSTYLED */

	ZFS_MODULE_PARAM_CALL(zfs_spa, spa_, slop_shift, param_set_slop_shift,
	param_get_uint, ZMOD_RW, "Reserved free space in pool");

	ZFS_MODULE_PARAM(zfs, spa_, num_allocators, INT, ZMOD_RW,
	"Number of allocators per spa");

	ZFS_MODULE_PARAM(zfs, spa_, cpus_per_allocator, INT, ZMOD_RW,
	"Minimum number of CPUs per allocators");
	diff --git a/sys/contrib/openzfs/module/zfs/vdev.c b/sys/contrib/openzfs/module/zfs/vdev.c
	index c5551eb6cf6e..c74f72159dc9 100644
	--- a/sys/contrib/openzfs/module/zfs/vdev.c
	+++ b/sys/contrib/openzfs/module/zfs/vdev.c
	@@ -1,6505 +1,6514 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2021 by Delphix. All rights reserved.
	* Copyright 2017 Nexenta Systems, Inc.
	* Copyright (c) 2014 Integros [integros.com]
	* Copyright 2016 Toomas Soome <tsoome@me.com>
	* Copyright 2017 Joyent, Inc.
	* Copyright (c) 2017, Intel Corporation.
	* Copyright (c) 2019, Datto Inc. All rights reserved.
	* Copyright (c) 2021, Klara Inc.
	* Copyright (c) 2021, 2023 Hewlett Packard Enterprise Development LP.
	*/

	#include <sys/zfs_context.h>
	#include <sys/fm/fs/zfs.h>
	#include <sys/spa.h>
	#include <sys/spa_impl.h>
	#include <sys/bpobj.h>
	#include <sys/dmu.h>
	#include <sys/dmu_tx.h>
	#include <sys/dsl_dir.h>
	#include <sys/vdev_impl.h>
	#include <sys/vdev_rebuild.h>
	#include <sys/vdev_draid.h>
	#include <sys/uberblock_impl.h>
	#include <sys/metaslab.h>
	#include <sys/metaslab_impl.h>
	#include <sys/space_map.h>
	#include <sys/space_reftree.h>
	#include <sys/zio.h>
	#include <sys/zap.h>
	#include <sys/fs/zfs.h>
	#include <sys/arc.h>
	#include <sys/zil.h>
	#include <sys/dsl_scan.h>
	#include <sys/vdev_raidz.h>
	#include <sys/abd.h>
	#include <sys/vdev_initialize.h>
	#include <sys/vdev_trim.h>
	#include <sys/vdev_raidz.h>
	#include <sys/zvol.h>
	#include <sys/zfs_ratelimit.h>
	#include "zfs_prop.h"

	/*
	* One metaslab from each (normal-class) vdev is used by the ZIL. These are
	* called "embedded slog metaslabs", are referenced by vdev_log_mg, and are
	* part of the spa_embedded_log_class. The metaslab with the most free space
	* in each vdev is selected for this purpose when the pool is opened (or a
	* vdev is added). See vdev_metaslab_init().
	*
	* Log blocks can be allocated from the following locations. Each one is tried
	* in order until the allocation succeeds:
	* 1. dedicated log vdevs, aka "slog" (spa_log_class)
	* 2. embedded slog metaslabs (spa_embedded_log_class)
	* 3. other metaslabs in normal vdevs (spa_normal_class)
	*
	* zfs_embedded_slog_min_ms disables the embedded slog if there are fewer
	* than this number of metaslabs in the vdev. This ensures that we don't set
	* aside an unreasonable amount of space for the ZIL. If set to less than
	* 1 << (spa_slop_shift + 1), on small pools the usable space may be reduced
	* (by more than 1<<spa_slop_shift) due to the embedded slog metaslab.
	*/
	static uint_t zfs_embedded_slog_min_ms = 64;

	/* default target for number of metaslabs per top-level vdev */
	static uint_t zfs_vdev_default_ms_count = 200;

	/* minimum number of metaslabs per top-level vdev */
	static uint_t zfs_vdev_min_ms_count = 16;

	/* practical upper limit of total metaslabs per top-level vdev */
	static uint_t zfs_vdev_ms_count_limit = 1ULL << 17;

	/* lower limit for metaslab size (512M) */
	static uint_t zfs_vdev_default_ms_shift = 29;

	/* upper limit for metaslab size (16G) */
	static uint_t zfs_vdev_max_ms_shift = 34;

	int vdev_validate_skip = B_FALSE;

	/*
	* Since the DTL space map of a vdev is not expected to have a lot of
	* entries, we default its block size to 4K.
	*/
	int zfs_vdev_dtl_sm_blksz = (1 << 12);

	/*
	* Rate limit slow IO (delay) events to this many per second.
	*/
	static unsigned int zfs_slow_io_events_per_second = 20;

	+/*
	+ * Rate limit deadman "hung IO" events to this many per second.
	+ */
	+static unsigned int zfs_deadman_events_per_second = 1;
	+
	/*
	* Rate limit checksum events after this many checksum errors per second.
	*/
	static unsigned int zfs_checksum_events_per_second = 20;

	/*
	* Ignore errors during scrub/resilver. Allows to work around resilver
	* upon import when there are pool errors.
	*/
	static int zfs_scan_ignore_errors = 0;

	/*
	* vdev-wide space maps that have lots of entries written to them at
	* the end of each transaction can benefit from a higher I/O bandwidth
	* (e.g. vdev_obsolete_sm), thus we default their block size to 128K.
	*/
	int zfs_vdev_standard_sm_blksz = (1 << 17);

	/*
	* Tunable parameter for debugging or performance analysis. Setting this
	* will cause pool corruption on power loss if a volatile out-of-order
	* write cache is enabled.
	*/
	int zfs_nocacheflush = 0;

	/*
	* Maximum and minimum ashift values that can be automatically set based on
	* vdev's physical ashift (disk's physical sector size). While ASHIFT_MAX
	* is higher than the maximum value, it is intentionally limited here to not
	* excessively impact pool space efficiency. Higher ashift values may still
	* be forced by vdev logical ashift or by user via ashift property, but won't
	* be set automatically as a performance optimization.
	*/
	uint_t zfs_vdev_max_auto_ashift = 14;
	uint_t zfs_vdev_min_auto_ashift = ASHIFT_MIN;

	void
	vdev_dbgmsg(vdev_t vd, const char fmt, ...)
	{
	va_list adx;
	char buf[256];

	va_start(adx, fmt);
	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
	va_end(adx);

	if (vd->vdev_path != NULL) {
	zfs_dbgmsg("%s vdev '%s': %s", vd->vdev_ops->vdev_op_type,
	vd->vdev_path, buf);
	} else {
	zfs_dbgmsg("%s-%llu vdev (guid %llu): %s",
	vd->vdev_ops->vdev_op_type,
	(u_longlong_t)vd->vdev_id,
	(u_longlong_t)vd->vdev_guid, buf);
	}
	}

	void
	vdev_dbgmsg_print_tree(vdev_t *vd, int indent)
	{
	char state[20];

	if (vd->vdev_ishole \|\| vd->vdev_ops == &vdev_missing_ops) {
	zfs_dbgmsg("%*svdev %llu: %s", indent, "",
	(u_longlong_t)vd->vdev_id,
	vd->vdev_ops->vdev_op_type);
	return;
	}

	switch (vd->vdev_state) {
	case VDEV_STATE_UNKNOWN:
	(void) snprintf(state, sizeof (state), "unknown");
	break;
	case VDEV_STATE_CLOSED:
	(void) snprintf(state, sizeof (state), "closed");
	break;
	case VDEV_STATE_OFFLINE:
	(void) snprintf(state, sizeof (state), "offline");
	break;
	case VDEV_STATE_REMOVED:
	(void) snprintf(state, sizeof (state), "removed");
	break;
	case VDEV_STATE_CANT_OPEN:
	(void) snprintf(state, sizeof (state), "can't open");
	break;
	case VDEV_STATE_FAULTED:
	(void) snprintf(state, sizeof (state), "faulted");
	break;
	case VDEV_STATE_DEGRADED:
	(void) snprintf(state, sizeof (state), "degraded");
	break;
	case VDEV_STATE_HEALTHY:
	(void) snprintf(state, sizeof (state), "healthy");
	break;
	default:
	(void) snprintf(state, sizeof (state), "<state %u>",
	(uint_t)vd->vdev_state);
	}

	zfs_dbgmsg("%*svdev %u: %s%s, guid: %llu, path: %s, %s", indent,
	"", (int)vd->vdev_id, vd->vdev_ops->vdev_op_type,
	vd->vdev_islog ? " (log)" : "",
	(u_longlong_t)vd->vdev_guid,
	vd->vdev_path ? vd->vdev_path : "N/A", state);

	for (uint64_t i = 0; i < vd->vdev_children; i++)
	vdev_dbgmsg_print_tree(vd->vdev_child[i], indent + 2);
	}

	/*
	* Virtual device management.
	*/

	static vdev_ops_t *const vdev_ops_table[] = {
	&vdev_root_ops,
	&vdev_raidz_ops,
	&vdev_draid_ops,
	&vdev_draid_spare_ops,
	&vdev_mirror_ops,
	&vdev_replacing_ops,
	&vdev_spare_ops,
	&vdev_disk_ops,
	&vdev_file_ops,
	&vdev_missing_ops,
	&vdev_hole_ops,
	&vdev_indirect_ops,
	NULL
	};

	/*
	* Given a vdev type, return the appropriate ops vector.
	*/
	static vdev_ops_t *
	vdev_getops(const char *type)
	{
	vdev_ops_t ops, const *opspp;

	for (opspp = vdev_ops_table; (ops = *opspp) != NULL; opspp++)
	if (strcmp(ops->vdev_op_type, type) == 0)
	break;

	return (ops);
	}

	/*
	* Given a vdev and a metaslab class, find which metaslab group we're
	* interested in. All vdevs may belong to two different metaslab classes.
	* Dedicated slog devices use only the primary metaslab group, rather than a
	* separate log group. For embedded slogs, the vdev_log_mg will be non-NULL.
	*/
	metaslab_group_t *
	vdev_get_mg(vdev_t vd, metaslab_class_t mc)
	{
	if (mc == spa_embedded_log_class(vd->vdev_spa) &&
	vd->vdev_log_mg != NULL)
	return (vd->vdev_log_mg);
	else
	return (vd->vdev_mg);
	}

	void
	vdev_default_xlate(vdev_t vd, const range_seg64_t logical_rs,
	range_seg64_t physical_rs, range_seg64_t remain_rs)
	{
	(void) vd, (void) remain_rs;

	physical_rs->rs_start = logical_rs->rs_start;
	physical_rs->rs_end = logical_rs->rs_end;
	}

	/*
	* Derive the enumerated allocation bias from string input.
	* String origin is either the per-vdev zap or zpool(8).
	*/
	static vdev_alloc_bias_t
	vdev_derive_alloc_bias(const char *bias)
	{
	vdev_alloc_bias_t alloc_bias = VDEV_BIAS_NONE;

	if (strcmp(bias, VDEV_ALLOC_BIAS_LOG) == 0)
	alloc_bias = VDEV_BIAS_LOG;
	else if (strcmp(bias, VDEV_ALLOC_BIAS_SPECIAL) == 0)
	alloc_bias = VDEV_BIAS_SPECIAL;
	else if (strcmp(bias, VDEV_ALLOC_BIAS_DEDUP) == 0)
	alloc_bias = VDEV_BIAS_DEDUP;

	return (alloc_bias);
	}

	/*
	* Default asize function: return the MAX of psize with the asize of
	* all children. This is what's used by anything other than RAID-Z.
	*/
	uint64_t
	vdev_default_asize(vdev_t *vd, uint64_t psize, uint64_t txg)
	{
	uint64_t asize = P2ROUNDUP(psize, 1ULL << vd->vdev_top->vdev_ashift);
	uint64_t csize;

	for (int c = 0; c < vd->vdev_children; c++) {
	csize = vdev_psize_to_asize_txg(vd->vdev_child[c], psize, txg);
	asize = MAX(asize, csize);
	}

	return (asize);
	}

	uint64_t
	vdev_default_min_asize(vdev_t *vd)
	{
	return (vd->vdev_min_asize);
	}

	/*
	* Get the minimum allocatable size. We define the allocatable size as
	* the vdev's asize rounded to the nearest metaslab. This allows us to
	* replace or attach devices which don't have the same physical size but
	* can still satisfy the same number of allocations.
	*/
	uint64_t
	vdev_get_min_asize(vdev_t *vd)
	{
	vdev_t *pvd = vd->vdev_parent;

	/*
	* If our parent is NULL (inactive spare or cache) or is the root,
	* just return our own asize.
	*/
	if (pvd == NULL)
	return (vd->vdev_asize);

	/*
	* The top-level vdev just returns the allocatable size rounded
	* to the nearest metaslab.
	*/
	if (vd == vd->vdev_top)
	- return (P2ALIGN(vd->vdev_asize, 1ULL << vd->vdev_ms_shift));
	+ return (P2ALIGN_TYPED(vd->vdev_asize, 1ULL << vd->vdev_ms_shift,
	+ uint64_t));

	return (pvd->vdev_ops->vdev_op_min_asize(pvd));
	}

	void
	vdev_set_min_asize(vdev_t *vd)
	{
	vd->vdev_min_asize = vdev_get_min_asize(vd);

	for (int c = 0; c < vd->vdev_children; c++)
	vdev_set_min_asize(vd->vdev_child[c]);
	}

	/*
	* Get the minimal allocation size for the top-level vdev.
	*/
	uint64_t
	vdev_get_min_alloc(vdev_t *vd)
	{
	uint64_t min_alloc = 1ULL << vd->vdev_ashift;

	if (vd->vdev_ops->vdev_op_min_alloc != NULL)
	min_alloc = vd->vdev_ops->vdev_op_min_alloc(vd);

	return (min_alloc);
	}

	/*
	* Get the parity level for a top-level vdev.
	*/
	uint64_t
	vdev_get_nparity(vdev_t *vd)
	{
	uint64_t nparity = 0;

	if (vd->vdev_ops->vdev_op_nparity != NULL)
	nparity = vd->vdev_ops->vdev_op_nparity(vd);

	return (nparity);
	}

	static int
	vdev_prop_get_int(vdev_t vd, vdev_prop_t prop, uint64_t value)
	{
	spa_t *spa = vd->vdev_spa;
	objset_t *mos = spa->spa_meta_objset;
	uint64_t objid;
	int err;

	if (vd->vdev_root_zap != 0) {
	objid = vd->vdev_root_zap;
	} else if (vd->vdev_top_zap != 0) {
	objid = vd->vdev_top_zap;
	} else if (vd->vdev_leaf_zap != 0) {
	objid = vd->vdev_leaf_zap;
	} else {
	return (EINVAL);
	}

	err = zap_lookup(mos, objid, vdev_prop_to_name(prop),
	sizeof (uint64_t), 1, value);

	if (err == ENOENT)
	*value = vdev_prop_default_numeric(prop);

	return (err);
	}

	/*
	* Get the number of data disks for a top-level vdev.
	*/
	uint64_t
	vdev_get_ndisks(vdev_t *vd)
	{
	uint64_t ndisks = 1;

	if (vd->vdev_ops->vdev_op_ndisks != NULL)
	ndisks = vd->vdev_ops->vdev_op_ndisks(vd);

	return (ndisks);
	}

	vdev_t *
	vdev_lookup_top(spa_t *spa, uint64_t vdev)
	{
	vdev_t *rvd = spa->spa_root_vdev;

	ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0);

	if (vdev < rvd->vdev_children) {
	ASSERT(rvd->vdev_child[vdev] != NULL);
	return (rvd->vdev_child[vdev]);
	}

	return (NULL);
	}

	vdev_t *
	vdev_lookup_by_guid(vdev_t *vd, uint64_t guid)
	{
	vdev_t *mvd;

	if (vd->vdev_guid == guid)
	return (vd);

	for (int c = 0; c < vd->vdev_children; c++)
	if ((mvd = vdev_lookup_by_guid(vd->vdev_child[c], guid)) !=
	NULL)
	return (mvd);

	return (NULL);
	}

	static int
	vdev_count_leaves_impl(vdev_t *vd)
	{
	int n = 0;

	if (vd->vdev_ops->vdev_op_leaf)
	return (1);

	for (int c = 0; c < vd->vdev_children; c++)
	n += vdev_count_leaves_impl(vd->vdev_child[c]);

	return (n);
	}

	int
	vdev_count_leaves(spa_t *spa)
	{
	int rc;

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
	rc = vdev_count_leaves_impl(spa->spa_root_vdev);
	spa_config_exit(spa, SCL_VDEV, FTAG);

	return (rc);
	}

	void
	vdev_add_child(vdev_t pvd, vdev_t cvd)
	{
	size_t oldsize, newsize;
	uint64_t id = cvd->vdev_id;
	vdev_t **newchild;

	ASSERT(spa_config_held(cvd->vdev_spa, SCL_ALL, RW_WRITER) == SCL_ALL);
	ASSERT(cvd->vdev_parent == NULL);

	cvd->vdev_parent = pvd;

	if (pvd == NULL)
	return;

	ASSERT(id >= pvd->vdev_children \|\| pvd->vdev_child[id] == NULL);

	oldsize = pvd->vdev_children * sizeof (vdev_t *);
	pvd->vdev_children = MAX(pvd->vdev_children, id + 1);
	newsize = pvd->vdev_children * sizeof (vdev_t *);

	newchild = kmem_alloc(newsize, KM_SLEEP);
	if (pvd->vdev_child != NULL) {
	memcpy(newchild, pvd->vdev_child, oldsize);
	kmem_free(pvd->vdev_child, oldsize);
	}

	pvd->vdev_child = newchild;
	pvd->vdev_child[id] = cvd;

	cvd->vdev_top = (pvd->vdev_top ? pvd->vdev_top: cvd);
	ASSERT(cvd->vdev_top->vdev_parent->vdev_parent == NULL);

	/*
	* Walk up all ancestors to update guid sum.
	*/
	for (; pvd != NULL; pvd = pvd->vdev_parent)
	pvd->vdev_guid_sum += cvd->vdev_guid_sum;

	if (cvd->vdev_ops->vdev_op_leaf) {
	list_insert_head(&cvd->vdev_spa->spa_leaf_list, cvd);
	cvd->vdev_spa->spa_leaf_list_gen++;
	}
	}

	void
	vdev_remove_child(vdev_t pvd, vdev_t cvd)
	{
	int c;
	uint_t id = cvd->vdev_id;

	ASSERT(cvd->vdev_parent == pvd);

	if (pvd == NULL)
	return;

	ASSERT(id < pvd->vdev_children);
	ASSERT(pvd->vdev_child[id] == cvd);

	pvd->vdev_child[id] = NULL;
	cvd->vdev_parent = NULL;

	for (c = 0; c < pvd->vdev_children; c++)
	if (pvd->vdev_child[c])
	break;

	if (c == pvd->vdev_children) {
	kmem_free(pvd->vdev_child, c * sizeof (vdev_t *));
	pvd->vdev_child = NULL;
	pvd->vdev_children = 0;
	}

	if (cvd->vdev_ops->vdev_op_leaf) {
	spa_t *spa = cvd->vdev_spa;
	list_remove(&spa->spa_leaf_list, cvd);
	spa->spa_leaf_list_gen++;
	}

	/*
	* Walk up all ancestors to update guid sum.
	*/
	for (; pvd != NULL; pvd = pvd->vdev_parent)
	pvd->vdev_guid_sum -= cvd->vdev_guid_sum;
	}

	/*
	* Remove any holes in the child array.
	*/
	void
	vdev_compact_children(vdev_t *pvd)
	{
	vdev_t *newchild, cvd;
	int oldc = pvd->vdev_children;
	int newc;

	ASSERT(spa_config_held(pvd->vdev_spa, SCL_ALL, RW_WRITER) == SCL_ALL);

	if (oldc == 0)
	return;

	for (int c = newc = 0; c < oldc; c++)
	if (pvd->vdev_child[c])
	newc++;

	if (newc > 0) {
	newchild = kmem_zalloc(newc * sizeof (vdev_t *), KM_SLEEP);

	for (int c = newc = 0; c < oldc; c++) {
	if ((cvd = pvd->vdev_child[c]) != NULL) {
	newchild[newc] = cvd;
	cvd->vdev_id = newc++;
	}
	}
	} else {
	newchild = NULL;
	}

	kmem_free(pvd->vdev_child, oldc * sizeof (vdev_t *));
	pvd->vdev_child = newchild;
	pvd->vdev_children = newc;
	}

	/*
	* Allocate and minimally initialize a vdev_t.
	*/
	vdev_t *
	vdev_alloc_common(spa_t spa, uint_t id, uint64_t guid, vdev_ops_t ops)
	{
	vdev_t *vd;
	vdev_indirect_config_t *vic;

	vd = kmem_zalloc(sizeof (vdev_t), KM_SLEEP);
	vic = &vd->vdev_indirect_config;

	if (spa->spa_root_vdev == NULL) {
	ASSERT(ops == &vdev_root_ops);
	spa->spa_root_vdev = vd;
	spa->spa_load_guid = spa_generate_guid(NULL);
	}

	if (guid == 0 && ops != &vdev_hole_ops) {
	if (spa->spa_root_vdev == vd) {
	/*
	* The root vdev's guid will also be the pool guid,
	* which must be unique among all pools.
	*/
	guid = spa_generate_guid(NULL);
	} else {
	/*
	* Any other vdev's guid must be unique within the pool.
	*/
	guid = spa_generate_guid(spa);
	}
	ASSERT(!spa_guid_exists(spa_guid(spa), guid));
	}

	vd->vdev_spa = spa;
	vd->vdev_id = id;
	vd->vdev_guid = guid;
	vd->vdev_guid_sum = guid;
	vd->vdev_ops = ops;
	vd->vdev_state = VDEV_STATE_CLOSED;
	vd->vdev_ishole = (ops == &vdev_hole_ops);
	vic->vic_prev_indirect_vdev = UINT64_MAX;

	rw_init(&vd->vdev_indirect_rwlock, NULL, RW_DEFAULT, NULL);
	mutex_init(&vd->vdev_obsolete_lock, NULL, MUTEX_DEFAULT, NULL);
	vd->vdev_obsolete_segments = range_tree_create(NULL, RANGE_SEG64, NULL,
	0, 0);

	/*
	* Initialize rate limit structs for events. We rate limit ZIO delay
	* and checksum events so that we don't overwhelm ZED with thousands
	* of events when a disk is acting up.
	*/
	zfs_ratelimit_init(&vd->vdev_delay_rl, &zfs_slow_io_events_per_second,
	1);
	- zfs_ratelimit_init(&vd->vdev_deadman_rl, &zfs_slow_io_events_per_second,
	+ zfs_ratelimit_init(&vd->vdev_deadman_rl, &zfs_deadman_events_per_second,
	1);
	zfs_ratelimit_init(&vd->vdev_checksum_rl,
	&zfs_checksum_events_per_second, 1);

	/*
	* Default Thresholds for tuning ZED
	*/
	vd->vdev_checksum_n = vdev_prop_default_numeric(VDEV_PROP_CHECKSUM_N);
	vd->vdev_checksum_t = vdev_prop_default_numeric(VDEV_PROP_CHECKSUM_T);
	vd->vdev_io_n = vdev_prop_default_numeric(VDEV_PROP_IO_N);
	vd->vdev_io_t = vdev_prop_default_numeric(VDEV_PROP_IO_T);
	vd->vdev_slow_io_n = vdev_prop_default_numeric(VDEV_PROP_SLOW_IO_N);
	vd->vdev_slow_io_t = vdev_prop_default_numeric(VDEV_PROP_SLOW_IO_T);

	list_link_init(&vd->vdev_config_dirty_node);
	list_link_init(&vd->vdev_state_dirty_node);
	list_link_init(&vd->vdev_initialize_node);
	list_link_init(&vd->vdev_leaf_node);
	list_link_init(&vd->vdev_trim_node);

	mutex_init(&vd->vdev_dtl_lock, NULL, MUTEX_NOLOCKDEP, NULL);
	mutex_init(&vd->vdev_stat_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&vd->vdev_probe_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&vd->vdev_scan_io_queue_lock, NULL, MUTEX_DEFAULT, NULL);

	mutex_init(&vd->vdev_initialize_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&vd->vdev_initialize_io_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&vd->vdev_initialize_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&vd->vdev_initialize_io_cv, NULL, CV_DEFAULT, NULL);

	mutex_init(&vd->vdev_trim_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&vd->vdev_autotrim_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&vd->vdev_trim_io_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&vd->vdev_trim_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&vd->vdev_autotrim_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&vd->vdev_autotrim_kick_cv, NULL, CV_DEFAULT, NULL);
	cv_init(&vd->vdev_trim_io_cv, NULL, CV_DEFAULT, NULL);

	mutex_init(&vd->vdev_rebuild_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&vd->vdev_rebuild_cv, NULL, CV_DEFAULT, NULL);

	for (int t = 0; t < DTL_TYPES; t++) {
	vd->vdev_dtl[t] = range_tree_create(NULL, RANGE_SEG64, NULL, 0,
	0);
	}

	txg_list_create(&vd->vdev_ms_list, spa,
	offsetof(struct metaslab, ms_txg_node));
	txg_list_create(&vd->vdev_dtl_list, spa,
	offsetof(struct vdev, vdev_dtl_node));
	vd->vdev_stat.vs_timestamp = gethrtime();
	vdev_queue_init(vd);

	return (vd);
	}

	/*
	* Allocate a new vdev. The 'alloctype' is used to control whether we are
	* creating a new vdev or loading an existing one - the behavior is slightly
	* different for each case.
	*/
	int
	vdev_alloc(spa_t spa, vdev_t vdp, nvlist_t nv, vdev_t *parent, uint_t id,
	int alloctype)
	{
	vdev_ops_t *ops;
	const char *type;
	uint64_t guid = 0, islog;
	vdev_t *vd;
	vdev_indirect_config_t *vic;
	const char *tmp = NULL;
	int rc;
	vdev_alloc_bias_t alloc_bias = VDEV_BIAS_NONE;
	boolean_t top_level = (parent && !parent->vdev_parent);

	ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);

	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0)
	return (SET_ERROR(EINVAL));

	if ((ops = vdev_getops(type)) == NULL)
	return (SET_ERROR(EINVAL));

	/*
	* If this is a load, get the vdev guid from the nvlist.
	* Otherwise, vdev_alloc_common() will generate one for us.
	*/
	if (alloctype == VDEV_ALLOC_LOAD) {
	uint64_t label_id;

	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID, &label_id) \|\|
	label_id != id)
	return (SET_ERROR(EINVAL));

	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) != 0)
	return (SET_ERROR(EINVAL));
	} else if (alloctype == VDEV_ALLOC_SPARE) {
	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) != 0)
	return (SET_ERROR(EINVAL));
	} else if (alloctype == VDEV_ALLOC_L2CACHE) {
	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) != 0)
	return (SET_ERROR(EINVAL));
	} else if (alloctype == VDEV_ALLOC_ROOTPOOL) {
	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) != 0)
	return (SET_ERROR(EINVAL));
	}

	/*
	* The first allocated vdev must be of type 'root'.
	*/
	if (ops != &vdev_root_ops && spa->spa_root_vdev == NULL)
	return (SET_ERROR(EINVAL));

	/*
	* Determine whether we're a log vdev.
	*/
	islog = 0;
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG, &islog);
	if (islog && spa_version(spa) < SPA_VERSION_SLOGS)
	return (SET_ERROR(ENOTSUP));

	if (ops == &vdev_hole_ops && spa_version(spa) < SPA_VERSION_HOLES)
	return (SET_ERROR(ENOTSUP));

	if (top_level && alloctype == VDEV_ALLOC_ADD) {
	const char *bias;

	/*
	* If creating a top-level vdev, check for allocation
	* classes input.
	*/
	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_ALLOCATION_BIAS,
	&bias) == 0) {
	alloc_bias = vdev_derive_alloc_bias(bias);

	/* spa_vdev_add() expects feature to be enabled */
	if (spa->spa_load_state != SPA_LOAD_CREATE &&
	!spa_feature_is_enabled(spa,
	SPA_FEATURE_ALLOCATION_CLASSES)) {
	return (SET_ERROR(ENOTSUP));
	}
	}

	/* spa_vdev_add() expects feature to be enabled */
	if (ops == &vdev_draid_ops &&
	spa->spa_load_state != SPA_LOAD_CREATE &&
	!spa_feature_is_enabled(spa, SPA_FEATURE_DRAID)) {
	return (SET_ERROR(ENOTSUP));
	}
	}

	/*
	* Initialize the vdev specific data. This is done before calling
	* vdev_alloc_common() since it may fail and this simplifies the
	* error reporting and cleanup code paths.
	*/
	void *tsd = NULL;
	if (ops->vdev_op_init != NULL) {
	rc = ops->vdev_op_init(spa, nv, &tsd);
	if (rc != 0) {
	return (rc);
	}
	}

	vd = vdev_alloc_common(spa, id, guid, ops);
	vd->vdev_tsd = tsd;
	vd->vdev_islog = islog;

	if (top_level && alloc_bias != VDEV_BIAS_NONE)
	vd->vdev_alloc_bias = alloc_bias;

	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &tmp) == 0)
	vd->vdev_path = spa_strdup(tmp);

	/*
	* ZPOOL_CONFIG_AUX_STATE = "external" means we previously forced a
	* fault on a vdev and want it to persist across imports (like with
	* zpool offline -f).
	*/
	rc = nvlist_lookup_string(nv, ZPOOL_CONFIG_AUX_STATE, &tmp);
	if (rc == 0 && tmp != NULL && strcmp(tmp, "external") == 0) {
	vd->vdev_stat.vs_aux = VDEV_AUX_EXTERNAL;
	vd->vdev_faulted = 1;
	vd->vdev_label_aux = VDEV_AUX_EXTERNAL;
	}

	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &tmp) == 0)
	vd->vdev_devid = spa_strdup(tmp);
	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PHYS_PATH, &tmp) == 0)
	vd->vdev_physpath = spa_strdup(tmp);

	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_VDEV_ENC_SYSFS_PATH,
	&tmp) == 0)
	vd->vdev_enc_sysfs_path = spa_strdup(tmp);

	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_FRU, &tmp) == 0)
	vd->vdev_fru = spa_strdup(tmp);

	/*
	* Set the whole_disk property. If it's not specified, leave the value
	* as -1.
	*/
	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
	&vd->vdev_wholedisk) != 0)
	vd->vdev_wholedisk = -1ULL;

	vic = &vd->vdev_indirect_config;

	ASSERT0(vic->vic_mapping_object);
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_INDIRECT_OBJECT,
	&vic->vic_mapping_object);
	ASSERT0(vic->vic_births_object);
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_INDIRECT_BIRTHS,
	&vic->vic_births_object);
	ASSERT3U(vic->vic_prev_indirect_vdev, ==, UINT64_MAX);
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_PREV_INDIRECT_VDEV,
	&vic->vic_prev_indirect_vdev);

	/*
	* Look for the 'not present' flag. This will only be set if the device
	* was not present at the time of import.
	*/
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
	&vd->vdev_not_present);

	/*
	* Get the alignment requirement. Ignore pool ashift for vdev
	* attach case.
	*/
	if (alloctype != VDEV_ALLOC_ATTACH) {
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ASHIFT,
	&vd->vdev_ashift);
	} else {
	vd->vdev_attaching = B_TRUE;
	}

	/*
	* Retrieve the vdev creation time.
	*/
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_CREATE_TXG,
	&vd->vdev_crtxg);

	if (vd->vdev_ops == &vdev_root_ops &&
	(alloctype == VDEV_ALLOC_LOAD \|\|
	alloctype == VDEV_ALLOC_SPLIT \|\|
	alloctype == VDEV_ALLOC_ROOTPOOL)) {
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_VDEV_ROOT_ZAP,
	&vd->vdev_root_zap);
	}

	/*
	* If we're a top-level vdev, try to load the allocation parameters.
	*/
	if (top_level &&
	(alloctype == VDEV_ALLOC_LOAD \|\| alloctype == VDEV_ALLOC_SPLIT)) {
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_METASLAB_ARRAY,
	&vd->vdev_ms_array);
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_METASLAB_SHIFT,
	&vd->vdev_ms_shift);
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ASIZE,
	&vd->vdev_asize);
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NONALLOCATING,
	&vd->vdev_noalloc);
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVING,
	&vd->vdev_removing);
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_VDEV_TOP_ZAP,
	&vd->vdev_top_zap);
	vd->vdev_rz_expanding = nvlist_exists(nv,
	ZPOOL_CONFIG_RAIDZ_EXPANDING);
	} else {
	ASSERT0(vd->vdev_top_zap);
	}

	if (top_level && alloctype != VDEV_ALLOC_ATTACH) {
	ASSERT(alloctype == VDEV_ALLOC_LOAD \|\|
	alloctype == VDEV_ALLOC_ADD \|\|
	alloctype == VDEV_ALLOC_SPLIT \|\|
	alloctype == VDEV_ALLOC_ROOTPOOL);
	/* Note: metaslab_group_create() is now deferred */
	}

	if (vd->vdev_ops->vdev_op_leaf &&
	(alloctype == VDEV_ALLOC_LOAD \|\| alloctype == VDEV_ALLOC_SPLIT)) {
	(void) nvlist_lookup_uint64(nv,
	ZPOOL_CONFIG_VDEV_LEAF_ZAP, &vd->vdev_leaf_zap);
	} else {
	ASSERT0(vd->vdev_leaf_zap);
	}

	/*
	* If we're a leaf vdev, try to load the DTL object and other state.
	*/

	if (vd->vdev_ops->vdev_op_leaf &&
	(alloctype == VDEV_ALLOC_LOAD \|\| alloctype == VDEV_ALLOC_L2CACHE \|\|
	alloctype == VDEV_ALLOC_ROOTPOOL)) {
	if (alloctype == VDEV_ALLOC_LOAD) {
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_DTL,
	&vd->vdev_dtl_object);
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_UNSPARE,
	&vd->vdev_unspare);
	}

	if (alloctype == VDEV_ALLOC_ROOTPOOL) {
	uint64_t spare = 0;

	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE,
	&spare) == 0 && spare)
	spa_spare_add(vd);
	}

	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE,
	&vd->vdev_offline);

	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_RESILVER_TXG,
	&vd->vdev_resilver_txg);

	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REBUILD_TXG,
	&vd->vdev_rebuild_txg);

	if (nvlist_exists(nv, ZPOOL_CONFIG_RESILVER_DEFER))
	vdev_defer_resilver(vd);

	/*
	* In general, when importing a pool we want to ignore the
	* persistent fault state, as the diagnosis made on another
	* system may not be valid in the current context. The only
	* exception is if we forced a vdev to a persistently faulted
	* state with 'zpool offline -f'. The persistent fault will
	* remain across imports until cleared.
	*
	* Local vdevs will remain in the faulted state.
	*/
	if (spa_load_state(spa) == SPA_LOAD_OPEN \|\|
	spa_load_state(spa) == SPA_LOAD_IMPORT) {
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED,
	&vd->vdev_faulted);
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_DEGRADED,
	&vd->vdev_degraded);
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED,
	&vd->vdev_removed);

	if (vd->vdev_faulted \|\| vd->vdev_degraded) {
	const char *aux;

	vd->vdev_label_aux =
	VDEV_AUX_ERR_EXCEEDED;
	if (nvlist_lookup_string(nv,
	ZPOOL_CONFIG_AUX_STATE, &aux) == 0 &&
	strcmp(aux, "external") == 0)
	vd->vdev_label_aux = VDEV_AUX_EXTERNAL;
	else
	vd->vdev_faulted = 0ULL;
	}
	}
	}

	/*
	* Add ourselves to the parent's list of children.
	*/
	vdev_add_child(parent, vd);

	*vdp = vd;

	return (0);
	}

	void
	vdev_free(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;

	ASSERT3P(vd->vdev_initialize_thread, ==, NULL);
	ASSERT3P(vd->vdev_trim_thread, ==, NULL);
	ASSERT3P(vd->vdev_autotrim_thread, ==, NULL);
	ASSERT3P(vd->vdev_rebuild_thread, ==, NULL);

	/*
	* Scan queues are normally destroyed at the end of a scan. If the
	* queue exists here, that implies the vdev is being removed while
	* the scan is still running.
	*/
	if (vd->vdev_scan_io_queue != NULL) {
	mutex_enter(&vd->vdev_scan_io_queue_lock);
	dsl_scan_io_queue_destroy(vd->vdev_scan_io_queue);
	vd->vdev_scan_io_queue = NULL;
	mutex_exit(&vd->vdev_scan_io_queue_lock);
	}

	/*
	* vdev_free() implies closing the vdev first. This is simpler than
	* trying to ensure complicated semantics for all callers.
	*/
	vdev_close(vd);

	ASSERT(!list_link_active(&vd->vdev_config_dirty_node));
	ASSERT(!list_link_active(&vd->vdev_state_dirty_node));

	/*
	* Free all children.
	*/
	for (int c = 0; c < vd->vdev_children; c++)
	vdev_free(vd->vdev_child[c]);

	ASSERT(vd->vdev_child == NULL);
	ASSERT(vd->vdev_guid_sum == vd->vdev_guid);

	if (vd->vdev_ops->vdev_op_fini != NULL)
	vd->vdev_ops->vdev_op_fini(vd);

	/*
	* Discard allocation state.
	*/
	if (vd->vdev_mg != NULL) {
	vdev_metaslab_fini(vd);
	metaslab_group_destroy(vd->vdev_mg);
	vd->vdev_mg = NULL;
	}
	if (vd->vdev_log_mg != NULL) {
	ASSERT0(vd->vdev_ms_count);
	metaslab_group_destroy(vd->vdev_log_mg);
	vd->vdev_log_mg = NULL;
	}

	ASSERT0(vd->vdev_stat.vs_space);
	ASSERT0(vd->vdev_stat.vs_dspace);
	ASSERT0(vd->vdev_stat.vs_alloc);

	/*
	* Remove this vdev from its parent's child list.
	*/
	vdev_remove_child(vd->vdev_parent, vd);

	ASSERT(vd->vdev_parent == NULL);
	ASSERT(!list_link_active(&vd->vdev_leaf_node));

	/*
	* Clean up vdev structure.
	*/
	vdev_queue_fini(vd);

	if (vd->vdev_path)
	spa_strfree(vd->vdev_path);
	if (vd->vdev_devid)
	spa_strfree(vd->vdev_devid);
	if (vd->vdev_physpath)
	spa_strfree(vd->vdev_physpath);

	if (vd->vdev_enc_sysfs_path)
	spa_strfree(vd->vdev_enc_sysfs_path);

	if (vd->vdev_fru)
	spa_strfree(vd->vdev_fru);

	if (vd->vdev_isspare)
	spa_spare_remove(vd);
	if (vd->vdev_isl2cache)
	spa_l2cache_remove(vd);

	txg_list_destroy(&vd->vdev_ms_list);
	txg_list_destroy(&vd->vdev_dtl_list);

	mutex_enter(&vd->vdev_dtl_lock);
	space_map_close(vd->vdev_dtl_sm);
	for (int t = 0; t < DTL_TYPES; t++) {
	range_tree_vacate(vd->vdev_dtl[t], NULL, NULL);
	range_tree_destroy(vd->vdev_dtl[t]);
	}
	mutex_exit(&vd->vdev_dtl_lock);

	EQUIV(vd->vdev_indirect_births != NULL,
	vd->vdev_indirect_mapping != NULL);
	if (vd->vdev_indirect_births != NULL) {
	vdev_indirect_mapping_close(vd->vdev_indirect_mapping);
	vdev_indirect_births_close(vd->vdev_indirect_births);
	}

	if (vd->vdev_obsolete_sm != NULL) {
	ASSERT(vd->vdev_removing \|\|
	vd->vdev_ops == &vdev_indirect_ops);
	space_map_close(vd->vdev_obsolete_sm);
	vd->vdev_obsolete_sm = NULL;
	}
	range_tree_destroy(vd->vdev_obsolete_segments);
	rw_destroy(&vd->vdev_indirect_rwlock);
	mutex_destroy(&vd->vdev_obsolete_lock);

	mutex_destroy(&vd->vdev_dtl_lock);
	mutex_destroy(&vd->vdev_stat_lock);
	mutex_destroy(&vd->vdev_probe_lock);
	mutex_destroy(&vd->vdev_scan_io_queue_lock);

	mutex_destroy(&vd->vdev_initialize_lock);
	mutex_destroy(&vd->vdev_initialize_io_lock);
	cv_destroy(&vd->vdev_initialize_io_cv);
	cv_destroy(&vd->vdev_initialize_cv);

	mutex_destroy(&vd->vdev_trim_lock);
	mutex_destroy(&vd->vdev_autotrim_lock);
	mutex_destroy(&vd->vdev_trim_io_lock);
	cv_destroy(&vd->vdev_trim_cv);
	cv_destroy(&vd->vdev_autotrim_cv);
	cv_destroy(&vd->vdev_autotrim_kick_cv);
	cv_destroy(&vd->vdev_trim_io_cv);

	mutex_destroy(&vd->vdev_rebuild_lock);
	cv_destroy(&vd->vdev_rebuild_cv);

	zfs_ratelimit_fini(&vd->vdev_delay_rl);
	zfs_ratelimit_fini(&vd->vdev_deadman_rl);
	zfs_ratelimit_fini(&vd->vdev_checksum_rl);

	if (vd == spa->spa_root_vdev)
	spa->spa_root_vdev = NULL;

	kmem_free(vd, sizeof (vdev_t));
	}

	/*
	* Transfer top-level vdev state from svd to tvd.
	*/
	static void
	vdev_top_transfer(vdev_t svd, vdev_t tvd)
	{
	spa_t *spa = svd->vdev_spa;
	metaslab_t *msp;
	vdev_t *vd;
	int t;

	ASSERT(tvd == tvd->vdev_top);

	tvd->vdev_ms_array = svd->vdev_ms_array;
	tvd->vdev_ms_shift = svd->vdev_ms_shift;
	tvd->vdev_ms_count = svd->vdev_ms_count;
	tvd->vdev_top_zap = svd->vdev_top_zap;

	svd->vdev_ms_array = 0;
	svd->vdev_ms_shift = 0;
	svd->vdev_ms_count = 0;
	svd->vdev_top_zap = 0;

	if (tvd->vdev_mg)
	ASSERT3P(tvd->vdev_mg, ==, svd->vdev_mg);
	if (tvd->vdev_log_mg)
	ASSERT3P(tvd->vdev_log_mg, ==, svd->vdev_log_mg);
	tvd->vdev_mg = svd->vdev_mg;
	tvd->vdev_log_mg = svd->vdev_log_mg;
	tvd->vdev_ms = svd->vdev_ms;

	svd->vdev_mg = NULL;
	svd->vdev_log_mg = NULL;
	svd->vdev_ms = NULL;

	if (tvd->vdev_mg != NULL)
	tvd->vdev_mg->mg_vd = tvd;
	if (tvd->vdev_log_mg != NULL)
	tvd->vdev_log_mg->mg_vd = tvd;

	tvd->vdev_checkpoint_sm = svd->vdev_checkpoint_sm;
	svd->vdev_checkpoint_sm = NULL;

	tvd->vdev_alloc_bias = svd->vdev_alloc_bias;
	svd->vdev_alloc_bias = VDEV_BIAS_NONE;

	tvd->vdev_stat.vs_alloc = svd->vdev_stat.vs_alloc;
	tvd->vdev_stat.vs_space = svd->vdev_stat.vs_space;
	tvd->vdev_stat.vs_dspace = svd->vdev_stat.vs_dspace;

	svd->vdev_stat.vs_alloc = 0;
	svd->vdev_stat.vs_space = 0;
	svd->vdev_stat.vs_dspace = 0;

	/*
	* State which may be set on a top-level vdev that's in the
	* process of being removed.
	*/
	ASSERT0(tvd->vdev_indirect_config.vic_births_object);
	ASSERT0(tvd->vdev_indirect_config.vic_mapping_object);
	ASSERT3U(tvd->vdev_indirect_config.vic_prev_indirect_vdev, ==, -1ULL);
	ASSERT3P(tvd->vdev_indirect_mapping, ==, NULL);
	ASSERT3P(tvd->vdev_indirect_births, ==, NULL);
	ASSERT3P(tvd->vdev_obsolete_sm, ==, NULL);
	ASSERT0(tvd->vdev_noalloc);
	ASSERT0(tvd->vdev_removing);
	ASSERT0(tvd->vdev_rebuilding);
	tvd->vdev_noalloc = svd->vdev_noalloc;
	tvd->vdev_removing = svd->vdev_removing;
	tvd->vdev_rebuilding = svd->vdev_rebuilding;
	tvd->vdev_rebuild_config = svd->vdev_rebuild_config;
	tvd->vdev_indirect_config = svd->vdev_indirect_config;
	tvd->vdev_indirect_mapping = svd->vdev_indirect_mapping;
	tvd->vdev_indirect_births = svd->vdev_indirect_births;
	range_tree_swap(&svd->vdev_obsolete_segments,
	&tvd->vdev_obsolete_segments);
	tvd->vdev_obsolete_sm = svd->vdev_obsolete_sm;
	svd->vdev_indirect_config.vic_mapping_object = 0;
	svd->vdev_indirect_config.vic_births_object = 0;
	svd->vdev_indirect_config.vic_prev_indirect_vdev = -1ULL;
	svd->vdev_indirect_mapping = NULL;
	svd->vdev_indirect_births = NULL;
	svd->vdev_obsolete_sm = NULL;
	svd->vdev_noalloc = 0;
	svd->vdev_removing = 0;
	svd->vdev_rebuilding = 0;

	for (t = 0; t < TXG_SIZE; t++) {
	while ((msp = txg_list_remove(&svd->vdev_ms_list, t)) != NULL)
	(void) txg_list_add(&tvd->vdev_ms_list, msp, t);
	while ((vd = txg_list_remove(&svd->vdev_dtl_list, t)) != NULL)
	(void) txg_list_add(&tvd->vdev_dtl_list, vd, t);
	if (txg_list_remove_this(&spa->spa_vdev_txg_list, svd, t))
	(void) txg_list_add(&spa->spa_vdev_txg_list, tvd, t);
	}

	if (list_link_active(&svd->vdev_config_dirty_node)) {
	vdev_config_clean(svd);
	vdev_config_dirty(tvd);
	}

	if (list_link_active(&svd->vdev_state_dirty_node)) {
	vdev_state_clean(svd);
	vdev_state_dirty(tvd);
	}

	tvd->vdev_deflate_ratio = svd->vdev_deflate_ratio;
	svd->vdev_deflate_ratio = 0;

	tvd->vdev_islog = svd->vdev_islog;
	svd->vdev_islog = 0;

	dsl_scan_io_queue_vdev_xfer(svd, tvd);
	}

	static void
	vdev_top_update(vdev_t tvd, vdev_t vd)
	{
	if (vd == NULL)
	return;

	vd->vdev_top = tvd;

	for (int c = 0; c < vd->vdev_children; c++)
	vdev_top_update(tvd, vd->vdev_child[c]);
	}

	/*
	* Add a mirror/replacing vdev above an existing vdev. There is no need to
	* call .vdev_op_init() since mirror/replacing vdevs do not have private state.
	*/
	vdev_t *
	vdev_add_parent(vdev_t cvd, vdev_ops_t ops)
	{
	spa_t *spa = cvd->vdev_spa;
	vdev_t *pvd = cvd->vdev_parent;
	vdev_t *mvd;

	ASSERT(spa_config_held(spa, SCL_ALL, RW_WRITER) == SCL_ALL);

	mvd = vdev_alloc_common(spa, cvd->vdev_id, 0, ops);

	mvd->vdev_asize = cvd->vdev_asize;
	mvd->vdev_min_asize = cvd->vdev_min_asize;
	mvd->vdev_max_asize = cvd->vdev_max_asize;
	mvd->vdev_psize = cvd->vdev_psize;
	mvd->vdev_ashift = cvd->vdev_ashift;
	mvd->vdev_logical_ashift = cvd->vdev_logical_ashift;
	mvd->vdev_physical_ashift = cvd->vdev_physical_ashift;
	mvd->vdev_state = cvd->vdev_state;
	mvd->vdev_crtxg = cvd->vdev_crtxg;

	vdev_remove_child(pvd, cvd);
	vdev_add_child(pvd, mvd);
	cvd->vdev_id = mvd->vdev_children;
	vdev_add_child(mvd, cvd);
	vdev_top_update(cvd->vdev_top, cvd->vdev_top);

	if (mvd == mvd->vdev_top)
	vdev_top_transfer(cvd, mvd);

	return (mvd);
	}

	/*
	* Remove a 1-way mirror/replacing vdev from the tree.
	*/
	void
	vdev_remove_parent(vdev_t *cvd)
	{
	vdev_t *mvd = cvd->vdev_parent;
	vdev_t *pvd = mvd->vdev_parent;

	ASSERT(spa_config_held(cvd->vdev_spa, SCL_ALL, RW_WRITER) == SCL_ALL);

	ASSERT(mvd->vdev_children == 1);
	ASSERT(mvd->vdev_ops == &vdev_mirror_ops \|\|
	mvd->vdev_ops == &vdev_replacing_ops \|\|
	mvd->vdev_ops == &vdev_spare_ops);
	cvd->vdev_ashift = mvd->vdev_ashift;
	cvd->vdev_logical_ashift = mvd->vdev_logical_ashift;
	cvd->vdev_physical_ashift = mvd->vdev_physical_ashift;
	vdev_remove_child(mvd, cvd);
	vdev_remove_child(pvd, mvd);

	/*
	* If cvd will replace mvd as a top-level vdev, preserve mvd's guid.
	* Otherwise, we could have detached an offline device, and when we
	* go to import the pool we'll think we have two top-level vdevs,
	* instead of a different version of the same top-level vdev.
	*/
	if (mvd->vdev_top == mvd) {
	uint64_t guid_delta = mvd->vdev_guid - cvd->vdev_guid;
	cvd->vdev_orig_guid = cvd->vdev_guid;
	cvd->vdev_guid += guid_delta;
	cvd->vdev_guid_sum += guid_delta;

	/*
	* If pool not set for autoexpand, we need to also preserve
	* mvd's asize to prevent automatic expansion of cvd.
	* Otherwise if we are adjusting the mirror by attaching and
	* detaching children of non-uniform sizes, the mirror could
	* autoexpand, unexpectedly requiring larger devices to
	* re-establish the mirror.
	*/
	if (!cvd->vdev_spa->spa_autoexpand)
	cvd->vdev_asize = mvd->vdev_asize;
	}
	cvd->vdev_id = mvd->vdev_id;
	vdev_add_child(pvd, cvd);
	vdev_top_update(cvd->vdev_top, cvd->vdev_top);

	if (cvd == cvd->vdev_top)
	vdev_top_transfer(mvd, cvd);

	ASSERT(mvd->vdev_children == 0);
	vdev_free(mvd);
	}

	/*
	* Choose GCD for spa_gcd_alloc.
	*/
	static uint64_t
	vdev_gcd(uint64_t a, uint64_t b)
	{
	while (b != 0) {
	uint64_t t = b;
	b = a % b;
	a = t;
	}
	return (a);
	}

	/*
	* Set spa_min_alloc and spa_gcd_alloc.
	*/
	static void
	vdev_spa_set_alloc(spa_t *spa, uint64_t min_alloc)
	{
	if (min_alloc < spa->spa_min_alloc)
	spa->spa_min_alloc = min_alloc;
	if (spa->spa_gcd_alloc == INT_MAX) {
	spa->spa_gcd_alloc = min_alloc;
	} else {
	spa->spa_gcd_alloc = vdev_gcd(min_alloc,
	spa->spa_gcd_alloc);
	}
	}

	void
	vdev_metaslab_group_create(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;

	/*
	* metaslab_group_create was delayed until allocation bias was available
	*/
	if (vd->vdev_mg == NULL) {
	metaslab_class_t *mc;

	if (vd->vdev_islog && vd->vdev_alloc_bias == VDEV_BIAS_NONE)
	vd->vdev_alloc_bias = VDEV_BIAS_LOG;

	ASSERT3U(vd->vdev_islog, ==,
	(vd->vdev_alloc_bias == VDEV_BIAS_LOG));

	switch (vd->vdev_alloc_bias) {
	case VDEV_BIAS_LOG:
	mc = spa_log_class(spa);
	break;
	case VDEV_BIAS_SPECIAL:
	mc = spa_special_class(spa);
	break;
	case VDEV_BIAS_DEDUP:
	mc = spa_dedup_class(spa);
	break;
	default:
	mc = spa_normal_class(spa);
	}

	vd->vdev_mg = metaslab_group_create(mc, vd,
	spa->spa_alloc_count);

	if (!vd->vdev_islog) {
	vd->vdev_log_mg = metaslab_group_create(
	spa_embedded_log_class(spa), vd, 1);
	}

	/*
	* The spa ashift min/max only apply for the normal metaslab
	* class. Class destination is late binding so ashift boundary
	* setting had to wait until now.
	*/
	if (vd->vdev_top == vd && vd->vdev_ashift != 0 &&
	mc == spa_normal_class(spa) && vd->vdev_aux == NULL) {
	if (vd->vdev_ashift > spa->spa_max_ashift)
	spa->spa_max_ashift = vd->vdev_ashift;
	if (vd->vdev_ashift < spa->spa_min_ashift)
	spa->spa_min_ashift = vd->vdev_ashift;

	uint64_t min_alloc = vdev_get_min_alloc(vd);
	vdev_spa_set_alloc(spa, min_alloc);
	}
	}
	}

	int
	vdev_metaslab_init(vdev_t *vd, uint64_t txg)
	{
	spa_t *spa = vd->vdev_spa;
	uint64_t oldc = vd->vdev_ms_count;
	uint64_t newc = vd->vdev_asize >> vd->vdev_ms_shift;
	metaslab_t **mspp;
	int error;
	boolean_t expanding = (oldc != 0);

	ASSERT(txg == 0 \|\| spa_config_held(spa, SCL_ALLOC, RW_WRITER));

	/*
	* This vdev is not being allocated from yet or is a hole.
	*/
	if (vd->vdev_ms_shift == 0)
	return (0);

	ASSERT(!vd->vdev_ishole);

	ASSERT(oldc <= newc);

	mspp = vmem_zalloc(newc * sizeof (*mspp), KM_SLEEP);

	if (expanding) {
	memcpy(mspp, vd->vdev_ms, oldc * sizeof (*mspp));
	vmem_free(vd->vdev_ms, oldc * sizeof (*mspp));
	}

	vd->vdev_ms = mspp;
	vd->vdev_ms_count = newc;

	for (uint64_t m = oldc; m < newc; m++) {
	uint64_t object = 0;
	/*
	* vdev_ms_array may be 0 if we are creating the "fake"
	* metaslabs for an indirect vdev for zdb's leak detection.
	* See zdb_leak_init().
	*/
	if (txg == 0 && vd->vdev_ms_array != 0) {
	error = dmu_read(spa->spa_meta_objset,
	vd->vdev_ms_array,
	m * sizeof (uint64_t), sizeof (uint64_t), &object,
	DMU_READ_PREFETCH);
	if (error != 0) {
	vdev_dbgmsg(vd, "unable to read the metaslab "
	"array [error=%d]", error);
	return (error);
	}
	}

	error = metaslab_init(vd->vdev_mg, m, object, txg,
	&(vd->vdev_ms[m]));
	if (error != 0) {
	vdev_dbgmsg(vd, "metaslab_init failed [error=%d]",
	error);
	return (error);
	}
	}

	/*
	* Find the emptiest metaslab on the vdev and mark it for use for
	* embedded slog by moving it from the regular to the log metaslab
	* group.
	*/
	if (vd->vdev_mg->mg_class == spa_normal_class(spa) &&
	vd->vdev_ms_count > zfs_embedded_slog_min_ms &&
	avl_is_empty(&vd->vdev_log_mg->mg_metaslab_tree)) {
	uint64_t slog_msid = 0;
	uint64_t smallest = UINT64_MAX;

	/*
	* Note, we only search the new metaslabs, because the old
	* (pre-existing) ones may be active (e.g. have non-empty
	* range_tree's), and we don't move them to the new
	* metaslab_t.
	*/
	for (uint64_t m = oldc; m < newc; m++) {
	uint64_t alloc =
	space_map_allocated(vd->vdev_ms[m]->ms_sm);
	if (alloc < smallest) {
	slog_msid = m;
	smallest = alloc;
	}
	}
	metaslab_t *slog_ms = vd->vdev_ms[slog_msid];
	/*
	* The metaslab was marked as dirty at the end of
	* metaslab_init(). Remove it from the dirty list so that we
	* can uninitialize and reinitialize it to the new class.
	*/
	if (txg != 0) {
	(void) txg_list_remove_this(&vd->vdev_ms_list,
	slog_ms, txg);
	}
	uint64_t sm_obj = space_map_object(slog_ms->ms_sm);
	metaslab_fini(slog_ms);
	VERIFY0(metaslab_init(vd->vdev_log_mg, slog_msid, sm_obj, txg,
	&vd->vdev_ms[slog_msid]));
	}

	if (txg == 0)
	spa_config_enter(spa, SCL_ALLOC, FTAG, RW_WRITER);

	/*
	* If the vdev is marked as non-allocating then don't
	* activate the metaslabs since we want to ensure that
	* no allocations are performed on this device.
	*/
	if (vd->vdev_noalloc) {
	/* track non-allocating vdev space */
	spa->spa_nonallocating_dspace += spa_deflate(spa) ?
	vd->vdev_stat.vs_dspace : vd->vdev_stat.vs_space;
	} else if (!expanding) {
	metaslab_group_activate(vd->vdev_mg);
	if (vd->vdev_log_mg != NULL)
	metaslab_group_activate(vd->vdev_log_mg);
	}

	if (txg == 0)
	spa_config_exit(spa, SCL_ALLOC, FTAG);

	return (0);
	}

	void
	vdev_metaslab_fini(vdev_t *vd)
	{
	if (vd->vdev_checkpoint_sm != NULL) {
	ASSERT(spa_feature_is_active(vd->vdev_spa,
	SPA_FEATURE_POOL_CHECKPOINT));
	space_map_close(vd->vdev_checkpoint_sm);
	/*
	* Even though we close the space map, we need to set its
	* pointer to NULL. The reason is that vdev_metaslab_fini()
	* may be called multiple times for certain operations
	* (i.e. when destroying a pool) so we need to ensure that
	* this clause never executes twice. This logic is similar
	* to the one used for the vdev_ms clause below.
	*/
	vd->vdev_checkpoint_sm = NULL;
	}

	if (vd->vdev_ms != NULL) {
	metaslab_group_t *mg = vd->vdev_mg;

	metaslab_group_passivate(mg);
	if (vd->vdev_log_mg != NULL) {
	ASSERT(!vd->vdev_islog);
	metaslab_group_passivate(vd->vdev_log_mg);
	}

	uint64_t count = vd->vdev_ms_count;
	for (uint64_t m = 0; m < count; m++) {
	metaslab_t *msp = vd->vdev_ms[m];
	if (msp != NULL)
	metaslab_fini(msp);
	}
	vmem_free(vd->vdev_ms, count * sizeof (metaslab_t *));
	vd->vdev_ms = NULL;
	vd->vdev_ms_count = 0;

	for (int i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i++) {
	ASSERT0(mg->mg_histogram[i]);
	if (vd->vdev_log_mg != NULL)
	ASSERT0(vd->vdev_log_mg->mg_histogram[i]);
	}
	}
	ASSERT0(vd->vdev_ms_count);
	}

	typedef struct vdev_probe_stats {
	boolean_t vps_readable;
	boolean_t vps_writeable;
	boolean_t vps_zio_done_probe;
	int vps_flags;
	} vdev_probe_stats_t;

	static void
	vdev_probe_done(zio_t *zio)
	{
	spa_t *spa = zio->io_spa;
	vdev_t *vd = zio->io_vd;
	vdev_probe_stats_t *vps = zio->io_private;

	ASSERT(vd->vdev_probe_zio != NULL);

	if (zio->io_type == ZIO_TYPE_READ) {
	if (zio->io_error == 0)
	vps->vps_readable = 1;
	if (zio->io_error == 0 && spa_writeable(spa)) {
	zio_nowait(zio_write_phys(vd->vdev_probe_zio, vd,
	zio->io_offset, zio->io_size, zio->io_abd,
	ZIO_CHECKSUM_OFF, vdev_probe_done, vps,
	ZIO_PRIORITY_SYNC_WRITE, vps->vps_flags, B_TRUE));
	} else {
	abd_free(zio->io_abd);
	}
	} else if (zio->io_type == ZIO_TYPE_WRITE) {
	if (zio->io_error == 0)
	vps->vps_writeable = 1;
	abd_free(zio->io_abd);
	} else if (zio->io_type == ZIO_TYPE_NULL) {
	zio_t *pio;
	zio_link_t *zl;

	vd->vdev_cant_read \|= !vps->vps_readable;
	vd->vdev_cant_write \|= !vps->vps_writeable;
	vdev_dbgmsg(vd, "probe done, cant_read=%u cant_write=%u",
	vd->vdev_cant_read, vd->vdev_cant_write);

	if (vdev_readable(vd) &&
	(vdev_writeable(vd) \|\| !spa_writeable(spa))) {
	zio->io_error = 0;
	} else {
	ASSERT(zio->io_error != 0);
	vdev_dbgmsg(vd, "failed probe");
	(void) zfs_ereport_post(FM_EREPORT_ZFS_PROBE_FAILURE,
	spa, vd, NULL, NULL, 0);
	zio->io_error = SET_ERROR(ENXIO);

	/*
	* If this probe was initiated from zio pipeline, then
	* change the state in a spa_async_request. Probes that
	* were initiated from a vdev_open can change the state
	* as part of the open call.
	*/
	if (vps->vps_zio_done_probe) {
	vd->vdev_fault_wanted = B_TRUE;
	spa_async_request(spa, SPA_ASYNC_FAULT_VDEV);
	}
	}

	mutex_enter(&vd->vdev_probe_lock);
	ASSERT(vd->vdev_probe_zio == zio);
	vd->vdev_probe_zio = NULL;
	mutex_exit(&vd->vdev_probe_lock);

	zl = NULL;
	while ((pio = zio_walk_parents(zio, &zl)) != NULL)
	if (!vdev_accessible(vd, pio))
	pio->io_error = SET_ERROR(ENXIO);

	kmem_free(vps, sizeof (*vps));
	}
	}

	/*
	* Determine whether this device is accessible.
	*
	* Read and write to several known locations: the pad regions of each
	* vdev label but the first, which we leave alone in case it contains
	* a VTOC.
	*/
	zio_t *
	vdev_probe(vdev_t vd, zio_t zio)
	{
	spa_t *spa = vd->vdev_spa;
	vdev_probe_stats_t *vps = NULL;
	zio_t *pio;

	ASSERT(vd->vdev_ops->vdev_op_leaf);

	/*
	* Don't probe the probe.
	*/
	if (zio && (zio->io_flags & ZIO_FLAG_PROBE))
	return (NULL);

	/*
	* To prevent 'probe storms' when a device fails, we create
	* just one probe i/o at a time. All zios that want to probe
	* this vdev will become parents of the probe io.
	*/
	mutex_enter(&vd->vdev_probe_lock);

	if ((pio = vd->vdev_probe_zio) == NULL) {
	vps = kmem_zalloc(sizeof (*vps), KM_SLEEP);

	vps->vps_flags = ZIO_FLAG_CANFAIL \| ZIO_FLAG_PROBE \|
	ZIO_FLAG_DONT_AGGREGATE \| ZIO_FLAG_TRYHARD;
	vps->vps_zio_done_probe = (zio != NULL);

	if (spa_config_held(spa, SCL_ZIO, RW_WRITER)) {
	/*
	* vdev_cant_read and vdev_cant_write can only
	* transition from TRUE to FALSE when we have the
	* SCL_ZIO lock as writer; otherwise they can only
	* transition from FALSE to TRUE. This ensures that
	* any zio looking at these values can assume that
	* failures persist for the life of the I/O. That's
	* important because when a device has intermittent
	* connectivity problems, we want to ensure that
	* they're ascribed to the device (ENXIO) and not
	* the zio (EIO).
	*
	* Since we hold SCL_ZIO as writer here, clear both
	* values so the probe can reevaluate from first
	* principles.
	*/
	vps->vps_flags \|= ZIO_FLAG_CONFIG_WRITER;
	vd->vdev_cant_read = B_FALSE;
	vd->vdev_cant_write = B_FALSE;
	}

	vd->vdev_probe_zio = pio = zio_null(NULL, spa, vd,
	vdev_probe_done, vps,
	vps->vps_flags \| ZIO_FLAG_DONT_PROPAGATE);
	}

	if (zio != NULL)
	zio_add_child(zio, pio);

	mutex_exit(&vd->vdev_probe_lock);

	if (vps == NULL) {
	ASSERT(zio != NULL);
	return (NULL);
	}

	for (int l = 1; l < VDEV_LABELS; l++) {
	zio_nowait(zio_read_phys(pio, vd,
	vdev_label_offset(vd->vdev_psize, l,
	offsetof(vdev_label_t, vl_be)), VDEV_PAD_SIZE,
	abd_alloc_for_io(VDEV_PAD_SIZE, B_TRUE),
	ZIO_CHECKSUM_OFF, vdev_probe_done, vps,
	ZIO_PRIORITY_SYNC_READ, vps->vps_flags, B_TRUE));
	}

	if (zio == NULL)
	return (pio);

	zio_nowait(pio);
	return (NULL);
	}

	static void
	vdev_load_child(void *arg)
	{
	vdev_t *vd = arg;

	vd->vdev_load_error = vdev_load(vd);
	}

	static void
	vdev_open_child(void *arg)
	{
	vdev_t *vd = arg;

	vd->vdev_open_thread = curthread;
	vd->vdev_open_error = vdev_open(vd);
	vd->vdev_open_thread = NULL;
	}

	static boolean_t
	vdev_uses_zvols(vdev_t *vd)
	{
	#ifdef _KERNEL
	if (zvol_is_zvol(vd->vdev_path))
	return (B_TRUE);
	#endif

	for (int c = 0; c < vd->vdev_children; c++)
	if (vdev_uses_zvols(vd->vdev_child[c]))
	return (B_TRUE);

	return (B_FALSE);
	}

	/*
	* Returns B_TRUE if the passed child should be opened.
	*/
	static boolean_t
	vdev_default_open_children_func(vdev_t *vd)
	{
	(void) vd;
	return (B_TRUE);
	}

	/*
	* Open the requested child vdevs. If any of the leaf vdevs are using
	* a ZFS volume then do the opens in a single thread. This avoids a
	* deadlock when the current thread is holding the spa_namespace_lock.
	*/
	static void
	vdev_open_children_impl(vdev_t vd, vdev_open_children_func_t open_func)
	{
	int children = vd->vdev_children;

	taskq_t *tq = taskq_create("vdev_open", children, minclsyspri,
	children, children, TASKQ_PREPOPULATE);
	vd->vdev_nonrot = B_TRUE;

	for (int c = 0; c < children; c++) {
	vdev_t *cvd = vd->vdev_child[c];

	if (open_func(cvd) == B_FALSE)
	continue;

	if (tq == NULL \|\| vdev_uses_zvols(vd)) {
	cvd->vdev_open_error = vdev_open(cvd);
	} else {
	VERIFY(taskq_dispatch(tq, vdev_open_child,
	cvd, TQ_SLEEP) != TASKQID_INVALID);
	}

	vd->vdev_nonrot &= cvd->vdev_nonrot;
	}

	if (tq != NULL) {
	taskq_wait(tq);
	taskq_destroy(tq);
	}
	}

	/*
	* Open all child vdevs.
	*/
	void
	vdev_open_children(vdev_t *vd)
	{
	vdev_open_children_impl(vd, vdev_default_open_children_func);
	}

	/*
	* Conditionally open a subset of child vdevs.
	*/
	void
	vdev_open_children_subset(vdev_t vd, vdev_open_children_func_t open_func)
	{
	vdev_open_children_impl(vd, open_func);
	}

	/*
	* Compute the raidz-deflation ratio. Note, we hard-code 128k (1 << 17)
	* because it is the "typical" blocksize. Even though SPA_MAXBLOCKSIZE
	* changed, this algorithm can not change, otherwise it would inconsistently
	* account for existing bp's. We also hard-code txg 0 for the same reason
	* since expanded RAIDZ vdevs can use a different asize for different birth
	* txg's.
	*/
	static void
	vdev_set_deflate_ratio(vdev_t *vd)
	{
	if (vd == vd->vdev_top && !vd->vdev_ishole && vd->vdev_ashift != 0) {
	vd->vdev_deflate_ratio = (1 << 17) /
	(vdev_psize_to_asize_txg(vd, 1 << 17, 0) >>
	SPA_MINBLOCKSHIFT);
	}
	}

	/*
	* Choose the best of two ashifts, preferring one between logical ashift
	* (absolute minimum) and administrator defined maximum, otherwise take
	* the biggest of the two.
	*/
	uint64_t
	vdev_best_ashift(uint64_t logical, uint64_t a, uint64_t b)
	{
	if (a > logical && a <= zfs_vdev_max_auto_ashift) {
	if (b <= logical \|\| b > zfs_vdev_max_auto_ashift)
	return (a);
	else
	return (MAX(a, b));
	} else if (b <= logical \|\| b > zfs_vdev_max_auto_ashift)
	return (MAX(a, b));
	return (b);
	}

	/*
	* Maximize performance by inflating the configured ashift for top level
	* vdevs to be as close to the physical ashift as possible while maintaining
	* administrator defined limits and ensuring it doesn't go below the
	* logical ashift.
	*/
	static void
	vdev_ashift_optimize(vdev_t *vd)
	{
	ASSERT(vd == vd->vdev_top);

	if (vd->vdev_ashift < vd->vdev_physical_ashift &&
	vd->vdev_physical_ashift <= zfs_vdev_max_auto_ashift) {
	vd->vdev_ashift = MIN(
	MAX(zfs_vdev_max_auto_ashift, vd->vdev_ashift),
	MAX(zfs_vdev_min_auto_ashift,
	vd->vdev_physical_ashift));
	} else {
	/*
	* If the logical and physical ashifts are the same, then
	* we ensure that the top-level vdev's ashift is not smaller
	* than our minimum ashift value. For the unusual case
	* where logical ashift > physical ashift, we can't cap
	* the calculated ashift based on max ashift as that
	* would cause failures.
	* We still check if we need to increase it to match
	* the min ashift.
	*/
	vd->vdev_ashift = MAX(zfs_vdev_min_auto_ashift,
	vd->vdev_ashift);
	}
	}

	/*
	* Prepare a virtual device for access.
	*/
	int
	vdev_open(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;
	int error;
	uint64_t osize = 0;
	uint64_t max_osize = 0;
	uint64_t asize, max_asize, psize;
	uint64_t logical_ashift = 0;
	uint64_t physical_ashift = 0;

	ASSERT(vd->vdev_open_thread == curthread \|\|
	spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);
	ASSERT(vd->vdev_state == VDEV_STATE_CLOSED \|\|
	vd->vdev_state == VDEV_STATE_CANT_OPEN \|\|
	vd->vdev_state == VDEV_STATE_OFFLINE);

	vd->vdev_stat.vs_aux = VDEV_AUX_NONE;
	vd->vdev_cant_read = B_FALSE;
	vd->vdev_cant_write = B_FALSE;
	vd->vdev_min_asize = vdev_get_min_asize(vd);

	/*
	* If this vdev is not removed, check its fault status. If it's
	* faulted, bail out of the open.
	*/
	if (!vd->vdev_removed && vd->vdev_faulted) {
	ASSERT(vd->vdev_children == 0);
	ASSERT(vd->vdev_label_aux == VDEV_AUX_ERR_EXCEEDED \|\|
	vd->vdev_label_aux == VDEV_AUX_EXTERNAL);
	vdev_set_state(vd, B_TRUE, VDEV_STATE_FAULTED,
	vd->vdev_label_aux);
	return (SET_ERROR(ENXIO));
	} else if (vd->vdev_offline) {
	ASSERT(vd->vdev_children == 0);
	vdev_set_state(vd, B_TRUE, VDEV_STATE_OFFLINE, VDEV_AUX_NONE);
	return (SET_ERROR(ENXIO));
	}

	error = vd->vdev_ops->vdev_op_open(vd, &osize, &max_osize,
	&logical_ashift, &physical_ashift);

	/* Keep the device in removed state if unplugged */
	if (error == ENOENT && vd->vdev_removed) {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_REMOVED,
	VDEV_AUX_NONE);
	return (error);
	}

	/*
	* Physical volume size should never be larger than its max size, unless
	* the disk has shrunk while we were reading it or the device is buggy
	* or damaged: either way it's not safe for use, bail out of the open.
	*/
	if (osize > max_osize) {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_OPEN_FAILED);
	return (SET_ERROR(ENXIO));
	}

	/*
	* Reset the vdev_reopening flag so that we actually close
	* the vdev on error.
	*/
	vd->vdev_reopening = B_FALSE;
	if (zio_injection_enabled && error == 0)
	error = zio_handle_device_injection(vd, NULL, SET_ERROR(ENXIO));

	if (error) {
	if (vd->vdev_removed &&
	vd->vdev_stat.vs_aux != VDEV_AUX_OPEN_FAILED)
	vd->vdev_removed = B_FALSE;

	if (vd->vdev_stat.vs_aux == VDEV_AUX_CHILDREN_OFFLINE) {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_OFFLINE,
	vd->vdev_stat.vs_aux);
	} else {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
	vd->vdev_stat.vs_aux);
	}
	return (error);
	}

	vd->vdev_removed = B_FALSE;

	/*
	* Recheck the faulted flag now that we have confirmed that
	* the vdev is accessible. If we're faulted, bail.
	*/
	if (vd->vdev_faulted) {
	ASSERT(vd->vdev_children == 0);
	ASSERT(vd->vdev_label_aux == VDEV_AUX_ERR_EXCEEDED \|\|
	vd->vdev_label_aux == VDEV_AUX_EXTERNAL);
	vdev_set_state(vd, B_TRUE, VDEV_STATE_FAULTED,
	vd->vdev_label_aux);
	return (SET_ERROR(ENXIO));
	}

	if (vd->vdev_degraded) {
	ASSERT(vd->vdev_children == 0);
	vdev_set_state(vd, B_TRUE, VDEV_STATE_DEGRADED,
	VDEV_AUX_ERR_EXCEEDED);
	} else {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_HEALTHY, 0);
	}

	/*
	* For hole or missing vdevs we just return success.
	*/
	if (vd->vdev_ishole \|\| vd->vdev_ops == &vdev_missing_ops)
	return (0);

	for (int c = 0; c < vd->vdev_children; c++) {
	if (vd->vdev_child[c]->vdev_state != VDEV_STATE_HEALTHY) {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_DEGRADED,
	VDEV_AUX_NONE);
	break;
	}
	}

	- osize = P2ALIGN(osize, (uint64_t)sizeof (vdev_label_t));
	- max_osize = P2ALIGN(max_osize, (uint64_t)sizeof (vdev_label_t));
	+ osize = P2ALIGN_TYPED(osize, sizeof (vdev_label_t), uint64_t);
	+ max_osize = P2ALIGN_TYPED(max_osize, sizeof (vdev_label_t), uint64_t);

	if (vd->vdev_children == 0) {
	if (osize < SPA_MINDEVSIZE) {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_TOO_SMALL);
	return (SET_ERROR(EOVERFLOW));
	}
	psize = osize;
	asize = osize - (VDEV_LABEL_START_SIZE + VDEV_LABEL_END_SIZE);
	max_asize = max_osize - (VDEV_LABEL_START_SIZE +
	VDEV_LABEL_END_SIZE);
	} else {
	if (vd->vdev_parent != NULL && osize < SPA_MINDEVSIZE -
	(VDEV_LABEL_START_SIZE + VDEV_LABEL_END_SIZE)) {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_TOO_SMALL);
	return (SET_ERROR(EOVERFLOW));
	}
	psize = 0;
	asize = osize;
	max_asize = max_osize;
	}

	/*
	* If the vdev was expanded, record this so that we can re-create the
	* uberblock rings in labels {2,3}, during the next sync.
	*/
	if ((psize > vd->vdev_psize) && (vd->vdev_psize != 0))
	vd->vdev_copy_uberblocks = B_TRUE;

	vd->vdev_psize = psize;

	/*
	* Make sure the allocatable size hasn't shrunk too much.
	*/
	if (asize < vd->vdev_min_asize) {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_BAD_LABEL);
	return (SET_ERROR(EINVAL));
	}

	/*
	* We can always set the logical/physical ashift members since
	* their values are only used to calculate the vdev_ashift when
	* the device is first added to the config. These values should
	* not be used for anything else since they may change whenever
	* the device is reopened and we don't store them in the label.
	*/
	vd->vdev_physical_ashift =
	MAX(physical_ashift, vd->vdev_physical_ashift);
	vd->vdev_logical_ashift = MAX(logical_ashift,
	vd->vdev_logical_ashift);

	if (vd->vdev_asize == 0) {
	/*
	* This is the first-ever open, so use the computed values.
	* For compatibility, a different ashift can be requested.
	*/
	vd->vdev_asize = asize;
	vd->vdev_max_asize = max_asize;

	/*
	* If the vdev_ashift was not overridden at creation time,
	* then set it the logical ashift and optimize the ashift.
	*/
	if (vd->vdev_ashift == 0) {
	vd->vdev_ashift = vd->vdev_logical_ashift;

	if (vd->vdev_logical_ashift > ASHIFT_MAX) {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_ASHIFT_TOO_BIG);
	return (SET_ERROR(EDOM));
	}

	if (vd->vdev_top == vd && vd->vdev_attaching == B_FALSE)
	vdev_ashift_optimize(vd);
	vd->vdev_attaching = B_FALSE;
	}
	if (vd->vdev_ashift != 0 && (vd->vdev_ashift < ASHIFT_MIN \|\|
	vd->vdev_ashift > ASHIFT_MAX)) {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_BAD_ASHIFT);
	return (SET_ERROR(EDOM));
	}
	} else {
	/*
	* Make sure the alignment required hasn't increased.
	*/
	if (vd->vdev_ashift > vd->vdev_top->vdev_ashift &&
	vd->vdev_ops->vdev_op_leaf) {
	(void) zfs_ereport_post(
	FM_EREPORT_ZFS_DEVICE_BAD_ASHIFT,
	spa, vd, NULL, NULL, 0);
	vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_BAD_LABEL);
	return (SET_ERROR(EDOM));
	}
	vd->vdev_max_asize = max_asize;
	}

	/*
	* If all children are healthy we update asize if either:
	* The asize has increased, due to a device expansion caused by dynamic
	* LUN growth or vdev replacement, and automatic expansion is enabled;
	* making the additional space available.
	*
	* The asize has decreased, due to a device shrink usually caused by a
	* vdev replace with a smaller device. This ensures that calculations
	* based of max_asize and asize e.g. esize are always valid. It's safe
	* to do this as we've already validated that asize is greater than
	* vdev_min_asize.
	*/
	if (vd->vdev_state == VDEV_STATE_HEALTHY &&
	((asize > vd->vdev_asize &&
	(vd->vdev_expanding \|\| spa->spa_autoexpand)) \|\|
	(asize < vd->vdev_asize)))
	vd->vdev_asize = asize;

	vdev_set_min_asize(vd);

	/*
	* Ensure we can issue some IO before declaring the
	* vdev open for business.
	*/
	if (vd->vdev_ops->vdev_op_leaf &&
	(error = zio_wait(vdev_probe(vd, NULL))) != 0) {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_FAULTED,
	VDEV_AUX_ERR_EXCEEDED);
	return (error);
	}

	/*
	* Track the minimum allocation size.
	*/
	if (vd->vdev_top == vd && vd->vdev_ashift != 0 &&
	vd->vdev_islog == 0 && vd->vdev_aux == NULL) {
	uint64_t min_alloc = vdev_get_min_alloc(vd);
	vdev_spa_set_alloc(spa, min_alloc);
	}

	/*
	* If this is a leaf vdev, assess whether a resilver is needed.
	* But don't do this if we are doing a reopen for a scrub, since
	* this would just restart the scrub we are already doing.
	*/
	if (vd->vdev_ops->vdev_op_leaf && !spa->spa_scrub_reopen)
	dsl_scan_assess_vdev(spa->spa_dsl_pool, vd);

	return (0);
	}

	static void
	vdev_validate_child(void *arg)
	{
	vdev_t *vd = arg;

	vd->vdev_validate_thread = curthread;
	vd->vdev_validate_error = vdev_validate(vd);
	vd->vdev_validate_thread = NULL;
	}

	/*
	* Called once the vdevs are all opened, this routine validates the label
	* contents. This needs to be done before vdev_load() so that we don't
	* inadvertently do repair I/Os to the wrong device.
	*
	* This function will only return failure if one of the vdevs indicates that it
	* has since been destroyed or exported. This is only possible if
	* /etc/zfs/zpool.cache was readonly at the time. Otherwise, the vdev state
	* will be updated but the function will return 0.
	*/
	int
	vdev_validate(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;
	taskq_t *tq = NULL;
	nvlist_t *label;
	uint64_t guid = 0, aux_guid = 0, top_guid;
	uint64_t state;
	nvlist_t *nvl;
	uint64_t txg;
	int children = vd->vdev_children;

	if (vdev_validate_skip)
	return (0);

	if (children > 0) {
	tq = taskq_create("vdev_validate", children, minclsyspri,
	children, children, TASKQ_PREPOPULATE);
	}

	for (uint64_t c = 0; c < children; c++) {
	vdev_t *cvd = vd->vdev_child[c];

	if (tq == NULL \|\| vdev_uses_zvols(cvd)) {
	vdev_validate_child(cvd);
	} else {
	VERIFY(taskq_dispatch(tq, vdev_validate_child, cvd,
	TQ_SLEEP) != TASKQID_INVALID);
	}
	}
	if (tq != NULL) {
	taskq_wait(tq);
	taskq_destroy(tq);
	}
	for (int c = 0; c < children; c++) {
	int error = vd->vdev_child[c]->vdev_validate_error;

	if (error != 0)
	return (SET_ERROR(EBADF));
	}


	/*
	* If the device has already failed, or was marked offline, don't do
	* any further validation. Otherwise, label I/O will fail and we will
	* overwrite the previous state.
	*/
	if (!vd->vdev_ops->vdev_op_leaf \|\| !vdev_readable(vd))
	return (0);

	/*
	* If we are performing an extreme rewind, we allow for a label that
	* was modified at a point after the current txg.
	* If config lock is not held do not check for the txg. spa_sync could
	* be updating the vdev's label before updating spa_last_synced_txg.
	*/
	if (spa->spa_extreme_rewind \|\| spa_last_synced_txg(spa) == 0 \|\|
	spa_config_held(spa, SCL_CONFIG, RW_WRITER) != SCL_CONFIG)
	txg = UINT64_MAX;
	else
	txg = spa_last_synced_txg(spa);

	if ((label = vdev_label_read_config(vd, txg)) == NULL) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_BAD_LABEL);
	vdev_dbgmsg(vd, "vdev_validate: failed reading config for "
	"txg %llu", (u_longlong_t)txg);
	return (0);
	}

	/*
	* Determine if this vdev has been split off into another
	* pool. If so, then refuse to open it.
	*/
	if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_SPLIT_GUID,
	&aux_guid) == 0 && aux_guid == spa_guid(spa)) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_SPLIT_POOL);
	nvlist_free(label);
	vdev_dbgmsg(vd, "vdev_validate: vdev split into other pool");
	return (0);
	}

	if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_GUID, &guid) != 0) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	nvlist_free(label);
	vdev_dbgmsg(vd, "vdev_validate: '%s' missing from label",
	ZPOOL_CONFIG_POOL_GUID);
	return (0);
	}

	/*
	* If config is not trusted then ignore the spa guid check. This is
	* necessary because if the machine crashed during a re-guid the new
	* guid might have been written to all of the vdev labels, but not the
	* cached config. The check will be performed again once we have the
	* trusted config from the MOS.
	*/
	if (spa->spa_trust_config && guid != spa_guid(spa)) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	nvlist_free(label);
	vdev_dbgmsg(vd, "vdev_validate: vdev label pool_guid doesn't "
	"match config (%llu != %llu)", (u_longlong_t)guid,
	(u_longlong_t)spa_guid(spa));
	return (0);
	}

	if (nvlist_lookup_nvlist(label, ZPOOL_CONFIG_VDEV_TREE, &nvl)
	!= 0 \|\| nvlist_lookup_uint64(nvl, ZPOOL_CONFIG_ORIG_GUID,
	&aux_guid) != 0)
	aux_guid = 0;

	if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID, &guid) != 0) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	nvlist_free(label);
	vdev_dbgmsg(vd, "vdev_validate: '%s' missing from label",
	ZPOOL_CONFIG_GUID);
	return (0);
	}

	if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_TOP_GUID, &top_guid)
	!= 0) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	nvlist_free(label);
	vdev_dbgmsg(vd, "vdev_validate: '%s' missing from label",
	ZPOOL_CONFIG_TOP_GUID);
	return (0);
	}

	/*
	* If this vdev just became a top-level vdev because its sibling was
	* detached, it will have adopted the parent's vdev guid -- but the
	* label may or may not be on disk yet. Fortunately, either version
	* of the label will have the same top guid, so if we're a top-level
	* vdev, we can safely compare to that instead.
	* However, if the config comes from a cachefile that failed to update
	* after the detach, a top-level vdev will appear as a non top-level
	* vdev in the config. Also relax the constraints if we perform an
	* extreme rewind.
	*
	* If we split this vdev off instead, then we also check the
	* original pool's guid. We don't want to consider the vdev
	* corrupt if it is partway through a split operation.
	*/
	if (vd->vdev_guid != guid && vd->vdev_guid != aux_guid) {
	boolean_t mismatch = B_FALSE;
	if (spa->spa_trust_config && !spa->spa_extreme_rewind) {
	if (vd != vd->vdev_top \|\| vd->vdev_guid != top_guid)
	mismatch = B_TRUE;
	} else {
	if (vd->vdev_guid != top_guid &&
	vd->vdev_top->vdev_guid != guid)
	mismatch = B_TRUE;
	}

	if (mismatch) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	nvlist_free(label);
	vdev_dbgmsg(vd, "vdev_validate: config guid "
	"doesn't match label guid");
	vdev_dbgmsg(vd, "CONFIG: guid %llu, top_guid %llu",
	(u_longlong_t)vd->vdev_guid,
	(u_longlong_t)vd->vdev_top->vdev_guid);
	vdev_dbgmsg(vd, "LABEL: guid %llu, top_guid %llu, "
	"aux_guid %llu", (u_longlong_t)guid,
	(u_longlong_t)top_guid, (u_longlong_t)aux_guid);
	return (0);
	}
	}

	if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_STATE,
	&state) != 0) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	nvlist_free(label);
	vdev_dbgmsg(vd, "vdev_validate: '%s' missing from label",
	ZPOOL_CONFIG_POOL_STATE);
	return (0);
	}

	nvlist_free(label);

	/*
	* If this is a verbatim import, no need to check the
	* state of the pool.
	*/
	if (!(spa->spa_import_flags & ZFS_IMPORT_VERBATIM) &&
	spa_load_state(spa) == SPA_LOAD_OPEN &&
	state != POOL_STATE_ACTIVE) {
	vdev_dbgmsg(vd, "vdev_validate: invalid pool state (%llu) "
	"for spa %s", (u_longlong_t)state, spa->spa_name);
	return (SET_ERROR(EBADF));
	}

	/*
	* If we were able to open and validate a vdev that was
	* previously marked permanently unavailable, clear that state
	* now.
	*/
	if (vd->vdev_not_present)
	vd->vdev_not_present = 0;

	return (0);
	}

	static void
	vdev_update_path(const char prefix, char svd, char **dvd, uint64_t guid)
	{
	if (svd != NULL && *dvd != NULL) {
	if (strcmp(svd, *dvd) != 0) {
	zfs_dbgmsg("vdev_copy_path: vdev %llu: %s changed "
	"from '%s' to '%s'", (u_longlong_t)guid, prefix,
	*dvd, svd);
	spa_strfree(*dvd);
	*dvd = spa_strdup(svd);
	}
	} else if (svd != NULL) {
	*dvd = spa_strdup(svd);
	zfs_dbgmsg("vdev_copy_path: vdev %llu: path set to '%s'",
	(u_longlong_t)guid, *dvd);
	}
	}

	static void
	vdev_copy_path_impl(vdev_t svd, vdev_t dvd)
	{
	char old, new;

	vdev_update_path("vdev_path", svd->vdev_path, &dvd->vdev_path,
	dvd->vdev_guid);

	vdev_update_path("vdev_devid", svd->vdev_devid, &dvd->vdev_devid,
	dvd->vdev_guid);

	vdev_update_path("vdev_physpath", svd->vdev_physpath,
	&dvd->vdev_physpath, dvd->vdev_guid);

	/*
	* Our enclosure sysfs path may have changed between imports
	*/
	old = dvd->vdev_enc_sysfs_path;
	new = svd->vdev_enc_sysfs_path;
	if ((old != NULL && new == NULL) \|\|
	(old == NULL && new != NULL) \|\|
	((old != NULL && new != NULL) && strcmp(new, old) != 0)) {
	zfs_dbgmsg("vdev_copy_path: vdev %llu: vdev_enc_sysfs_path "
	"changed from '%s' to '%s'", (u_longlong_t)dvd->vdev_guid,
	old, new);

	if (dvd->vdev_enc_sysfs_path)
	spa_strfree(dvd->vdev_enc_sysfs_path);

	if (svd->vdev_enc_sysfs_path) {
	dvd->vdev_enc_sysfs_path = spa_strdup(
	svd->vdev_enc_sysfs_path);
	} else {
	dvd->vdev_enc_sysfs_path = NULL;
	}
	}
	}

	/*
	* Recursively copy vdev paths from one vdev to another. Source and destination
	* vdev trees must have same geometry otherwise return error. Intended to copy
	* paths from userland config into MOS config.
	*/
	int
	vdev_copy_path_strict(vdev_t svd, vdev_t dvd)
	{
	if ((svd->vdev_ops == &vdev_missing_ops) \|\|
	(svd->vdev_ishole && dvd->vdev_ishole) \|\|
	(dvd->vdev_ops == &vdev_indirect_ops))
	return (0);

	if (svd->vdev_ops != dvd->vdev_ops) {
	vdev_dbgmsg(svd, "vdev_copy_path: vdev type mismatch: %s != %s",
	svd->vdev_ops->vdev_op_type, dvd->vdev_ops->vdev_op_type);
	return (SET_ERROR(EINVAL));
	}

	if (svd->vdev_guid != dvd->vdev_guid) {
	vdev_dbgmsg(svd, "vdev_copy_path: guids mismatch (%llu != "
	"%llu)", (u_longlong_t)svd->vdev_guid,
	(u_longlong_t)dvd->vdev_guid);
	return (SET_ERROR(EINVAL));
	}

	if (svd->vdev_children != dvd->vdev_children) {
	vdev_dbgmsg(svd, "vdev_copy_path: children count mismatch: "
	"%llu != %llu", (u_longlong_t)svd->vdev_children,
	(u_longlong_t)dvd->vdev_children);
	return (SET_ERROR(EINVAL));
	}

	for (uint64_t i = 0; i < svd->vdev_children; i++) {
	int error = vdev_copy_path_strict(svd->vdev_child[i],
	dvd->vdev_child[i]);
	if (error != 0)
	return (error);
	}

	if (svd->vdev_ops->vdev_op_leaf)
	vdev_copy_path_impl(svd, dvd);

	return (0);
	}

	static void
	vdev_copy_path_search(vdev_t stvd, vdev_t dvd)
	{
	ASSERT(stvd->vdev_top == stvd);
	ASSERT3U(stvd->vdev_id, ==, dvd->vdev_top->vdev_id);

	for (uint64_t i = 0; i < dvd->vdev_children; i++) {
	vdev_copy_path_search(stvd, dvd->vdev_child[i]);
	}

	if (!dvd->vdev_ops->vdev_op_leaf \|\| !vdev_is_concrete(dvd))
	return;

	/*
	* The idea here is that while a vdev can shift positions within
	* a top vdev (when replacing, attaching mirror, etc.) it cannot
	* step outside of it.
	*/
	vdev_t *vd = vdev_lookup_by_guid(stvd, dvd->vdev_guid);

	if (vd == NULL \|\| vd->vdev_ops != dvd->vdev_ops)
	return;

	ASSERT(vd->vdev_ops->vdev_op_leaf);

	vdev_copy_path_impl(vd, dvd);
	}

	/*
	* Recursively copy vdev paths from one root vdev to another. Source and
	* destination vdev trees may differ in geometry. For each destination leaf
	* vdev, search a vdev with the same guid and top vdev id in the source.
	* Intended to copy paths from userland config into MOS config.
	*/
	void
	vdev_copy_path_relaxed(vdev_t srvd, vdev_t drvd)
	{
	uint64_t children = MIN(srvd->vdev_children, drvd->vdev_children);
	ASSERT(srvd->vdev_ops == &vdev_root_ops);
	ASSERT(drvd->vdev_ops == &vdev_root_ops);

	for (uint64_t i = 0; i < children; i++) {
	vdev_copy_path_search(srvd->vdev_child[i],
	drvd->vdev_child[i]);
	}
	}

	/*
	* Close a virtual device.
	*/
	void
	vdev_close(vdev_t *vd)
	{
	vdev_t *pvd = vd->vdev_parent;
	spa_t *spa __maybe_unused = vd->vdev_spa;

	ASSERT(vd != NULL);
	ASSERT(vd->vdev_open_thread == curthread \|\|
	spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);

	/*
	* If our parent is reopening, then we are as well, unless we are
	* going offline.
	*/
	if (pvd != NULL && pvd->vdev_reopening)
	vd->vdev_reopening = (pvd->vdev_reopening && !vd->vdev_offline);

	vd->vdev_ops->vdev_op_close(vd);

	/*
	* We record the previous state before we close it, so that if we are
	* doing a reopen(), we don't generate FMA ereports if we notice that
	* it's still faulted.
	*/
	vd->vdev_prevstate = vd->vdev_state;

	if (vd->vdev_offline)
	vd->vdev_state = VDEV_STATE_OFFLINE;
	else
	vd->vdev_state = VDEV_STATE_CLOSED;
	vd->vdev_stat.vs_aux = VDEV_AUX_NONE;
	}

	void
	vdev_hold(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;

	ASSERT(spa_is_root(spa));
	if (spa->spa_state == POOL_STATE_UNINITIALIZED)
	return;

	for (int c = 0; c < vd->vdev_children; c++)
	vdev_hold(vd->vdev_child[c]);

	if (vd->vdev_ops->vdev_op_leaf && vd->vdev_ops->vdev_op_hold != NULL)
	vd->vdev_ops->vdev_op_hold(vd);
	}

	void
	vdev_rele(vdev_t *vd)
	{
	ASSERT(spa_is_root(vd->vdev_spa));
	for (int c = 0; c < vd->vdev_children; c++)
	vdev_rele(vd->vdev_child[c]);

	if (vd->vdev_ops->vdev_op_leaf && vd->vdev_ops->vdev_op_rele != NULL)
	vd->vdev_ops->vdev_op_rele(vd);
	}

	/*
	* Reopen all interior vdevs and any unopened leaves. We don't actually
	* reopen leaf vdevs which had previously been opened as they might deadlock
	* on the spa_config_lock. Instead we only obtain the leaf's physical size.
	* If the leaf has never been opened then open it, as usual.
	*/
	void
	vdev_reopen(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;

	ASSERT(spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);

	/* set the reopening flag unless we're taking the vdev offline */
	vd->vdev_reopening = !vd->vdev_offline;
	vdev_close(vd);
	(void) vdev_open(vd);

	/*
	* Call vdev_validate() here to make sure we have the same device.
	* Otherwise, a device with an invalid label could be successfully
	* opened in response to vdev_reopen().
	*/
	if (vd->vdev_aux) {
	(void) vdev_validate_aux(vd);
	if (vdev_readable(vd) && vdev_writeable(vd) &&
	vd->vdev_aux == &spa->spa_l2cache) {
	/*
	* In case the vdev is present we should evict all ARC
	* buffers and pointers to log blocks and reclaim their
	* space before restoring its contents to L2ARC.
	*/
	if (l2arc_vdev_present(vd)) {
	l2arc_rebuild_vdev(vd, B_TRUE);
	} else {
	l2arc_add_vdev(spa, vd);
	}
	spa_async_request(spa, SPA_ASYNC_L2CACHE_REBUILD);
	spa_async_request(spa, SPA_ASYNC_L2CACHE_TRIM);
	}
	} else {
	(void) vdev_validate(vd);
	}

	/*
	* Recheck if resilver is still needed and cancel any
	* scheduled resilver if resilver is unneeded.
	*/
	if (!vdev_resilver_needed(spa->spa_root_vdev, NULL, NULL) &&
	spa->spa_async_tasks & SPA_ASYNC_RESILVER) {
	mutex_enter(&spa->spa_async_lock);
	spa->spa_async_tasks &= ~SPA_ASYNC_RESILVER;
	mutex_exit(&spa->spa_async_lock);
	}

	/*
	* Reassess parent vdev's health.
	*/
	vdev_propagate_state(vd);
	}

	int
	vdev_create(vdev_t *vd, uint64_t txg, boolean_t isreplacing)
	{
	int error;

	/*
	* Normally, partial opens (e.g. of a mirror) are allowed.
	* For a create, however, we want to fail the request if
	* there are any components we can't open.
	*/
	error = vdev_open(vd);

	if (error \|\| vd->vdev_state != VDEV_STATE_HEALTHY) {
	vdev_close(vd);
	return (error ? error : SET_ERROR(ENXIO));
	}

	/*
	* Recursively load DTLs and initialize all labels.
	*/
	if ((error = vdev_dtl_load(vd)) != 0 \|\|
	(error = vdev_label_init(vd, txg, isreplacing ?
	VDEV_LABEL_REPLACE : VDEV_LABEL_CREATE)) != 0) {
	vdev_close(vd);
	return (error);
	}

	return (0);
	}

	void
	vdev_metaslab_set_size(vdev_t *vd)
	{
	uint64_t asize = vd->vdev_asize;
	uint64_t ms_count = asize >> zfs_vdev_default_ms_shift;
	uint64_t ms_shift;

	/*
	* There are two dimensions to the metaslab sizing calculation:
	* the size of the metaslab and the count of metaslabs per vdev.
	*
	* The default values used below are a good balance between memory
	* usage (larger metaslab size means more memory needed for loaded
	* metaslabs; more metaslabs means more memory needed for the
	* metaslab_t structs), metaslab load time (larger metaslabs take
	* longer to load), and metaslab sync time (more metaslabs means
	* more time spent syncing all of them).
	*
	* In general, we aim for zfs_vdev_default_ms_count (200) metaslabs.
	* The range of the dimensions are as follows:
	*
	* 2^29 <= ms_size <= 2^34
	* 16 <= ms_count <= 131,072
	*
	* On the lower end of vdev sizes, we aim for metaslabs sizes of
	* at least 512MB (2^29) to minimize fragmentation effects when
	* testing with smaller devices. However, the count constraint
	* of at least 16 metaslabs will override this minimum size goal.
	*
	* On the upper end of vdev sizes, we aim for a maximum metaslab
	* size of 16GB. However, we will cap the total count to 2^17
	* metaslabs to keep our memory footprint in check and let the
	* metaslab size grow from there if that limit is hit.
	*
	* The net effect of applying above constrains is summarized below.
	*
	* vdev size metaslab count
	* --------------\|-----------------
	* < 8GB ~16
	* 8GB - 100GB one per 512MB
	* 100GB - 3TB ~200
	* 3TB - 2PB one per 16GB
	* > 2PB ~131,072
	* --------------------------------
	*
	* Finally, note that all of the above calculate the initial
	* number of metaslabs. Expanding a top-level vdev will result
	* in additional metaslabs being allocated making it possible
	* to exceed the zfs_vdev_ms_count_limit.
	*/

	if (ms_count < zfs_vdev_min_ms_count)
	ms_shift = highbit64(asize / zfs_vdev_min_ms_count);
	else if (ms_count > zfs_vdev_default_ms_count)
	ms_shift = highbit64(asize / zfs_vdev_default_ms_count);
	else
	ms_shift = zfs_vdev_default_ms_shift;

	if (ms_shift < SPA_MAXBLOCKSHIFT) {
	ms_shift = SPA_MAXBLOCKSHIFT;
	} else if (ms_shift > zfs_vdev_max_ms_shift) {
	ms_shift = zfs_vdev_max_ms_shift;
	/* cap the total count to constrain memory footprint */
	if ((asize >> ms_shift) > zfs_vdev_ms_count_limit)
	ms_shift = highbit64(asize / zfs_vdev_ms_count_limit);
	}

	vd->vdev_ms_shift = ms_shift;
	ASSERT3U(vd->vdev_ms_shift, >=, SPA_MAXBLOCKSHIFT);
	}

	void
	vdev_dirty(vdev_t vd, int flags, void arg, uint64_t txg)
	{
	ASSERT(vd == vd->vdev_top);
	/* indirect vdevs don't have metaslabs or dtls */
	ASSERT(vdev_is_concrete(vd) \|\| flags == 0);
	ASSERT(ISP2(flags));
	ASSERT(spa_writeable(vd->vdev_spa));

	if (flags & VDD_METASLAB)
	(void) txg_list_add(&vd->vdev_ms_list, arg, txg);

	if (flags & VDD_DTL)
	(void) txg_list_add(&vd->vdev_dtl_list, arg, txg);

	(void) txg_list_add(&vd->vdev_spa->spa_vdev_txg_list, vd, txg);
	}

	void
	vdev_dirty_leaves(vdev_t *vd, int flags, uint64_t txg)
	{
	for (int c = 0; c < vd->vdev_children; c++)
	vdev_dirty_leaves(vd->vdev_child[c], flags, txg);

	if (vd->vdev_ops->vdev_op_leaf)
	vdev_dirty(vd->vdev_top, flags, vd, txg);
	}

	/*
	* DTLs.
	*
	* A vdev's DTL (dirty time log) is the set of transaction groups for which
	* the vdev has less than perfect replication. There are four kinds of DTL:
	*
	* DTL_MISSING: txgs for which the vdev has no valid copies of the data
	*
	* DTL_PARTIAL: txgs for which data is available, but not fully replicated
	*
	* DTL_SCRUB: the txgs that could not be repaired by the last scrub; upon
	* scrub completion, DTL_SCRUB replaces DTL_MISSING in the range of
	* txgs that was scrubbed.
	*
	* DTL_OUTAGE: txgs which cannot currently be read, whether due to
	* persistent errors or just some device being offline.
	* Unlike the other three, the DTL_OUTAGE map is not generally
	* maintained; it's only computed when needed, typically to
	* determine whether a device can be detached.
	*
	* For leaf vdevs, DTL_MISSING and DTL_PARTIAL are identical: the device
	* either has the data or it doesn't.
	*
	* For interior vdevs such as mirror and RAID-Z the picture is more complex.
	* A vdev's DTL_PARTIAL is the union of its children's DTL_PARTIALs, because
	* if any child is less than fully replicated, then so is its parent.
	* A vdev's DTL_MISSING is a modified union of its children's DTL_MISSINGs,
	* comprising only those txgs which appear in 'maxfaults' or more children;
	* those are the txgs we don't have enough replication to read. For example,
	* double-parity RAID-Z can tolerate up to two missing devices (maxfaults == 2);
	* thus, its DTL_MISSING consists of the set of txgs that appear in more than
	* two child DTL_MISSING maps.
	*
	* It should be clear from the above that to compute the DTLs and outage maps
	* for all vdevs, it suffices to know just the leaf vdevs' DTL_MISSING maps.
	* Therefore, that is all we keep on disk. When loading the pool, or after
	* a configuration change, we generate all other DTLs from first principles.
	*/
	void
	vdev_dtl_dirty(vdev_t *vd, vdev_dtl_type_t t, uint64_t txg, uint64_t size)
	{
	range_tree_t *rt = vd->vdev_dtl[t];

	ASSERT(t < DTL_TYPES);
	ASSERT(vd != vd->vdev_spa->spa_root_vdev);
	ASSERT(spa_writeable(vd->vdev_spa));

	mutex_enter(&vd->vdev_dtl_lock);
	if (!range_tree_contains(rt, txg, size))
	range_tree_add(rt, txg, size);
	mutex_exit(&vd->vdev_dtl_lock);
	}

	boolean_t
	vdev_dtl_contains(vdev_t *vd, vdev_dtl_type_t t, uint64_t txg, uint64_t size)
	{
	range_tree_t *rt = vd->vdev_dtl[t];
	boolean_t dirty = B_FALSE;

	ASSERT(t < DTL_TYPES);
	ASSERT(vd != vd->vdev_spa->spa_root_vdev);

	/*
	* While we are loading the pool, the DTLs have not been loaded yet.
	* This isn't a problem but it can result in devices being tried
	* which are known to not have the data. In which case, the import
	* is relying on the checksum to ensure that we get the right data.
	* Note that while importing we are only reading the MOS, which is
	* always checksummed.
	*/
	mutex_enter(&vd->vdev_dtl_lock);
	if (!range_tree_is_empty(rt))
	dirty = range_tree_contains(rt, txg, size);
	mutex_exit(&vd->vdev_dtl_lock);

	return (dirty);
	}

	boolean_t
	vdev_dtl_empty(vdev_t *vd, vdev_dtl_type_t t)
	{
	range_tree_t *rt = vd->vdev_dtl[t];
	boolean_t empty;

	mutex_enter(&vd->vdev_dtl_lock);
	empty = range_tree_is_empty(rt);
	mutex_exit(&vd->vdev_dtl_lock);

	return (empty);
	}

	/*
	* Check if the txg falls within the range which must be
	* resilvered. DVAs outside this range can always be skipped.
	*/
	boolean_t
	vdev_default_need_resilver(vdev_t vd, const dva_t dva, size_t psize,
	uint64_t phys_birth)
	{
	(void) dva, (void) psize;

	/* Set by sequential resilver. */
	if (phys_birth == TXG_UNKNOWN)
	return (B_TRUE);

	return (vdev_dtl_contains(vd, DTL_PARTIAL, phys_birth, 1));
	}

	/*
	* Returns B_TRUE if the vdev determines the DVA needs to be resilvered.
	*/
	boolean_t
	vdev_dtl_need_resilver(vdev_t vd, const dva_t dva, size_t psize,
	uint64_t phys_birth)
	{
	ASSERT(vd != vd->vdev_spa->spa_root_vdev);

	if (vd->vdev_ops->vdev_op_need_resilver == NULL \|\|
	vd->vdev_ops->vdev_op_leaf)
	return (B_TRUE);

	return (vd->vdev_ops->vdev_op_need_resilver(vd, dva, psize,
	phys_birth));
	}

	/*
	* Returns the lowest txg in the DTL range.
	*/
	static uint64_t
	vdev_dtl_min(vdev_t *vd)
	{
	ASSERT(MUTEX_HELD(&vd->vdev_dtl_lock));
	ASSERT3U(range_tree_space(vd->vdev_dtl[DTL_MISSING]), !=, 0);
	ASSERT0(vd->vdev_children);

	return (range_tree_min(vd->vdev_dtl[DTL_MISSING]) - 1);
	}

	/*
	* Returns the highest txg in the DTL.
	*/
	static uint64_t
	vdev_dtl_max(vdev_t *vd)
	{
	ASSERT(MUTEX_HELD(&vd->vdev_dtl_lock));
	ASSERT3U(range_tree_space(vd->vdev_dtl[DTL_MISSING]), !=, 0);
	ASSERT0(vd->vdev_children);

	return (range_tree_max(vd->vdev_dtl[DTL_MISSING]));
	}

	/*
	* Determine if a resilvering vdev should remove any DTL entries from
	* its range. If the vdev was resilvering for the entire duration of the
	* scan then it should excise that range from its DTLs. Otherwise, this
	* vdev is considered partially resilvered and should leave its DTL
	* entries intact. The comment in vdev_dtl_reassess() describes how we
	* excise the DTLs.
	*/
	static boolean_t
	vdev_dtl_should_excise(vdev_t *vd, boolean_t rebuild_done)
	{
	ASSERT0(vd->vdev_children);

	if (vd->vdev_state < VDEV_STATE_DEGRADED)
	return (B_FALSE);

	if (vd->vdev_resilver_deferred)
	return (B_FALSE);

	if (range_tree_is_empty(vd->vdev_dtl[DTL_MISSING]))
	return (B_TRUE);

	if (rebuild_done) {
	vdev_rebuild_t *vr = &vd->vdev_top->vdev_rebuild_config;
	vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;

	/* Rebuild not initiated by attach */
	if (vd->vdev_rebuild_txg == 0)
	return (B_TRUE);

	/*
	* When a rebuild completes without error then all missing data
	* up to the rebuild max txg has been reconstructed and the DTL
	* is eligible for excision.
	*/
	if (vrp->vrp_rebuild_state == VDEV_REBUILD_COMPLETE &&
	vdev_dtl_max(vd) <= vrp->vrp_max_txg) {
	ASSERT3U(vrp->vrp_min_txg, <=, vdev_dtl_min(vd));
	ASSERT3U(vrp->vrp_min_txg, <, vd->vdev_rebuild_txg);
	ASSERT3U(vd->vdev_rebuild_txg, <=, vrp->vrp_max_txg);
	return (B_TRUE);
	}
	} else {
	dsl_scan_t *scn = vd->vdev_spa->spa_dsl_pool->dp_scan;
	dsl_scan_phys_t *scnp __maybe_unused = &scn->scn_phys;

	/* Resilver not initiated by attach */
	if (vd->vdev_resilver_txg == 0)
	return (B_TRUE);

	/*
	* When a resilver is initiated the scan will assign the
	* scn_max_txg value to the highest txg value that exists
	* in all DTLs. If this device's max DTL is not part of this
	* scan (i.e. it is not in the range (scn_min_txg, scn_max_txg]
	* then it is not eligible for excision.
	*/
	if (vdev_dtl_max(vd) <= scn->scn_phys.scn_max_txg) {
	ASSERT3U(scnp->scn_min_txg, <=, vdev_dtl_min(vd));
	ASSERT3U(scnp->scn_min_txg, <, vd->vdev_resilver_txg);
	ASSERT3U(vd->vdev_resilver_txg, <=, scnp->scn_max_txg);
	return (B_TRUE);
	}
	}

	return (B_FALSE);
	}

	/*
	* Reassess DTLs after a config change or scrub completion. If txg == 0 no
	* write operations will be issued to the pool.
	*/
	void
	vdev_dtl_reassess(vdev_t *vd, uint64_t txg, uint64_t scrub_txg,
	boolean_t scrub_done, boolean_t rebuild_done)
	{
	spa_t *spa = vd->vdev_spa;
	avl_tree_t reftree;
	int minref;

	ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0);

	for (int c = 0; c < vd->vdev_children; c++)
	vdev_dtl_reassess(vd->vdev_child[c], txg,
	scrub_txg, scrub_done, rebuild_done);

	if (vd == spa->spa_root_vdev \|\| !vdev_is_concrete(vd) \|\| vd->vdev_aux)
	return;

	if (vd->vdev_ops->vdev_op_leaf) {
	dsl_scan_t *scn = spa->spa_dsl_pool->dp_scan;
	vdev_rebuild_t *vr = &vd->vdev_top->vdev_rebuild_config;
	boolean_t check_excise = B_FALSE;
	boolean_t wasempty = B_TRUE;

	mutex_enter(&vd->vdev_dtl_lock);

	/*
	* If requested, pretend the scan or rebuild completed cleanly.
	*/
	if (zfs_scan_ignore_errors) {
	if (scn != NULL)
	scn->scn_phys.scn_errors = 0;
	if (vr != NULL)
	vr->vr_rebuild_phys.vrp_errors = 0;
	}

	if (scrub_txg != 0 &&
	!range_tree_is_empty(vd->vdev_dtl[DTL_MISSING])) {
	wasempty = B_FALSE;
	zfs_dbgmsg("guid:%llu txg:%llu scrub:%llu started:%d "
	"dtl:%llu/%llu errors:%llu",
	(u_longlong_t)vd->vdev_guid, (u_longlong_t)txg,
	(u_longlong_t)scrub_txg, spa->spa_scrub_started,
	(u_longlong_t)vdev_dtl_min(vd),
	(u_longlong_t)vdev_dtl_max(vd),
	(u_longlong_t)(scn ? scn->scn_phys.scn_errors : 0));
	}

	/*
	* If we've completed a scrub/resilver or a rebuild cleanly
	* then determine if this vdev should remove any DTLs. We
	* only want to excise regions on vdevs that were available
	* during the entire duration of this scan.
	*/
	if (rebuild_done &&
	vr != NULL && vr->vr_rebuild_phys.vrp_errors == 0) {
	check_excise = B_TRUE;
	} else {
	if (spa->spa_scrub_started \|\|
	(scn != NULL && scn->scn_phys.scn_errors == 0)) {
	check_excise = B_TRUE;
	}
	}

	if (scrub_txg && check_excise &&
	vdev_dtl_should_excise(vd, rebuild_done)) {
	/*
	* We completed a scrub, resilver or rebuild up to
	* scrub_txg. If we did it without rebooting, then
	* the scrub dtl will be valid, so excise the old
	* region and fold in the scrub dtl. Otherwise,
	* leave the dtl as-is if there was an error.
	*
	* There's little trick here: to excise the beginning
	* of the DTL_MISSING map, we put it into a reference
	* tree and then add a segment with refcnt -1 that
	* covers the range [0, scrub_txg). This means
	* that each txg in that range has refcnt -1 or 0.
	* We then add DTL_SCRUB with a refcnt of 2, so that
	* entries in the range [0, scrub_txg) will have a
	* positive refcnt -- either 1 or 2. We then convert
	* the reference tree into the new DTL_MISSING map.
	*/
	space_reftree_create(&reftree);
	space_reftree_add_map(&reftree,
	vd->vdev_dtl[DTL_MISSING], 1);
	space_reftree_add_seg(&reftree, 0, scrub_txg, -1);
	space_reftree_add_map(&reftree,
	vd->vdev_dtl[DTL_SCRUB], 2);
	space_reftree_generate_map(&reftree,
	vd->vdev_dtl[DTL_MISSING], 1);
	space_reftree_destroy(&reftree);

	if (!range_tree_is_empty(vd->vdev_dtl[DTL_MISSING])) {
	zfs_dbgmsg("update DTL_MISSING:%llu/%llu",
	(u_longlong_t)vdev_dtl_min(vd),
	(u_longlong_t)vdev_dtl_max(vd));
	} else if (!wasempty) {
	zfs_dbgmsg("DTL_MISSING is now empty");
	}
	}
	range_tree_vacate(vd->vdev_dtl[DTL_PARTIAL], NULL, NULL);
	range_tree_walk(vd->vdev_dtl[DTL_MISSING],
	range_tree_add, vd->vdev_dtl[DTL_PARTIAL]);
	if (scrub_done)
	range_tree_vacate(vd->vdev_dtl[DTL_SCRUB], NULL, NULL);
	range_tree_vacate(vd->vdev_dtl[DTL_OUTAGE], NULL, NULL);
	if (!vdev_readable(vd))
	range_tree_add(vd->vdev_dtl[DTL_OUTAGE], 0, -1ULL);
	else
	range_tree_walk(vd->vdev_dtl[DTL_MISSING],
	range_tree_add, vd->vdev_dtl[DTL_OUTAGE]);

	/*
	* If the vdev was resilvering or rebuilding and no longer
	* has any DTLs then reset the appropriate flag and dirty
	* the top level so that we persist the change.
	*/
	if (txg != 0 &&
	range_tree_is_empty(vd->vdev_dtl[DTL_MISSING]) &&
	range_tree_is_empty(vd->vdev_dtl[DTL_OUTAGE])) {
	if (vd->vdev_rebuild_txg != 0) {
	vd->vdev_rebuild_txg = 0;
	vdev_config_dirty(vd->vdev_top);
	} else if (vd->vdev_resilver_txg != 0) {
	vd->vdev_resilver_txg = 0;
	vdev_config_dirty(vd->vdev_top);
	}
	}

	mutex_exit(&vd->vdev_dtl_lock);

	if (txg != 0)
	vdev_dirty(vd->vdev_top, VDD_DTL, vd, txg);
	} else {
	mutex_enter(&vd->vdev_dtl_lock);
	for (int t = 0; t < DTL_TYPES; t++) {
	/* account for child's outage in parent's missing map */
	int s = (t == DTL_MISSING) ? DTL_OUTAGE: t;
	if (t == DTL_SCRUB) {
	/* leaf vdevs only */
	continue;
	}
	if (t == DTL_PARTIAL) {
	/* i.e. non-zero */
	minref = 1;
	} else if (vdev_get_nparity(vd) != 0) {
	/* RAIDZ, DRAID */
	minref = vdev_get_nparity(vd) + 1;
	} else {
	/* any kind of mirror */
	minref = vd->vdev_children;
	}
	space_reftree_create(&reftree);
	for (int c = 0; c < vd->vdev_children; c++) {
	vdev_t *cvd = vd->vdev_child[c];
	mutex_enter(&cvd->vdev_dtl_lock);
	space_reftree_add_map(&reftree,
	cvd->vdev_dtl[s], 1);
	mutex_exit(&cvd->vdev_dtl_lock);
	}
	space_reftree_generate_map(&reftree,
	vd->vdev_dtl[t], minref);
	space_reftree_destroy(&reftree);
	}
	mutex_exit(&vd->vdev_dtl_lock);
	}

	if (vd->vdev_top->vdev_ops == &vdev_raidz_ops) {
	raidz_dtl_reassessed(vd);
	}
	}

	/*
	* Iterate over all the vdevs except spare, and post kobj events
	*/
	void
	vdev_post_kobj_evt(vdev_t *vd)
	{
	if (vd->vdev_ops->vdev_op_kobj_evt_post &&
	vd->vdev_kobj_flag == B_FALSE) {
	vd->vdev_kobj_flag = B_TRUE;
	vd->vdev_ops->vdev_op_kobj_evt_post(vd);
	}

	for (int c = 0; c < vd->vdev_children; c++)
	vdev_post_kobj_evt(vd->vdev_child[c]);
	}

	/*
	* Iterate over all the vdevs except spare, and clear kobj events
	*/
	void
	vdev_clear_kobj_evt(vdev_t *vd)
	{
	vd->vdev_kobj_flag = B_FALSE;

	for (int c = 0; c < vd->vdev_children; c++)
	vdev_clear_kobj_evt(vd->vdev_child[c]);
	}

	int
	vdev_dtl_load(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;
	objset_t *mos = spa->spa_meta_objset;
	range_tree_t *rt;
	int error = 0;

	if (vd->vdev_ops->vdev_op_leaf && vd->vdev_dtl_object != 0) {
	ASSERT(vdev_is_concrete(vd));

	/*
	* If the dtl cannot be sync'd there is no need to open it.
	*/
	if (spa->spa_mode == SPA_MODE_READ && !spa->spa_read_spacemaps)
	return (0);

	error = space_map_open(&vd->vdev_dtl_sm, mos,
	vd->vdev_dtl_object, 0, -1ULL, 0);
	if (error)
	return (error);
	ASSERT(vd->vdev_dtl_sm != NULL);

	rt = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0);
	error = space_map_load(vd->vdev_dtl_sm, rt, SM_ALLOC);
	if (error == 0) {
	mutex_enter(&vd->vdev_dtl_lock);
	range_tree_walk(rt, range_tree_add,
	vd->vdev_dtl[DTL_MISSING]);
	mutex_exit(&vd->vdev_dtl_lock);
	}

	range_tree_vacate(rt, NULL, NULL);
	range_tree_destroy(rt);

	return (error);
	}

	for (int c = 0; c < vd->vdev_children; c++) {
	error = vdev_dtl_load(vd->vdev_child[c]);
	if (error != 0)
	break;
	}

	return (error);
	}

	static void
	vdev_zap_allocation_data(vdev_t vd, dmu_tx_t tx)
	{
	spa_t *spa = vd->vdev_spa;
	objset_t *mos = spa->spa_meta_objset;
	vdev_alloc_bias_t alloc_bias = vd->vdev_alloc_bias;
	const char *string;

	ASSERT(alloc_bias != VDEV_BIAS_NONE);

	string =
	(alloc_bias == VDEV_BIAS_LOG) ? VDEV_ALLOC_BIAS_LOG :
	(alloc_bias == VDEV_BIAS_SPECIAL) ? VDEV_ALLOC_BIAS_SPECIAL :
	(alloc_bias == VDEV_BIAS_DEDUP) ? VDEV_ALLOC_BIAS_DEDUP : NULL;

	ASSERT(string != NULL);
	VERIFY0(zap_add(mos, vd->vdev_top_zap, VDEV_TOP_ZAP_ALLOCATION_BIAS,
	1, strlen(string) + 1, string, tx));

	if (alloc_bias == VDEV_BIAS_SPECIAL \|\| alloc_bias == VDEV_BIAS_DEDUP) {
	spa_activate_allocation_classes(spa, tx);
	}
	}

	void
	vdev_destroy_unlink_zap(vdev_t vd, uint64_t zapobj, dmu_tx_t tx)
	{
	spa_t *spa = vd->vdev_spa;

	VERIFY0(zap_destroy(spa->spa_meta_objset, zapobj, tx));
	VERIFY0(zap_remove_int(spa->spa_meta_objset, spa->spa_all_vdev_zaps,
	zapobj, tx));
	}

	uint64_t
	vdev_create_link_zap(vdev_t vd, dmu_tx_t tx)
	{
	spa_t *spa = vd->vdev_spa;
	uint64_t zap = zap_create(spa->spa_meta_objset, DMU_OTN_ZAP_METADATA,
	DMU_OT_NONE, 0, tx);

	ASSERT(zap != 0);
	VERIFY0(zap_add_int(spa->spa_meta_objset, spa->spa_all_vdev_zaps,
	zap, tx));

	return (zap);
	}

	void
	vdev_construct_zaps(vdev_t vd, dmu_tx_t tx)
	{
	if (vd->vdev_ops != &vdev_hole_ops &&
	vd->vdev_ops != &vdev_missing_ops &&
	vd->vdev_ops != &vdev_root_ops &&
	!vd->vdev_top->vdev_removing) {
	if (vd->vdev_ops->vdev_op_leaf && vd->vdev_leaf_zap == 0) {
	vd->vdev_leaf_zap = vdev_create_link_zap(vd, tx);
	}
	if (vd == vd->vdev_top && vd->vdev_top_zap == 0) {
	vd->vdev_top_zap = vdev_create_link_zap(vd, tx);
	if (vd->vdev_alloc_bias != VDEV_BIAS_NONE)
	vdev_zap_allocation_data(vd, tx);
	}
	}
	if (vd->vdev_ops == &vdev_root_ops && vd->vdev_root_zap == 0 &&
	spa_feature_is_enabled(vd->vdev_spa, SPA_FEATURE_AVZ_V2)) {
	if (!spa_feature_is_active(vd->vdev_spa, SPA_FEATURE_AVZ_V2))
	spa_feature_incr(vd->vdev_spa, SPA_FEATURE_AVZ_V2, tx);
	vd->vdev_root_zap = vdev_create_link_zap(vd, tx);
	}

	for (uint64_t i = 0; i < vd->vdev_children; i++) {
	vdev_construct_zaps(vd->vdev_child[i], tx);
	}
	}

	static void
	vdev_dtl_sync(vdev_t *vd, uint64_t txg)
	{
	spa_t *spa = vd->vdev_spa;
	range_tree_t *rt = vd->vdev_dtl[DTL_MISSING];
	objset_t *mos = spa->spa_meta_objset;
	range_tree_t *rtsync;
	dmu_tx_t *tx;
	uint64_t object = space_map_object(vd->vdev_dtl_sm);

	ASSERT(vdev_is_concrete(vd));
	ASSERT(vd->vdev_ops->vdev_op_leaf);

	tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);

	if (vd->vdev_detached \|\| vd->vdev_top->vdev_removing) {
	mutex_enter(&vd->vdev_dtl_lock);
	space_map_free(vd->vdev_dtl_sm, tx);
	space_map_close(vd->vdev_dtl_sm);
	vd->vdev_dtl_sm = NULL;
	mutex_exit(&vd->vdev_dtl_lock);

	/*
	* We only destroy the leaf ZAP for detached leaves or for
	* removed log devices. Removed data devices handle leaf ZAP
	* cleanup later, once cancellation is no longer possible.
	*/
	if (vd->vdev_leaf_zap != 0 && (vd->vdev_detached \|\|
	vd->vdev_top->vdev_islog)) {
	vdev_destroy_unlink_zap(vd, vd->vdev_leaf_zap, tx);
	vd->vdev_leaf_zap = 0;
	}

	dmu_tx_commit(tx);
	return;
	}

	if (vd->vdev_dtl_sm == NULL) {
	uint64_t new_object;

	new_object = space_map_alloc(mos, zfs_vdev_dtl_sm_blksz, tx);
	VERIFY3U(new_object, !=, 0);

	VERIFY0(space_map_open(&vd->vdev_dtl_sm, mos, new_object,
	0, -1ULL, 0));
	ASSERT(vd->vdev_dtl_sm != NULL);
	}

	rtsync = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0);

	mutex_enter(&vd->vdev_dtl_lock);
	range_tree_walk(rt, range_tree_add, rtsync);
	mutex_exit(&vd->vdev_dtl_lock);

	space_map_truncate(vd->vdev_dtl_sm, zfs_vdev_dtl_sm_blksz, tx);
	space_map_write(vd->vdev_dtl_sm, rtsync, SM_ALLOC, SM_NO_VDEVID, tx);
	range_tree_vacate(rtsync, NULL, NULL);

	range_tree_destroy(rtsync);

	/*
	* If the object for the space map has changed then dirty
	* the top level so that we update the config.
	*/
	if (object != space_map_object(vd->vdev_dtl_sm)) {
	vdev_dbgmsg(vd, "txg %llu, spa %s, DTL old object %llu, "
	"new object %llu", (u_longlong_t)txg, spa_name(spa),
	(u_longlong_t)object,
	(u_longlong_t)space_map_object(vd->vdev_dtl_sm));
	vdev_config_dirty(vd->vdev_top);
	}

	dmu_tx_commit(tx);
	}

	/*
	* Determine whether the specified vdev can be offlined/detached/removed
	* without losing data.
	*/
	boolean_t
	vdev_dtl_required(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;
	vdev_t *tvd = vd->vdev_top;
	uint8_t cant_read = vd->vdev_cant_read;
	boolean_t required;

	ASSERT(spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);

	if (vd == spa->spa_root_vdev \|\| vd == tvd)
	return (B_TRUE);

	/*
	* Temporarily mark the device as unreadable, and then determine
	* whether this results in any DTL outages in the top-level vdev.
	* If not, we can safely offline/detach/remove the device.
	*/
	vd->vdev_cant_read = B_TRUE;
	vdev_dtl_reassess(tvd, 0, 0, B_FALSE, B_FALSE);
	required = !vdev_dtl_empty(tvd, DTL_OUTAGE);
	vd->vdev_cant_read = cant_read;
	vdev_dtl_reassess(tvd, 0, 0, B_FALSE, B_FALSE);

	if (!required && zio_injection_enabled) {
	required = !!zio_handle_device_injection(vd, NULL,
	SET_ERROR(ECHILD));
	}

	return (required);
	}

	/*
	* Determine if resilver is needed, and if so the txg range.
	*/
	boolean_t
	vdev_resilver_needed(vdev_t vd, uint64_t minp, uint64_t *maxp)
	{
	boolean_t needed = B_FALSE;
	uint64_t thismin = UINT64_MAX;
	uint64_t thismax = 0;

	if (vd->vdev_children == 0) {
	mutex_enter(&vd->vdev_dtl_lock);
	if (!range_tree_is_empty(vd->vdev_dtl[DTL_MISSING]) &&
	vdev_writeable(vd)) {

	thismin = vdev_dtl_min(vd);
	thismax = vdev_dtl_max(vd);
	needed = B_TRUE;
	}
	mutex_exit(&vd->vdev_dtl_lock);
	} else {
	for (int c = 0; c < vd->vdev_children; c++) {
	vdev_t *cvd = vd->vdev_child[c];
	uint64_t cmin, cmax;

	if (vdev_resilver_needed(cvd, &cmin, &cmax)) {
	thismin = MIN(thismin, cmin);
	thismax = MAX(thismax, cmax);
	needed = B_TRUE;
	}
	}
	}

	if (needed && minp) {
	*minp = thismin;
	*maxp = thismax;
	}
	return (needed);
	}

	/*
	* Gets the checkpoint space map object from the vdev's ZAP. On success sm_obj
	* will contain either the checkpoint spacemap object or zero if none exists.
	* All other errors are returned to the caller.
	*/
	int
	vdev_checkpoint_sm_object(vdev_t vd, uint64_t sm_obj)
	{
	ASSERT0(spa_config_held(vd->vdev_spa, SCL_ALL, RW_WRITER));

	if (vd->vdev_top_zap == 0) {
	*sm_obj = 0;
	return (0);
	}

	int error = zap_lookup(spa_meta_objset(vd->vdev_spa), vd->vdev_top_zap,
	VDEV_TOP_ZAP_POOL_CHECKPOINT_SM, sizeof (uint64_t), 1, sm_obj);
	if (error == ENOENT) {
	*sm_obj = 0;
	error = 0;
	}

	return (error);
	}

	int
	vdev_load(vdev_t *vd)
	{
	int children = vd->vdev_children;
	int error = 0;
	taskq_t *tq = NULL;

	/*
	* It's only worthwhile to use the taskq for the root vdev, because the
	* slow part is metaslab_init, and that only happens for top-level
	* vdevs.
	*/
	if (vd->vdev_ops == &vdev_root_ops && vd->vdev_children > 0) {
	tq = taskq_create("vdev_load", children, minclsyspri,
	children, children, TASKQ_PREPOPULATE);
	}

	/*
	* Recursively load all children.
	*/
	for (int c = 0; c < vd->vdev_children; c++) {
	vdev_t *cvd = vd->vdev_child[c];

	if (tq == NULL \|\| vdev_uses_zvols(cvd)) {
	cvd->vdev_load_error = vdev_load(cvd);
	} else {
	VERIFY(taskq_dispatch(tq, vdev_load_child,
	cvd, TQ_SLEEP) != TASKQID_INVALID);
	}
	}

	if (tq != NULL) {
	taskq_wait(tq);
	taskq_destroy(tq);
	}

	for (int c = 0; c < vd->vdev_children; c++) {
	int error = vd->vdev_child[c]->vdev_load_error;

	if (error != 0)
	return (error);
	}

	vdev_set_deflate_ratio(vd);

	if (vd->vdev_ops == &vdev_raidz_ops) {
	error = vdev_raidz_load(vd);
	if (error != 0)
	return (error);
	}

	/*
	* On spa_load path, grab the allocation bias from our zap
	*/
	if (vd == vd->vdev_top && vd->vdev_top_zap != 0) {
	spa_t *spa = vd->vdev_spa;
	char bias_str[64];

	error = zap_lookup(spa->spa_meta_objset, vd->vdev_top_zap,
	VDEV_TOP_ZAP_ALLOCATION_BIAS, 1, sizeof (bias_str),
	bias_str);
	if (error == 0) {
	ASSERT(vd->vdev_alloc_bias == VDEV_BIAS_NONE);
	vd->vdev_alloc_bias = vdev_derive_alloc_bias(bias_str);
	} else if (error != ENOENT) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	vdev_dbgmsg(vd, "vdev_load: zap_lookup(top_zap=%llu) "
	"failed [error=%d]",
	(u_longlong_t)vd->vdev_top_zap, error);
	return (error);
	}
	}

	if (vd == vd->vdev_top && vd->vdev_top_zap != 0) {
	spa_t *spa = vd->vdev_spa;
	uint64_t failfast;

	error = zap_lookup(spa->spa_meta_objset, vd->vdev_top_zap,
	vdev_prop_to_name(VDEV_PROP_FAILFAST), sizeof (failfast),
	1, &failfast);
	if (error == 0) {
	vd->vdev_failfast = failfast & 1;
	} else if (error == ENOENT) {
	vd->vdev_failfast = vdev_prop_default_numeric(
	VDEV_PROP_FAILFAST);
	} else {
	vdev_dbgmsg(vd,
	"vdev_load: zap_lookup(top_zap=%llu) "
	"failed [error=%d]",
	(u_longlong_t)vd->vdev_top_zap, error);
	}
	}

	/*
	* Load any rebuild state from the top-level vdev zap.
	*/
	if (vd == vd->vdev_top && vd->vdev_top_zap != 0) {
	error = vdev_rebuild_load(vd);
	if (error && error != ENOTSUP) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	vdev_dbgmsg(vd, "vdev_load: vdev_rebuild_load "
	"failed [error=%d]", error);
	return (error);
	}
	}

	if (vd->vdev_top_zap != 0 \|\| vd->vdev_leaf_zap != 0) {
	uint64_t zapobj;

	if (vd->vdev_top_zap != 0)
	zapobj = vd->vdev_top_zap;
	else
	zapobj = vd->vdev_leaf_zap;

	error = vdev_prop_get_int(vd, VDEV_PROP_CHECKSUM_N,
	&vd->vdev_checksum_n);
	if (error && error != ENOENT)
	vdev_dbgmsg(vd, "vdev_load: zap_lookup(zap=%llu) "
	"failed [error=%d]", (u_longlong_t)zapobj, error);

	error = vdev_prop_get_int(vd, VDEV_PROP_CHECKSUM_T,
	&vd->vdev_checksum_t);
	if (error && error != ENOENT)
	vdev_dbgmsg(vd, "vdev_load: zap_lookup(zap=%llu) "
	"failed [error=%d]", (u_longlong_t)zapobj, error);

	error = vdev_prop_get_int(vd, VDEV_PROP_IO_N,
	&vd->vdev_io_n);
	if (error && error != ENOENT)
	vdev_dbgmsg(vd, "vdev_load: zap_lookup(zap=%llu) "
	"failed [error=%d]", (u_longlong_t)zapobj, error);

	error = vdev_prop_get_int(vd, VDEV_PROP_IO_T,
	&vd->vdev_io_t);
	if (error && error != ENOENT)
	vdev_dbgmsg(vd, "vdev_load: zap_lookup(zap=%llu) "
	"failed [error=%d]", (u_longlong_t)zapobj, error);

	error = vdev_prop_get_int(vd, VDEV_PROP_SLOW_IO_N,
	&vd->vdev_slow_io_n);
	if (error && error != ENOENT)
	vdev_dbgmsg(vd, "vdev_load: zap_lookup(zap=%llu) "
	"failed [error=%d]", (u_longlong_t)zapobj, error);

	error = vdev_prop_get_int(vd, VDEV_PROP_SLOW_IO_T,
	&vd->vdev_slow_io_t);
	if (error && error != ENOENT)
	vdev_dbgmsg(vd, "vdev_load: zap_lookup(zap=%llu) "
	"failed [error=%d]", (u_longlong_t)zapobj, error);
	}

	/*
	* If this is a top-level vdev, initialize its metaslabs.
	*/
	if (vd == vd->vdev_top && vdev_is_concrete(vd)) {
	vdev_metaslab_group_create(vd);

	if (vd->vdev_ashift == 0 \|\| vd->vdev_asize == 0) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	vdev_dbgmsg(vd, "vdev_load: invalid size. ashift=%llu, "
	"asize=%llu", (u_longlong_t)vd->vdev_ashift,
	(u_longlong_t)vd->vdev_asize);
	return (SET_ERROR(ENXIO));
	}

	error = vdev_metaslab_init(vd, 0);
	if (error != 0) {
	vdev_dbgmsg(vd, "vdev_load: metaslab_init failed "
	"[error=%d]", error);
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	return (error);
	}

	uint64_t checkpoint_sm_obj;
	error = vdev_checkpoint_sm_object(vd, &checkpoint_sm_obj);
	if (error == 0 && checkpoint_sm_obj != 0) {
	objset_t *mos = spa_meta_objset(vd->vdev_spa);
	ASSERT(vd->vdev_asize != 0);
	ASSERT3P(vd->vdev_checkpoint_sm, ==, NULL);

	error = space_map_open(&vd->vdev_checkpoint_sm,
	mos, checkpoint_sm_obj, 0, vd->vdev_asize,
	vd->vdev_ashift);
	if (error != 0) {
	vdev_dbgmsg(vd, "vdev_load: space_map_open "
	"failed for checkpoint spacemap (obj %llu) "
	"[error=%d]",
	(u_longlong_t)checkpoint_sm_obj, error);
	return (error);
	}
	ASSERT3P(vd->vdev_checkpoint_sm, !=, NULL);

	/*
	* Since the checkpoint_sm contains free entries
	* exclusively we can use space_map_allocated() to
	* indicate the cumulative checkpointed space that
	* has been freed.
	*/
	vd->vdev_stat.vs_checkpoint_space =
	-space_map_allocated(vd->vdev_checkpoint_sm);
	vd->vdev_spa->spa_checkpoint_info.sci_dspace +=
	vd->vdev_stat.vs_checkpoint_space;
	} else if (error != 0) {
	vdev_dbgmsg(vd, "vdev_load: failed to retrieve "
	"checkpoint space map object from vdev ZAP "
	"[error=%d]", error);
	return (error);
	}
	}

	/*
	* If this is a leaf vdev, load its DTL.
	*/
	if (vd->vdev_ops->vdev_op_leaf && (error = vdev_dtl_load(vd)) != 0) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	vdev_dbgmsg(vd, "vdev_load: vdev_dtl_load failed "
	"[error=%d]", error);
	return (error);
	}

	uint64_t obsolete_sm_object;
	error = vdev_obsolete_sm_object(vd, &obsolete_sm_object);
	if (error == 0 && obsolete_sm_object != 0) {
	objset_t *mos = vd->vdev_spa->spa_meta_objset;
	ASSERT(vd->vdev_asize != 0);
	ASSERT3P(vd->vdev_obsolete_sm, ==, NULL);

	if ((error = space_map_open(&vd->vdev_obsolete_sm, mos,
	obsolete_sm_object, 0, vd->vdev_asize, 0))) {
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	vdev_dbgmsg(vd, "vdev_load: space_map_open failed for "
	"obsolete spacemap (obj %llu) [error=%d]",
	(u_longlong_t)obsolete_sm_object, error);
	return (error);
	}
	} else if (error != 0) {
	vdev_dbgmsg(vd, "vdev_load: failed to retrieve obsolete "
	"space map object from vdev ZAP [error=%d]", error);
	return (error);
	}

	return (0);
	}

	/*
	* The special vdev case is used for hot spares and l2cache devices. Its
	* sole purpose it to set the vdev state for the associated vdev. To do this,
	* we make sure that we can open the underlying device, then try to read the
	* label, and make sure that the label is sane and that it hasn't been
	* repurposed to another pool.
	*/
	int
	vdev_validate_aux(vdev_t *vd)
	{
	nvlist_t *label;
	uint64_t guid, version;
	uint64_t state;

	if (!vdev_readable(vd))
	return (0);

	if ((label = vdev_label_read_config(vd, -1ULL)) == NULL) {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	return (-1);
	}

	if (nvlist_lookup_uint64(label, ZPOOL_CONFIG_VERSION, &version) != 0 \|\|
	!SPA_VERSION_IS_SUPPORTED(version) \|\|
	nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID, &guid) != 0 \|\|
	guid != vd->vdev_guid \|\|
	nvlist_lookup_uint64(label, ZPOOL_CONFIG_POOL_STATE, &state) != 0) {
	vdev_set_state(vd, B_TRUE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	nvlist_free(label);
	return (-1);
	}

	/*
	* We don't actually check the pool state here. If it's in fact in
	* use by another pool, we update this fact on the fly when requested.
	*/
	nvlist_free(label);
	return (0);
	}

	static void
	vdev_destroy_ms_flush_data(vdev_t vd, dmu_tx_t tx)
	{
	objset_t *mos = spa_meta_objset(vd->vdev_spa);

	if (vd->vdev_top_zap == 0)
	return;

	uint64_t object = 0;
	int err = zap_lookup(mos, vd->vdev_top_zap,
	VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS, sizeof (uint64_t), 1, &object);
	if (err == ENOENT)
	return;
	VERIFY0(err);

	VERIFY0(dmu_object_free(mos, object, tx));
	VERIFY0(zap_remove(mos, vd->vdev_top_zap,
	VDEV_TOP_ZAP_MS_UNFLUSHED_PHYS_TXGS, tx));
	}

	/*
	* Free the objects used to store this vdev's spacemaps, and the array
	* that points to them.
	*/
	void
	vdev_destroy_spacemaps(vdev_t vd, dmu_tx_t tx)
	{
	if (vd->vdev_ms_array == 0)
	return;

	objset_t *mos = vd->vdev_spa->spa_meta_objset;
	uint64_t array_count = vd->vdev_asize >> vd->vdev_ms_shift;
	size_t array_bytes = array_count * sizeof (uint64_t);
	uint64_t *smobj_array = kmem_alloc(array_bytes, KM_SLEEP);
	VERIFY0(dmu_read(mos, vd->vdev_ms_array, 0,
	array_bytes, smobj_array, 0));

	for (uint64_t i = 0; i < array_count; i++) {
	uint64_t smobj = smobj_array[i];
	if (smobj == 0)
	continue;

	space_map_free_obj(mos, smobj, tx);
	}

	kmem_free(smobj_array, array_bytes);
	VERIFY0(dmu_object_free(mos, vd->vdev_ms_array, tx));
	vdev_destroy_ms_flush_data(vd, tx);
	vd->vdev_ms_array = 0;
	}

	static void
	vdev_remove_empty_log(vdev_t *vd, uint64_t txg)
	{
	spa_t *spa = vd->vdev_spa;

	ASSERT(vd->vdev_islog);
	ASSERT(vd == vd->vdev_top);
	ASSERT3U(txg, ==, spa_syncing_txg(spa));

	dmu_tx_t *tx = dmu_tx_create_assigned(spa_get_dsl(spa), txg);

	vdev_destroy_spacemaps(vd, tx);
	if (vd->vdev_top_zap != 0) {
	vdev_destroy_unlink_zap(vd, vd->vdev_top_zap, tx);
	vd->vdev_top_zap = 0;
	}

	dmu_tx_commit(tx);
	}

	void
	vdev_sync_done(vdev_t *vd, uint64_t txg)
	{
	metaslab_t *msp;
	boolean_t reassess = !txg_list_empty(&vd->vdev_ms_list, TXG_CLEAN(txg));

	ASSERT(vdev_is_concrete(vd));

	while ((msp = txg_list_remove(&vd->vdev_ms_list, TXG_CLEAN(txg)))
	!= NULL)
	metaslab_sync_done(msp, txg);

	if (reassess) {
	metaslab_sync_reassess(vd->vdev_mg);
	if (vd->vdev_log_mg != NULL)
	metaslab_sync_reassess(vd->vdev_log_mg);
	}
	}

	void
	vdev_sync(vdev_t *vd, uint64_t txg)
	{
	spa_t *spa = vd->vdev_spa;
	vdev_t *lvd;
	metaslab_t *msp;

	ASSERT3U(txg, ==, spa->spa_syncing_txg);
	dmu_tx_t *tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);
	if (range_tree_space(vd->vdev_obsolete_segments) > 0) {
	ASSERT(vd->vdev_removing \|\|
	vd->vdev_ops == &vdev_indirect_ops);

	vdev_indirect_sync_obsolete(vd, tx);

	/*
	* If the vdev is indirect, it can't have dirty
	* metaslabs or DTLs.
	*/
	if (vd->vdev_ops == &vdev_indirect_ops) {
	ASSERT(txg_list_empty(&vd->vdev_ms_list, txg));
	ASSERT(txg_list_empty(&vd->vdev_dtl_list, txg));
	dmu_tx_commit(tx);
	return;
	}
	}

	ASSERT(vdev_is_concrete(vd));

	if (vd->vdev_ms_array == 0 && vd->vdev_ms_shift != 0 &&
	!vd->vdev_removing) {
	ASSERT(vd == vd->vdev_top);
	ASSERT0(vd->vdev_indirect_config.vic_mapping_object);
	vd->vdev_ms_array = dmu_object_alloc(spa->spa_meta_objset,
	DMU_OT_OBJECT_ARRAY, 0, DMU_OT_NONE, 0, tx);
	ASSERT(vd->vdev_ms_array != 0);
	vdev_config_dirty(vd);
	}

	while ((msp = txg_list_remove(&vd->vdev_ms_list, txg)) != NULL) {
	metaslab_sync(msp, txg);
	(void) txg_list_add(&vd->vdev_ms_list, msp, TXG_CLEAN(txg));
	}

	while ((lvd = txg_list_remove(&vd->vdev_dtl_list, txg)) != NULL)
	vdev_dtl_sync(lvd, txg);

	/*
	* If this is an empty log device being removed, destroy the
	* metadata associated with it.
	*/
	if (vd->vdev_islog && vd->vdev_stat.vs_alloc == 0 && vd->vdev_removing)
	vdev_remove_empty_log(vd, txg);

	(void) txg_list_add(&spa->spa_vdev_txg_list, vd, TXG_CLEAN(txg));
	dmu_tx_commit(tx);
	}

	/*
	* Return the amount of space that should be (or was) allocated for the given
	* psize (compressed block size) in the given TXG. Note that for expanded
	* RAIDZ vdevs, the size allocated for older BP's may be larger. See
	* vdev_raidz_asize().
	*/
	uint64_t
	vdev_psize_to_asize_txg(vdev_t *vd, uint64_t psize, uint64_t txg)
	{
	return (vd->vdev_ops->vdev_op_asize(vd, psize, txg));
	}

	uint64_t
	vdev_psize_to_asize(vdev_t *vd, uint64_t psize)
	{
	return (vdev_psize_to_asize_txg(vd, psize, 0));
	}

	/*
	* Mark the given vdev faulted. A faulted vdev behaves as if the device could
	* not be opened, and no I/O is attempted.
	*/
	int
	vdev_fault(spa_t *spa, uint64_t guid, vdev_aux_t aux)
	{
	vdev_t vd, tvd;

	spa_vdev_state_enter(spa, SCL_NONE);

	if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL)
	return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENODEV)));

	if (!vd->vdev_ops->vdev_op_leaf)
	return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENOTSUP)));

	tvd = vd->vdev_top;

	/*
	* If user did a 'zpool offline -f' then make the fault persist across
	* reboots.
	*/
	if (aux == VDEV_AUX_EXTERNAL_PERSIST) {
	/*
	* There are two kinds of forced faults: temporary and
	* persistent. Temporary faults go away at pool import, while
	* persistent faults stay set. Both types of faults can be
	* cleared with a zpool clear.
	*
	* We tell if a vdev is persistently faulted by looking at the
	* ZPOOL_CONFIG_AUX_STATE nvpair. If it's set to "external" at
	* import then it's a persistent fault. Otherwise, it's
	* temporary. We get ZPOOL_CONFIG_AUX_STATE set to "external"
	* by setting vd.vdev_stat.vs_aux to VDEV_AUX_EXTERNAL. This
	* tells vdev_config_generate() (which gets run later) to set
	* ZPOOL_CONFIG_AUX_STATE to "external" in the nvlist.
	*/
	vd->vdev_stat.vs_aux = VDEV_AUX_EXTERNAL;
	vd->vdev_tmpoffline = B_FALSE;
	aux = VDEV_AUX_EXTERNAL;
	} else {
	vd->vdev_tmpoffline = B_TRUE;
	}

	/*
	* We don't directly use the aux state here, but if we do a
	* vdev_reopen(), we need this value to be present to remember why we
	* were faulted.
	*/
	vd->vdev_label_aux = aux;

	/*
	* Faulted state takes precedence over degraded.
	*/
	vd->vdev_delayed_close = B_FALSE;
	vd->vdev_faulted = 1ULL;
	vd->vdev_degraded = 0ULL;
	vdev_set_state(vd, B_FALSE, VDEV_STATE_FAULTED, aux);

	/*
	* If this device has the only valid copy of the data, then
	* back off and simply mark the vdev as degraded instead.
	*/
	if (!tvd->vdev_islog && vd->vdev_aux == NULL && vdev_dtl_required(vd)) {
	vd->vdev_degraded = 1ULL;
	vd->vdev_faulted = 0ULL;

	/*
	* If we reopen the device and it's not dead, only then do we
	* mark it degraded.
	*/
	vdev_reopen(tvd);

	if (vdev_readable(vd))
	vdev_set_state(vd, B_FALSE, VDEV_STATE_DEGRADED, aux);
	}

	return (spa_vdev_state_exit(spa, vd, 0));
	}

	/*
	* Mark the given vdev degraded. A degraded vdev is purely an indication to the
	* user that something is wrong. The vdev continues to operate as normal as far
	* as I/O is concerned.
	*/
	int
	vdev_degrade(spa_t *spa, uint64_t guid, vdev_aux_t aux)
	{
	vdev_t *vd;

	spa_vdev_state_enter(spa, SCL_NONE);

	if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL)
	return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENODEV)));

	if (!vd->vdev_ops->vdev_op_leaf)
	return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENOTSUP)));

	/*
	* If the vdev is already faulted, then don't do anything.
	*/
	if (vd->vdev_faulted \|\| vd->vdev_degraded)
	return (spa_vdev_state_exit(spa, NULL, 0));

	vd->vdev_degraded = 1ULL;
	if (!vdev_is_dead(vd))
	vdev_set_state(vd, B_FALSE, VDEV_STATE_DEGRADED,
	aux);

	return (spa_vdev_state_exit(spa, vd, 0));
	}

	int
	vdev_remove_wanted(spa_t *spa, uint64_t guid)
	{
	vdev_t *vd;

	spa_vdev_state_enter(spa, SCL_NONE);

	if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL)
	return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENODEV)));

	/*
	* If the vdev is already removed, or expanding which can trigger
	* repartition add/remove events, then don't do anything.
	*/
	if (vd->vdev_removed \|\| vd->vdev_expanding)
	return (spa_vdev_state_exit(spa, NULL, 0));

	/*
	* Confirm the vdev has been removed, otherwise don't do anything.
	*/
	if (vd->vdev_ops->vdev_op_leaf && !zio_wait(vdev_probe(vd, NULL)))
	return (spa_vdev_state_exit(spa, NULL, SET_ERROR(EEXIST)));

	vd->vdev_remove_wanted = B_TRUE;
	spa_async_request(spa, SPA_ASYNC_REMOVE);

	return (spa_vdev_state_exit(spa, vd, 0));
	}


	/*
	* Online the given vdev.
	*
	* If 'ZFS_ONLINE_UNSPARE' is set, it implies two things. First, any attached
	* spare device should be detached when the device finishes resilvering.
	* Second, the online should be treated like a 'test' online case, so no FMA
	* events are generated if the device fails to open.
	*/
	int
	vdev_online(spa_t spa, uint64_t guid, uint64_t flags, vdev_state_t newstate)
	{
	vdev_t vd, tvd, pvd, rvd = spa->spa_root_vdev;
	boolean_t wasoffline;
	vdev_state_t oldstate;

	spa_vdev_state_enter(spa, SCL_NONE);

	if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL)
	return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENODEV)));

	wasoffline = (vd->vdev_offline \|\| vd->vdev_tmpoffline);
	oldstate = vd->vdev_state;

	tvd = vd->vdev_top;
	vd->vdev_offline = B_FALSE;
	vd->vdev_tmpoffline = B_FALSE;
	vd->vdev_checkremove = !!(flags & ZFS_ONLINE_CHECKREMOVE);
	vd->vdev_forcefault = !!(flags & ZFS_ONLINE_FORCEFAULT);

	/* XXX - L2ARC 1.0 does not support expansion */
	if (!vd->vdev_aux) {
	for (pvd = vd; pvd != rvd; pvd = pvd->vdev_parent)
	pvd->vdev_expanding = !!((flags & ZFS_ONLINE_EXPAND) \|\|
	spa->spa_autoexpand);
	vd->vdev_expansion_time = gethrestime_sec();
	}

	vdev_reopen(tvd);
	vd->vdev_checkremove = vd->vdev_forcefault = B_FALSE;

	if (!vd->vdev_aux) {
	for (pvd = vd; pvd != rvd; pvd = pvd->vdev_parent)
	pvd->vdev_expanding = B_FALSE;
	}

	if (newstate)
	*newstate = vd->vdev_state;
	if ((flags & ZFS_ONLINE_UNSPARE) &&
	!vdev_is_dead(vd) && vd->vdev_parent &&
	vd->vdev_parent->vdev_ops == &vdev_spare_ops &&
	vd->vdev_parent->vdev_child[0] == vd)
	vd->vdev_unspare = B_TRUE;

	if ((flags & ZFS_ONLINE_EXPAND) \|\| spa->spa_autoexpand) {

	/* XXX - L2ARC 1.0 does not support expansion */
	if (vd->vdev_aux)
	return (spa_vdev_state_exit(spa, vd, ENOTSUP));
	spa->spa_ccw_fail_time = 0;
	spa_async_request(spa, SPA_ASYNC_CONFIG_UPDATE);
	}

	/* Restart initializing if necessary */
	mutex_enter(&vd->vdev_initialize_lock);
	if (vdev_writeable(vd) &&
	vd->vdev_initialize_thread == NULL &&
	vd->vdev_initialize_state == VDEV_INITIALIZE_ACTIVE) {
	(void) vdev_initialize(vd);
	}
	mutex_exit(&vd->vdev_initialize_lock);

	/*
	* Restart trimming if necessary. We do not restart trimming for cache
	* devices here. This is triggered by l2arc_rebuild_vdev()
	* asynchronously for the whole device or in l2arc_evict() as it evicts
	* space for upcoming writes.
	*/
	mutex_enter(&vd->vdev_trim_lock);
	if (vdev_writeable(vd) && !vd->vdev_isl2cache &&
	vd->vdev_trim_thread == NULL &&
	vd->vdev_trim_state == VDEV_TRIM_ACTIVE) {
	(void) vdev_trim(vd, vd->vdev_trim_rate, vd->vdev_trim_partial,
	vd->vdev_trim_secure);
	}
	mutex_exit(&vd->vdev_trim_lock);

	if (wasoffline \|\|
	(oldstate < VDEV_STATE_DEGRADED &&
	vd->vdev_state >= VDEV_STATE_DEGRADED)) {
	spa_event_notify(spa, vd, NULL, ESC_ZFS_VDEV_ONLINE);

	/*
	* Asynchronously detach spare vdev if resilver or
	* rebuild is not required
	*/
	if (vd->vdev_unspare &&
	!dsl_scan_resilvering(spa->spa_dsl_pool) &&
	!dsl_scan_resilver_scheduled(spa->spa_dsl_pool) &&
	!vdev_rebuild_active(tvd))
	spa_async_request(spa, SPA_ASYNC_DETACH_SPARE);
	}
	return (spa_vdev_state_exit(spa, vd, 0));
	}

	static int
	vdev_offline_locked(spa_t *spa, uint64_t guid, uint64_t flags)
	{
	vdev_t vd, tvd;
	int error = 0;
	uint64_t generation;
	metaslab_group_t *mg;

	top:
	spa_vdev_state_enter(spa, SCL_ALLOC);

	if ((vd = spa_lookup_by_guid(spa, guid, B_TRUE)) == NULL)
	return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENODEV)));

	if (!vd->vdev_ops->vdev_op_leaf)
	return (spa_vdev_state_exit(spa, NULL, SET_ERROR(ENOTSUP)));

	if (vd->vdev_ops == &vdev_draid_spare_ops)
	return (spa_vdev_state_exit(spa, NULL, ENOTSUP));

	tvd = vd->vdev_top;
	mg = tvd->vdev_mg;
	generation = spa->spa_config_generation + 1;

	/*
	* If the device isn't already offline, try to offline it.
	*/
	if (!vd->vdev_offline) {
	/*
	* If this device has the only valid copy of some data,
	* don't allow it to be offlined. Log devices are always
	* expendable.
	*/
	if (!tvd->vdev_islog && vd->vdev_aux == NULL &&
	vdev_dtl_required(vd))
	return (spa_vdev_state_exit(spa, NULL,
	SET_ERROR(EBUSY)));

	/*
	* If the top-level is a slog and it has had allocations
	* then proceed. We check that the vdev's metaslab group
	* is not NULL since it's possible that we may have just
	* added this vdev but not yet initialized its metaslabs.
	*/
	if (tvd->vdev_islog && mg != NULL) {
	/*
	* Prevent any future allocations.
	*/
	ASSERT3P(tvd->vdev_log_mg, ==, NULL);
	metaslab_group_passivate(mg);
	(void) spa_vdev_state_exit(spa, vd, 0);

	error = spa_reset_logs(spa);

	/*
	* If the log device was successfully reset but has
	* checkpointed data, do not offline it.
	*/
	if (error == 0 &&
	tvd->vdev_checkpoint_sm != NULL) {
	ASSERT3U(space_map_allocated(
	tvd->vdev_checkpoint_sm), !=, 0);
	error = ZFS_ERR_CHECKPOINT_EXISTS;
	}

	spa_vdev_state_enter(spa, SCL_ALLOC);

	/*
	* Check to see if the config has changed.
	*/
	if (error \|\| generation != spa->spa_config_generation) {
	metaslab_group_activate(mg);
	if (error)
	return (spa_vdev_state_exit(spa,
	vd, error));
	(void) spa_vdev_state_exit(spa, vd, 0);
	goto top;
	}
	ASSERT0(tvd->vdev_stat.vs_alloc);
	}

	/*
	* Offline this device and reopen its top-level vdev.
	* If the top-level vdev is a log device then just offline
	* it. Otherwise, if this action results in the top-level
	* vdev becoming unusable, undo it and fail the request.
	*/
	vd->vdev_offline = B_TRUE;
	vdev_reopen(tvd);

	if (!tvd->vdev_islog && vd->vdev_aux == NULL &&
	vdev_is_dead(tvd)) {
	vd->vdev_offline = B_FALSE;
	vdev_reopen(tvd);
	return (spa_vdev_state_exit(spa, NULL,
	SET_ERROR(EBUSY)));
	}

	/*
	* Add the device back into the metaslab rotor so that
	* once we online the device it's open for business.
	*/
	if (tvd->vdev_islog && mg != NULL)
	metaslab_group_activate(mg);
	}

	vd->vdev_tmpoffline = !!(flags & ZFS_OFFLINE_TEMPORARY);

	return (spa_vdev_state_exit(spa, vd, 0));
	}

	int
	vdev_offline(spa_t *spa, uint64_t guid, uint64_t flags)
	{
	int error;

	mutex_enter(&spa->spa_vdev_top_lock);
	error = vdev_offline_locked(spa, guid, flags);
	mutex_exit(&spa->spa_vdev_top_lock);

	return (error);
	}

	/*
	* Clear the error counts associated with this vdev. Unlike vdev_online() and
	* vdev_offline(), we assume the spa config is locked. We also clear all
	* children. If 'vd' is NULL, then the user wants to clear all vdevs.
	*/
	void
	vdev_clear(spa_t spa, vdev_t vd)
	{
	vdev_t *rvd = spa->spa_root_vdev;

	ASSERT(spa_config_held(spa, SCL_STATE_ALL, RW_WRITER) == SCL_STATE_ALL);

	if (vd == NULL)
	vd = rvd;

	vd->vdev_stat.vs_read_errors = 0;
	vd->vdev_stat.vs_write_errors = 0;
	vd->vdev_stat.vs_checksum_errors = 0;
	vd->vdev_stat.vs_slow_ios = 0;

	for (int c = 0; c < vd->vdev_children; c++)
	vdev_clear(spa, vd->vdev_child[c]);

	/*
	* It makes no sense to "clear" an indirect or removed vdev.
	*/
	if (!vdev_is_concrete(vd) \|\| vd->vdev_removed)
	return;

	/*
	* If we're in the FAULTED state or have experienced failed I/O, then
	* clear the persistent state and attempt to reopen the device. We
	* also mark the vdev config dirty, so that the new faulted state is
	* written out to disk.
	*/
	if (vd->vdev_faulted \|\| vd->vdev_degraded \|\|
	!vdev_readable(vd) \|\| !vdev_writeable(vd)) {
	/*
	* When reopening in response to a clear event, it may be due to
	* a fmadm repair request. In this case, if the device is
	* still broken, we want to still post the ereport again.
	*/
	vd->vdev_forcefault = B_TRUE;

	vd->vdev_faulted = vd->vdev_degraded = 0ULL;
	vd->vdev_cant_read = B_FALSE;
	vd->vdev_cant_write = B_FALSE;
	vd->vdev_stat.vs_aux = 0;

	vdev_reopen(vd == rvd ? rvd : vd->vdev_top);

	vd->vdev_forcefault = B_FALSE;

	if (vd != rvd && vdev_writeable(vd->vdev_top))
	vdev_state_dirty(vd->vdev_top);

	/* If a resilver isn't required, check if vdevs can be culled */
	if (vd->vdev_aux == NULL && !vdev_is_dead(vd) &&
	!dsl_scan_resilvering(spa->spa_dsl_pool) &&
	!dsl_scan_resilver_scheduled(spa->spa_dsl_pool))
	spa_async_request(spa, SPA_ASYNC_RESILVER_DONE);

	spa_event_notify(spa, vd, NULL, ESC_ZFS_VDEV_CLEAR);
	}

	/*
	* When clearing a FMA-diagnosed fault, we always want to
	* unspare the device, as we assume that the original spare was
	* done in response to the FMA fault.
	*/
	if (!vdev_is_dead(vd) && vd->vdev_parent != NULL &&
	vd->vdev_parent->vdev_ops == &vdev_spare_ops &&
	vd->vdev_parent->vdev_child[0] == vd)
	vd->vdev_unspare = B_TRUE;

	/* Clear recent error events cache (i.e. duplicate events tracking) */
	zfs_ereport_clear(spa, vd);
	}

	boolean_t
	vdev_is_dead(vdev_t *vd)
	{
	/*
	* Holes and missing devices are always considered "dead".
	* This simplifies the code since we don't have to check for
	* these types of devices in the various code paths.
	* Instead we rely on the fact that we skip over dead devices
	* before issuing I/O to them.
	*/
	return (vd->vdev_state < VDEV_STATE_DEGRADED \|\|
	vd->vdev_ops == &vdev_hole_ops \|\|
	vd->vdev_ops == &vdev_missing_ops);
	}

	boolean_t
	vdev_readable(vdev_t *vd)
	{
	return (!vdev_is_dead(vd) && !vd->vdev_cant_read);
	}

	boolean_t
	vdev_writeable(vdev_t *vd)
	{
	return (!vdev_is_dead(vd) && !vd->vdev_cant_write &&
	vdev_is_concrete(vd));
	}

	boolean_t
	vdev_allocatable(vdev_t *vd)
	{
	uint64_t state = vd->vdev_state;

	/*
	* We currently allow allocations from vdevs which may be in the
	* process of reopening (i.e. VDEV_STATE_CLOSED). If the device
	* fails to reopen then we'll catch it later when we're holding
	* the proper locks. Note that we have to get the vdev state
	* in a local variable because although it changes atomically,
	* we're asking two separate questions about it.
	*/
	return (!(state < VDEV_STATE_DEGRADED && state != VDEV_STATE_CLOSED) &&
	!vd->vdev_cant_write && vdev_is_concrete(vd) &&
	vd->vdev_mg->mg_initialized);
	}

	boolean_t
	vdev_accessible(vdev_t vd, zio_t zio)
	{
	ASSERT(zio->io_vd == vd);

	if (vdev_is_dead(vd) \|\| vd->vdev_remove_wanted)
	return (B_FALSE);

	if (zio->io_type == ZIO_TYPE_READ)
	return (!vd->vdev_cant_read);

	if (zio->io_type == ZIO_TYPE_WRITE)
	return (!vd->vdev_cant_write);

	return (B_TRUE);
	}

	static void
	vdev_get_child_stat(vdev_t cvd, vdev_stat_t vs, vdev_stat_t *cvs)
	{
	/*
	* Exclude the dRAID spare when aggregating to avoid double counting
	* the ops and bytes. These IOs are counted by the physical leaves.
	*/
	if (cvd->vdev_ops == &vdev_draid_spare_ops)
	return;

	for (int t = 0; t < VS_ZIO_TYPES; t++) {
	vs->vs_ops[t] += cvs->vs_ops[t];
	vs->vs_bytes[t] += cvs->vs_bytes[t];
	}

	cvs->vs_scan_removing = cvd->vdev_removing;
	}

	/*
	* Get extended stats
	*/
	static void
	vdev_get_child_stat_ex(vdev_t cvd, vdev_stat_ex_t vsx, vdev_stat_ex_t *cvsx)
	{
	(void) cvd;

	int t, b;
	for (t = 0; t < ZIO_TYPES; t++) {
	for (b = 0; b < ARRAY_SIZE(vsx->vsx_disk_histo[0]); b++)
	vsx->vsx_disk_histo[t][b] += cvsx->vsx_disk_histo[t][b];

	for (b = 0; b < ARRAY_SIZE(vsx->vsx_total_histo[0]); b++) {
	vsx->vsx_total_histo[t][b] +=
	cvsx->vsx_total_histo[t][b];
	}
	}

	for (t = 0; t < ZIO_PRIORITY_NUM_QUEUEABLE; t++) {
	for (b = 0; b < ARRAY_SIZE(vsx->vsx_queue_histo[0]); b++) {
	vsx->vsx_queue_histo[t][b] +=
	cvsx->vsx_queue_histo[t][b];
	}
	vsx->vsx_active_queue[t] += cvsx->vsx_active_queue[t];
	vsx->vsx_pend_queue[t] += cvsx->vsx_pend_queue[t];

	for (b = 0; b < ARRAY_SIZE(vsx->vsx_ind_histo[0]); b++)
	vsx->vsx_ind_histo[t][b] += cvsx->vsx_ind_histo[t][b];

	for (b = 0; b < ARRAY_SIZE(vsx->vsx_agg_histo[0]); b++)
	vsx->vsx_agg_histo[t][b] += cvsx->vsx_agg_histo[t][b];
	}

	}

	boolean_t
	vdev_is_spacemap_addressable(vdev_t *vd)
	{
	if (spa_feature_is_active(vd->vdev_spa, SPA_FEATURE_SPACEMAP_V2))
	return (B_TRUE);

	/*
	* If double-word space map entries are not enabled we assume
	* 47 bits of the space map entry are dedicated to the entry's
	* offset (see SM_OFFSET_BITS in space_map.h). We then use that
	* to calculate the maximum address that can be described by a
	* space map entry for the given device.
	*/
	uint64_t shift = vd->vdev_ashift + SM_OFFSET_BITS;

	if (shift >= 63) /* detect potential overflow */
	return (B_TRUE);

	return (vd->vdev_asize < (1ULL << shift));
	}

	/*
	* Get statistics for the given vdev.
	*/
	static void
	vdev_get_stats_ex_impl(vdev_t vd, vdev_stat_t vs, vdev_stat_ex_t *vsx)
	{
	int t;
	/*
	* If we're getting stats on the root vdev, aggregate the I/O counts
	* over all top-level vdevs (i.e. the direct children of the root).
	*/
	if (!vd->vdev_ops->vdev_op_leaf) {
	if (vs) {
	memset(vs->vs_ops, 0, sizeof (vs->vs_ops));
	memset(vs->vs_bytes, 0, sizeof (vs->vs_bytes));
	}
	if (vsx)
	memset(vsx, 0, sizeof (*vsx));

	for (int c = 0; c < vd->vdev_children; c++) {
	vdev_t *cvd = vd->vdev_child[c];
	vdev_stat_t *cvs = &cvd->vdev_stat;
	vdev_stat_ex_t *cvsx = &cvd->vdev_stat_ex;

	vdev_get_stats_ex_impl(cvd, cvs, cvsx);
	if (vs)
	vdev_get_child_stat(cvd, vs, cvs);
	if (vsx)
	vdev_get_child_stat_ex(cvd, vsx, cvsx);
	}
	} else {
	/*
	* We're a leaf. Just copy our ZIO active queue stats in. The
	* other leaf stats are updated in vdev_stat_update().
	*/
	if (!vsx)
	return;

	memcpy(vsx, &vd->vdev_stat_ex, sizeof (vd->vdev_stat_ex));

	for (t = 0; t < ZIO_PRIORITY_NUM_QUEUEABLE; t++) {
	vsx->vsx_active_queue[t] = vd->vdev_queue.vq_cactive[t];
	vsx->vsx_pend_queue[t] = vdev_queue_class_length(vd, t);
	}
	}
	}

	void
	vdev_get_stats_ex(vdev_t vd, vdev_stat_t vs, vdev_stat_ex_t *vsx)
	{
	vdev_t *tvd = vd->vdev_top;
	mutex_enter(&vd->vdev_stat_lock);
	if (vs) {
	memcpy(vs, &vd->vdev_stat, sizeof (*vs));
	vs->vs_timestamp = gethrtime() - vs->vs_timestamp;
	vs->vs_state = vd->vdev_state;
	vs->vs_rsize = vdev_get_min_asize(vd);

	if (vd->vdev_ops->vdev_op_leaf) {
	vs->vs_pspace = vd->vdev_psize;
	vs->vs_rsize += VDEV_LABEL_START_SIZE +
	VDEV_LABEL_END_SIZE;
	/*
	* Report initializing progress. Since we don't
	* have the initializing locks held, this is only
	* an estimate (although a fairly accurate one).
	*/
	vs->vs_initialize_bytes_done =
	vd->vdev_initialize_bytes_done;
	vs->vs_initialize_bytes_est =
	vd->vdev_initialize_bytes_est;
	vs->vs_initialize_state = vd->vdev_initialize_state;
	vs->vs_initialize_action_time =
	vd->vdev_initialize_action_time;

	/*
	* Report manual TRIM progress. Since we don't have
	* the manual TRIM locks held, this is only an
	* estimate (although fairly accurate one).
	*/
	vs->vs_trim_notsup = !vd->vdev_has_trim;
	vs->vs_trim_bytes_done = vd->vdev_trim_bytes_done;
	vs->vs_trim_bytes_est = vd->vdev_trim_bytes_est;
	vs->vs_trim_state = vd->vdev_trim_state;
	vs->vs_trim_action_time = vd->vdev_trim_action_time;

	/* Set when there is a deferred resilver. */
	vs->vs_resilver_deferred = vd->vdev_resilver_deferred;
	}

	/*
	* Report expandable space on top-level, non-auxiliary devices
	* only. The expandable space is reported in terms of metaslab
	* sized units since that determines how much space the pool
	* can expand.
	*/
	if (vd->vdev_aux == NULL && tvd != NULL) {
	- vs->vs_esize = P2ALIGN(
	+ vs->vs_esize = P2ALIGN_TYPED(
	vd->vdev_max_asize - vd->vdev_asize,
	- 1ULL << tvd->vdev_ms_shift);
	+ 1ULL << tvd->vdev_ms_shift, uint64_t);
	}

	vs->vs_configured_ashift = vd->vdev_top != NULL
	? vd->vdev_top->vdev_ashift : vd->vdev_ashift;
	vs->vs_logical_ashift = vd->vdev_logical_ashift;
	if (vd->vdev_physical_ashift <= ASHIFT_MAX)
	vs->vs_physical_ashift = vd->vdev_physical_ashift;
	else
	vs->vs_physical_ashift = 0;

	/*
	* Report fragmentation and rebuild progress for top-level,
	* non-auxiliary, concrete devices.
	*/
	if (vd->vdev_aux == NULL && vd == vd->vdev_top &&
	vdev_is_concrete(vd)) {
	/*
	* The vdev fragmentation rating doesn't take into
	* account the embedded slog metaslab (vdev_log_mg).
	* Since it's only one metaslab, it would have a tiny
	* impact on the overall fragmentation.
	*/
	vs->vs_fragmentation = (vd->vdev_mg != NULL) ?
	vd->vdev_mg->mg_fragmentation : 0;
	}
	vs->vs_noalloc = MAX(vd->vdev_noalloc,
	tvd ? tvd->vdev_noalloc : 0);
	}

	vdev_get_stats_ex_impl(vd, vs, vsx);
	mutex_exit(&vd->vdev_stat_lock);
	}

	void
	vdev_get_stats(vdev_t vd, vdev_stat_t vs)
	{
	return (vdev_get_stats_ex(vd, vs, NULL));
	}

	void
	vdev_clear_stats(vdev_t *vd)
	{
	mutex_enter(&vd->vdev_stat_lock);
	vd->vdev_stat.vs_space = 0;
	vd->vdev_stat.vs_dspace = 0;
	vd->vdev_stat.vs_alloc = 0;
	mutex_exit(&vd->vdev_stat_lock);
	}

	void
	vdev_scan_stat_init(vdev_t *vd)
	{
	vdev_stat_t *vs = &vd->vdev_stat;

	for (int c = 0; c < vd->vdev_children; c++)
	vdev_scan_stat_init(vd->vdev_child[c]);

	mutex_enter(&vd->vdev_stat_lock);
	vs->vs_scan_processed = 0;
	mutex_exit(&vd->vdev_stat_lock);
	}

	void
	vdev_stat_update(zio_t *zio, uint64_t psize)
	{
	spa_t *spa = zio->io_spa;
	vdev_t *rvd = spa->spa_root_vdev;
	vdev_t *vd = zio->io_vd ? zio->io_vd : rvd;
	vdev_t *pvd;
	uint64_t txg = zio->io_txg;
	/* Suppress ASAN false positive */
	#ifdef __SANITIZE_ADDRESS__
	vdev_stat_t *vs = vd ? &vd->vdev_stat : NULL;
	vdev_stat_ex_t *vsx = vd ? &vd->vdev_stat_ex : NULL;
	#else
	vdev_stat_t *vs = &vd->vdev_stat;
	vdev_stat_ex_t *vsx = &vd->vdev_stat_ex;
	#endif
	zio_type_t type = zio->io_type;
	int flags = zio->io_flags;

	/*
	* If this i/o is a gang leader, it didn't do any actual work.
	*/
	if (zio->io_gang_tree)
	return;

	if (zio->io_error == 0) {
	/*
	* If this is a root i/o, don't count it -- we've already
	* counted the top-level vdevs, and vdev_get_stats() will
	* aggregate them when asked. This reduces contention on
	* the root vdev_stat_lock and implicitly handles blocks
	* that compress away to holes, for which there is no i/o.
	* (Holes never create vdev children, so all the counters
	* remain zero, which is what we want.)
	*
	* Note: this only applies to successful i/o (io_error == 0)
	* because unlike i/o counts, errors are not additive.
	* When reading a ditto block, for example, failure of
	* one top-level vdev does not imply a root-level error.
	*/
	if (vd == rvd)
	return;

	ASSERT(vd == zio->io_vd);

	if (flags & ZIO_FLAG_IO_BYPASS)
	return;

	mutex_enter(&vd->vdev_stat_lock);

	if (flags & ZIO_FLAG_IO_REPAIR) {
	/*
	* Repair is the result of a resilver issued by the
	* scan thread (spa_sync).
	*/
	if (flags & ZIO_FLAG_SCAN_THREAD) {
	dsl_scan_t *scn = spa->spa_dsl_pool->dp_scan;
	dsl_scan_phys_t *scn_phys = &scn->scn_phys;
	uint64_t *processed = &scn_phys->scn_processed;

	if (vd->vdev_ops->vdev_op_leaf)
	atomic_add_64(processed, psize);
	vs->vs_scan_processed += psize;
	}

	/*
	* Repair is the result of a rebuild issued by the
	* rebuild thread (vdev_rebuild_thread). To avoid
	* double counting repaired bytes the virtual dRAID
	* spare vdev is excluded from the processed bytes.
	*/
	if (zio->io_priority == ZIO_PRIORITY_REBUILD) {
	vdev_t *tvd = vd->vdev_top;
	vdev_rebuild_t *vr = &tvd->vdev_rebuild_config;
	vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
	uint64_t *rebuilt = &vrp->vrp_bytes_rebuilt;

	if (vd->vdev_ops->vdev_op_leaf &&
	vd->vdev_ops != &vdev_draid_spare_ops) {
	atomic_add_64(rebuilt, psize);
	}
	vs->vs_rebuild_processed += psize;
	}

	if (flags & ZIO_FLAG_SELF_HEAL)
	vs->vs_self_healed += psize;
	}

	/*
	* The bytes/ops/histograms are recorded at the leaf level and
	* aggregated into the higher level vdevs in vdev_get_stats().
	*/
	if (vd->vdev_ops->vdev_op_leaf &&
	(zio->io_priority < ZIO_PRIORITY_NUM_QUEUEABLE)) {
	zio_type_t vs_type = type;
	zio_priority_t priority = zio->io_priority;

	/*
	* TRIM ops and bytes are reported to user space as
	* ZIO_TYPE_FLUSH. This is done to preserve the
	* vdev_stat_t structure layout for user space.
	*/
	if (type == ZIO_TYPE_TRIM)
	vs_type = ZIO_TYPE_FLUSH;

	/*
	* Solely for the purposes of 'zpool iostat -lqrw'
	* reporting use the priority to categorize the IO.
	* Only the following are reported to user space:
	*
	* ZIO_PRIORITY_SYNC_READ,
	* ZIO_PRIORITY_SYNC_WRITE,
	* ZIO_PRIORITY_ASYNC_READ,
	* ZIO_PRIORITY_ASYNC_WRITE,
	* ZIO_PRIORITY_SCRUB,
	* ZIO_PRIORITY_TRIM,
	* ZIO_PRIORITY_REBUILD.
	*/
	if (priority == ZIO_PRIORITY_INITIALIZING) {
	ASSERT3U(type, ==, ZIO_TYPE_WRITE);
	priority = ZIO_PRIORITY_ASYNC_WRITE;
	} else if (priority == ZIO_PRIORITY_REMOVAL) {
	priority = ((type == ZIO_TYPE_WRITE) ?
	ZIO_PRIORITY_ASYNC_WRITE :
	ZIO_PRIORITY_ASYNC_READ);
	}

	vs->vs_ops[vs_type]++;
	vs->vs_bytes[vs_type] += psize;

	if (flags & ZIO_FLAG_DELEGATED) {
	vsx->vsx_agg_histo[priority]
	[RQ_HISTO(zio->io_size)]++;
	} else {
	vsx->vsx_ind_histo[priority]
	[RQ_HISTO(zio->io_size)]++;
	}

	if (zio->io_delta && zio->io_delay) {
	vsx->vsx_queue_histo[priority]
	[L_HISTO(zio->io_delta - zio->io_delay)]++;
	vsx->vsx_disk_histo[type]
	[L_HISTO(zio->io_delay)]++;
	vsx->vsx_total_histo[type]
	[L_HISTO(zio->io_delta)]++;
	}
	}

	mutex_exit(&vd->vdev_stat_lock);
	return;
	}

	if (flags & ZIO_FLAG_SPECULATIVE)
	return;

	/*
	* If this is an I/O error that is going to be retried, then ignore the
	* error. Otherwise, the user may interpret B_FAILFAST I/O errors as
	* hard errors, when in reality they can happen for any number of
	* innocuous reasons (bus resets, MPxIO link failure, etc).
	*/
	if (zio->io_error == EIO &&
	!(zio->io_flags & ZIO_FLAG_IO_RETRY))
	return;

	/*
	* Intent logs writes won't propagate their error to the root
	* I/O so don't mark these types of failures as pool-level
	* errors.
	*/
	if (zio->io_vd == NULL && (zio->io_flags & ZIO_FLAG_DONT_PROPAGATE))
	return;

	if (type == ZIO_TYPE_WRITE && txg != 0 &&
	(!(flags & ZIO_FLAG_IO_REPAIR) \|\|
	(flags & ZIO_FLAG_SCAN_THREAD) \|\|
	spa->spa_claiming)) {
	/*
	* This is either a normal write (not a repair), or it's
	* a repair induced by the scrub thread, or it's a repair
	* made by zil_claim() during spa_load() in the first txg.
	* In the normal case, we commit the DTL change in the same
	* txg as the block was born. In the scrub-induced repair
	* case, we know that scrubs run in first-pass syncing context,
	* so we commit the DTL change in spa_syncing_txg(spa).
	* In the zil_claim() case, we commit in spa_first_txg(spa).
	*
	* We currently do not make DTL entries for failed spontaneous
	* self-healing writes triggered by normal (non-scrubbing)
	* reads, because we have no transactional context in which to
	* do so -- and it's not clear that it'd be desirable anyway.
	*/
	if (vd->vdev_ops->vdev_op_leaf) {
	uint64_t commit_txg = txg;
	if (flags & ZIO_FLAG_SCAN_THREAD) {
	ASSERT(flags & ZIO_FLAG_IO_REPAIR);
	ASSERT(spa_sync_pass(spa) == 1);
	vdev_dtl_dirty(vd, DTL_SCRUB, txg, 1);
	commit_txg = spa_syncing_txg(spa);
	} else if (spa->spa_claiming) {
	ASSERT(flags & ZIO_FLAG_IO_REPAIR);
	commit_txg = spa_first_txg(spa);
	}
	ASSERT(commit_txg >= spa_syncing_txg(spa));
	if (vdev_dtl_contains(vd, DTL_MISSING, txg, 1))
	return;
	for (pvd = vd; pvd != rvd; pvd = pvd->vdev_parent)
	vdev_dtl_dirty(pvd, DTL_PARTIAL, txg, 1);
	vdev_dirty(vd->vdev_top, VDD_DTL, vd, commit_txg);
	}
	if (vd != rvd)
	vdev_dtl_dirty(vd, DTL_MISSING, txg, 1);
	}
	}

	int64_t
	vdev_deflated_space(vdev_t *vd, int64_t space)
	{
	ASSERT((space & (SPA_MINBLOCKSIZE-1)) == 0);
	ASSERT(vd->vdev_deflate_ratio != 0 \|\| vd->vdev_isl2cache);

	return ((space >> SPA_MINBLOCKSHIFT) * vd->vdev_deflate_ratio);
	}

	/*
	* Update the in-core space usage stats for this vdev, its metaslab class,
	* and the root vdev.
	*/
	void
	vdev_space_update(vdev_t *vd, int64_t alloc_delta, int64_t defer_delta,
	int64_t space_delta)
	{
	(void) defer_delta;
	int64_t dspace_delta;
	spa_t *spa = vd->vdev_spa;
	vdev_t *rvd = spa->spa_root_vdev;

	ASSERT(vd == vd->vdev_top);

	/*
	* Apply the inverse of the psize-to-asize (ie. RAID-Z) space-expansion
	* factor. We must calculate this here and not at the root vdev
	* because the root vdev's psize-to-asize is simply the max of its
	* children's, thus not accurate enough for us.
	*/
	dspace_delta = vdev_deflated_space(vd, space_delta);

	mutex_enter(&vd->vdev_stat_lock);
	/* ensure we won't underflow */
	if (alloc_delta < 0) {
	ASSERT3U(vd->vdev_stat.vs_alloc, >=, -alloc_delta);
	}

	vd->vdev_stat.vs_alloc += alloc_delta;
	vd->vdev_stat.vs_space += space_delta;
	vd->vdev_stat.vs_dspace += dspace_delta;
	mutex_exit(&vd->vdev_stat_lock);

	/* every class but log contributes to root space stats */
	if (vd->vdev_mg != NULL && !vd->vdev_islog) {
	ASSERT(!vd->vdev_isl2cache);
	mutex_enter(&rvd->vdev_stat_lock);
	rvd->vdev_stat.vs_alloc += alloc_delta;
	rvd->vdev_stat.vs_space += space_delta;
	rvd->vdev_stat.vs_dspace += dspace_delta;
	mutex_exit(&rvd->vdev_stat_lock);
	}
	/* Note: metaslab_class_space_update moved to metaslab_space_update */
	}

	/*
	* Mark a top-level vdev's config as dirty, placing it on the dirty list
	* so that it will be written out next time the vdev configuration is synced.
	* If the root vdev is specified (vdev_top == NULL), dirty all top-level vdevs.
	*/
	void
	vdev_config_dirty(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;
	vdev_t *rvd = spa->spa_root_vdev;
	int c;

	ASSERT(spa_writeable(spa));

	/*
	* If this is an aux vdev (as with l2cache and spare devices), then we
	* update the vdev config manually and set the sync flag.
	*/
	if (vd->vdev_aux != NULL) {
	spa_aux_vdev_t *sav = vd->vdev_aux;
	nvlist_t **aux;
	uint_t naux;

	for (c = 0; c < sav->sav_count; c++) {
	if (sav->sav_vdevs[c] == vd)
	break;
	}

	if (c == sav->sav_count) {
	/*
	* We're being removed. There's nothing more to do.
	*/
	ASSERT(sav->sav_sync == B_TRUE);
	return;
	}

	sav->sav_sync = B_TRUE;

	if (nvlist_lookup_nvlist_array(sav->sav_config,
	ZPOOL_CONFIG_L2CACHE, &aux, &naux) != 0) {
	VERIFY(nvlist_lookup_nvlist_array(sav->sav_config,
	ZPOOL_CONFIG_SPARES, &aux, &naux) == 0);
	}

	ASSERT(c < naux);

	/*
	* Setting the nvlist in the middle if the array is a little
	* sketchy, but it will work.
	*/
	nvlist_free(aux[c]);
	aux[c] = vdev_config_generate(spa, vd, B_TRUE, 0);

	return;
	}

	/*
	* The dirty list is protected by the SCL_CONFIG lock. The caller
	* must either hold SCL_CONFIG as writer, or must be the sync thread
	* (which holds SCL_CONFIG as reader). There's only one sync thread,
	* so this is sufficient to ensure mutual exclusion.
	*/
	ASSERT(spa_config_held(spa, SCL_CONFIG, RW_WRITER) \|\|
	(dsl_pool_sync_context(spa_get_dsl(spa)) &&
	spa_config_held(spa, SCL_CONFIG, RW_READER)));

	if (vd == rvd) {
	for (c = 0; c < rvd->vdev_children; c++)
	vdev_config_dirty(rvd->vdev_child[c]);
	} else {
	ASSERT(vd == vd->vdev_top);

	if (!list_link_active(&vd->vdev_config_dirty_node) &&
	vdev_is_concrete(vd)) {
	list_insert_head(&spa->spa_config_dirty_list, vd);
	}
	}
	}

	void
	vdev_config_clean(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;

	ASSERT(spa_config_held(spa, SCL_CONFIG, RW_WRITER) \|\|
	(dsl_pool_sync_context(spa_get_dsl(spa)) &&
	spa_config_held(spa, SCL_CONFIG, RW_READER)));

	ASSERT(list_link_active(&vd->vdev_config_dirty_node));
	list_remove(&spa->spa_config_dirty_list, vd);
	}

	/*
	* Mark a top-level vdev's state as dirty, so that the next pass of
	* spa_sync() can convert this into vdev_config_dirty(). We distinguish
	* the state changes from larger config changes because they require
	* much less locking, and are often needed for administrative actions.
	*/
	void
	vdev_state_dirty(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;

	ASSERT(spa_writeable(spa));
	ASSERT(vd == vd->vdev_top);

	/*
	* The state list is protected by the SCL_STATE lock. The caller
	* must either hold SCL_STATE as writer, or must be the sync thread
	* (which holds SCL_STATE as reader). There's only one sync thread,
	* so this is sufficient to ensure mutual exclusion.
	*/
	ASSERT(spa_config_held(spa, SCL_STATE, RW_WRITER) \|\|
	(dsl_pool_sync_context(spa_get_dsl(spa)) &&
	spa_config_held(spa, SCL_STATE, RW_READER)));

	if (!list_link_active(&vd->vdev_state_dirty_node) &&
	vdev_is_concrete(vd))
	list_insert_head(&spa->spa_state_dirty_list, vd);
	}

	void
	vdev_state_clean(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;

	ASSERT(spa_config_held(spa, SCL_STATE, RW_WRITER) \|\|
	(dsl_pool_sync_context(spa_get_dsl(spa)) &&
	spa_config_held(spa, SCL_STATE, RW_READER)));

	ASSERT(list_link_active(&vd->vdev_state_dirty_node));
	list_remove(&spa->spa_state_dirty_list, vd);
	}

	/*
	* Propagate vdev state up from children to parent.
	*/
	void
	vdev_propagate_state(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;
	vdev_t *rvd = spa->spa_root_vdev;
	int degraded = 0, faulted = 0;
	int corrupted = 0;
	vdev_t *child;

	if (vd->vdev_children > 0) {
	for (int c = 0; c < vd->vdev_children; c++) {
	child = vd->vdev_child[c];

	/*
	* Don't factor holes or indirect vdevs into the
	* decision.
	*/
	if (!vdev_is_concrete(child))
	continue;

	if (!vdev_readable(child) \|\|
	(!vdev_writeable(child) && spa_writeable(spa))) {
	/*
	* Root special: if there is a top-level log
	* device, treat the root vdev as if it were
	* degraded.
	*/
	if (child->vdev_islog && vd == rvd)
	degraded++;
	else
	faulted++;
	} else if (child->vdev_state <= VDEV_STATE_DEGRADED) {
	degraded++;
	}

	if (child->vdev_stat.vs_aux == VDEV_AUX_CORRUPT_DATA)
	corrupted++;
	}

	vd->vdev_ops->vdev_op_state_change(vd, faulted, degraded);

	/*
	* Root special: if there is a top-level vdev that cannot be
	* opened due to corrupted metadata, then propagate the root
	* vdev's aux state as 'corrupt' rather than 'insufficient
	* replicas'.
	*/
	if (corrupted && vd == rvd &&
	rvd->vdev_state == VDEV_STATE_CANT_OPEN)
	vdev_set_state(rvd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_CORRUPT_DATA);
	}

	if (vd->vdev_parent)
	vdev_propagate_state(vd->vdev_parent);
	}

	/*
	* Set a vdev's state. If this is during an open, we don't update the parent
	* state, because we're in the process of opening children depth-first.
	* Otherwise, we propagate the change to the parent.
	*
	* If this routine places a device in a faulted state, an appropriate ereport is
	* generated.
	*/
	void
	vdev_set_state(vdev_t *vd, boolean_t isopen, vdev_state_t state, vdev_aux_t aux)
	{
	uint64_t save_state;
	spa_t *spa = vd->vdev_spa;

	if (state == vd->vdev_state) {
	/*
	* Since vdev_offline() code path is already in an offline
	* state we can miss a statechange event to OFFLINE. Check
	* the previous state to catch this condition.
	*/
	if (vd->vdev_ops->vdev_op_leaf &&
	(state == VDEV_STATE_OFFLINE) &&
	(vd->vdev_prevstate >= VDEV_STATE_FAULTED)) {
	/* post an offline state change */
	zfs_post_state_change(spa, vd, vd->vdev_prevstate);
	}
	vd->vdev_stat.vs_aux = aux;
	return;
	}

	save_state = vd->vdev_state;

	vd->vdev_state = state;
	vd->vdev_stat.vs_aux = aux;

	/*
	* If we are setting the vdev state to anything but an open state, then
	* always close the underlying device unless the device has requested
	* a delayed close (i.e. we're about to remove or fault the device).
	* Otherwise, we keep accessible but invalid devices open forever.
	* We don't call vdev_close() itself, because that implies some extra
	* checks (offline, etc) that we don't want here. This is limited to
	* leaf devices, because otherwise closing the device will affect other
	* children.
	*/
	if (!vd->vdev_delayed_close && vdev_is_dead(vd) &&
	vd->vdev_ops->vdev_op_leaf)
	vd->vdev_ops->vdev_op_close(vd);

	if (vd->vdev_removed &&
	state == VDEV_STATE_CANT_OPEN &&
	(aux == VDEV_AUX_OPEN_FAILED \|\| vd->vdev_checkremove)) {
	/*
	* If the previous state is set to VDEV_STATE_REMOVED, then this
	* device was previously marked removed and someone attempted to
	* reopen it. If this failed due to a nonexistent device, then
	* keep the device in the REMOVED state. We also let this be if
	* it is one of our special test online cases, which is only
	* attempting to online the device and shouldn't generate an FMA
	* fault.
	*/
	vd->vdev_state = VDEV_STATE_REMOVED;
	vd->vdev_stat.vs_aux = VDEV_AUX_NONE;
	} else if (state == VDEV_STATE_REMOVED) {
	vd->vdev_removed = B_TRUE;
	} else if (state == VDEV_STATE_CANT_OPEN) {
	/*
	* If we fail to open a vdev during an import or recovery, we
	* mark it as "not available", which signifies that it was
	* never there to begin with. Failure to open such a device
	* is not considered an error.
	*/
	if ((spa_load_state(spa) == SPA_LOAD_IMPORT \|\|
	spa_load_state(spa) == SPA_LOAD_RECOVER) &&
	vd->vdev_ops->vdev_op_leaf)
	vd->vdev_not_present = 1;

	/*
	* Post the appropriate ereport. If the 'prevstate' field is
	* set to something other than VDEV_STATE_UNKNOWN, it indicates
	* that this is part of a vdev_reopen(). In this case, we don't
	* want to post the ereport if the device was already in the
	* CANT_OPEN state beforehand.
	*
	* If the 'checkremove' flag is set, then this is an attempt to
	* online the device in response to an insertion event. If we
	* hit this case, then we have detected an insertion event for a
	* faulted or offline device that wasn't in the removed state.
	* In this scenario, we don't post an ereport because we are
	* about to replace the device, or attempt an online with
	* vdev_forcefault, which will generate the fault for us.
	*/
	if ((vd->vdev_prevstate != state \|\| vd->vdev_forcefault) &&
	!vd->vdev_not_present && !vd->vdev_checkremove &&
	vd != spa->spa_root_vdev) {
	const char *class;

	switch (aux) {
	case VDEV_AUX_OPEN_FAILED:
	class = FM_EREPORT_ZFS_DEVICE_OPEN_FAILED;
	break;
	case VDEV_AUX_CORRUPT_DATA:
	class = FM_EREPORT_ZFS_DEVICE_CORRUPT_DATA;
	break;
	case VDEV_AUX_NO_REPLICAS:
	class = FM_EREPORT_ZFS_DEVICE_NO_REPLICAS;
	break;
	case VDEV_AUX_BAD_GUID_SUM:
	class = FM_EREPORT_ZFS_DEVICE_BAD_GUID_SUM;
	break;
	case VDEV_AUX_TOO_SMALL:
	class = FM_EREPORT_ZFS_DEVICE_TOO_SMALL;
	break;
	case VDEV_AUX_BAD_LABEL:
	class = FM_EREPORT_ZFS_DEVICE_BAD_LABEL;
	break;
	case VDEV_AUX_BAD_ASHIFT:
	class = FM_EREPORT_ZFS_DEVICE_BAD_ASHIFT;
	break;
	default:
	class = FM_EREPORT_ZFS_DEVICE_UNKNOWN;
	}

	(void) zfs_ereport_post(class, spa, vd, NULL, NULL,
	save_state);
	}

	/* Erase any notion of persistent removed state */
	vd->vdev_removed = B_FALSE;
	} else {
	vd->vdev_removed = B_FALSE;
	}

	/*
	* Notify ZED of any significant state-change on a leaf vdev.
	*
	*/
	if (vd->vdev_ops->vdev_op_leaf) {
	/* preserve original state from a vdev_reopen() */
	if ((vd->vdev_prevstate != VDEV_STATE_UNKNOWN) &&
	(vd->vdev_prevstate != vd->vdev_state) &&
	(save_state <= VDEV_STATE_CLOSED))
	save_state = vd->vdev_prevstate;

	/* filter out state change due to initial vdev_open */
	if (save_state > VDEV_STATE_CLOSED)
	zfs_post_state_change(spa, vd, save_state);
	}

	if (!isopen && vd->vdev_parent)
	vdev_propagate_state(vd->vdev_parent);
	}

	boolean_t
	vdev_children_are_offline(vdev_t *vd)
	{
	ASSERT(!vd->vdev_ops->vdev_op_leaf);

	for (uint64_t i = 0; i < vd->vdev_children; i++) {
	if (vd->vdev_child[i]->vdev_state != VDEV_STATE_OFFLINE)
	return (B_FALSE);
	}

	return (B_TRUE);
	}

	/*
	* Check the vdev configuration to ensure that it's capable of supporting
	* a root pool. We do not support partial configuration.
	*/
	boolean_t
	vdev_is_bootable(vdev_t *vd)
	{
	if (!vd->vdev_ops->vdev_op_leaf) {
	const char *vdev_type = vd->vdev_ops->vdev_op_type;

	if (strcmp(vdev_type, VDEV_TYPE_MISSING) == 0)
	return (B_FALSE);
	}

	for (int c = 0; c < vd->vdev_children; c++) {
	if (!vdev_is_bootable(vd->vdev_child[c]))
	return (B_FALSE);
	}
	return (B_TRUE);
	}

	boolean_t
	vdev_is_concrete(vdev_t *vd)
	{
	vdev_ops_t *ops = vd->vdev_ops;
	if (ops == &vdev_indirect_ops \|\| ops == &vdev_hole_ops \|\|
	ops == &vdev_missing_ops \|\| ops == &vdev_root_ops) {
	return (B_FALSE);
	} else {
	return (B_TRUE);
	}
	}

	/*
	* Determine if a log device has valid content. If the vdev was
	* removed or faulted in the MOS config then we know that
	* the content on the log device has already been written to the pool.
	*/
	boolean_t
	vdev_log_state_valid(vdev_t *vd)
	{
	if (vd->vdev_ops->vdev_op_leaf && !vd->vdev_faulted &&
	!vd->vdev_removed)
	return (B_TRUE);

	for (int c = 0; c < vd->vdev_children; c++)
	if (vdev_log_state_valid(vd->vdev_child[c]))
	return (B_TRUE);

	return (B_FALSE);
	}

	/*
	* Expand a vdev if possible.
	*/
	void
	vdev_expand(vdev_t *vd, uint64_t txg)
	{
	ASSERT(vd->vdev_top == vd);
	ASSERT(spa_config_held(vd->vdev_spa, SCL_ALL, RW_WRITER) == SCL_ALL);
	ASSERT(vdev_is_concrete(vd));

	vdev_set_deflate_ratio(vd);

	if ((vd->vdev_spa->spa_raidz_expand == NULL \|\|
	vd->vdev_spa->spa_raidz_expand->vre_vdev_id != vd->vdev_id) &&
	(vd->vdev_asize >> vd->vdev_ms_shift) > vd->vdev_ms_count &&
	vdev_is_concrete(vd)) {
	vdev_metaslab_group_create(vd);
	VERIFY(vdev_metaslab_init(vd, txg) == 0);
	vdev_config_dirty(vd);
	}
	}

	/*
	* Split a vdev.
	*/
	void
	vdev_split(vdev_t *vd)
	{
	vdev_t cvd, pvd = vd->vdev_parent;

	VERIFY3U(pvd->vdev_children, >, 1);

	vdev_remove_child(pvd, vd);
	vdev_compact_children(pvd);

	ASSERT3P(pvd->vdev_child, !=, NULL);

	cvd = pvd->vdev_child[0];
	if (pvd->vdev_children == 1) {
	vdev_remove_parent(cvd);
	cvd->vdev_splitting = B_TRUE;
	}
	vdev_propagate_state(cvd);
	}

	void
	vdev_deadman(vdev_t vd, const char tag)
	{
	for (int c = 0; c < vd->vdev_children; c++) {
	vdev_t *cvd = vd->vdev_child[c];

	vdev_deadman(cvd, tag);
	}

	if (vd->vdev_ops->vdev_op_leaf) {
	vdev_queue_t *vq = &vd->vdev_queue;

	mutex_enter(&vq->vq_lock);
	if (vq->vq_active > 0) {
	spa_t *spa = vd->vdev_spa;
	zio_t *fio;
	uint64_t delta;

	zfs_dbgmsg("slow vdev: %s has %u active IOs",
	vd->vdev_path, vq->vq_active);

	/*
	* Look at the head of all the pending queues,
	* if any I/O has been outstanding for longer than
	* the spa_deadman_synctime invoke the deadman logic.
	*/
	fio = list_head(&vq->vq_active_list);
	delta = gethrtime() - fio->io_timestamp;
	if (delta > spa_deadman_synctime(spa))
	zio_deadman(fio, tag);
	}
	mutex_exit(&vq->vq_lock);
	}
	}

	void
	vdev_defer_resilver(vdev_t *vd)
	{
	ASSERT(vd->vdev_ops->vdev_op_leaf);

	vd->vdev_resilver_deferred = B_TRUE;
	vd->vdev_spa->spa_resilver_deferred = B_TRUE;
	}

	/*
	* Clears the resilver deferred flag on all leaf devs under vd. Returns
	* B_TRUE if we have devices that need to be resilvered and are available to
	* accept resilver I/Os.
	*/
	boolean_t
	vdev_clear_resilver_deferred(vdev_t vd, dmu_tx_t tx)
	{
	boolean_t resilver_needed = B_FALSE;
	spa_t *spa = vd->vdev_spa;

	for (int c = 0; c < vd->vdev_children; c++) {
	vdev_t *cvd = vd->vdev_child[c];
	resilver_needed \|= vdev_clear_resilver_deferred(cvd, tx);
	}

	if (vd == spa->spa_root_vdev &&
	spa_feature_is_active(spa, SPA_FEATURE_RESILVER_DEFER)) {
	spa_feature_decr(spa, SPA_FEATURE_RESILVER_DEFER, tx);
	vdev_config_dirty(vd);
	spa->spa_resilver_deferred = B_FALSE;
	return (resilver_needed);
	}

	if (!vdev_is_concrete(vd) \|\| vd->vdev_aux \|\|
	!vd->vdev_ops->vdev_op_leaf)
	return (resilver_needed);

	vd->vdev_resilver_deferred = B_FALSE;

	return (!vdev_is_dead(vd) && !vd->vdev_offline &&
	vdev_resilver_needed(vd, NULL, NULL));
	}

	boolean_t
	vdev_xlate_is_empty(range_seg64_t *rs)
	{
	return (rs->rs_start == rs->rs_end);
	}

	/*
	* Translate a logical range to the first contiguous physical range for the
	* specified vdev_t. This function is initially called with a leaf vdev and
	* will walk each parent vdev until it reaches a top-level vdev. Once the
	* top-level is reached the physical range is initialized and the recursive
	* function begins to unwind. As it unwinds it calls the parent's vdev
	* specific translation function to do the real conversion.
	*/
	void
	vdev_xlate(vdev_t vd, const range_seg64_t logical_rs,
	range_seg64_t physical_rs, range_seg64_t remain_rs)
	{
	/*
	* Walk up the vdev tree
	*/
	if (vd != vd->vdev_top) {
	vdev_xlate(vd->vdev_parent, logical_rs, physical_rs,
	remain_rs);
	} else {
	/*
	* We've reached the top-level vdev, initialize the physical
	* range to the logical range and set an empty remaining
	* range then start to unwind.
	*/
	physical_rs->rs_start = logical_rs->rs_start;
	physical_rs->rs_end = logical_rs->rs_end;

	remain_rs->rs_start = logical_rs->rs_start;
	remain_rs->rs_end = logical_rs->rs_start;

	return;
	}

	vdev_t *pvd = vd->vdev_parent;
	ASSERT3P(pvd, !=, NULL);
	ASSERT3P(pvd->vdev_ops->vdev_op_xlate, !=, NULL);

	/*
	* As this recursive function unwinds, translate the logical
	* range into its physical and any remaining components by calling
	* the vdev specific translate function.
	*/
	range_seg64_t intermediate = { 0 };
	pvd->vdev_ops->vdev_op_xlate(vd, physical_rs, &intermediate, remain_rs);

	physical_rs->rs_start = intermediate.rs_start;
	physical_rs->rs_end = intermediate.rs_end;
	}

	void
	vdev_xlate_walk(vdev_t vd, const range_seg64_t logical_rs,
	vdev_xlate_func_t func, void arg)
	{
	range_seg64_t iter_rs = *logical_rs;
	range_seg64_t physical_rs;
	range_seg64_t remain_rs;

	while (!vdev_xlate_is_empty(&iter_rs)) {

	vdev_xlate(vd, &iter_rs, &physical_rs, &remain_rs);

	/*
	* With raidz and dRAID, it's possible that the logical range
	* does not live on this leaf vdev. Only when there is a non-
	* zero physical size call the provided function.
	*/
	if (!vdev_xlate_is_empty(&physical_rs))
	func(arg, &physical_rs);

	iter_rs = remain_rs;
	}
	}

	static char *
	vdev_name(vdev_t vd, char buf, int buflen)
	{
	if (vd->vdev_path == NULL) {
	if (strcmp(vd->vdev_ops->vdev_op_type, "root") == 0) {
	strlcpy(buf, vd->vdev_spa->spa_name, buflen);
	} else if (!vd->vdev_ops->vdev_op_leaf) {
	snprintf(buf, buflen, "%s-%llu",
	vd->vdev_ops->vdev_op_type,
	(u_longlong_t)vd->vdev_id);
	}
	} else {
	strlcpy(buf, vd->vdev_path, buflen);
	}
	return (buf);
	}

	/*
	* Look at the vdev tree and determine whether any devices are currently being
	* replaced.
	*/
	boolean_t
	vdev_replace_in_progress(vdev_t *vdev)
	{
	ASSERT(spa_config_held(vdev->vdev_spa, SCL_ALL, RW_READER) != 0);

	if (vdev->vdev_ops == &vdev_replacing_ops)
	return (B_TRUE);

	/*
	* A 'spare' vdev indicates that we have a replace in progress, unless
	* it has exactly two children, and the second, the hot spare, has
	* finished being resilvered.
	*/
	if (vdev->vdev_ops == &vdev_spare_ops && (vdev->vdev_children > 2 \|\|
	!vdev_dtl_empty(vdev->vdev_child[1], DTL_MISSING)))
	return (B_TRUE);

	for (int i = 0; i < vdev->vdev_children; i++) {
	if (vdev_replace_in_progress(vdev->vdev_child[i]))
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	/*
	* Add a (source=src, propname=propval) list to an nvlist.
	*/
	static void
	vdev_prop_add_list(nvlist_t nvl, const char propname, const char *strval,
	uint64_t intval, zprop_source_t src)
	{
	nvlist_t *propval;

	propval = fnvlist_alloc();
	fnvlist_add_uint64(propval, ZPROP_SOURCE, src);

	if (strval != NULL)
	fnvlist_add_string(propval, ZPROP_VALUE, strval);
	else
	fnvlist_add_uint64(propval, ZPROP_VALUE, intval);

	fnvlist_add_nvlist(nvl, propname, propval);
	nvlist_free(propval);
	}

	static void
	vdev_props_set_sync(void arg, dmu_tx_t tx)
	{
	vdev_t *vd;
	nvlist_t *nvp = arg;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	objset_t *mos = spa->spa_meta_objset;
	nvpair_t *elem = NULL;
	uint64_t vdev_guid;
	uint64_t objid;
	nvlist_t *nvprops;

	vdev_guid = fnvlist_lookup_uint64(nvp, ZPOOL_VDEV_PROPS_SET_VDEV);
	nvprops = fnvlist_lookup_nvlist(nvp, ZPOOL_VDEV_PROPS_SET_PROPS);
	vd = spa_lookup_by_guid(spa, vdev_guid, B_TRUE);

	/* this vdev could get removed while waiting for this sync task */
	if (vd == NULL)
	return;

	/*
	* Set vdev property values in the vdev props mos object.
	*/
	if (vd->vdev_root_zap != 0) {
	objid = vd->vdev_root_zap;
	} else if (vd->vdev_top_zap != 0) {
	objid = vd->vdev_top_zap;
	} else if (vd->vdev_leaf_zap != 0) {
	objid = vd->vdev_leaf_zap;
	} else {
	panic("unexpected vdev type");
	}

	mutex_enter(&spa->spa_props_lock);

	while ((elem = nvlist_next_nvpair(nvprops, elem)) != NULL) {
	uint64_t intval;
	const char *strval;
	vdev_prop_t prop;
	const char *propname = nvpair_name(elem);
	zprop_type_t proptype;

	switch (prop = vdev_name_to_prop(propname)) {
	case VDEV_PROP_USERPROP:
	if (vdev_prop_user(propname)) {
	strval = fnvpair_value_string(elem);
	if (strlen(strval) == 0) {
	/* remove the property if value == "" */
	(void) zap_remove(mos, objid, propname,
	tx);
	} else {
	VERIFY0(zap_update(mos, objid, propname,
	1, strlen(strval) + 1, strval, tx));
	}
	spa_history_log_internal(spa, "vdev set", tx,
	"vdev_guid=%llu: %s=%s",
	(u_longlong_t)vdev_guid, nvpair_name(elem),
	strval);
	}
	break;
	default:
	/* normalize the property name */
	propname = vdev_prop_to_name(prop);
	proptype = vdev_prop_get_type(prop);

	if (nvpair_type(elem) == DATA_TYPE_STRING) {
	ASSERT(proptype == PROP_TYPE_STRING);
	strval = fnvpair_value_string(elem);
	VERIFY0(zap_update(mos, objid, propname,
	1, strlen(strval) + 1, strval, tx));
	spa_history_log_internal(spa, "vdev set", tx,
	"vdev_guid=%llu: %s=%s",
	(u_longlong_t)vdev_guid, nvpair_name(elem),
	strval);
	} else if (nvpair_type(elem) == DATA_TYPE_UINT64) {
	intval = fnvpair_value_uint64(elem);

	if (proptype == PROP_TYPE_INDEX) {
	const char *unused;
	VERIFY0(vdev_prop_index_to_string(
	prop, intval, &unused));
	}
	VERIFY0(zap_update(mos, objid, propname,
	sizeof (uint64_t), 1, &intval, tx));
	spa_history_log_internal(spa, "vdev set", tx,
	"vdev_guid=%llu: %s=%lld",
	(u_longlong_t)vdev_guid,
	nvpair_name(elem), (longlong_t)intval);
	} else {
	panic("invalid vdev property type %u",
	nvpair_type(elem));
	}
	}

	}

	mutex_exit(&spa->spa_props_lock);
	}

	int
	vdev_prop_set(vdev_t vd, nvlist_t innvl, nvlist_t *outnvl)
	{
	spa_t *spa = vd->vdev_spa;
	nvpair_t *elem = NULL;
	uint64_t vdev_guid;
	nvlist_t *nvprops;
	int error = 0;

	ASSERT(vd != NULL);

	/* Check that vdev has a zap we can use */
	if (vd->vdev_root_zap == 0 &&
	vd->vdev_top_zap == 0 &&
	vd->vdev_leaf_zap == 0)
	return (SET_ERROR(EINVAL));

	if (nvlist_lookup_uint64(innvl, ZPOOL_VDEV_PROPS_SET_VDEV,
	&vdev_guid) != 0)
	return (SET_ERROR(EINVAL));

	if (nvlist_lookup_nvlist(innvl, ZPOOL_VDEV_PROPS_SET_PROPS,
	&nvprops) != 0)
	return (SET_ERROR(EINVAL));

	if ((vd = spa_lookup_by_guid(spa, vdev_guid, B_TRUE)) == NULL)
	return (SET_ERROR(EINVAL));

	while ((elem = nvlist_next_nvpair(nvprops, elem)) != NULL) {
	const char *propname = nvpair_name(elem);
	vdev_prop_t prop = vdev_name_to_prop(propname);
	uint64_t intval = 0;
	const char *strval = NULL;

	if (prop == VDEV_PROP_USERPROP && !vdev_prop_user(propname)) {
	error = EINVAL;
	goto end;
	}

	if (vdev_prop_readonly(prop)) {
	error = EROFS;
	goto end;
	}

	/* Special Processing */
	switch (prop) {
	case VDEV_PROP_PATH:
	if (vd->vdev_path == NULL) {
	error = EROFS;
	break;
	}
	if (nvpair_value_string(elem, &strval) != 0) {
	error = EINVAL;
	break;
	}
	/* New path must start with /dev/ */
	if (strncmp(strval, "/dev/", 5)) {
	error = EINVAL;
	break;
	}
	error = spa_vdev_setpath(spa, vdev_guid, strval);
	break;
	case VDEV_PROP_ALLOCATING:
	if (nvpair_value_uint64(elem, &intval) != 0) {
	error = EINVAL;
	break;
	}
	if (intval != vd->vdev_noalloc)
	break;
	if (intval == 0)
	error = spa_vdev_noalloc(spa, vdev_guid);
	else
	error = spa_vdev_alloc(spa, vdev_guid);
	break;
	case VDEV_PROP_FAILFAST:
	if (nvpair_value_uint64(elem, &intval) != 0) {
	error = EINVAL;
	break;
	}
	vd->vdev_failfast = intval & 1;
	break;
	case VDEV_PROP_CHECKSUM_N:
	if (nvpair_value_uint64(elem, &intval) != 0) {
	error = EINVAL;
	break;
	}
	vd->vdev_checksum_n = intval;
	break;
	case VDEV_PROP_CHECKSUM_T:
	if (nvpair_value_uint64(elem, &intval) != 0) {
	error = EINVAL;
	break;
	}
	vd->vdev_checksum_t = intval;
	break;
	case VDEV_PROP_IO_N:
	if (nvpair_value_uint64(elem, &intval) != 0) {
	error = EINVAL;
	break;
	}
	vd->vdev_io_n = intval;
	break;
	case VDEV_PROP_IO_T:
	if (nvpair_value_uint64(elem, &intval) != 0) {
	error = EINVAL;
	break;
	}
	vd->vdev_io_t = intval;
	break;
	case VDEV_PROP_SLOW_IO_N:
	if (nvpair_value_uint64(elem, &intval) != 0) {
	error = EINVAL;
	break;
	}
	vd->vdev_slow_io_n = intval;
	break;
	case VDEV_PROP_SLOW_IO_T:
	if (nvpair_value_uint64(elem, &intval) != 0) {
	error = EINVAL;
	break;
	}
	vd->vdev_slow_io_t = intval;
	break;
	default:
	/* Most processing is done in vdev_props_set_sync */
	break;
	}
	end:
	if (error != 0) {
	intval = error;
	vdev_prop_add_list(outnvl, propname, strval, intval, 0);
	return (error);
	}
	}

	return (dsl_sync_task(spa->spa_name, NULL, vdev_props_set_sync,
	innvl, 6, ZFS_SPACE_CHECK_EXTRA_RESERVED));
	}

	int
	vdev_prop_get(vdev_t vd, nvlist_t innvl, nvlist_t *outnvl)
	{
	spa_t *spa = vd->vdev_spa;
	objset_t *mos = spa->spa_meta_objset;
	int err = 0;
	uint64_t objid;
	uint64_t vdev_guid;
	nvpair_t *elem = NULL;
	nvlist_t *nvprops = NULL;
	uint64_t intval = 0;
	char *strval = NULL;
	const char *propname = NULL;
	vdev_prop_t prop;

	ASSERT(vd != NULL);
	ASSERT(mos != NULL);

	if (nvlist_lookup_uint64(innvl, ZPOOL_VDEV_PROPS_GET_VDEV,
	&vdev_guid) != 0)
	return (SET_ERROR(EINVAL));

	nvlist_lookup_nvlist(innvl, ZPOOL_VDEV_PROPS_GET_PROPS, &nvprops);

	if (vd->vdev_root_zap != 0) {
	objid = vd->vdev_root_zap;
	} else if (vd->vdev_top_zap != 0) {
	objid = vd->vdev_top_zap;
	} else if (vd->vdev_leaf_zap != 0) {
	objid = vd->vdev_leaf_zap;
	} else {
	return (SET_ERROR(EINVAL));
	}
	ASSERT(objid != 0);

	mutex_enter(&spa->spa_props_lock);

	if (nvprops != NULL) {
	char namebuf[64] = { 0 };

	while ((elem = nvlist_next_nvpair(nvprops, elem)) != NULL) {
	intval = 0;
	strval = NULL;
	propname = nvpair_name(elem);
	prop = vdev_name_to_prop(propname);
	zprop_source_t src = ZPROP_SRC_DEFAULT;
	uint64_t integer_size, num_integers;

	switch (prop) {
	/* Special Read-only Properties */
	case VDEV_PROP_NAME:
	strval = vdev_name(vd, namebuf,
	sizeof (namebuf));
	if (strval == NULL)
	continue;
	vdev_prop_add_list(outnvl, propname, strval, 0,
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_CAPACITY:
	/* percent used */
	intval = (vd->vdev_stat.vs_dspace == 0) ? 0 :
	(vd->vdev_stat.vs_alloc * 100 /
	vd->vdev_stat.vs_dspace);
	vdev_prop_add_list(outnvl, propname, NULL,
	intval, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_STATE:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_state, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_GUID:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_guid, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_ASIZE:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_asize, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_PSIZE:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_psize, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_ASHIFT:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_ashift, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_SIZE:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_dspace, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_FREE:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_dspace -
	vd->vdev_stat.vs_alloc, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_ALLOCATED:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_alloc, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_EXPANDSZ:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_esize, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_FRAGMENTATION:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_fragmentation,
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_PARITY:
	vdev_prop_add_list(outnvl, propname, NULL,
	vdev_get_nparity(vd), ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_PATH:
	if (vd->vdev_path == NULL)
	continue;
	vdev_prop_add_list(outnvl, propname,
	vd->vdev_path, 0, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_DEVID:
	if (vd->vdev_devid == NULL)
	continue;
	vdev_prop_add_list(outnvl, propname,
	vd->vdev_devid, 0, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_PHYS_PATH:
	if (vd->vdev_physpath == NULL)
	continue;
	vdev_prop_add_list(outnvl, propname,
	vd->vdev_physpath, 0, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_ENC_PATH:
	if (vd->vdev_enc_sysfs_path == NULL)
	continue;
	vdev_prop_add_list(outnvl, propname,
	vd->vdev_enc_sysfs_path, 0, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_FRU:
	if (vd->vdev_fru == NULL)
	continue;
	vdev_prop_add_list(outnvl, propname,
	vd->vdev_fru, 0, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_PARENT:
	if (vd->vdev_parent != NULL) {
	strval = vdev_name(vd->vdev_parent,
	namebuf, sizeof (namebuf));
	vdev_prop_add_list(outnvl, propname,
	strval, 0, ZPROP_SRC_NONE);
	}
	continue;
	case VDEV_PROP_CHILDREN:
	if (vd->vdev_children > 0)
	strval = kmem_zalloc(ZAP_MAXVALUELEN,
	KM_SLEEP);
	for (uint64_t i = 0; i < vd->vdev_children;
	i++) {
	const char *vname;

	vname = vdev_name(vd->vdev_child[i],
	namebuf, sizeof (namebuf));
	if (vname == NULL)
	vname = "(unknown)";
	if (strlen(strval) > 0)
	strlcat(strval, ",",
	ZAP_MAXVALUELEN);
	strlcat(strval, vname, ZAP_MAXVALUELEN);
	}
	if (strval != NULL) {
	vdev_prop_add_list(outnvl, propname,
	strval, 0, ZPROP_SRC_NONE);
	kmem_free(strval, ZAP_MAXVALUELEN);
	}
	continue;
	case VDEV_PROP_NUMCHILDREN:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_children, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_READ_ERRORS:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_read_errors,
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_WRITE_ERRORS:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_write_errors,
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_CHECKSUM_ERRORS:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_checksum_errors,
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_INITIALIZE_ERRORS:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_initialize_errors,
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_OPS_NULL:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_ops[ZIO_TYPE_NULL],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_OPS_READ:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_ops[ZIO_TYPE_READ],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_OPS_WRITE:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_ops[ZIO_TYPE_WRITE],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_OPS_FREE:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_ops[ZIO_TYPE_FREE],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_OPS_CLAIM:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_ops[ZIO_TYPE_CLAIM],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_OPS_TRIM:
	/*
	* TRIM ops and bytes are reported to user
	* space as ZIO_TYPE_FLUSH. This is done to
	* preserve the vdev_stat_t structure layout
	* for user space.
	*/
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_ops[ZIO_TYPE_FLUSH],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_BYTES_NULL:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_bytes[ZIO_TYPE_NULL],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_BYTES_READ:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_bytes[ZIO_TYPE_READ],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_BYTES_WRITE:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_bytes[ZIO_TYPE_WRITE],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_BYTES_FREE:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_bytes[ZIO_TYPE_FREE],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_BYTES_CLAIM:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_bytes[ZIO_TYPE_CLAIM],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_BYTES_TRIM:
	/*
	* TRIM ops and bytes are reported to user
	* space as ZIO_TYPE_FLUSH. This is done to
	* preserve the vdev_stat_t structure layout
	* for user space.
	*/
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_stat.vs_bytes[ZIO_TYPE_FLUSH],
	ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_REMOVING:
	vdev_prop_add_list(outnvl, propname, NULL,
	vd->vdev_removing, ZPROP_SRC_NONE);
	continue;
	case VDEV_PROP_RAIDZ_EXPANDING:
	/* Only expose this for raidz */
	if (vd->vdev_ops == &vdev_raidz_ops) {
	vdev_prop_add_list(outnvl, propname,
	NULL, vd->vdev_rz_expanding,
	ZPROP_SRC_NONE);
	}
	continue;
	/* Numeric Properites */
	case VDEV_PROP_ALLOCATING:
	/* Leaf vdevs cannot have this property */
	if (vd->vdev_mg == NULL &&
	vd->vdev_top != NULL) {
	src = ZPROP_SRC_NONE;
	intval = ZPROP_BOOLEAN_NA;
	} else {
	err = vdev_prop_get_int(vd, prop,
	&intval);
	if (err && err != ENOENT)
	break;

	if (intval ==
	vdev_prop_default_numeric(prop))
	src = ZPROP_SRC_DEFAULT;
	else
	src = ZPROP_SRC_LOCAL;
	}

	vdev_prop_add_list(outnvl, propname, NULL,
	intval, src);
	break;
	case VDEV_PROP_FAILFAST:
	src = ZPROP_SRC_LOCAL;
	strval = NULL;

	err = zap_lookup(mos, objid, nvpair_name(elem),
	sizeof (uint64_t), 1, &intval);
	if (err == ENOENT) {
	intval = vdev_prop_default_numeric(
	prop);
	err = 0;
	} else if (err) {
	break;
	}
	if (intval == vdev_prop_default_numeric(prop))
	src = ZPROP_SRC_DEFAULT;

	vdev_prop_add_list(outnvl, propname, strval,
	intval, src);
	break;
	case VDEV_PROP_CHECKSUM_N:
	case VDEV_PROP_CHECKSUM_T:
	case VDEV_PROP_IO_N:
	case VDEV_PROP_IO_T:
	case VDEV_PROP_SLOW_IO_N:
	case VDEV_PROP_SLOW_IO_T:
	err = vdev_prop_get_int(vd, prop, &intval);
	if (err && err != ENOENT)
	break;

	if (intval == vdev_prop_default_numeric(prop))
	src = ZPROP_SRC_DEFAULT;
	else
	src = ZPROP_SRC_LOCAL;

	vdev_prop_add_list(outnvl, propname, NULL,
	intval, src);
	break;
	/* Text Properties */
	case VDEV_PROP_COMMENT:
	/* Exists in the ZAP below */
	/* FALLTHRU */
	case VDEV_PROP_USERPROP:
	/* User Properites */
	src = ZPROP_SRC_LOCAL;

	err = zap_length(mos, objid, nvpair_name(elem),
	&integer_size, &num_integers);
	if (err)
	break;

	switch (integer_size) {
	case 8:
	/* User properties cannot be integers */
	err = EINVAL;
	break;
	case 1:
	/* string property */
	strval = kmem_alloc(num_integers,
	KM_SLEEP);
	err = zap_lookup(mos, objid,
	nvpair_name(elem), 1,
	num_integers, strval);
	if (err) {
	kmem_free(strval,
	num_integers);
	break;
	}
	vdev_prop_add_list(outnvl, propname,
	strval, 0, src);
	kmem_free(strval, num_integers);
	break;
	}
	break;
	default:
	err = ENOENT;
	break;
	}
	if (err)
	break;
	}
	} else {
	/*
	* Get all properties from the MOS vdev property object.
	*/
	zap_cursor_t zc;
	zap_attribute_t za;
	for (zap_cursor_init(&zc, mos, objid);
	(err = zap_cursor_retrieve(&zc, &za)) == 0;
	zap_cursor_advance(&zc)) {
	intval = 0;
	strval = NULL;
	zprop_source_t src = ZPROP_SRC_DEFAULT;
	propname = za.za_name;

	switch (za.za_integer_length) {
	case 8:
	/* We do not allow integer user properties */
	/* This is likely an internal value */
	break;
	case 1:
	/* string property */
	strval = kmem_alloc(za.za_num_integers,
	KM_SLEEP);
	err = zap_lookup(mos, objid, za.za_name, 1,
	za.za_num_integers, strval);
	if (err) {
	kmem_free(strval, za.za_num_integers);
	break;
	}
	vdev_prop_add_list(outnvl, propname, strval, 0,
	src);
	kmem_free(strval, za.za_num_integers);
	break;

	default:
	break;
	}
	}
	zap_cursor_fini(&zc);
	}

	mutex_exit(&spa->spa_props_lock);
	if (err && err != ENOENT) {
	return (err);
	}

	return (0);
	}

	EXPORT_SYMBOL(vdev_fault);
	EXPORT_SYMBOL(vdev_degrade);
	EXPORT_SYMBOL(vdev_online);
	EXPORT_SYMBOL(vdev_offline);
	EXPORT_SYMBOL(vdev_clear);

	ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, default_ms_count, UINT, ZMOD_RW,
	"Target number of metaslabs per top-level vdev");

	ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, default_ms_shift, UINT, ZMOD_RW,
	"Default lower limit for metaslab size");

	ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, max_ms_shift, UINT, ZMOD_RW,
	"Default upper limit for metaslab size");

	ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, min_ms_count, UINT, ZMOD_RW,
	"Minimum number of metaslabs per top-level vdev");

	ZFS_MODULE_PARAM(zfs_vdev, zfs_vdev_, ms_count_limit, UINT, ZMOD_RW,
	"Practical upper limit of total metaslabs per top-level vdev");

	ZFS_MODULE_PARAM(zfs, zfs_, slow_io_events_per_second, UINT, ZMOD_RW,
	"Rate limit slow IO (delay) events to this many per second");

	+ZFS_MODULE_PARAM(zfs, zfs_, deadman_events_per_second, UINT, ZMOD_RW,
	+ "Rate limit hung IO (deadman) events to this many per second");
	+
	/* BEGIN CSTYLED */
	ZFS_MODULE_PARAM(zfs, zfs_, checksum_events_per_second, UINT, ZMOD_RW,
	"Rate limit checksum events to this many checksum errors per second "
	"(do not set below ZED threshold).");
	/* END CSTYLED */

	ZFS_MODULE_PARAM(zfs, zfs_, scan_ignore_errors, INT, ZMOD_RW,
	"Ignore errors during resilver/scrub");

	ZFS_MODULE_PARAM(zfs_vdev, vdev_, validate_skip, INT, ZMOD_RW,
	"Bypass vdev_validate()");

	ZFS_MODULE_PARAM(zfs, zfs_, nocacheflush, INT, ZMOD_RW,
	"Disable cache flushes");

	ZFS_MODULE_PARAM(zfs, zfs_, embedded_slog_min_ms, UINT, ZMOD_RW,
	"Minimum number of metaslabs required to dedicate one for log blocks");

	/* BEGIN CSTYLED */
	ZFS_MODULE_PARAM_CALL(zfs_vdev, zfs_vdev_, min_auto_ashift,
	param_set_min_auto_ashift, param_get_uint, ZMOD_RW,
	"Minimum ashift used when creating new top-level vdevs");

	ZFS_MODULE_PARAM_CALL(zfs_vdev, zfs_vdev_, max_auto_ashift,
	param_set_max_auto_ashift, param_get_uint, ZMOD_RW,
	"Maximum ashift used when optimizing for logical -> physical sector "
	"size on new top-level vdevs");
	/* END CSTYLED */
	diff --git a/sys/contrib/openzfs/module/zfs/vdev_initialize.c b/sys/contrib/openzfs/module/zfs/vdev_initialize.c
	index c5e16af16692..0a7323f58df2 100644
	--- a/sys/contrib/openzfs/module/zfs/vdev_initialize.c
	+++ b/sys/contrib/openzfs/module/zfs/vdev_initialize.c
	@@ -1,829 +1,832 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2016, 2024 by Delphix. All rights reserved.
	*/

	#include <sys/spa.h>
	#include <sys/spa_impl.h>
	#include <sys/txg.h>
	#include <sys/vdev_impl.h>
	#include <sys/metaslab_impl.h>
	#include <sys/dsl_synctask.h>
	#include <sys/zap.h>
	#include <sys/dmu_tx.h>
	#include <sys/vdev_initialize.h>

	/*
	* Value that is written to disk during initialization.
	*/
	static uint64_t zfs_initialize_value = 0xdeadbeefdeadbeeeULL;

	/* maximum number of I/Os outstanding per leaf vdev */
	static const int zfs_initialize_limit = 1;

	/* size of initializing writes; default 1MiB, see zfs_remove_max_segment */
	static uint64_t zfs_initialize_chunk_size = 1024 * 1024;

	static boolean_t
	vdev_initialize_should_stop(vdev_t *vd)
	{
	return (vd->vdev_initialize_exit_wanted \|\| !vdev_writeable(vd) \|\|
	vd->vdev_detached \|\| vd->vdev_top->vdev_removing \|\|
	vd->vdev_top->vdev_rz_expanding);
	}

	static void
	vdev_initialize_zap_update_sync(void arg, dmu_tx_t tx)
	{
	/*
	* We pass in the guid instead of the vdev_t since the vdev may
	* have been freed prior to the sync task being processed. This
	* happens when a vdev is detached as we call spa_config_vdev_exit(),
	* stop the initializing thread, schedule the sync task, and free
	* the vdev. Later when the scheduled sync task is invoked, it would
	* find that the vdev has been freed.
	*/
	uint64_t guid = (uint64_t )arg;
	uint64_t txg = dmu_tx_get_txg(tx);
	kmem_free(arg, sizeof (uint64_t));

	vdev_t *vd = spa_lookup_by_guid(tx->tx_pool->dp_spa, guid, B_FALSE);
	if (vd == NULL \|\| vd->vdev_top->vdev_removing \|\|
	!vdev_is_concrete(vd) \|\| vd->vdev_top->vdev_rz_expanding)
	return;

	uint64_t last_offset = vd->vdev_initialize_offset[txg & TXG_MASK];
	vd->vdev_initialize_offset[txg & TXG_MASK] = 0;

	VERIFY(vd->vdev_leaf_zap != 0);

	objset_t *mos = vd->vdev_spa->spa_meta_objset;

	if (last_offset > 0) {
	vd->vdev_initialize_last_offset = last_offset;
	VERIFY0(zap_update(mos, vd->vdev_leaf_zap,
	VDEV_LEAF_ZAP_INITIALIZE_LAST_OFFSET,
	sizeof (last_offset), 1, &last_offset, tx));
	}
	if (vd->vdev_initialize_action_time > 0) {
	uint64_t val = (uint64_t)vd->vdev_initialize_action_time;
	VERIFY0(zap_update(mos, vd->vdev_leaf_zap,
	VDEV_LEAF_ZAP_INITIALIZE_ACTION_TIME, sizeof (val),
	1, &val, tx));
	}

	uint64_t initialize_state = vd->vdev_initialize_state;
	VERIFY0(zap_update(mos, vd->vdev_leaf_zap,
	VDEV_LEAF_ZAP_INITIALIZE_STATE, sizeof (initialize_state), 1,
	&initialize_state, tx));
	}

	static void
	vdev_initialize_zap_remove_sync(void arg, dmu_tx_t tx)
	{
	uint64_t guid = (uint64_t )arg;

	kmem_free(arg, sizeof (uint64_t));

	vdev_t *vd = spa_lookup_by_guid(tx->tx_pool->dp_spa, guid, B_FALSE);
	if (vd == NULL \|\| vd->vdev_top->vdev_removing \|\| !vdev_is_concrete(vd))
	return;

	ASSERT3S(vd->vdev_initialize_state, ==, VDEV_INITIALIZE_NONE);
	ASSERT3U(vd->vdev_leaf_zap, !=, 0);

	vd->vdev_initialize_last_offset = 0;
	vd->vdev_initialize_action_time = 0;

	objset_t *mos = vd->vdev_spa->spa_meta_objset;
	int error;

	error = zap_remove(mos, vd->vdev_leaf_zap,
	VDEV_LEAF_ZAP_INITIALIZE_LAST_OFFSET, tx);
	VERIFY(error == 0 \|\| error == ENOENT);

	error = zap_remove(mos, vd->vdev_leaf_zap,
	VDEV_LEAF_ZAP_INITIALIZE_STATE, tx);
	VERIFY(error == 0 \|\| error == ENOENT);

	error = zap_remove(mos, vd->vdev_leaf_zap,
	VDEV_LEAF_ZAP_INITIALIZE_ACTION_TIME, tx);
	VERIFY(error == 0 \|\| error == ENOENT);
	}

	static void
	vdev_initialize_change_state(vdev_t *vd, vdev_initializing_state_t new_state)
	{
	ASSERT(MUTEX_HELD(&vd->vdev_initialize_lock));
	spa_t *spa = vd->vdev_spa;

	if (new_state == vd->vdev_initialize_state)
	return;

	/*
	* Copy the vd's guid, this will be freed by the sync task.
	*/
	uint64_t *guid = kmem_zalloc(sizeof (uint64_t), KM_SLEEP);
	*guid = vd->vdev_guid;

	/*
	* If we're suspending, then preserving the original start time.
	*/
	if (vd->vdev_initialize_state != VDEV_INITIALIZE_SUSPENDED) {
	vd->vdev_initialize_action_time = gethrestime_sec();
	}

	vdev_initializing_state_t old_state = vd->vdev_initialize_state;
	vd->vdev_initialize_state = new_state;

	dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
	VERIFY0(dmu_tx_assign(tx, TXG_WAIT));

	if (new_state != VDEV_INITIALIZE_NONE) {
	dsl_sync_task_nowait(spa_get_dsl(spa),
	vdev_initialize_zap_update_sync, guid, tx);
	} else {
	dsl_sync_task_nowait(spa_get_dsl(spa),
	vdev_initialize_zap_remove_sync, guid, tx);
	}

	switch (new_state) {
	case VDEV_INITIALIZE_ACTIVE:
	spa_history_log_internal(spa, "initialize", tx,
	"vdev=%s activated", vd->vdev_path);
	break;
	case VDEV_INITIALIZE_SUSPENDED:
	spa_history_log_internal(spa, "initialize", tx,
	"vdev=%s suspended", vd->vdev_path);
	break;
	case VDEV_INITIALIZE_CANCELED:
	if (old_state == VDEV_INITIALIZE_ACTIVE \|\|
	old_state == VDEV_INITIALIZE_SUSPENDED)
	spa_history_log_internal(spa, "initialize", tx,
	"vdev=%s canceled", vd->vdev_path);
	break;
	case VDEV_INITIALIZE_COMPLETE:
	spa_history_log_internal(spa, "initialize", tx,
	"vdev=%s complete", vd->vdev_path);
	break;
	case VDEV_INITIALIZE_NONE:
	spa_history_log_internal(spa, "uninitialize", tx,
	"vdev=%s", vd->vdev_path);
	break;
	default:
	panic("invalid state %llu", (unsigned long long)new_state);
	}

	dmu_tx_commit(tx);

	if (new_state != VDEV_INITIALIZE_ACTIVE)
	spa_notify_waiters(spa);
	}

	static void
	vdev_initialize_cb(zio_t *zio)
	{
	vdev_t *vd = zio->io_vd;
	mutex_enter(&vd->vdev_initialize_io_lock);
	if (zio->io_error == ENXIO && !vdev_writeable(vd)) {
	/*
	* The I/O failed because the vdev was unavailable; roll the
	* last offset back. (This works because spa_sync waits on
	* spa_txg_zio before it runs sync tasks.)
	*/
	uint64_t *off =
	&vd->vdev_initialize_offset[zio->io_txg & TXG_MASK];
	off = MIN(off, zio->io_offset);
	} else {
	/*
	* Since initializing is best-effort, we ignore I/O errors and
	* rely on vdev_probe to determine if the errors are more
	* critical.
	*/
	if (zio->io_error != 0)
	vd->vdev_stat.vs_initialize_errors++;

	vd->vdev_initialize_bytes_done += zio->io_orig_size;
	}
	ASSERT3U(vd->vdev_initialize_inflight, >, 0);
	vd->vdev_initialize_inflight--;
	cv_broadcast(&vd->vdev_initialize_io_cv);
	mutex_exit(&vd->vdev_initialize_io_lock);

	spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd);
	}

	/* Takes care of physical writing and limiting # of concurrent ZIOs. */
	static int
	vdev_initialize_write(vdev_t vd, uint64_t start, uint64_t size, abd_t data)
	{
	spa_t *spa = vd->vdev_spa;

	/* Limit inflight initializing I/Os */
	mutex_enter(&vd->vdev_initialize_io_lock);
	while (vd->vdev_initialize_inflight >= zfs_initialize_limit) {
	cv_wait(&vd->vdev_initialize_io_cv,
	&vd->vdev_initialize_io_lock);
	}
	vd->vdev_initialize_inflight++;
	mutex_exit(&vd->vdev_initialize_io_lock);

	dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
	VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
	uint64_t txg = dmu_tx_get_txg(tx);

	spa_config_enter(spa, SCL_STATE_ALL, vd, RW_READER);
	mutex_enter(&vd->vdev_initialize_lock);

	if (vd->vdev_initialize_offset[txg & TXG_MASK] == 0) {
	uint64_t *guid = kmem_zalloc(sizeof (uint64_t), KM_SLEEP);
	*guid = vd->vdev_guid;

	/* This is the first write of this txg. */
	dsl_sync_task_nowait(spa_get_dsl(spa),
	vdev_initialize_zap_update_sync, guid, tx);
	}

	/*
	* We know the vdev struct will still be around since all
	* consumers of vdev_free must stop the initialization first.
	*/
	if (vdev_initialize_should_stop(vd)) {
	mutex_enter(&vd->vdev_initialize_io_lock);
	ASSERT3U(vd->vdev_initialize_inflight, >, 0);
	vd->vdev_initialize_inflight--;
	mutex_exit(&vd->vdev_initialize_io_lock);
	spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd);
	mutex_exit(&vd->vdev_initialize_lock);
	dmu_tx_commit(tx);
	return (SET_ERROR(EINTR));
	}
	mutex_exit(&vd->vdev_initialize_lock);

	vd->vdev_initialize_offset[txg & TXG_MASK] = start + size;
	zio_nowait(zio_write_phys(spa->spa_txg_zio[txg & TXG_MASK], vd, start,
	size, data, ZIO_CHECKSUM_OFF, vdev_initialize_cb, NULL,
	ZIO_PRIORITY_INITIALIZING, ZIO_FLAG_CANFAIL, B_FALSE));
	/* vdev_initialize_cb releases SCL_STATE_ALL */

	dmu_tx_commit(tx);

	return (0);
	}

	/*
	* Callback to fill each ABD chunk with zfs_initialize_value. len must be
	* divisible by sizeof (uint64_t), and buf must be 8-byte aligned. The ABD
	* allocation will guarantee these for us.
	*/
	static int
	vdev_initialize_block_fill(void buf, size_t len, void unused)
	{
	(void) unused;

	ASSERT0(len % sizeof (uint64_t));
	for (uint64_t i = 0; i < len; i += sizeof (uint64_t)) {
	(uint64_t )((char *)(buf) + i) = zfs_initialize_value;
	}
	return (0);
	}

	static abd_t *
	vdev_initialize_block_alloc(void)
	{
	/* Allocate ABD for filler data */
	abd_t *data = abd_alloc_for_io(zfs_initialize_chunk_size, B_FALSE);

	ASSERT0(zfs_initialize_chunk_size % sizeof (uint64_t));
	(void) abd_iterate_func(data, 0, zfs_initialize_chunk_size,
	vdev_initialize_block_fill, NULL);

	return (data);
	}

	static void
	vdev_initialize_block_free(abd_t *data)
	{
	abd_free(data);
	}

	static int
	vdev_initialize_ranges(vdev_t vd, abd_t data)
	{
	range_tree_t *rt = vd->vdev_initialize_tree;
	zfs_btree_t *bt = &rt->rt_root;
	zfs_btree_index_t where;

	for (range_seg_t *rs = zfs_btree_first(bt, &where); rs != NULL;
	rs = zfs_btree_next(bt, &where, &where)) {
	uint64_t size = rs_get_end(rs, rt) - rs_get_start(rs, rt);

	/* Split range into legally-sized physical chunks */
	uint64_t writes_required =
	((size - 1) / zfs_initialize_chunk_size) + 1;

	for (uint64_t w = 0; w < writes_required; w++) {
	int error;

	error = vdev_initialize_write(vd,
	VDEV_LABEL_START_SIZE + rs_get_start(rs, rt) +
	(w * zfs_initialize_chunk_size),
	MIN(size - (w * zfs_initialize_chunk_size),
	zfs_initialize_chunk_size), data);
	if (error != 0)
	return (error);
	}
	}
	return (0);
	}

	static void
	vdev_initialize_xlate_last_rs_end(void arg, range_seg64_t physical_rs)
	{
	uint64_t last_rs_end = (uint64_t )arg;

	if (physical_rs->rs_end > *last_rs_end)
	*last_rs_end = physical_rs->rs_end;
	}

	static void
	vdev_initialize_xlate_progress(void arg, range_seg64_t physical_rs)
	{
	vdev_t vd = (vdev_t )arg;

	uint64_t size = physical_rs->rs_end - physical_rs->rs_start;
	vd->vdev_initialize_bytes_est += size;

	if (vd->vdev_initialize_last_offset > physical_rs->rs_end) {
	vd->vdev_initialize_bytes_done += size;
	} else if (vd->vdev_initialize_last_offset > physical_rs->rs_start &&
	vd->vdev_initialize_last_offset < physical_rs->rs_end) {
	vd->vdev_initialize_bytes_done +=
	vd->vdev_initialize_last_offset - physical_rs->rs_start;
	}
	}

	static void
	vdev_initialize_calculate_progress(vdev_t *vd)
	{
	ASSERT(spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_READER) \|\|
	spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_WRITER));
	ASSERT(vd->vdev_leaf_zap != 0);

	vd->vdev_initialize_bytes_est = 0;
	vd->vdev_initialize_bytes_done = 0;

	for (uint64_t i = 0; i < vd->vdev_top->vdev_ms_count; i++) {
	metaslab_t *msp = vd->vdev_top->vdev_ms[i];
	mutex_enter(&msp->ms_lock);

	uint64_t ms_free = (msp->ms_size -
	metaslab_allocated_space(msp)) /
	vdev_get_ndisks(vd->vdev_top);

	/*
	* Convert the metaslab range to a physical range
	* on our vdev. We use this to determine if we are
	* in the middle of this metaslab range.
	*/
	range_seg64_t logical_rs, physical_rs, remain_rs;
	logical_rs.rs_start = msp->ms_start;
	logical_rs.rs_end = msp->ms_start + msp->ms_size;

	/* Metaslab space after this offset has not been initialized */
	vdev_xlate(vd, &logical_rs, &physical_rs, &remain_rs);
	if (vd->vdev_initialize_last_offset <= physical_rs.rs_start) {
	vd->vdev_initialize_bytes_est += ms_free;
	mutex_exit(&msp->ms_lock);
	continue;
	}

	/* Metaslab space before this offset has been initialized */
	uint64_t last_rs_end = physical_rs.rs_end;
	if (!vdev_xlate_is_empty(&remain_rs)) {
	vdev_xlate_walk(vd, &remain_rs,
	vdev_initialize_xlate_last_rs_end, &last_rs_end);
	}

	if (vd->vdev_initialize_last_offset > last_rs_end) {
	vd->vdev_initialize_bytes_done += ms_free;
	vd->vdev_initialize_bytes_est += ms_free;
	mutex_exit(&msp->ms_lock);
	continue;
	}

	/*
	* If we get here, we're in the middle of initializing this
	* metaslab. Load it and walk the free tree for more accurate
	* progress estimation.
	*/
	VERIFY0(metaslab_load(msp));

	zfs_btree_index_t where;
	range_tree_t *rt = msp->ms_allocatable;
	for (range_seg_t *rs =
	zfs_btree_first(&rt->rt_root, &where); rs;
	rs = zfs_btree_next(&rt->rt_root, &where,
	&where)) {
	logical_rs.rs_start = rs_get_start(rs, rt);
	logical_rs.rs_end = rs_get_end(rs, rt);

	vdev_xlate_walk(vd, &logical_rs,
	vdev_initialize_xlate_progress, vd);
	}
	mutex_exit(&msp->ms_lock);
	}
	}

	static int
	vdev_initialize_load(vdev_t *vd)
	{
	int err = 0;
	ASSERT(spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_READER) \|\|
	spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_WRITER));
	ASSERT(vd->vdev_leaf_zap != 0);

	if (vd->vdev_initialize_state == VDEV_INITIALIZE_ACTIVE \|\|
	vd->vdev_initialize_state == VDEV_INITIALIZE_SUSPENDED) {
	err = zap_lookup(vd->vdev_spa->spa_meta_objset,
	vd->vdev_leaf_zap, VDEV_LEAF_ZAP_INITIALIZE_LAST_OFFSET,
	sizeof (vd->vdev_initialize_last_offset), 1,
	&vd->vdev_initialize_last_offset);
	if (err == ENOENT) {
	vd->vdev_initialize_last_offset = 0;
	err = 0;
	}
	}

	vdev_initialize_calculate_progress(vd);
	return (err);
	}

	static void
	vdev_initialize_xlate_range_add(void arg, range_seg64_t physical_rs)
	{
	vdev_t *vd = arg;

	/* Only add segments that we have not visited yet */
	if (physical_rs->rs_end <= vd->vdev_initialize_last_offset)
	return;

	/* Pick up where we left off mid-range. */
	if (vd->vdev_initialize_last_offset > physical_rs->rs_start) {
	zfs_dbgmsg("range write: vd %s changed (%llu, %llu) to "
	"(%llu, %llu)", vd->vdev_path,
	(u_longlong_t)physical_rs->rs_start,
	(u_longlong_t)physical_rs->rs_end,
	(u_longlong_t)vd->vdev_initialize_last_offset,
	(u_longlong_t)physical_rs->rs_end);
	ASSERT3U(physical_rs->rs_end, >,
	vd->vdev_initialize_last_offset);
	physical_rs->rs_start = vd->vdev_initialize_last_offset;
	}

	ASSERT3U(physical_rs->rs_end, >, physical_rs->rs_start);

	range_tree_add(vd->vdev_initialize_tree, physical_rs->rs_start,
	physical_rs->rs_end - physical_rs->rs_start);
	}

	/*
	* Convert the logical range into a physical range and add it to our
	* avl tree.
	*/
	static void
	vdev_initialize_range_add(void *arg, uint64_t start, uint64_t size)
	{
	vdev_t *vd = arg;
	range_seg64_t logical_rs;
	logical_rs.rs_start = start;
	logical_rs.rs_end = start + size;

	ASSERT(vd->vdev_ops->vdev_op_leaf);
	vdev_xlate_walk(vd, &logical_rs, vdev_initialize_xlate_range_add, arg);
	}

	static __attribute__((noreturn)) void
	vdev_initialize_thread(void *arg)
	{
	vdev_t *vd = arg;
	spa_t *spa = vd->vdev_spa;
	int error = 0;
	uint64_t ms_count = 0;

	ASSERT(vdev_is_concrete(vd));
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);

	vd->vdev_initialize_last_offset = 0;
	VERIFY0(vdev_initialize_load(vd));

	abd_t *deadbeef = vdev_initialize_block_alloc();

	vd->vdev_initialize_tree = range_tree_create(NULL, RANGE_SEG64, NULL,
	0, 0);

	for (uint64_t i = 0; !vd->vdev_detached &&
	i < vd->vdev_top->vdev_ms_count; i++) {
	metaslab_t *msp = vd->vdev_top->vdev_ms[i];
	boolean_t unload_when_done = B_FALSE;

	/*
	* If we've expanded the top-level vdev or it's our
	* first pass, calculate our progress.
	*/
	if (vd->vdev_top->vdev_ms_count != ms_count) {
	vdev_initialize_calculate_progress(vd);
	ms_count = vd->vdev_top->vdev_ms_count;
	}

	spa_config_exit(spa, SCL_CONFIG, FTAG);
	metaslab_disable(msp);
	mutex_enter(&msp->ms_lock);
	if (!msp->ms_loaded && !msp->ms_loading)
	unload_when_done = B_TRUE;
	VERIFY0(metaslab_load(msp));

	range_tree_walk(msp->ms_allocatable, vdev_initialize_range_add,
	vd);
	mutex_exit(&msp->ms_lock);

	error = vdev_initialize_ranges(vd, deadbeef);
	metaslab_enable(msp, B_TRUE, unload_when_done);
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);

	range_tree_vacate(vd->vdev_initialize_tree, NULL, NULL);
	if (error != 0)
	break;
	}

	spa_config_exit(spa, SCL_CONFIG, FTAG);
	mutex_enter(&vd->vdev_initialize_io_lock);
	while (vd->vdev_initialize_inflight > 0) {
	cv_wait(&vd->vdev_initialize_io_cv,
	&vd->vdev_initialize_io_lock);
	}
	mutex_exit(&vd->vdev_initialize_io_lock);

	range_tree_destroy(vd->vdev_initialize_tree);
	vdev_initialize_block_free(deadbeef);
	vd->vdev_initialize_tree = NULL;

	mutex_enter(&vd->vdev_initialize_lock);
	if (!vd->vdev_initialize_exit_wanted) {
	if (vdev_writeable(vd)) {
	vdev_initialize_change_state(vd,
	VDEV_INITIALIZE_COMPLETE);
	} else if (vd->vdev_faulted) {
	vdev_initialize_change_state(vd,
	VDEV_INITIALIZE_CANCELED);
	}
	}
	ASSERT(vd->vdev_initialize_thread != NULL \|\|
	vd->vdev_initialize_inflight == 0);

	/*
	* Drop the vdev_initialize_lock while we sync out the
	* txg since it's possible that a device might be trying to
	* come online and must check to see if it needs to restart an
	* initialization. That thread will be holding the spa_config_lock
	* which would prevent the txg_wait_synced from completing.
	*/
	mutex_exit(&vd->vdev_initialize_lock);
	txg_wait_synced(spa_get_dsl(spa), 0);
	mutex_enter(&vd->vdev_initialize_lock);

	vd->vdev_initialize_thread = NULL;
	cv_broadcast(&vd->vdev_initialize_cv);
	mutex_exit(&vd->vdev_initialize_lock);

	thread_exit();
	}

	/*
	* Initiates a device. Caller must hold vdev_initialize_lock.
	* Device must be a leaf and not already be initializing.
	*/
	void
	vdev_initialize(vdev_t *vd)
	{
	ASSERT(MUTEX_HELD(&vd->vdev_initialize_lock));
	ASSERT(vd->vdev_ops->vdev_op_leaf);
	ASSERT(vdev_is_concrete(vd));
	ASSERT3P(vd->vdev_initialize_thread, ==, NULL);
	ASSERT(!vd->vdev_detached);
	ASSERT(!vd->vdev_initialize_exit_wanted);
	ASSERT(!vd->vdev_top->vdev_removing);
	ASSERT(!vd->vdev_top->vdev_rz_expanding);

	vdev_initialize_change_state(vd, VDEV_INITIALIZE_ACTIVE);
	vd->vdev_initialize_thread = thread_create(NULL, 0,
	vdev_initialize_thread, vd, 0, &p0, TS_RUN, maxclsyspri);
	}

	/*
	* Uninitializes a device. Caller must hold vdev_initialize_lock.
	* Device must be a leaf and not already be initializing.
	*/
	void
	vdev_uninitialize(vdev_t *vd)
	{
	ASSERT(MUTEX_HELD(&vd->vdev_initialize_lock));
	ASSERT(vd->vdev_ops->vdev_op_leaf);
	ASSERT(vdev_is_concrete(vd));
	ASSERT3P(vd->vdev_initialize_thread, ==, NULL);
	ASSERT(!vd->vdev_detached);
	ASSERT(!vd->vdev_initialize_exit_wanted);
	ASSERT(!vd->vdev_top->vdev_removing);

	vdev_initialize_change_state(vd, VDEV_INITIALIZE_NONE);
	}

	/*
	* Wait for the initialize thread to be terminated (cancelled or stopped).
	*/
	static void
	vdev_initialize_stop_wait_impl(vdev_t *vd)
	{
	ASSERT(MUTEX_HELD(&vd->vdev_initialize_lock));

	while (vd->vdev_initialize_thread != NULL)
	cv_wait(&vd->vdev_initialize_cv, &vd->vdev_initialize_lock);

	ASSERT3P(vd->vdev_initialize_thread, ==, NULL);
	vd->vdev_initialize_exit_wanted = B_FALSE;
	}

	/*
	* Wait for vdev initialize threads which were either to cleanly exit.
	*/
	void
	vdev_initialize_stop_wait(spa_t spa, list_t vd_list)
	{
	(void) spa;
	vdev_t *vd;

	- ASSERT(MUTEX_HELD(&spa_namespace_lock));
	+ ASSERT(MUTEX_HELD(&spa_namespace_lock) \|\|
	+ spa->spa_export_thread == curthread);

	while ((vd = list_remove_head(vd_list)) != NULL) {
	mutex_enter(&vd->vdev_initialize_lock);
	vdev_initialize_stop_wait_impl(vd);
	mutex_exit(&vd->vdev_initialize_lock);
	}
	}

	/*
	* Stop initializing a device, with the resultant initializing state being
	* tgt_state. For blocking behavior pass NULL for vd_list. Otherwise, when
	* a list_t is provided the stopping vdev is inserted in to the list. Callers
	* are then required to call vdev_initialize_stop_wait() to block for all the
	* initialization threads to exit. The caller must hold vdev_initialize_lock
	* and must not be writing to the spa config, as the initializing thread may
	* try to enter the config as a reader before exiting.
	*/
	void
	vdev_initialize_stop(vdev_t *vd, vdev_initializing_state_t tgt_state,
	list_t *vd_list)
	{
	ASSERT(!spa_config_held(vd->vdev_spa, SCL_CONFIG\|SCL_STATE, RW_WRITER));
	ASSERT(MUTEX_HELD(&vd->vdev_initialize_lock));
	ASSERT(vd->vdev_ops->vdev_op_leaf);
	ASSERT(vdev_is_concrete(vd));

	/*
	* Allow cancel requests to proceed even if the initialize thread
	* has stopped.
	*/
	if (vd->vdev_initialize_thread == NULL &&
	tgt_state != VDEV_INITIALIZE_CANCELED) {
	return;
	}

	vdev_initialize_change_state(vd, tgt_state);
	vd->vdev_initialize_exit_wanted = B_TRUE;

	if (vd_list == NULL) {
	vdev_initialize_stop_wait_impl(vd);
	} else {
	- ASSERT(MUTEX_HELD(&spa_namespace_lock));
	+ ASSERT(MUTEX_HELD(&spa_namespace_lock) \|\|
	+ vd->vdev_spa->spa_export_thread == curthread);
	list_insert_tail(vd_list, vd);
	}
	}

	static void
	vdev_initialize_stop_all_impl(vdev_t *vd, vdev_initializing_state_t tgt_state,
	list_t *vd_list)
	{
	if (vd->vdev_ops->vdev_op_leaf && vdev_is_concrete(vd)) {
	mutex_enter(&vd->vdev_initialize_lock);
	vdev_initialize_stop(vd, tgt_state, vd_list);
	mutex_exit(&vd->vdev_initialize_lock);
	return;
	}

	for (uint64_t i = 0; i < vd->vdev_children; i++) {
	vdev_initialize_stop_all_impl(vd->vdev_child[i], tgt_state,
	vd_list);
	}
	}

	/*
	* Convenience function to stop initializing of a vdev tree and set all
	* initialize thread pointers to NULL.
	*/
	void
	vdev_initialize_stop_all(vdev_t *vd, vdev_initializing_state_t tgt_state)
	{
	spa_t *spa = vd->vdev_spa;
	list_t vd_list;

	- ASSERT(MUTEX_HELD(&spa_namespace_lock));
	+ ASSERT(MUTEX_HELD(&spa_namespace_lock) \|\|
	+ spa->spa_export_thread == curthread);

	list_create(&vd_list, sizeof (vdev_t),
	offsetof(vdev_t, vdev_initialize_node));

	vdev_initialize_stop_all_impl(vd, tgt_state, &vd_list);
	vdev_initialize_stop_wait(spa, &vd_list);

	if (vd->vdev_spa->spa_sync_on) {
	/* Make sure that our state has been synced to disk */
	txg_wait_synced(spa_get_dsl(vd->vdev_spa), 0);
	}

	list_destroy(&vd_list);
	}

	void
	vdev_initialize_restart(vdev_t *vd)
	{
	ASSERT(MUTEX_HELD(&spa_namespace_lock) \|\|
	vd->vdev_spa->spa_load_thread == curthread);
	ASSERT(!spa_config_held(vd->vdev_spa, SCL_ALL, RW_WRITER));

	if (vd->vdev_leaf_zap != 0) {
	mutex_enter(&vd->vdev_initialize_lock);
	uint64_t initialize_state = VDEV_INITIALIZE_NONE;
	int err = zap_lookup(vd->vdev_spa->spa_meta_objset,
	vd->vdev_leaf_zap, VDEV_LEAF_ZAP_INITIALIZE_STATE,
	sizeof (initialize_state), 1, &initialize_state);
	ASSERT(err == 0 \|\| err == ENOENT);
	vd->vdev_initialize_state = initialize_state;

	uint64_t timestamp = 0;
	err = zap_lookup(vd->vdev_spa->spa_meta_objset,
	vd->vdev_leaf_zap, VDEV_LEAF_ZAP_INITIALIZE_ACTION_TIME,
	sizeof (timestamp), 1, &timestamp);
	ASSERT(err == 0 \|\| err == ENOENT);
	vd->vdev_initialize_action_time = timestamp;

	if ((vd->vdev_initialize_state == VDEV_INITIALIZE_SUSPENDED \|\|
	vd->vdev_offline) && !vd->vdev_top->vdev_rz_expanding) {
	/* load progress for reporting, but don't resume */
	VERIFY0(vdev_initialize_load(vd));
	} else if (vd->vdev_initialize_state ==
	VDEV_INITIALIZE_ACTIVE && vdev_writeable(vd) &&
	!vd->vdev_top->vdev_removing &&
	!vd->vdev_top->vdev_rz_expanding &&
	vd->vdev_initialize_thread == NULL) {
	vdev_initialize(vd);
	}

	mutex_exit(&vd->vdev_initialize_lock);
	}

	for (uint64_t i = 0; i < vd->vdev_children; i++) {
	vdev_initialize_restart(vd->vdev_child[i]);
	}
	}

	EXPORT_SYMBOL(vdev_initialize);
	EXPORT_SYMBOL(vdev_uninitialize);
	EXPORT_SYMBOL(vdev_initialize_stop);
	EXPORT_SYMBOL(vdev_initialize_stop_all);
	EXPORT_SYMBOL(vdev_initialize_stop_wait);
	EXPORT_SYMBOL(vdev_initialize_restart);

	ZFS_MODULE_PARAM(zfs, zfs_, initialize_value, U64, ZMOD_RW,
	"Value written during zpool initialize");

	ZFS_MODULE_PARAM(zfs, zfs_, initialize_chunk_size, U64, ZMOD_RW,
	"Size in bytes of writes by zpool initialize");
	diff --git a/sys/contrib/openzfs/module/zfs/vdev_raidz.c b/sys/contrib/openzfs/module/zfs/vdev_raidz.c
	index de7d0fa79478..15c8b8ca6016 100644
	--- a/sys/contrib/openzfs/module/zfs/vdev_raidz.c
	+++ b/sys/contrib/openzfs/module/zfs/vdev_raidz.c
	@@ -1,5021 +1,5022 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2020 by Delphix. All rights reserved.
	* Copyright (c) 2016 Gvozden Nešković. All rights reserved.
	*/

	#include <sys/zfs_context.h>
	#include <sys/spa.h>
	#include <sys/spa_impl.h>
	#include <sys/zap.h>
	#include <sys/vdev_impl.h>
	#include <sys/metaslab_impl.h>
	#include <sys/zio.h>
	#include <sys/zio_checksum.h>
	#include <sys/dmu_tx.h>
	#include <sys/abd.h>
	#include <sys/zfs_rlock.h>
	#include <sys/fs/zfs.h>
	#include <sys/fm/fs/zfs.h>
	#include <sys/vdev_raidz.h>
	#include <sys/vdev_raidz_impl.h>
	#include <sys/vdev_draid.h>
	#include <sys/uberblock_impl.h>
	#include <sys/dsl_scan.h>

	#ifdef ZFS_DEBUG
	#include <sys/vdev.h> /* For vdev_xlate() in vdev_raidz_io_verify() */
	#endif

	/*
	* Virtual device vector for RAID-Z.
	*
	* This vdev supports single, double, and triple parity. For single parity,
	* we use a simple XOR of all the data columns. For double or triple parity,
	* we use a special case of Reed-Solomon coding. This extends the
	* technique described in "The mathematics of RAID-6" by H. Peter Anvin by
	* drawing on the system described in "A Tutorial on Reed-Solomon Coding for
	* Fault-Tolerance in RAID-like Systems" by James S. Plank on which the
	* former is also based. The latter is designed to provide higher performance
	* for writes.
	*
	* Note that the Plank paper claimed to support arbitrary N+M, but was then
	* amended six years later identifying a critical flaw that invalidates its
	* claims. Nevertheless, the technique can be adapted to work for up to
	* triple parity. For additional parity, the amendment "Note: Correction to
	* the 1997 Tutorial on Reed-Solomon Coding" by James S. Plank and Ying Ding
	* is viable, but the additional complexity means that write performance will
	* suffer.
	*
	* All of the methods above operate on a Galois field, defined over the
	* integers mod 2^N. In our case we choose N=8 for GF(8) so that all elements
	* can be expressed with a single byte. Briefly, the operations on the
	* field are defined as follows:
	*
	* o addition (+) is represented by a bitwise XOR
	* o subtraction (-) is therefore identical to addition: A + B = A - B
	* o multiplication of A by 2 is defined by the following bitwise expression:
	*
	* (A * 2)_7 = A_6
	* (A * 2)_6 = A_5
	* (A * 2)_5 = A_4
	* (A * 2)_4 = A_3 + A_7
	* (A * 2)_3 = A_2 + A_7
	* (A * 2)_2 = A_1 + A_7
	* (A * 2)_1 = A_0
	* (A * 2)_0 = A_7
	*
	* In C, multiplying by 2 is therefore ((a << 1) ^ ((a & 0x80) ? 0x1d : 0)).
	* As an aside, this multiplication is derived from the error correcting
	* primitive polynomial x^8 + x^4 + x^3 + x^2 + 1.
	*
	* Observe that any number in the field (except for 0) can be expressed as a
	* power of 2 -- a generator for the field. We store a table of the powers of
	* 2 and logs base 2 for quick look ups, and exploit the fact that A * B can
	* be rewritten as 2^(log_2(A) + log_2(B)) (where '+' is normal addition rather
	* than field addition). The inverse of a field element A (A^-1) is therefore
	* A ^ (255 - 1) = A^254.
	*
	* The up-to-three parity columns, P, Q, R over several data columns,
	* D_0, ... D_n-1, can be expressed by field operations:
	*
	* P = D_0 + D_1 + ... + D_n-2 + D_n-1
	* Q = 2^n-1 * D_0 + 2^n-2 * D_1 + ... + 2^1 * D_n-2 + 2^0 * D_n-1
	* = ((...((D_0) * 2 + D_1) * 2 + ...) * 2 + D_n-2) * 2 + D_n-1
	* R = 4^n-1 * D_0 + 4^n-2 * D_1 + ... + 4^1 * D_n-2 + 4^0 * D_n-1
	* = ((...((D_0) * 4 + D_1) * 4 + ...) * 4 + D_n-2) * 4 + D_n-1
	*
	* We chose 1, 2, and 4 as our generators because 1 corresponds to the trivial
	* XOR operation, and 2 and 4 can be computed quickly and generate linearly-
	* independent coefficients. (There are no additional coefficients that have
	* this property which is why the uncorrected Plank method breaks down.)
	*
	* See the reconstruction code below for how P, Q and R can used individually
	* or in concert to recover missing data columns.
	*/

	#define VDEV_RAIDZ_P 0
	#define VDEV_RAIDZ_Q 1
	#define VDEV_RAIDZ_R 2

	#define VDEV_RAIDZ_MUL_2(x) (((x) << 1) ^ (((x) & 0x80) ? 0x1d : 0))
	#define VDEV_RAIDZ_MUL_4(x) (VDEV_RAIDZ_MUL_2(VDEV_RAIDZ_MUL_2(x)))

	/*
	* We provide a mechanism to perform the field multiplication operation on a
	* 64-bit value all at once rather than a byte at a time. This works by
	* creating a mask from the top bit in each byte and using that to
	* conditionally apply the XOR of 0x1d.
	*/
	#define VDEV_RAIDZ_64MUL_2(x, mask) \
	{ \
	(mask) = (x) & 0x8080808080808080ULL; \
	(mask) = ((mask) << 1) - ((mask) >> 7); \
	(x) = (((x) << 1) & 0xfefefefefefefefeULL) ^ \
	((mask) & 0x1d1d1d1d1d1d1d1dULL); \
	}

	#define VDEV_RAIDZ_64MUL_4(x, mask) \
	{ \
	VDEV_RAIDZ_64MUL_2((x), mask); \
	VDEV_RAIDZ_64MUL_2((x), mask); \
	}


	/*
	* Big Theory Statement for how a RAIDZ VDEV is expanded
	*
	* An existing RAIDZ VDEV can be expanded by attaching a new disk. Expansion
	* works with all three RAIDZ parity choices, including RAIDZ1, 2, or 3. VDEVs
	* that have been previously expanded can be expanded again.
	*
	* The RAIDZ VDEV must be healthy (must be able to write to all the drives in
	* the VDEV) when an expansion starts. And the expansion will pause if any
	* disk in the VDEV fails, and resume once the VDEV is healthy again. All other
	* operations on the pool can continue while an expansion is in progress (e.g.
	* read/write, snapshot, zpool add, etc). Except zpool checkpoint, zpool trim,
	* and zpool initialize which can't be run during an expansion. Following a
	* reboot or export/import, the expansion resumes where it left off.
	*
	* == Reflowing the Data ==
	*
	* The expansion involves reflowing (copying) the data from the current set
	* of disks to spread it across the new set which now has one more disk. This
	* reflow operation is similar to reflowing text when the column width of a
	* text editor window is expanded. The text doesn’t change but the location of
	* the text changes to accommodate the new width. An example reflow result for
	* a 4-wide RAIDZ1 to a 5-wide is shown below.
	*
	* Reflow End State
	* Each letter indicates a parity group (logical stripe)
	*
	* Before expansion After Expansion
	* D1 D2 D3 D4 D1 D2 D3 D4 D5
	* +------+------+------+------+ +------+------+------+------+------+
	* \| \| \| \| \| \| \| \| \| \| \|
	* \| A \| A \| A \| A \| \| A \| A \| A \| A \| B \|
	* \| 1\| 2\| 3\| 4\| \| 1\| 2\| 3\| 4\| 5\|
	* +------+------+------+------+ +------+------+------+------+------+
	* \| \| \| \| \| \| \| \| \| \| \|
	* \| B \| B \| C \| C \| \| B \| C \| C \| C \| C \|
	* \| 5\| 6\| 7\| 8\| \| 6\| 7\| 8\| 9\| 10\|
	* +------+------+------+------+ +------+------+------+------+------+
	* \| \| \| \| \| \| \| \| \| \| \|
	* \| C \| C \| D \| D \| \| D \| D \| E \| E \| E \|
	* \| 9\| 10\| 11\| 12\| \| 11\| 12\| 13\| 14\| 15\|
	* +------+------+------+------+ +------+------+------+------+------+
	* \| \| \| \| \| \| \| \| \| \| \|
	* \| E \| E \| E \| E \| --> \| E \| F \| F \| G \| G \|
	* \| 13\| 14\| 15\| 16\| \| 16\| 17\| 18\|p 19\| 20\|
	* +------+------+------+------+ +------+------+------+------+------+
	* \| \| \| \| \| \| \| \| \| \| \|
	* \| F \| F \| G \| G \| \| G \| G \| H \| H \| H \|
	* \| 17\| 18\| 19\| 20\| \| 21\| 22\| 23\| 24\| 25\|
	* +------+------+------+------+ +------+------+------+------+------+
	* \| \| \| \| \| \| \| \| \| \| \|
	* \| G \| G \| H \| H \| \| H \| I \| I \| J \| J \|
	* \| 21\| 22\| 23\| 24\| \| 26\| 27\| 28\| 29\| 30\|
	* +------+------+------+------+ +------+------+------+------+------+
	* \| \| \| \| \| \| \| \| \| \| \|
	* \| H \| H \| I \| I \| \| J \| J \| \| \| K \|
	* \| 25\| 26\| 27\| 28\| \| 31\| 32\| 33\| 34\| 35\|
	* +------+------+------+------+ +------+------+------+------+------+
	*
	* This reflow approach has several advantages. There is no need to read or
	* modify the block pointers or recompute any block checksums. The reflow
	* doesn’t need to know where the parity sectors reside. We can read and write
	* data sequentially and the copy can occur in a background thread in open
	* context. The design also allows for fast discovery of what data to copy.
	*
	* The VDEV metaslabs are processed, one at a time, to copy the block data to
	* have it flow across all the disks. The metaslab is disabled for allocations
	* during the copy. As an optimization, we only copy the allocated data which
	* can be determined by looking at the metaslab range tree. During the copy we
	* must maintain the redundancy guarantees of the RAIDZ VDEV (i.e., we still
	* need to be able to survive losing parity count disks). This means we
	* cannot overwrite data during the reflow that would be needed if a disk is
	* lost.
	*
	* After the reflow completes, all newly-written blocks will have the new
	* layout, i.e., they will have the parity to data ratio implied by the new
	* number of disks in the RAIDZ group. Even though the reflow copies all of
	* the allocated space (data and parity), it is only rearranged, not changed.
	*
	* This act of reflowing the data has a few implications about blocks
	* that were written before the reflow completes:
	*
	* - Old blocks will still use the same amount of space (i.e., they will have
	* the parity to data ratio implied by the old number of disks in the RAIDZ
	* group).
	* - Reading old blocks will be slightly slower than before the reflow, for
	* two reasons. First, we will have to read from all disks in the RAIDZ
	* VDEV, rather than being able to skip the children that contain only
	* parity of this block (because the data of a single block is now spread
	* out across all the disks). Second, in most cases there will be an extra
	* bcopy, needed to rearrange the data back to its original layout in memory.
	*
	* == Scratch Area ==
	*
	* As we copy the block data, we can only progress to the point that writes
	* will not overlap with blocks whose progress has not yet been recorded on
	* disk. Since partially-copied rows are always read from the old location,
	* we need to stop one row before the sector-wise overlap, to prevent any
	* row-wise overlap. For example, in the diagram above, when we reflow sector
	* B6 it will overwite the original location for B5.
	*
	* To get around this, a scratch space is used so that we can start copying
	* without risking data loss by overlapping the row. As an added benefit, it
	* improves performance at the beginning of the reflow, but that small perf
	* boost wouldn't be worth the complexity on its own.
	*
	* Ideally we want to copy at least 2 * (new_width)^2 so that we have a
	* separation of 2*(new_width+1) and a chunk size of new_width+2. With the max
	* RAIDZ width of 255 and 4K sectors this would be 2MB per disk. In practice
	* the widths will likely be single digits so we can get a substantial chuck
	* size using only a few MB of scratch per disk.
	*
	* The scratch area is persisted to disk which holds a large amount of reflowed
	* state. We can always read the partially written stripes when a disk fails or
	* the copy is interrupted (crash) during the initial copying phase and also
	* get past a small chunk size restriction. At a minimum, the scratch space
	* must be large enough to get us to the point that one row does not overlap
	* itself when moved (i.e new_width^2). But going larger is even better. We
	* use the 3.5 MiB reserved "boot" space that resides after the ZFS disk labels
	* as our scratch space to handle overwriting the initial part of the VDEV.
	*
	* 0 256K 512K 4M
	* +------+------+-----------------------+-----------------------------
	* \| VDEV \| VDEV \| Boot Block (3.5M) \| Allocatable space ...
	* \| L0 \| L1 \| Reserved \| (Metaslabs)
	* +------+------+-----------------------+-------------------------------
	* Scratch Area
	*
	* == Reflow Progress Updates ==
	* After the initial scratch-based reflow, the expansion process works
	* similarly to device removal. We create a new open context thread which
	* reflows the data, and periodically kicks off sync tasks to update logical
	* state. In this case, state is the committed progress (offset of next data
	* to copy). We need to persist the completed offset on disk, so that if we
	* crash we know which format each VDEV offset is in.
	*
	* == Time Dependent Geometry ==
	*
	* In non-expanded RAIDZ, blocks are read from disk in a column by column
	* fashion. For a multi-row block, the second sector is in the first column
	* not in the second column. This allows us to issue full reads for each
	* column directly into the request buffer. The block data is thus laid out
	* sequentially in a column-by-column fashion.
	*
	* For example, in the before expansion diagram above, one logical block might
	* be sectors G19-H26. The parity is in G19,H23; and the data is in
	* G20,H24,G21,H25,G22,H26.
	*
	* After a block is reflowed, the sectors that were all in the original column
	* data can now reside in different columns. When reading from an expanded
	* VDEV, we need to know the logical stripe width for each block so we can
	* reconstitute the block’s data after the reads are completed. Likewise,
	* when we perform the combinatorial reconstruction we need to know the
	* original width so we can retry combinations from the past layouts.
	*
	* Time dependent geometry is what we call having blocks with different layouts
	* (stripe widths) in the same VDEV. This time-dependent geometry uses the
	* block’s birth time (+ the time expansion ended) to establish the correct
	* width for a given block. After an expansion completes, we record the time
	* for blocks written with a particular width (geometry).
	*
	* == On Disk Format Changes ==
	*
	* New pool feature flag, 'raidz_expansion' whose reference count is the number
	* of RAIDZ VDEVs that have been expanded.
	*
	* The blocks on expanded RAIDZ VDEV can have different logical stripe widths.
	*
	* Since the uberblock can point to arbitrary blocks, which might be on the
	* expanding RAIDZ, and might or might not have been expanded. We need to know
	* which way a block is laid out before reading it. This info is the next
	* offset that needs to be reflowed and we persist that in the uberblock, in
	* the new ub_raidz_reflow_info field, as opposed to the MOS or the vdev label.
	* After the expansion is complete, we then use the raidz_expand_txgs array
	* (see below) to determine how to read a block and the ub_raidz_reflow_info
	* field no longer required.
	*
	* The uberblock's ub_raidz_reflow_info field also holds the scratch space
	* state (i.e., active or not) which is also required before reading a block
	* during the initial phase of reflowing the data.
	*
	* The top-level RAIDZ VDEV has two new entries in the nvlist:
	*
	* 'raidz_expand_txgs' array: logical stripe widths by txg are recorded here
	* and used after the expansion is complete to
	* determine how to read a raidz block
	* 'raidz_expanding' boolean: present during reflow and removed after completion
	* used during a spa import to resume an unfinished
	* expansion
	*
	* And finally the VDEVs top zap adds the following informational entries:
	* VDEV_TOP_ZAP_RAIDZ_EXPAND_STATE
	* VDEV_TOP_ZAP_RAIDZ_EXPAND_START_TIME
	* VDEV_TOP_ZAP_RAIDZ_EXPAND_END_TIME
	* VDEV_TOP_ZAP_RAIDZ_EXPAND_BYTES_COPIED
	*/

	/*
	* For testing only: pause the raidz expansion after reflowing this amount.
	* (accessed by ZTS and ztest)
	*/
	#ifdef _KERNEL
	static
	#endif /* _KERNEL */
	unsigned long raidz_expand_max_reflow_bytes = 0;

	/*
	* For testing only: pause the raidz expansion at a certain point.
	*/
	uint_t raidz_expand_pause_point = 0;

	/*
	* Maximum amount of copy io's outstanding at once.
	*/
	static unsigned long raidz_expand_max_copy_bytes = 10 * SPA_MAXBLOCKSIZE;

	/*
	* Apply raidz map abds aggregation if the number of rows in the map is equal
	* or greater than the value below.
	*/
	static unsigned long raidz_io_aggregate_rows = 4;

	/*
	* Automatically start a pool scrub when a RAIDZ expansion completes in
	* order to verify the checksums of all blocks which have been copied
	* during the expansion. Automatic scrubbing is enabled by default and
	* is strongly recommended.
	*/
	static int zfs_scrub_after_expand = 1;

	static void
	vdev_raidz_row_free(raidz_row_t *rr)
	{
	for (int c = 0; c < rr->rr_cols; c++) {
	raidz_col_t *rc = &rr->rr_col[c];

	if (rc->rc_size != 0)
	abd_free(rc->rc_abd);
	if (rc->rc_orig_data != NULL)
	abd_free(rc->rc_orig_data);
	}

	if (rr->rr_abd_empty != NULL)
	abd_free(rr->rr_abd_empty);

	kmem_free(rr, offsetof(raidz_row_t, rr_col[rr->rr_scols]));
	}

	void
	vdev_raidz_map_free(raidz_map_t *rm)
	{
	for (int i = 0; i < rm->rm_nrows; i++)
	vdev_raidz_row_free(rm->rm_row[i]);

	if (rm->rm_nphys_cols) {
	for (int i = 0; i < rm->rm_nphys_cols; i++) {
	if (rm->rm_phys_col[i].rc_abd != NULL)
	abd_free(rm->rm_phys_col[i].rc_abd);
	}

	kmem_free(rm->rm_phys_col, sizeof (raidz_col_t) *
	rm->rm_nphys_cols);
	}

	ASSERT3P(rm->rm_lr, ==, NULL);
	kmem_free(rm, offsetof(raidz_map_t, rm_row[rm->rm_nrows]));
	}

	static void
	vdev_raidz_map_free_vsd(zio_t *zio)
	{
	raidz_map_t *rm = zio->io_vsd;

	vdev_raidz_map_free(rm);
	}

	static int
	vdev_raidz_reflow_compare(const void x1, const void x2)
	{
	const reflow_node_t *l = x1;
	const reflow_node_t *r = x2;

	return (TREE_CMP(l->re_txg, r->re_txg));
	}

	const zio_vsd_ops_t vdev_raidz_vsd_ops = {
	.vsd_free = vdev_raidz_map_free_vsd,
	};

	raidz_row_t *
	vdev_raidz_row_alloc(int cols)
	{
	raidz_row_t *rr =
	kmem_zalloc(offsetof(raidz_row_t, rr_col[cols]), KM_SLEEP);

	rr->rr_cols = cols;
	rr->rr_scols = cols;

	for (int c = 0; c < cols; c++) {
	raidz_col_t *rc = &rr->rr_col[c];
	rc->rc_shadow_devidx = INT_MAX;
	rc->rc_shadow_offset = UINT64_MAX;
	rc->rc_allow_repair = 1;
	}
	return (rr);
	}

	static void
	vdev_raidz_map_alloc_write(zio_t zio, raidz_map_t rm, uint64_t ashift)
	{
	int c;
	int nwrapped = 0;
	uint64_t off = 0;
	raidz_row_t *rr = rm->rm_row[0];

	ASSERT3U(zio->io_type, ==, ZIO_TYPE_WRITE);
	ASSERT3U(rm->rm_nrows, ==, 1);

	/*
	* Pad any parity columns with additional space to account for skip
	* sectors.
	*/
	if (rm->rm_skipstart < rr->rr_firstdatacol) {
	ASSERT0(rm->rm_skipstart);
	nwrapped = rm->rm_nskip;
	} else if (rr->rr_scols < (rm->rm_skipstart + rm->rm_nskip)) {
	nwrapped =
	(rm->rm_skipstart + rm->rm_nskip) % rr->rr_scols;
	}

	/*
	* Optional single skip sectors (rc_size == 0) will be handled in
	* vdev_raidz_io_start_write().
	*/
	int skipped = rr->rr_scols - rr->rr_cols;

	/* Allocate buffers for the parity columns */
	for (c = 0; c < rr->rr_firstdatacol; c++) {
	raidz_col_t *rc = &rr->rr_col[c];

	/*
	* Parity columns will pad out a linear ABD to account for
	* the skip sector. A linear ABD is used here because
	* parity calculations use the ABD buffer directly to calculate
	* parity. This avoids doing a memcpy back to the ABD after the
	* parity has been calculated. By issuing the parity column
	* with the skip sector we can reduce contention on the child
	* VDEV queue locks (vq_lock).
	*/
	if (c < nwrapped) {
	rc->rc_abd = abd_alloc_linear(
	rc->rc_size + (1ULL << ashift), B_FALSE);
	abd_zero_off(rc->rc_abd, rc->rc_size, 1ULL << ashift);
	skipped++;
	} else {
	rc->rc_abd = abd_alloc_linear(rc->rc_size, B_FALSE);
	}
	}

	for (off = 0; c < rr->rr_cols; c++) {
	raidz_col_t *rc = &rr->rr_col[c];
	abd_t *abd = abd_get_offset_struct(&rc->rc_abdstruct,
	zio->io_abd, off, rc->rc_size);

	/*
	* Generate I/O for skip sectors to improve aggregation
	* continuity. We will use gang ABD's to reduce contention
	* on the child VDEV queue locks (vq_lock) by issuing
	* a single I/O that contains the data and skip sector.
	*
	* It is important to make sure that rc_size is not updated
	* even though we are adding a skip sector to the ABD. When
	* calculating the parity in vdev_raidz_generate_parity_row()
	* the rc_size is used to iterate through the ABD's. We can
	* not have zero'd out skip sectors used for calculating
	* parity for raidz, because those same sectors are not used
	* during reconstruction.
	*/
	if (c >= rm->rm_skipstart && skipped < rm->rm_nskip) {
	rc->rc_abd = abd_alloc_gang();
	abd_gang_add(rc->rc_abd, abd, B_TRUE);
	abd_gang_add(rc->rc_abd,
	abd_get_zeros(1ULL << ashift), B_TRUE);
	skipped++;
	} else {
	rc->rc_abd = abd;
	}
	off += rc->rc_size;
	}

	ASSERT3U(off, ==, zio->io_size);
	ASSERT3S(skipped, ==, rm->rm_nskip);
	}

	static void
	vdev_raidz_map_alloc_read(zio_t zio, raidz_map_t rm)
	{
	int c;
	raidz_row_t *rr = rm->rm_row[0];

	ASSERT3U(rm->rm_nrows, ==, 1);

	/* Allocate buffers for the parity columns */
	for (c = 0; c < rr->rr_firstdatacol; c++)
	rr->rr_col[c].rc_abd =
	abd_alloc_linear(rr->rr_col[c].rc_size, B_FALSE);

	for (uint64_t off = 0; c < rr->rr_cols; c++) {
	raidz_col_t *rc = &rr->rr_col[c];
	rc->rc_abd = abd_get_offset_struct(&rc->rc_abdstruct,
	zio->io_abd, off, rc->rc_size);
	off += rc->rc_size;
	}
	}

	/*
	* Divides the IO evenly across all child vdevs; usually, dcols is
	* the number of children in the target vdev.
	*
	* Avoid inlining the function to keep vdev_raidz_io_start(), which
	* is this functions only caller, as small as possible on the stack.
	*/
	noinline raidz_map_t *
	vdev_raidz_map_alloc(zio_t *zio, uint64_t ashift, uint64_t dcols,
	uint64_t nparity)
	{
	raidz_row_t *rr;
	/* The starting RAIDZ (parent) vdev sector of the block. */
	uint64_t b = zio->io_offset >> ashift;
	/* The zio's size in units of the vdev's minimum sector size. */
	uint64_t s = zio->io_size >> ashift;
	/* The first column for this stripe. */
	uint64_t f = b % dcols;
	/* The starting byte offset on each child vdev. */
	uint64_t o = (b / dcols) << ashift;
	uint64_t acols, scols;

	raidz_map_t *rm =
	kmem_zalloc(offsetof(raidz_map_t, rm_row[1]), KM_SLEEP);
	rm->rm_nrows = 1;

	/*
	* "Quotient": The number of data sectors for this stripe on all but
	* the "big column" child vdevs that also contain "remainder" data.
	*/
	uint64_t q = s / (dcols - nparity);

	/*
	* "Remainder": The number of partial stripe data sectors in this I/O.
	* This will add a sector to some, but not all, child vdevs.
	*/
	uint64_t r = s - q * (dcols - nparity);

	/* The number of "big columns" - those which contain remainder data. */
	uint64_t bc = (r == 0 ? 0 : r + nparity);

	/*
	* The total number of data and parity sectors associated with
	* this I/O.
	*/
	uint64_t tot = s + nparity * (q + (r == 0 ? 0 : 1));

	/*
	* acols: The columns that will be accessed.
	* scols: The columns that will be accessed or skipped.
	*/
	if (q == 0) {
	/* Our I/O request doesn't span all child vdevs. */
	acols = bc;
	scols = MIN(dcols, roundup(bc, nparity + 1));
	} else {
	acols = dcols;
	scols = dcols;
	}

	ASSERT3U(acols, <=, scols);
	rr = vdev_raidz_row_alloc(scols);
	rm->rm_row[0] = rr;
	rr->rr_cols = acols;
	rr->rr_bigcols = bc;
	rr->rr_firstdatacol = nparity;
	#ifdef ZFS_DEBUG
	rr->rr_offset = zio->io_offset;
	rr->rr_size = zio->io_size;
	#endif

	uint64_t asize = 0;

	for (uint64_t c = 0; c < scols; c++) {
	raidz_col_t *rc = &rr->rr_col[c];
	uint64_t col = f + c;
	uint64_t coff = o;
	if (col >= dcols) {
	col -= dcols;
	coff += 1ULL << ashift;
	}
	rc->rc_devidx = col;
	rc->rc_offset = coff;

	if (c >= acols)
	rc->rc_size = 0;
	else if (c < bc)
	rc->rc_size = (q + 1) << ashift;
	else
	rc->rc_size = q << ashift;

	asize += rc->rc_size;
	}

	ASSERT3U(asize, ==, tot << ashift);
	rm->rm_nskip = roundup(tot, nparity + 1) - tot;
	rm->rm_skipstart = bc;

	/*
	* If all data stored spans all columns, there's a danger that parity
	* will always be on the same device and, since parity isn't read
	* during normal operation, that device's I/O bandwidth won't be
	* used effectively. We therefore switch the parity every 1MB.
	*
	* ... at least that was, ostensibly, the theory. As a practical
	* matter unless we juggle the parity between all devices evenly, we
	* won't see any benefit. Further, occasional writes that aren't a
	* multiple of the LCM of the number of children and the minimum
	* stripe width are sufficient to avoid pessimal behavior.
	* Unfortunately, this decision created an implicit on-disk format
	* requirement that we need to support for all eternity, but only
	* for single-parity RAID-Z.
	*
	* If we intend to skip a sector in the zeroth column for padding
	* we must make sure to note this swap. We will never intend to
	* skip the first column since at least one data and one parity
	* column must appear in each row.
	*/
	ASSERT(rr->rr_cols >= 2);
	ASSERT(rr->rr_col[0].rc_size == rr->rr_col[1].rc_size);

	if (rr->rr_firstdatacol == 1 && (zio->io_offset & (1ULL << 20))) {
	uint64_t devidx = rr->rr_col[0].rc_devidx;
	o = rr->rr_col[0].rc_offset;
	rr->rr_col[0].rc_devidx = rr->rr_col[1].rc_devidx;
	rr->rr_col[0].rc_offset = rr->rr_col[1].rc_offset;
	rr->rr_col[1].rc_devidx = devidx;
	rr->rr_col[1].rc_offset = o;
	if (rm->rm_skipstart == 0)
	rm->rm_skipstart = 1;
	}

	if (zio->io_type == ZIO_TYPE_WRITE) {
	vdev_raidz_map_alloc_write(zio, rm, ashift);
	} else {
	vdev_raidz_map_alloc_read(zio, rm);
	}
	/* init RAIDZ parity ops */
	rm->rm_ops = vdev_raidz_math_get_ops();

	return (rm);
	}

	/*
	* Everything before reflow_offset_synced should have been moved to the new
	* location (read and write completed). However, this may not yet be reflected
	* in the on-disk format (e.g. raidz_reflow_sync() has been called but the
	* uberblock has not yet been written). If reflow is not in progress,
	* reflow_offset_synced should be UINT64_MAX. For each row, if the row is
	* entirely before reflow_offset_synced, it will come from the new location.
	* Otherwise this row will come from the old location. Therefore, rows that
	* straddle the reflow_offset_synced will come from the old location.
	*
	* For writes, reflow_offset_next is the next offset to copy. If a sector has
	* been copied, but not yet reflected in the on-disk progress
	* (reflow_offset_synced), it will also be written to the new (already copied)
	* offset.
	*/
	noinline raidz_map_t *
	vdev_raidz_map_alloc_expanded(zio_t *zio,
	uint64_t ashift, uint64_t physical_cols, uint64_t logical_cols,
	uint64_t nparity, uint64_t reflow_offset_synced,
	uint64_t reflow_offset_next, boolean_t use_scratch)
	{
	abd_t *abd = zio->io_abd;
	uint64_t offset = zio->io_offset;
	uint64_t size = zio->io_size;

	/* The zio's size in units of the vdev's minimum sector size. */
	uint64_t s = size >> ashift;

	/*
	* "Quotient": The number of data sectors for this stripe on all but
	* the "big column" child vdevs that also contain "remainder" data.
	* AKA "full rows"
	*/
	uint64_t q = s / (logical_cols - nparity);

	/*
	* "Remainder": The number of partial stripe data sectors in this I/O.
	* This will add a sector to some, but not all, child vdevs.
	*/
	uint64_t r = s - q * (logical_cols - nparity);

	/* The number of "big columns" - those which contain remainder data. */
	uint64_t bc = (r == 0 ? 0 : r + nparity);

	/*
	* The total number of data and parity sectors associated with
	* this I/O.
	*/
	uint64_t tot = s + nparity * (q + (r == 0 ? 0 : 1));

	/* How many rows contain data (not skip) */
	uint64_t rows = howmany(tot, logical_cols);
	int cols = MIN(tot, logical_cols);

	raidz_map_t *rm =
	kmem_zalloc(offsetof(raidz_map_t, rm_row[rows]),
	KM_SLEEP);
	rm->rm_nrows = rows;
	rm->rm_nskip = roundup(tot, nparity + 1) - tot;
	rm->rm_skipstart = bc;
	uint64_t asize = 0;

	for (uint64_t row = 0; row < rows; row++) {
	boolean_t row_use_scratch = B_FALSE;
	raidz_row_t *rr = vdev_raidz_row_alloc(cols);
	rm->rm_row[row] = rr;

	/* The starting RAIDZ (parent) vdev sector of the row. */
	uint64_t b = (offset >> ashift) + row * logical_cols;

	/*
	* If we are in the middle of a reflow, and the copying has
	* not yet completed for any part of this row, then use the
	* old location of this row. Note that reflow_offset_synced
	* reflects the i/o that's been completed, because it's
	* updated by a synctask, after zio_wait(spa_txg_zio[]).
	* This is sufficient for our check, even if that progress
	* has not yet been recorded to disk (reflected in
	* spa_ubsync). Also note that we consider the last row to
	* be "full width" (`cols`-wide rather than `bc`-wide) for
	* this calculation. This causes a tiny bit of unnecessary
	* double-writes but is safe and simpler to calculate.
	*/
	int row_phys_cols = physical_cols;
	if (b + cols > reflow_offset_synced >> ashift)
	row_phys_cols--;
	else if (use_scratch)
	row_use_scratch = B_TRUE;

	/* starting child of this row */
	uint64_t child_id = b % row_phys_cols;
	/* The starting byte offset on each child vdev. */
	uint64_t child_offset = (b / row_phys_cols) << ashift;

	/*
	* Note, rr_cols is the entire width of the block, even
	* if this row is shorter. This is needed because parity
	* generation (for Q and R) needs to know the entire width,
	* because it treats the short row as though it was
	* full-width (and the "phantom" sectors were zero-filled).
	*
	* Another approach to this would be to set cols shorter
	* (to just the number of columns that we might do i/o to)
	* and have another mechanism to tell the parity generation
	* about the "entire width". Reconstruction (at least
	* vdev_raidz_reconstruct_general()) would also need to
	* know about the "entire width".
	*/
	rr->rr_firstdatacol = nparity;
	#ifdef ZFS_DEBUG
	/*
	* note: rr_size is PSIZE, not ASIZE
	*/
	rr->rr_offset = b << ashift;
	rr->rr_size = (rr->rr_cols - rr->rr_firstdatacol) << ashift;
	#endif

	for (int c = 0; c < rr->rr_cols; c++, child_id++) {
	if (child_id >= row_phys_cols) {
	child_id -= row_phys_cols;
	child_offset += 1ULL << ashift;
	}
	raidz_col_t *rc = &rr->rr_col[c];
	rc->rc_devidx = child_id;
	rc->rc_offset = child_offset;

	/*
	* Get this from the scratch space if appropriate.
	* This only happens if we crashed in the middle of
	* raidz_reflow_scratch_sync() (while it's running,
	* the rangelock prevents us from doing concurrent
	* io), and even then only during zpool import or
	* when the pool is imported readonly.
	*/
	if (row_use_scratch)
	rc->rc_offset -= VDEV_BOOT_SIZE;

	uint64_t dc = c - rr->rr_firstdatacol;
	if (c < rr->rr_firstdatacol) {
	rc->rc_size = 1ULL << ashift;

	/*
	* Parity sectors' rc_abd's are set below
	* after determining if this is an aggregation.
	*/
	} else if (row == rows - 1 && bc != 0 && c >= bc) {
	/*
	* Past the end of the block (even including
	* skip sectors). This sector is part of the
	* map so that we have full rows for p/q parity
	* generation.
	*/
	rc->rc_size = 0;
	rc->rc_abd = NULL;
	} else {
	/* "data column" (col excluding parity) */
	uint64_t off;

	if (c < bc \|\| r == 0) {
	off = dc * rows + row;
	} else {
	off = r * rows +
	(dc - r) * (rows - 1) + row;
	}
	rc->rc_size = 1ULL << ashift;
	rc->rc_abd = abd_get_offset_struct(
	&rc->rc_abdstruct, abd, off << ashift,
	rc->rc_size);
	}

	if (rc->rc_size == 0)
	continue;

	/*
	* If any part of this row is in both old and new
	* locations, the primary location is the old
	* location. If this sector was already copied to the
	* new location, we need to also write to the new,
	* "shadow" location.
	*
	* Note, `row_phys_cols != physical_cols` indicates
	* that the primary location is the old location.
	* `b+c < reflow_offset_next` indicates that the copy
	* to the new location has been initiated. We know
	* that the copy has completed because we have the
	* rangelock, which is held exclusively while the
	* copy is in progress.
	*/
	if (row_use_scratch \|\|
	(row_phys_cols != physical_cols &&
	b + c < reflow_offset_next >> ashift)) {
	rc->rc_shadow_devidx = (b + c) % physical_cols;
	rc->rc_shadow_offset =
	((b + c) / physical_cols) << ashift;
	if (row_use_scratch)
	rc->rc_shadow_offset -= VDEV_BOOT_SIZE;
	}

	asize += rc->rc_size;
	}

	/*
	* See comment in vdev_raidz_map_alloc()
	*/
	if (rr->rr_firstdatacol == 1 && rr->rr_cols > 1 &&
	(offset & (1ULL << 20))) {
	ASSERT(rr->rr_cols >= 2);
	ASSERT(rr->rr_col[0].rc_size == rr->rr_col[1].rc_size);

	int devidx0 = rr->rr_col[0].rc_devidx;
	uint64_t offset0 = rr->rr_col[0].rc_offset;
	int shadow_devidx0 = rr->rr_col[0].rc_shadow_devidx;
	uint64_t shadow_offset0 =
	rr->rr_col[0].rc_shadow_offset;

	rr->rr_col[0].rc_devidx = rr->rr_col[1].rc_devidx;
	rr->rr_col[0].rc_offset = rr->rr_col[1].rc_offset;
	rr->rr_col[0].rc_shadow_devidx =
	rr->rr_col[1].rc_shadow_devidx;
	rr->rr_col[0].rc_shadow_offset =
	rr->rr_col[1].rc_shadow_offset;

	rr->rr_col[1].rc_devidx = devidx0;
	rr->rr_col[1].rc_offset = offset0;
	rr->rr_col[1].rc_shadow_devidx = shadow_devidx0;
	rr->rr_col[1].rc_shadow_offset = shadow_offset0;
	}
	}
	ASSERT3U(asize, ==, tot << ashift);

	/*
	* Determine if the block is contiguous, in which case we can use
	* an aggregation.
	*/
	if (rows >= raidz_io_aggregate_rows) {
	rm->rm_nphys_cols = physical_cols;
	rm->rm_phys_col =
	kmem_zalloc(sizeof (raidz_col_t) * rm->rm_nphys_cols,
	KM_SLEEP);

	/*
	* Determine the aggregate io's offset and size, and check
	* that the io is contiguous.
	*/
	for (int i = 0;
	i < rm->rm_nrows && rm->rm_phys_col != NULL; i++) {
	raidz_row_t *rr = rm->rm_row[i];
	for (int c = 0; c < rr->rr_cols; c++) {
	raidz_col_t *rc = &rr->rr_col[c];
	raidz_col_t *prc =
	&rm->rm_phys_col[rc->rc_devidx];

	if (rc->rc_size == 0)
	continue;

	if (prc->rc_size == 0) {
	ASSERT0(prc->rc_offset);
	prc->rc_offset = rc->rc_offset;
	} else if (prc->rc_offset + prc->rc_size !=
	rc->rc_offset) {
	/*
	* This block is not contiguous and
	* therefore can't be aggregated.
	* This is expected to be rare, so
	* the cost of allocating and then
	* freeing rm_phys_col is not
	* significant.
	*/
	kmem_free(rm->rm_phys_col,
	sizeof (raidz_col_t) *
	rm->rm_nphys_cols);
	rm->rm_phys_col = NULL;
	rm->rm_nphys_cols = 0;
	break;
	}
	prc->rc_size += rc->rc_size;
	}
	}
	}
	if (rm->rm_phys_col != NULL) {
	/*
	* Allocate aggregate ABD's.
	*/
	for (int i = 0; i < rm->rm_nphys_cols; i++) {
	raidz_col_t *prc = &rm->rm_phys_col[i];

	prc->rc_devidx = i;

	if (prc->rc_size == 0)
	continue;

	prc->rc_abd =
	abd_alloc_linear(rm->rm_phys_col[i].rc_size,
	B_FALSE);
	}

	/*
	* Point the parity abd's into the aggregate abd's.
	*/
	for (int i = 0; i < rm->rm_nrows; i++) {
	raidz_row_t *rr = rm->rm_row[i];
	for (int c = 0; c < rr->rr_firstdatacol; c++) {
	raidz_col_t *rc = &rr->rr_col[c];
	raidz_col_t *prc =
	&rm->rm_phys_col[rc->rc_devidx];
	rc->rc_abd =
	abd_get_offset_struct(&rc->rc_abdstruct,
	prc->rc_abd,
	rc->rc_offset - prc->rc_offset,
	rc->rc_size);
	}
	}
	} else {
	/*
	* Allocate new abd's for the parity sectors.
	*/
	for (int i = 0; i < rm->rm_nrows; i++) {
	raidz_row_t *rr = rm->rm_row[i];
	for (int c = 0; c < rr->rr_firstdatacol; c++) {
	raidz_col_t *rc = &rr->rr_col[c];
	rc->rc_abd =
	abd_alloc_linear(rc->rc_size,
	B_TRUE);
	}
	}
	}
	/* init RAIDZ parity ops */
	rm->rm_ops = vdev_raidz_math_get_ops();

	return (rm);
	}

	struct pqr_struct {
	uint64_t *p;
	uint64_t *q;
	uint64_t *r;
	};

	static int
	vdev_raidz_p_func(void buf, size_t size, void private)
	{
	struct pqr_struct *pqr = private;
	const uint64_t *src = buf;
	int cnt = size / sizeof (src[0]);

	ASSERT(pqr->p && !pqr->q && !pqr->r);

	for (int i = 0; i < cnt; i++, src++, pqr->p++)
	pqr->p ^= src;

	return (0);
	}

	static int
	vdev_raidz_pq_func(void buf, size_t size, void private)
	{
	struct pqr_struct *pqr = private;
	const uint64_t *src = buf;
	uint64_t mask;
	int cnt = size / sizeof (src[0]);

	ASSERT(pqr->p && pqr->q && !pqr->r);

	for (int i = 0; i < cnt; i++, src++, pqr->p++, pqr->q++) {
	pqr->p ^= src;
	VDEV_RAIDZ_64MUL_2(*pqr->q, mask);
	pqr->q ^= src;
	}

	return (0);
	}

	static int
	vdev_raidz_pqr_func(void buf, size_t size, void private)
	{
	struct pqr_struct *pqr = private;
	const uint64_t *src = buf;
	uint64_t mask;
	int cnt = size / sizeof (src[0]);

	ASSERT(pqr->p && pqr->q && pqr->r);

	for (int i = 0; i < cnt; i++, src++, pqr->p++, pqr->q++, pqr->r++) {
	pqr->p ^= src;
	VDEV_RAIDZ_64MUL_2(*pqr->q, mask);
	pqr->q ^= src;
	VDEV_RAIDZ_64MUL_4(*pqr->r, mask);
	pqr->r ^= src;
	}

	return (0);
	}

	static void
	vdev_raidz_generate_parity_p(raidz_row_t *rr)
	{
	uint64_t *p = abd_to_buf(rr->rr_col[VDEV_RAIDZ_P].rc_abd);

	for (int c = rr->rr_firstdatacol; c < rr->rr_cols; c++) {
	abd_t *src = rr->rr_col[c].rc_abd;

	if (c == rr->rr_firstdatacol) {
	abd_copy_to_buf(p, src, rr->rr_col[c].rc_size);
	} else {
	struct pqr_struct pqr = { p, NULL, NULL };
	(void) abd_iterate_func(src, 0, rr->rr_col[c].rc_size,
	vdev_raidz_p_func, &pqr);
	}
	}
	}

	static void
	vdev_raidz_generate_parity_pq(raidz_row_t *rr)
	{
	uint64_t *p = abd_to_buf(rr->rr_col[VDEV_RAIDZ_P].rc_abd);
	uint64_t *q = abd_to_buf(rr->rr_col[VDEV_RAIDZ_Q].rc_abd);
	uint64_t pcnt = rr->rr_col[VDEV_RAIDZ_P].rc_size / sizeof (p[0]);
	ASSERT(rr->rr_col[VDEV_RAIDZ_P].rc_size ==
	rr->rr_col[VDEV_RAIDZ_Q].rc_size);

	for (int c = rr->rr_firstdatacol; c < rr->rr_cols; c++) {
	abd_t *src = rr->rr_col[c].rc_abd;

	uint64_t ccnt = rr->rr_col[c].rc_size / sizeof (p[0]);

	if (c == rr->rr_firstdatacol) {
	ASSERT(ccnt == pcnt \|\| ccnt == 0);
	abd_copy_to_buf(p, src, rr->rr_col[c].rc_size);
	(void) memcpy(q, p, rr->rr_col[c].rc_size);

	for (uint64_t i = ccnt; i < pcnt; i++) {
	p[i] = 0;
	q[i] = 0;
	}
	} else {
	struct pqr_struct pqr = { p, q, NULL };

	ASSERT(ccnt <= pcnt);
	(void) abd_iterate_func(src, 0, rr->rr_col[c].rc_size,
	vdev_raidz_pq_func, &pqr);

	/*
	* Treat short columns as though they are full of 0s.
	* Note that there's therefore nothing needed for P.
	*/
	uint64_t mask;
	for (uint64_t i = ccnt; i < pcnt; i++) {
	VDEV_RAIDZ_64MUL_2(q[i], mask);
	}
	}
	}
	}

	static void
	vdev_raidz_generate_parity_pqr(raidz_row_t *rr)
	{
	uint64_t *p = abd_to_buf(rr->rr_col[VDEV_RAIDZ_P].rc_abd);
	uint64_t *q = abd_to_buf(rr->rr_col[VDEV_RAIDZ_Q].rc_abd);
	uint64_t *r = abd_to_buf(rr->rr_col[VDEV_RAIDZ_R].rc_abd);
	uint64_t pcnt = rr->rr_col[VDEV_RAIDZ_P].rc_size / sizeof (p[0]);
	ASSERT(rr->rr_col[VDEV_RAIDZ_P].rc_size ==
	rr->rr_col[VDEV_RAIDZ_Q].rc_size);
	ASSERT(rr->rr_col[VDEV_RAIDZ_P].rc_size ==
	rr->rr_col[VDEV_RAIDZ_R].rc_size);

	for (int c = rr->rr_firstdatacol; c < rr->rr_cols; c++) {
	abd_t *src = rr->rr_col[c].rc_abd;

	uint64_t ccnt = rr->rr_col[c].rc_size / sizeof (p[0]);

	if (c == rr->rr_firstdatacol) {
	ASSERT(ccnt == pcnt \|\| ccnt == 0);
	abd_copy_to_buf(p, src, rr->rr_col[c].rc_size);
	(void) memcpy(q, p, rr->rr_col[c].rc_size);
	(void) memcpy(r, p, rr->rr_col[c].rc_size);

	for (uint64_t i = ccnt; i < pcnt; i++) {
	p[i] = 0;
	q[i] = 0;
	r[i] = 0;
	}
	} else {
	struct pqr_struct pqr = { p, q, r };

	ASSERT(ccnt <= pcnt);
	(void) abd_iterate_func(src, 0, rr->rr_col[c].rc_size,
	vdev_raidz_pqr_func, &pqr);

	/*
	* Treat short columns as though they are full of 0s.
	* Note that there's therefore nothing needed for P.
	*/
	uint64_t mask;
	for (uint64_t i = ccnt; i < pcnt; i++) {
	VDEV_RAIDZ_64MUL_2(q[i], mask);
	VDEV_RAIDZ_64MUL_4(r[i], mask);
	}
	}
	}
	}

	/*
	* Generate RAID parity in the first virtual columns according to the number of
	* parity columns available.
	*/
	void
	vdev_raidz_generate_parity_row(raidz_map_t rm, raidz_row_t rr)
	{
	if (rr->rr_cols == 0) {
	/*
	* We are handling this block one row at a time (because
	* this block has a different logical vs physical width,
	* due to RAIDZ expansion), and this is a pad-only row,
	* which has no parity.
	*/
	return;
	}

	/* Generate using the new math implementation */
	if (vdev_raidz_math_generate(rm, rr) != RAIDZ_ORIGINAL_IMPL)
	return;

	switch (rr->rr_firstdatacol) {
	case 1:
	vdev_raidz_generate_parity_p(rr);
	break;
	case 2:
	vdev_raidz_generate_parity_pq(rr);
	break;
	case 3:
	vdev_raidz_generate_parity_pqr(rr);
	break;
	default:
	cmn_err(CE_PANIC, "invalid RAID-Z configuration");
	}
	}

	void
	vdev_raidz_generate_parity(raidz_map_t *rm)
	{
	for (int i = 0; i < rm->rm_nrows; i++) {
	raidz_row_t *rr = rm->rm_row[i];
	vdev_raidz_generate_parity_row(rm, rr);
	}
	}

	static int
	vdev_raidz_reconst_p_func(void dbuf, void sbuf, size_t size, void *private)
	{
	(void) private;
	uint64_t *dst = dbuf;
	uint64_t *src = sbuf;
	int cnt = size / sizeof (src[0]);

	for (int i = 0; i < cnt; i++) {
	dst[i] ^= src[i];
	}

	return (0);
	}

	static int
	vdev_raidz_reconst_q_pre_func(void dbuf, void sbuf, size_t size,
	void *private)
	{
	(void) private;
	uint64_t *dst = dbuf;
	uint64_t *src = sbuf;
	uint64_t mask;
	int cnt = size / sizeof (dst[0]);

	for (int i = 0; i < cnt; i++, dst++, src++) {
	VDEV_RAIDZ_64MUL_2(*dst, mask);
	dst ^= src;
	}

	return (0);
	}

	static int
	vdev_raidz_reconst_q_pre_tail_func(void buf, size_t size, void private)
	{
	(void) private;
	uint64_t *dst = buf;
	uint64_t mask;
	int cnt = size / sizeof (dst[0]);

	for (int i = 0; i < cnt; i++, dst++) {
	/* same operation as vdev_raidz_reconst_q_pre_func() on dst */
	VDEV_RAIDZ_64MUL_2(*dst, mask);
	}

	return (0);
	}

	struct reconst_q_struct {
	uint64_t *q;
	int exp;
	};

	static int
	vdev_raidz_reconst_q_post_func(void buf, size_t size, void private)
	{
	struct reconst_q_struct *rq = private;
	uint64_t *dst = buf;
	int cnt = size / sizeof (dst[0]);

	for (int i = 0; i < cnt; i++, dst++, rq->q++) {
	int j;
	uint8_t *b;

	dst ^= rq->q;
	for (j = 0, b = (uint8_t *)dst; j < 8; j++, b++) {
	b = vdev_raidz_exp2(b, rq->exp);
	}
	}

	return (0);
	}

	struct reconst_pq_struct {
	uint8_t *p;
	uint8_t *q;
	uint8_t *pxy;
	uint8_t *qxy;
	int aexp;
	int bexp;
	};

	static int
	vdev_raidz_reconst_pq_func(void xbuf, void ybuf, size_t size, void *private)
	{
	struct reconst_pq_struct *rpq = private;
	uint8_t *xd = xbuf;
	uint8_t *yd = ybuf;

	for (int i = 0; i < size;
	i++, rpq->p++, rpq->q++, rpq->pxy++, rpq->qxy++, xd++, yd++) {
	xd = vdev_raidz_exp2(rpq->p ^ *rpq->pxy, rpq->aexp) ^
	vdev_raidz_exp2(rpq->q ^ rpq->qxy, rpq->bexp);
	yd = rpq->p ^ rpq->pxy ^ xd;
	}

	return (0);
	}

	static int
	vdev_raidz_reconst_pq_tail_func(void xbuf, size_t size, void private)
	{
	struct reconst_pq_struct *rpq = private;
	uint8_t *xd = xbuf;

	for (int i = 0; i < size;
	i++, rpq->p++, rpq->q++, rpq->pxy++, rpq->qxy++, xd++) {
	/* same operation as vdev_raidz_reconst_pq_func() on xd */
	xd = vdev_raidz_exp2(rpq->p ^ *rpq->pxy, rpq->aexp) ^
	vdev_raidz_exp2(rpq->q ^ rpq->qxy, rpq->bexp);
	}

	return (0);
	}

	static void
	vdev_raidz_reconstruct_p(raidz_row_t rr, int tgts, int ntgts)
	{
	int x = tgts[0];
	abd_t dst, src;

	if (zfs_flags & ZFS_DEBUG_RAIDZ_RECONSTRUCT)
	zfs_dbgmsg("reconstruct_p(rm=%px x=%u)", rr, x);

	ASSERT3U(ntgts, ==, 1);
	ASSERT3U(x, >=, rr->rr_firstdatacol);
	ASSERT3U(x, <, rr->rr_cols);

	ASSERT3U(rr->rr_col[x].rc_size, <=, rr->rr_col[VDEV_RAIDZ_P].rc_size);

	src = rr->rr_col[VDEV_RAIDZ_P].rc_abd;
	dst = rr->rr_col[x].rc_abd;

	abd_copy_from_buf(dst, abd_to_buf(src), rr->rr_col[x].rc_size);

	for (int c = rr->rr_firstdatacol; c < rr->rr_cols; c++) {
	uint64_t size = MIN(rr->rr_col[x].rc_size,
	rr->rr_col[c].rc_size);

	src = rr->rr_col[c].rc_abd;

	if (c == x)
	continue;

	(void) abd_iterate_func2(dst, src, 0, 0, size,
	vdev_raidz_reconst_p_func, NULL);
	}
	}

	static void
	vdev_raidz_reconstruct_q(raidz_row_t rr, int tgts, int ntgts)
	{
	int x = tgts[0];
	int c, exp;
	abd_t dst, src;

	if (zfs_flags & ZFS_DEBUG_RAIDZ_RECONSTRUCT)
	zfs_dbgmsg("reconstruct_q(rm=%px x=%u)", rr, x);

	ASSERT(ntgts == 1);

	ASSERT(rr->rr_col[x].rc_size <= rr->rr_col[VDEV_RAIDZ_Q].rc_size);

	for (c = rr->rr_firstdatacol; c < rr->rr_cols; c++) {
	uint64_t size = (c == x) ? 0 : MIN(rr->rr_col[x].rc_size,
	rr->rr_col[c].rc_size);

	src = rr->rr_col[c].rc_abd;
	dst = rr->rr_col[x].rc_abd;

	if (c == rr->rr_firstdatacol) {
	abd_copy(dst, src, size);
	if (rr->rr_col[x].rc_size > size) {
	abd_zero_off(dst, size,
	rr->rr_col[x].rc_size - size);
	}
	} else {
	ASSERT3U(size, <=, rr->rr_col[x].rc_size);
	(void) abd_iterate_func2(dst, src, 0, 0, size,
	vdev_raidz_reconst_q_pre_func, NULL);
	(void) abd_iterate_func(dst,
	size, rr->rr_col[x].rc_size - size,
	vdev_raidz_reconst_q_pre_tail_func, NULL);
	}
	}

	src = rr->rr_col[VDEV_RAIDZ_Q].rc_abd;
	dst = rr->rr_col[x].rc_abd;
	exp = 255 - (rr->rr_cols - 1 - x);

	struct reconst_q_struct rq = { abd_to_buf(src), exp };
	(void) abd_iterate_func(dst, 0, rr->rr_col[x].rc_size,
	vdev_raidz_reconst_q_post_func, &rq);
	}

	static void
	vdev_raidz_reconstruct_pq(raidz_row_t rr, int tgts, int ntgts)
	{
	uint8_t p, q, pxy, qxy, tmp, a, b, aexp, bexp;
	abd_t pdata, qdata;
	uint64_t xsize, ysize;
	int x = tgts[0];
	int y = tgts[1];
	abd_t xd, yd;

	if (zfs_flags & ZFS_DEBUG_RAIDZ_RECONSTRUCT)
	zfs_dbgmsg("reconstruct_pq(rm=%px x=%u y=%u)", rr, x, y);

	ASSERT(ntgts == 2);
	ASSERT(x < y);
	ASSERT(x >= rr->rr_firstdatacol);
	ASSERT(y < rr->rr_cols);

	ASSERT(rr->rr_col[x].rc_size >= rr->rr_col[y].rc_size);

	/*
	* Move the parity data aside -- we're going to compute parity as
	* though columns x and y were full of zeros -- Pxy and Qxy. We want to
	* reuse the parity generation mechanism without trashing the actual
	* parity so we make those columns appear to be full of zeros by
	* setting their lengths to zero.
	*/
	pdata = rr->rr_col[VDEV_RAIDZ_P].rc_abd;
	qdata = rr->rr_col[VDEV_RAIDZ_Q].rc_abd;
	xsize = rr->rr_col[x].rc_size;
	ysize = rr->rr_col[y].rc_size;

	rr->rr_col[VDEV_RAIDZ_P].rc_abd =
	abd_alloc_linear(rr->rr_col[VDEV_RAIDZ_P].rc_size, B_TRUE);
	rr->rr_col[VDEV_RAIDZ_Q].rc_abd =
	abd_alloc_linear(rr->rr_col[VDEV_RAIDZ_Q].rc_size, B_TRUE);
	rr->rr_col[x].rc_size = 0;
	rr->rr_col[y].rc_size = 0;

	vdev_raidz_generate_parity_pq(rr);

	rr->rr_col[x].rc_size = xsize;
	rr->rr_col[y].rc_size = ysize;

	p = abd_to_buf(pdata);
	q = abd_to_buf(qdata);
	pxy = abd_to_buf(rr->rr_col[VDEV_RAIDZ_P].rc_abd);
	qxy = abd_to_buf(rr->rr_col[VDEV_RAIDZ_Q].rc_abd);
	xd = rr->rr_col[x].rc_abd;
	yd = rr->rr_col[y].rc_abd;

	/*
	* We now have:
	* Pxy = P + D_x + D_y
	* Qxy = Q + 2^(ndevs - 1 - x) * D_x + 2^(ndevs - 1 - y) * D_y
	*
	* We can then solve for D_x:
	* D_x = A * (P + Pxy) + B * (Q + Qxy)
	* where
	* A = 2^(x - y) * (2^(x - y) + 1)^-1
	* B = 2^(ndevs - 1 - x) * (2^(x - y) + 1)^-1
	*
	* With D_x in hand, we can easily solve for D_y:
	* D_y = P + Pxy + D_x
	*/

	a = vdev_raidz_pow2[255 + x - y];
	b = vdev_raidz_pow2[255 - (rr->rr_cols - 1 - x)];
	tmp = 255 - vdev_raidz_log2[a ^ 1];

	aexp = vdev_raidz_log2[vdev_raidz_exp2(a, tmp)];
	bexp = vdev_raidz_log2[vdev_raidz_exp2(b, tmp)];

	ASSERT3U(xsize, >=, ysize);
	struct reconst_pq_struct rpq = { p, q, pxy, qxy, aexp, bexp };

	(void) abd_iterate_func2(xd, yd, 0, 0, ysize,
	vdev_raidz_reconst_pq_func, &rpq);
	(void) abd_iterate_func(xd, ysize, xsize - ysize,
	vdev_raidz_reconst_pq_tail_func, &rpq);

	abd_free(rr->rr_col[VDEV_RAIDZ_P].rc_abd);
	abd_free(rr->rr_col[VDEV_RAIDZ_Q].rc_abd);

	/*
	* Restore the saved parity data.
	*/
	rr->rr_col[VDEV_RAIDZ_P].rc_abd = pdata;
	rr->rr_col[VDEV_RAIDZ_Q].rc_abd = qdata;
	}

	/*
	* In the general case of reconstruction, we must solve the system of linear
	* equations defined by the coefficients used to generate parity as well as
	* the contents of the data and parity disks. This can be expressed with
	* vectors for the original data (D) and the actual data (d) and parity (p)
	* and a matrix composed of the identity matrix (I) and a dispersal matrix (V):
	*
	* __ __ __ __
	* \| \| __ __ \| p_0 \|
	* \| V \| \| D_0 \| \| p_m-1 \|
	* \| \| x \| : \| = \| d_0 \|
	* \| I \| \| D_n-1 \| \| : \|
	* \| \| ~~ ~~ \| d_n-1 \|
	* ~~ ~~ ~~ ~~
	*
	* I is simply a square identity matrix of size n, and V is a vandermonde
	* matrix defined by the coefficients we chose for the various parity columns
	* (1, 2, 4). Note that these values were chosen both for simplicity, speedy
	* computation as well as linear separability.
	*
	* __ __ __ __
	* \| 1 .. 1 1 1 \| \| p_0 \|
	* \| 2^n-1 .. 4 2 1 \| __ __ \| : \|
	* \| 4^n-1 .. 16 4 1 \| \| D_0 \| \| p_m-1 \|
	* \| 1 .. 0 0 0 \| \| D_1 \| \| d_0 \|
	* \| 0 .. 0 0 0 \| x \| D_2 \| = \| d_1 \|
	* \| : : : : \| \| : \| \| d_2 \|
	* \| 0 .. 1 0 0 \| \| D_n-1 \| \| : \|
	* \| 0 .. 0 1 0 \| ~~ ~~ \| : \|
	* \| 0 .. 0 0 1 \| \| d_n-1 \|
	* ~~ ~~ ~~ ~~
	*
	* Note that I, V, d, and p are known. To compute D, we must invert the
	* matrix and use the known data and parity values to reconstruct the unknown
	* data values. We begin by removing the rows in V\|I and d\|p that correspond
	* to failed or missing columns; we then make V\|I square (n x n) and d\|p
	* sized n by removing rows corresponding to unused parity from the bottom up
	* to generate (V\|I)' and (d\|p)'. We can then generate the inverse of (V\|I)'
	* using Gauss-Jordan elimination. In the example below we use m=3 parity
	* columns, n=8 data columns, with errors in d_1, d_2, and p_1:
	* __ __
	* \| 1 1 1 1 1 1 1 1 \|
	* \| 128 64 32 16 8 4 2 1 \| <-----+-+-- missing disks
	* \| 19 205 116 29 64 16 4 1 \| / /
	* \| 1 0 0 0 0 0 0 0 \| / /
	* \| 0 1 0 0 0 0 0 0 \| <--' /
	* (V\|I) = \| 0 0 1 0 0 0 0 0 \| <---'
	* \| 0 0 0 1 0 0 0 0 \|
	* \| 0 0 0 0 1 0 0 0 \|
	* \| 0 0 0 0 0 1 0 0 \|
	* \| 0 0 0 0 0 0 1 0 \|
	* \| 0 0 0 0 0 0 0 1 \|
	* ~~ ~~
	* __ __
	* \| 1 1 1 1 1 1 1 1 \|
	* \| 128 64 32 16 8 4 2 1 \|
	* \| 19 205 116 29 64 16 4 1 \|
	* \| 1 0 0 0 0 0 0 0 \|
	* \| 0 1 0 0 0 0 0 0 \|
	* (V\|I)' = \| 0 0 1 0 0 0 0 0 \|
	* \| 0 0 0 1 0 0 0 0 \|
	* \| 0 0 0 0 1 0 0 0 \|
	* \| 0 0 0 0 0 1 0 0 \|
	* \| 0 0 0 0 0 0 1 0 \|
	* \| 0 0 0 0 0 0 0 1 \|
	* ~~ ~~
	*
	* Here we employ Gauss-Jordan elimination to find the inverse of (V\|I)'. We
	* have carefully chosen the seed values 1, 2, and 4 to ensure that this
	* matrix is not singular.
	* __ __
	* \| 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 \|
	* \| 19 205 116 29 64 16 4 1 0 1 0 0 0 0 0 0 \|
	* \| 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 \|
	* \| 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 \|
	* \| 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 \|
	* \| 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 \|
	* \| 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 \|
	* \| 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 \|
	* ~~ ~~
	* __ __
	* \| 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 \|
	* \| 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 \|
	* \| 19 205 116 29 64 16 4 1 0 1 0 0 0 0 0 0 \|
	* \| 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 \|
	* \| 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 \|
	* \| 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 \|
	* \| 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 \|
	* \| 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 \|
	* ~~ ~~
	* __ __
	* \| 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 \|
	* \| 0 1 1 0 0 0 0 0 1 0 1 1 1 1 1 1 \|
	* \| 0 205 116 0 0 0 0 0 0 1 19 29 64 16 4 1 \|
	* \| 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 \|
	* \| 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 \|
	* \| 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 \|
	* \| 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 \|
	* \| 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 \|
	* ~~ ~~
	* __ __
	* \| 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 \|
	* \| 0 1 1 0 0 0 0 0 1 0 1 1 1 1 1 1 \|
	* \| 0 0 185 0 0 0 0 0 205 1 222 208 141 221 201 204 \|
	* \| 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 \|
	* \| 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 \|
	* \| 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 \|
	* \| 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 \|
	* \| 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 \|
	* ~~ ~~
	* __ __
	* \| 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 \|
	* \| 0 1 1 0 0 0 0 0 1 0 1 1 1 1 1 1 \|
	* \| 0 0 1 0 0 0 0 0 166 100 4 40 158 168 216 209 \|
	* \| 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 \|
	* \| 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 \|
	* \| 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 \|
	* \| 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 \|
	* \| 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 \|
	* ~~ ~~
	* __ __
	* \| 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 \|
	* \| 0 1 0 0 0 0 0 0 167 100 5 41 159 169 217 208 \|
	* \| 0 0 1 0 0 0 0 0 166 100 4 40 158 168 216 209 \|
	* \| 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 \|
	* \| 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 \|
	* \| 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 \|
	* \| 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 \|
	* \| 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 \|
	* ~~ ~~
	* __ __
	* \| 0 0 1 0 0 0 0 0 \|
	* \| 167 100 5 41 159 169 217 208 \|
	* \| 166 100 4 40 158 168 216 209 \|
	* (V\|I)'^-1 = \| 0 0 0 1 0 0 0 0 \|
	* \| 0 0 0 0 1 0 0 0 \|
	* \| 0 0 0 0 0 1 0 0 \|
	* \| 0 0 0 0 0 0 1 0 \|
	* \| 0 0 0 0 0 0 0 1 \|
	* ~~ ~~
	*
	* We can then simply compute D = (V\|I)'^-1 x (d\|p)' to discover the values
	* of the missing data.
	*
	* As is apparent from the example above, the only non-trivial rows in the
	* inverse matrix correspond to the data disks that we're trying to
	* reconstruct. Indeed, those are the only rows we need as the others would
	* only be useful for reconstructing data known or assumed to be valid. For
	* that reason, we only build the coefficients in the rows that correspond to
	* targeted columns.
	*/

	static void
	vdev_raidz_matrix_init(raidz_row_t rr, int n, int nmap, int map,
	uint8_t **rows)
	{
	int i, j;
	int pow;

	ASSERT(n == rr->rr_cols - rr->rr_firstdatacol);

	/*
	* Fill in the missing rows of interest.
	*/
	for (i = 0; i < nmap; i++) {
	ASSERT3S(0, <=, map[i]);
	ASSERT3S(map[i], <=, 2);

	pow = map[i] * n;
	if (pow > 255)
	pow -= 255;
	ASSERT(pow <= 255);

	for (j = 0; j < n; j++) {
	pow -= map[i];
	if (pow < 0)
	pow += 255;
	rows[i][j] = vdev_raidz_pow2[pow];
	}
	}
	}

	static void
	vdev_raidz_matrix_invert(raidz_row_t rr, int n, int nmissing, int missing,
	uint8_t rows, uint8_t invrows, const uint8_t *used)
	{
	int i, j, ii, jj;
	uint8_t log;

	/*
	* Assert that the first nmissing entries from the array of used
	* columns correspond to parity columns and that subsequent entries
	* correspond to data columns.
	*/
	for (i = 0; i < nmissing; i++) {
	ASSERT3S(used[i], <, rr->rr_firstdatacol);
	}
	for (; i < n; i++) {
	ASSERT3S(used[i], >=, rr->rr_firstdatacol);
	}

	/*
	* First initialize the storage where we'll compute the inverse rows.
	*/
	for (i = 0; i < nmissing; i++) {
	for (j = 0; j < n; j++) {
	invrows[i][j] = (i == j) ? 1 : 0;
	}
	}

	/*
	* Subtract all trivial rows from the rows of consequence.
	*/
	for (i = 0; i < nmissing; i++) {
	for (j = nmissing; j < n; j++) {
	ASSERT3U(used[j], >=, rr->rr_firstdatacol);
	jj = used[j] - rr->rr_firstdatacol;
	ASSERT3S(jj, <, n);
	invrows[i][j] = rows[i][jj];
	rows[i][jj] = 0;
	}
	}

	/*
	* For each of the rows of interest, we must normalize it and subtract
	* a multiple of it from the other rows.
	*/
	for (i = 0; i < nmissing; i++) {
	for (j = 0; j < missing[i]; j++) {
	ASSERT0(rows[i][j]);
	}
	ASSERT3U(rows[i][missing[i]], !=, 0);

	/*
	* Compute the inverse of the first element and multiply each
	* element in the row by that value.
	*/
	log = 255 - vdev_raidz_log2[rows[i][missing[i]]];

	for (j = 0; j < n; j++) {
	rows[i][j] = vdev_raidz_exp2(rows[i][j], log);
	invrows[i][j] = vdev_raidz_exp2(invrows[i][j], log);
	}

	for (ii = 0; ii < nmissing; ii++) {
	if (i == ii)
	continue;

	ASSERT3U(rows[ii][missing[i]], !=, 0);

	log = vdev_raidz_log2[rows[ii][missing[i]]];

	for (j = 0; j < n; j++) {
	rows[ii][j] ^=
	vdev_raidz_exp2(rows[i][j], log);
	invrows[ii][j] ^=
	vdev_raidz_exp2(invrows[i][j], log);
	}
	}
	}

	/*
	* Verify that the data that is left in the rows are properly part of
	* an identity matrix.
	*/
	for (i = 0; i < nmissing; i++) {
	for (j = 0; j < n; j++) {
	if (j == missing[i]) {
	ASSERT3U(rows[i][j], ==, 1);
	} else {
	ASSERT0(rows[i][j]);
	}
	}
	}
	}

	static void
	vdev_raidz_matrix_reconstruct(raidz_row_t *rr, int n, int nmissing,
	int missing, uint8_t invrows, const uint8_t used)
	{
	int i, j, x, cc, c;
	uint8_t *src;
	uint64_t ccount;
	uint8_t *dst[VDEV_RAIDZ_MAXPARITY] = { NULL };
	uint64_t dcount[VDEV_RAIDZ_MAXPARITY] = { 0 };
	uint8_t log = 0;
	uint8_t val;
	int ll;
	uint8_t *invlog[VDEV_RAIDZ_MAXPARITY];
	uint8_t p, pp;
	size_t psize;

	psize = sizeof (invlog[0][0]) * n * nmissing;
	p = kmem_alloc(psize, KM_SLEEP);

	for (pp = p, i = 0; i < nmissing; i++) {
	invlog[i] = pp;
	pp += n;
	}

	for (i = 0; i < nmissing; i++) {
	for (j = 0; j < n; j++) {
	ASSERT3U(invrows[i][j], !=, 0);
	invlog[i][j] = vdev_raidz_log2[invrows[i][j]];
	}
	}

	for (i = 0; i < n; i++) {
	c = used[i];
	ASSERT3U(c, <, rr->rr_cols);

	ccount = rr->rr_col[c].rc_size;
	ASSERT(ccount >= rr->rr_col[missing[0]].rc_size \|\| i > 0);
	if (ccount == 0)
	continue;
	src = abd_to_buf(rr->rr_col[c].rc_abd);
	for (j = 0; j < nmissing; j++) {
	cc = missing[j] + rr->rr_firstdatacol;
	ASSERT3U(cc, >=, rr->rr_firstdatacol);
	ASSERT3U(cc, <, rr->rr_cols);
	ASSERT3U(cc, !=, c);

	dcount[j] = rr->rr_col[cc].rc_size;
	if (dcount[j] != 0)
	dst[j] = abd_to_buf(rr->rr_col[cc].rc_abd);
	}

	for (x = 0; x < ccount; x++, src++) {
	if (*src != 0)
	log = vdev_raidz_log2[*src];

	for (cc = 0; cc < nmissing; cc++) {
	if (x >= dcount[cc])
	continue;

	if (*src == 0) {
	val = 0;
	} else {
	if ((ll = log + invlog[cc][i]) >= 255)
	ll -= 255;
	val = vdev_raidz_pow2[ll];
	}

	if (i == 0)
	dst[cc][x] = val;
	else
	dst[cc][x] ^= val;
	}
	}
	}

	kmem_free(p, psize);
	}

	static void
	vdev_raidz_reconstruct_general(raidz_row_t rr, int tgts, int ntgts)
	{
	int i, c, t, tt;
	unsigned int n;
	unsigned int nmissing_rows;
	int missing_rows[VDEV_RAIDZ_MAXPARITY];
	int parity_map[VDEV_RAIDZ_MAXPARITY];
	uint8_t p, pp;
	size_t psize;
	uint8_t *rows[VDEV_RAIDZ_MAXPARITY];
	uint8_t *invrows[VDEV_RAIDZ_MAXPARITY];
	uint8_t *used;

	abd_t **bufs = NULL;

	if (zfs_flags & ZFS_DEBUG_RAIDZ_RECONSTRUCT)
	zfs_dbgmsg("reconstruct_general(rm=%px ntgts=%u)", rr, ntgts);
	/*
	* Matrix reconstruction can't use scatter ABDs yet, so we allocate
	* temporary linear ABDs if any non-linear ABDs are found.
	*/
	for (i = rr->rr_firstdatacol; i < rr->rr_cols; i++) {
	ASSERT(rr->rr_col[i].rc_abd != NULL);
	if (!abd_is_linear(rr->rr_col[i].rc_abd)) {
	bufs = kmem_alloc(rr->rr_cols * sizeof (abd_t *),
	KM_PUSHPAGE);

	for (c = rr->rr_firstdatacol; c < rr->rr_cols; c++) {
	raidz_col_t *col = &rr->rr_col[c];

	bufs[c] = col->rc_abd;
	if (bufs[c] != NULL) {
	col->rc_abd = abd_alloc_linear(
	col->rc_size, B_TRUE);
	abd_copy(col->rc_abd, bufs[c],
	col->rc_size);
	}
	}

	break;
	}
	}

	n = rr->rr_cols - rr->rr_firstdatacol;

	/*
	* Figure out which data columns are missing.
	*/
	nmissing_rows = 0;
	for (t = 0; t < ntgts; t++) {
	if (tgts[t] >= rr->rr_firstdatacol) {
	missing_rows[nmissing_rows++] =
	tgts[t] - rr->rr_firstdatacol;
	}
	}

	/*
	* Figure out which parity columns to use to help generate the missing
	* data columns.
	*/
	for (tt = 0, c = 0, i = 0; i < nmissing_rows; c++) {
	ASSERT(tt < ntgts);
	ASSERT(c < rr->rr_firstdatacol);

	/*
	* Skip any targeted parity columns.
	*/
	if (c == tgts[tt]) {
	tt++;
	continue;
	}

	parity_map[i] = c;
	i++;
	}

	psize = (sizeof (rows[0][0]) + sizeof (invrows[0][0])) *
	nmissing_rows * n + sizeof (used[0]) * n;
	p = kmem_alloc(psize, KM_SLEEP);

	for (pp = p, i = 0; i < nmissing_rows; i++) {
	rows[i] = pp;
	pp += n;
	invrows[i] = pp;
	pp += n;
	}
	used = pp;

	for (i = 0; i < nmissing_rows; i++) {
	used[i] = parity_map[i];
	}

	for (tt = 0, c = rr->rr_firstdatacol; c < rr->rr_cols; c++) {
	if (tt < nmissing_rows &&
	c == missing_rows[tt] + rr->rr_firstdatacol) {
	tt++;
	continue;
	}

	ASSERT3S(i, <, n);
	used[i] = c;
	i++;
	}

	/*
	* Initialize the interesting rows of the matrix.
	*/
	vdev_raidz_matrix_init(rr, n, nmissing_rows, parity_map, rows);

	/*
	* Invert the matrix.
	*/
	vdev_raidz_matrix_invert(rr, n, nmissing_rows, missing_rows, rows,
	invrows, used);

	/*
	* Reconstruct the missing data using the generated matrix.
	*/
	vdev_raidz_matrix_reconstruct(rr, n, nmissing_rows, missing_rows,
	invrows, used);

	kmem_free(p, psize);

	/*
	* copy back from temporary linear abds and free them
	*/
	if (bufs) {
	for (c = rr->rr_firstdatacol; c < rr->rr_cols; c++) {
	raidz_col_t *col = &rr->rr_col[c];

	if (bufs[c] != NULL) {
	abd_copy(bufs[c], col->rc_abd, col->rc_size);
	abd_free(col->rc_abd);
	}
	col->rc_abd = bufs[c];
	}
	kmem_free(bufs, rr->rr_cols * sizeof (abd_t *));
	}
	}

	static void
	vdev_raidz_reconstruct_row(raidz_map_t rm, raidz_row_t rr,
	const int *t, int nt)
	{
	int tgts[VDEV_RAIDZ_MAXPARITY], *dt;
	int ntgts;
	int i, c, ret;
	int nbadparity, nbaddata;
	int parity_valid[VDEV_RAIDZ_MAXPARITY];

	if (zfs_flags & ZFS_DEBUG_RAIDZ_RECONSTRUCT) {
	zfs_dbgmsg("reconstruct(rm=%px nt=%u cols=%u md=%u mp=%u)",
	rr, nt, (int)rr->rr_cols, (int)rr->rr_missingdata,
	(int)rr->rr_missingparity);
	}

	nbadparity = rr->rr_firstdatacol;
	nbaddata = rr->rr_cols - nbadparity;
	ntgts = 0;
	for (i = 0, c = 0; c < rr->rr_cols; c++) {
	if (zfs_flags & ZFS_DEBUG_RAIDZ_RECONSTRUCT) {
	zfs_dbgmsg("reconstruct(rm=%px col=%u devid=%u "
	"offset=%llx error=%u)",
	rr, c, (int)rr->rr_col[c].rc_devidx,
	(long long)rr->rr_col[c].rc_offset,
	(int)rr->rr_col[c].rc_error);
	}
	if (c < rr->rr_firstdatacol)
	parity_valid[c] = B_FALSE;

	if (i < nt && c == t[i]) {
	tgts[ntgts++] = c;
	i++;
	} else if (rr->rr_col[c].rc_error != 0) {
	tgts[ntgts++] = c;
	} else if (c >= rr->rr_firstdatacol) {
	nbaddata--;
	} else {
	parity_valid[c] = B_TRUE;
	nbadparity--;
	}
	}

	ASSERT(ntgts >= nt);
	ASSERT(nbaddata >= 0);
	ASSERT(nbaddata + nbadparity == ntgts);

	dt = &tgts[nbadparity];

	/* Reconstruct using the new math implementation */
	ret = vdev_raidz_math_reconstruct(rm, rr, parity_valid, dt, nbaddata);
	if (ret != RAIDZ_ORIGINAL_IMPL)
	return;

	/*
	* See if we can use any of our optimized reconstruction routines.
	*/
	switch (nbaddata) {
	case 1:
	if (parity_valid[VDEV_RAIDZ_P]) {
	vdev_raidz_reconstruct_p(rr, dt, 1);
	return;
	}

	ASSERT(rr->rr_firstdatacol > 1);

	if (parity_valid[VDEV_RAIDZ_Q]) {
	vdev_raidz_reconstruct_q(rr, dt, 1);
	return;
	}

	ASSERT(rr->rr_firstdatacol > 2);
	break;

	case 2:
	ASSERT(rr->rr_firstdatacol > 1);

	if (parity_valid[VDEV_RAIDZ_P] &&
	parity_valid[VDEV_RAIDZ_Q]) {
	vdev_raidz_reconstruct_pq(rr, dt, 2);
	return;
	}

	ASSERT(rr->rr_firstdatacol > 2);

	break;
	}

	vdev_raidz_reconstruct_general(rr, tgts, ntgts);
	}

	static int
	vdev_raidz_open(vdev_t vd, uint64_t asize, uint64_t *max_asize,
	uint64_t logical_ashift, uint64_t physical_ashift)
	{
	vdev_raidz_t *vdrz = vd->vdev_tsd;
	uint64_t nparity = vdrz->vd_nparity;
	int c;
	int lasterror = 0;
	int numerrors = 0;

	ASSERT(nparity > 0);

	if (nparity > VDEV_RAIDZ_MAXPARITY \|\|
	vd->vdev_children < nparity + 1) {
	vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
	return (SET_ERROR(EINVAL));
	}

	vdev_open_children(vd);

	for (c = 0; c < vd->vdev_children; c++) {
	vdev_t *cvd = vd->vdev_child[c];

	if (cvd->vdev_open_error != 0) {
	lasterror = cvd->vdev_open_error;
	numerrors++;
	continue;
	}

	asize = MIN(asize - 1, cvd->vdev_asize - 1) + 1;
	max_asize = MIN(max_asize - 1, cvd->vdev_max_asize - 1) + 1;
	logical_ashift = MAX(logical_ashift, cvd->vdev_ashift);
	}
	for (c = 0; c < vd->vdev_children; c++) {
	vdev_t *cvd = vd->vdev_child[c];

	if (cvd->vdev_open_error != 0)
	continue;
	physical_ashift = vdev_best_ashift(logical_ashift,
	*physical_ashift, cvd->vdev_physical_ashift);
	}

	if (vd->vdev_rz_expanding) {
	asize = vd->vdev_children - 1;
	max_asize = vd->vdev_children - 1;

	vd->vdev_min_asize = *asize;
	} else {
	asize = vd->vdev_children;
	max_asize = vd->vdev_children;
	}

	if (numerrors > nparity) {
	vd->vdev_stat.vs_aux = VDEV_AUX_NO_REPLICAS;
	return (lasterror);
	}

	return (0);
	}

	static void
	vdev_raidz_close(vdev_t *vd)
	{
	for (int c = 0; c < vd->vdev_children; c++) {
	if (vd->vdev_child[c] != NULL)
	vdev_close(vd->vdev_child[c]);
	}
	}

	/*
	* Return the logical width to use, given the txg in which the allocation
	* happened. Note that BP_GET_BIRTH() is usually the txg in which the
	* BP was allocated. Remapped BP's (that were relocated due to device
	* removal, see remap_blkptr_cb()), will have a more recent physical birth
	* which reflects when the BP was relocated, but we can ignore these because
	* they can't be on RAIDZ (device removal doesn't support RAIDZ).
	*/
	static uint64_t
	vdev_raidz_get_logical_width(vdev_raidz_t *vdrz, uint64_t txg)
	{
	reflow_node_t lookup = {
	.re_txg = txg,
	};
	avl_index_t where;

	uint64_t width;
	mutex_enter(&vdrz->vd_expand_lock);
	reflow_node_t *re = avl_find(&vdrz->vd_expand_txgs, &lookup, &where);
	if (re != NULL) {
	width = re->re_logical_width;
	} else {
	re = avl_nearest(&vdrz->vd_expand_txgs, where, AVL_BEFORE);
	if (re != NULL)
	width = re->re_logical_width;
	else
	width = vdrz->vd_original_width;
	}
	mutex_exit(&vdrz->vd_expand_lock);
	return (width);
	}

	/*
	* Note: If the RAIDZ vdev has been expanded, older BP's may have allocated
	* more space due to the lower data-to-parity ratio. In this case it's
	* important to pass in the correct txg. Note that vdev_gang_header_asize()
	* relies on a constant asize for psize=SPA_GANGBLOCKSIZE=SPA_MINBLOCKSIZE,
	* regardless of txg. This is assured because for a single data sector, we
	* allocate P+1 sectors regardless of width ("cols", which is at least P+1).
	*/
	static uint64_t
	vdev_raidz_asize(vdev_t *vd, uint64_t psize, uint64_t txg)
	{
	vdev_raidz_t *vdrz = vd->vdev_tsd;
	uint64_t asize;
	uint64_t ashift = vd->vdev_top->vdev_ashift;
	uint64_t cols = vdrz->vd_original_width;
	uint64_t nparity = vdrz->vd_nparity;

	cols = vdev_raidz_get_logical_width(vdrz, txg);

	asize = ((psize - 1) >> ashift) + 1;
	asize += nparity * ((asize + cols - nparity - 1) / (cols - nparity));
	asize = roundup(asize, nparity + 1) << ashift;

	#ifdef ZFS_DEBUG
	uint64_t asize_new = ((psize - 1) >> ashift) + 1;
	uint64_t ncols_new = vdrz->vd_physical_width;
	asize_new += nparity * ((asize_new + ncols_new - nparity - 1) /
	(ncols_new - nparity));
	asize_new = roundup(asize_new, nparity + 1) << ashift;
	VERIFY3U(asize_new, <=, asize);
	#endif

	return (asize);
	}

	/*
	* The allocatable space for a raidz vdev is N * sizeof(smallest child)
	* so each child must provide at least 1/Nth of its asize.
	*/
	static uint64_t
	vdev_raidz_min_asize(vdev_t *vd)
	{
	return ((vd->vdev_min_asize + vd->vdev_children - 1) /
	vd->vdev_children);
	}

	void
	vdev_raidz_child_done(zio_t *zio)
	{
	raidz_col_t *rc = zio->io_private;

	ASSERT3P(rc->rc_abd, !=, NULL);
	rc->rc_error = zio->io_error;
	rc->rc_tried = 1;
	rc->rc_skipped = 0;
	}

	static void
	vdev_raidz_shadow_child_done(zio_t *zio)
	{
	raidz_col_t *rc = zio->io_private;

	rc->rc_shadow_error = zio->io_error;
	}

	static void
	vdev_raidz_io_verify(zio_t zio, raidz_map_t rm, raidz_row_t *rr, int col)
	{
	(void) rm;
	#ifdef ZFS_DEBUG
	range_seg64_t logical_rs, physical_rs, remain_rs;
	logical_rs.rs_start = rr->rr_offset;
	logical_rs.rs_end = logical_rs.rs_start +
	vdev_raidz_asize(zio->io_vd, rr->rr_size,
	BP_GET_BIRTH(zio->io_bp));

	raidz_col_t *rc = &rr->rr_col[col];
	vdev_t *cvd = zio->io_vd->vdev_child[rc->rc_devidx];

	vdev_xlate(cvd, &logical_rs, &physical_rs, &remain_rs);
	ASSERT(vdev_xlate_is_empty(&remain_rs));
	if (vdev_xlate_is_empty(&physical_rs)) {
	/*
	* If we are in the middle of expansion, the
	* physical->logical mapping is changing so vdev_xlate()
	* can't give us a reliable answer.
	*/
	return;
	}
	ASSERT3U(rc->rc_offset, ==, physical_rs.rs_start);
	ASSERT3U(rc->rc_offset, <, physical_rs.rs_end);
	/*
	* It would be nice to assert that rs_end is equal
	* to rc_offset + rc_size but there might be an
	* optional I/O at the end that is not accounted in
	* rc_size.
	*/
	if (physical_rs.rs_end > rc->rc_offset + rc->rc_size) {
	ASSERT3U(physical_rs.rs_end, ==, rc->rc_offset +
	rc->rc_size + (1 << zio->io_vd->vdev_top->vdev_ashift));
	} else {
	ASSERT3U(physical_rs.rs_end, ==, rc->rc_offset + rc->rc_size);
	}
	#endif
	}

	static void
	vdev_raidz_io_start_write(zio_t zio, raidz_row_t rr)
	{
	vdev_t *vd = zio->io_vd;
	raidz_map_t *rm = zio->io_vsd;

	vdev_raidz_generate_parity_row(rm, rr);

	for (int c = 0; c < rr->rr_scols; c++) {
	raidz_col_t *rc = &rr->rr_col[c];
	vdev_t *cvd = vd->vdev_child[rc->rc_devidx];

	/* Verify physical to logical translation */
	vdev_raidz_io_verify(zio, rm, rr, c);

	if (rc->rc_size == 0)
	continue;

	ASSERT3U(rc->rc_offset + rc->rc_size, <,
	cvd->vdev_psize - VDEV_LABEL_END_SIZE);

	ASSERT3P(rc->rc_abd, !=, NULL);
	zio_nowait(zio_vdev_child_io(zio, NULL, cvd,
	rc->rc_offset, rc->rc_abd,
	abd_get_size(rc->rc_abd), zio->io_type,
	zio->io_priority, 0, vdev_raidz_child_done, rc));

	if (rc->rc_shadow_devidx != INT_MAX) {
	vdev_t *cvd2 = vd->vdev_child[rc->rc_shadow_devidx];

	ASSERT3U(
	rc->rc_shadow_offset + abd_get_size(rc->rc_abd), <,
	cvd2->vdev_psize - VDEV_LABEL_END_SIZE);

	zio_nowait(zio_vdev_child_io(zio, NULL, cvd2,
	rc->rc_shadow_offset, rc->rc_abd,
	abd_get_size(rc->rc_abd),
	zio->io_type, zio->io_priority, 0,
	vdev_raidz_shadow_child_done, rc));
	}
	}
	}

	/*
	* Generate optional I/Os for skip sectors to improve aggregation contiguity.
	* This only works for vdev_raidz_map_alloc() (not _expanded()).
	*/
	static void
	raidz_start_skip_writes(zio_t *zio)
	{
	vdev_t *vd = zio->io_vd;
	uint64_t ashift = vd->vdev_top->vdev_ashift;
	raidz_map_t *rm = zio->io_vsd;
	ASSERT3U(rm->rm_nrows, ==, 1);
	raidz_row_t *rr = rm->rm_row[0];
	for (int c = 0; c < rr->rr_scols; c++) {
	raidz_col_t *rc = &rr->rr_col[c];
	vdev_t *cvd = vd->vdev_child[rc->rc_devidx];
	if (rc->rc_size != 0)
	continue;
	ASSERT3P(rc->rc_abd, ==, NULL);

	ASSERT3U(rc->rc_offset, <,
	cvd->vdev_psize - VDEV_LABEL_END_SIZE);

	zio_nowait(zio_vdev_child_io(zio, NULL, cvd, rc->rc_offset,
	NULL, 1ULL << ashift, zio->io_type, zio->io_priority,
	ZIO_FLAG_NODATA \| ZIO_FLAG_OPTIONAL, NULL, NULL));
	}
	}

	static void
	vdev_raidz_io_start_read_row(zio_t zio, raidz_row_t rr, boolean_t forceparity)
	{
	vdev_t *vd = zio->io_vd;

	/*
	* Iterate over the columns in reverse order so that we hit the parity
	* last -- any errors along the way will force us to read the parity.
	*/
	for (int c = rr->rr_cols - 1; c >= 0; c--) {
	raidz_col_t *rc = &rr->rr_col[c];
	if (rc->rc_size == 0)
	continue;
	vdev_t *cvd = vd->vdev_child[rc->rc_devidx];
	if (!vdev_readable(cvd)) {
	if (c >= rr->rr_firstdatacol)
	rr->rr_missingdata++;
	else
	rr->rr_missingparity++;
	rc->rc_error = SET_ERROR(ENXIO);
	rc->rc_tried = 1; /* don't even try */
	rc->rc_skipped = 1;
	continue;
	}
	if (vdev_dtl_contains(cvd, DTL_MISSING, zio->io_txg, 1)) {
	if (c >= rr->rr_firstdatacol)
	rr->rr_missingdata++;
	else
	rr->rr_missingparity++;
	rc->rc_error = SET_ERROR(ESTALE);
	rc->rc_skipped = 1;
	continue;
	}
	if (forceparity \|\|
	c >= rr->rr_firstdatacol \|\| rr->rr_missingdata > 0 \|\|
	(zio->io_flags & (ZIO_FLAG_SCRUB \| ZIO_FLAG_RESILVER))) {
	zio_nowait(zio_vdev_child_io(zio, NULL, cvd,
	rc->rc_offset, rc->rc_abd, rc->rc_size,
	zio->io_type, zio->io_priority, 0,
	vdev_raidz_child_done, rc));
	}
	}
	}

	static void
	vdev_raidz_io_start_read_phys_cols(zio_t zio, raidz_map_t rm)
	{
	vdev_t *vd = zio->io_vd;

	for (int i = 0; i < rm->rm_nphys_cols; i++) {
	raidz_col_t *prc = &rm->rm_phys_col[i];
	if (prc->rc_size == 0)
	continue;

	ASSERT3U(prc->rc_devidx, ==, i);
	vdev_t *cvd = vd->vdev_child[i];
	if (!vdev_readable(cvd)) {
	prc->rc_error = SET_ERROR(ENXIO);
	prc->rc_tried = 1; /* don't even try */
	prc->rc_skipped = 1;
	continue;
	}
	if (vdev_dtl_contains(cvd, DTL_MISSING, zio->io_txg, 1)) {
	prc->rc_error = SET_ERROR(ESTALE);
	prc->rc_skipped = 1;
	continue;
	}
	zio_nowait(zio_vdev_child_io(zio, NULL, cvd,
	prc->rc_offset, prc->rc_abd, prc->rc_size,
	zio->io_type, zio->io_priority, 0,
	vdev_raidz_child_done, prc));
	}
	}

	static void
	vdev_raidz_io_start_read(zio_t zio, raidz_map_t rm)
	{
	/*
	* If there are multiple rows, we will be hitting
	* all disks, so go ahead and read the parity so
	* that we are reading in decent size chunks.
	*/
	boolean_t forceparity = rm->rm_nrows > 1;

	if (rm->rm_phys_col) {
	vdev_raidz_io_start_read_phys_cols(zio, rm);
	} else {
	for (int i = 0; i < rm->rm_nrows; i++) {
	raidz_row_t *rr = rm->rm_row[i];
	vdev_raidz_io_start_read_row(zio, rr, forceparity);
	}
	}
	}

	/*
	* Start an IO operation on a RAIDZ VDev
	*
	* Outline:
	* - For write operations:
	* 1. Generate the parity data
	* 2. Create child zio write operations to each column's vdev, for both
	* data and parity.
	* 3. If the column skips any sectors for padding, create optional dummy
	* write zio children for those areas to improve aggregation continuity.
	* - For read operations:
	* 1. Create child zio read operations to each data column's vdev to read
	* the range of data required for zio.
	* 2. If this is a scrub or resilver operation, or if any of the data
	* vdevs have had errors, then create zio read operations to the parity
	* columns' VDevs as well.
	*/
	static void
	vdev_raidz_io_start(zio_t *zio)
	{
	vdev_t *vd = zio->io_vd;
	vdev_t *tvd = vd->vdev_top;
	vdev_raidz_t *vdrz = vd->vdev_tsd;
	raidz_map_t *rm;

	uint64_t logical_width = vdev_raidz_get_logical_width(vdrz,
	BP_GET_BIRTH(zio->io_bp));
	if (logical_width != vdrz->vd_physical_width) {
	zfs_locked_range_t *lr = NULL;
	uint64_t synced_offset = UINT64_MAX;
	uint64_t next_offset = UINT64_MAX;
	boolean_t use_scratch = B_FALSE;
	/*
	* Note: when the expansion is completing, we set
	* vre_state=DSS_FINISHED (in raidz_reflow_complete_sync())
	* in a later txg than when we last update spa_ubsync's state
	* (see the end of spa_raidz_expand_thread()). Therefore we
	* may see vre_state!=SCANNING before
	* VDEV_TOP_ZAP_RAIDZ_EXPAND_STATE=DSS_FINISHED is reflected
	* on disk, but the copying progress has been synced to disk
	* (and reflected in spa_ubsync). In this case it's fine to
	* treat the expansion as completed, since if we crash there's
	* no additional copying to do.
	*/
	if (vdrz->vn_vre.vre_state == DSS_SCANNING) {
	ASSERT3P(vd->vdev_spa->spa_raidz_expand, ==,
	&vdrz->vn_vre);
	lr = zfs_rangelock_enter(&vdrz->vn_vre.vre_rangelock,
	zio->io_offset, zio->io_size, RL_READER);
	use_scratch =
	(RRSS_GET_STATE(&vd->vdev_spa->spa_ubsync) ==
	RRSS_SCRATCH_VALID);
	synced_offset =
	RRSS_GET_OFFSET(&vd->vdev_spa->spa_ubsync);
	next_offset = vdrz->vn_vre.vre_offset;
	/*
	* If we haven't resumed expanding since importing the
	* pool, vre_offset won't have been set yet. In
	* this case the next offset to be copied is the same
	* as what was synced.
	*/
	if (next_offset == UINT64_MAX) {
	next_offset = synced_offset;
	}
	}
	if (use_scratch) {
	zfs_dbgmsg("zio=%px %s io_offset=%llu offset_synced="
	"%lld next_offset=%lld use_scratch=%u",
	zio,
	zio->io_type == ZIO_TYPE_WRITE ? "WRITE" : "READ",
	(long long)zio->io_offset,
	(long long)synced_offset,
	(long long)next_offset,
	use_scratch);
	}

	rm = vdev_raidz_map_alloc_expanded(zio,
	tvd->vdev_ashift, vdrz->vd_physical_width,
	logical_width, vdrz->vd_nparity,
	synced_offset, next_offset, use_scratch);
	rm->rm_lr = lr;
	} else {
	rm = vdev_raidz_map_alloc(zio,
	tvd->vdev_ashift, logical_width, vdrz->vd_nparity);
	}
	rm->rm_original_width = vdrz->vd_original_width;

	zio->io_vsd = rm;
	zio->io_vsd_ops = &vdev_raidz_vsd_ops;
	if (zio->io_type == ZIO_TYPE_WRITE) {
	for (int i = 0; i < rm->rm_nrows; i++) {
	vdev_raidz_io_start_write(zio, rm->rm_row[i]);
	}

	if (logical_width == vdrz->vd_physical_width) {
	raidz_start_skip_writes(zio);
	}
	} else {
	ASSERT(zio->io_type == ZIO_TYPE_READ);
	vdev_raidz_io_start_read(zio, rm);
	}

	zio_execute(zio);
	}

	/*
	* Report a checksum error for a child of a RAID-Z device.
	*/
	void
	vdev_raidz_checksum_error(zio_t zio, raidz_col_t rc, abd_t *bad_data)
	{
	vdev_t *vd = zio->io_vd->vdev_child[rc->rc_devidx];

	if (!(zio->io_flags & ZIO_FLAG_SPECULATIVE) &&
	zio->io_priority != ZIO_PRIORITY_REBUILD) {
	zio_bad_cksum_t zbc;
	raidz_map_t *rm = zio->io_vsd;

	zbc.zbc_has_cksum = 0;
	zbc.zbc_injected = rm->rm_ecksuminjected;

	mutex_enter(&vd->vdev_stat_lock);
	vd->vdev_stat.vs_checksum_errors++;
	mutex_exit(&vd->vdev_stat_lock);
	(void) zfs_ereport_post_checksum(zio->io_spa, vd,
	&zio->io_bookmark, zio, rc->rc_offset, rc->rc_size,
	rc->rc_abd, bad_data, &zbc);
	}
	}

	/*
	* We keep track of whether or not there were any injected errors, so that
	* any ereports we generate can note it.
	*/
	static int
	raidz_checksum_verify(zio_t *zio)
	{
	zio_bad_cksum_t zbc = {0};
	raidz_map_t *rm = zio->io_vsd;

	int ret = zio_checksum_error(zio, &zbc);
	if (ret != 0 && zbc.zbc_injected != 0)
	rm->rm_ecksuminjected = 1;

	return (ret);
	}

	/*
	* Generate the parity from the data columns. If we tried and were able to
	* read the parity without error, verify that the generated parity matches the
	* data we read. If it doesn't, we fire off a checksum error. Return the
	* number of such failures.
	*/
	static int
	raidz_parity_verify(zio_t zio, raidz_row_t rr)
	{
	abd_t *orig[VDEV_RAIDZ_MAXPARITY];
	int c, ret = 0;
	raidz_map_t *rm = zio->io_vsd;
	raidz_col_t *rc;

	blkptr_t *bp = zio->io_bp;
	enum zio_checksum checksum = (bp == NULL ? zio->io_prop.zp_checksum :
	(BP_IS_GANG(bp) ? ZIO_CHECKSUM_GANG_HEADER : BP_GET_CHECKSUM(bp)));

	if (checksum == ZIO_CHECKSUM_NOPARITY)
	return (ret);

	for (c = 0; c < rr->rr_firstdatacol; c++) {
	rc = &rr->rr_col[c];
	if (!rc->rc_tried \|\| rc->rc_error != 0)
	continue;

	orig[c] = rc->rc_abd;
	ASSERT3U(abd_get_size(rc->rc_abd), ==, rc->rc_size);
	rc->rc_abd = abd_alloc_linear(rc->rc_size, B_FALSE);
	}

	/*
	* Verify any empty sectors are zero filled to ensure the parity
	* is calculated correctly even if these non-data sectors are damaged.
	*/
	if (rr->rr_nempty && rr->rr_abd_empty != NULL)
	ret += vdev_draid_map_verify_empty(zio, rr);

	/*
	* Regenerates parity even for !tried\|\|rc_error!=0 columns. This
	* isn't harmful but it does have the side effect of fixing stuff
	* we didn't realize was necessary (i.e. even if we return 0).
	*/
	vdev_raidz_generate_parity_row(rm, rr);

	for (c = 0; c < rr->rr_firstdatacol; c++) {
	rc = &rr->rr_col[c];

	if (!rc->rc_tried \|\| rc->rc_error != 0)
	continue;

	if (abd_cmp(orig[c], rc->rc_abd) != 0) {
	zfs_dbgmsg("found error on col=%u devidx=%u off %llx",
	c, (int)rc->rc_devidx, (u_longlong_t)rc->rc_offset);
	vdev_raidz_checksum_error(zio, rc, orig[c]);
	rc->rc_error = SET_ERROR(ECKSUM);
	ret++;
	}
	abd_free(orig[c]);
	}

	return (ret);
	}

	static int
	vdev_raidz_worst_error(raidz_row_t *rr)
	{
	int error = 0;

	for (int c = 0; c < rr->rr_cols; c++) {
	error = zio_worst_error(error, rr->rr_col[c].rc_error);
	error = zio_worst_error(error, rr->rr_col[c].rc_shadow_error);
	}

	return (error);
	}

	static void
	vdev_raidz_io_done_verified(zio_t zio, raidz_row_t rr)
	{
	int unexpected_errors = 0;
	int parity_errors = 0;
	int parity_untried = 0;
	int data_errors = 0;

	ASSERT3U(zio->io_type, ==, ZIO_TYPE_READ);

	for (int c = 0; c < rr->rr_cols; c++) {
	raidz_col_t *rc = &rr->rr_col[c];

	if (rc->rc_error) {
	if (c < rr->rr_firstdatacol)
	parity_errors++;
	else
	data_errors++;

	if (!rc->rc_skipped)
	unexpected_errors++;
	} else if (c < rr->rr_firstdatacol && !rc->rc_tried) {
	parity_untried++;
	}

	if (rc->rc_force_repair)
	unexpected_errors++;
	}

	/*
	* If we read more parity disks than were used for
	* reconstruction, confirm that the other parity disks produced
	* correct data.
	*
	* Note that we also regenerate parity when resilvering so we
	* can write it out to failed devices later.
	*/
	if (parity_errors + parity_untried <
	rr->rr_firstdatacol - data_errors \|\|
	(zio->io_flags & ZIO_FLAG_RESILVER)) {
	int n = raidz_parity_verify(zio, rr);
	unexpected_errors += n;
	}

	if (zio->io_error == 0 && spa_writeable(zio->io_spa) &&
	(unexpected_errors > 0 \|\| (zio->io_flags & ZIO_FLAG_RESILVER))) {
	/*
	* Use the good data we have in hand to repair damaged children.
	*/
	for (int c = 0; c < rr->rr_cols; c++) {
	raidz_col_t *rc = &rr->rr_col[c];
	vdev_t *vd = zio->io_vd;
	vdev_t *cvd = vd->vdev_child[rc->rc_devidx];

	if (!rc->rc_allow_repair) {
	continue;
	} else if (!rc->rc_force_repair &&
	(rc->rc_error == 0 \|\| rc->rc_size == 0)) {
	continue;
	}

	zfs_dbgmsg("zio=%px repairing c=%u devidx=%u "
	"offset=%llx",
	zio, c, rc->rc_devidx, (long long)rc->rc_offset);

	zio_nowait(zio_vdev_child_io(zio, NULL, cvd,
	rc->rc_offset, rc->rc_abd, rc->rc_size,
	ZIO_TYPE_WRITE,
	zio->io_priority == ZIO_PRIORITY_REBUILD ?
	ZIO_PRIORITY_REBUILD : ZIO_PRIORITY_ASYNC_WRITE,
	ZIO_FLAG_IO_REPAIR \| (unexpected_errors ?
	ZIO_FLAG_SELF_HEAL : 0), NULL, NULL));
	}
	}

	/*
	* Scrub or resilver i/o's: overwrite any shadow locations with the
	* good data. This ensures that if we've already copied this sector,
	* it will be corrected if it was damaged. This writes more than is
	* necessary, but since expansion is paused during scrub/resilver, at
	* most a single row will have a shadow location.
	*/
	if (zio->io_error == 0 && spa_writeable(zio->io_spa) &&
	(zio->io_flags & (ZIO_FLAG_RESILVER \| ZIO_FLAG_SCRUB))) {
	for (int c = 0; c < rr->rr_cols; c++) {
	raidz_col_t *rc = &rr->rr_col[c];
	vdev_t *vd = zio->io_vd;

	if (rc->rc_shadow_devidx == INT_MAX \|\| rc->rc_size == 0)
	continue;
	vdev_t *cvd = vd->vdev_child[rc->rc_shadow_devidx];

	/*
	* Note: We don't want to update the repair stats
	* because that would incorrectly indicate that there
	* was bad data to repair, which we aren't sure about.
	* By clearing the SCAN_THREAD flag, we prevent this
	* from happening, despite having the REPAIR flag set.
	* We need to set SELF_HEAL so that this i/o can't be
	* bypassed by zio_vdev_io_start().
	*/
	zio_t *cio = zio_vdev_child_io(zio, NULL, cvd,
	rc->rc_shadow_offset, rc->rc_abd, rc->rc_size,
	ZIO_TYPE_WRITE, ZIO_PRIORITY_ASYNC_WRITE,
	ZIO_FLAG_IO_REPAIR \| ZIO_FLAG_SELF_HEAL,
	NULL, NULL);
	cio->io_flags &= ~ZIO_FLAG_SCAN_THREAD;
	zio_nowait(cio);
	}
	}
	}

	static void
	raidz_restore_orig_data(raidz_map_t *rm)
	{
	for (int i = 0; i < rm->rm_nrows; i++) {
	raidz_row_t *rr = rm->rm_row[i];
	for (int c = 0; c < rr->rr_cols; c++) {
	raidz_col_t *rc = &rr->rr_col[c];
	if (rc->rc_need_orig_restore) {
	abd_copy(rc->rc_abd,
	rc->rc_orig_data, rc->rc_size);
	rc->rc_need_orig_restore = B_FALSE;
	}
	}
	}
	}

	/*
	* During raidz_reconstruct() for expanded VDEV, we need special consideration
	* failure simulations. See note in raidz_reconstruct() on simulating failure
	* of a pre-expansion device.
	*
	* Treating logical child i as failed, return TRUE if the given column should
	* be treated as failed. The idea of logical children allows us to imagine
	* that a disk silently failed before a RAIDZ expansion (reads from this disk
	* succeed but return the wrong data). Since the expansion doesn't verify
	* checksums, the incorrect data will be moved to new locations spread among
	* the children (going diagonally across them).
	*
	* Higher "logical child failures" (values of `i`) indicate these
	* "pre-expansion failures". The first physical_width values imagine that a
	* current child failed; the next physical_width-1 values imagine that a
	* child failed before the most recent expansion; the next physical_width-2
	* values imagine a child failed in the expansion before that, etc.
	*/
	static boolean_t
	raidz_simulate_failure(int physical_width, int original_width, int ashift,
	int i, raidz_col_t *rc)
	{
	uint64_t sector_id =
	physical_width * (rc->rc_offset >> ashift) +
	rc->rc_devidx;

	for (int w = physical_width; w >= original_width; w--) {
	if (i < w) {
	return (sector_id % w == i);
	} else {
	i -= w;
	}
	}
	ASSERT(!"invalid logical child id");
	return (B_FALSE);
	}

	/*
	* returns EINVAL if reconstruction of the block will not be possible
	* returns ECKSUM if this specific reconstruction failed
	* returns 0 on successful reconstruction
	*/
	static int
	raidz_reconstruct(zio_t zio, int ltgts, int ntgts, int nparity)
	{
	raidz_map_t *rm = zio->io_vsd;
	int physical_width = zio->io_vd->vdev_children;
	int original_width = (rm->rm_original_width != 0) ?
	rm->rm_original_width : physical_width;
	int dbgmsg = zfs_flags & ZFS_DEBUG_RAIDZ_RECONSTRUCT;

	if (dbgmsg) {
	zfs_dbgmsg("raidz_reconstruct_expanded(zio=%px ltgts=%u,%u,%u "
	"ntgts=%u", zio, ltgts[0], ltgts[1], ltgts[2], ntgts);
	}

	/* Reconstruct each row */
	for (int r = 0; r < rm->rm_nrows; r++) {
	raidz_row_t *rr = rm->rm_row[r];
	int my_tgts[VDEV_RAIDZ_MAXPARITY]; /* value is child id */
	int t = 0;
	int dead = 0;
	int dead_data = 0;

	if (dbgmsg)
	zfs_dbgmsg("raidz_reconstruct_expanded(row=%u)", r);

	for (int c = 0; c < rr->rr_cols; c++) {
	raidz_col_t *rc = &rr->rr_col[c];
	ASSERT0(rc->rc_need_orig_restore);
	if (rc->rc_error != 0) {
	dead++;
	if (c >= nparity)
	dead_data++;
	continue;
	}
	if (rc->rc_size == 0)
	continue;
	for (int lt = 0; lt < ntgts; lt++) {
	if (raidz_simulate_failure(physical_width,
	original_width,
	zio->io_vd->vdev_top->vdev_ashift,
	ltgts[lt], rc)) {
	if (rc->rc_orig_data == NULL) {
	rc->rc_orig_data =
	abd_alloc_linear(
	rc->rc_size, B_TRUE);
	abd_copy(rc->rc_orig_data,
	rc->rc_abd, rc->rc_size);
	}
	rc->rc_need_orig_restore = B_TRUE;

	dead++;
	if (c >= nparity)
	dead_data++;
	/*
	* Note: simulating failure of a
	* pre-expansion device can hit more
	* than one column, in which case we
	* might try to simulate more failures
	* than can be reconstructed, which is
	* also more than the size of my_tgts.
	* This check prevents accessing past
	* the end of my_tgts. The "dead >
	* nparity" check below will fail this
	* reconstruction attempt.
	*/
	if (t < VDEV_RAIDZ_MAXPARITY) {
	my_tgts[t++] = c;
	if (dbgmsg) {
	zfs_dbgmsg("simulating "
	"failure of col %u "
	"devidx %u", c,
	(int)rc->rc_devidx);
	}
	}
	break;
	}
	}
	}
	if (dead > nparity) {
	/* reconstruction not possible */
	if (dbgmsg) {
	zfs_dbgmsg("reconstruction not possible; "
	"too many failures");
	}
	raidz_restore_orig_data(rm);
	return (EINVAL);
	}
	if (dead_data > 0)
	vdev_raidz_reconstruct_row(rm, rr, my_tgts, t);
	}

	/* Check for success */
	if (raidz_checksum_verify(zio) == 0) {

	/* Reconstruction succeeded - report errors */
	for (int i = 0; i < rm->rm_nrows; i++) {
	raidz_row_t *rr = rm->rm_row[i];

	for (int c = 0; c < rr->rr_cols; c++) {
	raidz_col_t *rc = &rr->rr_col[c];
	if (rc->rc_need_orig_restore) {
	/*
	* Note: if this is a parity column,
	* we don't really know if it's wrong.
	* We need to let
	* vdev_raidz_io_done_verified() check
	* it, and if we set rc_error, it will
	* think that it is a "known" error
	* that doesn't need to be checked
	* or corrected.
	*/
	if (rc->rc_error == 0 &&
	c >= rr->rr_firstdatacol) {
	vdev_raidz_checksum_error(zio,
	rc, rc->rc_orig_data);
	rc->rc_error =
	SET_ERROR(ECKSUM);
	}
	rc->rc_need_orig_restore = B_FALSE;
	}
	}

	vdev_raidz_io_done_verified(zio, rr);
	}

	zio_checksum_verified(zio);

	if (dbgmsg) {
	zfs_dbgmsg("reconstruction successful "
	"(checksum verified)");
	}
	return (0);
	}

	/* Reconstruction failed - restore original data */
	raidz_restore_orig_data(rm);
	if (dbgmsg) {
	zfs_dbgmsg("raidz_reconstruct_expanded(zio=%px) checksum "
	"failed", zio);
	}
	return (ECKSUM);
	}

	/*
	* Iterate over all combinations of N bad vdevs and attempt a reconstruction.
	* Note that the algorithm below is non-optimal because it doesn't take into
	* account how reconstruction is actually performed. For example, with
	* triple-parity RAID-Z the reconstruction procedure is the same if column 4
	* is targeted as invalid as if columns 1 and 4 are targeted since in both
	* cases we'd only use parity information in column 0.
	*
	* The order that we find the various possible combinations of failed
	* disks is dictated by these rules:
	* - Examine each "slot" (the "i" in tgts[i])
	* - Try to increment this slot (tgts[i] += 1)
	* - if we can't increment because it runs into the next slot,
	* reset our slot to the minimum, and examine the next slot
	*
	* For example, with a 6-wide RAIDZ3, and no known errors (so we have to choose
	* 3 columns to reconstruct), we will generate the following sequence:
	*
	* STATE ACTION
	* 0 1 2 special case: skip since these are all parity
	* 0 1 3 first slot: reset to 0; middle slot: increment to 2
	* 0 2 3 first slot: increment to 1
	* 1 2 3 first: reset to 0; middle: reset to 1; last: increment to 4
	* 0 1 4 first: reset to 0; middle: increment to 2
	* 0 2 4 first: increment to 1
	* 1 2 4 first: reset to 0; middle: increment to 3
	* 0 3 4 first: increment to 1
	* 1 3 4 first: increment to 2
	* 2 3 4 first: reset to 0; middle: reset to 1; last: increment to 5
	* 0 1 5 first: reset to 0; middle: increment to 2
	* 0 2 5 first: increment to 1
	* 1 2 5 first: reset to 0; middle: increment to 3
	* 0 3 5 first: increment to 1
	* 1 3 5 first: increment to 2
	* 2 3 5 first: reset to 0; middle: increment to 4
	* 0 4 5 first: increment to 1
	* 1 4 5 first: increment to 2
	* 2 4 5 first: increment to 3
	* 3 4 5 done
	*
	* This strategy works for dRAID but is less efficient when there are a large
	* number of child vdevs and therefore permutations to check. Furthermore,
	* since the raidz_map_t rows likely do not overlap, reconstruction would be
	* possible as long as there are no more than nparity data errors per row.
	* These additional permutations are not currently checked but could be as
	* a future improvement.
	*
	* Returns 0 on success, ECKSUM on failure.
	*/
	static int
	vdev_raidz_combrec(zio_t *zio)
	{
	int nparity = vdev_get_nparity(zio->io_vd);
	raidz_map_t *rm = zio->io_vsd;
	int physical_width = zio->io_vd->vdev_children;
	int original_width = (rm->rm_original_width != 0) ?
	rm->rm_original_width : physical_width;

	for (int i = 0; i < rm->rm_nrows; i++) {
	raidz_row_t *rr = rm->rm_row[i];
	int total_errors = 0;

	for (int c = 0; c < rr->rr_cols; c++) {
	if (rr->rr_col[c].rc_error)
	total_errors++;
	}

	if (total_errors > nparity)
	return (vdev_raidz_worst_error(rr));
	}

	for (int num_failures = 1; num_failures <= nparity; num_failures++) {
	int tstore[VDEV_RAIDZ_MAXPARITY + 2];
	int ltgts = &tstore[1]; / value is logical child ID */


	/*
	* Determine number of logical children, n. See comment
	* above raidz_simulate_failure().
	*/
	int n = 0;
	for (int w = physical_width;
	w >= original_width; w--) {
	n += w;
	}

	ASSERT3U(num_failures, <=, nparity);
	ASSERT3U(num_failures, <=, VDEV_RAIDZ_MAXPARITY);

	/* Handle corner cases in combrec logic */
	ltgts[-1] = -1;
	for (int i = 0; i < num_failures; i++) {
	ltgts[i] = i;
	}
	ltgts[num_failures] = n;

	for (;;) {
	int err = raidz_reconstruct(zio, ltgts, num_failures,
	nparity);
	if (err == EINVAL) {
	/*
	* Reconstruction not possible with this #
	* failures; try more failures.
	*/
	break;
	} else if (err == 0)
	return (0);

	/* Compute next targets to try */
	for (int t = 0; ; t++) {
	ASSERT3U(t, <, num_failures);
	ltgts[t]++;
	if (ltgts[t] == n) {
	/* try more failures */
	ASSERT3U(t, ==, num_failures - 1);
	if (zfs_flags &
	ZFS_DEBUG_RAIDZ_RECONSTRUCT) {
	zfs_dbgmsg("reconstruction "
	"failed for num_failures="
	"%u; tried all "
	"combinations",
	num_failures);
	}
	break;
	}

	ASSERT3U(ltgts[t], <, n);
	ASSERT3U(ltgts[t], <=, ltgts[t + 1]);

	/*
	* If that spot is available, we're done here.
	* Try the next combination.
	*/
	if (ltgts[t] != ltgts[t + 1])
	break; // found next combination

	/*
	* Otherwise, reset this tgt to the minimum,
	* and move on to the next tgt.
	*/
	ltgts[t] = ltgts[t - 1] + 1;
	ASSERT3U(ltgts[t], ==, t);
	}

	/* Increase the number of failures and keep trying. */
	if (ltgts[num_failures - 1] == n)
	break;
	}
	}
	if (zfs_flags & ZFS_DEBUG_RAIDZ_RECONSTRUCT)
	zfs_dbgmsg("reconstruction failed for all num_failures");
	return (ECKSUM);
	}

	void
	vdev_raidz_reconstruct(raidz_map_t rm, const int t, int nt)
	{
	for (uint64_t row = 0; row < rm->rm_nrows; row++) {
	raidz_row_t *rr = rm->rm_row[row];
	vdev_raidz_reconstruct_row(rm, rr, t, nt);
	}
	}

	/*
	* Complete a write IO operation on a RAIDZ VDev
	*
	* Outline:
	* 1. Check for errors on the child IOs.
	* 2. Return, setting an error code if too few child VDevs were written
	* to reconstruct the data later. Note that partial writes are
	* considered successful if they can be reconstructed at all.
	*/
	static void
	vdev_raidz_io_done_write_impl(zio_t zio, raidz_row_t rr)
	{
	int normal_errors = 0;
	int shadow_errors = 0;

	ASSERT3U(rr->rr_missingparity, <=, rr->rr_firstdatacol);
	ASSERT3U(rr->rr_missingdata, <=, rr->rr_cols - rr->rr_firstdatacol);
	ASSERT3U(zio->io_type, ==, ZIO_TYPE_WRITE);

	for (int c = 0; c < rr->rr_cols; c++) {
	raidz_col_t *rc = &rr->rr_col[c];

	if (rc->rc_error != 0) {
	ASSERT(rc->rc_error != ECKSUM); /* child has no bp */
	normal_errors++;
	}
	if (rc->rc_shadow_error != 0) {
	ASSERT(rc->rc_shadow_error != ECKSUM);
	shadow_errors++;
	}
	}

	/*
	* Treat partial writes as a success. If we couldn't write enough
	* columns to reconstruct the data, the I/O failed. Otherwise, good
	* enough. Note that in the case of a shadow write (during raidz
	* expansion), depending on if we crash, either the normal (old) or
	* shadow (new) location may become the "real" version of the block,
	* so both locations must have sufficient redundancy.
	*
	* Now that we support write reallocation, it would be better
	* to treat partial failure as real failure unless there are
	* no non-degraded top-level vdevs left, and not update DTLs
	* if we intend to reallocate.
	*/
	if (normal_errors > rr->rr_firstdatacol \|\|
	shadow_errors > rr->rr_firstdatacol) {
	zio->io_error = zio_worst_error(zio->io_error,
	vdev_raidz_worst_error(rr));
	}
	}

	static void
	vdev_raidz_io_done_reconstruct_known_missing(zio_t zio, raidz_map_t rm,
	raidz_row_t *rr)
	{
	int parity_errors = 0;
	int parity_untried = 0;
	int data_errors = 0;
	int total_errors = 0;

	ASSERT3U(rr->rr_missingparity, <=, rr->rr_firstdatacol);
	ASSERT3U(rr->rr_missingdata, <=, rr->rr_cols - rr->rr_firstdatacol);

	for (int c = 0; c < rr->rr_cols; c++) {
	raidz_col_t *rc = &rr->rr_col[c];

	/*
	* If scrubbing and a replacing/sparing child vdev determined
	* that not all of its children have an identical copy of the
	* data, then clear the error so the column is treated like
	* any other read and force a repair to correct the damage.
	*/
	if (rc->rc_error == ECKSUM) {
	ASSERT(zio->io_flags & ZIO_FLAG_SCRUB);
	vdev_raidz_checksum_error(zio, rc, rc->rc_abd);
	rc->rc_force_repair = 1;
	rc->rc_error = 0;
	}

	if (rc->rc_error) {
	if (c < rr->rr_firstdatacol)
	parity_errors++;
	else
	data_errors++;

	total_errors++;
	} else if (c < rr->rr_firstdatacol && !rc->rc_tried) {
	parity_untried++;
	}
	}

	/*
	* If there were data errors and the number of errors we saw was
	* correctable -- less than or equal to the number of parity disks read
	* -- reconstruct based on the missing data.
	*/
	if (data_errors != 0 &&
	total_errors <= rr->rr_firstdatacol - parity_untried) {
	/*
	* We either attempt to read all the parity columns or
	* none of them. If we didn't try to read parity, we
	* wouldn't be here in the correctable case. There must
	* also have been fewer parity errors than parity
	* columns or, again, we wouldn't be in this code path.
	*/
	ASSERT(parity_untried == 0);
	ASSERT(parity_errors < rr->rr_firstdatacol);

	/*
	* Identify the data columns that reported an error.
	*/
	int n = 0;
	int tgts[VDEV_RAIDZ_MAXPARITY];
	for (int c = rr->rr_firstdatacol; c < rr->rr_cols; c++) {
	raidz_col_t *rc = &rr->rr_col[c];
	if (rc->rc_error != 0) {
	ASSERT(n < VDEV_RAIDZ_MAXPARITY);
	tgts[n++] = c;
	}
	}

	ASSERT(rr->rr_firstdatacol >= n);

	vdev_raidz_reconstruct_row(rm, rr, tgts, n);
	}
	}

	/*
	* Return the number of reads issued.
	*/
	static int
	vdev_raidz_read_all(zio_t zio, raidz_row_t rr)
	{
	vdev_t *vd = zio->io_vd;
	int nread = 0;

	rr->rr_missingdata = 0;
	rr->rr_missingparity = 0;

	/*
	* If this rows contains empty sectors which are not required
	* for a normal read then allocate an ABD for them now so they
	* may be read, verified, and any needed repairs performed.
	*/
	if (rr->rr_nempty != 0 && rr->rr_abd_empty == NULL)
	vdev_draid_map_alloc_empty(zio, rr);

	for (int c = 0; c < rr->rr_cols; c++) {
	raidz_col_t *rc = &rr->rr_col[c];
	if (rc->rc_tried \|\| rc->rc_size == 0)
	continue;

	zio_nowait(zio_vdev_child_io(zio, NULL,
	vd->vdev_child[rc->rc_devidx],
	rc->rc_offset, rc->rc_abd, rc->rc_size,
	zio->io_type, zio->io_priority, 0,
	vdev_raidz_child_done, rc));
	nread++;
	}
	return (nread);
	}

	/*
	* We're here because either there were too many errors to even attempt
	* reconstruction (total_errors == rm_first_datacol), or vdev_*_combrec()
	* failed. In either case, there is enough bad data to prevent reconstruction.
	* Start checksum ereports for all children which haven't failed.
	*/
	static void
	vdev_raidz_io_done_unrecoverable(zio_t *zio)
	{
	raidz_map_t *rm = zio->io_vsd;

	for (int i = 0; i < rm->rm_nrows; i++) {
	raidz_row_t *rr = rm->rm_row[i];

	for (int c = 0; c < rr->rr_cols; c++) {
	raidz_col_t *rc = &rr->rr_col[c];
	vdev_t *cvd = zio->io_vd->vdev_child[rc->rc_devidx];

	if (rc->rc_error != 0)
	continue;

	zio_bad_cksum_t zbc;
	zbc.zbc_has_cksum = 0;
	zbc.zbc_injected = rm->rm_ecksuminjected;

	mutex_enter(&cvd->vdev_stat_lock);
	cvd->vdev_stat.vs_checksum_errors++;
	mutex_exit(&cvd->vdev_stat_lock);
	(void) zfs_ereport_start_checksum(zio->io_spa,
	cvd, &zio->io_bookmark, zio, rc->rc_offset,
	rc->rc_size, &zbc);
	}
	}
	}

	void
	vdev_raidz_io_done(zio_t *zio)
	{
	raidz_map_t *rm = zio->io_vsd;

	ASSERT(zio->io_bp != NULL);
	if (zio->io_type == ZIO_TYPE_WRITE) {
	for (int i = 0; i < rm->rm_nrows; i++) {
	vdev_raidz_io_done_write_impl(zio, rm->rm_row[i]);
	}
	} else {
	if (rm->rm_phys_col) {
	/*
	* This is an aggregated read. Copy the data and status
	* from the aggregate abd's to the individual rows.
	*/
	for (int i = 0; i < rm->rm_nrows; i++) {
	raidz_row_t *rr = rm->rm_row[i];

	for (int c = 0; c < rr->rr_cols; c++) {
	raidz_col_t *rc = &rr->rr_col[c];
	if (rc->rc_tried \|\| rc->rc_size == 0)
	continue;

	raidz_col_t *prc =
	&rm->rm_phys_col[rc->rc_devidx];
	rc->rc_error = prc->rc_error;
	rc->rc_tried = prc->rc_tried;
	rc->rc_skipped = prc->rc_skipped;
	if (c >= rr->rr_firstdatacol) {
	/*
	* Note: this is slightly faster
	* than using abd_copy_off().
	*/
	char *physbuf = abd_to_buf(
	prc->rc_abd);
	void *physloc = physbuf +
	rc->rc_offset -
	prc->rc_offset;

	abd_copy_from_buf(rc->rc_abd,
	physloc, rc->rc_size);
	}
	}
	}
	}

	for (int i = 0; i < rm->rm_nrows; i++) {
	raidz_row_t *rr = rm->rm_row[i];
	vdev_raidz_io_done_reconstruct_known_missing(zio,
	rm, rr);
	}

	if (raidz_checksum_verify(zio) == 0) {
	for (int i = 0; i < rm->rm_nrows; i++) {
	raidz_row_t *rr = rm->rm_row[i];
	vdev_raidz_io_done_verified(zio, rr);
	}
	zio_checksum_verified(zio);
	} else {
	/*
	* A sequential resilver has no checksum which makes
	* combinatoral reconstruction impossible. This code
	* path is unreachable since raidz_checksum_verify()
	* has no checksum to verify and must succeed.
	*/
	ASSERT3U(zio->io_priority, !=, ZIO_PRIORITY_REBUILD);

	/*
	* This isn't a typical situation -- either we got a
	* read error or a child silently returned bad data.
	* Read every block so we can try again with as much
	* data and parity as we can track down. If we've
	* already been through once before, all children will
	* be marked as tried so we'll proceed to combinatorial
	* reconstruction.
	*/
	int nread = 0;
	for (int i = 0; i < rm->rm_nrows; i++) {
	nread += vdev_raidz_read_all(zio,
	rm->rm_row[i]);
	}
	if (nread != 0) {
	/*
	* Normally our stage is VDEV_IO_DONE, but if
	* we've already called redone(), it will have
	* changed to VDEV_IO_START, in which case we
	* don't want to call redone() again.
	*/
	if (zio->io_stage != ZIO_STAGE_VDEV_IO_START)
	zio_vdev_io_redone(zio);
	return;
	}
	/*
	* It would be too expensive to try every possible
	* combination of failed sectors in every row, so
	* instead we try every combination of failed current or
	* past physical disk. This means that if the incorrect
	* sectors were all on Nparity disks at any point in the
	* past, we will find the correct data. The only known
	* case where this is less durable than a non-expanded
	* RAIDZ, is if we have a silent failure during
	* expansion. In that case, one block could be
	* partially in the old format and partially in the
	* new format, so we'd lost some sectors from the old
	* format and some from the new format.
	*
	* e.g. logical_width=4 physical_width=6
	* the 15 (6+5+4) possible failed disks are:
	* width=6 child=0
	* width=6 child=1
	* width=6 child=2
	* width=6 child=3
	* width=6 child=4
	* width=6 child=5
	* width=5 child=0
	* width=5 child=1
	* width=5 child=2
	* width=5 child=3
	* width=5 child=4
	* width=4 child=0
	* width=4 child=1
	* width=4 child=2
	* width=4 child=3
	* And we will try every combination of Nparity of these
	* failing.
	*
	* As a first pass, we can generate every combo,
	* and try reconstructing, ignoring any known
	* failures. If any row has too many known + simulated
	* failures, then we bail on reconstructing with this
	* number of simulated failures. As an improvement,
	* we could detect the number of whole known failures
	* (i.e. we have known failures on these disks for
	* every row; the disks never succeeded), and
	* subtract that from the max # failures to simulate.
	* We could go even further like the current
	* combrec code, but that doesn't seem like it
	* gains us very much. If we simulate a failure
	* that is also a known failure, that's fine.
	*/
	zio->io_error = vdev_raidz_combrec(zio);
	if (zio->io_error == ECKSUM &&
	!(zio->io_flags & ZIO_FLAG_SPECULATIVE)) {
	vdev_raidz_io_done_unrecoverable(zio);
	}
	}
	}
	if (rm->rm_lr != NULL) {
	zfs_rangelock_exit(rm->rm_lr);
	rm->rm_lr = NULL;
	}
	}

	static void
	vdev_raidz_state_change(vdev_t *vd, int faulted, int degraded)
	{
	vdev_raidz_t *vdrz = vd->vdev_tsd;
	if (faulted > vdrz->vd_nparity)
	vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
	VDEV_AUX_NO_REPLICAS);
	else if (degraded + faulted != 0)
	vdev_set_state(vd, B_FALSE, VDEV_STATE_DEGRADED, VDEV_AUX_NONE);
	else
	vdev_set_state(vd, B_FALSE, VDEV_STATE_HEALTHY, VDEV_AUX_NONE);
	}

	/*
	* Determine if any portion of the provided block resides on a child vdev
	* with a dirty DTL and therefore needs to be resilvered. The function
	* assumes that at least one DTL is dirty which implies that full stripe
	* width blocks must be resilvered.
	*/
	static boolean_t
	vdev_raidz_need_resilver(vdev_t vd, const dva_t dva, size_t psize,
	uint64_t phys_birth)
	{
	vdev_raidz_t *vdrz = vd->vdev_tsd;

	/*
	* If we're in the middle of a RAIDZ expansion, this block may be in
	* the old and/or new location. For simplicity, always resilver it.
	*/
	if (vdrz->vn_vre.vre_state == DSS_SCANNING)
	return (B_TRUE);

	uint64_t dcols = vd->vdev_children;
	uint64_t nparity = vdrz->vd_nparity;
	uint64_t ashift = vd->vdev_top->vdev_ashift;
	/* The starting RAIDZ (parent) vdev sector of the block. */
	uint64_t b = DVA_GET_OFFSET(dva) >> ashift;
	/* The zio's size in units of the vdev's minimum sector size. */
	uint64_t s = ((psize - 1) >> ashift) + 1;
	/* The first column for this stripe. */
	uint64_t f = b % dcols;

	/* Unreachable by sequential resilver. */
	ASSERT3U(phys_birth, !=, TXG_UNKNOWN);

	if (!vdev_dtl_contains(vd, DTL_PARTIAL, phys_birth, 1))
	return (B_FALSE);

	if (s + nparity >= dcols)
	return (B_TRUE);

	for (uint64_t c = 0; c < s + nparity; c++) {
	uint64_t devidx = (f + c) % dcols;
	vdev_t *cvd = vd->vdev_child[devidx];

	/*
	* dsl_scan_need_resilver() already checked vd with
	* vdev_dtl_contains(). So here just check cvd with
	* vdev_dtl_empty(), cheaper and a good approximation.
	*/
	if (!vdev_dtl_empty(cvd, DTL_PARTIAL))
	return (B_TRUE);
	}

	return (B_FALSE);
	}

	static void
	vdev_raidz_xlate(vdev_t cvd, const range_seg64_t logical_rs,
	range_seg64_t physical_rs, range_seg64_t remain_rs)
	{
	(void) remain_rs;

	vdev_t *raidvd = cvd->vdev_parent;
	ASSERT(raidvd->vdev_ops == &vdev_raidz_ops);

	vdev_raidz_t *vdrz = raidvd->vdev_tsd;

	if (vdrz->vn_vre.vre_state == DSS_SCANNING) {
	/*
	* We're in the middle of expansion, in which case the
	* translation is in flux. Any answer we give may be wrong
	* by the time we return, so it isn't safe for the caller to
	* act on it. Therefore we say that this range isn't present
	* on any children. The only consumers of this are "zpool
	* initialize" and trimming, both of which are "best effort"
	* anyway.
	*/
	physical_rs->rs_start = physical_rs->rs_end = 0;
	remain_rs->rs_start = remain_rs->rs_end = 0;
	return;
	}

	uint64_t width = vdrz->vd_physical_width;
	uint64_t tgt_col = cvd->vdev_id;
	uint64_t ashift = raidvd->vdev_top->vdev_ashift;

	/* make sure the offsets are block-aligned */
	ASSERT0(logical_rs->rs_start % (1 << ashift));
	ASSERT0(logical_rs->rs_end % (1 << ashift));
	uint64_t b_start = logical_rs->rs_start >> ashift;
	uint64_t b_end = logical_rs->rs_end >> ashift;

	uint64_t start_row = 0;
	if (b_start > tgt_col) /* avoid underflow */
	start_row = ((b_start - tgt_col - 1) / width) + 1;

	uint64_t end_row = 0;
	if (b_end > tgt_col)
	end_row = ((b_end - tgt_col - 1) / width) + 1;

	physical_rs->rs_start = start_row << ashift;
	physical_rs->rs_end = end_row << ashift;

	ASSERT3U(physical_rs->rs_start, <=, logical_rs->rs_start);
	ASSERT3U(physical_rs->rs_end - physical_rs->rs_start, <=,
	logical_rs->rs_end - logical_rs->rs_start);
	}

	static void
	raidz_reflow_sync(void arg, dmu_tx_t tx)
	{
	spa_t *spa = arg;
	int txgoff = dmu_tx_get_txg(tx) & TXG_MASK;
	vdev_raidz_expand_t *vre = spa->spa_raidz_expand;

	/*
	* Ensure there are no i/os to the range that is being committed.
	*/
	uint64_t old_offset = RRSS_GET_OFFSET(&spa->spa_uberblock);
	ASSERT3U(vre->vre_offset_pertxg[txgoff], >=, old_offset);

	mutex_enter(&vre->vre_lock);
	uint64_t new_offset =
	MIN(vre->vre_offset_pertxg[txgoff], vre->vre_failed_offset);
	/*
	* We should not have committed anything that failed.
	*/
	VERIFY3U(vre->vre_failed_offset, >=, old_offset);
	mutex_exit(&vre->vre_lock);

	zfs_locked_range_t *lr = zfs_rangelock_enter(&vre->vre_rangelock,
	old_offset, new_offset - old_offset,
	RL_WRITER);

	/*
	* Update the uberblock that will be written when this txg completes.
	*/
	RAIDZ_REFLOW_SET(&spa->spa_uberblock,
	RRSS_SCRATCH_INVALID_SYNCED_REFLOW, new_offset);
	vre->vre_offset_pertxg[txgoff] = 0;
	zfs_rangelock_exit(lr);

	mutex_enter(&vre->vre_lock);
	vre->vre_bytes_copied += vre->vre_bytes_copied_pertxg[txgoff];
	vre->vre_bytes_copied_pertxg[txgoff] = 0;
	mutex_exit(&vre->vre_lock);

	vdev_t *vd = vdev_lookup_top(spa, vre->vre_vdev_id);
	VERIFY0(zap_update(spa->spa_meta_objset,
	vd->vdev_top_zap, VDEV_TOP_ZAP_RAIDZ_EXPAND_BYTES_COPIED,
	sizeof (vre->vre_bytes_copied), 1, &vre->vre_bytes_copied, tx));
	}

	static void
	raidz_reflow_complete_sync(void arg, dmu_tx_t tx)
	{
	spa_t *spa = arg;
	vdev_raidz_expand_t *vre = spa->spa_raidz_expand;
	vdev_t *raidvd = vdev_lookup_top(spa, vre->vre_vdev_id);
	vdev_raidz_t *vdrz = raidvd->vdev_tsd;

	for (int i = 0; i < TXG_SIZE; i++)
	VERIFY0(vre->vre_offset_pertxg[i]);

	reflow_node_t re = kmem_zalloc(sizeof (re), KM_SLEEP);
	re->re_txg = tx->tx_txg + TXG_CONCURRENT_STATES;
	re->re_logical_width = vdrz->vd_physical_width;
	mutex_enter(&vdrz->vd_expand_lock);
	avl_add(&vdrz->vd_expand_txgs, re);
	mutex_exit(&vdrz->vd_expand_lock);

	vdev_t *vd = vdev_lookup_top(spa, vre->vre_vdev_id);

	/*
	* Dirty the config so that the updated ZPOOL_CONFIG_RAIDZ_EXPAND_TXGS
	* will get written (based on vd_expand_txgs).
	*/
	vdev_config_dirty(vd);

	/*
	* Before we change vre_state, the on-disk state must reflect that we
	* have completed all copying, so that vdev_raidz_io_start() can use
	* vre_state to determine if the reflow is in progress. See also the
	* end of spa_raidz_expand_thread().
	*/
	VERIFY3U(RRSS_GET_OFFSET(&spa->spa_ubsync), ==,
	raidvd->vdev_ms_count << raidvd->vdev_ms_shift);

	vre->vre_end_time = gethrestime_sec();
	vre->vre_state = DSS_FINISHED;

	uint64_t state = vre->vre_state;
	VERIFY0(zap_update(spa->spa_meta_objset,
	vd->vdev_top_zap, VDEV_TOP_ZAP_RAIDZ_EXPAND_STATE,
	sizeof (state), 1, &state, tx));

	uint64_t end_time = vre->vre_end_time;
	VERIFY0(zap_update(spa->spa_meta_objset,
	vd->vdev_top_zap, VDEV_TOP_ZAP_RAIDZ_EXPAND_END_TIME,
	sizeof (end_time), 1, &end_time, tx));

	spa->spa_uberblock.ub_raidz_reflow_info = 0;

	spa_history_log_internal(spa, "raidz vdev expansion completed", tx,
	"%s vdev %llu new width %llu", spa_name(spa),
	(unsigned long long)vd->vdev_id,
	(unsigned long long)vd->vdev_children);

	spa->spa_raidz_expand = NULL;
	raidvd->vdev_rz_expanding = B_FALSE;

	spa_async_request(spa, SPA_ASYNC_INITIALIZE_RESTART);
	spa_async_request(spa, SPA_ASYNC_TRIM_RESTART);
	spa_async_request(spa, SPA_ASYNC_AUTOTRIM_RESTART);

	spa_notify_waiters(spa);

	/*
	* While we're in syncing context take the opportunity to
	* setup a scrub. All the data has been sucessfully copied
	* but we have not validated any checksums.
	*/
	pool_scan_func_t func = POOL_SCAN_SCRUB;
	if (zfs_scrub_after_expand && dsl_scan_setup_check(&func, tx) == 0)
	dsl_scan_setup_sync(&func, tx);
	}

	/*
	* Struct for one copy zio.
	*/
	typedef struct raidz_reflow_arg {
	vdev_raidz_expand_t *rra_vre;
	zfs_locked_range_t *rra_lr;
	uint64_t rra_txg;
	} raidz_reflow_arg_t;

	/*
	* The write of the new location is done.
	*/
	static void
	raidz_reflow_write_done(zio_t *zio)
	{
	raidz_reflow_arg_t *rra = zio->io_private;
	vdev_raidz_expand_t *vre = rra->rra_vre;

	abd_free(zio->io_abd);

	mutex_enter(&vre->vre_lock);
	if (zio->io_error != 0) {
	/* Force a reflow pause on errors */
	vre->vre_failed_offset =
	MIN(vre->vre_failed_offset, rra->rra_lr->lr_offset);
	}
	ASSERT3U(vre->vre_outstanding_bytes, >=, zio->io_size);
	vre->vre_outstanding_bytes -= zio->io_size;
	if (rra->rra_lr->lr_offset + rra->rra_lr->lr_length <
	vre->vre_failed_offset) {
	vre->vre_bytes_copied_pertxg[rra->rra_txg & TXG_MASK] +=
	zio->io_size;
	}
	cv_signal(&vre->vre_cv);
	mutex_exit(&vre->vre_lock);

	zfs_rangelock_exit(rra->rra_lr);

	kmem_free(rra, sizeof (*rra));
	spa_config_exit(zio->io_spa, SCL_STATE, zio->io_spa);
	}

	/*
	* The read of the old location is done. The parent zio is the write to
	* the new location. Allow it to start.
	*/
	static void
	raidz_reflow_read_done(zio_t *zio)
	{
	raidz_reflow_arg_t *rra = zio->io_private;
	vdev_raidz_expand_t *vre = rra->rra_vre;

	/*
	* If the read failed, or if it was done on a vdev that is not fully
	* healthy (e.g. a child that has a resilver in progress), we may not
	* have the correct data. Note that it's OK if the write proceeds.
	* It may write garbage but the location is otherwise unused and we
	* will retry later due to vre_failed_offset.
	*/
	if (zio->io_error != 0 \|\| !vdev_dtl_empty(zio->io_vd, DTL_MISSING)) {
	zfs_dbgmsg("reflow read failed off=%llu size=%llu txg=%llu "
	"err=%u partial_dtl_empty=%u missing_dtl_empty=%u",
	(long long)rra->rra_lr->lr_offset,
	(long long)rra->rra_lr->lr_length,
	(long long)rra->rra_txg,
	zio->io_error,
	vdev_dtl_empty(zio->io_vd, DTL_PARTIAL),
	vdev_dtl_empty(zio->io_vd, DTL_MISSING));
	mutex_enter(&vre->vre_lock);
	/* Force a reflow pause on errors */
	vre->vre_failed_offset =
	MIN(vre->vre_failed_offset, rra->rra_lr->lr_offset);
	mutex_exit(&vre->vre_lock);
	}

	zio_nowait(zio_unique_parent(zio));
	}

	static void
	raidz_reflow_record_progress(vdev_raidz_expand_t *vre, uint64_t offset,
	dmu_tx_t *tx)
	{
	int txgoff = dmu_tx_get_txg(tx) & TXG_MASK;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;

	if (offset == 0)
	return;

	mutex_enter(&vre->vre_lock);
	ASSERT3U(vre->vre_offset, <=, offset);
	vre->vre_offset = offset;
	mutex_exit(&vre->vre_lock);

	if (vre->vre_offset_pertxg[txgoff] == 0) {
	dsl_sync_task_nowait(dmu_tx_pool(tx), raidz_reflow_sync,
	spa, tx);
	}
	vre->vre_offset_pertxg[txgoff] = offset;
	}

	static boolean_t
	vdev_raidz_expand_child_replacing(vdev_t *raidz_vd)
	{
	for (int i = 0; i < raidz_vd->vdev_children; i++) {
	/* Quick check if a child is being replaced */
	if (!raidz_vd->vdev_child[i]->vdev_ops->vdev_op_leaf)
	return (B_TRUE);
	}
	return (B_FALSE);
	}

	static boolean_t
	raidz_reflow_impl(vdev_t vd, vdev_raidz_expand_t vre, range_tree_t *rt,
	dmu_tx_t *tx)
	{
	spa_t *spa = vd->vdev_spa;
	int ashift = vd->vdev_top->vdev_ashift;
	uint64_t offset, size;

	if (!range_tree_find_in(rt, 0, vd->vdev_top->vdev_asize,
	&offset, &size)) {
	return (B_FALSE);
	}
	ASSERT(IS_P2ALIGNED(offset, 1 << ashift));
	ASSERT3U(size, >=, 1 << ashift);
	uint64_t length = 1 << ashift;
	int txgoff = dmu_tx_get_txg(tx) & TXG_MASK;

	uint64_t blkid = offset >> ashift;

	int old_children = vd->vdev_children - 1;

	/*
	* We can only progress to the point that writes will not overlap
	* with blocks whose progress has not yet been recorded on disk.
	* Since partially-copied rows are still read from the old location,
	* we need to stop one row before the sector-wise overlap, to prevent
	* row-wise overlap.
	*
	* Note that even if we are skipping over a large unallocated region,
	* we can't move the on-disk progress to `offset`, because concurrent
	* writes/allocations could still use the currently-unallocated
	* region.
	*/
	uint64_t ubsync_blkid =
	RRSS_GET_OFFSET(&spa->spa_ubsync) >> ashift;
	uint64_t next_overwrite_blkid = ubsync_blkid +
	ubsync_blkid / old_children - old_children;
	VERIFY3U(next_overwrite_blkid, >, ubsync_blkid);

	if (blkid >= next_overwrite_blkid) {
	raidz_reflow_record_progress(vre,
	next_overwrite_blkid << ashift, tx);
	return (B_TRUE);
	}

	range_tree_remove(rt, offset, length);

	raidz_reflow_arg_t rra = kmem_zalloc(sizeof (rra), KM_SLEEP);
	rra->rra_vre = vre;
	rra->rra_lr = zfs_rangelock_enter(&vre->vre_rangelock,
	offset, length, RL_WRITER);
	rra->rra_txg = dmu_tx_get_txg(tx);

	raidz_reflow_record_progress(vre, offset + length, tx);

	mutex_enter(&vre->vre_lock);
	vre->vre_outstanding_bytes += length;
	mutex_exit(&vre->vre_lock);

	/*
	* SCL_STATE will be released when the read and write are done,
	* by raidz_reflow_write_done().
	*/
	spa_config_enter(spa, SCL_STATE, spa, RW_READER);

	/* check if a replacing vdev was added, if so treat it as an error */
	if (vdev_raidz_expand_child_replacing(vd)) {
	zfs_dbgmsg("replacing vdev encountered, reflow paused at "
	"offset=%llu txg=%llu",
	(long long)rra->rra_lr->lr_offset,
	(long long)rra->rra_txg);

	mutex_enter(&vre->vre_lock);
	vre->vre_failed_offset =
	MIN(vre->vre_failed_offset, rra->rra_lr->lr_offset);
	cv_signal(&vre->vre_cv);
	mutex_exit(&vre->vre_lock);

	/* drop everything we acquired */
	zfs_rangelock_exit(rra->rra_lr);
	kmem_free(rra, sizeof (*rra));
	spa_config_exit(spa, SCL_STATE, spa);
	return (B_TRUE);
	}

	zio_t *pio = spa->spa_txg_zio[txgoff];
	abd_t *abd = abd_alloc_for_io(length, B_FALSE);
	zio_t *write_zio = zio_vdev_child_io(pio, NULL,
	vd->vdev_child[blkid % vd->vdev_children],
	(blkid / vd->vdev_children) << ashift,
	abd, length,
	ZIO_TYPE_WRITE, ZIO_PRIORITY_REMOVAL,
	ZIO_FLAG_CANFAIL,
	raidz_reflow_write_done, rra);

	zio_nowait(zio_vdev_child_io(write_zio, NULL,
	vd->vdev_child[blkid % old_children],
	(blkid / old_children) << ashift,
	abd, length,
	ZIO_TYPE_READ, ZIO_PRIORITY_REMOVAL,
	ZIO_FLAG_CANFAIL,
	raidz_reflow_read_done, rra));

	return (B_FALSE);
	}

	/*
	* For testing (ztest specific)
	*/
	static void
	raidz_expand_pause(uint_t pause_point)
	{
	while (raidz_expand_pause_point != 0 &&
	raidz_expand_pause_point <= pause_point)
	delay(hz);
	}

	static void
	raidz_scratch_child_done(zio_t *zio)
	{
	zio_t *pio = zio->io_private;

	mutex_enter(&pio->io_lock);
	pio->io_error = zio_worst_error(pio->io_error, zio->io_error);
	mutex_exit(&pio->io_lock);
	}

	/*
	* Reflow the beginning portion of the vdev into an intermediate scratch area
	* in memory and on disk. This operation must be persisted on disk before we
	* proceed to overwrite the beginning portion with the reflowed data.
	*
	* This multi-step task can fail to complete if disk errors are encountered
	* and we can return here after a pause (waiting for disk to become healthy).
	*/
	static void
	raidz_reflow_scratch_sync(void arg, dmu_tx_t tx)
	{
	vdev_raidz_expand_t *vre = arg;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	zio_t *pio;
	int error;

	spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
	vdev_t *raidvd = vdev_lookup_top(spa, vre->vre_vdev_id);
	int ashift = raidvd->vdev_ashift;
	- uint64_t write_size = P2ALIGN(VDEV_BOOT_SIZE, 1 << ashift);
	+ uint64_t write_size = P2ALIGN_TYPED(VDEV_BOOT_SIZE, 1 << ashift,
	+ uint64_t);
	uint64_t logical_size = write_size * raidvd->vdev_children;
	uint64_t read_size =
	P2ROUNDUP(DIV_ROUND_UP(logical_size, (raidvd->vdev_children - 1)),
	1 << ashift);

	/*
	* The scratch space must be large enough to get us to the point
	* that one row does not overlap itself when moved. This is checked
	* by vdev_raidz_attach_check().
	*/
	VERIFY3U(write_size, >=, raidvd->vdev_children << ashift);
	VERIFY3U(write_size, <=, VDEV_BOOT_SIZE);
	VERIFY3U(write_size, <=, read_size);

	zfs_locked_range_t *lr = zfs_rangelock_enter(&vre->vre_rangelock,
	0, logical_size, RL_WRITER);

	abd_t *abds = kmem_alloc(raidvd->vdev_children sizeof (abd_t *),
	KM_SLEEP);
	for (int i = 0; i < raidvd->vdev_children; i++) {
	abds[i] = abd_alloc_linear(read_size, B_FALSE);
	}

	raidz_expand_pause(RAIDZ_EXPAND_PAUSE_PRE_SCRATCH_1);

	/*
	* If we have already written the scratch area then we must read from
	* there, since new writes were redirected there while we were paused
	* or the original location may have been partially overwritten with
	* reflowed data.
	*/
	if (RRSS_GET_STATE(&spa->spa_ubsync) == RRSS_SCRATCH_VALID) {
	VERIFY3U(RRSS_GET_OFFSET(&spa->spa_ubsync), ==, logical_size);
	/*
	* Read from scratch space.
	*/
	pio = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL);
	for (int i = 0; i < raidvd->vdev_children; i++) {
	/*
	* Note: zio_vdev_child_io() adds VDEV_LABEL_START_SIZE
	* to the offset to calculate the physical offset to
	* write to. Passing in a negative offset makes us
	* access the scratch area.
	*/
	zio_nowait(zio_vdev_child_io(pio, NULL,
	raidvd->vdev_child[i],
	VDEV_BOOT_OFFSET - VDEV_LABEL_START_SIZE, abds[i],
	write_size, ZIO_TYPE_READ, ZIO_PRIORITY_ASYNC_READ,
	ZIO_FLAG_CANFAIL, raidz_scratch_child_done, pio));
	}
	error = zio_wait(pio);
	if (error != 0) {
	zfs_dbgmsg("reflow: error %d reading scratch location",
	error);
	goto io_error_exit;
	}
	goto overwrite;
	}

	/*
	* Read from original location.
	*/
	pio = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL);
	for (int i = 0; i < raidvd->vdev_children - 1; i++) {
	ASSERT0(vdev_is_dead(raidvd->vdev_child[i]));
	zio_nowait(zio_vdev_child_io(pio, NULL, raidvd->vdev_child[i],
	0, abds[i], read_size, ZIO_TYPE_READ,
	ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL,
	raidz_scratch_child_done, pio));
	}
	error = zio_wait(pio);
	if (error != 0) {
	zfs_dbgmsg("reflow: error %d reading original location", error);
	io_error_exit:
	for (int i = 0; i < raidvd->vdev_children; i++)
	abd_free(abds[i]);
	kmem_free(abds, raidvd->vdev_children * sizeof (abd_t *));
	zfs_rangelock_exit(lr);
	spa_config_exit(spa, SCL_STATE, FTAG);
	return;
	}

	raidz_expand_pause(RAIDZ_EXPAND_PAUSE_PRE_SCRATCH_2);

	/*
	* Reflow in memory.
	*/
	uint64_t logical_sectors = logical_size >> ashift;
	for (int i = raidvd->vdev_children - 1; i < logical_sectors; i++) {
	int oldchild = i % (raidvd->vdev_children - 1);
	uint64_t oldoff = (i / (raidvd->vdev_children - 1)) << ashift;

	int newchild = i % raidvd->vdev_children;
	uint64_t newoff = (i / raidvd->vdev_children) << ashift;

	/* a single sector should not be copying over itself */
	ASSERT(!(newchild == oldchild && newoff == oldoff));

	abd_copy_off(abds[newchild], abds[oldchild],
	newoff, oldoff, 1 << ashift);
	}

	/*
	* Verify that we filled in everything we intended to (write_size on
	* each child).
	*/
	VERIFY0(logical_sectors % raidvd->vdev_children);
	VERIFY3U((logical_sectors / raidvd->vdev_children) << ashift, ==,
	write_size);

	/*
	* Write to scratch location (boot area).
	*/
	pio = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL);
	for (int i = 0; i < raidvd->vdev_children; i++) {
	/*
	* Note: zio_vdev_child_io() adds VDEV_LABEL_START_SIZE to
	* the offset to calculate the physical offset to write to.
	* Passing in a negative offset lets us access the boot area.
	*/
	zio_nowait(zio_vdev_child_io(pio, NULL, raidvd->vdev_child[i],
	VDEV_BOOT_OFFSET - VDEV_LABEL_START_SIZE, abds[i],
	write_size, ZIO_TYPE_WRITE, ZIO_PRIORITY_ASYNC_WRITE,
	ZIO_FLAG_CANFAIL, raidz_scratch_child_done, pio));
	}
	error = zio_wait(pio);
	if (error != 0) {
	zfs_dbgmsg("reflow: error %d writing scratch location", error);
	goto io_error_exit;
	}
	pio = zio_root(spa, NULL, NULL, 0);
	zio_flush(pio, raidvd);
	zio_wait(pio);

	zfs_dbgmsg("reflow: wrote %llu bytes (logical) to scratch area",
	(long long)logical_size);

	raidz_expand_pause(RAIDZ_EXPAND_PAUSE_PRE_SCRATCH_3);

	/*
	* Update uberblock to indicate that scratch space is valid. This is
	* needed because after this point, the real location may be
	* overwritten. If we crash, we need to get the data from the
	* scratch space, rather than the real location.
	*
	* Note: ub_timestamp is bumped so that vdev_uberblock_compare()
	* will prefer this uberblock.
	*/
	RAIDZ_REFLOW_SET(&spa->spa_ubsync, RRSS_SCRATCH_VALID, logical_size);
	spa->spa_ubsync.ub_timestamp++;
	ASSERT0(vdev_uberblock_sync_list(&spa->spa_root_vdev, 1,
	&spa->spa_ubsync, ZIO_FLAG_CONFIG_WRITER));
	if (spa_multihost(spa))
	mmp_update_uberblock(spa, &spa->spa_ubsync);

	zfs_dbgmsg("reflow: uberblock updated "
	"(txg %llu, SCRATCH_VALID, size %llu, ts %llu)",
	(long long)spa->spa_ubsync.ub_txg,
	(long long)logical_size,
	(long long)spa->spa_ubsync.ub_timestamp);

	raidz_expand_pause(RAIDZ_EXPAND_PAUSE_SCRATCH_VALID);

	/*
	* Overwrite with reflow'ed data.
	*/
	overwrite:
	pio = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL);
	for (int i = 0; i < raidvd->vdev_children; i++) {
	zio_nowait(zio_vdev_child_io(pio, NULL, raidvd->vdev_child[i],
	0, abds[i], write_size, ZIO_TYPE_WRITE,
	ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_CANFAIL,
	raidz_scratch_child_done, pio));
	}
	error = zio_wait(pio);
	if (error != 0) {
	/*
	* When we exit early here and drop the range lock, new
	* writes will go into the scratch area so we'll need to
	* read from there when we return after pausing.
	*/
	zfs_dbgmsg("reflow: error %d writing real location", error);
	/*
	* Update the uberblock that is written when this txg completes.
	*/
	RAIDZ_REFLOW_SET(&spa->spa_uberblock, RRSS_SCRATCH_VALID,
	logical_size);
	goto io_error_exit;
	}
	pio = zio_root(spa, NULL, NULL, 0);
	zio_flush(pio, raidvd);
	zio_wait(pio);

	zfs_dbgmsg("reflow: overwrote %llu bytes (logical) to real location",
	(long long)logical_size);
	for (int i = 0; i < raidvd->vdev_children; i++)
	abd_free(abds[i]);
	kmem_free(abds, raidvd->vdev_children * sizeof (abd_t *));

	raidz_expand_pause(RAIDZ_EXPAND_PAUSE_SCRATCH_REFLOWED);

	/*
	* Update uberblock to indicate that the initial part has been
	* reflow'ed. This is needed because after this point (when we exit
	* the rangelock), we allow regular writes to this region, which will
	* be written to the new location only (because reflow_offset_next ==
	* reflow_offset_synced). If we crashed and re-copied from the
	* scratch space, we would lose the regular writes.
	*/
	RAIDZ_REFLOW_SET(&spa->spa_ubsync, RRSS_SCRATCH_INVALID_SYNCED,
	logical_size);
	spa->spa_ubsync.ub_timestamp++;
	ASSERT0(vdev_uberblock_sync_list(&spa->spa_root_vdev, 1,
	&spa->spa_ubsync, ZIO_FLAG_CONFIG_WRITER));
	if (spa_multihost(spa))
	mmp_update_uberblock(spa, &spa->spa_ubsync);

	zfs_dbgmsg("reflow: uberblock updated "
	"(txg %llu, SCRATCH_NOT_IN_USE, size %llu, ts %llu)",
	(long long)spa->spa_ubsync.ub_txg,
	(long long)logical_size,
	(long long)spa->spa_ubsync.ub_timestamp);

	raidz_expand_pause(RAIDZ_EXPAND_PAUSE_SCRATCH_POST_REFLOW_1);

	/*
	* Update progress.
	*/
	vre->vre_offset = logical_size;
	zfs_rangelock_exit(lr);
	spa_config_exit(spa, SCL_STATE, FTAG);

	int txgoff = dmu_tx_get_txg(tx) & TXG_MASK;
	vre->vre_offset_pertxg[txgoff] = vre->vre_offset;
	vre->vre_bytes_copied_pertxg[txgoff] = vre->vre_bytes_copied;
	/*
	* Note - raidz_reflow_sync() will update the uberblock state to
	* RRSS_SCRATCH_INVALID_SYNCED_REFLOW
	*/
	raidz_reflow_sync(spa, tx);

	raidz_expand_pause(RAIDZ_EXPAND_PAUSE_SCRATCH_POST_REFLOW_2);
	}

	/*
	* We crashed in the middle of raidz_reflow_scratch_sync(); complete its work
	* here. No other i/o can be in progress, so we don't need the vre_rangelock.
	*/
	void
	vdev_raidz_reflow_copy_scratch(spa_t *spa)
	{
	vdev_raidz_expand_t *vre = spa->spa_raidz_expand;
	uint64_t logical_size = RRSS_GET_OFFSET(&spa->spa_uberblock);
	ASSERT3U(RRSS_GET_STATE(&spa->spa_uberblock), ==, RRSS_SCRATCH_VALID);

	spa_config_enter(spa, SCL_STATE, FTAG, RW_READER);
	vdev_t *raidvd = vdev_lookup_top(spa, vre->vre_vdev_id);
	ASSERT0(logical_size % raidvd->vdev_children);
	uint64_t write_size = logical_size / raidvd->vdev_children;

	zio_t *pio;

	/*
	* Read from scratch space.
	*/
	abd_t *abds = kmem_alloc(raidvd->vdev_children sizeof (abd_t *),
	KM_SLEEP);
	for (int i = 0; i < raidvd->vdev_children; i++) {
	abds[i] = abd_alloc_linear(write_size, B_FALSE);
	}

	pio = zio_root(spa, NULL, NULL, 0);
	for (int i = 0; i < raidvd->vdev_children; i++) {
	/*
	* Note: zio_vdev_child_io() adds VDEV_LABEL_START_SIZE to
	* the offset to calculate the physical offset to write to.
	* Passing in a negative offset lets us access the boot area.
	*/
	zio_nowait(zio_vdev_child_io(pio, NULL, raidvd->vdev_child[i],
	VDEV_BOOT_OFFSET - VDEV_LABEL_START_SIZE, abds[i],
	write_size, ZIO_TYPE_READ,
	ZIO_PRIORITY_ASYNC_READ, 0,
	raidz_scratch_child_done, pio));
	}
	zio_wait(pio);

	/*
	* Overwrite real location with reflow'ed data.
	*/
	pio = zio_root(spa, NULL, NULL, 0);
	for (int i = 0; i < raidvd->vdev_children; i++) {
	zio_nowait(zio_vdev_child_io(pio, NULL, raidvd->vdev_child[i],
	0, abds[i], write_size, ZIO_TYPE_WRITE,
	ZIO_PRIORITY_ASYNC_WRITE, 0,
	raidz_scratch_child_done, pio));
	}
	zio_wait(pio);
	pio = zio_root(spa, NULL, NULL, 0);
	zio_flush(pio, raidvd);
	zio_wait(pio);

	zfs_dbgmsg("reflow recovery: overwrote %llu bytes (logical) "
	"to real location", (long long)logical_size);

	for (int i = 0; i < raidvd->vdev_children; i++)
	abd_free(abds[i]);
	kmem_free(abds, raidvd->vdev_children * sizeof (abd_t *));

	/*
	* Update uberblock.
	*/
	RAIDZ_REFLOW_SET(&spa->spa_ubsync,
	RRSS_SCRATCH_INVALID_SYNCED_ON_IMPORT, logical_size);
	spa->spa_ubsync.ub_timestamp++;
	VERIFY0(vdev_uberblock_sync_list(&spa->spa_root_vdev, 1,
	&spa->spa_ubsync, ZIO_FLAG_CONFIG_WRITER));
	if (spa_multihost(spa))
	mmp_update_uberblock(spa, &spa->spa_ubsync);

	zfs_dbgmsg("reflow recovery: uberblock updated "
	"(txg %llu, SCRATCH_NOT_IN_USE, size %llu, ts %llu)",
	(long long)spa->spa_ubsync.ub_txg,
	(long long)logical_size,
	(long long)spa->spa_ubsync.ub_timestamp);

	dmu_tx_t *tx = dmu_tx_create_assigned(spa->spa_dsl_pool,
	spa_first_txg(spa));
	int txgoff = dmu_tx_get_txg(tx) & TXG_MASK;
	vre->vre_offset = logical_size;
	vre->vre_offset_pertxg[txgoff] = vre->vre_offset;
	vre->vre_bytes_copied_pertxg[txgoff] = vre->vre_bytes_copied;
	/*
	* Note that raidz_reflow_sync() will update the uberblock once more
	*/
	raidz_reflow_sync(spa, tx);

	dmu_tx_commit(tx);

	spa_config_exit(spa, SCL_STATE, FTAG);
	}

	static boolean_t
	spa_raidz_expand_thread_check(void arg, zthr_t zthr)
	{
	(void) zthr;
	spa_t *spa = arg;

	return (spa->spa_raidz_expand != NULL &&
	!spa->spa_raidz_expand->vre_waiting_for_resilver);
	}

	/*
	* RAIDZ expansion background thread
	*
	* Can be called multiple times if the reflow is paused
	*/
	static void
	spa_raidz_expand_thread(void arg, zthr_t zthr)
	{
	spa_t *spa = arg;
	vdev_raidz_expand_t *vre = spa->spa_raidz_expand;

	if (RRSS_GET_STATE(&spa->spa_ubsync) == RRSS_SCRATCH_VALID)
	vre->vre_offset = 0;
	else
	vre->vre_offset = RRSS_GET_OFFSET(&spa->spa_ubsync);

	/* Reflow the begining portion using the scratch area */
	if (vre->vre_offset == 0) {
	VERIFY0(dsl_sync_task(spa_name(spa),
	NULL, raidz_reflow_scratch_sync,
	vre, 0, ZFS_SPACE_CHECK_NONE));

	/* if we encountered errors then pause */
	if (vre->vre_offset == 0) {
	mutex_enter(&vre->vre_lock);
	vre->vre_waiting_for_resilver = B_TRUE;
	mutex_exit(&vre->vre_lock);
	return;
	}
	}

	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	vdev_t *raidvd = vdev_lookup_top(spa, vre->vre_vdev_id);

	uint64_t guid = raidvd->vdev_guid;

	/* Iterate over all the remaining metaslabs */
	for (uint64_t i = vre->vre_offset >> raidvd->vdev_ms_shift;
	i < raidvd->vdev_ms_count &&
	!zthr_iscancelled(zthr) &&
	vre->vre_failed_offset == UINT64_MAX; i++) {
	metaslab_t *msp = raidvd->vdev_ms[i];

	metaslab_disable(msp);
	mutex_enter(&msp->ms_lock);

	/*
	* The metaslab may be newly created (for the expanded
	* space), in which case its trees won't exist yet,
	* so we need to bail out early.
	*/
	if (msp->ms_new) {
	mutex_exit(&msp->ms_lock);
	metaslab_enable(msp, B_FALSE, B_FALSE);
	continue;
	}

	VERIFY0(metaslab_load(msp));

	/*
	* We want to copy everything except the free (allocatable)
	* space. Note that there may be a little bit more free
	* space (e.g. in ms_defer), and it's fine to copy that too.
	*/
	range_tree_t *rt = range_tree_create(NULL, RANGE_SEG64,
	NULL, 0, 0);
	range_tree_add(rt, msp->ms_start, msp->ms_size);
	range_tree_walk(msp->ms_allocatable, range_tree_remove, rt);
	mutex_exit(&msp->ms_lock);

	/*
	* Force the last sector of each metaslab to be copied. This
	* ensures that we advance the on-disk progress to the end of
	* this metaslab while the metaslab is disabled. Otherwise, we
	* could move past this metaslab without advancing the on-disk
	* progress, and then an allocation to this metaslab would not
	* be copied.
	*/
	int sectorsz = 1 << raidvd->vdev_ashift;
	uint64_t ms_last_offset = msp->ms_start +
	msp->ms_size - sectorsz;
	if (!range_tree_contains(rt, ms_last_offset, sectorsz)) {
	range_tree_add(rt, ms_last_offset, sectorsz);
	}

	/*
	* When we are resuming from a paused expansion (i.e.
	* when importing a pool with a expansion in progress),
	* discard any state that we have already processed.
	*/
	range_tree_clear(rt, 0, vre->vre_offset);

	while (!zthr_iscancelled(zthr) &&
	!range_tree_is_empty(rt) &&
	vre->vre_failed_offset == UINT64_MAX) {

	/*
	* We need to periodically drop the config lock so that
	* writers can get in. Additionally, we can't wait
	* for a txg to sync while holding a config lock
	* (since a waiting writer could cause a 3-way deadlock
	* with the sync thread, which also gets a config
	* lock for reader). So we can't hold the config lock
	* while calling dmu_tx_assign().
	*/
	spa_config_exit(spa, SCL_CONFIG, FTAG);

	/*
	* If requested, pause the reflow when the amount
	* specified by raidz_expand_max_reflow_bytes is reached
	*
	* This pause is only used during testing or debugging.
	*/
	while (raidz_expand_max_reflow_bytes != 0 &&
	raidz_expand_max_reflow_bytes <=
	vre->vre_bytes_copied && !zthr_iscancelled(zthr)) {
	delay(hz);
	}

	mutex_enter(&vre->vre_lock);
	while (vre->vre_outstanding_bytes >
	raidz_expand_max_copy_bytes) {
	cv_wait(&vre->vre_cv, &vre->vre_lock);
	}
	mutex_exit(&vre->vre_lock);

	dmu_tx_t *tx =
	dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);

	VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
	uint64_t txg = dmu_tx_get_txg(tx);

	/*
	* Reacquire the vdev_config lock. Theoretically, the
	* vdev_t that we're expanding may have changed.
	*/
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	raidvd = vdev_lookup_top(spa, vre->vre_vdev_id);

	boolean_t needsync =
	raidz_reflow_impl(raidvd, vre, rt, tx);

	dmu_tx_commit(tx);

	if (needsync) {
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	txg_wait_synced(spa->spa_dsl_pool, txg);
	spa_config_enter(spa, SCL_CONFIG, FTAG,
	RW_READER);
	}
	}

	spa_config_exit(spa, SCL_CONFIG, FTAG);

	metaslab_enable(msp, B_FALSE, B_FALSE);
	range_tree_vacate(rt, NULL, NULL);
	range_tree_destroy(rt);

	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	raidvd = vdev_lookup_top(spa, vre->vre_vdev_id);
	}

	spa_config_exit(spa, SCL_CONFIG, FTAG);

	/*
	* The txg_wait_synced() here ensures that all reflow zio's have
	* completed, and vre_failed_offset has been set if necessary. It
	* also ensures that the progress of the last raidz_reflow_sync() is
	* written to disk before raidz_reflow_complete_sync() changes the
	* in-memory vre_state. vdev_raidz_io_start() uses vre_state to
	* determine if a reflow is in progress, in which case we may need to
	* write to both old and new locations. Therefore we can only change
	* vre_state once this is not necessary, which is once the on-disk
	* progress (in spa_ubsync) has been set past any possible writes (to
	* the end of the last metaslab).
	*/
	txg_wait_synced(spa->spa_dsl_pool, 0);

	if (!zthr_iscancelled(zthr) &&
	vre->vre_offset == raidvd->vdev_ms_count << raidvd->vdev_ms_shift) {
	/*
	* We are not being canceled or paused, so the reflow must be
	* complete. In that case also mark it as completed on disk.
	*/
	ASSERT3U(vre->vre_failed_offset, ==, UINT64_MAX);
	VERIFY0(dsl_sync_task(spa_name(spa), NULL,
	raidz_reflow_complete_sync, spa,
	0, ZFS_SPACE_CHECK_NONE));
	(void) vdev_online(spa, guid, ZFS_ONLINE_EXPAND, NULL);
	} else {
	/*
	* Wait for all copy zio's to complete and for all the
	* raidz_reflow_sync() synctasks to be run.
	*/
	spa_history_log_internal(spa, "reflow pause",
	NULL, "offset=%llu failed_offset=%lld",
	(long long)vre->vre_offset,
	(long long)vre->vre_failed_offset);
	mutex_enter(&vre->vre_lock);
	if (vre->vre_failed_offset != UINT64_MAX) {
	/*
	* Reset progress so that we will retry everything
	* after the point that something failed.
	*/
	vre->vre_offset = vre->vre_failed_offset;
	vre->vre_failed_offset = UINT64_MAX;
	vre->vre_waiting_for_resilver = B_TRUE;
	}
	mutex_exit(&vre->vre_lock);
	}
	}

	void
	spa_start_raidz_expansion_thread(spa_t *spa)
	{
	ASSERT3P(spa->spa_raidz_expand_zthr, ==, NULL);
	spa->spa_raidz_expand_zthr = zthr_create("raidz_expand",
	spa_raidz_expand_thread_check, spa_raidz_expand_thread,
	spa, defclsyspri);
	}

	void
	raidz_dtl_reassessed(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;
	if (spa->spa_raidz_expand != NULL) {
	vdev_raidz_expand_t *vre = spa->spa_raidz_expand;
	/*
	* we get called often from vdev_dtl_reassess() so make
	* sure it's our vdev and any replacing is complete
	*/
	if (vd->vdev_top->vdev_id == vre->vre_vdev_id &&
	!vdev_raidz_expand_child_replacing(vd->vdev_top)) {
	mutex_enter(&vre->vre_lock);
	if (vre->vre_waiting_for_resilver) {
	vdev_dbgmsg(vd, "DTL reassessed, "
	"continuing raidz expansion");
	vre->vre_waiting_for_resilver = B_FALSE;
	zthr_wakeup(spa->spa_raidz_expand_zthr);
	}
	mutex_exit(&vre->vre_lock);
	}
	}
	}

	int
	vdev_raidz_attach_check(vdev_t *new_child)
	{
	vdev_t *raidvd = new_child->vdev_parent;
	uint64_t new_children = raidvd->vdev_children;

	/*
	* We use the "boot" space as scratch space to handle overwriting the
	* initial part of the vdev. If it is too small, then this expansion
	* is not allowed. This would be very unusual (e.g. ashift > 13 and
	* >200 children).
	*/
	if (new_children << raidvd->vdev_ashift > VDEV_BOOT_SIZE) {
	return (EINVAL);
	}
	return (0);
	}

	void
	vdev_raidz_attach_sync(void arg, dmu_tx_t tx)
	{
	vdev_t *new_child = arg;
	spa_t *spa = new_child->vdev_spa;
	vdev_t *raidvd = new_child->vdev_parent;
	vdev_raidz_t *vdrz = raidvd->vdev_tsd;
	ASSERT3P(raidvd->vdev_ops, ==, &vdev_raidz_ops);
	ASSERT3P(raidvd->vdev_top, ==, raidvd);
	ASSERT3U(raidvd->vdev_children, >, vdrz->vd_original_width);
	ASSERT3U(raidvd->vdev_children, ==, vdrz->vd_physical_width + 1);
	ASSERT3P(raidvd->vdev_child[raidvd->vdev_children - 1], ==,
	new_child);

	spa_feature_incr(spa, SPA_FEATURE_RAIDZ_EXPANSION, tx);

	vdrz->vd_physical_width++;

	VERIFY0(spa->spa_uberblock.ub_raidz_reflow_info);
	vdrz->vn_vre.vre_vdev_id = raidvd->vdev_id;
	vdrz->vn_vre.vre_offset = 0;
	vdrz->vn_vre.vre_failed_offset = UINT64_MAX;
	spa->spa_raidz_expand = &vdrz->vn_vre;
	zthr_wakeup(spa->spa_raidz_expand_zthr);

	/*
	* Dirty the config so that ZPOOL_CONFIG_RAIDZ_EXPANDING will get
	* written to the config.
	*/
	vdev_config_dirty(raidvd);

	vdrz->vn_vre.vre_start_time = gethrestime_sec();
	vdrz->vn_vre.vre_end_time = 0;
	vdrz->vn_vre.vre_state = DSS_SCANNING;
	vdrz->vn_vre.vre_bytes_copied = 0;

	uint64_t state = vdrz->vn_vre.vre_state;
	VERIFY0(zap_update(spa->spa_meta_objset,
	raidvd->vdev_top_zap, VDEV_TOP_ZAP_RAIDZ_EXPAND_STATE,
	sizeof (state), 1, &state, tx));

	uint64_t start_time = vdrz->vn_vre.vre_start_time;
	VERIFY0(zap_update(spa->spa_meta_objset,
	raidvd->vdev_top_zap, VDEV_TOP_ZAP_RAIDZ_EXPAND_START_TIME,
	sizeof (start_time), 1, &start_time, tx));

	(void) zap_remove(spa->spa_meta_objset,
	raidvd->vdev_top_zap, VDEV_TOP_ZAP_RAIDZ_EXPAND_END_TIME, tx);
	(void) zap_remove(spa->spa_meta_objset,
	raidvd->vdev_top_zap, VDEV_TOP_ZAP_RAIDZ_EXPAND_BYTES_COPIED, tx);

	spa_history_log_internal(spa, "raidz vdev expansion started", tx,
	"%s vdev %llu new width %llu", spa_name(spa),
	(unsigned long long)raidvd->vdev_id,
	(unsigned long long)raidvd->vdev_children);
	}

	int
	vdev_raidz_load(vdev_t *vd)
	{
	vdev_raidz_t *vdrz = vd->vdev_tsd;
	int err;

	uint64_t state = DSS_NONE;
	uint64_t start_time = 0;
	uint64_t end_time = 0;
	uint64_t bytes_copied = 0;

	if (vd->vdev_top_zap != 0) {
	err = zap_lookup(vd->vdev_spa->spa_meta_objset,
	vd->vdev_top_zap, VDEV_TOP_ZAP_RAIDZ_EXPAND_STATE,
	sizeof (state), 1, &state);
	if (err != 0 && err != ENOENT)
	return (err);

	err = zap_lookup(vd->vdev_spa->spa_meta_objset,
	vd->vdev_top_zap, VDEV_TOP_ZAP_RAIDZ_EXPAND_START_TIME,
	sizeof (start_time), 1, &start_time);
	if (err != 0 && err != ENOENT)
	return (err);

	err = zap_lookup(vd->vdev_spa->spa_meta_objset,
	vd->vdev_top_zap, VDEV_TOP_ZAP_RAIDZ_EXPAND_END_TIME,
	sizeof (end_time), 1, &end_time);
	if (err != 0 && err != ENOENT)
	return (err);

	err = zap_lookup(vd->vdev_spa->spa_meta_objset,
	vd->vdev_top_zap, VDEV_TOP_ZAP_RAIDZ_EXPAND_BYTES_COPIED,
	sizeof (bytes_copied), 1, &bytes_copied);
	if (err != 0 && err != ENOENT)
	return (err);
	}

	/*
	* If we are in the middle of expansion, vre_state should have
	* already been set by vdev_raidz_init().
	*/
	EQUIV(vdrz->vn_vre.vre_state == DSS_SCANNING, state == DSS_SCANNING);
	vdrz->vn_vre.vre_state = (dsl_scan_state_t)state;
	vdrz->vn_vre.vre_start_time = start_time;
	vdrz->vn_vre.vre_end_time = end_time;
	vdrz->vn_vre.vre_bytes_copied = bytes_copied;

	return (0);
	}

	int
	spa_raidz_expand_get_stats(spa_t spa, pool_raidz_expand_stat_t pres)
	{
	vdev_raidz_expand_t *vre = spa->spa_raidz_expand;

	if (vre == NULL) {
	/* no removal in progress; find most recent completed */
	for (int c = 0; c < spa->spa_root_vdev->vdev_children; c++) {
	vdev_t *vd = spa->spa_root_vdev->vdev_child[c];
	if (vd->vdev_ops == &vdev_raidz_ops) {
	vdev_raidz_t *vdrz = vd->vdev_tsd;

	if (vdrz->vn_vre.vre_end_time != 0 &&
	(vre == NULL \|\|
	vdrz->vn_vre.vre_end_time >
	vre->vre_end_time)) {
	vre = &vdrz->vn_vre;
	}
	}
	}
	}

	if (vre == NULL) {
	return (SET_ERROR(ENOENT));
	}

	pres->pres_state = vre->vre_state;
	pres->pres_expanding_vdev = vre->vre_vdev_id;

	vdev_t *vd = vdev_lookup_top(spa, vre->vre_vdev_id);
	pres->pres_to_reflow = vd->vdev_stat.vs_alloc;

	mutex_enter(&vre->vre_lock);
	pres->pres_reflowed = vre->vre_bytes_copied;
	for (int i = 0; i < TXG_SIZE; i++)
	pres->pres_reflowed += vre->vre_bytes_copied_pertxg[i];
	mutex_exit(&vre->vre_lock);

	pres->pres_start_time = vre->vre_start_time;
	pres->pres_end_time = vre->vre_end_time;
	pres->pres_waiting_for_resilver = vre->vre_waiting_for_resilver;

	return (0);
	}

	/*
	* Initialize private RAIDZ specific fields from the nvlist.
	*/
	static int
	vdev_raidz_init(spa_t spa, nvlist_t nv, void **tsd)
	{
	uint_t children;
	nvlist_t **child;
	int error = nvlist_lookup_nvlist_array(nv,
	ZPOOL_CONFIG_CHILDREN, &child, &children);
	if (error != 0)
	return (SET_ERROR(EINVAL));

	uint64_t nparity;
	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY, &nparity) == 0) {
	if (nparity == 0 \|\| nparity > VDEV_RAIDZ_MAXPARITY)
	return (SET_ERROR(EINVAL));

	/*
	* Previous versions could only support 1 or 2 parity
	* device.
	*/
	if (nparity > 1 && spa_version(spa) < SPA_VERSION_RAIDZ2)
	return (SET_ERROR(EINVAL));
	else if (nparity > 2 && spa_version(spa) < SPA_VERSION_RAIDZ3)
	return (SET_ERROR(EINVAL));
	} else {
	/*
	* We require the parity to be specified for SPAs that
	* support multiple parity levels.
	*/
	if (spa_version(spa) >= SPA_VERSION_RAIDZ2)
	return (SET_ERROR(EINVAL));

	/*
	* Otherwise, we default to 1 parity device for RAID-Z.
	*/
	nparity = 1;
	}

	vdev_raidz_t vdrz = kmem_zalloc(sizeof (vdrz), KM_SLEEP);
	vdrz->vn_vre.vre_vdev_id = -1;
	vdrz->vn_vre.vre_offset = UINT64_MAX;
	vdrz->vn_vre.vre_failed_offset = UINT64_MAX;
	mutex_init(&vdrz->vn_vre.vre_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&vdrz->vn_vre.vre_cv, NULL, CV_DEFAULT, NULL);
	zfs_rangelock_init(&vdrz->vn_vre.vre_rangelock, NULL, NULL);
	mutex_init(&vdrz->vd_expand_lock, NULL, MUTEX_DEFAULT, NULL);
	avl_create(&vdrz->vd_expand_txgs, vdev_raidz_reflow_compare,
	sizeof (reflow_node_t), offsetof(reflow_node_t, re_link));

	vdrz->vd_physical_width = children;
	vdrz->vd_nparity = nparity;

	/* note, the ID does not exist when creating a pool */
	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
	&vdrz->vn_vre.vre_vdev_id);

	boolean_t reflow_in_progress =
	nvlist_exists(nv, ZPOOL_CONFIG_RAIDZ_EXPANDING);
	if (reflow_in_progress) {
	spa->spa_raidz_expand = &vdrz->vn_vre;
	vdrz->vn_vre.vre_state = DSS_SCANNING;
	}

	vdrz->vd_original_width = children;
	uint64_t *txgs;
	unsigned int txgs_size = 0;
	error = nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_RAIDZ_EXPAND_TXGS,
	&txgs, &txgs_size);
	if (error == 0) {
	for (int i = 0; i < txgs_size; i++) {
	reflow_node_t re = kmem_zalloc(sizeof (re), KM_SLEEP);
	re->re_txg = txgs[txgs_size - i - 1];
	re->re_logical_width = vdrz->vd_physical_width - i;

	if (reflow_in_progress)
	re->re_logical_width--;

	avl_add(&vdrz->vd_expand_txgs, re);
	}

	vdrz->vd_original_width = vdrz->vd_physical_width - txgs_size;
	}
	if (reflow_in_progress) {
	vdrz->vd_original_width--;
	zfs_dbgmsg("reflow_in_progress, %u wide, %d prior expansions",
	children, txgs_size);
	}

	*tsd = vdrz;

	return (0);
	}

	static void
	vdev_raidz_fini(vdev_t *vd)
	{
	vdev_raidz_t *vdrz = vd->vdev_tsd;
	if (vd->vdev_spa->spa_raidz_expand == &vdrz->vn_vre)
	vd->vdev_spa->spa_raidz_expand = NULL;
	reflow_node_t *re;
	void *cookie = NULL;
	avl_tree_t *tree = &vdrz->vd_expand_txgs;
	while ((re = avl_destroy_nodes(tree, &cookie)) != NULL)
	kmem_free(re, sizeof (*re));
	avl_destroy(&vdrz->vd_expand_txgs);
	mutex_destroy(&vdrz->vd_expand_lock);
	mutex_destroy(&vdrz->vn_vre.vre_lock);
	cv_destroy(&vdrz->vn_vre.vre_cv);
	zfs_rangelock_fini(&vdrz->vn_vre.vre_rangelock);
	kmem_free(vdrz, sizeof (*vdrz));
	}

	/*
	* Add RAIDZ specific fields to the config nvlist.
	*/
	static void
	vdev_raidz_config_generate(vdev_t vd, nvlist_t nv)
	{
	ASSERT3P(vd->vdev_ops, ==, &vdev_raidz_ops);
	vdev_raidz_t *vdrz = vd->vdev_tsd;

	/*
	* Make sure someone hasn't managed to sneak a fancy new vdev
	* into a crufty old storage pool.
	*/
	ASSERT(vdrz->vd_nparity == 1 \|\|
	(vdrz->vd_nparity <= 2 &&
	spa_version(vd->vdev_spa) >= SPA_VERSION_RAIDZ2) \|\|
	(vdrz->vd_nparity <= 3 &&
	spa_version(vd->vdev_spa) >= SPA_VERSION_RAIDZ3));

	/*
	* Note that we'll add these even on storage pools where they
	* aren't strictly required -- older software will just ignore
	* it.
	*/
	fnvlist_add_uint64(nv, ZPOOL_CONFIG_NPARITY, vdrz->vd_nparity);

	if (vdrz->vn_vre.vre_state == DSS_SCANNING) {
	fnvlist_add_boolean(nv, ZPOOL_CONFIG_RAIDZ_EXPANDING);
	}

	mutex_enter(&vdrz->vd_expand_lock);
	if (!avl_is_empty(&vdrz->vd_expand_txgs)) {
	uint64_t count = avl_numnodes(&vdrz->vd_expand_txgs);
	uint64_t txgs = kmem_alloc(sizeof (uint64_t) count,
	KM_SLEEP);
	uint64_t i = 0;

	for (reflow_node_t *re = avl_first(&vdrz->vd_expand_txgs);
	re != NULL; re = AVL_NEXT(&vdrz->vd_expand_txgs, re)) {
	txgs[i++] = re->re_txg;
	}

	fnvlist_add_uint64_array(nv, ZPOOL_CONFIG_RAIDZ_EXPAND_TXGS,
	txgs, count);

	kmem_free(txgs, sizeof (uint64_t) * count);
	}
	mutex_exit(&vdrz->vd_expand_lock);
	}

	static uint64_t
	vdev_raidz_nparity(vdev_t *vd)
	{
	vdev_raidz_t *vdrz = vd->vdev_tsd;
	return (vdrz->vd_nparity);
	}

	static uint64_t
	vdev_raidz_ndisks(vdev_t *vd)
	{
	return (vd->vdev_children);
	}

	vdev_ops_t vdev_raidz_ops = {
	.vdev_op_init = vdev_raidz_init,
	.vdev_op_fini = vdev_raidz_fini,
	.vdev_op_open = vdev_raidz_open,
	.vdev_op_close = vdev_raidz_close,
	.vdev_op_asize = vdev_raidz_asize,
	.vdev_op_min_asize = vdev_raidz_min_asize,
	.vdev_op_min_alloc = NULL,
	.vdev_op_io_start = vdev_raidz_io_start,
	.vdev_op_io_done = vdev_raidz_io_done,
	.vdev_op_state_change = vdev_raidz_state_change,
	.vdev_op_need_resilver = vdev_raidz_need_resilver,
	.vdev_op_hold = NULL,
	.vdev_op_rele = NULL,
	.vdev_op_remap = NULL,
	.vdev_op_xlate = vdev_raidz_xlate,
	.vdev_op_rebuild_asize = NULL,
	.vdev_op_metaslab_init = NULL,
	.vdev_op_config_generate = vdev_raidz_config_generate,
	.vdev_op_nparity = vdev_raidz_nparity,
	.vdev_op_ndisks = vdev_raidz_ndisks,
	.vdev_op_type = VDEV_TYPE_RAIDZ, /* name of this vdev type */
	.vdev_op_leaf = B_FALSE /* not a leaf vdev */
	};

	/* BEGIN CSTYLED */
	ZFS_MODULE_PARAM(zfs_vdev, raidz_, expand_max_reflow_bytes, ULONG, ZMOD_RW,
	"For testing, pause RAIDZ expansion after reflowing this many bytes");
	ZFS_MODULE_PARAM(zfs_vdev, raidz_, expand_max_copy_bytes, ULONG, ZMOD_RW,
	"Max amount of concurrent i/o for RAIDZ expansion");
	ZFS_MODULE_PARAM(zfs_vdev, raidz_, io_aggregate_rows, ULONG, ZMOD_RW,
	"For expanded RAIDZ, aggregate reads that have more rows than this");
	ZFS_MODULE_PARAM(zfs, zfs_, scrub_after_expand, INT, ZMOD_RW,
	"For expanded RAIDZ, automatically start a pool scrub when expansion "
	"completes");
	/* END CSTYLED */
	diff --git a/sys/contrib/openzfs/module/zfs/vdev_rebuild.c b/sys/contrib/openzfs/module/zfs/vdev_rebuild.c
	index 00ebd4c9fca4..8a8b02cab5c6 100644
	--- a/sys/contrib/openzfs/module/zfs/vdev_rebuild.c
	+++ b/sys/contrib/openzfs/module/zfs/vdev_rebuild.c
	@@ -1,1175 +1,1176 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	*
	* Copyright (c) 2018, Intel Corporation.
	* Copyright (c) 2020 by Lawrence Livermore National Security, LLC.
	* Copyright (c) 2022 Hewlett Packard Enterprise Development LP.
	* Copyright (c) 2024 by Delphix. All rights reserved.
	*/

	#include <sys/vdev_impl.h>
	#include <sys/vdev_draid.h>
	#include <sys/dsl_scan.h>
	#include <sys/spa_impl.h>
	#include <sys/metaslab_impl.h>
	#include <sys/vdev_rebuild.h>
	#include <sys/zio.h>
	#include <sys/dmu_tx.h>
	#include <sys/arc.h>
	#include <sys/arc_impl.h>
	#include <sys/zap.h>

	/*
	* This file contains the sequential reconstruction implementation for
	* resilvering. This form of resilvering is internally referred to as device
	* rebuild to avoid conflating it with the traditional healing reconstruction
	* performed by the dsl scan code.
	*
	* When replacing a device, or scrubbing the pool, ZFS has historically used
	* a process called resilvering which is a form of healing reconstruction.
	* This approach has the advantage that as blocks are read from disk their
	* checksums can be immediately verified and the data repaired. Unfortunately,
	* it also results in a random IO pattern to the disk even when extra care
	* is taken to sequentialize the IO as much as possible. This substantially
	* increases the time required to resilver the pool and restore redundancy.
	*
	* For mirrored devices it's possible to implement an alternate sequential
	* reconstruction strategy when resilvering. Sequential reconstruction
	* behaves like a traditional RAID rebuild and reconstructs a device in LBA
	* order without verifying the checksum. After this phase completes a second
	* scrub phase is started to verify all of the checksums. This two phase
	* process will take longer than the healing reconstruction described above.
	* However, it has that advantage that after the reconstruction first phase
	* completes redundancy has been restored. At this point the pool can incur
	* another device failure without risking data loss.
	*
	* There are a few noteworthy limitations and other advantages of resilvering
	* using sequential reconstruction vs healing reconstruction.
	*
	* Limitations:
	*
	* - Sequential reconstruction is not possible on RAIDZ due to its
	* variable stripe width. Note dRAID uses a fixed stripe width which
	* avoids this issue, but comes at the expense of some usable capacity.
	*
	* - Block checksums are not verified during sequential reconstruction.
	* Similar to traditional RAID the parity/mirror data is reconstructed
	* but cannot be immediately double checked. For this reason when the
	* last active resilver completes the pool is automatically scrubbed
	* by default.
	*
	* - Deferred resilvers using sequential reconstruction are not currently
	* supported. When adding another vdev to an active top-level resilver
	* it must be restarted.
	*
	* Advantages:
	*
	* - Sequential reconstruction is performed in LBA order which may be faster
	* than healing reconstruction particularly when using HDDs (or
	* especially with SMR devices). Only allocated capacity is resilvered.
	*
	* - Sequential reconstruction is not constrained by ZFS block boundaries.
	* This allows it to issue larger IOs to disk which span multiple blocks
	* allowing all of these logical blocks to be repaired with a single IO.
	*
	* - Unlike a healing resilver or scrub which are pool wide operations,
	* sequential reconstruction is handled by the top-level vdevs. This
	* allows for it to be started or canceled on a top-level vdev without
	* impacting any other top-level vdevs in the pool.
	*
	* - Data only referenced by a pool checkpoint will be repaired because
	* that space is reflected in the space maps. This differs for a
	* healing resilver or scrub which will not repair that data.
	*/


	/*
	* Size of rebuild reads; defaults to 1MiB per data disk and is capped at
	* SPA_MAXBLOCKSIZE.
	*/
	static uint64_t zfs_rebuild_max_segment = 1024 * 1024;

	/*
	* Maximum number of parallelly executed bytes per leaf vdev caused by a
	* sequential resilver. We attempt to strike a balance here between keeping
	* the vdev queues full of I/Os at all times and not overflowing the queues
	* to cause long latency, which would cause long txg sync times.
	*
	* A large default value can be safely used here because the default target
	* segment size is also large (zfs_rebuild_max_segment=1M). This helps keep
	* the queue depth short.
	*
	* 64MB was observed to deliver the best performance and set as the default.
	* Testing was performed with a 106-drive dRAID HDD pool (draid2:11d:106c)
	* and a rebuild rate of 1.2GB/s was measured to the distribute spare.
	* Smaller values were unable to fully saturate the available pool I/O.
	*/
	static uint64_t zfs_rebuild_vdev_limit = 64 << 20;

	/*
	* Automatically start a pool scrub when the last active sequential resilver
	* completes in order to verify the checksums of all blocks which have been
	* resilvered. This option is enabled by default and is strongly recommended.
	*/
	static int zfs_rebuild_scrub_enabled = 1;

	/*
	* For vdev_rebuild_initiate_sync() and vdev_rebuild_reset_sync().
	*/
	static __attribute__((noreturn)) void vdev_rebuild_thread(void *arg);
	static void vdev_rebuild_reset_sync(void arg, dmu_tx_t tx);

	/*
	* Clear the per-vdev rebuild bytes value for a vdev tree.
	*/
	static void
	clear_rebuild_bytes(vdev_t *vd)
	{
	vdev_stat_t *vs = &vd->vdev_stat;

	for (uint64_t i = 0; i < vd->vdev_children; i++)
	clear_rebuild_bytes(vd->vdev_child[i]);

	mutex_enter(&vd->vdev_stat_lock);
	vs->vs_rebuild_processed = 0;
	mutex_exit(&vd->vdev_stat_lock);
	}

	/*
	* Determines whether a vdev_rebuild_thread() should be stopped.
	*/
	static boolean_t
	vdev_rebuild_should_stop(vdev_t *vd)
	{
	return (!vdev_writeable(vd) \|\| vd->vdev_removing \|\|
	vd->vdev_rebuild_exit_wanted \|\|
	vd->vdev_rebuild_cancel_wanted \|\|
	vd->vdev_rebuild_reset_wanted);
	}

	/*
	* Determine if the rebuild should be canceled. This may happen when all
	* vdevs with MISSING DTLs are detached.
	*/
	static boolean_t
	vdev_rebuild_should_cancel(vdev_t *vd)
	{
	vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
	vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;

	if (!vdev_resilver_needed(vd, &vrp->vrp_min_txg, &vrp->vrp_max_txg))
	return (B_TRUE);

	return (B_FALSE);
	}

	/*
	* The sync task for updating the on-disk state of a rebuild. This is
	* scheduled by vdev_rebuild_range().
	*/
	static void
	vdev_rebuild_update_sync(void arg, dmu_tx_t tx)
	{
	int vdev_id = (uintptr_t)arg;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	vdev_t *vd = vdev_lookup_top(spa, vdev_id);
	vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
	vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
	uint64_t txg = dmu_tx_get_txg(tx);

	mutex_enter(&vd->vdev_rebuild_lock);

	if (vr->vr_scan_offset[txg & TXG_MASK] > 0) {
	vrp->vrp_last_offset = vr->vr_scan_offset[txg & TXG_MASK];
	vr->vr_scan_offset[txg & TXG_MASK] = 0;
	}

	vrp->vrp_scan_time_ms = vr->vr_prev_scan_time_ms +
	NSEC2MSEC(gethrtime() - vr->vr_pass_start_time);

	VERIFY0(zap_update(vd->vdev_spa->spa_meta_objset, vd->vdev_top_zap,
	VDEV_TOP_ZAP_VDEV_REBUILD_PHYS, sizeof (uint64_t),
	REBUILD_PHYS_ENTRIES, vrp, tx));

	mutex_exit(&vd->vdev_rebuild_lock);
	}

	/*
	* Initialize the on-disk state for a new rebuild, start the rebuild thread.
	*/
	static void
	vdev_rebuild_initiate_sync(void arg, dmu_tx_t tx)
	{
	int vdev_id = (uintptr_t)arg;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	vdev_t *vd = vdev_lookup_top(spa, vdev_id);
	vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
	vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;

	ASSERT(vd->vdev_rebuilding);

	spa_feature_incr(vd->vdev_spa, SPA_FEATURE_DEVICE_REBUILD, tx);

	mutex_enter(&vd->vdev_rebuild_lock);
	memset(vrp, 0, sizeof (uint64_t) * REBUILD_PHYS_ENTRIES);
	vrp->vrp_rebuild_state = VDEV_REBUILD_ACTIVE;
	vrp->vrp_min_txg = 0;
	vrp->vrp_max_txg = dmu_tx_get_txg(tx);
	vrp->vrp_start_time = gethrestime_sec();
	vrp->vrp_scan_time_ms = 0;
	vr->vr_prev_scan_time_ms = 0;

	/*
	* Rebuilds are currently only used when replacing a device, in which
	* case there must be DTL_MISSING entries. In the future, we could
	* allow rebuilds to be used in a way similar to a scrub. This would
	* be useful because it would allow us to rebuild the space used by
	* pool checkpoints.
	*/
	VERIFY(vdev_resilver_needed(vd, &vrp->vrp_min_txg, &vrp->vrp_max_txg));

	VERIFY0(zap_update(vd->vdev_spa->spa_meta_objset, vd->vdev_top_zap,
	VDEV_TOP_ZAP_VDEV_REBUILD_PHYS, sizeof (uint64_t),
	REBUILD_PHYS_ENTRIES, vrp, tx));

	spa_history_log_internal(spa, "rebuild", tx,
	"vdev_id=%llu vdev_guid=%llu started",
	(u_longlong_t)vd->vdev_id, (u_longlong_t)vd->vdev_guid);

	ASSERT3P(vd->vdev_rebuild_thread, ==, NULL);
	vd->vdev_rebuild_thread = thread_create(NULL, 0,
	vdev_rebuild_thread, vd, 0, &p0, TS_RUN, maxclsyspri);

	mutex_exit(&vd->vdev_rebuild_lock);
	}

	static void
	vdev_rebuild_log_notify(spa_t spa, vdev_t vd, const char *name)
	{
	nvlist_t *aux = fnvlist_alloc();

	fnvlist_add_string(aux, ZFS_EV_RESILVER_TYPE, "sequential");
	spa_event_notify(spa, vd, aux, name);
	nvlist_free(aux);
	}

	/*
	* Called to request that a new rebuild be started. The feature will remain
	* active for the duration of the rebuild, then revert to the enabled state.
	*/
	static void
	vdev_rebuild_initiate(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;

	ASSERT(vd->vdev_top == vd);
	ASSERT(MUTEX_HELD(&vd->vdev_rebuild_lock));
	ASSERT(!vd->vdev_rebuilding);

	dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
	VERIFY0(dmu_tx_assign(tx, TXG_WAIT));

	vd->vdev_rebuilding = B_TRUE;

	dsl_sync_task_nowait(spa_get_dsl(spa), vdev_rebuild_initiate_sync,
	(void *)(uintptr_t)vd->vdev_id, tx);
	dmu_tx_commit(tx);

	vdev_rebuild_log_notify(spa, vd, ESC_ZFS_RESILVER_START);
	}

	/*
	* Update the on-disk state to completed when a rebuild finishes.
	*/
	static void
	vdev_rebuild_complete_sync(void arg, dmu_tx_t tx)
	{
	int vdev_id = (uintptr_t)arg;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	vdev_t *vd = vdev_lookup_top(spa, vdev_id);
	vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
	vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;

	mutex_enter(&vd->vdev_rebuild_lock);

	/*
	* Handle a second device failure if it occurs after all rebuild I/O
	* has completed but before this sync task has been executed.
	*/
	if (vd->vdev_rebuild_reset_wanted) {
	mutex_exit(&vd->vdev_rebuild_lock);
	vdev_rebuild_reset_sync(arg, tx);
	return;
	}

	vrp->vrp_rebuild_state = VDEV_REBUILD_COMPLETE;
	vrp->vrp_end_time = gethrestime_sec();

	VERIFY0(zap_update(vd->vdev_spa->spa_meta_objset, vd->vdev_top_zap,
	VDEV_TOP_ZAP_VDEV_REBUILD_PHYS, sizeof (uint64_t),
	REBUILD_PHYS_ENTRIES, vrp, tx));

	vdev_dtl_reassess(vd, tx->tx_txg, vrp->vrp_max_txg, B_TRUE, B_TRUE);
	spa_feature_decr(vd->vdev_spa, SPA_FEATURE_DEVICE_REBUILD, tx);

	spa_history_log_internal(spa, "rebuild", tx,
	"vdev_id=%llu vdev_guid=%llu complete",
	(u_longlong_t)vd->vdev_id, (u_longlong_t)vd->vdev_guid);
	vdev_rebuild_log_notify(spa, vd, ESC_ZFS_RESILVER_FINISH);

	/* Handles detaching of spares */
	spa_async_request(spa, SPA_ASYNC_REBUILD_DONE);
	vd->vdev_rebuilding = B_FALSE;
	mutex_exit(&vd->vdev_rebuild_lock);

	/*
	* While we're in syncing context take the opportunity to
	* setup the scrub when there are no more active rebuilds.
	*/
	pool_scan_func_t func = POOL_SCAN_SCRUB;
	if (dsl_scan_setup_check(&func, tx) == 0 &&
	zfs_rebuild_scrub_enabled) {
	dsl_scan_setup_sync(&func, tx);
	}

	cv_broadcast(&vd->vdev_rebuild_cv);

	/* Clear recent error events (i.e. duplicate events tracking) */
	zfs_ereport_clear(spa, NULL);
	}

	/*
	* Update the on-disk state to canceled when a rebuild finishes.
	*/
	static void
	vdev_rebuild_cancel_sync(void arg, dmu_tx_t tx)
	{
	int vdev_id = (uintptr_t)arg;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	vdev_t *vd = vdev_lookup_top(spa, vdev_id);
	vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
	vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;

	mutex_enter(&vd->vdev_rebuild_lock);
	vrp->vrp_rebuild_state = VDEV_REBUILD_CANCELED;
	vrp->vrp_end_time = gethrestime_sec();

	VERIFY0(zap_update(vd->vdev_spa->spa_meta_objset, vd->vdev_top_zap,
	VDEV_TOP_ZAP_VDEV_REBUILD_PHYS, sizeof (uint64_t),
	REBUILD_PHYS_ENTRIES, vrp, tx));

	spa_feature_decr(vd->vdev_spa, SPA_FEATURE_DEVICE_REBUILD, tx);

	spa_history_log_internal(spa, "rebuild", tx,
	"vdev_id=%llu vdev_guid=%llu canceled",
	(u_longlong_t)vd->vdev_id, (u_longlong_t)vd->vdev_guid);
	vdev_rebuild_log_notify(spa, vd, ESC_ZFS_RESILVER_FINISH);

	vd->vdev_rebuild_cancel_wanted = B_FALSE;
	vd->vdev_rebuilding = B_FALSE;
	mutex_exit(&vd->vdev_rebuild_lock);

	spa_notify_waiters(spa);
	cv_broadcast(&vd->vdev_rebuild_cv);
	}

	/*
	* Resets the progress of a running rebuild. This will occur when a new
	* vdev is added to rebuild.
	*/
	static void
	vdev_rebuild_reset_sync(void arg, dmu_tx_t tx)
	{
	int vdev_id = (uintptr_t)arg;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	vdev_t *vd = vdev_lookup_top(spa, vdev_id);
	vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
	vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;

	mutex_enter(&vd->vdev_rebuild_lock);

	ASSERT(vrp->vrp_rebuild_state == VDEV_REBUILD_ACTIVE);
	ASSERT3P(vd->vdev_rebuild_thread, ==, NULL);

	vrp->vrp_last_offset = 0;
	vrp->vrp_min_txg = 0;
	vrp->vrp_max_txg = dmu_tx_get_txg(tx);
	vrp->vrp_bytes_scanned = 0;
	vrp->vrp_bytes_issued = 0;
	vrp->vrp_bytes_rebuilt = 0;
	vrp->vrp_bytes_est = 0;
	vrp->vrp_scan_time_ms = 0;
	vr->vr_prev_scan_time_ms = 0;

	/* See vdev_rebuild_initiate_sync comment */
	VERIFY(vdev_resilver_needed(vd, &vrp->vrp_min_txg, &vrp->vrp_max_txg));

	VERIFY0(zap_update(vd->vdev_spa->spa_meta_objset, vd->vdev_top_zap,
	VDEV_TOP_ZAP_VDEV_REBUILD_PHYS, sizeof (uint64_t),
	REBUILD_PHYS_ENTRIES, vrp, tx));

	spa_history_log_internal(spa, "rebuild", tx,
	"vdev_id=%llu vdev_guid=%llu reset",
	(u_longlong_t)vd->vdev_id, (u_longlong_t)vd->vdev_guid);

	vd->vdev_rebuild_reset_wanted = B_FALSE;
	ASSERT(vd->vdev_rebuilding);

	vd->vdev_rebuild_thread = thread_create(NULL, 0,
	vdev_rebuild_thread, vd, 0, &p0, TS_RUN, maxclsyspri);

	mutex_exit(&vd->vdev_rebuild_lock);
	}

	/*
	* Clear the last rebuild status.
	*/
	void
	vdev_rebuild_clear_sync(void arg, dmu_tx_t tx)
	{
	int vdev_id = (uintptr_t)arg;
	spa_t *spa = dmu_tx_pool(tx)->dp_spa;
	vdev_t *vd = vdev_lookup_top(spa, vdev_id);
	vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
	vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
	objset_t *mos = spa_meta_objset(spa);

	mutex_enter(&vd->vdev_rebuild_lock);

	if (!spa_feature_is_enabled(spa, SPA_FEATURE_DEVICE_REBUILD) \|\|
	vrp->vrp_rebuild_state == VDEV_REBUILD_ACTIVE) {
	mutex_exit(&vd->vdev_rebuild_lock);
	return;
	}

	clear_rebuild_bytes(vd);
	memset(vrp, 0, sizeof (uint64_t) * REBUILD_PHYS_ENTRIES);

	if (vd->vdev_top_zap != 0 && zap_contains(mos, vd->vdev_top_zap,
	VDEV_TOP_ZAP_VDEV_REBUILD_PHYS) == 0) {
	VERIFY0(zap_update(mos, vd->vdev_top_zap,
	VDEV_TOP_ZAP_VDEV_REBUILD_PHYS, sizeof (uint64_t),
	REBUILD_PHYS_ENTRIES, vrp, tx));
	}

	mutex_exit(&vd->vdev_rebuild_lock);
	}

	/*
	* The zio_done_func_t callback for each rebuild I/O issued. It's responsible
	* for updating the rebuild stats and limiting the number of in flight I/Os.
	*/
	static void
	vdev_rebuild_cb(zio_t *zio)
	{
	vdev_rebuild_t *vr = zio->io_private;
	vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
	vdev_t *vd = vr->vr_top_vdev;

	mutex_enter(&vr->vr_io_lock);
	if (zio->io_error == ENXIO && !vdev_writeable(vd)) {
	/*
	* The I/O failed because the top-level vdev was unavailable.
	* Attempt to roll back to the last completed offset, in order
	* resume from the correct location if the pool is resumed.
	* (This works because spa_sync waits on spa_txg_zio before
	* it runs sync tasks.)
	*/
	uint64_t *off = &vr->vr_scan_offset[zio->io_txg & TXG_MASK];
	off = MIN(off, zio->io_offset);
	} else if (zio->io_error) {
	vrp->vrp_errors++;
	}

	abd_free(zio->io_abd);

	ASSERT3U(vr->vr_bytes_inflight, >, 0);
	vr->vr_bytes_inflight -= zio->io_size;
	cv_broadcast(&vr->vr_io_cv);
	mutex_exit(&vr->vr_io_lock);

	spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd);
	}

	/*
	* Initialize a block pointer that can be used to read the given segment
	* for sequential rebuild.
	*/
	static void
	vdev_rebuild_blkptr_init(blkptr_t bp, vdev_t vd, uint64_t start,
	uint64_t asize)
	{
	ASSERT(vd->vdev_ops == &vdev_draid_ops \|\|
	vd->vdev_ops == &vdev_mirror_ops \|\|
	vd->vdev_ops == &vdev_replacing_ops \|\|
	vd->vdev_ops == &vdev_spare_ops);

	uint64_t psize = vd->vdev_ops == &vdev_draid_ops ?
	vdev_draid_asize_to_psize(vd, asize) : asize;

	BP_ZERO(bp);

	DVA_SET_VDEV(&bp->blk_dva[0], vd->vdev_id);
	DVA_SET_OFFSET(&bp->blk_dva[0], start);
	DVA_SET_GANG(&bp->blk_dva[0], 0);
	DVA_SET_ASIZE(&bp->blk_dva[0], asize);

	BP_SET_BIRTH(bp, TXG_INITIAL, TXG_INITIAL);
	BP_SET_LSIZE(bp, psize);
	BP_SET_PSIZE(bp, psize);
	BP_SET_COMPRESS(bp, ZIO_COMPRESS_OFF);
	BP_SET_CHECKSUM(bp, ZIO_CHECKSUM_OFF);
	BP_SET_TYPE(bp, DMU_OT_NONE);
	BP_SET_LEVEL(bp, 0);
	BP_SET_DEDUP(bp, 0);
	BP_SET_BYTEORDER(bp, ZFS_HOST_BYTEORDER);
	}

	/*
	* Issues a rebuild I/O and takes care of rate limiting the number of queued
	* rebuild I/Os. The provided start and size must be properly aligned for the
	* top-level vdev type being rebuilt.
	*/
	static int
	vdev_rebuild_range(vdev_rebuild_t *vr, uint64_t start, uint64_t size)
	{
	uint64_t ms_id __maybe_unused = vr->vr_scan_msp->ms_id;
	vdev_t *vd = vr->vr_top_vdev;
	spa_t *spa = vd->vdev_spa;
	blkptr_t blk;

	ASSERT3U(ms_id, ==, start >> vd->vdev_ms_shift);
	ASSERT3U(ms_id, ==, (start + size - 1) >> vd->vdev_ms_shift);

	vr->vr_pass_bytes_scanned += size;
	vr->vr_rebuild_phys.vrp_bytes_scanned += size;

	/*
	* Rebuild the data in this range by constructing a special block
	* pointer. It has no relation to any existing blocks in the pool.
	* However, by disabling checksum verification and issuing a scrub IO
	* we can reconstruct and repair any children with missing data.
	*/
	vdev_rebuild_blkptr_init(&blk, vd, start, size);
	uint64_t psize = BP_GET_PSIZE(&blk);

	if (!vdev_dtl_need_resilver(vd, &blk.blk_dva[0], psize, TXG_UNKNOWN)) {
	vr->vr_pass_bytes_skipped += size;
	return (0);
	}

	mutex_enter(&vr->vr_io_lock);

	/* Limit in flight rebuild I/Os */
	while (vr->vr_bytes_inflight >= vr->vr_bytes_inflight_max)
	cv_wait(&vr->vr_io_cv, &vr->vr_io_lock);

	vr->vr_bytes_inflight += psize;
	mutex_exit(&vr->vr_io_lock);

	dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
	VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
	uint64_t txg = dmu_tx_get_txg(tx);

	spa_config_enter(spa, SCL_STATE_ALL, vd, RW_READER);
	mutex_enter(&vd->vdev_rebuild_lock);

	/* This is the first I/O for this txg. */
	if (vr->vr_scan_offset[txg & TXG_MASK] == 0) {
	vr->vr_scan_offset[txg & TXG_MASK] = start;
	dsl_sync_task_nowait(spa_get_dsl(spa),
	vdev_rebuild_update_sync,
	(void *)(uintptr_t)vd->vdev_id, tx);
	}

	/* When exiting write out our progress. */
	if (vdev_rebuild_should_stop(vd)) {
	mutex_enter(&vr->vr_io_lock);
	vr->vr_bytes_inflight -= psize;
	mutex_exit(&vr->vr_io_lock);
	spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd);
	mutex_exit(&vd->vdev_rebuild_lock);
	dmu_tx_commit(tx);
	return (SET_ERROR(EINTR));
	}
	mutex_exit(&vd->vdev_rebuild_lock);
	dmu_tx_commit(tx);

	vr->vr_scan_offset[txg & TXG_MASK] = start + size;
	vr->vr_pass_bytes_issued += size;
	vr->vr_rebuild_phys.vrp_bytes_issued += size;

	zio_nowait(zio_read(spa->spa_txg_zio[txg & TXG_MASK], spa, &blk,
	abd_alloc(psize, B_FALSE), psize, vdev_rebuild_cb, vr,
	ZIO_PRIORITY_REBUILD, ZIO_FLAG_RAW \| ZIO_FLAG_CANFAIL \|
	ZIO_FLAG_RESILVER, NULL));

	return (0);
	}

	/*
	* Issues rebuild I/Os for all ranges in the provided vr->vr_tree range tree.
	*/
	static int
	vdev_rebuild_ranges(vdev_rebuild_t *vr)
	{
	vdev_t *vd = vr->vr_top_vdev;
	zfs_btree_t *t = &vr->vr_scan_tree->rt_root;
	zfs_btree_index_t idx;
	int error;

	for (range_seg_t *rs = zfs_btree_first(t, &idx); rs != NULL;
	rs = zfs_btree_next(t, &idx, &idx)) {
	uint64_t start = rs_get_start(rs, vr->vr_scan_tree);
	uint64_t size = rs_get_end(rs, vr->vr_scan_tree) - start;

	/*
	* zfs_scan_suspend_progress can be set to disable rebuild
	* progress for testing. See comment in dsl_scan_sync().
	*/
	while (zfs_scan_suspend_progress &&
	!vdev_rebuild_should_stop(vd)) {
	delay(hz);
	}

	while (size > 0) {
	uint64_t chunk_size;

	/*
	* Split range into legally-sized logical chunks
	* given the constraints of the top-level vdev
	* being rebuilt (dRAID or mirror).
	*/
	ASSERT3P(vd->vdev_ops, !=, NULL);
	chunk_size = vd->vdev_ops->vdev_op_rebuild_asize(vd,
	start, size, zfs_rebuild_max_segment);

	error = vdev_rebuild_range(vr, start, chunk_size);
	if (error != 0)
	return (error);

	size -= chunk_size;
	start += chunk_size;
	}
	}

	return (0);
	}

	/*
	* Calculates the estimated capacity which remains to be scanned. Since
	* we traverse the pool in metaslab order only allocated capacity beyond
	* the vrp_last_offset need be considered. All lower offsets must have
	* already been rebuilt and are thus already included in vrp_bytes_scanned.
	*/
	static void
	vdev_rebuild_update_bytes_est(vdev_t *vd, uint64_t ms_id)
	{
	vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
	vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
	uint64_t bytes_est = vrp->vrp_bytes_scanned;

	if (vrp->vrp_last_offset < vd->vdev_ms[ms_id]->ms_start)
	return;

	for (uint64_t i = ms_id; i < vd->vdev_ms_count; i++) {
	metaslab_t *msp = vd->vdev_ms[i];

	mutex_enter(&msp->ms_lock);
	bytes_est += metaslab_allocated_space(msp);
	mutex_exit(&msp->ms_lock);
	}

	vrp->vrp_bytes_est = bytes_est;
	}

	/*
	* Load from disk the top-level vdev's rebuild information.
	*/
	int
	vdev_rebuild_load(vdev_t *vd)
	{
	vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
	vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
	spa_t *spa = vd->vdev_spa;
	int err = 0;

	mutex_enter(&vd->vdev_rebuild_lock);
	vd->vdev_rebuilding = B_FALSE;

	if (!spa_feature_is_enabled(spa, SPA_FEATURE_DEVICE_REBUILD)) {
	memset(vrp, 0, sizeof (uint64_t) * REBUILD_PHYS_ENTRIES);
	mutex_exit(&vd->vdev_rebuild_lock);
	return (SET_ERROR(ENOTSUP));
	}

	ASSERT(vd->vdev_top == vd);

	err = zap_lookup(spa->spa_meta_objset, vd->vdev_top_zap,
	VDEV_TOP_ZAP_VDEV_REBUILD_PHYS, sizeof (uint64_t),
	REBUILD_PHYS_ENTRIES, vrp);

	/*
	* A missing or damaged VDEV_TOP_ZAP_VDEV_REBUILD_PHYS should
	* not prevent a pool from being imported. Clear the rebuild
	* status allowing a new resilver/rebuild to be started.
	*/
	if (err == ENOENT \|\| err == EOVERFLOW \|\| err == ECKSUM) {
	memset(vrp, 0, sizeof (uint64_t) * REBUILD_PHYS_ENTRIES);
	} else if (err) {
	mutex_exit(&vd->vdev_rebuild_lock);
	return (err);
	}

	vr->vr_prev_scan_time_ms = vrp->vrp_scan_time_ms;
	vr->vr_top_vdev = vd;

	mutex_exit(&vd->vdev_rebuild_lock);

	return (0);
	}

	/*
	* Each scan thread is responsible for rebuilding a top-level vdev. The
	* rebuild progress in tracked on-disk in VDEV_TOP_ZAP_VDEV_REBUILD_PHYS.
	*/
	static __attribute__((noreturn)) void
	vdev_rebuild_thread(void *arg)
	{
	vdev_t *vd = arg;
	spa_t *spa = vd->vdev_spa;
	vdev_t *rvd = spa->spa_root_vdev;
	int error = 0;

	/*
	* If there's a scrub in process request that it be stopped. This
	* is not required for a correct rebuild, but we do want rebuilds to
	* emulate the resilver behavior as much as possible.
	*/
	dsl_pool_t *dsl = spa_get_dsl(spa);
	if (dsl_scan_scrubbing(dsl))
	dsl_scan_cancel(dsl);

	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	mutex_enter(&vd->vdev_rebuild_lock);

	ASSERT3P(vd->vdev_top, ==, vd);
	ASSERT3P(vd->vdev_rebuild_thread, !=, NULL);
	ASSERT(vd->vdev_rebuilding);
	ASSERT(spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REBUILD));
	ASSERT3B(vd->vdev_rebuild_cancel_wanted, ==, B_FALSE);

	vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
	vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;
	vr->vr_top_vdev = vd;
	vr->vr_scan_msp = NULL;
	vr->vr_scan_tree = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0);
	mutex_init(&vr->vr_io_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&vr->vr_io_cv, NULL, CV_DEFAULT, NULL);

	vr->vr_pass_start_time = gethrtime();
	vr->vr_pass_bytes_scanned = 0;
	vr->vr_pass_bytes_issued = 0;
	vr->vr_pass_bytes_skipped = 0;

	uint64_t update_est_time = gethrtime();
	vdev_rebuild_update_bytes_est(vd, 0);

	clear_rebuild_bytes(vr->vr_top_vdev);

	mutex_exit(&vd->vdev_rebuild_lock);

	/*
	* Systematically walk the metaslabs and issue rebuild I/Os for
	* all ranges in the allocated space map.
	*/
	for (uint64_t i = 0; i < vd->vdev_ms_count; i++) {
	metaslab_t *msp = vd->vdev_ms[i];
	vr->vr_scan_msp = msp;

	/*
	* Calculate the max number of in-flight bytes for top-level
	* vdev scanning operations (minimum 1MB, maximum 1/2 of
	* arc_c_max shared by all top-level vdevs). Limits for the
	* issuing phase are done per top-level vdev and are handled
	* separately.
	*/
	uint64_t limit = (arc_c_max / 2) / MAX(rvd->vdev_children, 1);
	vr->vr_bytes_inflight_max = MIN(limit, MAX(1ULL << 20,
	zfs_rebuild_vdev_limit * vd->vdev_children));

	/*
	* Removal of vdevs from the vdev tree may eliminate the need
	* for the rebuild, in which case it should be canceled. The
	* vdev_rebuild_cancel_wanted flag is set until the sync task
	* completes. This may be after the rebuild thread exits.
	*/
	if (vdev_rebuild_should_cancel(vd)) {
	vd->vdev_rebuild_cancel_wanted = B_TRUE;
	error = EINTR;
	break;
	}

	ASSERT0(range_tree_space(vr->vr_scan_tree));

	/* Disable any new allocations to this metaslab */
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	metaslab_disable(msp);

	mutex_enter(&msp->ms_sync_lock);
	mutex_enter(&msp->ms_lock);

	/*
	* If there are outstanding allocations wait for them to be
	* synced. This is needed to ensure all allocated ranges are
	* on disk and therefore will be rebuilt.
	*/
	for (int j = 0; j < TXG_SIZE; j++) {
	if (range_tree_space(msp->ms_allocating[j])) {
	mutex_exit(&msp->ms_lock);
	mutex_exit(&msp->ms_sync_lock);
	txg_wait_synced(dsl, 0);
	mutex_enter(&msp->ms_sync_lock);
	mutex_enter(&msp->ms_lock);
	break;
	}
	}

	/*
	* When a metaslab has been allocated from read its allocated
	* ranges from the space map object into the vr_scan_tree.
	* Then add inflight / unflushed ranges and remove inflight /
	* unflushed frees. This is the minimum range to be rebuilt.
	*/
	if (msp->ms_sm != NULL) {
	VERIFY0(space_map_load(msp->ms_sm,
	vr->vr_scan_tree, SM_ALLOC));

	for (int i = 0; i < TXG_SIZE; i++) {
	ASSERT0(range_tree_space(
	msp->ms_allocating[i]));
	}

	range_tree_walk(msp->ms_unflushed_allocs,
	range_tree_add, vr->vr_scan_tree);
	range_tree_walk(msp->ms_unflushed_frees,
	range_tree_remove, vr->vr_scan_tree);

	/*
	* Remove ranges which have already been rebuilt based
	* on the last offset. This can happen when restarting
	* a scan after exporting and re-importing the pool.
	*/
	range_tree_clear(vr->vr_scan_tree, 0,
	vrp->vrp_last_offset);
	}

	mutex_exit(&msp->ms_lock);
	mutex_exit(&msp->ms_sync_lock);

	/*
	* To provide an accurate estimate re-calculate the estimated
	* size every 5 minutes to account for recent allocations and
	* frees made to space maps which have not yet been rebuilt.
	*/
	if (gethrtime() > update_est_time + SEC2NSEC(300)) {
	update_est_time = gethrtime();
	vdev_rebuild_update_bytes_est(vd, i);
	}

	/*
	* Walk the allocated space map and issue the rebuild I/O.
	*/
	error = vdev_rebuild_ranges(vr);
	range_tree_vacate(vr->vr_scan_tree, NULL, NULL);

	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	metaslab_enable(msp, B_FALSE, B_FALSE);

	if (error != 0)
	break;
	}

	range_tree_destroy(vr->vr_scan_tree);
	spa_config_exit(spa, SCL_CONFIG, FTAG);

	/* Wait for any remaining rebuild I/O to complete */
	mutex_enter(&vr->vr_io_lock);
	while (vr->vr_bytes_inflight > 0)
	cv_wait(&vr->vr_io_cv, &vr->vr_io_lock);

	mutex_exit(&vr->vr_io_lock);

	mutex_destroy(&vr->vr_io_lock);
	cv_destroy(&vr->vr_io_cv);

	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);

	dsl_pool_t *dp = spa_get_dsl(spa);
	dmu_tx_t *tx = dmu_tx_create_dd(dp->dp_mos_dir);
	VERIFY0(dmu_tx_assign(tx, TXG_WAIT));

	mutex_enter(&vd->vdev_rebuild_lock);
	if (error == 0) {
	/*
	* After a successful rebuild clear the DTLs of all ranges
	* which were missing when the rebuild was started. These
	* ranges must have been rebuilt as a consequence of rebuilding
	* all allocated space. Note that unlike a scrub or resilver
	* the rebuild operation will reconstruct data only referenced
	* by a pool checkpoint. See the dsl_scan_done() comments.
	*/
	dsl_sync_task_nowait(dp, vdev_rebuild_complete_sync,
	(void *)(uintptr_t)vd->vdev_id, tx);
	} else if (vd->vdev_rebuild_cancel_wanted) {
	/*
	* The rebuild operation was canceled. This will occur when
	* a device participating in the rebuild is detached.
	*/
	dsl_sync_task_nowait(dp, vdev_rebuild_cancel_sync,
	(void *)(uintptr_t)vd->vdev_id, tx);
	} else if (vd->vdev_rebuild_reset_wanted) {
	/*
	* Reset the running rebuild without canceling and restarting
	* it. This will occur when a new device is attached and must
	* participate in the rebuild.
	*/
	dsl_sync_task_nowait(dp, vdev_rebuild_reset_sync,
	(void *)(uintptr_t)vd->vdev_id, tx);
	} else {
	/*
	* The rebuild operation should be suspended. This may occur
	* when detaching a child vdev or when exporting the pool. The
	* rebuild is left in the active state so it will be resumed.
	*/
	ASSERT(vrp->vrp_rebuild_state == VDEV_REBUILD_ACTIVE);
	vd->vdev_rebuilding = B_FALSE;
	}

	dmu_tx_commit(tx);

	vd->vdev_rebuild_thread = NULL;
	mutex_exit(&vd->vdev_rebuild_lock);
	spa_config_exit(spa, SCL_CONFIG, FTAG);

	cv_broadcast(&vd->vdev_rebuild_cv);

	thread_exit();
	}

	/*
	* Returns B_TRUE if any top-level vdev are rebuilding.
	*/
	boolean_t
	vdev_rebuild_active(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;
	boolean_t ret = B_FALSE;

	if (vd == spa->spa_root_vdev) {
	for (uint64_t i = 0; i < vd->vdev_children; i++) {
	ret = vdev_rebuild_active(vd->vdev_child[i]);
	if (ret)
	return (ret);
	}
	} else if (vd->vdev_top_zap != 0) {
	vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
	vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;

	mutex_enter(&vd->vdev_rebuild_lock);
	ret = (vrp->vrp_rebuild_state == VDEV_REBUILD_ACTIVE);
	mutex_exit(&vd->vdev_rebuild_lock);
	}

	return (ret);
	}

	/*
	* Start a rebuild operation. The rebuild may be restarted when the
	* top-level vdev is currently actively rebuilding.
	*/
	void
	vdev_rebuild(vdev_t *vd)
	{
	vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
	vdev_rebuild_phys_t *vrp __maybe_unused = &vr->vr_rebuild_phys;

	ASSERT(vd->vdev_top == vd);
	ASSERT(vdev_is_concrete(vd));
	ASSERT(!vd->vdev_removing);
	ASSERT(spa_feature_is_enabled(vd->vdev_spa,
	SPA_FEATURE_DEVICE_REBUILD));

	mutex_enter(&vd->vdev_rebuild_lock);
	if (vd->vdev_rebuilding) {
	ASSERT3U(vrp->vrp_rebuild_state, ==, VDEV_REBUILD_ACTIVE);

	/*
	* Signal a running rebuild operation that it should restart
	* from the beginning because a new device was attached. The
	* vdev_rebuild_reset_wanted flag is set until the sync task
	* completes. This may be after the rebuild thread exits.
	*/
	if (!vd->vdev_rebuild_reset_wanted)
	vd->vdev_rebuild_reset_wanted = B_TRUE;
	} else {
	vdev_rebuild_initiate(vd);
	}
	mutex_exit(&vd->vdev_rebuild_lock);
	}

	static void
	vdev_rebuild_restart_impl(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;

	if (vd == spa->spa_root_vdev) {
	for (uint64_t i = 0; i < vd->vdev_children; i++)
	vdev_rebuild_restart_impl(vd->vdev_child[i]);

	} else if (vd->vdev_top_zap != 0) {
	vdev_rebuild_t *vr = &vd->vdev_rebuild_config;
	vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;

	mutex_enter(&vd->vdev_rebuild_lock);
	if (vrp->vrp_rebuild_state == VDEV_REBUILD_ACTIVE &&
	vdev_writeable(vd) && !vd->vdev_rebuilding) {
	ASSERT(spa_feature_is_active(spa,
	SPA_FEATURE_DEVICE_REBUILD));
	vd->vdev_rebuilding = B_TRUE;
	vd->vdev_rebuild_thread = thread_create(NULL, 0,
	vdev_rebuild_thread, vd, 0, &p0, TS_RUN,
	maxclsyspri);
	}
	mutex_exit(&vd->vdev_rebuild_lock);
	}
	}

	/*
	* Conditionally restart all of the vdev_rebuild_thread's for a pool. The
	* feature flag must be active and the rebuild in the active state. This
	* cannot be used to start a new rebuild.
	*/
	void
	vdev_rebuild_restart(spa_t *spa)
	{
	ASSERT(MUTEX_HELD(&spa_namespace_lock) \|\|
	spa->spa_load_thread == curthread);

	vdev_rebuild_restart_impl(spa->spa_root_vdev);
	}

	/*
	* Stop and wait for all of the vdev_rebuild_thread's associated with the
	* vdev tree provide to be terminated (canceled or stopped).
	*/
	void
	vdev_rebuild_stop_wait(vdev_t *vd)
	{
	spa_t *spa = vd->vdev_spa;

	- ASSERT(MUTEX_HELD(&spa_namespace_lock));
	+ ASSERT(MUTEX_HELD(&spa_namespace_lock) \|\|
	+ spa->spa_export_thread == curthread);

	if (vd == spa->spa_root_vdev) {
	for (uint64_t i = 0; i < vd->vdev_children; i++)
	vdev_rebuild_stop_wait(vd->vdev_child[i]);

	} else if (vd->vdev_top_zap != 0) {
	ASSERT(vd == vd->vdev_top);

	mutex_enter(&vd->vdev_rebuild_lock);
	if (vd->vdev_rebuild_thread != NULL) {
	vd->vdev_rebuild_exit_wanted = B_TRUE;
	while (vd->vdev_rebuilding) {
	cv_wait(&vd->vdev_rebuild_cv,
	&vd->vdev_rebuild_lock);
	}
	vd->vdev_rebuild_exit_wanted = B_FALSE;
	}
	mutex_exit(&vd->vdev_rebuild_lock);
	}
	}

	/*
	* Stop all rebuild operations but leave them in the active state so they
	* will be resumed when importing the pool.
	*/
	void
	vdev_rebuild_stop_all(spa_t *spa)
	{
	vdev_rebuild_stop_wait(spa->spa_root_vdev);
	}

	/*
	* Rebuild statistics reported per top-level vdev.
	*/
	int
	vdev_rebuild_get_stats(vdev_t tvd, vdev_rebuild_stat_t vrs)
	{
	spa_t *spa = tvd->vdev_spa;

	if (!spa_feature_is_enabled(spa, SPA_FEATURE_DEVICE_REBUILD))
	return (SET_ERROR(ENOTSUP));

	if (tvd != tvd->vdev_top \|\| tvd->vdev_top_zap == 0)
	return (SET_ERROR(EINVAL));

	int error = zap_contains(spa_meta_objset(spa),
	tvd->vdev_top_zap, VDEV_TOP_ZAP_VDEV_REBUILD_PHYS);

	if (error == ENOENT) {
	memset(vrs, 0, sizeof (vdev_rebuild_stat_t));
	vrs->vrs_state = VDEV_REBUILD_NONE;
	error = 0;
	} else if (error == 0) {
	vdev_rebuild_t *vr = &tvd->vdev_rebuild_config;
	vdev_rebuild_phys_t *vrp = &vr->vr_rebuild_phys;

	mutex_enter(&tvd->vdev_rebuild_lock);
	vrs->vrs_state = vrp->vrp_rebuild_state;
	vrs->vrs_start_time = vrp->vrp_start_time;
	vrs->vrs_end_time = vrp->vrp_end_time;
	vrs->vrs_scan_time_ms = vrp->vrp_scan_time_ms;
	vrs->vrs_bytes_scanned = vrp->vrp_bytes_scanned;
	vrs->vrs_bytes_issued = vrp->vrp_bytes_issued;
	vrs->vrs_bytes_rebuilt = vrp->vrp_bytes_rebuilt;
	vrs->vrs_bytes_est = vrp->vrp_bytes_est;
	vrs->vrs_errors = vrp->vrp_errors;
	vrs->vrs_pass_time_ms = NSEC2MSEC(gethrtime() -
	vr->vr_pass_start_time);
	vrs->vrs_pass_bytes_scanned = vr->vr_pass_bytes_scanned;
	vrs->vrs_pass_bytes_issued = vr->vr_pass_bytes_issued;
	vrs->vrs_pass_bytes_skipped = vr->vr_pass_bytes_skipped;
	mutex_exit(&tvd->vdev_rebuild_lock);
	}

	return (error);
	}

	ZFS_MODULE_PARAM(zfs, zfs_, rebuild_max_segment, U64, ZMOD_RW,
	"Max segment size in bytes of rebuild reads");

	ZFS_MODULE_PARAM(zfs, zfs_, rebuild_vdev_limit, U64, ZMOD_RW,
	"Max bytes in flight per leaf vdev for sequential resilvers");

	ZFS_MODULE_PARAM(zfs, zfs_, rebuild_scrub_enabled, INT, ZMOD_RW,
	"Automatically scrub after sequential resilver completes");
	diff --git a/sys/contrib/openzfs/module/zfs/vdev_trim.c b/sys/contrib/openzfs/module/zfs/vdev_trim.c
	index 9753d5a1ea04..9cf10332e8bf 100644
	--- a/sys/contrib/openzfs/module/zfs/vdev_trim.c
	+++ b/sys/contrib/openzfs/module/zfs/vdev_trim.c
	@@ -1,1785 +1,1788 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2016, 2024 by Delphix. All rights reserved.
	* Copyright (c) 2019 by Lawrence Livermore National Security, LLC.
	* Copyright (c) 2021 Hewlett Packard Enterprise Development LP
	* Copyright 2023 RackTop Systems, Inc.
	*/

	#include <sys/spa.h>
	#include <sys/spa_impl.h>
	#include <sys/txg.h>
	#include <sys/vdev_impl.h>
	#include <sys/vdev_trim.h>
	#include <sys/metaslab_impl.h>
	#include <sys/dsl_synctask.h>
	#include <sys/zap.h>
	#include <sys/dmu_tx.h>
	#include <sys/arc_impl.h>

	/*
	* TRIM is a feature which is used to notify a SSD that some previously
	* written space is no longer allocated by the pool. This is useful because
	* writes to a SSD must be performed to blocks which have first been erased.
	* Ensuring the SSD always has a supply of erased blocks for new writes
	* helps prevent the performance from deteriorating.
	*
	* There are two supported TRIM methods; manual and automatic.
	*
	* Manual TRIM:
	*
	* A manual TRIM is initiated by running the 'zpool trim' command. A single
	* 'vdev_trim' thread is created for each leaf vdev, and it is responsible for
	* managing that vdev TRIM process. This involves iterating over all the
	* metaslabs, calculating the unallocated space ranges, and then issuing the
	* required TRIM I/Os.
	*
	* While a metaslab is being actively trimmed it is not eligible to perform
	* new allocations. After traversing all of the metaslabs the thread is
	* terminated. Finally, both the requested options and current progress of
	* the TRIM are regularly written to the pool. This allows the TRIM to be
	* suspended and resumed as needed.
	*
	* Automatic TRIM:
	*
	* An automatic TRIM is enabled by setting the 'autotrim' pool property
	* to 'on'. When enabled, a `vdev_autotrim' thread is created for each
	* top-level (not leaf) vdev in the pool. These threads perform the same
	* core TRIM process as a manual TRIM, but with a few key differences.
	*
	* 1) Automatic TRIM happens continuously in the background and operates
	* solely on recently freed blocks (ms_trim not ms_allocatable).
	*
	* 2) Each thread is associated with a top-level (not leaf) vdev. This has
	* the benefit of simplifying the threading model, it makes it easier
	* to coordinate administrative commands, and it ensures only a single
	* metaslab is disabled at a time. Unlike manual TRIM, this means each
	* 'vdev_autotrim' thread is responsible for issuing TRIM I/Os for its
	* children.
	*
	* 3) There is no automatic TRIM progress information stored on disk, nor
	* is it reported by 'zpool status'.
	*
	* While the automatic TRIM process is highly effective it is more likely
	* than a manual TRIM to encounter tiny ranges. Ranges less than or equal to
	* 'zfs_trim_extent_bytes_min' (32k) are considered too small to efficiently
	* TRIM and are skipped. This means small amounts of freed space may not
	* be automatically trimmed.
	*
	* Furthermore, devices with attached hot spares and devices being actively
	* replaced are skipped. This is done to avoid adding additional stress to
	* a potentially unhealthy device and to minimize the required rebuild time.
	*
	* For this reason it may be beneficial to occasionally manually TRIM a pool
	* even when automatic TRIM is enabled.
	*/

	/*
	* Maximum size of TRIM I/O, ranges will be chunked in to 128MiB lengths.
	*/
	static unsigned int zfs_trim_extent_bytes_max = 128 * 1024 * 1024;

	/*
	* Minimum size of TRIM I/O, extents smaller than 32Kib will be skipped.
	*/
	static unsigned int zfs_trim_extent_bytes_min = 32 * 1024;

	/*
	* Skip uninitialized metaslabs during the TRIM process. This option is
	* useful for pools constructed from large thinly-provisioned devices where
	* TRIM operations are slow. As a pool ages an increasing fraction of
	* the pools metaslabs will be initialized progressively degrading the
	* usefulness of this option. This setting is stored when starting a
	* manual TRIM and will persist for the duration of the requested TRIM.
	*/
	unsigned int zfs_trim_metaslab_skip = 0;

	/*
	* Maximum number of queued TRIM I/Os per leaf vdev. The number of
	* concurrent TRIM I/Os issued to the device is controlled by the
	* zfs_vdev_trim_min_active and zfs_vdev_trim_max_active module options.
	*/
	static unsigned int zfs_trim_queue_limit = 10;

	/*
	* The minimum number of transaction groups between automatic trims of a
	* metaslab. This setting represents a trade-off between issuing more
	* efficient TRIM operations, by allowing them to be aggregated longer,
	* and issuing them promptly so the trimmed space is available. Note
	* that this value is a minimum; metaslabs can be trimmed less frequently
	* when there are a large number of ranges which need to be trimmed.
	*
	* Increasing this value will allow frees to be aggregated for a longer
	* time. This can result is larger TRIM operations, and increased memory
	* usage in order to track the ranges to be trimmed. Decreasing this value
	* has the opposite effect. The default value of 32 was determined though
	* testing to be a reasonable compromise.
	*/
	static unsigned int zfs_trim_txg_batch = 32;

	/*
	* The trim_args are a control structure which describe how a leaf vdev
	* should be trimmed. The core elements are the vdev, the metaslab being
	* trimmed and a range tree containing the extents to TRIM. All provided
	* ranges must be within the metaslab.
	*/
	typedef struct trim_args {
	/*
	* These fields are set by the caller of vdev_trim_ranges().
	*/
	vdev_t trim_vdev; / Leaf vdev to TRIM */
	metaslab_t trim_msp; / Disabled metaslab */
	range_tree_t trim_tree; / TRIM ranges (in metaslab) */
	trim_type_t trim_type; /* Manual or auto TRIM */
	uint64_t trim_extent_bytes_max; /* Maximum TRIM I/O size */
	uint64_t trim_extent_bytes_min; /* Minimum TRIM I/O size */
	enum trim_flag trim_flags; /* TRIM flags (secure) */

	/*
	* These fields are updated by vdev_trim_ranges().
	*/
	hrtime_t trim_start_time; /* Start time */
	uint64_t trim_bytes_done; /* Bytes trimmed */
	} trim_args_t;

	/*
	* Determines whether a vdev_trim_thread() should be stopped.
	*/
	static boolean_t
	vdev_trim_should_stop(vdev_t *vd)
	{
	return (vd->vdev_trim_exit_wanted \|\| !vdev_writeable(vd) \|\|
	vd->vdev_detached \|\| vd->vdev_top->vdev_removing \|\|
	vd->vdev_top->vdev_rz_expanding);
	}

	/*
	* Determines whether a vdev_autotrim_thread() should be stopped.
	*/
	static boolean_t
	vdev_autotrim_should_stop(vdev_t *tvd)
	{
	return (tvd->vdev_autotrim_exit_wanted \|\|
	!vdev_writeable(tvd) \|\| tvd->vdev_removing \|\|
	tvd->vdev_rz_expanding \|\|
	spa_get_autotrim(tvd->vdev_spa) == SPA_AUTOTRIM_OFF);
	}

	/*
	* Wait for given number of kicks, return true if the wait is aborted due to
	* vdev_autotrim_exit_wanted.
	*/
	static boolean_t
	vdev_autotrim_wait_kick(vdev_t *vd, int num_of_kick)
	{
	mutex_enter(&vd->vdev_autotrim_lock);
	for (int i = 0; i < num_of_kick; i++) {
	if (vd->vdev_autotrim_exit_wanted)
	break;
	cv_wait_idle(&vd->vdev_autotrim_kick_cv,
	&vd->vdev_autotrim_lock);
	}
	boolean_t exit_wanted = vd->vdev_autotrim_exit_wanted;
	mutex_exit(&vd->vdev_autotrim_lock);

	return (exit_wanted);
	}

	/*
	* The sync task for updating the on-disk state of a manual TRIM. This
	* is scheduled by vdev_trim_change_state().
	*/
	static void
	vdev_trim_zap_update_sync(void arg, dmu_tx_t tx)
	{
	/*
	* We pass in the guid instead of the vdev_t since the vdev may
	* have been freed prior to the sync task being processed. This
	* happens when a vdev is detached as we call spa_config_vdev_exit(),
	* stop the trimming thread, schedule the sync task, and free
	* the vdev. Later when the scheduled sync task is invoked, it would
	* find that the vdev has been freed.
	*/
	uint64_t guid = (uint64_t )arg;
	uint64_t txg = dmu_tx_get_txg(tx);
	kmem_free(arg, sizeof (uint64_t));

	vdev_t *vd = spa_lookup_by_guid(tx->tx_pool->dp_spa, guid, B_FALSE);
	if (vd == NULL \|\| vd->vdev_top->vdev_removing \|\|
	!vdev_is_concrete(vd) \|\| vd->vdev_top->vdev_rz_expanding)
	return;

	uint64_t last_offset = vd->vdev_trim_offset[txg & TXG_MASK];
	vd->vdev_trim_offset[txg & TXG_MASK] = 0;

	VERIFY3U(vd->vdev_leaf_zap, !=, 0);

	objset_t *mos = vd->vdev_spa->spa_meta_objset;

	if (last_offset > 0 \|\| vd->vdev_trim_last_offset == UINT64_MAX) {

	if (vd->vdev_trim_last_offset == UINT64_MAX)
	last_offset = 0;

	vd->vdev_trim_last_offset = last_offset;
	VERIFY0(zap_update(mos, vd->vdev_leaf_zap,
	VDEV_LEAF_ZAP_TRIM_LAST_OFFSET,
	sizeof (last_offset), 1, &last_offset, tx));
	}

	if (vd->vdev_trim_action_time > 0) {
	uint64_t val = (uint64_t)vd->vdev_trim_action_time;
	VERIFY0(zap_update(mos, vd->vdev_leaf_zap,
	VDEV_LEAF_ZAP_TRIM_ACTION_TIME, sizeof (val),
	1, &val, tx));
	}

	if (vd->vdev_trim_rate > 0) {
	uint64_t rate = (uint64_t)vd->vdev_trim_rate;

	if (rate == UINT64_MAX)
	rate = 0;

	VERIFY0(zap_update(mos, vd->vdev_leaf_zap,
	VDEV_LEAF_ZAP_TRIM_RATE, sizeof (rate), 1, &rate, tx));
	}

	uint64_t partial = vd->vdev_trim_partial;
	if (partial == UINT64_MAX)
	partial = 0;

	VERIFY0(zap_update(mos, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_PARTIAL,
	sizeof (partial), 1, &partial, tx));

	uint64_t secure = vd->vdev_trim_secure;
	if (secure == UINT64_MAX)
	secure = 0;

	VERIFY0(zap_update(mos, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_SECURE,
	sizeof (secure), 1, &secure, tx));


	uint64_t trim_state = vd->vdev_trim_state;
	VERIFY0(zap_update(mos, vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_STATE,
	sizeof (trim_state), 1, &trim_state, tx));
	}

	/*
	* Update the on-disk state of a manual TRIM. This is called to request
	* that a TRIM be started/suspended/canceled, or to change one of the
	* TRIM options (partial, secure, rate).
	*/
	static void
	vdev_trim_change_state(vdev_t *vd, vdev_trim_state_t new_state,
	uint64_t rate, boolean_t partial, boolean_t secure)
	{
	ASSERT(MUTEX_HELD(&vd->vdev_trim_lock));
	spa_t *spa = vd->vdev_spa;

	if (new_state == vd->vdev_trim_state)
	return;

	/*
	* Copy the vd's guid, this will be freed by the sync task.
	*/
	uint64_t *guid = kmem_zalloc(sizeof (uint64_t), KM_SLEEP);
	*guid = vd->vdev_guid;

	/*
	* If we're suspending, then preserve the original start time.
	*/
	if (vd->vdev_trim_state != VDEV_TRIM_SUSPENDED) {
	vd->vdev_trim_action_time = gethrestime_sec();
	}

	/*
	* If we're activating, then preserve the requested rate and trim
	* method. Setting the last offset and rate to UINT64_MAX is used
	* as a sentinel to indicate they should be reset to default values.
	*/
	if (new_state == VDEV_TRIM_ACTIVE) {
	if (vd->vdev_trim_state == VDEV_TRIM_COMPLETE \|\|
	vd->vdev_trim_state == VDEV_TRIM_CANCELED) {
	vd->vdev_trim_last_offset = UINT64_MAX;
	vd->vdev_trim_rate = UINT64_MAX;
	vd->vdev_trim_partial = UINT64_MAX;
	vd->vdev_trim_secure = UINT64_MAX;
	}

	if (rate != 0)
	vd->vdev_trim_rate = rate;

	if (partial != 0)
	vd->vdev_trim_partial = partial;

	if (secure != 0)
	vd->vdev_trim_secure = secure;
	}

	vdev_trim_state_t old_state = vd->vdev_trim_state;
	boolean_t resumed = (old_state == VDEV_TRIM_SUSPENDED);
	vd->vdev_trim_state = new_state;

	dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
	VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
	dsl_sync_task_nowait(spa_get_dsl(spa), vdev_trim_zap_update_sync,
	guid, tx);

	switch (new_state) {
	case VDEV_TRIM_ACTIVE:
	spa_event_notify(spa, vd, NULL,
	resumed ? ESC_ZFS_TRIM_RESUME : ESC_ZFS_TRIM_START);
	spa_history_log_internal(spa, "trim", tx,
	"vdev=%s activated", vd->vdev_path);
	break;
	case VDEV_TRIM_SUSPENDED:
	spa_event_notify(spa, vd, NULL, ESC_ZFS_TRIM_SUSPEND);
	spa_history_log_internal(spa, "trim", tx,
	"vdev=%s suspended", vd->vdev_path);
	break;
	case VDEV_TRIM_CANCELED:
	if (old_state == VDEV_TRIM_ACTIVE \|\|
	old_state == VDEV_TRIM_SUSPENDED) {
	spa_event_notify(spa, vd, NULL, ESC_ZFS_TRIM_CANCEL);
	spa_history_log_internal(spa, "trim", tx,
	"vdev=%s canceled", vd->vdev_path);
	}
	break;
	case VDEV_TRIM_COMPLETE:
	spa_event_notify(spa, vd, NULL, ESC_ZFS_TRIM_FINISH);
	spa_history_log_internal(spa, "trim", tx,
	"vdev=%s complete", vd->vdev_path);
	break;
	default:
	panic("invalid state %llu", (unsigned long long)new_state);
	}

	dmu_tx_commit(tx);

	if (new_state != VDEV_TRIM_ACTIVE)
	spa_notify_waiters(spa);
	}

	/*
	* The zio_done_func_t done callback for each manual TRIM issued. It is
	* responsible for updating the TRIM stats, reissuing failed TRIM I/Os,
	* and limiting the number of in flight TRIM I/Os.
	*/
	static void
	vdev_trim_cb(zio_t *zio)
	{
	vdev_t *vd = zio->io_vd;

	mutex_enter(&vd->vdev_trim_io_lock);
	if (zio->io_error == ENXIO && !vdev_writeable(vd)) {
	/*
	* The I/O failed because the vdev was unavailable; roll the
	* last offset back. (This works because spa_sync waits on
	* spa_txg_zio before it runs sync tasks.)
	*/
	uint64_t *offset =
	&vd->vdev_trim_offset[zio->io_txg & TXG_MASK];
	offset = MIN(offset, zio->io_offset);
	} else {
	if (zio->io_error != 0) {
	vd->vdev_stat.vs_trim_errors++;
	spa_iostats_trim_add(vd->vdev_spa, TRIM_TYPE_MANUAL,
	0, 0, 0, 0, 1, zio->io_orig_size);
	} else {
	spa_iostats_trim_add(vd->vdev_spa, TRIM_TYPE_MANUAL,
	1, zio->io_orig_size, 0, 0, 0, 0);
	}

	vd->vdev_trim_bytes_done += zio->io_orig_size;
	}

	ASSERT3U(vd->vdev_trim_inflight[TRIM_TYPE_MANUAL], >, 0);
	vd->vdev_trim_inflight[TRIM_TYPE_MANUAL]--;
	cv_broadcast(&vd->vdev_trim_io_cv);
	mutex_exit(&vd->vdev_trim_io_lock);

	spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd);
	}

	/*
	* The zio_done_func_t done callback for each automatic TRIM issued. It
	* is responsible for updating the TRIM stats and limiting the number of
	* in flight TRIM I/Os. Automatic TRIM I/Os are best effort and are
	* never reissued on failure.
	*/
	static void
	vdev_autotrim_cb(zio_t *zio)
	{
	vdev_t *vd = zio->io_vd;

	mutex_enter(&vd->vdev_trim_io_lock);

	if (zio->io_error != 0) {
	vd->vdev_stat.vs_trim_errors++;
	spa_iostats_trim_add(vd->vdev_spa, TRIM_TYPE_AUTO,
	0, 0, 0, 0, 1, zio->io_orig_size);
	} else {
	spa_iostats_trim_add(vd->vdev_spa, TRIM_TYPE_AUTO,
	1, zio->io_orig_size, 0, 0, 0, 0);
	}

	ASSERT3U(vd->vdev_trim_inflight[TRIM_TYPE_AUTO], >, 0);
	vd->vdev_trim_inflight[TRIM_TYPE_AUTO]--;
	cv_broadcast(&vd->vdev_trim_io_cv);
	mutex_exit(&vd->vdev_trim_io_lock);

	spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd);
	}

	/*
	* The zio_done_func_t done callback for each TRIM issued via
	* vdev_trim_simple(). It is responsible for updating the TRIM stats and
	* limiting the number of in flight TRIM I/Os. Simple TRIM I/Os are best
	* effort and are never reissued on failure.
	*/
	static void
	vdev_trim_simple_cb(zio_t *zio)
	{
	vdev_t *vd = zio->io_vd;

	mutex_enter(&vd->vdev_trim_io_lock);

	if (zio->io_error != 0) {
	vd->vdev_stat.vs_trim_errors++;
	spa_iostats_trim_add(vd->vdev_spa, TRIM_TYPE_SIMPLE,
	0, 0, 0, 0, 1, zio->io_orig_size);
	} else {
	spa_iostats_trim_add(vd->vdev_spa, TRIM_TYPE_SIMPLE,
	1, zio->io_orig_size, 0, 0, 0, 0);
	}

	ASSERT3U(vd->vdev_trim_inflight[TRIM_TYPE_SIMPLE], >, 0);
	vd->vdev_trim_inflight[TRIM_TYPE_SIMPLE]--;
	cv_broadcast(&vd->vdev_trim_io_cv);
	mutex_exit(&vd->vdev_trim_io_lock);

	spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd);
	}
	/*
	* Returns the average trim rate in bytes/sec for the ta->trim_vdev.
	*/
	static uint64_t
	vdev_trim_calculate_rate(trim_args_t *ta)
	{
	return (ta->trim_bytes_done * 1000 /
	(NSEC2MSEC(gethrtime() - ta->trim_start_time) + 1));
	}

	/*
	* Issues a physical TRIM and takes care of rate limiting (bytes/sec)
	* and number of concurrent TRIM I/Os.
	*/
	static int
	vdev_trim_range(trim_args_t *ta, uint64_t start, uint64_t size)
	{
	vdev_t *vd = ta->trim_vdev;
	spa_t *spa = vd->vdev_spa;
	void *cb;

	mutex_enter(&vd->vdev_trim_io_lock);

	/*
	* Limit manual TRIM I/Os to the requested rate. This does not
	* apply to automatic TRIM since no per vdev rate can be specified.
	*/
	if (ta->trim_type == TRIM_TYPE_MANUAL) {
	while (vd->vdev_trim_rate != 0 && !vdev_trim_should_stop(vd) &&
	vdev_trim_calculate_rate(ta) > vd->vdev_trim_rate) {
	cv_timedwait_idle(&vd->vdev_trim_io_cv,
	&vd->vdev_trim_io_lock, ddi_get_lbolt() +
	MSEC_TO_TICK(10));
	}
	}
	ta->trim_bytes_done += size;

	/* Limit in flight trimming I/Os */
	while (vd->vdev_trim_inflight[0] + vd->vdev_trim_inflight[1] +
	vd->vdev_trim_inflight[2] >= zfs_trim_queue_limit) {
	cv_wait(&vd->vdev_trim_io_cv, &vd->vdev_trim_io_lock);
	}
	vd->vdev_trim_inflight[ta->trim_type]++;
	mutex_exit(&vd->vdev_trim_io_lock);

	dmu_tx_t *tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
	VERIFY0(dmu_tx_assign(tx, TXG_WAIT));
	uint64_t txg = dmu_tx_get_txg(tx);

	spa_config_enter(spa, SCL_STATE_ALL, vd, RW_READER);
	mutex_enter(&vd->vdev_trim_lock);

	if (ta->trim_type == TRIM_TYPE_MANUAL &&
	vd->vdev_trim_offset[txg & TXG_MASK] == 0) {
	uint64_t *guid = kmem_zalloc(sizeof (uint64_t), KM_SLEEP);
	*guid = vd->vdev_guid;

	/* This is the first write of this txg. */
	dsl_sync_task_nowait(spa_get_dsl(spa),
	vdev_trim_zap_update_sync, guid, tx);
	}

	/*
	* We know the vdev_t will still be around since all consumers of
	* vdev_free must stop the trimming first.
	*/
	if ((ta->trim_type == TRIM_TYPE_MANUAL &&
	vdev_trim_should_stop(vd)) \|\|
	(ta->trim_type == TRIM_TYPE_AUTO &&
	vdev_autotrim_should_stop(vd->vdev_top))) {
	mutex_enter(&vd->vdev_trim_io_lock);
	vd->vdev_trim_inflight[ta->trim_type]--;
	mutex_exit(&vd->vdev_trim_io_lock);
	spa_config_exit(vd->vdev_spa, SCL_STATE_ALL, vd);
	mutex_exit(&vd->vdev_trim_lock);
	dmu_tx_commit(tx);
	return (SET_ERROR(EINTR));
	}
	mutex_exit(&vd->vdev_trim_lock);

	if (ta->trim_type == TRIM_TYPE_MANUAL)
	vd->vdev_trim_offset[txg & TXG_MASK] = start + size;

	if (ta->trim_type == TRIM_TYPE_MANUAL) {
	cb = vdev_trim_cb;
	} else if (ta->trim_type == TRIM_TYPE_AUTO) {
	cb = vdev_autotrim_cb;
	} else {
	cb = vdev_trim_simple_cb;
	}

	zio_nowait(zio_trim(spa->spa_txg_zio[txg & TXG_MASK], vd,
	start, size, cb, NULL, ZIO_PRIORITY_TRIM, ZIO_FLAG_CANFAIL,
	ta->trim_flags));
	/* vdev_trim_cb and vdev_autotrim_cb release SCL_STATE_ALL */

	dmu_tx_commit(tx);

	return (0);
	}

	/*
	* Issues TRIM I/Os for all ranges in the provided ta->trim_tree range tree.
	* Additional parameters describing how the TRIM should be performed must
	* be set in the trim_args structure. See the trim_args definition for
	* additional information.
	*/
	static int
	vdev_trim_ranges(trim_args_t *ta)
	{
	vdev_t *vd = ta->trim_vdev;
	zfs_btree_t *t = &ta->trim_tree->rt_root;
	zfs_btree_index_t idx;
	uint64_t extent_bytes_max = ta->trim_extent_bytes_max;
	uint64_t extent_bytes_min = ta->trim_extent_bytes_min;
	spa_t *spa = vd->vdev_spa;
	int error = 0;

	ta->trim_start_time = gethrtime();
	ta->trim_bytes_done = 0;

	for (range_seg_t *rs = zfs_btree_first(t, &idx); rs != NULL;
	rs = zfs_btree_next(t, &idx, &idx)) {
	uint64_t size = rs_get_end(rs, ta->trim_tree) - rs_get_start(rs,
	ta->trim_tree);

	if (extent_bytes_min && size < extent_bytes_min) {
	spa_iostats_trim_add(spa, ta->trim_type,
	0, 0, 1, size, 0, 0);
	continue;
	}

	/* Split range into legally-sized physical chunks */
	uint64_t writes_required = ((size - 1) / extent_bytes_max) + 1;

	for (uint64_t w = 0; w < writes_required; w++) {
	error = vdev_trim_range(ta, VDEV_LABEL_START_SIZE +
	rs_get_start(rs, ta->trim_tree) +
	(w *extent_bytes_max), MIN(size -
	(w * extent_bytes_max), extent_bytes_max));
	if (error != 0) {
	goto done;
	}
	}
	}

	done:
	/*
	* Make sure all TRIMs for this metaslab have completed before
	* returning. TRIM zios have lower priority over regular or syncing
	* zios, so all TRIM zios for this metaslab must complete before the
	* metaslab is re-enabled. Otherwise it's possible write zios to
	* this metaslab could cut ahead of still queued TRIM zios for this
	* metaslab causing corruption if the ranges overlap.
	*/
	mutex_enter(&vd->vdev_trim_io_lock);
	while (vd->vdev_trim_inflight[0] > 0) {
	cv_wait(&vd->vdev_trim_io_cv, &vd->vdev_trim_io_lock);
	}
	mutex_exit(&vd->vdev_trim_io_lock);

	return (error);
	}

	static void
	vdev_trim_xlate_last_rs_end(void arg, range_seg64_t physical_rs)
	{
	uint64_t last_rs_end = (uint64_t )arg;

	if (physical_rs->rs_end > *last_rs_end)
	*last_rs_end = physical_rs->rs_end;
	}

	static void
	vdev_trim_xlate_progress(void arg, range_seg64_t physical_rs)
	{
	vdev_t vd = (vdev_t )arg;

	uint64_t size = physical_rs->rs_end - physical_rs->rs_start;
	vd->vdev_trim_bytes_est += size;

	if (vd->vdev_trim_last_offset >= physical_rs->rs_end) {
	vd->vdev_trim_bytes_done += size;
	} else if (vd->vdev_trim_last_offset > physical_rs->rs_start &&
	vd->vdev_trim_last_offset <= physical_rs->rs_end) {
	vd->vdev_trim_bytes_done +=
	vd->vdev_trim_last_offset - physical_rs->rs_start;
	}
	}

	/*
	* Calculates the completion percentage of a manual TRIM.
	*/
	static void
	vdev_trim_calculate_progress(vdev_t *vd)
	{
	ASSERT(spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_READER) \|\|
	spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_WRITER));
	ASSERT(vd->vdev_leaf_zap != 0);

	vd->vdev_trim_bytes_est = 0;
	vd->vdev_trim_bytes_done = 0;

	for (uint64_t i = 0; i < vd->vdev_top->vdev_ms_count; i++) {
	metaslab_t *msp = vd->vdev_top->vdev_ms[i];
	mutex_enter(&msp->ms_lock);

	uint64_t ms_free = (msp->ms_size -
	metaslab_allocated_space(msp)) /
	vdev_get_ndisks(vd->vdev_top);

	/*
	* Convert the metaslab range to a physical range
	* on our vdev. We use this to determine if we are
	* in the middle of this metaslab range.
	*/
	range_seg64_t logical_rs, physical_rs, remain_rs;
	logical_rs.rs_start = msp->ms_start;
	logical_rs.rs_end = msp->ms_start + msp->ms_size;

	/* Metaslab space after this offset has not been trimmed. */
	vdev_xlate(vd, &logical_rs, &physical_rs, &remain_rs);
	if (vd->vdev_trim_last_offset <= physical_rs.rs_start) {
	vd->vdev_trim_bytes_est += ms_free;
	mutex_exit(&msp->ms_lock);
	continue;
	}

	/* Metaslab space before this offset has been trimmed */
	uint64_t last_rs_end = physical_rs.rs_end;
	if (!vdev_xlate_is_empty(&remain_rs)) {
	vdev_xlate_walk(vd, &remain_rs,
	vdev_trim_xlate_last_rs_end, &last_rs_end);
	}

	if (vd->vdev_trim_last_offset > last_rs_end) {
	vd->vdev_trim_bytes_done += ms_free;
	vd->vdev_trim_bytes_est += ms_free;
	mutex_exit(&msp->ms_lock);
	continue;
	}

	/*
	* If we get here, we're in the middle of trimming this
	* metaslab. Load it and walk the free tree for more
	* accurate progress estimation.
	*/
	VERIFY0(metaslab_load(msp));

	range_tree_t *rt = msp->ms_allocatable;
	zfs_btree_t *bt = &rt->rt_root;
	zfs_btree_index_t idx;
	for (range_seg_t *rs = zfs_btree_first(bt, &idx);
	rs != NULL; rs = zfs_btree_next(bt, &idx, &idx)) {
	logical_rs.rs_start = rs_get_start(rs, rt);
	logical_rs.rs_end = rs_get_end(rs, rt);

	vdev_xlate_walk(vd, &logical_rs,
	vdev_trim_xlate_progress, vd);
	}
	mutex_exit(&msp->ms_lock);
	}
	}

	/*
	* Load from disk the vdev's manual TRIM information. This includes the
	* state, progress, and options provided when initiating the manual TRIM.
	*/
	static int
	vdev_trim_load(vdev_t *vd)
	{
	int err = 0;
	ASSERT(spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_READER) \|\|
	spa_config_held(vd->vdev_spa, SCL_CONFIG, RW_WRITER));
	ASSERT(vd->vdev_leaf_zap != 0);

	if (vd->vdev_trim_state == VDEV_TRIM_ACTIVE \|\|
	vd->vdev_trim_state == VDEV_TRIM_SUSPENDED) {
	err = zap_lookup(vd->vdev_spa->spa_meta_objset,
	vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_LAST_OFFSET,
	sizeof (vd->vdev_trim_last_offset), 1,
	&vd->vdev_trim_last_offset);
	if (err == ENOENT) {
	vd->vdev_trim_last_offset = 0;
	err = 0;
	}

	if (err == 0) {
	err = zap_lookup(vd->vdev_spa->spa_meta_objset,
	vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_RATE,
	sizeof (vd->vdev_trim_rate), 1,
	&vd->vdev_trim_rate);
	if (err == ENOENT) {
	vd->vdev_trim_rate = 0;
	err = 0;
	}
	}

	if (err == 0) {
	err = zap_lookup(vd->vdev_spa->spa_meta_objset,
	vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_PARTIAL,
	sizeof (vd->vdev_trim_partial), 1,
	&vd->vdev_trim_partial);
	if (err == ENOENT) {
	vd->vdev_trim_partial = 0;
	err = 0;
	}
	}

	if (err == 0) {
	err = zap_lookup(vd->vdev_spa->spa_meta_objset,
	vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_SECURE,
	sizeof (vd->vdev_trim_secure), 1,
	&vd->vdev_trim_secure);
	if (err == ENOENT) {
	vd->vdev_trim_secure = 0;
	err = 0;
	}
	}
	}

	vdev_trim_calculate_progress(vd);

	return (err);
	}

	static void
	vdev_trim_xlate_range_add(void arg, range_seg64_t physical_rs)
	{
	trim_args_t *ta = arg;
	vdev_t *vd = ta->trim_vdev;

	/*
	* Only a manual trim will be traversing the vdev sequentially.
	* For an auto trim all valid ranges should be added.
	*/
	if (ta->trim_type == TRIM_TYPE_MANUAL) {

	/* Only add segments that we have not visited yet */
	if (physical_rs->rs_end <= vd->vdev_trim_last_offset)
	return;

	/* Pick up where we left off mid-range. */
	if (vd->vdev_trim_last_offset > physical_rs->rs_start) {
	ASSERT3U(physical_rs->rs_end, >,
	vd->vdev_trim_last_offset);
	physical_rs->rs_start = vd->vdev_trim_last_offset;
	}
	}

	ASSERT3U(physical_rs->rs_end, >, physical_rs->rs_start);

	range_tree_add(ta->trim_tree, physical_rs->rs_start,
	physical_rs->rs_end - physical_rs->rs_start);
	}

	/*
	* Convert the logical range into physical ranges and add them to the
	* range tree passed in the trim_args_t.
	*/
	static void
	vdev_trim_range_add(void *arg, uint64_t start, uint64_t size)
	{
	trim_args_t *ta = arg;
	vdev_t *vd = ta->trim_vdev;
	range_seg64_t logical_rs;
	logical_rs.rs_start = start;
	logical_rs.rs_end = start + size;

	/*
	* Every range to be trimmed must be part of ms_allocatable.
	* When ZFS_DEBUG_TRIM is set load the metaslab to verify this
	* is always the case.
	*/
	if (zfs_flags & ZFS_DEBUG_TRIM) {
	metaslab_t *msp = ta->trim_msp;
	VERIFY0(metaslab_load(msp));
	VERIFY3B(msp->ms_loaded, ==, B_TRUE);
	VERIFY(range_tree_contains(msp->ms_allocatable, start, size));
	}

	ASSERT(vd->vdev_ops->vdev_op_leaf);
	vdev_xlate_walk(vd, &logical_rs, vdev_trim_xlate_range_add, arg);
	}

	/*
	* Each manual TRIM thread is responsible for trimming the unallocated
	* space for each leaf vdev. This is accomplished by sequentially iterating
	* over its top-level metaslabs and issuing TRIM I/O for the space described
	* by its ms_allocatable. While a metaslab is undergoing trimming it is
	* not eligible for new allocations.
	*/
	static __attribute__((noreturn)) void
	vdev_trim_thread(void *arg)
	{
	vdev_t *vd = arg;
	spa_t *spa = vd->vdev_spa;
	trim_args_t ta;
	int error = 0;

	/*
	* The VDEV_LEAF_ZAP_TRIM_* entries may have been updated by
	* vdev_trim(). Wait for the updated values to be reflected
	* in the zap in order to start with the requested settings.
	*/
	txg_wait_synced(spa_get_dsl(vd->vdev_spa), 0);

	ASSERT(vdev_is_concrete(vd));
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);

	vd->vdev_trim_last_offset = 0;
	vd->vdev_trim_rate = 0;
	vd->vdev_trim_partial = 0;
	vd->vdev_trim_secure = 0;

	VERIFY0(vdev_trim_load(vd));

	ta.trim_vdev = vd;
	ta.trim_extent_bytes_max = zfs_trim_extent_bytes_max;
	ta.trim_extent_bytes_min = zfs_trim_extent_bytes_min;
	ta.trim_tree = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0);
	ta.trim_type = TRIM_TYPE_MANUAL;
	ta.trim_flags = 0;

	/*
	* When a secure TRIM has been requested infer that the intent
	* is that everything must be trimmed. Override the default
	* minimum TRIM size to prevent ranges from being skipped.
	*/
	if (vd->vdev_trim_secure) {
	ta.trim_flags \|= ZIO_TRIM_SECURE;
	ta.trim_extent_bytes_min = SPA_MINBLOCKSIZE;
	}

	uint64_t ms_count = 0;
	for (uint64_t i = 0; !vd->vdev_detached &&
	i < vd->vdev_top->vdev_ms_count; i++) {
	metaslab_t *msp = vd->vdev_top->vdev_ms[i];

	/*
	* If we've expanded the top-level vdev or it's our
	* first pass, calculate our progress.
	*/
	if (vd->vdev_top->vdev_ms_count != ms_count) {
	vdev_trim_calculate_progress(vd);
	ms_count = vd->vdev_top->vdev_ms_count;
	}

	spa_config_exit(spa, SCL_CONFIG, FTAG);
	metaslab_disable(msp);
	mutex_enter(&msp->ms_lock);
	VERIFY0(metaslab_load(msp));

	/*
	* If a partial TRIM was requested skip metaslabs which have
	* never been initialized and thus have never been written.
	*/
	if (msp->ms_sm == NULL && vd->vdev_trim_partial) {
	mutex_exit(&msp->ms_lock);
	metaslab_enable(msp, B_FALSE, B_FALSE);
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	vdev_trim_calculate_progress(vd);
	continue;
	}

	ta.trim_msp = msp;
	range_tree_walk(msp->ms_allocatable, vdev_trim_range_add, &ta);
	range_tree_vacate(msp->ms_trim, NULL, NULL);
	mutex_exit(&msp->ms_lock);

	error = vdev_trim_ranges(&ta);
	metaslab_enable(msp, B_TRUE, B_FALSE);
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);

	range_tree_vacate(ta.trim_tree, NULL, NULL);
	if (error != 0)
	break;
	}

	spa_config_exit(spa, SCL_CONFIG, FTAG);

	range_tree_destroy(ta.trim_tree);

	mutex_enter(&vd->vdev_trim_lock);
	if (!vd->vdev_trim_exit_wanted) {
	if (vdev_writeable(vd)) {
	vdev_trim_change_state(vd, VDEV_TRIM_COMPLETE,
	vd->vdev_trim_rate, vd->vdev_trim_partial,
	vd->vdev_trim_secure);
	} else if (vd->vdev_faulted) {
	vdev_trim_change_state(vd, VDEV_TRIM_CANCELED,
	vd->vdev_trim_rate, vd->vdev_trim_partial,
	vd->vdev_trim_secure);
	}
	}
	ASSERT(vd->vdev_trim_thread != NULL \|\| vd->vdev_trim_inflight[0] == 0);

	/*
	* Drop the vdev_trim_lock while we sync out the txg since it's
	* possible that a device might be trying to come online and must
	* check to see if it needs to restart a trim. That thread will be
	* holding the spa_config_lock which would prevent the txg_wait_synced
	* from completing.
	*/
	mutex_exit(&vd->vdev_trim_lock);
	txg_wait_synced(spa_get_dsl(spa), 0);
	mutex_enter(&vd->vdev_trim_lock);

	vd->vdev_trim_thread = NULL;
	cv_broadcast(&vd->vdev_trim_cv);
	mutex_exit(&vd->vdev_trim_lock);

	thread_exit();
	}

	/*
	* Initiates a manual TRIM for the vdev_t. Callers must hold vdev_trim_lock,
	* the vdev_t must be a leaf and cannot already be manually trimming.
	*/
	void
	vdev_trim(vdev_t *vd, uint64_t rate, boolean_t partial, boolean_t secure)
	{
	ASSERT(MUTEX_HELD(&vd->vdev_trim_lock));
	ASSERT(vd->vdev_ops->vdev_op_leaf);
	ASSERT(vdev_is_concrete(vd));
	ASSERT3P(vd->vdev_trim_thread, ==, NULL);
	ASSERT(!vd->vdev_detached);
	ASSERT(!vd->vdev_trim_exit_wanted);
	ASSERT(!vd->vdev_top->vdev_removing);
	ASSERT(!vd->vdev_rz_expanding);

	vdev_trim_change_state(vd, VDEV_TRIM_ACTIVE, rate, partial, secure);
	vd->vdev_trim_thread = thread_create(NULL, 0,
	vdev_trim_thread, vd, 0, &p0, TS_RUN, maxclsyspri);
	}

	/*
	* Wait for the trimming thread to be terminated (canceled or stopped).
	*/
	static void
	vdev_trim_stop_wait_impl(vdev_t *vd)
	{
	ASSERT(MUTEX_HELD(&vd->vdev_trim_lock));

	while (vd->vdev_trim_thread != NULL)
	cv_wait(&vd->vdev_trim_cv, &vd->vdev_trim_lock);

	ASSERT3P(vd->vdev_trim_thread, ==, NULL);
	vd->vdev_trim_exit_wanted = B_FALSE;
	}

	/*
	* Wait for vdev trim threads which were listed to cleanly exit.
	*/
	void
	vdev_trim_stop_wait(spa_t spa, list_t vd_list)
	{
	(void) spa;
	vdev_t *vd;

	- ASSERT(MUTEX_HELD(&spa_namespace_lock));
	+ ASSERT(MUTEX_HELD(&spa_namespace_lock) \|\|
	+ spa->spa_export_thread == curthread);

	while ((vd = list_remove_head(vd_list)) != NULL) {
	mutex_enter(&vd->vdev_trim_lock);
	vdev_trim_stop_wait_impl(vd);
	mutex_exit(&vd->vdev_trim_lock);
	}
	}

	/*
	* Stop trimming a device, with the resultant trimming state being tgt_state.
	* For blocking behavior pass NULL for vd_list. Otherwise, when a list_t is
	* provided the stopping vdev is inserted in to the list. Callers are then
	* required to call vdev_trim_stop_wait() to block for all the trim threads
	* to exit. The caller must hold vdev_trim_lock and must not be writing to
	* the spa config, as the trimming thread may try to enter the config as a
	* reader before exiting.
	*/
	void
	vdev_trim_stop(vdev_t vd, vdev_trim_state_t tgt_state, list_t vd_list)
	{
	ASSERT(!spa_config_held(vd->vdev_spa, SCL_CONFIG\|SCL_STATE, RW_WRITER));
	ASSERT(MUTEX_HELD(&vd->vdev_trim_lock));
	ASSERT(vd->vdev_ops->vdev_op_leaf);
	ASSERT(vdev_is_concrete(vd));

	/*
	* Allow cancel requests to proceed even if the trim thread has
	* stopped.
	*/
	if (vd->vdev_trim_thread == NULL && tgt_state != VDEV_TRIM_CANCELED)
	return;

	vdev_trim_change_state(vd, tgt_state, 0, 0, 0);
	vd->vdev_trim_exit_wanted = B_TRUE;

	if (vd_list == NULL) {
	vdev_trim_stop_wait_impl(vd);
	} else {
	- ASSERT(MUTEX_HELD(&spa_namespace_lock));
	+ ASSERT(MUTEX_HELD(&spa_namespace_lock) \|\|
	+ vd->vdev_spa->spa_export_thread == curthread);
	list_insert_tail(vd_list, vd);
	}
	}

	/*
	* Requests that all listed vdevs stop trimming.
	*/
	static void
	vdev_trim_stop_all_impl(vdev_t *vd, vdev_trim_state_t tgt_state,
	list_t *vd_list)
	{
	if (vd->vdev_ops->vdev_op_leaf && vdev_is_concrete(vd)) {
	mutex_enter(&vd->vdev_trim_lock);
	vdev_trim_stop(vd, tgt_state, vd_list);
	mutex_exit(&vd->vdev_trim_lock);
	return;
	}

	for (uint64_t i = 0; i < vd->vdev_children; i++) {
	vdev_trim_stop_all_impl(vd->vdev_child[i], tgt_state,
	vd_list);
	}
	}

	/*
	* Convenience function to stop trimming of a vdev tree and set all trim
	* thread pointers to NULL.
	*/
	void
	vdev_trim_stop_all(vdev_t *vd, vdev_trim_state_t tgt_state)
	{
	spa_t *spa = vd->vdev_spa;
	list_t vd_list;
	vdev_t *vd_l2cache;

	- ASSERT(MUTEX_HELD(&spa_namespace_lock));
	+ ASSERT(MUTEX_HELD(&spa_namespace_lock) \|\|
	+ spa->spa_export_thread == curthread);

	list_create(&vd_list, sizeof (vdev_t),
	offsetof(vdev_t, vdev_trim_node));

	vdev_trim_stop_all_impl(vd, tgt_state, &vd_list);

	/*
	* Iterate over cache devices and request stop trimming the
	* whole device in case we export the pool or remove the cache
	* device prematurely.
	*/
	for (int i = 0; i < spa->spa_l2cache.sav_count; i++) {
	vd_l2cache = spa->spa_l2cache.sav_vdevs[i];
	vdev_trim_stop_all_impl(vd_l2cache, tgt_state, &vd_list);
	}

	vdev_trim_stop_wait(spa, &vd_list);

	if (vd->vdev_spa->spa_sync_on) {
	/* Make sure that our state has been synced to disk */
	txg_wait_synced(spa_get_dsl(vd->vdev_spa), 0);
	}

	list_destroy(&vd_list);
	}

	/*
	* Conditionally restarts a manual TRIM given its on-disk state.
	*/
	void
	vdev_trim_restart(vdev_t *vd)
	{
	ASSERT(MUTEX_HELD(&spa_namespace_lock) \|\|
	vd->vdev_spa->spa_load_thread == curthread);
	ASSERT(!spa_config_held(vd->vdev_spa, SCL_ALL, RW_WRITER));

	if (vd->vdev_leaf_zap != 0) {
	mutex_enter(&vd->vdev_trim_lock);
	uint64_t trim_state = VDEV_TRIM_NONE;
	int err = zap_lookup(vd->vdev_spa->spa_meta_objset,
	vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_STATE,
	sizeof (trim_state), 1, &trim_state);
	ASSERT(err == 0 \|\| err == ENOENT);
	vd->vdev_trim_state = trim_state;

	uint64_t timestamp = 0;
	err = zap_lookup(vd->vdev_spa->spa_meta_objset,
	vd->vdev_leaf_zap, VDEV_LEAF_ZAP_TRIM_ACTION_TIME,
	sizeof (timestamp), 1, &timestamp);
	ASSERT(err == 0 \|\| err == ENOENT);
	vd->vdev_trim_action_time = timestamp;

	if ((vd->vdev_trim_state == VDEV_TRIM_SUSPENDED \|\|
	vd->vdev_offline) && !vd->vdev_top->vdev_rz_expanding) {
	/* load progress for reporting, but don't resume */
	VERIFY0(vdev_trim_load(vd));
	} else if (vd->vdev_trim_state == VDEV_TRIM_ACTIVE &&
	vdev_writeable(vd) && !vd->vdev_top->vdev_removing &&
	!vd->vdev_top->vdev_rz_expanding &&
	vd->vdev_trim_thread == NULL) {
	VERIFY0(vdev_trim_load(vd));
	vdev_trim(vd, vd->vdev_trim_rate,
	vd->vdev_trim_partial, vd->vdev_trim_secure);
	}

	mutex_exit(&vd->vdev_trim_lock);
	}

	for (uint64_t i = 0; i < vd->vdev_children; i++) {
	vdev_trim_restart(vd->vdev_child[i]);
	}
	}

	/*
	* Used by the automatic TRIM when ZFS_DEBUG_TRIM is set to verify that
	* every TRIM range is contained within ms_allocatable.
	*/
	static void
	vdev_trim_range_verify(void *arg, uint64_t start, uint64_t size)
	{
	trim_args_t *ta = arg;
	metaslab_t *msp = ta->trim_msp;

	VERIFY3B(msp->ms_loaded, ==, B_TRUE);
	VERIFY3U(msp->ms_disabled, >, 0);
	VERIFY(range_tree_contains(msp->ms_allocatable, start, size));
	}

	/*
	* Each automatic TRIM thread is responsible for managing the trimming of a
	* top-level vdev in the pool. No automatic TRIM state is maintained on-disk.
	*
	* N.B. This behavior is different from a manual TRIM where a thread
	* is created for each leaf vdev, instead of each top-level vdev.
	*/
	static __attribute__((noreturn)) void
	vdev_autotrim_thread(void *arg)
	{
	vdev_t *vd = arg;
	spa_t *spa = vd->vdev_spa;
	int shift = 0;

	mutex_enter(&vd->vdev_autotrim_lock);
	ASSERT3P(vd->vdev_top, ==, vd);
	ASSERT3P(vd->vdev_autotrim_thread, !=, NULL);
	mutex_exit(&vd->vdev_autotrim_lock);
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);

	while (!vdev_autotrim_should_stop(vd)) {
	int txgs_per_trim = MAX(zfs_trim_txg_batch, 1);
	uint64_t extent_bytes_max = zfs_trim_extent_bytes_max;
	uint64_t extent_bytes_min = zfs_trim_extent_bytes_min;

	/*
	* All of the metaslabs are divided in to groups of size
	* num_metaslabs / zfs_trim_txg_batch. Each of these groups
	* is composed of metaslabs which are spread evenly over the
	* device.
	*
	* For example, when zfs_trim_txg_batch = 32 (default) then
	* group 0 will contain metaslabs 0, 32, 64, ...;
	* group 1 will contain metaslabs 1, 33, 65, ...;
	* group 2 will contain metaslabs 2, 34, 66, ...; and so on.
	*
	* On each pass through the while() loop one of these groups
	* is selected. This is accomplished by using a shift value
	* to select the starting metaslab, then striding over the
	* metaslabs using the zfs_trim_txg_batch size. This is
	* done to accomplish two things.
	*
	* 1) By dividing the metaslabs in to groups, and making sure
	* that each group takes a minimum of one txg to process.
	* Then zfs_trim_txg_batch controls the minimum number of
	* txgs which must occur before a metaslab is revisited.
	*
	* 2) Selecting non-consecutive metaslabs distributes the
	* TRIM commands for a group evenly over the entire device.
	* This can be advantageous for certain types of devices.
	*/
	for (uint64_t i = shift % txgs_per_trim; i < vd->vdev_ms_count;
	i += txgs_per_trim) {
	metaslab_t *msp = vd->vdev_ms[i];
	range_tree_t *trim_tree;
	boolean_t issued_trim = B_FALSE;
	boolean_t wait_aborted = B_FALSE;

	spa_config_exit(spa, SCL_CONFIG, FTAG);
	metaslab_disable(msp);
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);

	mutex_enter(&msp->ms_lock);

	/*
	* Skip the metaslab when it has never been allocated
	* or when there are no recent frees to trim.
	*/
	if (msp->ms_sm == NULL \|\|
	range_tree_is_empty(msp->ms_trim)) {
	mutex_exit(&msp->ms_lock);
	metaslab_enable(msp, B_FALSE, B_FALSE);
	continue;
	}

	/*
	* Skip the metaslab when it has already been disabled.
	* This may happen when a manual TRIM or initialize
	* operation is running concurrently. In the case
	* of a manual TRIM, the ms_trim tree will have been
	* vacated. Only ranges added after the manual TRIM
	* disabled the metaslab will be included in the tree.
	* These will be processed when the automatic TRIM
	* next revisits this metaslab.
	*/
	if (msp->ms_disabled > 1) {
	mutex_exit(&msp->ms_lock);
	metaslab_enable(msp, B_FALSE, B_FALSE);
	continue;
	}

	/*
	* Allocate an empty range tree which is swapped in
	* for the existing ms_trim tree while it is processed.
	*/
	trim_tree = range_tree_create(NULL, RANGE_SEG64, NULL,
	0, 0);
	range_tree_swap(&msp->ms_trim, &trim_tree);
	ASSERT(range_tree_is_empty(msp->ms_trim));

	/*
	* There are two cases when constructing the per-vdev
	* trim trees for a metaslab. If the top-level vdev
	* has no children then it is also a leaf and should
	* be trimmed. Otherwise our children are the leaves
	* and a trim tree should be constructed for each.
	*/
	trim_args_t *tap;
	uint64_t children = vd->vdev_children;
	if (children == 0) {
	children = 1;
	tap = kmem_zalloc(sizeof (trim_args_t) *
	children, KM_SLEEP);
	tap[0].trim_vdev = vd;
	} else {
	tap = kmem_zalloc(sizeof (trim_args_t) *
	children, KM_SLEEP);

	for (uint64_t c = 0; c < children; c++) {
	tap[c].trim_vdev = vd->vdev_child[c];
	}
	}

	for (uint64_t c = 0; c < children; c++) {
	trim_args_t *ta = &tap[c];
	vdev_t *cvd = ta->trim_vdev;

	ta->trim_msp = msp;
	ta->trim_extent_bytes_max = extent_bytes_max;
	ta->trim_extent_bytes_min = extent_bytes_min;
	ta->trim_type = TRIM_TYPE_AUTO;
	ta->trim_flags = 0;

	if (cvd->vdev_detached \|\|
	!vdev_writeable(cvd) \|\|
	!cvd->vdev_has_trim \|\|
	cvd->vdev_trim_thread != NULL) {
	continue;
	}

	/*
	* When a device has an attached hot spare, or
	* is being replaced it will not be trimmed.
	* This is done to avoid adding additional
	* stress to a potentially unhealthy device,
	* and to minimize the required rebuild time.
	*/
	if (!cvd->vdev_ops->vdev_op_leaf)
	continue;

	ta->trim_tree = range_tree_create(NULL,
	RANGE_SEG64, NULL, 0, 0);
	range_tree_walk(trim_tree,
	vdev_trim_range_add, ta);
	}

	mutex_exit(&msp->ms_lock);
	spa_config_exit(spa, SCL_CONFIG, FTAG);

	/*
	* Issue the TRIM I/Os for all ranges covered by the
	* TRIM trees. These ranges are safe to TRIM because
	* no new allocations will be performed until the call
	* to metaslab_enabled() below.
	*/
	for (uint64_t c = 0; c < children; c++) {
	trim_args_t *ta = &tap[c];

	/*
	* Always yield to a manual TRIM if one has
	* been started for the child vdev.
	*/
	if (ta->trim_tree == NULL \|\|
	ta->trim_vdev->vdev_trim_thread != NULL) {
	continue;
	}

	/*
	* After this point metaslab_enable() must be
	* called with the sync flag set. This is done
	* here because vdev_trim_ranges() is allowed
	* to be interrupted (EINTR) before issuing all
	* of the required TRIM I/Os.
	*/
	issued_trim = B_TRUE;

	int error = vdev_trim_ranges(ta);
	if (error)
	break;
	}

	/*
	* Verify every range which was trimmed is still
	* contained within the ms_allocatable tree.
	*/
	if (zfs_flags & ZFS_DEBUG_TRIM) {
	mutex_enter(&msp->ms_lock);
	VERIFY0(metaslab_load(msp));
	VERIFY3P(tap[0].trim_msp, ==, msp);
	range_tree_walk(trim_tree,
	vdev_trim_range_verify, &tap[0]);
	mutex_exit(&msp->ms_lock);
	}

	range_tree_vacate(trim_tree, NULL, NULL);
	range_tree_destroy(trim_tree);

	/*
	* Wait for couples of kicks, to ensure the trim io is
	* synced. If the wait is aborted due to
	* vdev_autotrim_exit_wanted, we need to signal
	* metaslab_enable() to wait for sync.
	*/
	if (issued_trim) {
	wait_aborted = vdev_autotrim_wait_kick(vd,
	TXG_CONCURRENT_STATES + TXG_DEFER_SIZE);
	}

	metaslab_enable(msp, wait_aborted, B_FALSE);
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);

	for (uint64_t c = 0; c < children; c++) {
	trim_args_t *ta = &tap[c];

	if (ta->trim_tree == NULL)
	continue;

	range_tree_vacate(ta->trim_tree, NULL, NULL);
	range_tree_destroy(ta->trim_tree);
	}

	kmem_free(tap, sizeof (trim_args_t) * children);

	if (vdev_autotrim_should_stop(vd))
	break;
	}

	spa_config_exit(spa, SCL_CONFIG, FTAG);

	vdev_autotrim_wait_kick(vd, 1);

	shift++;
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);
	}

	for (uint64_t c = 0; c < vd->vdev_children; c++) {
	vdev_t *cvd = vd->vdev_child[c];
	mutex_enter(&cvd->vdev_trim_io_lock);

	while (cvd->vdev_trim_inflight[1] > 0) {
	cv_wait(&cvd->vdev_trim_io_cv,
	&cvd->vdev_trim_io_lock);
	}
	mutex_exit(&cvd->vdev_trim_io_lock);
	}

	spa_config_exit(spa, SCL_CONFIG, FTAG);

	/*
	* When exiting because the autotrim property was set to off, then
	* abandon any unprocessed ms_trim ranges to reclaim the memory.
	*/
	if (spa_get_autotrim(spa) == SPA_AUTOTRIM_OFF) {
	for (uint64_t i = 0; i < vd->vdev_ms_count; i++) {
	metaslab_t *msp = vd->vdev_ms[i];

	mutex_enter(&msp->ms_lock);
	range_tree_vacate(msp->ms_trim, NULL, NULL);
	mutex_exit(&msp->ms_lock);
	}
	}

	mutex_enter(&vd->vdev_autotrim_lock);
	ASSERT(vd->vdev_autotrim_thread != NULL);
	vd->vdev_autotrim_thread = NULL;
	cv_broadcast(&vd->vdev_autotrim_cv);
	mutex_exit(&vd->vdev_autotrim_lock);

	thread_exit();
	}

	/*
	* Starts an autotrim thread, if needed, for each top-level vdev which can be
	* trimmed. A top-level vdev which has been evacuated will never be trimmed.
	*/
	void
	vdev_autotrim(spa_t *spa)
	{
	vdev_t *root_vd = spa->spa_root_vdev;

	for (uint64_t i = 0; i < root_vd->vdev_children; i++) {
	vdev_t *tvd = root_vd->vdev_child[i];

	mutex_enter(&tvd->vdev_autotrim_lock);
	if (vdev_writeable(tvd) && !tvd->vdev_removing &&
	tvd->vdev_autotrim_thread == NULL &&
	!tvd->vdev_rz_expanding) {
	ASSERT3P(tvd->vdev_top, ==, tvd);

	tvd->vdev_autotrim_thread = thread_create(NULL, 0,
	vdev_autotrim_thread, tvd, 0, &p0, TS_RUN,
	maxclsyspri);
	ASSERT(tvd->vdev_autotrim_thread != NULL);
	}
	mutex_exit(&tvd->vdev_autotrim_lock);
	}
	}

	/*
	* Wait for the vdev_autotrim_thread associated with the passed top-level
	* vdev to be terminated (canceled or stopped).
	*/
	void
	vdev_autotrim_stop_wait(vdev_t *tvd)
	{
	mutex_enter(&tvd->vdev_autotrim_lock);
	if (tvd->vdev_autotrim_thread != NULL) {
	tvd->vdev_autotrim_exit_wanted = B_TRUE;
	cv_broadcast(&tvd->vdev_autotrim_kick_cv);
	cv_wait(&tvd->vdev_autotrim_cv,
	&tvd->vdev_autotrim_lock);

	ASSERT3P(tvd->vdev_autotrim_thread, ==, NULL);
	tvd->vdev_autotrim_exit_wanted = B_FALSE;
	}
	mutex_exit(&tvd->vdev_autotrim_lock);
	}

	void
	vdev_autotrim_kick(spa_t *spa)
	{
	ASSERT(spa_config_held(spa, SCL_CONFIG, RW_READER));

	vdev_t *root_vd = spa->spa_root_vdev;
	vdev_t *tvd;

	for (uint64_t i = 0; i < root_vd->vdev_children; i++) {
	tvd = root_vd->vdev_child[i];

	mutex_enter(&tvd->vdev_autotrim_lock);
	if (tvd->vdev_autotrim_thread != NULL)
	cv_broadcast(&tvd->vdev_autotrim_kick_cv);
	mutex_exit(&tvd->vdev_autotrim_lock);
	}
	}

	/*
	* Wait for all of the vdev_autotrim_thread associated with the pool to
	* be terminated (canceled or stopped).
	*/
	void
	vdev_autotrim_stop_all(spa_t *spa)
	{
	vdev_t *root_vd = spa->spa_root_vdev;

	for (uint64_t i = 0; i < root_vd->vdev_children; i++)
	vdev_autotrim_stop_wait(root_vd->vdev_child[i]);
	}

	/*
	* Conditionally restart all of the vdev_autotrim_thread's for the pool.
	*/
	void
	vdev_autotrim_restart(spa_t *spa)
	{
	ASSERT(MUTEX_HELD(&spa_namespace_lock) \|\|
	spa->spa_load_thread == curthread);
	if (spa->spa_autotrim)
	vdev_autotrim(spa);
	}

	static __attribute__((noreturn)) void
	vdev_trim_l2arc_thread(void *arg)
	{
	vdev_t *vd = arg;
	spa_t *spa = vd->vdev_spa;
	l2arc_dev_t *dev = l2arc_vdev_get(vd);
	trim_args_t ta = {0};
	range_seg64_t physical_rs;

	ASSERT(vdev_is_concrete(vd));
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER);

	vd->vdev_trim_last_offset = 0;
	vd->vdev_trim_rate = 0;
	vd->vdev_trim_partial = 0;
	vd->vdev_trim_secure = 0;

	ta.trim_vdev = vd;
	ta.trim_tree = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0);
	ta.trim_type = TRIM_TYPE_MANUAL;
	ta.trim_extent_bytes_max = zfs_trim_extent_bytes_max;
	ta.trim_extent_bytes_min = SPA_MINBLOCKSIZE;
	ta.trim_flags = 0;

	physical_rs.rs_start = vd->vdev_trim_bytes_done = 0;
	physical_rs.rs_end = vd->vdev_trim_bytes_est =
	vdev_get_min_asize(vd);

	range_tree_add(ta.trim_tree, physical_rs.rs_start,
	physical_rs.rs_end - physical_rs.rs_start);

	mutex_enter(&vd->vdev_trim_lock);
	vdev_trim_change_state(vd, VDEV_TRIM_ACTIVE, 0, 0, 0);
	mutex_exit(&vd->vdev_trim_lock);

	(void) vdev_trim_ranges(&ta);

	spa_config_exit(spa, SCL_CONFIG, FTAG);
	mutex_enter(&vd->vdev_trim_io_lock);
	while (vd->vdev_trim_inflight[TRIM_TYPE_MANUAL] > 0) {
	cv_wait(&vd->vdev_trim_io_cv, &vd->vdev_trim_io_lock);
	}
	mutex_exit(&vd->vdev_trim_io_lock);

	range_tree_vacate(ta.trim_tree, NULL, NULL);
	range_tree_destroy(ta.trim_tree);

	mutex_enter(&vd->vdev_trim_lock);
	if (!vd->vdev_trim_exit_wanted && vdev_writeable(vd)) {
	vdev_trim_change_state(vd, VDEV_TRIM_COMPLETE,
	vd->vdev_trim_rate, vd->vdev_trim_partial,
	vd->vdev_trim_secure);
	}
	ASSERT(vd->vdev_trim_thread != NULL \|\|
	vd->vdev_trim_inflight[TRIM_TYPE_MANUAL] == 0);

	/*
	* Drop the vdev_trim_lock while we sync out the txg since it's
	* possible that a device might be trying to come online and
	* must check to see if it needs to restart a trim. That thread
	* will be holding the spa_config_lock which would prevent the
	* txg_wait_synced from completing. Same strategy as in
	* vdev_trim_thread().
	*/
	mutex_exit(&vd->vdev_trim_lock);
	txg_wait_synced(spa_get_dsl(vd->vdev_spa), 0);
	mutex_enter(&vd->vdev_trim_lock);

	/*
	* Update the header of the cache device here, before
	* broadcasting vdev_trim_cv which may lead to the removal
	* of the device. The same applies for setting l2ad_trim_all to
	* false.
	*/
	spa_config_enter(vd->vdev_spa, SCL_L2ARC, vd,
	RW_READER);
	memset(dev->l2ad_dev_hdr, 0, dev->l2ad_dev_hdr_asize);
	l2arc_dev_hdr_update(dev);
	spa_config_exit(vd->vdev_spa, SCL_L2ARC, vd);

	vd->vdev_trim_thread = NULL;
	if (vd->vdev_trim_state == VDEV_TRIM_COMPLETE)
	dev->l2ad_trim_all = B_FALSE;

	cv_broadcast(&vd->vdev_trim_cv);
	mutex_exit(&vd->vdev_trim_lock);

	thread_exit();
	}

	/*
	* Punches out TRIM threads for the L2ARC devices in a spa and assigns them
	* to vd->vdev_trim_thread variable. This facilitates the management of
	* trimming the whole cache device using TRIM_TYPE_MANUAL upon addition
	* to a pool or pool creation or when the header of the device is invalid.
	*/
	void
	vdev_trim_l2arc(spa_t *spa)
	{
	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	/*
	* Locate the spa's l2arc devices and kick off TRIM threads.
	*/
	for (int i = 0; i < spa->spa_l2cache.sav_count; i++) {
	vdev_t *vd = spa->spa_l2cache.sav_vdevs[i];
	l2arc_dev_t *dev = l2arc_vdev_get(vd);

	if (dev == NULL \|\| !dev->l2ad_trim_all) {
	/*
	* Don't attempt TRIM if the vdev is UNAVAIL or if the
	* cache device was not marked for whole device TRIM
	* (ie l2arc_trim_ahead = 0, or the L2ARC device header
	* is valid with trim_state = VDEV_TRIM_COMPLETE and
	* l2ad_log_entries > 0).
	*/
	continue;
	}

	mutex_enter(&vd->vdev_trim_lock);
	ASSERT(vd->vdev_ops->vdev_op_leaf);
	ASSERT(vdev_is_concrete(vd));
	ASSERT3P(vd->vdev_trim_thread, ==, NULL);
	ASSERT(!vd->vdev_detached);
	ASSERT(!vd->vdev_trim_exit_wanted);
	ASSERT(!vd->vdev_top->vdev_removing);
	vdev_trim_change_state(vd, VDEV_TRIM_ACTIVE, 0, 0, 0);
	vd->vdev_trim_thread = thread_create(NULL, 0,
	vdev_trim_l2arc_thread, vd, 0, &p0, TS_RUN, maxclsyspri);
	mutex_exit(&vd->vdev_trim_lock);
	}
	}

	/*
	* A wrapper which calls vdev_trim_ranges(). It is intended to be called
	* on leaf vdevs.
	*/
	int
	vdev_trim_simple(vdev_t *vd, uint64_t start, uint64_t size)
	{
	trim_args_t ta = {0};
	range_seg64_t physical_rs;
	int error;
	physical_rs.rs_start = start;
	physical_rs.rs_end = start + size;

	ASSERT(vdev_is_concrete(vd));
	ASSERT(vd->vdev_ops->vdev_op_leaf);
	ASSERT(!vd->vdev_detached);
	ASSERT(!vd->vdev_top->vdev_removing);
	ASSERT(!vd->vdev_top->vdev_rz_expanding);

	ta.trim_vdev = vd;
	ta.trim_tree = range_tree_create(NULL, RANGE_SEG64, NULL, 0, 0);
	ta.trim_type = TRIM_TYPE_SIMPLE;
	ta.trim_extent_bytes_max = zfs_trim_extent_bytes_max;
	ta.trim_extent_bytes_min = SPA_MINBLOCKSIZE;
	ta.trim_flags = 0;

	ASSERT3U(physical_rs.rs_end, >=, physical_rs.rs_start);

	if (physical_rs.rs_end > physical_rs.rs_start) {
	range_tree_add(ta.trim_tree, physical_rs.rs_start,
	physical_rs.rs_end - physical_rs.rs_start);
	} else {
	ASSERT3U(physical_rs.rs_end, ==, physical_rs.rs_start);
	}

	error = vdev_trim_ranges(&ta);

	mutex_enter(&vd->vdev_trim_io_lock);
	while (vd->vdev_trim_inflight[TRIM_TYPE_SIMPLE] > 0) {
	cv_wait(&vd->vdev_trim_io_cv, &vd->vdev_trim_io_lock);
	}
	mutex_exit(&vd->vdev_trim_io_lock);

	range_tree_vacate(ta.trim_tree, NULL, NULL);
	range_tree_destroy(ta.trim_tree);

	return (error);
	}

	EXPORT_SYMBOL(vdev_trim);
	EXPORT_SYMBOL(vdev_trim_stop);
	EXPORT_SYMBOL(vdev_trim_stop_all);
	EXPORT_SYMBOL(vdev_trim_stop_wait);
	EXPORT_SYMBOL(vdev_trim_restart);
	EXPORT_SYMBOL(vdev_autotrim);
	EXPORT_SYMBOL(vdev_autotrim_stop_all);
	EXPORT_SYMBOL(vdev_autotrim_stop_wait);
	EXPORT_SYMBOL(vdev_autotrim_restart);
	EXPORT_SYMBOL(vdev_trim_l2arc);
	EXPORT_SYMBOL(vdev_trim_simple);

	ZFS_MODULE_PARAM(zfs_trim, zfs_trim_, extent_bytes_max, UINT, ZMOD_RW,
	"Max size of TRIM commands, larger will be split");

	ZFS_MODULE_PARAM(zfs_trim, zfs_trim_, extent_bytes_min, UINT, ZMOD_RW,
	"Min size of TRIM commands, smaller will be skipped");

	ZFS_MODULE_PARAM(zfs_trim, zfs_trim_, metaslab_skip, UINT, ZMOD_RW,
	"Skip metaslabs which have never been initialized");

	ZFS_MODULE_PARAM(zfs_trim, zfs_trim_, txg_batch, UINT, ZMOD_RW,
	"Min number of txgs to aggregate frees before issuing TRIM");

	ZFS_MODULE_PARAM(zfs_trim, zfs_trim_, queue_limit, UINT, ZMOD_RW,
	"Max queued TRIMs outstanding per leaf vdev");
	diff --git a/sys/contrib/openzfs/module/zfs/zap.c b/sys/contrib/openzfs/module/zfs/zap.c
	index 1b6b16fc6662..03b76ea1b7bf 100644
	--- a/sys/contrib/openzfs/module/zfs/zap.c
	+++ b/sys/contrib/openzfs/module/zfs/zap.c
	@@ -1,1692 +1,1709 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2018 by Delphix. All rights reserved.
	* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
	* Copyright 2023 Alexander Stetsenko <alex.stetsenko@gmail.com>
	* Copyright (c) 2023, Klara Inc.
	*/

	/*
	* This file contains the top half of the zfs directory structure
	* implementation. The bottom half is in zap_leaf.c.
	*
	* The zdir is an extendable hash data structure. There is a table of
	* pointers to buckets (zap_t->zd_data->zd_leafs). The buckets are
	* each a constant size and hold a variable number of directory entries.
	* The buckets (aka "leaf nodes") are implemented in zap_leaf.c.
	*
	* The pointer table holds a power of 2 number of pointers.
	* (1<<zap_t->zd_data->zd_phys->zd_prefix_len). The bucket pointed to
	* by the pointer at index i in the table holds entries whose hash value
	* has a zd_prefix_len - bit prefix
	*/

	#include <sys/spa.h>
	#include <sys/dmu.h>
	#include <sys/dnode.h>
	#include <sys/zfs_context.h>
	#include <sys/zfs_znode.h>
	#include <sys/fs/zfs.h>
	#include <sys/zap.h>
	#include <sys/zap_impl.h>
	#include <sys/zap_leaf.h>

	/*
	* If zap_iterate_prefetch is set, we will prefetch the entire ZAP object
	* (all leaf blocks) when we start iterating over it.
	*
	* For zap_cursor_init(), the callers all intend to iterate through all the
	* entries. There are a few cases where an error (typically i/o error) could
	* cause it to bail out early.
	*
	* For zap_cursor_init_serialized(), there are callers that do the iteration
	* outside of ZFS. Typically they would iterate over everything, but we
	* don't have control of that. E.g. zfs_ioc_snapshot_list_next(),
	* zcp_snapshots_iter(), and other iterators over things in the MOS - these
	* are called by /sbin/zfs and channel programs. The other example is
	* zfs_readdir() which iterates over directory entries for the getdents()
	* syscall. /sbin/ls iterates to the end (unless it receives a signal), but
	* userland doesn't have to.
	*
	* Given that the ZAP entries aren't returned in a specific order, the only
	* legitimate use cases for partial iteration would be:
	*
	* 1. Pagination: e.g. you only want to display 100 entries at a time, so you
	* get the first 100 and then wait for the user to hit "next page", which
	* they may never do).
	*
	* 2. You want to know if there are more than X entries, without relying on
	* the zfs-specific implementation of the directory's st_size (which is
	* the number of entries).
	*/
	static int zap_iterate_prefetch = B_TRUE;

	/*
	* Enable ZAP shrinking. When enabled, empty sibling leaf blocks will be
	* collapsed into a single block.
	*/
	int zap_shrink_enabled = B_TRUE;

	int fzap_default_block_shift = 14; /* 16k blocksize */

	static uint64_t zap_allocate_blocks(zap_t *zap, int nblocks);
	static int zap_shrink(zap_name_t zn, zap_leaf_t l, dmu_tx_t *tx);

	void
	fzap_byteswap(void *vbuf, size_t size)
	{
	uint64_t block_type = (uint64_t )vbuf;

	if (block_type == ZBT_LEAF \|\| block_type == BSWAP_64(ZBT_LEAF))
	zap_leaf_byteswap(vbuf, size);
	else {
	/* it's a ptrtbl block */
	byteswap_uint64_array(vbuf, size);
	}
	}

	void
	fzap_upgrade(zap_t zap, dmu_tx_t tx, zap_flags_t flags)
	{
	ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
	zap->zap_ismicro = FALSE;

	zap->zap_dbu.dbu_evict_func_sync = zap_evict_sync;
	zap->zap_dbu.dbu_evict_func_async = NULL;

	mutex_init(&zap->zap_f.zap_num_entries_mtx, 0, MUTEX_DEFAULT, 0);
	zap->zap_f.zap_block_shift = highbit64(zap->zap_dbuf->db_size) - 1;

	zap_phys_t *zp = zap_f_phys(zap);
	/*
	* explicitly zero it since it might be coming from an
	* initialized microzap
	*/
	memset(zap->zap_dbuf->db_data, 0, zap->zap_dbuf->db_size);
	zp->zap_block_type = ZBT_HEADER;
	zp->zap_magic = ZAP_MAGIC;

	zp->zap_ptrtbl.zt_shift = ZAP_EMBEDDED_PTRTBL_SHIFT(zap);

	zp->zap_freeblk = 2; /* block 1 will be the first leaf */
	zp->zap_num_leafs = 1;
	zp->zap_num_entries = 0;
	zp->zap_salt = zap->zap_salt;
	zp->zap_normflags = zap->zap_normflags;
	zp->zap_flags = flags;

	/* block 1 will be the first leaf */
	for (int i = 0; i < (1<<zp->zap_ptrtbl.zt_shift); i++)
	ZAP_EMBEDDED_PTRTBL_ENT(zap, i) = 1;

	/*
	* set up block 1 - the first leaf
	*/
	dmu_buf_t *db;
	VERIFY0(dmu_buf_hold_by_dnode(zap->zap_dnode,
	1<<FZAP_BLOCK_SHIFT(zap), FTAG, &db, DMU_READ_NO_PREFETCH));
	dmu_buf_will_dirty(db, tx);

	zap_leaf_t *l = kmem_zalloc(sizeof (zap_leaf_t), KM_SLEEP);
	l->l_dbuf = db;

	zap_leaf_init(l, zp->zap_normflags != 0);

	kmem_free(l, sizeof (zap_leaf_t));
	dmu_buf_rele(db, FTAG);
	}

	static int
	zap_tryupgradedir(zap_t zap, dmu_tx_t tx)
	{
	if (RW_WRITE_HELD(&zap->zap_rwlock))
	return (1);
	if (rw_tryupgrade(&zap->zap_rwlock)) {
	dmu_buf_will_dirty(zap->zap_dbuf, tx);
	return (1);
	}
	return (0);
	}

	/*
	* Generic routines for dealing with the pointer & cookie tables.
	*/

	static int
	zap_table_grow(zap_t zap, zap_table_phys_t tbl,
	void (transfer_func)(const uint64_t src, uint64_t *dst, int n),
	dmu_tx_t *tx)
	{
	uint64_t newblk;
	int bs = FZAP_BLOCK_SHIFT(zap);
	int hepb = 1<<(bs-4);
	/* hepb = half the number of entries in a block */

	ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
	ASSERT(tbl->zt_blk != 0);
	ASSERT(tbl->zt_numblks > 0);

	if (tbl->zt_nextblk != 0) {
	newblk = tbl->zt_nextblk;
	} else {
	newblk = zap_allocate_blocks(zap, tbl->zt_numblks * 2);
	tbl->zt_nextblk = newblk;
	ASSERT0(tbl->zt_blks_copied);
	dmu_prefetch_by_dnode(zap->zap_dnode, 0,
	tbl->zt_blk << bs, tbl->zt_numblks << bs,
	ZIO_PRIORITY_SYNC_READ);
	}

	/*
	* Copy the ptrtbl from the old to new location.
	*/

	uint64_t b = tbl->zt_blks_copied;
	dmu_buf_t *db_old;
	int err = dmu_buf_hold_by_dnode(zap->zap_dnode,
	(tbl->zt_blk + b) << bs, FTAG, &db_old, DMU_READ_NO_PREFETCH);
	if (err != 0)
	return (err);

	/* first half of entries in old[b] go to new[2b+0] /
	dmu_buf_t *db_new;
	VERIFY0(dmu_buf_hold_by_dnode(zap->zap_dnode,
	(newblk + 2*b+0) << bs, FTAG, &db_new, DMU_READ_NO_PREFETCH));
	dmu_buf_will_dirty(db_new, tx);
	transfer_func(db_old->db_data, db_new->db_data, hepb);
	dmu_buf_rele(db_new, FTAG);

	/* second half of entries in old[b] go to new[2b+1] /
	VERIFY0(dmu_buf_hold_by_dnode(zap->zap_dnode,
	(newblk + 2*b+1) << bs, FTAG, &db_new, DMU_READ_NO_PREFETCH));
	dmu_buf_will_dirty(db_new, tx);
	transfer_func((uint64_t *)db_old->db_data + hepb,
	db_new->db_data, hepb);
	dmu_buf_rele(db_new, FTAG);

	dmu_buf_rele(db_old, FTAG);

	tbl->zt_blks_copied++;

	dprintf("copied block %llu of %llu\n",
	(u_longlong_t)tbl->zt_blks_copied,
	(u_longlong_t)tbl->zt_numblks);

	if (tbl->zt_blks_copied == tbl->zt_numblks) {
	(void) dmu_free_range(zap->zap_objset, zap->zap_object,
	tbl->zt_blk << bs, tbl->zt_numblks << bs, tx);

	tbl->zt_blk = newblk;
	tbl->zt_numblks *= 2;
	tbl->zt_shift++;
	tbl->zt_nextblk = 0;
	tbl->zt_blks_copied = 0;

	dprintf("finished; numblocks now %llu (%uk entries)\n",
	(u_longlong_t)tbl->zt_numblks, 1<<(tbl->zt_shift-10));
	}

	return (0);
	}

	static int
	zap_table_store(zap_t zap, zap_table_phys_t tbl, uint64_t idx, uint64_t val,
	dmu_tx_t *tx)
	{
	int bs = FZAP_BLOCK_SHIFT(zap);

	ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));
	ASSERT(tbl->zt_blk != 0);

	dprintf("storing %llx at index %llx\n", (u_longlong_t)val,
	(u_longlong_t)idx);

	uint64_t blk = idx >> (bs-3);
	uint64_t off = idx & ((1<<(bs-3))-1);

	dmu_buf_t *db;
	int err = dmu_buf_hold_by_dnode(zap->zap_dnode,
	(tbl->zt_blk + blk) << bs, FTAG, &db, DMU_READ_NO_PREFETCH);
	if (err != 0)
	return (err);
	dmu_buf_will_dirty(db, tx);

	if (tbl->zt_nextblk != 0) {
	uint64_t idx2 = idx * 2;
	uint64_t blk2 = idx2 >> (bs-3);
	uint64_t off2 = idx2 & ((1<<(bs-3))-1);
	dmu_buf_t *db2;

	err = dmu_buf_hold_by_dnode(zap->zap_dnode,
	(tbl->zt_nextblk + blk2) << bs, FTAG, &db2,
	DMU_READ_NO_PREFETCH);
	if (err != 0) {
	dmu_buf_rele(db, FTAG);
	return (err);
	}
	dmu_buf_will_dirty(db2, tx);
	((uint64_t *)db2->db_data)[off2] = val;
	((uint64_t *)db2->db_data)[off2+1] = val;
	dmu_buf_rele(db2, FTAG);
	}

	((uint64_t *)db->db_data)[off] = val;
	dmu_buf_rele(db, FTAG);

	return (0);
	}

	static int
	zap_table_load(zap_t zap, zap_table_phys_t tbl, uint64_t idx, uint64_t *valp)
	{
	int bs = FZAP_BLOCK_SHIFT(zap);

	ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));

	uint64_t blk = idx >> (bs-3);
	uint64_t off = idx & ((1<<(bs-3))-1);

	dmu_buf_t *db;
	int err = dmu_buf_hold_by_dnode(zap->zap_dnode,
	(tbl->zt_blk + blk) << bs, FTAG, &db, DMU_READ_NO_PREFETCH);
	if (err != 0)
	return (err);
	valp = ((uint64_t )db->db_data)[off];
	dmu_buf_rele(db, FTAG);

	if (tbl->zt_nextblk != 0) {
	/*
	* read the nextblk for the sake of i/o error checking,
	* so that zap_table_load() will catch errors for
	* zap_table_store.
	*/
	blk = (idx*2) >> (bs-3);

	err = dmu_buf_hold_by_dnode(zap->zap_dnode,
	(tbl->zt_nextblk + blk) << bs, FTAG, &db,
	DMU_READ_NO_PREFETCH);
	if (err == 0)
	dmu_buf_rele(db, FTAG);
	}
	return (err);
	}

	/*
	* Routines for growing the ptrtbl.
	*/

	static void
	zap_ptrtbl_transfer(const uint64_t src, uint64_t dst, int n)
	{
	for (int i = 0; i < n; i++) {
	uint64_t lb = src[i];
	dst[2 * i + 0] = lb;
	dst[2 * i + 1] = lb;
	}
	}

	static int
	zap_grow_ptrtbl(zap_t zap, dmu_tx_t tx)
	{
	/*
	* The pointer table should never use more hash bits than we
	* have (otherwise we'd be using useless zero bits to index it).
	* If we are within 2 bits of running out, stop growing, since
	* this is already an aberrant condition.
	*/
	if (zap_f_phys(zap)->zap_ptrtbl.zt_shift >= zap_hashbits(zap) - 2)
	return (SET_ERROR(ENOSPC));

	if (zap_f_phys(zap)->zap_ptrtbl.zt_numblks == 0) {
	/*
	* We are outgrowing the "embedded" ptrtbl (the one
	* stored in the header block). Give it its own entire
	* block, which will double the size of the ptrtbl.
	*/
	ASSERT3U(zap_f_phys(zap)->zap_ptrtbl.zt_shift, ==,
	ZAP_EMBEDDED_PTRTBL_SHIFT(zap));
	ASSERT0(zap_f_phys(zap)->zap_ptrtbl.zt_blk);

	uint64_t newblk = zap_allocate_blocks(zap, 1);
	dmu_buf_t *db_new;
	int err = dmu_buf_hold_by_dnode(zap->zap_dnode,
	newblk << FZAP_BLOCK_SHIFT(zap), FTAG, &db_new,
	DMU_READ_NO_PREFETCH);
	if (err != 0)
	return (err);
	dmu_buf_will_dirty(db_new, tx);
	zap_ptrtbl_transfer(&ZAP_EMBEDDED_PTRTBL_ENT(zap, 0),
	db_new->db_data, 1 << ZAP_EMBEDDED_PTRTBL_SHIFT(zap));
	dmu_buf_rele(db_new, FTAG);

	zap_f_phys(zap)->zap_ptrtbl.zt_blk = newblk;
	zap_f_phys(zap)->zap_ptrtbl.zt_numblks = 1;
	zap_f_phys(zap)->zap_ptrtbl.zt_shift++;

	ASSERT3U(1ULL << zap_f_phys(zap)->zap_ptrtbl.zt_shift, ==,
	zap_f_phys(zap)->zap_ptrtbl.zt_numblks <<
	(FZAP_BLOCK_SHIFT(zap)-3));

	return (0);
	} else {
	return (zap_table_grow(zap, &zap_f_phys(zap)->zap_ptrtbl,
	zap_ptrtbl_transfer, tx));
	}
	}

	static void
	zap_increment_num_entries(zap_t zap, int delta, dmu_tx_t tx)
	{
	dmu_buf_will_dirty(zap->zap_dbuf, tx);
	mutex_enter(&zap->zap_f.zap_num_entries_mtx);
	ASSERT(delta > 0 \|\| zap_f_phys(zap)->zap_num_entries >= -delta);
	zap_f_phys(zap)->zap_num_entries += delta;
	mutex_exit(&zap->zap_f.zap_num_entries_mtx);
	}

	static uint64_t
	zap_allocate_blocks(zap_t *zap, int nblocks)
	{
	ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
	uint64_t newblk = zap_f_phys(zap)->zap_freeblk;
	zap_f_phys(zap)->zap_freeblk += nblocks;
	return (newblk);
	}

	static void
	zap_leaf_evict_sync(void *dbu)
	{
	zap_leaf_t *l = dbu;

	rw_destroy(&l->l_rwlock);
	kmem_free(l, sizeof (zap_leaf_t));
	}

	static zap_leaf_t *
	zap_create_leaf(zap_t zap, dmu_tx_t tx)
	{
	- zap_leaf_t *l = kmem_zalloc(sizeof (zap_leaf_t), KM_SLEEP);
	-
	ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));

	- rw_init(&l->l_rwlock, NULL, RW_NOLOCKDEP, NULL);
	- rw_enter(&l->l_rwlock, RW_WRITER);
	- l->l_blkid = zap_allocate_blocks(zap, 1);
	- l->l_dbuf = NULL;
	+ uint64_t blkid = zap_allocate_blocks(zap, 1);
	+ dmu_buf_t *db = NULL;

	VERIFY0(dmu_buf_hold_by_dnode(zap->zap_dnode,
	- l->l_blkid << FZAP_BLOCK_SHIFT(zap), NULL, &l->l_dbuf,
	+ blkid << FZAP_BLOCK_SHIFT(zap), NULL, &db,
	DMU_READ_NO_PREFETCH));
	- dmu_buf_init_user(&l->l_dbu, zap_leaf_evict_sync, NULL, &l->l_dbuf);
	- VERIFY3P(NULL, ==, dmu_buf_set_user(l->l_dbuf, &l->l_dbu));
	+
	+ /*
	+ * Create the leaf structure and stash it on the dbuf. If zap was
	+ * recent shrunk or truncated, the dbuf might have been sitting in the
	+ * cache waiting to be evicted, and so still have the old leaf attached
	+ * to it. If so, just reuse it.
	+ */
	+ zap_leaf_t *l = dmu_buf_get_user(db);
	+ if (l == NULL) {
	+ l = kmem_zalloc(sizeof (zap_leaf_t), KM_SLEEP);
	+ l->l_blkid = blkid;
	+ l->l_dbuf = db;
	+ rw_init(&l->l_rwlock, NULL, RW_NOLOCKDEP, NULL);
	+ dmu_buf_init_user(&l->l_dbu, zap_leaf_evict_sync, NULL,
	+ &l->l_dbuf);
	+ dmu_buf_set_user(l->l_dbuf, &l->l_dbu);
	+ } else {
	+ ASSERT3U(l->l_blkid, ==, blkid);
	+ ASSERT3P(l->l_dbuf, ==, db);
	+ }
	+
	+ rw_enter(&l->l_rwlock, RW_WRITER);
	dmu_buf_will_dirty(l->l_dbuf, tx);

	zap_leaf_init(l, zap->zap_normflags != 0);

	zap_f_phys(zap)->zap_num_leafs++;

	return (l);
	}

	int
	fzap_count(zap_t zap, uint64_t count)
	{
	ASSERT(!zap->zap_ismicro);
	mutex_enter(&zap->zap_f.zap_num_entries_mtx); /* unnecessary */
	*count = zap_f_phys(zap)->zap_num_entries;
	mutex_exit(&zap->zap_f.zap_num_entries_mtx);
	return (0);
	}

	/*
	* Routines for obtaining zap_leaf_t's
	*/

	void
	zap_put_leaf(zap_leaf_t *l)
	{
	rw_exit(&l->l_rwlock);
	dmu_buf_rele(l->l_dbuf, NULL);
	}

	static zap_leaf_t *
	zap_open_leaf(uint64_t blkid, dmu_buf_t *db)
	{
	ASSERT(blkid != 0);

	zap_leaf_t *l = kmem_zalloc(sizeof (zap_leaf_t), KM_SLEEP);
	rw_init(&l->l_rwlock, NULL, RW_DEFAULT, NULL);
	rw_enter(&l->l_rwlock, RW_WRITER);
	l->l_blkid = blkid;
	l->l_bs = highbit64(db->db_size) - 1;
	l->l_dbuf = db;

	dmu_buf_init_user(&l->l_dbu, zap_leaf_evict_sync, NULL, &l->l_dbuf);
	zap_leaf_t *winner = dmu_buf_set_user(db, &l->l_dbu);

	rw_exit(&l->l_rwlock);
	if (winner != NULL) {
	/* someone else set it first */
	zap_leaf_evict_sync(&l->l_dbu);
	l = winner;
	}

	/*
	* lhr_pad was previously used for the next leaf in the leaf
	* chain. There should be no chained leafs (as we have removed
	* support for them).
	*/
	ASSERT0(zap_leaf_phys(l)->l_hdr.lh_pad1);

	/*
	* There should be more hash entries than there can be
	* chunks to put in the hash table
	*/
	ASSERT3U(ZAP_LEAF_HASH_NUMENTRIES(l), >, ZAP_LEAF_NUMCHUNKS(l) / 3);

	/* The chunks should begin at the end of the hash table */
	ASSERT3P(&ZAP_LEAF_CHUNK(l, 0), ==, (zap_leaf_chunk_t *)
	&zap_leaf_phys(l)->l_hash[ZAP_LEAF_HASH_NUMENTRIES(l)]);

	/* The chunks should end at the end of the block */
	ASSERT3U((uintptr_t)&ZAP_LEAF_CHUNK(l, ZAP_LEAF_NUMCHUNKS(l)) -
	(uintptr_t)zap_leaf_phys(l), ==, l->l_dbuf->db_size);

	return (l);
	}

	static int
	zap_get_leaf_byblk(zap_t zap, uint64_t blkid, dmu_tx_t tx, krw_t lt,
	zap_leaf_t **lp)
	{
	dmu_buf_t *db;

	ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));

	/*
	* If system crashed just after dmu_free_long_range in zfs_rmnode, we
	* would be left with an empty xattr dir in delete queue. blkid=0
	* would be passed in when doing zfs_purgedir. If that's the case we
	* should just return immediately. The underlying objects should
	* already be freed, so this should be perfectly fine.
	*/
	if (blkid == 0)
	return (SET_ERROR(ENOENT));

	int bs = FZAP_BLOCK_SHIFT(zap);
	int err = dmu_buf_hold_by_dnode(zap->zap_dnode,
	blkid << bs, NULL, &db, DMU_READ_NO_PREFETCH);
	if (err != 0)
	return (err);

	ASSERT3U(db->db_object, ==, zap->zap_object);
	ASSERT3U(db->db_offset, ==, blkid << bs);
	ASSERT3U(db->db_size, ==, 1 << bs);
	ASSERT(blkid != 0);

	zap_leaf_t *l = dmu_buf_get_user(db);

	if (l == NULL)
	l = zap_open_leaf(blkid, db);

	rw_enter(&l->l_rwlock, lt);
	/*
	* Must lock before dirtying, otherwise zap_leaf_phys(l) could change,
	* causing ASSERT below to fail.
	*/
	if (lt == RW_WRITER)
	dmu_buf_will_dirty(db, tx);
	ASSERT3U(l->l_blkid, ==, blkid);
	ASSERT3P(l->l_dbuf, ==, db);
	ASSERT3U(zap_leaf_phys(l)->l_hdr.lh_block_type, ==, ZBT_LEAF);
	ASSERT3U(zap_leaf_phys(l)->l_hdr.lh_magic, ==, ZAP_LEAF_MAGIC);

	*lp = l;
	return (0);
	}

	static int
	zap_idx_to_blk(zap_t zap, uint64_t idx, uint64_t valp)
	{
	ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));

	if (zap_f_phys(zap)->zap_ptrtbl.zt_numblks == 0) {
	ASSERT3U(idx, <,
	(1ULL << zap_f_phys(zap)->zap_ptrtbl.zt_shift));
	*valp = ZAP_EMBEDDED_PTRTBL_ENT(zap, idx);
	return (0);
	} else {
	return (zap_table_load(zap, &zap_f_phys(zap)->zap_ptrtbl,
	idx, valp));
	}
	}

	static int
	zap_set_idx_to_blk(zap_t zap, uint64_t idx, uint64_t blk, dmu_tx_t tx)
	{
	ASSERT(tx != NULL);
	ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));

	if (zap_f_phys(zap)->zap_ptrtbl.zt_blk == 0) {
	ZAP_EMBEDDED_PTRTBL_ENT(zap, idx) = blk;
	return (0);
	} else {
	return (zap_table_store(zap, &zap_f_phys(zap)->zap_ptrtbl,
	idx, blk, tx));
	}
	}

	static int
	zap_set_idx_range_to_blk(zap_t *zap, uint64_t idx, uint64_t nptrs, uint64_t blk,
	dmu_tx_t *tx)
	{
	int bs = FZAP_BLOCK_SHIFT(zap);
	int epb = bs >> 3; /* entries per block */
	int err = 0;

	ASSERT(tx != NULL);
	ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));

	/*
	* Check for i/o errors
	*/
	for (int i = 0; i < nptrs; i += epb) {
	uint64_t blk;
	err = zap_idx_to_blk(zap, idx + i, &blk);
	if (err != 0) {
	return (err);
	}
	}

	for (int i = 0; i < nptrs; i++) {
	err = zap_set_idx_to_blk(zap, idx + i, blk, tx);
	ASSERT0(err); /* we checked for i/o errors above */
	if (err != 0)
	break;
	}

	return (err);
	}

	#define ZAP_PREFIX_HASH(pref, pref_len) ((pref) << (64 - (pref_len)))

	/*
	* Each leaf has single range of entries (block pointers) in the ZAP ptrtbl.
	* If two leaves are siblings, their ranges are adjecent and contain the same
	* number of entries. In order to find out if a leaf has a sibling, we need to
	* check the range corresponding to the sibling leaf. There is no need to check
	* all entries in the range, we only need to check the frist and the last one.
	*/
	static uint64_t
	check_sibling_ptrtbl_range(zap_t *zap, uint64_t prefix, uint64_t prefix_len)
	{
	ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));

	uint64_t h = ZAP_PREFIX_HASH(prefix, prefix_len);
	uint64_t idx = ZAP_HASH_IDX(h, zap_f_phys(zap)->zap_ptrtbl.zt_shift);
	uint64_t pref_diff = zap_f_phys(zap)->zap_ptrtbl.zt_shift - prefix_len;
	uint64_t nptrs = (1 << pref_diff);
	uint64_t first;
	uint64_t last;

	ASSERT3U(idx+nptrs, <=, (1UL << zap_f_phys(zap)->zap_ptrtbl.zt_shift));

	if (zap_idx_to_blk(zap, idx, &first) != 0)
	return (0);

	if (zap_idx_to_blk(zap, idx + nptrs - 1, &last) != 0)
	return (0);

	if (first != last)
	return (0);
	return (first);
	}

	static int
	zap_deref_leaf(zap_t zap, uint64_t h, dmu_tx_t tx, krw_t lt, zap_leaf_t **lp)
	{
	uint64_t blk;

	ASSERT(zap->zap_dbuf == NULL \|\|
	zap_f_phys(zap) == zap->zap_dbuf->db_data);

	/* Reality check for corrupt zap objects (leaf or header). */
	if ((zap_f_phys(zap)->zap_block_type != ZBT_LEAF &&
	zap_f_phys(zap)->zap_block_type != ZBT_HEADER) \|\|
	zap_f_phys(zap)->zap_magic != ZAP_MAGIC) {
	return (SET_ERROR(EIO));
	}

	uint64_t idx = ZAP_HASH_IDX(h, zap_f_phys(zap)->zap_ptrtbl.zt_shift);
	int err = zap_idx_to_blk(zap, idx, &blk);
	if (err != 0)
	return (err);
	err = zap_get_leaf_byblk(zap, blk, tx, lt, lp);

	ASSERT(err \|\|
	ZAP_HASH_IDX(h, zap_leaf_phys(*lp)->l_hdr.lh_prefix_len) ==
	zap_leaf_phys(*lp)->l_hdr.lh_prefix);
	return (err);
	}

	static int
	zap_expand_leaf(zap_name_t zn, zap_leaf_t l,
	const void tag, dmu_tx_t tx, zap_leaf_t **lp)
	{
	zap_t *zap = zn->zn_zap;
	uint64_t hash = zn->zn_hash;
	int err;
	int old_prefix_len = zap_leaf_phys(l)->l_hdr.lh_prefix_len;

	ASSERT3U(old_prefix_len, <=, zap_f_phys(zap)->zap_ptrtbl.zt_shift);
	ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));

	ASSERT3U(ZAP_HASH_IDX(hash, old_prefix_len), ==,
	zap_leaf_phys(l)->l_hdr.lh_prefix);

	if (zap_tryupgradedir(zap, tx) == 0 \|\|
	old_prefix_len == zap_f_phys(zap)->zap_ptrtbl.zt_shift) {
	/* We failed to upgrade, or need to grow the pointer table */
	objset_t *os = zap->zap_objset;
	uint64_t object = zap->zap_object;

	zap_put_leaf(l);
	+ *lp = l = NULL;
	zap_unlockdir(zap, tag);
	err = zap_lockdir(os, object, tx, RW_WRITER,
	FALSE, FALSE, tag, &zn->zn_zap);
	zap = zn->zn_zap;
	if (err != 0)
	return (err);
	ASSERT(!zap->zap_ismicro);

	while (old_prefix_len ==
	zap_f_phys(zap)->zap_ptrtbl.zt_shift) {
	err = zap_grow_ptrtbl(zap, tx);
	if (err != 0)
	return (err);
	}

	err = zap_deref_leaf(zap, hash, tx, RW_WRITER, &l);
	if (err != 0)
	return (err);

	if (zap_leaf_phys(l)->l_hdr.lh_prefix_len != old_prefix_len) {
	/* it split while our locks were down */
	*lp = l;
	return (0);
	}
	}
	ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
	ASSERT3U(old_prefix_len, <, zap_f_phys(zap)->zap_ptrtbl.zt_shift);
	ASSERT3U(ZAP_HASH_IDX(hash, old_prefix_len), ==,
	zap_leaf_phys(l)->l_hdr.lh_prefix);

	int prefix_diff = zap_f_phys(zap)->zap_ptrtbl.zt_shift -
	(old_prefix_len + 1);
	uint64_t sibling =
	(ZAP_HASH_IDX(hash, old_prefix_len + 1) \| 1) << prefix_diff;

	/* check for i/o errors before doing zap_leaf_split */
	for (int i = 0; i < (1ULL << prefix_diff); i++) {
	uint64_t blk;
	err = zap_idx_to_blk(zap, sibling + i, &blk);
	if (err != 0)
	return (err);
	ASSERT3U(blk, ==, l->l_blkid);
	}

	zap_leaf_t *nl = zap_create_leaf(zap, tx);
	zap_leaf_split(l, nl, zap->zap_normflags != 0);

	/* set sibling pointers */
	for (int i = 0; i < (1ULL << prefix_diff); i++) {
	err = zap_set_idx_to_blk(zap, sibling + i, nl->l_blkid, tx);
	ASSERT0(err); /* we checked for i/o errors above */
	}

	ASSERT3U(zap_leaf_phys(l)->l_hdr.lh_prefix_len, >, 0);

	if (hash & (1ULL << (64 - zap_leaf_phys(l)->l_hdr.lh_prefix_len))) {
	/* we want the sibling */
	zap_put_leaf(l);
	*lp = nl;
	} else {
	zap_put_leaf(nl);
	*lp = l;
	}

	return (0);
	}

	static void
	zap_put_leaf_maybe_grow_ptrtbl(zap_name_t zn, zap_leaf_t l,
	const void tag, dmu_tx_t tx)
	{
	zap_t *zap = zn->zn_zap;
	int shift = zap_f_phys(zap)->zap_ptrtbl.zt_shift;
	int leaffull = (zap_leaf_phys(l)->l_hdr.lh_prefix_len == shift &&
	zap_leaf_phys(l)->l_hdr.lh_nfree < ZAP_LEAF_LOW_WATER);

	zap_put_leaf(l);

	if (leaffull \|\| zap_f_phys(zap)->zap_ptrtbl.zt_nextblk) {
	/*
	* We are in the middle of growing the pointer table, or
	* this leaf will soon make us grow it.
	*/
	if (zap_tryupgradedir(zap, tx) == 0) {
	objset_t *os = zap->zap_objset;
	uint64_t zapobj = zap->zap_object;

	zap_unlockdir(zap, tag);
	int err = zap_lockdir(os, zapobj, tx,
	RW_WRITER, FALSE, FALSE, tag, &zn->zn_zap);
	zap = zn->zn_zap;
	if (err != 0)
	return;
	}

	/* could have finished growing while our locks were down */
	if (zap_f_phys(zap)->zap_ptrtbl.zt_shift == shift)
	(void) zap_grow_ptrtbl(zap, tx);
	}
	}

	static int
	fzap_checkname(zap_name_t *zn)
	{
	if (zn->zn_key_orig_numints * zn->zn_key_intlen > ZAP_MAXNAMELEN)
	return (SET_ERROR(ENAMETOOLONG));
	return (0);
	}

	static int
	fzap_checksize(uint64_t integer_size, uint64_t num_integers)
	{
	/* Only integer sizes supported by C */
	switch (integer_size) {
	case 1:
	case 2:
	case 4:
	case 8:
	break;
	default:
	return (SET_ERROR(EINVAL));
	}

	if (integer_size * num_integers > ZAP_MAXVALUELEN)
	return (SET_ERROR(E2BIG));

	return (0);
	}

	static int
	fzap_check(zap_name_t *zn, uint64_t integer_size, uint64_t num_integers)
	{
	int err = fzap_checkname(zn);
	if (err != 0)
	return (err);
	return (fzap_checksize(integer_size, num_integers));
	}

	/*
	* Routines for manipulating attributes.
	*/
	int
	fzap_lookup(zap_name_t *zn,
	uint64_t integer_size, uint64_t num_integers, void *buf,
	char realname, int rn_len, boolean_t ncp)
	{
	zap_leaf_t *l;
	zap_entry_handle_t zeh;

	int err = fzap_checkname(zn);
	if (err != 0)
	return (err);

	err = zap_deref_leaf(zn->zn_zap, zn->zn_hash, NULL, RW_READER, &l);
	if (err != 0)
	return (err);
	err = zap_leaf_lookup(l, zn, &zeh);
	if (err == 0) {
	if ((err = fzap_checksize(integer_size, num_integers)) != 0) {
	zap_put_leaf(l);
	return (err);
	}

	err = zap_entry_read(&zeh, integer_size, num_integers, buf);
	(void) zap_entry_read_name(zn->zn_zap, &zeh, rn_len, realname);
	if (ncp) {
	*ncp = zap_entry_normalization_conflict(&zeh,
	zn, NULL, zn->zn_zap);
	}
	}

	zap_put_leaf(l);
	return (err);
	}

	int
	fzap_add_cd(zap_name_t *zn,
	uint64_t integer_size, uint64_t num_integers,
	const void val, uint32_t cd, const void tag, dmu_tx_t *tx)
	{
	zap_leaf_t *l;
	int err;
	zap_entry_handle_t zeh;
	zap_t *zap = zn->zn_zap;

	ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));
	ASSERT(!zap->zap_ismicro);
	ASSERT(fzap_check(zn, integer_size, num_integers) == 0);

	err = zap_deref_leaf(zap, zn->zn_hash, tx, RW_WRITER, &l);
	if (err != 0)
	return (err);
	retry:
	err = zap_leaf_lookup(l, zn, &zeh);
	if (err == 0) {
	err = SET_ERROR(EEXIST);
	goto out;
	}
	if (err != ENOENT)
	goto out;

	err = zap_entry_create(l, zn, cd,
	integer_size, num_integers, val, &zeh);

	if (err == 0) {
	zap_increment_num_entries(zap, 1, tx);
	} else if (err == EAGAIN) {
	err = zap_expand_leaf(zn, l, tag, tx, &l);
	zap = zn->zn_zap; /* zap_expand_leaf() may change zap */
	- if (err == 0) {
	+ if (err == 0)
	goto retry;
	- } else if (err == ENOSPC) {
	- /*
	- * If we failed to expand the leaf, then bailout
	- * as there is no point trying
	- * zap_put_leaf_maybe_grow_ptrtbl().
	- */
	- return (err);
	- }
	}

	out:
	- if (zap != NULL)
	- zap_put_leaf_maybe_grow_ptrtbl(zn, l, tag, tx);
	+ if (l != NULL) {
	+ if (err == ENOSPC)
	+ zap_put_leaf(l);
	+ else
	+ zap_put_leaf_maybe_grow_ptrtbl(zn, l, tag, tx);
	+ }
	return (err);
	}

	int
	fzap_add(zap_name_t *zn,
	uint64_t integer_size, uint64_t num_integers,
	const void val, const void tag, dmu_tx_t *tx)
	{
	int err = fzap_check(zn, integer_size, num_integers);
	if (err != 0)
	return (err);

	return (fzap_add_cd(zn, integer_size, num_integers,
	val, ZAP_NEED_CD, tag, tx));
	}

	int
	fzap_update(zap_name_t *zn,
	int integer_size, uint64_t num_integers, const void *val,
	const void tag, dmu_tx_t tx)
	{
	zap_leaf_t *l;
	int err;
	boolean_t create;
	zap_entry_handle_t zeh;
	zap_t *zap = zn->zn_zap;

	ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));
	err = fzap_check(zn, integer_size, num_integers);
	if (err != 0)
	return (err);

	err = zap_deref_leaf(zap, zn->zn_hash, tx, RW_WRITER, &l);
	if (err != 0)
	return (err);
	retry:
	err = zap_leaf_lookup(l, zn, &zeh);
	create = (err == ENOENT);
	ASSERT(err == 0 \|\| err == ENOENT);

	if (create) {
	err = zap_entry_create(l, zn, ZAP_NEED_CD,
	integer_size, num_integers, val, &zeh);
	if (err == 0)
	zap_increment_num_entries(zap, 1, tx);
	} else {
	err = zap_entry_update(&zeh, integer_size, num_integers, val);
	}

	if (err == EAGAIN) {
	err = zap_expand_leaf(zn, l, tag, tx, &l);
	zap = zn->zn_zap; /* zap_expand_leaf() may change zap */
	if (err == 0)
	goto retry;
	}

	- if (zap != NULL)
	- zap_put_leaf_maybe_grow_ptrtbl(zn, l, tag, tx);
	+ if (l != NULL) {
	+ if (err == ENOSPC)
	+ zap_put_leaf(l);
	+ else
	+ zap_put_leaf_maybe_grow_ptrtbl(zn, l, tag, tx);
	+ }
	return (err);
	}

	int
	fzap_length(zap_name_t *zn,
	uint64_t integer_size, uint64_t num_integers)
	{
	zap_leaf_t *l;
	int err;
	zap_entry_handle_t zeh;

	err = zap_deref_leaf(zn->zn_zap, zn->zn_hash, NULL, RW_READER, &l);
	if (err != 0)
	return (err);
	err = zap_leaf_lookup(l, zn, &zeh);
	if (err != 0)
	goto out;

	if (integer_size != NULL)
	*integer_size = zeh.zeh_integer_size;
	if (num_integers != NULL)
	*num_integers = zeh.zeh_num_integers;
	out:
	zap_put_leaf(l);
	return (err);
	}

	int
	fzap_remove(zap_name_t zn, dmu_tx_t tx)
	{
	zap_leaf_t *l;
	int err;
	zap_entry_handle_t zeh;

	err = zap_deref_leaf(zn->zn_zap, zn->zn_hash, tx, RW_WRITER, &l);
	if (err != 0)
	return (err);
	err = zap_leaf_lookup(l, zn, &zeh);
	if (err == 0) {
	zap_entry_remove(&zeh);
	zap_increment_num_entries(zn->zn_zap, -1, tx);

	if (zap_leaf_phys(l)->l_hdr.lh_nentries == 0 &&
	zap_shrink_enabled)
	return (zap_shrink(zn, l, tx));
	}
	zap_put_leaf(l);
	return (err);
	}

	void
	fzap_prefetch(zap_name_t *zn)
	{
	uint64_t blk;
	zap_t *zap = zn->zn_zap;

	uint64_t idx = ZAP_HASH_IDX(zn->zn_hash,
	zap_f_phys(zap)->zap_ptrtbl.zt_shift);
	if (zap_idx_to_blk(zap, idx, &blk) != 0)
	return;
	int bs = FZAP_BLOCK_SHIFT(zap);
	dmu_prefetch_by_dnode(zap->zap_dnode, 0, blk << bs, 1 << bs,
	ZIO_PRIORITY_SYNC_READ);
	}

	/*
	* Helper functions for consumers.
	*/

	uint64_t
	zap_create_link(objset_t *os, dmu_object_type_t ot, uint64_t parent_obj,
	const char name, dmu_tx_t tx)
	{
	return (zap_create_link_dnsize(os, ot, parent_obj, name, 0, tx));
	}

	uint64_t
	zap_create_link_dnsize(objset_t *os, dmu_object_type_t ot, uint64_t parent_obj,
	const char name, int dnodesize, dmu_tx_t tx)
	{
	uint64_t new_obj;

	new_obj = zap_create_dnsize(os, ot, DMU_OT_NONE, 0, dnodesize, tx);
	VERIFY(new_obj != 0);
	VERIFY0(zap_add(os, parent_obj, name, sizeof (uint64_t), 1, &new_obj,
	tx));

	return (new_obj);
	}

	int
	zap_value_search(objset_t *os, uint64_t zapobj, uint64_t value, uint64_t mask,
	char *name)
	{
	zap_cursor_t zc;
	int err;

	if (mask == 0)
	mask = -1ULL;

	zap_attribute_t za = kmem_alloc(sizeof (za), KM_SLEEP);
	for (zap_cursor_init(&zc, os, zapobj);
	(err = zap_cursor_retrieve(&zc, za)) == 0;
	zap_cursor_advance(&zc)) {
	if ((za->za_first_integer & mask) == (value & mask)) {
	(void) strlcpy(name, za->za_name, MAXNAMELEN);
	break;
	}
	}
	zap_cursor_fini(&zc);
	kmem_free(za, sizeof (*za));
	return (err);
	}

	int
	zap_join(objset_t os, uint64_t fromobj, uint64_t intoobj, dmu_tx_t tx)
	{
	zap_cursor_t zc;
	int err = 0;

	zap_attribute_t za = kmem_alloc(sizeof (za), KM_SLEEP);
	for (zap_cursor_init(&zc, os, fromobj);
	zap_cursor_retrieve(&zc, za) == 0;
	(void) zap_cursor_advance(&zc)) {
	if (za->za_integer_length != 8 \|\| za->za_num_integers != 1) {
	err = SET_ERROR(EINVAL);
	break;
	}
	err = zap_add(os, intoobj, za->za_name,
	8, 1, &za->za_first_integer, tx);
	if (err != 0)
	break;
	}
	zap_cursor_fini(&zc);
	kmem_free(za, sizeof (*za));
	return (err);
	}

	int
	zap_join_key(objset_t *os, uint64_t fromobj, uint64_t intoobj,
	uint64_t value, dmu_tx_t *tx)
	{
	zap_cursor_t zc;
	int err = 0;

	zap_attribute_t za = kmem_alloc(sizeof (za), KM_SLEEP);
	for (zap_cursor_init(&zc, os, fromobj);
	zap_cursor_retrieve(&zc, za) == 0;
	(void) zap_cursor_advance(&zc)) {
	if (za->za_integer_length != 8 \|\| za->za_num_integers != 1) {
	err = SET_ERROR(EINVAL);
	break;
	}
	err = zap_add(os, intoobj, za->za_name,
	8, 1, &value, tx);
	if (err != 0)
	break;
	}
	zap_cursor_fini(&zc);
	kmem_free(za, sizeof (*za));
	return (err);
	}

	int
	zap_join_increment(objset_t *os, uint64_t fromobj, uint64_t intoobj,
	dmu_tx_t *tx)
	{
	zap_cursor_t zc;
	int err = 0;

	zap_attribute_t za = kmem_alloc(sizeof (za), KM_SLEEP);
	for (zap_cursor_init(&zc, os, fromobj);
	zap_cursor_retrieve(&zc, za) == 0;
	(void) zap_cursor_advance(&zc)) {
	uint64_t delta = 0;

	if (za->za_integer_length != 8 \|\| za->za_num_integers != 1) {
	err = SET_ERROR(EINVAL);
	break;
	}

	err = zap_lookup(os, intoobj, za->za_name, 8, 1, &delta);
	if (err != 0 && err != ENOENT)
	break;
	delta += za->za_first_integer;
	err = zap_update(os, intoobj, za->za_name, 8, 1, &delta, tx);
	if (err != 0)
	break;
	}
	zap_cursor_fini(&zc);
	kmem_free(za, sizeof (*za));
	return (err);
	}

	int
	zap_add_int(objset_t os, uint64_t obj, uint64_t value, dmu_tx_t tx)
	{
	char name[20];

	(void) snprintf(name, sizeof (name), "%llx", (longlong_t)value);
	return (zap_add(os, obj, name, 8, 1, &value, tx));
	}

	int
	zap_remove_int(objset_t os, uint64_t obj, uint64_t value, dmu_tx_t tx)
	{
	char name[20];

	(void) snprintf(name, sizeof (name), "%llx", (longlong_t)value);
	return (zap_remove(os, obj, name, tx));
	}

	int
	zap_lookup_int(objset_t *os, uint64_t obj, uint64_t value)
	{
	char name[20];

	(void) snprintf(name, sizeof (name), "%llx", (longlong_t)value);
	return (zap_lookup(os, obj, name, 8, 1, &value));
	}

	int
	zap_add_int_key(objset_t *os, uint64_t obj,
	uint64_t key, uint64_t value, dmu_tx_t *tx)
	{
	char name[20];

	(void) snprintf(name, sizeof (name), "%llx", (longlong_t)key);
	return (zap_add(os, obj, name, 8, 1, &value, tx));
	}

	int
	zap_update_int_key(objset_t *os, uint64_t obj,
	uint64_t key, uint64_t value, dmu_tx_t *tx)
	{
	char name[20];

	(void) snprintf(name, sizeof (name), "%llx", (longlong_t)key);
	return (zap_update(os, obj, name, 8, 1, &value, tx));
	}

	int
	zap_lookup_int_key(objset_t os, uint64_t obj, uint64_t key, uint64_t valuep)
	{
	char name[20];

	(void) snprintf(name, sizeof (name), "%llx", (longlong_t)key);
	return (zap_lookup(os, obj, name, 8, 1, valuep));
	}

	int
	zap_increment(objset_t os, uint64_t obj, const char name, int64_t delta,
	dmu_tx_t *tx)
	{
	uint64_t value = 0;

	if (delta == 0)
	return (0);

	int err = zap_lookup(os, obj, name, 8, 1, &value);
	if (err != 0 && err != ENOENT)
	return (err);
	value += delta;
	if (value == 0)
	err = zap_remove(os, obj, name, tx);
	else
	err = zap_update(os, obj, name, 8, 1, &value, tx);
	return (err);
	}

	int
	zap_increment_int(objset_t *os, uint64_t obj, uint64_t key, int64_t delta,
	dmu_tx_t *tx)
	{
	char name[20];

	(void) snprintf(name, sizeof (name), "%llx", (longlong_t)key);
	return (zap_increment(os, obj, name, delta, tx));
	}

	/*
	* Routines for iterating over the attributes.
	*/

	int
	fzap_cursor_retrieve(zap_t zap, zap_cursor_t zc, zap_attribute_t *za)
	{
	int err = ENOENT;
	zap_entry_handle_t zeh;
	zap_leaf_t *l;

	/* retrieve the next entry at or after zc_hash/zc_cd */
	/* if no entry, return ENOENT */

	/*
	* If we are reading from the beginning, we're almost certain to
	* iterate over the entire ZAP object. If there are multiple leaf
	* blocks (freeblk > 2), prefetch the whole object (up to
	* dmu_prefetch_max bytes), so that we read the leaf blocks
	* concurrently. (Unless noprefetch was requested via
	* zap_cursor_init_noprefetch()).
	*/
	if (zc->zc_hash == 0 && zap_iterate_prefetch &&
	zc->zc_prefetch && zap_f_phys(zap)->zap_freeblk > 2) {
	dmu_prefetch_by_dnode(zap->zap_dnode, 0, 0,
	zap_f_phys(zap)->zap_freeblk << FZAP_BLOCK_SHIFT(zap),
	ZIO_PRIORITY_ASYNC_READ);
	}

	if (zc->zc_leaf) {
	rw_enter(&zc->zc_leaf->l_rwlock, RW_READER);

	/*
	* The leaf was either shrunk or split.
	*/
	if ((zap_leaf_phys(zc->zc_leaf)->l_hdr.lh_block_type == 0) \|\|
	(ZAP_HASH_IDX(zc->zc_hash,
	zap_leaf_phys(zc->zc_leaf)->l_hdr.lh_prefix_len) !=
	zap_leaf_phys(zc->zc_leaf)->l_hdr.lh_prefix)) {
	zap_put_leaf(zc->zc_leaf);
	zc->zc_leaf = NULL;
	}
	}

	again:
	if (zc->zc_leaf == NULL) {
	err = zap_deref_leaf(zap, zc->zc_hash, NULL, RW_READER,
	&zc->zc_leaf);
	if (err != 0)
	return (err);
	}
	l = zc->zc_leaf;

	err = zap_leaf_lookup_closest(l, zc->zc_hash, zc->zc_cd, &zeh);

	if (err == ENOENT) {
	if (zap_leaf_phys(l)->l_hdr.lh_prefix_len == 0) {
	zc->zc_hash = -1ULL;
	zc->zc_cd = 0;
	} else {
	uint64_t nocare = (1ULL <<
	(64 - zap_leaf_phys(l)->l_hdr.lh_prefix_len)) - 1;

	zc->zc_hash = (zc->zc_hash & ~nocare) + nocare + 1;
	zc->zc_cd = 0;

	if (zc->zc_hash == 0) {
	zc->zc_hash = -1ULL;
	} else {
	zap_put_leaf(zc->zc_leaf);
	zc->zc_leaf = NULL;
	goto again;
	}
	}
	}

	if (err == 0) {
	zc->zc_hash = zeh.zeh_hash;
	zc->zc_cd = zeh.zeh_cd;
	za->za_integer_length = zeh.zeh_integer_size;
	za->za_num_integers = zeh.zeh_num_integers;
	if (zeh.zeh_num_integers == 0) {
	za->za_first_integer = 0;
	} else {
	err = zap_entry_read(&zeh, 8, 1, &za->za_first_integer);
	ASSERT(err == 0 \|\| err == EOVERFLOW);
	}
	err = zap_entry_read_name(zap, &zeh,
	sizeof (za->za_name), za->za_name);
	ASSERT(err == 0);

	za->za_normalization_conflict =
	zap_entry_normalization_conflict(&zeh,
	NULL, za->za_name, zap);
	}
	rw_exit(&zc->zc_leaf->l_rwlock);
	return (err);
	}

	static void
	zap_stats_ptrtbl(zap_t zap, uint64_t tbl, int len, zap_stats_t *zs)
	{
	uint64_t lastblk = 0;

	/*
	* NB: if a leaf has more pointers than an entire ptrtbl block
	* can hold, then it'll be accounted for more than once, since
	* we won't have lastblk.
	*/
	for (int i = 0; i < len; i++) {
	zap_leaf_t *l;

	if (tbl[i] == lastblk)
	continue;
	lastblk = tbl[i];

	int err = zap_get_leaf_byblk(zap, tbl[i], NULL, RW_READER, &l);
	if (err == 0) {
	zap_leaf_stats(zap, l, zs);
	zap_put_leaf(l);
	}
	}
	}

	void
	fzap_get_stats(zap_t zap, zap_stats_t zs)
	{
	int bs = FZAP_BLOCK_SHIFT(zap);
	zs->zs_blocksize = 1ULL << bs;

	/*
	* Set zap_phys_t fields
	*/
	zs->zs_num_leafs = zap_f_phys(zap)->zap_num_leafs;
	zs->zs_num_entries = zap_f_phys(zap)->zap_num_entries;
	zs->zs_num_blocks = zap_f_phys(zap)->zap_freeblk;
	zs->zs_block_type = zap_f_phys(zap)->zap_block_type;
	zs->zs_magic = zap_f_phys(zap)->zap_magic;
	zs->zs_salt = zap_f_phys(zap)->zap_salt;

	/*
	* Set zap_ptrtbl fields
	*/
	zs->zs_ptrtbl_len = 1ULL << zap_f_phys(zap)->zap_ptrtbl.zt_shift;
	zs->zs_ptrtbl_nextblk = zap_f_phys(zap)->zap_ptrtbl.zt_nextblk;
	zs->zs_ptrtbl_blks_copied =
	zap_f_phys(zap)->zap_ptrtbl.zt_blks_copied;
	zs->zs_ptrtbl_zt_blk = zap_f_phys(zap)->zap_ptrtbl.zt_blk;
	zs->zs_ptrtbl_zt_numblks = zap_f_phys(zap)->zap_ptrtbl.zt_numblks;
	zs->zs_ptrtbl_zt_shift = zap_f_phys(zap)->zap_ptrtbl.zt_shift;

	if (zap_f_phys(zap)->zap_ptrtbl.zt_numblks == 0) {
	/* the ptrtbl is entirely in the header block. */
	zap_stats_ptrtbl(zap, &ZAP_EMBEDDED_PTRTBL_ENT(zap, 0),
	1 << ZAP_EMBEDDED_PTRTBL_SHIFT(zap), zs);
	} else {
	dmu_prefetch_by_dnode(zap->zap_dnode, 0,
	zap_f_phys(zap)->zap_ptrtbl.zt_blk << bs,
	zap_f_phys(zap)->zap_ptrtbl.zt_numblks << bs,
	ZIO_PRIORITY_SYNC_READ);

	for (int b = 0; b < zap_f_phys(zap)->zap_ptrtbl.zt_numblks;
	b++) {
	dmu_buf_t *db;
	int err;

	err = dmu_buf_hold_by_dnode(zap->zap_dnode,
	(zap_f_phys(zap)->zap_ptrtbl.zt_blk + b) << bs,
	FTAG, &db, DMU_READ_NO_PREFETCH);
	if (err == 0) {
	zap_stats_ptrtbl(zap, db->db_data,
	1<<(bs-3), zs);
	dmu_buf_rele(db, FTAG);
	}
	}
	}
	}

	/*
	* Find last allocated block and update freeblk.
	*/
	static void
	zap_trunc(zap_t *zap)
	{
	uint64_t nentries;
	uint64_t lastblk;

	ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));

	if (zap_f_phys(zap)->zap_ptrtbl.zt_blk > 0) {
	/* External ptrtbl */
	nentries = (1 << zap_f_phys(zap)->zap_ptrtbl.zt_shift);
	lastblk = zap_f_phys(zap)->zap_ptrtbl.zt_blk +
	zap_f_phys(zap)->zap_ptrtbl.zt_numblks - 1;
	} else {
	/* Embedded ptrtbl */
	nentries = (1 << ZAP_EMBEDDED_PTRTBL_SHIFT(zap));
	lastblk = 0;
	}

	for (uint64_t idx = 0; idx < nentries; idx++) {
	uint64_t blk;
	if (zap_idx_to_blk(zap, idx, &blk) != 0)
	return;
	if (blk > lastblk)
	lastblk = blk;
	}

	ASSERT3U(lastblk, <, zap_f_phys(zap)->zap_freeblk);

	zap_f_phys(zap)->zap_freeblk = lastblk + 1;
	}

	/*
	* ZAP shrinking algorithm.
	*
	* We shrink ZAP recuresively removing empty leaves. We can remove an empty leaf
	* only if it has a sibling. Sibling leaves have the same prefix length and
	* their prefixes differ only by the least significant (sibling) bit. We require
	* both siblings to be empty. This eliminates a need to rehash the non-empty
	* remaining leaf. When we have removed one of two empty sibling, we set ptrtbl
	* entries of the removed leaf to point out to the remaining leaf. Prefix length
	* of the remaining leaf is decremented. As a result, it has a new prefix and it
	* might have a new sibling. So, we repeat the process.
	*
	* Steps:
	* 1. Check if a sibling leaf (sl) exists and it is empty.
	* 2. Release the leaf (l) if it has the sibling bit (slbit) equal to 1.
	* 3. Release the sibling (sl) to derefer it again with WRITER lock.
	* 4. Upgrade zapdir lock to WRITER (once).
	* 5. Derefer released leaves again.
	* 6. If it is needed, recheck whether both leaves are still siblings and empty.
	* 7. Set ptrtbl pointers of the removed leaf (slbit 1) to point out to blkid of
	* the remaining leaf (slbit 0).
	* 8. Free disk block of the removed leaf (dmu_free_range).
	* 9. Decrement prefix_len of the remaining leaf.
	* 10. Repeat the steps.
	*/
	static int
	zap_shrink(zap_name_t zn, zap_leaf_t l, dmu_tx_t *tx)
	{
	zap_t *zap = zn->zn_zap;
	int64_t zt_shift = zap_f_phys(zap)->zap_ptrtbl.zt_shift;
	uint64_t hash = zn->zn_hash;
	uint64_t prefix = zap_leaf_phys(l)->l_hdr.lh_prefix;
	uint64_t prefix_len = zap_leaf_phys(l)->l_hdr.lh_prefix_len;
	boolean_t trunc = B_FALSE;
	int err = 0;

	ASSERT3U(zap_leaf_phys(l)->l_hdr.lh_nentries, ==, 0);
	ASSERT3U(prefix_len, <=, zap_f_phys(zap)->zap_ptrtbl.zt_shift);
	ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));
	ASSERT3U(ZAP_HASH_IDX(hash, prefix_len), ==, prefix);

	boolean_t writer = B_FALSE;

	/*
	* To avoid deadlock always deref leaves in the same order -
	* sibling 0 first, then sibling 1.
	*/
	while (prefix_len) {
	zap_leaf_t *sl;
	int64_t prefix_diff = zt_shift - prefix_len;
	uint64_t sl_prefix = prefix ^ 1;
	uint64_t sl_hash = ZAP_PREFIX_HASH(sl_prefix, prefix_len);
	int slbit = prefix & 1;

	ASSERT3U(zap_leaf_phys(l)->l_hdr.lh_nentries, ==, 0);

	/*
	* Check if there is a sibling by reading ptrtbl ptrs.
	*/
	if (check_sibling_ptrtbl_range(zap, sl_prefix, prefix_len) == 0)
	break;

	/*
	* sibling 1, unlock it - we haven't yet dereferenced sibling 0.
	*/
	if (slbit == 1) {
	zap_put_leaf(l);
	l = NULL;
	}

	/*
	* Dereference sibling leaf and check if it is empty.
	*/
	if ((err = zap_deref_leaf(zap, sl_hash, tx, RW_READER,
	&sl)) != 0)
	break;

	ASSERT3U(ZAP_HASH_IDX(sl_hash, prefix_len), ==, sl_prefix);

	/*
	* Check if we have a sibling and it is empty.
	*/
	if (zap_leaf_phys(sl)->l_hdr.lh_prefix_len != prefix_len \|\|
	zap_leaf_phys(sl)->l_hdr.lh_nentries != 0) {
	zap_put_leaf(sl);
	break;
	}

	zap_put_leaf(sl);

	/*
	* If there two empty sibling, we have work to do, so
	* we need to lock ZAP ptrtbl as WRITER.
	*/
	if (!writer && (writer = zap_tryupgradedir(zap, tx)) == 0) {
	/* We failed to upgrade */
	if (l != NULL) {
	zap_put_leaf(l);
	l = NULL;
	}

	/*
	* Usually, the right way to upgrade from a READER lock
	* to a WRITER lock is to call zap_unlockdir() and
	* zap_lockdir(), but we do not have a tag. Instead,
	* we do it in more sophisticated way.
	*/
	rw_exit(&zap->zap_rwlock);
	rw_enter(&zap->zap_rwlock, RW_WRITER);
	dmu_buf_will_dirty(zap->zap_dbuf, tx);

	zt_shift = zap_f_phys(zap)->zap_ptrtbl.zt_shift;
	writer = B_TRUE;
	}

	/*
	* Here we have WRITER lock for ptrtbl.
	* Now, we need a WRITER lock for both siblings leaves.
	* Also, we have to recheck if the leaves are still siblings
	* and still empty.
	*/
	if (l == NULL) {
	/* sibling 0 */
	if ((err = zap_deref_leaf(zap, (slbit ? sl_hash : hash),
	tx, RW_WRITER, &l)) != 0)
	break;

	/*
	* The leaf isn't empty anymore or
	* it was shrunk/split while our locks were down.
	*/
	if (zap_leaf_phys(l)->l_hdr.lh_nentries != 0 \|\|
	zap_leaf_phys(l)->l_hdr.lh_prefix_len != prefix_len)
	break;
	}

	/* sibling 1 */
	if ((err = zap_deref_leaf(zap, (slbit ? hash : sl_hash), tx,
	RW_WRITER, &sl)) != 0)
	break;

	/*
	* The leaf isn't empty anymore or
	* it was shrunk/split while our locks were down.
	*/
	if (zap_leaf_phys(sl)->l_hdr.lh_nentries != 0 \|\|
	zap_leaf_phys(sl)->l_hdr.lh_prefix_len != prefix_len) {
	zap_put_leaf(sl);
	break;
	}

	/* If we have gotten here, we have a leaf to collapse */
	uint64_t idx = (slbit ? prefix : sl_prefix) << prefix_diff;
	uint64_t nptrs = (1ULL << prefix_diff);
	uint64_t sl_blkid = sl->l_blkid;

	/*
	* Set ptrtbl entries to point out to the slibling 0 blkid
	*/
	if ((err = zap_set_idx_range_to_blk(zap, idx, nptrs, l->l_blkid,
	tx)) != 0) {
	zap_put_leaf(sl);
	break;
	}

	/*
	* Free sibling 1 disk block.
	*/
	int bs = FZAP_BLOCK_SHIFT(zap);
	if (sl_blkid == zap_f_phys(zap)->zap_freeblk - 1)
	trunc = B_TRUE;

	(void) dmu_free_range(zap->zap_objset, zap->zap_object,
	sl_blkid << bs, 1 << bs, tx);
	zap_put_leaf(sl);

	zap_f_phys(zap)->zap_num_leafs--;

	/*
	* Update prefix and prefix_len.
	*/
	zap_leaf_phys(l)->l_hdr.lh_prefix >>= 1;
	zap_leaf_phys(l)->l_hdr.lh_prefix_len--;

	prefix = zap_leaf_phys(l)->l_hdr.lh_prefix;
	prefix_len = zap_leaf_phys(l)->l_hdr.lh_prefix_len;
	}

	if (trunc)
	zap_trunc(zap);

	if (l != NULL)
	zap_put_leaf(l);

	return (err);
	}

	/* CSTYLED */
	ZFS_MODULE_PARAM(zfs, , zap_iterate_prefetch, INT, ZMOD_RW,
	"When iterating ZAP object, prefetch it");

	/* CSTYLED */
	ZFS_MODULE_PARAM(zfs, , zap_shrink_enabled, INT, ZMOD_RW,
	"Enable ZAP shrinking");
	diff --git a/sys/contrib/openzfs/module/zfs/zcp.c b/sys/contrib/openzfs/module/zfs/zcp.c
	index 959404f665ab..7c279162a9d1 100644
	--- a/sys/contrib/openzfs/module/zfs/zcp.c
	+++ b/sys/contrib/openzfs/module/zfs/zcp.c
	@@ -1,1450 +1,1449 @@
	/*
	* CDDL HEADER START
	*
	* This file and its contents are supplied under the terms of the
	* Common Development and Distribution License ("CDDL"), version 1.0.
	* You may only use this file in accordance with the terms of version
	* 1.0 of the CDDL.
	*
	* A full copy of the text of the CDDL should have accompanied this
	* source. A copy of the CDDL is also available via the Internet at
	* http://www.illumos.org/license/CDDL.
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2016, 2018 by Delphix. All rights reserved.
	*/

	/*
	* ZFS Channel Programs (ZCP)
	*
	* The ZCP interface allows various ZFS commands and operations ZFS
	* administrative operations (e.g. creating and destroying snapshots, typically
	* performed via an ioctl to /dev/zfs by the zfs(8) command and
	* libzfs/libzfs_core) to be run * programmatically as a Lua script. A ZCP
	* script is run as a dsl_sync_task and fully executed during one transaction
	* group sync. This ensures that no other changes can be written concurrently
	* with a running Lua script. Combining multiple calls to the exposed ZFS
	* functions into one script gives a number of benefits:
	*
	* 1. Atomicity. For some compound or iterative operations, it's useful to be
	* able to guarantee that the state of a pool has not changed between calls to
	* ZFS.
	*
	* 2. Performance. If a large number of changes need to be made (e.g. deleting
	* many filesystems), there can be a significant performance penalty as a
	* result of the need to wait for a transaction group sync to pass for every
	* single operation. When expressed as a single ZCP script, all these changes
	* can be performed at once in one txg sync.
	*
	* A modified version of the Lua 5.2 interpreter is used to run channel program
	* scripts. The Lua 5.2 manual can be found at:
	*
	* http://www.lua.org/manual/5.2/
	*
	* If being run by a user (via an ioctl syscall), executing a ZCP script
	* requires root privileges in the global zone.
	*
	* Scripts are passed to zcp_eval() as a string, then run in a synctask by
	* zcp_eval_sync(). Arguments can be passed into the Lua script as an nvlist,
	* which will be converted to a Lua table. Similarly, values returned from
	* a ZCP script will be converted to an nvlist. See zcp_lua_to_nvlist_impl()
	* for details on exact allowed types and conversion.
	*
	* ZFS functionality is exposed to a ZCP script as a library of function calls.
	* These calls are sorted into submodules, such as zfs.list and zfs.sync, for
	* iterators and synctasks, respectively. Each of these submodules resides in
	* its own source file, with a zcp_*_info structure describing each library
	* call in the submodule.
	*
	* Error handling in ZCP scripts is handled by a number of different methods
	* based on severity:
	*
	* 1. Memory and time limits are in place to prevent a channel program from
	* consuming excessive system or running forever. If one of these limits is
	* hit, the channel program will be stopped immediately and return from
	* zcp_eval() with an error code. No attempt will be made to roll back or undo
	* any changes made by the channel program before the error occurred.
	* Consumers invoking zcp_eval() from elsewhere in the kernel may pass a time
	* limit of 0, disabling the time limit.
	*
	* 2. Internal Lua errors can occur as a result of a syntax error, calling a
	* library function with incorrect arguments, invoking the error() function,
	* failing an assert(), or other runtime errors. In these cases the channel
	* program will stop executing and return from zcp_eval() with an error code.
	* In place of a return value, an error message will also be returned in the
	* 'result' nvlist containing information about the error. No attempt will be
	* made to roll back or undo any changes made by the channel program before the
	* error occurred.
	*
	* 3. If an error occurs inside a ZFS library call which returns an error code,
	* the error is returned to the Lua script to be handled as desired.
	*
	* In the first two cases, Lua's error-throwing mechanism is used, which
	* longjumps out of the script execution with luaL_error() and returns with the
	* error.
	*
	* See zfs-program(8) for more information on high level usage.
	*/

	#include <sys/lua/lua.h>
	#include <sys/lua/lualib.h>
	#include <sys/lua/lauxlib.h>

	#include <sys/dsl_prop.h>
	#include <sys/dsl_synctask.h>
	#include <sys/dsl_dataset.h>
	#include <sys/zcp.h>
	#include <sys/zcp_iter.h>
	#include <sys/zcp_prop.h>
	#include <sys/zcp_global.h>
	#include <sys/zvol.h>

	#ifndef KM_NORMALPRI
	#define KM_NORMALPRI 0
	#endif

	#define ZCP_NVLIST_MAX_DEPTH 20

	static const uint64_t zfs_lua_check_instrlimit_interval = 100;
	uint64_t zfs_lua_max_instrlimit = ZCP_MAX_INSTRLIMIT;
	uint64_t zfs_lua_max_memlimit = ZCP_MAX_MEMLIMIT;

	/*
	* Forward declarations for mutually recursive functions
	*/
	static int zcp_nvpair_value_to_lua(lua_State , nvpair_t , char *, int);
	static int zcp_lua_to_nvlist_impl(lua_State , int, nvlist_t , const char *,
	int);

	/*
	* The outer-most error callback handler for use with lua_pcall(). On
	* error Lua will call this callback with a single argument that
	* represents the error value. In most cases this will be a string
	* containing an error message, but channel programs can use Lua's
	* error() function to return arbitrary objects as errors. This callback
	* returns (on the Lua stack) the original error object along with a traceback.
	*
	* Fatal Lua errors can occur while resources are held, so we also call any
	* registered cleanup function here.
	*/
	static int
	zcp_error_handler(lua_State *state)
	{
	const char *msg;

	zcp_cleanup(state);

	VERIFY3U(1, ==, lua_gettop(state));
	msg = lua_tostring(state, 1);
	luaL_traceback(state, state, msg, 1);
	return (1);
	}

	int
	zcp_argerror(lua_State state, int narg, const char msg, ...)
	{
	va_list alist;

	va_start(alist, msg);
	const char *buf = lua_pushvfstring(state, msg, alist);
	va_end(alist);

	return (luaL_argerror(state, narg, buf));
	}

	/*
	* Install a new cleanup function, which will be invoked with the given
	* opaque argument if a fatal error causes the Lua interpreter to longjump out
	* of a function call.
	*
	* If an error occurs, the cleanup function will be invoked exactly once and
	* then unregistered.
	*
	* Returns the registered cleanup handler so the caller can deregister it
	* if no error occurs.
	*/
	zcp_cleanup_handler_t *
	zcp_register_cleanup(lua_State state, zcp_cleanup_t cleanfunc, void cleanarg)
	{
	zcp_run_info_t *ri = zcp_run_info(state);

	zcp_cleanup_handler_t zch = kmem_alloc(sizeof (zch), KM_SLEEP);
	zch->zch_cleanup_func = cleanfunc;
	zch->zch_cleanup_arg = cleanarg;
	list_insert_head(&ri->zri_cleanup_handlers, zch);

	return (zch);
	}

	void
	zcp_deregister_cleanup(lua_State state, zcp_cleanup_handler_t zch)
	{
	zcp_run_info_t *ri = zcp_run_info(state);
	list_remove(&ri->zri_cleanup_handlers, zch);
	kmem_free(zch, sizeof (*zch));
	}

	/*
	* Execute the currently registered cleanup handlers then free them and
	* destroy the handler list.
	*/
	void
	zcp_cleanup(lua_State *state)
	{
	zcp_run_info_t *ri = zcp_run_info(state);

	for (zcp_cleanup_handler_t *zch =
	list_remove_head(&ri->zri_cleanup_handlers); zch != NULL;
	zch = list_remove_head(&ri->zri_cleanup_handlers)) {
	zch->zch_cleanup_func(zch->zch_cleanup_arg);
	kmem_free(zch, sizeof (*zch));
	}
	}

	/*
	* Convert the lua table at the given index on the Lua stack to an nvlist
	* and return it.
	*
	* If the table can not be converted for any reason, NULL is returned and
	* an error message is pushed onto the Lua stack.
	*/
	static nvlist_t *
	zcp_table_to_nvlist(lua_State *state, int index, int depth)
	{
	nvlist_t *nvl;
	/*
	* Converting a Lua table to an nvlist with key uniqueness checking is
	* O(n^2) in the number of keys in the nvlist, which can take a long
	* time when we return a large table from a channel program.
	* Furthermore, Lua's table interface almost guarantees unique keys
	* on its own (details below). Therefore, we don't use fnvlist_alloc()
	* here to avoid the built-in uniqueness checking.
	*
	* The almost is because it's possible to have key collisions between
	* e.g. the string "1" and the number 1, or the string "true" and the
	* boolean true, so we explicitly check that when we're looking at a
	* key which is an integer / boolean or a string that can be parsed as
	* one of those types. In the worst case this could still devolve into
	* O(n^2), so we only start doing these checks on boolean/integer keys
	* once we've seen a string key which fits this weird usage pattern.
	*
	* Ultimately, we still want callers to know that the keys in this
	* nvlist are unique, so before we return this we set the nvlist's
	* flags to reflect that.
	*/
	VERIFY0(nvlist_alloc(&nvl, 0, KM_SLEEP));

	/*
	* Push an empty stack slot where lua_next() will store each
	* table key.
	*/
	lua_pushnil(state);
	boolean_t saw_str_could_collide = B_FALSE;
	while (lua_next(state, index) != 0) {
	/*
	* The next key-value pair from the table at index is
	* now on the stack, with the key at stack slot -2 and
	* the value at slot -1.
	*/
	int err = 0;
	char buf[32];
	const char *key = NULL;
	boolean_t key_could_collide = B_FALSE;

	switch (lua_type(state, -2)) {
	case LUA_TSTRING:
	key = lua_tostring(state, -2);

	/* check if this could collide with a number or bool */
	long long tmp;
	int parselen;
	if ((sscanf(key, "%lld%n", &tmp, &parselen) > 0 &&
	parselen == strlen(key)) \|\|
	strcmp(key, "true") == 0 \|\|
	strcmp(key, "false") == 0) {
	key_could_collide = B_TRUE;
	saw_str_could_collide = B_TRUE;
	}
	break;
	case LUA_TBOOLEAN:
	key = (lua_toboolean(state, -2) == B_TRUE ?
	"true" : "false");
	if (saw_str_could_collide) {
	key_could_collide = B_TRUE;
	}
	break;
	case LUA_TNUMBER:
	(void) snprintf(buf, sizeof (buf), "%lld",
	(longlong_t)lua_tonumber(state, -2));

	key = buf;
	if (saw_str_could_collide) {
	key_could_collide = B_TRUE;
	}
	break;
	default:
	fnvlist_free(nvl);
	(void) lua_pushfstring(state, "Invalid key "
	"type '%s' in table",
	lua_typename(state, lua_type(state, -2)));
	return (NULL);
	}
	/*
	* Check for type-mismatched key collisions, and throw an error.
	*/
	if (key_could_collide && nvlist_exists(nvl, key)) {
	fnvlist_free(nvl);
	(void) lua_pushfstring(state, "Collision of "
	"key '%s' in table", key);
	return (NULL);
	}
	/*
	* Recursively convert the table value and insert into
	* the new nvlist with the parsed key. To prevent
	* stack overflow on circular or heavily nested tables,
	* we track the current nvlist depth.
	*/
	if (depth >= ZCP_NVLIST_MAX_DEPTH) {
	fnvlist_free(nvl);
	(void) lua_pushfstring(state, "Maximum table "
	"depth (%d) exceeded for table",
	ZCP_NVLIST_MAX_DEPTH);
	return (NULL);
	}
	err = zcp_lua_to_nvlist_impl(state, -1, nvl, key,
	depth + 1);
	if (err != 0) {
	fnvlist_free(nvl);
	/*
	* Error message has been pushed to the lua
	* stack by the recursive call.
	*/
	return (NULL);
	}
	/*
	* Pop the value pushed by lua_next().
	*/
	lua_pop(state, 1);
	}

	/*
	* Mark the nvlist as having unique keys. This is a little ugly, but we
	* ensured above that there are no duplicate keys in the nvlist.
	*/
	nvl->nvl_nvflag \|= NV_UNIQUE_NAME;

	return (nvl);
	}

	/*
	* Convert a value from the given index into the lua stack to an nvpair, adding
	* it to an nvlist with the given key.
	*
	* Values are converted as follows:
	*
	* string -> string
	* number -> int64
	* boolean -> boolean
	* nil -> boolean (no value)
	*
	* Lua tables are converted to nvlists and then inserted. The table's keys
	* are converted to strings then used as keys in the nvlist to store each table
	* element. Keys are converted as follows:
	*
	* string -> no change
	* number -> "%lld"
	* boolean -> "true" \| "false"
	* nil -> error
	*
	* In the case of a key collision, an error is thrown.
	*
	* If an error is encountered, a nonzero error code is returned, and an error
	* string will be pushed onto the Lua stack.
	*/
	static int
	zcp_lua_to_nvlist_impl(lua_State state, int index, nvlist_t nvl,
	const char *key, int depth)
	{
	/*
	* Verify that we have enough remaining space in the lua stack to parse
	* a key-value pair and push an error.
	*/
	if (!lua_checkstack(state, 3)) {
	(void) lua_pushstring(state, "Lua stack overflow");
	return (1);
	}

	index = lua_absindex(state, index);

	switch (lua_type(state, index)) {
	case LUA_TNIL:
	fnvlist_add_boolean(nvl, key);
	break;
	case LUA_TBOOLEAN:
	fnvlist_add_boolean_value(nvl, key,
	lua_toboolean(state, index));
	break;
	case LUA_TNUMBER:
	fnvlist_add_int64(nvl, key, lua_tonumber(state, index));
	break;
	case LUA_TSTRING:
	fnvlist_add_string(nvl, key, lua_tostring(state, index));
	break;
	case LUA_TTABLE: {
	nvlist_t *value_nvl = zcp_table_to_nvlist(state, index, depth);
	if (value_nvl == NULL)
	return (SET_ERROR(EINVAL));

	fnvlist_add_nvlist(nvl, key, value_nvl);
	fnvlist_free(value_nvl);
	break;
	}
	default:
	(void) lua_pushfstring(state,
	"Invalid value type '%s' for key '%s'",
	lua_typename(state, lua_type(state, index)), key);
	return (SET_ERROR(EINVAL));
	}

	return (0);
	}

	/*
	* Convert a lua value to an nvpair, adding it to an nvlist with the given key.
	*/
	static void
	zcp_lua_to_nvlist(lua_State state, int index, nvlist_t nvl, const char *key)
	{
	/*
	* On error, zcp_lua_to_nvlist_impl pushes an error string onto the Lua
	* stack before returning with a nonzero error code. If an error is
	* returned, throw a fatal lua error with the given string.
	*/
	if (zcp_lua_to_nvlist_impl(state, index, nvl, key, 0) != 0)
	(void) lua_error(state);
	}

	static int
	zcp_lua_to_nvlist_helper(lua_State *state)
	{
	nvlist_t nv = (nvlist_t )lua_touserdata(state, 2);
	const char key = (const char )lua_touserdata(state, 1);
	zcp_lua_to_nvlist(state, 3, nv, key);
	return (0);
	}

	static void
	zcp_convert_return_values(lua_State state, nvlist_t nvl,
	const char key, int result)
	{
	int err;
	VERIFY3U(1, ==, lua_gettop(state));
	lua_pushcfunction(state, zcp_lua_to_nvlist_helper);
	lua_pushlightuserdata(state, (char *)key);
	lua_pushlightuserdata(state, nvl);
	lua_pushvalue(state, 1);
	lua_remove(state, 1);
	err = lua_pcall(state, 3, 0, 0); /* zcp_lua_to_nvlist_helper */
	if (err != 0) {
	zcp_lua_to_nvlist(state, 1, nvl, ZCP_RET_ERROR);
	*result = SET_ERROR(ECHRNG);
	}
	}

	/*
	* Push a Lua table representing nvl onto the stack. If it can't be
	* converted, return EINVAL, fill in errbuf, and push nothing. errbuf may
	* be specified as NULL, in which case no error string will be output.
	*
	* Most nvlists are converted as simple key->value Lua tables, but we make
	* an exception for the case where all nvlist entries are BOOLEANs (a string
	* key without a value). In Lua, a table key pointing to a value of Nil
	* (no value) is equivalent to the key not existing, so a BOOLEAN nvlist
	* entry can't be directly converted to a Lua table entry. Nvlists of entirely
	* BOOLEAN entries are frequently used to pass around lists of datasets, so for
	* convenience we check for this case, and convert it to a simple Lua array of
	* strings.
	*/
	int
	zcp_nvlist_to_lua(lua_State state, nvlist_t nvl,
	char *errbuf, int errbuf_len)
	{
	nvpair_t *pair;
	lua_newtable(state);
	boolean_t has_values = B_FALSE;
	/*
	* If the list doesn't have any values, just convert it to a string
	* array.
	*/
	for (pair = nvlist_next_nvpair(nvl, NULL);
	pair != NULL; pair = nvlist_next_nvpair(nvl, pair)) {
	if (nvpair_type(pair) != DATA_TYPE_BOOLEAN) {
	has_values = B_TRUE;
	break;
	}
	}
	if (!has_values) {
	int i = 1;
	for (pair = nvlist_next_nvpair(nvl, NULL);
	pair != NULL; pair = nvlist_next_nvpair(nvl, pair)) {
	(void) lua_pushinteger(state, i);
	(void) lua_pushstring(state, nvpair_name(pair));
	(void) lua_settable(state, -3);
	i++;
	}
	} else {
	for (pair = nvlist_next_nvpair(nvl, NULL);
	pair != NULL; pair = nvlist_next_nvpair(nvl, pair)) {
	int err = zcp_nvpair_value_to_lua(state, pair,
	errbuf, errbuf_len);
	if (err != 0) {
	lua_pop(state, 1);
	return (err);
	}
	(void) lua_setfield(state, -2, nvpair_name(pair));
	}
	}
	return (0);
	}

	/*
	* Push a Lua object representing the value of "pair" onto the stack.
	*
	* Only understands boolean_value, string, int64, nvlist,
	* string_array, and int64_array type values. For other
	* types, returns EINVAL, fills in errbuf, and pushes nothing.
	*/
	static int
	zcp_nvpair_value_to_lua(lua_State state, nvpair_t pair,
	char *errbuf, int errbuf_len)
	{
	int err = 0;

	if (pair == NULL) {
	lua_pushnil(state);
	return (0);
	}

	switch (nvpair_type(pair)) {
	case DATA_TYPE_BOOLEAN_VALUE:
	(void) lua_pushboolean(state,
	fnvpair_value_boolean_value(pair));
	break;
	case DATA_TYPE_STRING:
	(void) lua_pushstring(state, fnvpair_value_string(pair));
	break;
	case DATA_TYPE_INT64:
	(void) lua_pushinteger(state, fnvpair_value_int64(pair));
	break;
	case DATA_TYPE_NVLIST:
	err = zcp_nvlist_to_lua(state,
	fnvpair_value_nvlist(pair), errbuf, errbuf_len);
	break;
	case DATA_TYPE_STRING_ARRAY: {
	const char **strarr;
	uint_t nelem;
	(void) nvpair_value_string_array(pair, &strarr, &nelem);
	lua_newtable(state);
	for (int i = 0; i < nelem; i++) {
	(void) lua_pushinteger(state, i + 1);
	(void) lua_pushstring(state, strarr[i]);
	(void) lua_settable(state, -3);
	}
	break;
	}
	case DATA_TYPE_UINT64_ARRAY: {
	uint64_t *intarr;
	uint_t nelem;
	(void) nvpair_value_uint64_array(pair, &intarr, &nelem);
	lua_newtable(state);
	for (int i = 0; i < nelem; i++) {
	(void) lua_pushinteger(state, i + 1);
	(void) lua_pushinteger(state, intarr[i]);
	(void) lua_settable(state, -3);
	}
	break;
	}
	case DATA_TYPE_INT64_ARRAY: {
	int64_t *intarr;
	uint_t nelem;
	(void) nvpair_value_int64_array(pair, &intarr, &nelem);
	lua_newtable(state);
	for (int i = 0; i < nelem; i++) {
	(void) lua_pushinteger(state, i + 1);
	(void) lua_pushinteger(state, intarr[i]);
	(void) lua_settable(state, -3);
	}
	break;
	}
	default: {
	if (errbuf != NULL) {
	(void) snprintf(errbuf, errbuf_len,
	"Unhandled nvpair type %d for key '%s'",
	nvpair_type(pair), nvpair_name(pair));
	}
	return (SET_ERROR(EINVAL));
	}
	}
	return (err);
	}

	int
	zcp_dataset_hold_error(lua_State state, dsl_pool_t dp, const char *dsname,
	int error)
	{
	if (error == ENOENT) {
	(void) zcp_argerror(state, 1, "no such dataset '%s'", dsname);
	return (0); /* not reached; zcp_argerror will longjmp */
	} else if (error == EXDEV) {
	(void) zcp_argerror(state, 1,
	"dataset '%s' is not in the target pool '%s'",
	dsname, spa_name(dp->dp_spa));
	return (0); /* not reached; zcp_argerror will longjmp */
	} else if (error == EIO) {
	(void) luaL_error(state,
	"I/O error while accessing dataset '%s'", dsname);
	return (0); /* not reached; luaL_error will longjmp */
	} else if (error != 0) {
	(void) luaL_error(state,
	"unexpected error %d while accessing dataset '%s'",
	error, dsname);
	return (0); /* not reached; luaL_error will longjmp */
	}
	return (0);
	}

	/*
	* Note: will longjmp (via lua_error()) on error.
	* Assumes that the dsname is argument #1 (for error reporting purposes).
	*/
	dsl_dataset_t *
	zcp_dataset_hold(lua_State state, dsl_pool_t dp, const char *dsname,
	const void *tag)
	{
	dsl_dataset_t *ds;
	int error = dsl_dataset_hold(dp, dsname, tag, &ds);
	(void) zcp_dataset_hold_error(state, dp, dsname, error);
	return (ds);
	}

	static int zcp_debug(lua_State *);
	static const zcp_lib_info_t zcp_debug_info = {
	.name = "debug",
	.func = zcp_debug,
	.pargs = {
	{ .za_name = "debug string", .za_lua_type = LUA_TSTRING },
	{NULL, 0}
	},
	.kwargs = {
	{NULL, 0}
	}
	};

	static int
	zcp_debug(lua_State *state)
	{
	const char *dbgstring;
	zcp_run_info_t *ri = zcp_run_info(state);
	const zcp_lib_info_t *libinfo = &zcp_debug_info;

	zcp_parse_args(state, libinfo->name, libinfo->pargs, libinfo->kwargs);

	dbgstring = lua_tostring(state, 1);

	zfs_dbgmsg("txg %lld ZCP: %s", (longlong_t)ri->zri_tx->tx_txg,
	dbgstring);

	return (0);
	}

	static int zcp_exists(lua_State *);
	static const zcp_lib_info_t zcp_exists_info = {
	.name = "exists",
	.func = zcp_exists,
	.pargs = {
	{ .za_name = "dataset", .za_lua_type = LUA_TSTRING },
	{NULL, 0}
	},
	.kwargs = {
	{NULL, 0}
	}
	};

	static int
	zcp_exists(lua_State *state)
	{
	zcp_run_info_t *ri = zcp_run_info(state);
	dsl_pool_t *dp = ri->zri_pool;
	const zcp_lib_info_t *libinfo = &zcp_exists_info;

	zcp_parse_args(state, libinfo->name, libinfo->pargs, libinfo->kwargs);

	const char *dsname = lua_tostring(state, 1);

	dsl_dataset_t *ds;
	int error = dsl_dataset_hold(dp, dsname, FTAG, &ds);
	if (error == 0) {
	dsl_dataset_rele(ds, FTAG);
	lua_pushboolean(state, B_TRUE);
	} else if (error == ENOENT) {
	lua_pushboolean(state, B_FALSE);
	} else if (error == EXDEV) {
	return (luaL_error(state, "dataset '%s' is not in the "
	"target pool", dsname));
	} else if (error == EIO) {
	return (luaL_error(state, "I/O error opening dataset '%s'",
	dsname));
	} else if (error != 0) {
	return (luaL_error(state, "unexpected error %d", error));
	}

	return (1);
	}

	/*
	* Allocate/realloc/free a buffer for the lua interpreter.
	*
	* When nsize is 0, behaves as free() and returns NULL.
	*
	* If ptr is NULL, behaves as malloc() and returns an allocated buffer of size
	* at least nsize.
	*
	* Otherwise, behaves as realloc(), changing the allocation from osize to nsize.
	* Shrinking the buffer size never fails.
	*
	* The original allocated buffer size is stored as a uint64 at the beginning of
	* the buffer to avoid actually reallocating when shrinking a buffer, since lua
	* requires that this operation never fail.
	*/
	static void *
	zcp_lua_alloc(void ud, void ptr, size_t osize, size_t nsize)
	{
	zcp_alloc_arg_t *allocargs = ud;

	if (nsize == 0) {
	if (ptr != NULL) {
	int64_t allocbuf = (int64_t )ptr - 1;
	int64_t allocsize = *allocbuf;
	ASSERT3S(allocsize, >, 0);
	ASSERT3S(allocargs->aa_alloc_remaining + allocsize, <=,
	allocargs->aa_alloc_limit);
	allocargs->aa_alloc_remaining += allocsize;
	vmem_free(allocbuf, allocsize);
	}
	return (NULL);
	} else if (ptr == NULL) {
	int64_t *allocbuf;
	int64_t allocsize = nsize + sizeof (int64_t);

	if (!allocargs->aa_must_succeed &&
	(allocsize <= 0 \|\|
	allocsize > allocargs->aa_alloc_remaining)) {
	return (NULL);
	}

	allocbuf = vmem_alloc(allocsize, KM_SLEEP);
	allocargs->aa_alloc_remaining -= allocsize;

	*allocbuf = allocsize;
	return (allocbuf + 1);
	} else if (nsize <= osize) {
	/*
	* If shrinking the buffer, lua requires that the reallocation
	* never fail.
	*/
	return (ptr);
	} else {
	ASSERT3U(nsize, >, osize);

	uint64_t *luabuf = zcp_lua_alloc(ud, NULL, 0, nsize);
	if (luabuf == NULL) {
	return (NULL);
	}
	(void) memcpy(luabuf, ptr, osize);
	VERIFY3P(zcp_lua_alloc(ud, ptr, osize, 0), ==, NULL);
	return (luabuf);
	}
	}

	static void
	zcp_lua_counthook(lua_State state, lua_Debug ar)
	{
	(void) ar;
	lua_getfield(state, LUA_REGISTRYINDEX, ZCP_RUN_INFO_KEY);
	zcp_run_info_t *ri = lua_touserdata(state, -1);

	/*
	* Check if we were canceled while waiting for the
	* txg to sync or from our open context thread
	*/
	- if (ri->zri_canceled \|\|
	- (!ri->zri_sync && issig(JUSTLOOKING) && issig(FORREAL))) {
	+ if (ri->zri_canceled \|\| (!ri->zri_sync && issig())) {
	ri->zri_canceled = B_TRUE;
	(void) lua_pushstring(state, "Channel program was canceled.");
	(void) lua_error(state);
	/* Unreachable */
	}

	/*
	* Check how many instructions the channel program has
	* executed so far, and compare against the limit.
	*/
	ri->zri_curinstrs += zfs_lua_check_instrlimit_interval;
	if (ri->zri_maxinstrs != 0 && ri->zri_curinstrs > ri->zri_maxinstrs) {
	ri->zri_timed_out = B_TRUE;
	(void) lua_pushstring(state,
	"Channel program timed out.");
	(void) lua_error(state);
	/* Unreachable */
	}
	}

	static int
	zcp_panic_cb(lua_State *state)
	{
	panic("unprotected error in call to Lua API (%s)\n",
	lua_tostring(state, -1));
	return (0);
	}

	static void
	zcp_eval_impl(dmu_tx_t tx, zcp_run_info_t ri)
	{
	int err;
	lua_State *state = ri->zri_state;

	VERIFY3U(3, ==, lua_gettop(state));

	/* finish initializing our runtime state */
	ri->zri_pool = dmu_tx_pool(tx);
	ri->zri_tx = tx;
	list_create(&ri->zri_cleanup_handlers, sizeof (zcp_cleanup_handler_t),
	offsetof(zcp_cleanup_handler_t, zch_node));

	/*
	* Store the zcp_run_info_t struct for this run in the Lua registry.
	* Registry entries are not directly accessible by the Lua scripts but
	* can be accessed by our callbacks.
	*/
	lua_pushlightuserdata(state, ri);
	lua_setfield(state, LUA_REGISTRYINDEX, ZCP_RUN_INFO_KEY);
	VERIFY3U(3, ==, lua_gettop(state));

	/*
	* Tell the Lua interpreter to call our handler every count
	* instructions. Channel programs that execute too many instructions
	* should die with ETIME.
	*/
	(void) lua_sethook(state, zcp_lua_counthook, LUA_MASKCOUNT,
	zfs_lua_check_instrlimit_interval);

	/*
	* Tell the Lua memory allocator to stop using KM_SLEEP before handing
	* off control to the channel program. Channel programs that use too
	* much memory should die with ENOSPC.
	*/
	ri->zri_allocargs->aa_must_succeed = B_FALSE;

	/*
	* Call the Lua function that open-context passed us. This pops the
	* function and its input from the stack and pushes any return
	* or error values.
	*/
	err = lua_pcall(state, 1, LUA_MULTRET, 1);

	/*
	* Let Lua use KM_SLEEP while we interpret the return values.
	*/
	ri->zri_allocargs->aa_must_succeed = B_TRUE;

	/*
	* Remove the error handler callback from the stack. At this point,
	* there shouldn't be any cleanup handler registered in the handler
	* list (zri_cleanup_handlers), regardless of whether it ran or not.
	*/
	list_destroy(&ri->zri_cleanup_handlers);
	lua_remove(state, 1);

	switch (err) {
	case LUA_OK: {
	/*
	* Lua supports returning multiple values in a single return
	* statement. Return values will have been pushed onto the
	* stack:
	* 1: Return value 1
	* 2: Return value 2
	* 3: etc...
	* To simplify the process of retrieving a return value from a
	* channel program, we disallow returning more than one value
	* to ZFS from the Lua script, yielding a singleton return
	* nvlist of the form { "return": Return value 1 }.
	*/
	int return_count = lua_gettop(state);

	if (return_count == 1) {
	ri->zri_result = 0;
	zcp_convert_return_values(state, ri->zri_outnvl,
	ZCP_RET_RETURN, &ri->zri_result);
	} else if (return_count > 1) {
	ri->zri_result = SET_ERROR(ECHRNG);
	lua_settop(state, 0);
	(void) lua_pushfstring(state, "Multiple return "
	"values not supported");
	zcp_convert_return_values(state, ri->zri_outnvl,
	ZCP_RET_ERROR, &ri->zri_result);
	}
	break;
	}
	case LUA_ERRRUN:
	case LUA_ERRGCMM: {
	/*
	* The channel program encountered a fatal error within the
	* script, such as failing an assertion, or calling a function
	* with incompatible arguments. The error value and the
	* traceback generated by zcp_error_handler() should be on the
	* stack.
	*/
	VERIFY3U(1, ==, lua_gettop(state));
	if (ri->zri_timed_out) {
	ri->zri_result = SET_ERROR(ETIME);
	} else if (ri->zri_canceled) {
	ri->zri_result = SET_ERROR(EINTR);
	} else {
	ri->zri_result = SET_ERROR(ECHRNG);
	}

	zcp_convert_return_values(state, ri->zri_outnvl,
	ZCP_RET_ERROR, &ri->zri_result);

	if (ri->zri_result == ETIME && ri->zri_outnvl != NULL) {
	(void) nvlist_add_uint64(ri->zri_outnvl,
	ZCP_ARG_INSTRLIMIT, ri->zri_curinstrs);
	}
	break;
	}
	case LUA_ERRERR: {
	/*
	* The channel program encountered a fatal error within the
	* script, and we encountered another error while trying to
	* compute the traceback in zcp_error_handler(). We can only
	* return the error message.
	*/
	VERIFY3U(1, ==, lua_gettop(state));
	if (ri->zri_timed_out) {
	ri->zri_result = SET_ERROR(ETIME);
	} else if (ri->zri_canceled) {
	ri->zri_result = SET_ERROR(EINTR);
	} else {
	ri->zri_result = SET_ERROR(ECHRNG);
	}

	zcp_convert_return_values(state, ri->zri_outnvl,
	ZCP_RET_ERROR, &ri->zri_result);
	break;
	}
	case LUA_ERRMEM:
	/*
	* Lua ran out of memory while running the channel program.
	* There's not much we can do.
	*/
	ri->zri_result = SET_ERROR(ENOSPC);
	break;
	default:
	VERIFY0(err);
	}
	}

	static void
	zcp_pool_error(zcp_run_info_t ri, const char poolname, int error)
	{
	ri->zri_result = SET_ERROR(ECHRNG);
	lua_settop(ri->zri_state, 0);
	(void) lua_pushfstring(ri->zri_state, "Could not open pool: %s "
	"errno: %d", poolname, error);
	zcp_convert_return_values(ri->zri_state, ri->zri_outnvl,
	ZCP_RET_ERROR, &ri->zri_result);

	}

	/*
	* This callback is called when txg_wait_synced_sig encountered a signal.
	* The txg_wait_synced_sig will continue to wait for the txg to complete
	* after calling this callback.
	*/
	static void
	zcp_eval_sig(void arg, dmu_tx_t tx)
	{
	(void) tx;
	zcp_run_info_t *ri = arg;

	ri->zri_canceled = B_TRUE;
	}

	static void
	zcp_eval_sync(void arg, dmu_tx_t tx)
	{
	zcp_run_info_t *ri = arg;

	/*
	* Open context should have setup the stack to contain:
	* 1: Error handler callback
	* 2: Script to run (converted to a Lua function)
	* 3: nvlist input to function (converted to Lua table or nil)
	*/
	VERIFY3U(3, ==, lua_gettop(ri->zri_state));

	zcp_eval_impl(tx, ri);
	}

	static void
	zcp_eval_open(zcp_run_info_t ri, const char poolname)
	{
	int error;
	dsl_pool_t *dp;
	dmu_tx_t *tx;

	/*
	* See comment from the same assertion in zcp_eval_sync().
	*/
	VERIFY3U(3, ==, lua_gettop(ri->zri_state));

	error = dsl_pool_hold(poolname, FTAG, &dp);
	if (error != 0) {
	zcp_pool_error(ri, poolname, error);
	return;
	}

	/*
	* As we are running in open-context, we have no transaction associated
	* with the channel program. At the same time, functions from the
	* zfs.check submodule need to be associated with a transaction as
	* they are basically dry-runs of their counterparts in the zfs.sync
	* submodule. These functions should be able to run in open-context.
	* Therefore we create a new transaction that we later abort once
	* the channel program has been evaluated.
	*/
	tx = dmu_tx_create_dd(dp->dp_mos_dir);

	zcp_eval_impl(tx, ri);

	dmu_tx_abort(tx);

	dsl_pool_rele(dp, FTAG);
	}

	int
	zcp_eval(const char poolname, const char program, boolean_t sync,
	uint64_t instrlimit, uint64_t memlimit, nvpair_t nvarg, nvlist_t outnvl)
	{
	int err;
	lua_State *state;
	zcp_run_info_t runinfo;

	if (instrlimit > zfs_lua_max_instrlimit)
	return (SET_ERROR(EINVAL));
	if (memlimit == 0 \|\| memlimit > zfs_lua_max_memlimit)
	return (SET_ERROR(EINVAL));

	zcp_alloc_arg_t allocargs = {
	.aa_must_succeed = B_TRUE,
	.aa_alloc_remaining = (int64_t)memlimit,
	.aa_alloc_limit = (int64_t)memlimit,
	};

	/*
	* Creates a Lua state with a memory allocator that uses KM_SLEEP.
	* This should never fail.
	*/
	state = lua_newstate(zcp_lua_alloc, &allocargs);
	VERIFY(state != NULL);
	(void) lua_atpanic(state, zcp_panic_cb);

	/*
	* Load core Lua libraries we want access to.
	*/
	VERIFY3U(1, ==, luaopen_base(state));
	lua_pop(state, 1);
	VERIFY3U(1, ==, luaopen_coroutine(state));
	lua_setglobal(state, LUA_COLIBNAME);
	VERIFY0(lua_gettop(state));
	VERIFY3U(1, ==, luaopen_string(state));
	lua_setglobal(state, LUA_STRLIBNAME);
	VERIFY0(lua_gettop(state));
	VERIFY3U(1, ==, luaopen_table(state));
	lua_setglobal(state, LUA_TABLIBNAME);
	VERIFY0(lua_gettop(state));

	/*
	* Load globally visible variables such as errno aliases.
	*/
	zcp_load_globals(state);
	VERIFY0(lua_gettop(state));

	/*
	* Load ZFS-specific modules.
	*/
	lua_newtable(state);
	VERIFY3U(1, ==, zcp_load_list_lib(state));
	lua_setfield(state, -2, "list");
	VERIFY3U(1, ==, zcp_load_synctask_lib(state, B_FALSE));
	lua_setfield(state, -2, "check");
	VERIFY3U(1, ==, zcp_load_synctask_lib(state, B_TRUE));
	lua_setfield(state, -2, "sync");
	VERIFY3U(1, ==, zcp_load_get_lib(state));
	lua_pushcclosure(state, zcp_debug_info.func, 0);
	lua_setfield(state, -2, zcp_debug_info.name);
	lua_pushcclosure(state, zcp_exists_info.func, 0);
	lua_setfield(state, -2, zcp_exists_info.name);
	lua_setglobal(state, "zfs");
	VERIFY0(lua_gettop(state));

	/*
	* Push the error-callback that calculates Lua stack traces on
	* unexpected failures.
	*/
	lua_pushcfunction(state, zcp_error_handler);
	VERIFY3U(1, ==, lua_gettop(state));

	/*
	* Load the actual script as a function onto the stack as text ("t").
	* The only valid error condition is a syntax error in the script.
	* ERRMEM should not be possible because our allocator is using
	* KM_SLEEP. ERRGCMM should not be possible because we have not added
	* any objects with __gc metamethods to the interpreter that could
	* fail.
	*/
	err = luaL_loadbufferx(state, program, strlen(program),
	"channel program", "t");
	if (err == LUA_ERRSYNTAX) {
	fnvlist_add_string(outnvl, ZCP_RET_ERROR,
	lua_tostring(state, -1));
	lua_close(state);
	return (SET_ERROR(EINVAL));
	}
	VERIFY0(err);
	VERIFY3U(2, ==, lua_gettop(state));

	/*
	* Convert the input nvlist to a Lua object and put it on top of the
	* stack.
	*/
	char errmsg[128];
	err = zcp_nvpair_value_to_lua(state, nvarg,
	errmsg, sizeof (errmsg));
	if (err != 0) {
	fnvlist_add_string(outnvl, ZCP_RET_ERROR, errmsg);
	lua_close(state);
	return (SET_ERROR(EINVAL));
	}
	VERIFY3U(3, ==, lua_gettop(state));

	runinfo.zri_state = state;
	runinfo.zri_allocargs = &allocargs;
	runinfo.zri_outnvl = outnvl;
	runinfo.zri_result = 0;
	runinfo.zri_cred = CRED();
	runinfo.zri_proc = curproc;
	runinfo.zri_timed_out = B_FALSE;
	runinfo.zri_canceled = B_FALSE;
	runinfo.zri_sync = sync;
	runinfo.zri_space_used = 0;
	runinfo.zri_curinstrs = 0;
	runinfo.zri_maxinstrs = instrlimit;
	runinfo.zri_new_zvols = fnvlist_alloc();

	if (sync) {
	err = dsl_sync_task_sig(poolname, NULL, zcp_eval_sync,
	zcp_eval_sig, &runinfo, 0, ZFS_SPACE_CHECK_ZCP_EVAL);
	if (err != 0)
	zcp_pool_error(&runinfo, poolname, err);
	} else {
	zcp_eval_open(&runinfo, poolname);
	}
	lua_close(state);

	/*
	* Create device minor nodes for any new zvols.
	*/
	for (nvpair_t *pair = nvlist_next_nvpair(runinfo.zri_new_zvols, NULL);
	pair != NULL;
	pair = nvlist_next_nvpair(runinfo.zri_new_zvols, pair)) {
	zvol_create_minor(nvpair_name(pair));
	}
	fnvlist_free(runinfo.zri_new_zvols);

	return (runinfo.zri_result);
	}

	/*
	* Retrieve metadata about the currently running channel program.
	*/
	zcp_run_info_t *
	zcp_run_info(lua_State *state)
	{
	zcp_run_info_t *ri;

	lua_getfield(state, LUA_REGISTRYINDEX, ZCP_RUN_INFO_KEY);
	ri = lua_touserdata(state, -1);
	lua_pop(state, 1);
	return (ri);
	}

	/*
	* Argument Parsing
	* ================
	*
	* The Lua language allows methods to be called with any number
	* of arguments of any type. When calling back into ZFS we need to sanitize
	* arguments from channel programs to make sure unexpected arguments or
	* arguments of the wrong type result in clear error messages. To do this
	* in a uniform way all callbacks from channel programs should use the
	* zcp_parse_args() function to interpret inputs.
	*
	* Positional vs Keyword Arguments
	* ===============================
	*
	* Every callback function takes a fixed set of required positional arguments
	* and optional keyword arguments. For example, the destroy function takes
	* a single positional string argument (the name of the dataset to destroy)
	* and an optional "defer" keyword boolean argument. When calling lua functions
	* with parentheses, only positional arguments can be used:
	*
	* zfs.sync.snapshot("rpool@snap")
	*
	* To use keyword arguments functions should be called with a single argument
	* that is a lua table containing mappings of integer -> positional arguments
	* and string -> keyword arguments:
	*
	* zfs.sync.snapshot({1="rpool@snap", defer=true})
	*
	* The lua language allows curly braces to be used in place of parenthesis as
	* syntactic sugar for this calling convention:
	*
	* zfs.sync.snapshot{"rpool@snap", defer=true}
	*/

	/*
	* Throw an error and print the given arguments. If there are too many
	* arguments to fit in the output buffer, only the error format string is
	* output.
	*/
	static void
	zcp_args_error(lua_State state, const char fname, const zcp_arg_t *pargs,
	const zcp_arg_t kwargs, const char fmt, ...)
	{
	int i;
	char errmsg[512];
	size_t len = sizeof (errmsg);
	size_t msglen = 0;
	va_list argp;

	va_start(argp, fmt);
	VERIFY3U(len, >, vsnprintf(errmsg, len, fmt, argp));
	va_end(argp);

	/*
	* Calculate the total length of the final string, including extra
	* formatting characters. If the argument dump would be too large,
	* only print the error string.
	*/
	msglen = strlen(errmsg);
	msglen += strlen(fname) + 4; /* : + {} + null terminator */
	for (i = 0; pargs[i].za_name != NULL; i++) {
	msglen += strlen(pargs[i].za_name);
	msglen += strlen(lua_typename(state, pargs[i].za_lua_type));
	if (pargs[i + 1].za_name != NULL \|\| kwargs[0].za_name != NULL)
	msglen += 5; /* < + ( + )> + , */
	else
	msglen += 4; /* < + ( + )> */
	}
	for (i = 0; kwargs[i].za_name != NULL; i++) {
	msglen += strlen(kwargs[i].za_name);
	msglen += strlen(lua_typename(state, kwargs[i].za_lua_type));
	if (kwargs[i + 1].za_name != NULL)
	msglen += 4; /* =( + ) + , */
	else
	msglen += 3; /* =( + ) */
	}

	if (msglen >= len)
	(void) luaL_error(state, errmsg);

	VERIFY3U(len, >, strlcat(errmsg, ": ", len));
	VERIFY3U(len, >, strlcat(errmsg, fname, len));
	VERIFY3U(len, >, strlcat(errmsg, "{", len));
	for (i = 0; pargs[i].za_name != NULL; i++) {
	VERIFY3U(len, >, strlcat(errmsg, "<", len));
	VERIFY3U(len, >, strlcat(errmsg, pargs[i].za_name, len));
	VERIFY3U(len, >, strlcat(errmsg, "(", len));
	VERIFY3U(len, >, strlcat(errmsg,
	lua_typename(state, pargs[i].za_lua_type), len));
	VERIFY3U(len, >, strlcat(errmsg, ")>", len));
	if (pargs[i + 1].za_name != NULL \|\| kwargs[0].za_name != NULL) {
	VERIFY3U(len, >, strlcat(errmsg, ", ", len));
	}
	}
	for (i = 0; kwargs[i].za_name != NULL; i++) {
	VERIFY3U(len, >, strlcat(errmsg, kwargs[i].za_name, len));
	VERIFY3U(len, >, strlcat(errmsg, "=(", len));
	VERIFY3U(len, >, strlcat(errmsg,
	lua_typename(state, kwargs[i].za_lua_type), len));
	VERIFY3U(len, >, strlcat(errmsg, ")", len));
	if (kwargs[i + 1].za_name != NULL) {
	VERIFY3U(len, >, strlcat(errmsg, ", ", len));
	}
	}
	VERIFY3U(len, >, strlcat(errmsg, "}", len));

	(void) luaL_error(state, errmsg);
	panic("unreachable code");
	}

	static void
	zcp_parse_table_args(lua_State state, const char fname,
	const zcp_arg_t pargs, const zcp_arg_t kwargs)
	{
	int i;
	int type;

	for (i = 0; pargs[i].za_name != NULL; i++) {
	/*
	* Check the table for this positional argument, leaving it
	* on the top of the stack once we finish validating it.
	*/
	lua_pushinteger(state, i + 1);
	lua_gettable(state, 1);

	type = lua_type(state, -1);
	if (type == LUA_TNIL) {
	zcp_args_error(state, fname, pargs, kwargs,
	"too few arguments");
	panic("unreachable code");
	} else if (type != pargs[i].za_lua_type) {
	zcp_args_error(state, fname, pargs, kwargs,
	"arg %d wrong type (is '%s', expected '%s')",
	i + 1, lua_typename(state, type),
	lua_typename(state, pargs[i].za_lua_type));
	panic("unreachable code");
	}

	/*
	* Remove the positional argument from the table.
	*/
	lua_pushinteger(state, i + 1);
	lua_pushnil(state);
	lua_settable(state, 1);
	}

	for (i = 0; kwargs[i].za_name != NULL; i++) {
	/*
	* Check the table for this keyword argument, which may be
	* nil if it was omitted. Leave the value on the top of
	* the stack after validating it.
	*/
	lua_getfield(state, 1, kwargs[i].za_name);

	type = lua_type(state, -1);
	if (type != LUA_TNIL && type != kwargs[i].za_lua_type) {
	zcp_args_error(state, fname, pargs, kwargs,
	"kwarg '%s' wrong type (is '%s', expected '%s')",
	kwargs[i].za_name, lua_typename(state, type),
	lua_typename(state, kwargs[i].za_lua_type));
	panic("unreachable code");
	}

	/*
	* Remove the keyword argument from the table.
	*/
	lua_pushnil(state);
	lua_setfield(state, 1, kwargs[i].za_name);
	}

	/*
	* Any entries remaining in the table are invalid inputs, print
	* an error message based on what the entry is.
	*/
	lua_pushnil(state);
	if (lua_next(state, 1)) {
	if (lua_isnumber(state, -2) && lua_tointeger(state, -2) > 0) {
	zcp_args_error(state, fname, pargs, kwargs,
	"too many positional arguments");
	} else if (lua_isstring(state, -2)) {
	zcp_args_error(state, fname, pargs, kwargs,
	"invalid kwarg '%s'", lua_tostring(state, -2));
	} else {
	zcp_args_error(state, fname, pargs, kwargs,
	"kwarg keys must be strings");
	}
	panic("unreachable code");
	}

	lua_remove(state, 1);
	}

	static void
	zcp_parse_pos_args(lua_State state, const char fname, const zcp_arg_t *pargs,
	const zcp_arg_t *kwargs)
	{
	int i;
	int type;

	for (i = 0; pargs[i].za_name != NULL; i++) {
	type = lua_type(state, i + 1);
	if (type == LUA_TNONE) {
	zcp_args_error(state, fname, pargs, kwargs,
	"too few arguments");
	panic("unreachable code");
	} else if (type != pargs[i].za_lua_type) {
	zcp_args_error(state, fname, pargs, kwargs,
	"arg %d wrong type (is '%s', expected '%s')",
	i + 1, lua_typename(state, type),
	lua_typename(state, pargs[i].za_lua_type));
	panic("unreachable code");
	}
	}
	if (lua_gettop(state) != i) {
	zcp_args_error(state, fname, pargs, kwargs,
	"too many positional arguments");
	panic("unreachable code");
	}

	for (i = 0; kwargs[i].za_name != NULL; i++) {
	lua_pushnil(state);
	}
	}

	/*
	* Checks the current Lua stack against an expected set of positional and
	* keyword arguments. If the stack does not match the expected arguments
	* aborts the current channel program with a useful error message, otherwise
	* it re-arranges the stack so that it contains the positional arguments
	* followed by the keyword argument values in declaration order. Any missing
	* keyword argument will be represented by a nil value on the stack.
	*
	* If the stack contains exactly one argument of type LUA_TTABLE the curly
	* braces calling convention is assumed, otherwise the stack is parsed for
	* positional arguments only.
	*
	* This function should be used by every function callback. It should be called
	* before the callback manipulates the Lua stack as it assumes the stack
	* represents the function arguments.
	*/
	void
	zcp_parse_args(lua_State state, const char fname, const zcp_arg_t *pargs,
	const zcp_arg_t *kwargs)
	{
	if (lua_gettop(state) == 1 && lua_istable(state, 1)) {
	zcp_parse_table_args(state, fname, pargs, kwargs);
	} else {
	zcp_parse_pos_args(state, fname, pargs, kwargs);
	}
	}

	ZFS_MODULE_PARAM(zfs_lua, zfs_lua_, max_instrlimit, U64, ZMOD_RW,
	"Max instruction limit that can be specified for a channel program");

	ZFS_MODULE_PARAM(zfs_lua, zfs_lua_, max_memlimit, U64, ZMOD_RW,
	"Max memory limit that can be specified for a channel program");
	diff --git a/sys/contrib/openzfs/module/zfs/zfs_ioctl.c b/sys/contrib/openzfs/module/zfs/zfs_ioctl.c
	index 908b9efc1813..7b527eb75e83 100644
	--- a/sys/contrib/openzfs/module/zfs/zfs_ioctl.c
	+++ b/sys/contrib/openzfs/module/zfs/zfs_ioctl.c
	@@ -1,7955 +1,7992 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Portions Copyright 2011 Martin Matuska
	* Copyright 2015, OmniTI Computer Consulting, Inc. All rights reserved.
	* Copyright (c) 2012 Pawel Jakub Dawidek
	* Copyright (c) 2014, 2016 Joyent, Inc. All rights reserved.
	* Copyright 2016 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2014, Joyent, Inc. All rights reserved.
	* Copyright (c) 2011, 2024 by Delphix. All rights reserved.
	* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
	* Copyright (c) 2013 Steven Hartland. All rights reserved.
	* Copyright (c) 2014 Integros [integros.com]
	* Copyright 2016 Toomas Soome <tsoome@me.com>
	* Copyright (c) 2016 Actifio, Inc. All rights reserved.
	* Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>. All rights reserved.
	* Copyright 2017 RackTop Systems.
	* Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
	* Copyright (c) 2019 Datto Inc.
	* Copyright (c) 2019, 2020 by Christian Schwarz. All rights reserved.
	- * Copyright (c) 2019, 2021, Klara Inc.
	+ * Copyright (c) 2019, 2021, 2024, Klara Inc.
	* Copyright (c) 2019, Allan Jude
	* Copyright 2024 Oxide Computer Company
	*/

	/*
	* ZFS ioctls.
	*
	* This file handles the ioctls to /dev/zfs, used for configuring ZFS storage
	* pools and filesystems, e.g. with /sbin/zfs and /sbin/zpool.
	*
	* There are two ways that we handle ioctls: the legacy way where almost
	* all of the logic is in the ioctl callback, and the new way where most
	* of the marshalling is handled in the common entry point, zfsdev_ioctl().
	*
	* Non-legacy ioctls should be registered by calling
	* zfs_ioctl_register() from zfs_ioctl_init(). The ioctl is invoked
	* from userland by lzc_ioctl().
	*
	* The registration arguments are as follows:
	*
	* const char *name
	* The name of the ioctl. This is used for history logging. If the
	* ioctl returns successfully (the callback returns 0), and allow_log
	* is true, then a history log entry will be recorded with the input &
	* output nvlists. The log entry can be printed with "zpool history -i".
	*
	* zfs_ioc_t ioc
	* The ioctl request number, which userland will pass to ioctl(2).
	* We want newer versions of libzfs and libzfs_core to run against
	* existing zfs kernel modules (i.e. a deferred reboot after an update).
	* Therefore the ioctl numbers cannot change from release to release.
	*
	* zfs_secpolicy_func_t *secpolicy
	* This function will be called before the zfs_ioc_func_t, to
	* determine if this operation is permitted. It should return EPERM
	* on failure, and 0 on success. Checks include determining if the
	* dataset is visible in this zone, and if the user has either all
	* zfs privileges in the zone (SYS_MOUNT), or has been granted permission
	* to do this operation on this dataset with "zfs allow".
	*
	* zfs_ioc_namecheck_t namecheck
	* This specifies what to expect in the zfs_cmd_t:zc_name -- a pool
	* name, a dataset name, or nothing. If the name is not well-formed,
	* the ioctl will fail and the callback will not be called.
	* Therefore, the callback can assume that the name is well-formed
	* (e.g. is null-terminated, doesn't have more than one '@' character,
	* doesn't have invalid characters).
	*
	* zfs_ioc_poolcheck_t pool_check
	* This specifies requirements on the pool state. If the pool does
	* not meet them (is suspended or is readonly), the ioctl will fail
	* and the callback will not be called. If any checks are specified
	* (i.e. it is not POOL_CHECK_NONE), namecheck must not be NO_NAME.
	* Multiple checks can be or-ed together (e.g. POOL_CHECK_SUSPENDED \|
	* POOL_CHECK_READONLY).
	*
	* zfs_ioc_key_t *nvl_keys
	* The list of expected/allowable innvl input keys. This list is used
	* to validate the nvlist input to the ioctl.
	*
	* boolean_t smush_outnvlist
	* If smush_outnvlist is true, then the output is presumed to be a
	* list of errors, and it will be "smushed" down to fit into the
	* caller's buffer, by removing some entries and replacing them with a
	* single "N_MORE_ERRORS" entry indicating how many were removed. See
	* nvlist_smush() for details. If smush_outnvlist is false, and the
	* outnvlist does not fit into the userland-provided buffer, then the
	* ioctl will fail with ENOMEM.
	*
	* zfs_ioc_func_t *func
	* The callback function that will perform the operation.
	*
	* The callback should return 0 on success, or an error number on
	* failure. If the function fails, the userland ioctl will return -1,
	* and errno will be set to the callback's return value. The callback
	* will be called with the following arguments:
	*
	* const char *name
	* The name of the pool or dataset to operate on, from
	* zfs_cmd_t:zc_name. The 'namecheck' argument specifies the
	* expected type (pool, dataset, or none).
	*
	* nvlist_t *innvl
	* The input nvlist, deserialized from zfs_cmd_t:zc_nvlist_src. Or
	* NULL if no input nvlist was provided. Changes to this nvlist are
	* ignored. If the input nvlist could not be deserialized, the
	* ioctl will fail and the callback will not be called.
	*
	* nvlist_t *outnvl
	* The output nvlist, initially empty. The callback can fill it in,
	* and it will be returned to userland by serializing it into
	* zfs_cmd_t:zc_nvlist_dst. If it is non-empty, and serialization
	* fails (e.g. because the caller didn't supply a large enough
	* buffer), then the overall ioctl will fail. See the
	* 'smush_nvlist' argument above for additional behaviors.
	*
	* There are two typical uses of the output nvlist:
	* - To return state, e.g. property values. In this case,
	* smush_outnvlist should be false. If the buffer was not large
	* enough, the caller will reallocate a larger buffer and try
	* the ioctl again.
	*
	* - To return multiple errors from an ioctl which makes on-disk
	* changes. In this case, smush_outnvlist should be true.
	* Ioctls which make on-disk modifications should generally not
	* use the outnvl if they succeed, because the caller can not
	* distinguish between the operation failing, and
	* deserialization failing.
	*
	* IOCTL Interface Errors
	*
	* The following ioctl input errors can be returned:
	* ZFS_ERR_IOC_CMD_UNAVAIL the ioctl number is not supported by kernel
	* ZFS_ERR_IOC_ARG_UNAVAIL an input argument is not supported by kernel
	* ZFS_ERR_IOC_ARG_REQUIRED a required input argument is missing
	* ZFS_ERR_IOC_ARG_BADTYPE an input argument has an invalid type
	*/

	#include <sys/types.h>
	#include <sys/param.h>
	#include <sys/errno.h>
	#include <sys/uio_impl.h>
	#include <sys/file.h>
	#include <sys/kmem.h>
	#include <sys/cmn_err.h>
	#include <sys/stat.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/zfs_quota.h>
	#include <sys/zfs_vfsops.h>
	#include <sys/zfs_znode.h>
	#include <sys/zap.h>
	#include <sys/spa.h>
	#include <sys/spa_impl.h>
	#include <sys/vdev.h>
	#include <sys/vdev_impl.h>
	#include <sys/dmu.h>
	#include <sys/dsl_dir.h>
	#include <sys/dsl_dataset.h>
	#include <sys/dsl_prop.h>
	#include <sys/dsl_deleg.h>
	#include <sys/dmu_objset.h>
	#include <sys/dmu_impl.h>
	#include <sys/dmu_redact.h>
	#include <sys/dmu_tx.h>
	#include <sys/sunddi.h>
	#include <sys/policy.h>
	#include <sys/zone.h>
	#include <sys/nvpair.h>
	#include <sys/pathname.h>
	#include <sys/fs/zfs.h>
	#include <sys/zfs_ctldir.h>
	#include <sys/zfs_dir.h>
	#include <sys/zfs_onexit.h>
	#include <sys/zvol.h>
	#include <sys/dsl_scan.h>
	#include <sys/fm/util.h>
	#include <sys/dsl_crypt.h>
	#include <sys/rrwlock.h>
	#include <sys/zfs_file.h>

	#include <sys/dmu_recv.h>
	#include <sys/dmu_send.h>
	#include <sys/dmu_recv.h>
	#include <sys/dsl_destroy.h>
	#include <sys/dsl_bookmark.h>
	#include <sys/dsl_userhold.h>
	#include <sys/zfeature.h>
	#include <sys/zcp.h>
	#include <sys/zio_checksum.h>
	#include <sys/vdev_removal.h>
	#include <sys/vdev_impl.h>
	#include <sys/vdev_initialize.h>
	#include <sys/vdev_trim.h>

	#include "zfs_namecheck.h"
	#include "zfs_prop.h"
	#include "zfs_deleg.h"
	#include "zfs_comutil.h"

	#include <sys/lua/lua.h>
	#include <sys/lua/lauxlib.h>
	#include <sys/zfs_ioctl_impl.h>

	kmutex_t zfsdev_state_lock;
	static zfsdev_state_t zfsdev_state_listhead;

	/*
	* Limit maximum nvlist size. We don't want users passing in insane values
	* for zc->zc_nvlist_src_size, since we will need to allocate that much memory.
	* Defaults to 0=auto which is handled by platform code.
	*/
	uint64_t zfs_max_nvlist_src_size = 0;

	/*
	* When logging the output nvlist of an ioctl in the on-disk history, limit
	* the logged size to this many bytes. This must be less than DMU_MAX_ACCESS.
	* This applies primarily to zfs_ioc_channel_program().
	*/
	static uint64_t zfs_history_output_max = 1024 * 1024;

	uint_t zfs_allow_log_key;

	/* DATA_TYPE_ANY is used when zkey_type can vary. */
	#define DATA_TYPE_ANY DATA_TYPE_UNKNOWN

	typedef struct zfs_ioc_vec {
	zfs_ioc_legacy_func_t *zvec_legacy_func;
	zfs_ioc_func_t *zvec_func;
	zfs_secpolicy_func_t *zvec_secpolicy;
	zfs_ioc_namecheck_t zvec_namecheck;
	boolean_t zvec_allow_log;
	zfs_ioc_poolcheck_t zvec_pool_check;
	boolean_t zvec_smush_outnvlist;
	const char *zvec_name;
	const zfs_ioc_key_t *zvec_nvl_keys;
	size_t zvec_nvl_key_count;
	} zfs_ioc_vec_t;

	/* This array is indexed by zfs_userquota_prop_t */
	static const char *userquota_perms[] = {
	ZFS_DELEG_PERM_USERUSED,
	ZFS_DELEG_PERM_USERQUOTA,
	ZFS_DELEG_PERM_GROUPUSED,
	ZFS_DELEG_PERM_GROUPQUOTA,
	ZFS_DELEG_PERM_USEROBJUSED,
	ZFS_DELEG_PERM_USEROBJQUOTA,
	ZFS_DELEG_PERM_GROUPOBJUSED,
	ZFS_DELEG_PERM_GROUPOBJQUOTA,
	ZFS_DELEG_PERM_PROJECTUSED,
	ZFS_DELEG_PERM_PROJECTQUOTA,
	ZFS_DELEG_PERM_PROJECTOBJUSED,
	ZFS_DELEG_PERM_PROJECTOBJQUOTA,
	};

	static int zfs_ioc_userspace_upgrade(zfs_cmd_t *zc);
	static int zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc);
	static int zfs_check_settable(const char name, nvpair_t property,
	cred_t *cr);
	static int zfs_check_clearable(const char dataset, nvlist_t props,
	nvlist_t **errors);
	static int zfs_fill_zplprops_root(uint64_t, nvlist_t , nvlist_t ,
	boolean_t *);
	int zfs_set_prop_nvlist(const char , zprop_source_t, nvlist_t , nvlist_t *);
	static int get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp);

	static void
	history_str_free(char *buf)
	{
	kmem_free(buf, HIS_MAX_RECORD_LEN);
	}

	static char *
	history_str_get(zfs_cmd_t *zc)
	{
	char *buf;

	if (zc->zc_history == 0)
	return (NULL);

	buf = kmem_alloc(HIS_MAX_RECORD_LEN, KM_SLEEP);
	if (copyinstr((void *)(uintptr_t)zc->zc_history,
	buf, HIS_MAX_RECORD_LEN, NULL) != 0) {
	history_str_free(buf);
	return (NULL);
	}

	buf[HIS_MAX_RECORD_LEN -1] = '\0';

	return (buf);
	}

	/*
	* Return non-zero if the spa version is less than requested version.
	*/
	static int
	zfs_earlier_version(const char *name, int version)
	{
	spa_t *spa;

	if (spa_open(name, &spa, FTAG) == 0) {
	if (spa_version(spa) < version) {
	spa_close(spa, FTAG);
	return (1);
	}
	spa_close(spa, FTAG);
	}
	return (0);
	}

	/*
	* Return TRUE if the ZPL version is less than requested version.
	*/
	static boolean_t
	zpl_earlier_version(const char *name, int version)
	{
	objset_t *os;
	boolean_t rc = B_TRUE;

	if (dmu_objset_hold(name, FTAG, &os) == 0) {
	uint64_t zplversion;

	if (dmu_objset_type(os) != DMU_OST_ZFS) {
	dmu_objset_rele(os, FTAG);
	return (B_TRUE);
	}
	/* XXX reading from non-owned objset */
	if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &zplversion) == 0)
	rc = zplversion < version;
	dmu_objset_rele(os, FTAG);
	}
	return (rc);
	}

	static void
	zfs_log_history(zfs_cmd_t *zc)
	{
	spa_t *spa;
	char *buf;

	if ((buf = history_str_get(zc)) == NULL)
	return;

	if (spa_open(zc->zc_name, &spa, FTAG) == 0) {
	if (spa_version(spa) >= SPA_VERSION_ZPOOL_HISTORY)
	(void) spa_history_log(spa, buf);
	spa_close(spa, FTAG);
	}
	history_str_free(buf);
	}

	/*
	* Policy for top-level read operations (list pools). Requires no privileges,
	* and can be used in the local zone, as there is no associated dataset.
	*/
	static int
	zfs_secpolicy_none(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc, (void) innvl, (void) cr;
	return (0);
	}

	/*
	* Policy for dataset read operations (list children, get statistics). Requires
	* no privileges, but must be visible in the local zone.
	*/
	static int
	zfs_secpolicy_read(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) innvl, (void) cr;
	if (INGLOBALZONE(curproc) \|\|
	zone_dataset_visible(zc->zc_name, NULL))
	return (0);

	return (SET_ERROR(ENOENT));
	}

	static int
	zfs_dozonecheck_impl(const char dataset, uint64_t zoned, cred_t cr)
	{
	int writable = 1;

	/*
	* The dataset must be visible by this zone -- check this first
	* so they don't see EPERM on something they shouldn't know about.
	*/
	if (!INGLOBALZONE(curproc) &&
	!zone_dataset_visible(dataset, &writable))
	return (SET_ERROR(ENOENT));

	if (INGLOBALZONE(curproc)) {
	/*
	* If the fs is zoned, only root can access it from the
	* global zone.
	*/
	if (secpolicy_zfs(cr) && zoned)
	return (SET_ERROR(EPERM));
	} else {
	/*
	* If we are in a local zone, the 'zoned' property must be set.
	*/
	if (!zoned)
	return (SET_ERROR(EPERM));

	/* must be writable by this zone */
	if (!writable)
	return (SET_ERROR(EPERM));
	}
	return (0);
	}

	static int
	zfs_dozonecheck(const char dataset, cred_t cr)
	{
	uint64_t zoned;

	if (dsl_prop_get_integer(dataset, zfs_prop_to_name(ZFS_PROP_ZONED),
	&zoned, NULL))
	return (SET_ERROR(ENOENT));

	return (zfs_dozonecheck_impl(dataset, zoned, cr));
	}

	static int
	zfs_dozonecheck_ds(const char dataset, dsl_dataset_t ds, cred_t *cr)
	{
	uint64_t zoned;

	if (dsl_prop_get_int_ds(ds, zfs_prop_to_name(ZFS_PROP_ZONED), &zoned))
	return (SET_ERROR(ENOENT));

	return (zfs_dozonecheck_impl(dataset, zoned, cr));
	}

	static int
	zfs_secpolicy_write_perms_ds(const char name, dsl_dataset_t ds,
	const char perm, cred_t cr)
	{
	int error;

	error = zfs_dozonecheck_ds(name, ds, cr);
	if (error == 0) {
	error = secpolicy_zfs(cr);
	if (error != 0)
	error = dsl_deleg_access_impl(ds, perm, cr);
	}
	return (error);
	}

	static int
	zfs_secpolicy_write_perms(const char name, const char perm, cred_t *cr)
	{
	int error;
	dsl_dataset_t *ds;
	dsl_pool_t *dp;

	/*
	* First do a quick check for root in the global zone, which
	* is allowed to do all write_perms. This ensures that zfs_ioc_*
	* will get to handle nonexistent datasets.
	*/
	if (INGLOBALZONE(curproc) && secpolicy_zfs(cr) == 0)
	return (0);

	error = dsl_pool_hold(name, FTAG, &dp);
	if (error != 0)
	return (error);

	error = dsl_dataset_hold(dp, name, FTAG, &ds);
	if (error != 0) {
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	error = zfs_secpolicy_write_perms_ds(name, ds, perm, cr);

	dsl_dataset_rele(ds, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	/*
	* Policy for setting the security label property.
	*
	* Returns 0 for success, non-zero for access and other errors.
	*/
	static int
	zfs_set_slabel_policy(const char name, const char strval, cred_t *cr)
	{
	#ifdef HAVE_MLSLABEL
	char ds_hexsl[MAXNAMELEN];
	bslabel_t ds_sl, new_sl;
	boolean_t new_default = FALSE;
	uint64_t zoned;
	int needed_priv = -1;
	int error;

	/* First get the existing dataset label. */
	error = dsl_prop_get(name, zfs_prop_to_name(ZFS_PROP_MLSLABEL),
	1, sizeof (ds_hexsl), &ds_hexsl, NULL);
	if (error != 0)
	return (SET_ERROR(EPERM));

	if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
	new_default = TRUE;

	/* The label must be translatable */
	if (!new_default && (hexstr_to_label(strval, &new_sl) != 0))
	return (SET_ERROR(EINVAL));

	/*
	* In a non-global zone, disallow attempts to set a label that
	* doesn't match that of the zone; otherwise no other checks
	* are needed.
	*/
	if (!INGLOBALZONE(curproc)) {
	if (new_default \|\| !blequal(&new_sl, CR_SL(CRED())))
	return (SET_ERROR(EPERM));
	return (0);
	}

	/*
	* For global-zone datasets (i.e., those whose zoned property is
	* "off", verify that the specified new label is valid for the
	* global zone.
	*/
	if (dsl_prop_get_integer(name,
	zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL))
	return (SET_ERROR(EPERM));
	if (!zoned) {
	if (zfs_check_global_label(name, strval) != 0)
	return (SET_ERROR(EPERM));
	}

	/*
	* If the existing dataset label is nondefault, check if the
	* dataset is mounted (label cannot be changed while mounted).
	* Get the zfsvfs_t; if there isn't one, then the dataset isn't
	* mounted (or isn't a dataset, doesn't exist, ...).
	*/
	if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) != 0) {
	objset_t *os;
	static const char *setsl_tag = "setsl_tag";

	/*
	* Try to own the dataset; abort if there is any error,
	* (e.g., already mounted, in use, or other error).
	*/
	error = dmu_objset_own(name, DMU_OST_ZFS, B_TRUE, B_TRUE,
	setsl_tag, &os);
	if (error != 0)
	return (SET_ERROR(EPERM));

	dmu_objset_disown(os, B_TRUE, setsl_tag);

	if (new_default) {
	needed_priv = PRIV_FILE_DOWNGRADE_SL;
	goto out_check;
	}

	if (hexstr_to_label(strval, &new_sl) != 0)
	return (SET_ERROR(EPERM));

	if (blstrictdom(&ds_sl, &new_sl))
	needed_priv = PRIV_FILE_DOWNGRADE_SL;
	else if (blstrictdom(&new_sl, &ds_sl))
	needed_priv = PRIV_FILE_UPGRADE_SL;
	} else {
	/* dataset currently has a default label */
	if (!new_default)
	needed_priv = PRIV_FILE_UPGRADE_SL;
	}

	out_check:
	if (needed_priv != -1)
	return (PRIV_POLICY(cr, needed_priv, B_FALSE, EPERM, NULL));
	return (0);
	#else
	return (SET_ERROR(ENOTSUP));
	#endif /* HAVE_MLSLABEL */
	}

	static int
	zfs_secpolicy_setprop(const char dsname, zfs_prop_t prop, nvpair_t propval,
	cred_t *cr)
	{
	const char *strval;

	/*
	* Check permissions for special properties.
	*/
	switch (prop) {
	default:
	break;
	case ZFS_PROP_ZONED:
	/*
	* Disallow setting of 'zoned' from within a local zone.
	*/
	if (!INGLOBALZONE(curproc))
	return (SET_ERROR(EPERM));
	break;

	case ZFS_PROP_QUOTA:
	case ZFS_PROP_FILESYSTEM_LIMIT:
	case ZFS_PROP_SNAPSHOT_LIMIT:
	if (!INGLOBALZONE(curproc)) {
	uint64_t zoned;
	char setpoint[ZFS_MAX_DATASET_NAME_LEN];
	/*
	* Unprivileged users are allowed to modify the
	* limit on things under (ie. contained by)
	* the thing they own.
	*/
	if (dsl_prop_get_integer(dsname,
	zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, setpoint))
	return (SET_ERROR(EPERM));
	if (!zoned \|\| strlen(dsname) <= strlen(setpoint))
	return (SET_ERROR(EPERM));
	}
	break;

	case ZFS_PROP_MLSLABEL:
	if (!is_system_labeled())
	return (SET_ERROR(EPERM));

	if (nvpair_value_string(propval, &strval) == 0) {
	int err;

	err = zfs_set_slabel_policy(dsname, strval, CRED());
	if (err != 0)
	return (err);
	}
	break;
	}

	return (zfs_secpolicy_write_perms(dsname, zfs_prop_to_name(prop), cr));
	}

	static int
	zfs_secpolicy_set_fsacl(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	/*
	* permission to set permissions will be evaluated later in
	* dsl_deleg_can_allow()
	*/
	(void) innvl;
	return (zfs_dozonecheck(zc->zc_name, cr));
	}

	static int
	zfs_secpolicy_rollback(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) innvl;
	return (zfs_secpolicy_write_perms(zc->zc_name,
	ZFS_DELEG_PERM_ROLLBACK, cr));
	}

	static int
	zfs_secpolicy_send(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) innvl;
	dsl_pool_t *dp;
	dsl_dataset_t *ds;
	const char *cp;
	int error;

	/*
	* Generate the current snapshot name from the given objsetid, then
	* use that name for the secpolicy/zone checks.
	*/
	cp = strchr(zc->zc_name, '@');
	if (cp == NULL)
	return (SET_ERROR(EINVAL));
	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
	if (error != 0)
	return (error);

	error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &ds);
	if (error != 0) {
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	dsl_dataset_name(ds, zc->zc_name);

	error = zfs_secpolicy_write_perms_ds(zc->zc_name, ds,
	ZFS_DELEG_PERM_SEND, cr);
	dsl_dataset_rele(ds, FTAG);
	dsl_pool_rele(dp, FTAG);

	return (error);
	}

	static int
	zfs_secpolicy_send_new(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) innvl;
	return (zfs_secpolicy_write_perms(zc->zc_name,
	ZFS_DELEG_PERM_SEND, cr));
	}

	static int
	zfs_secpolicy_share(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc, (void) innvl, (void) cr;
	return (SET_ERROR(ENOTSUP));
	}

	static int
	zfs_secpolicy_smb_acl(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc, (void) innvl, (void) cr;
	return (SET_ERROR(ENOTSUP));
	}

	static int
	zfs_get_parent(const char datasetname, char parent, int parentsize)
	{
	char *cp;

	/*
	* Remove the @bla or /bla from the end of the name to get the parent.
	*/
	(void) strlcpy(parent, datasetname, parentsize);
	cp = strrchr(parent, '@');
	if (cp != NULL) {
	cp[0] = '\0';
	} else {
	cp = strrchr(parent, '/');
	if (cp == NULL)
	return (SET_ERROR(ENOENT));
	cp[0] = '\0';
	}

	return (0);
	}

	int
	zfs_secpolicy_destroy_perms(const char name, cred_t cr)
	{
	int error;

	if ((error = zfs_secpolicy_write_perms(name,
	ZFS_DELEG_PERM_MOUNT, cr)) != 0)
	return (error);

	return (zfs_secpolicy_write_perms(name, ZFS_DELEG_PERM_DESTROY, cr));
	}

	static int
	zfs_secpolicy_destroy(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) innvl;
	return (zfs_secpolicy_destroy_perms(zc->zc_name, cr));
	}

	/*
	* Destroying snapshots with delegated permissions requires
	* descendant mount and destroy permissions.
	*/
	static int
	zfs_secpolicy_destroy_snaps(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc;
	nvlist_t *snaps;
	nvpair_t pair, nextpair;
	int error = 0;

	snaps = fnvlist_lookup_nvlist(innvl, "snaps");

	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
	pair = nextpair) {
	nextpair = nvlist_next_nvpair(snaps, pair);
	error = zfs_secpolicy_destroy_perms(nvpair_name(pair), cr);
	if (error == ENOENT) {
	/*
	* Ignore any snapshots that don't exist (we consider
	* them "already destroyed"). Remove the name from the
	* nvl here in case the snapshot is created between
	* now and when we try to destroy it (in which case
	* we don't want to destroy it since we haven't
	* checked for permission).
	*/
	fnvlist_remove_nvpair(snaps, pair);
	error = 0;
	}
	if (error != 0)
	break;
	}

	return (error);
	}

	int
	zfs_secpolicy_rename_perms(const char from, const char to, cred_t *cr)
	{
	char parentname[ZFS_MAX_DATASET_NAME_LEN];
	int error;

	if ((error = zfs_secpolicy_write_perms(from,
	ZFS_DELEG_PERM_RENAME, cr)) != 0)
	return (error);

	if ((error = zfs_secpolicy_write_perms(from,
	ZFS_DELEG_PERM_MOUNT, cr)) != 0)
	return (error);

	if ((error = zfs_get_parent(to, parentname,
	sizeof (parentname))) != 0)
	return (error);

	if ((error = zfs_secpolicy_write_perms(parentname,
	ZFS_DELEG_PERM_CREATE, cr)) != 0)
	return (error);

	if ((error = zfs_secpolicy_write_perms(parentname,
	ZFS_DELEG_PERM_MOUNT, cr)) != 0)
	return (error);

	return (error);
	}

	static int
	zfs_secpolicy_rename(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) innvl;
	return (zfs_secpolicy_rename_perms(zc->zc_name, zc->zc_value, cr));
	}

	static int
	zfs_secpolicy_promote(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) innvl;
	dsl_pool_t *dp;
	dsl_dataset_t *clone;
	int error;

	error = zfs_secpolicy_write_perms(zc->zc_name,
	ZFS_DELEG_PERM_PROMOTE, cr);
	if (error != 0)
	return (error);

	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
	if (error != 0)
	return (error);

	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &clone);

	if (error == 0) {
	char parentname[ZFS_MAX_DATASET_NAME_LEN];
	dsl_dataset_t *origin = NULL;
	dsl_dir_t *dd;
	dd = clone->ds_dir;

	error = dsl_dataset_hold_obj(dd->dd_pool,
	dsl_dir_phys(dd)->dd_origin_obj, FTAG, &origin);
	if (error != 0) {
	dsl_dataset_rele(clone, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	error = zfs_secpolicy_write_perms_ds(zc->zc_name, clone,
	ZFS_DELEG_PERM_MOUNT, cr);

	dsl_dataset_name(origin, parentname);
	if (error == 0) {
	error = zfs_secpolicy_write_perms_ds(parentname, origin,
	ZFS_DELEG_PERM_PROMOTE, cr);
	}
	dsl_dataset_rele(clone, FTAG);
	dsl_dataset_rele(origin, FTAG);
	}
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	static int
	zfs_secpolicy_recv(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) innvl;
	int error;

	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
	ZFS_DELEG_PERM_RECEIVE, cr)) != 0)
	return (error);

	if ((error = zfs_secpolicy_write_perms(zc->zc_name,
	ZFS_DELEG_PERM_MOUNT, cr)) != 0)
	return (error);

	return (zfs_secpolicy_write_perms(zc->zc_name,
	ZFS_DELEG_PERM_CREATE, cr));
	}

	int
	zfs_secpolicy_snapshot_perms(const char name, cred_t cr)
	{
	return (zfs_secpolicy_write_perms(name,
	ZFS_DELEG_PERM_SNAPSHOT, cr));
	}

	/*
	* Check for permission to create each snapshot in the nvlist.
	*/
	static int
	zfs_secpolicy_snapshot(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc;
	nvlist_t *snaps;
	int error = 0;
	nvpair_t *pair;

	snaps = fnvlist_lookup_nvlist(innvl, "snaps");

	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
	pair = nvlist_next_nvpair(snaps, pair)) {
	char name = (char )nvpair_name(pair);
	char *atp = strchr(name, '@');

	if (atp == NULL) {
	error = SET_ERROR(EINVAL);
	break;
	}
	*atp = '\0';
	error = zfs_secpolicy_snapshot_perms(name, cr);
	*atp = '@';
	if (error != 0)
	break;
	}
	return (error);
	}

	/*
	* Check for permission to create each bookmark in the nvlist.
	*/
	static int
	zfs_secpolicy_bookmark(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc;
	int error = 0;

	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
	pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
	char name = (char )nvpair_name(pair);
	char *hashp = strchr(name, '#');

	if (hashp == NULL) {
	error = SET_ERROR(EINVAL);
	break;
	}
	*hashp = '\0';
	error = zfs_secpolicy_write_perms(name,
	ZFS_DELEG_PERM_BOOKMARK, cr);
	*hashp = '#';
	if (error != 0)
	break;
	}
	return (error);
	}

	static int
	zfs_secpolicy_destroy_bookmarks(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc;
	nvpair_t pair, nextpair;
	int error = 0;

	for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
	pair = nextpair) {
	char name = (char )nvpair_name(pair);
	char *hashp = strchr(name, '#');
	nextpair = nvlist_next_nvpair(innvl, pair);

	if (hashp == NULL) {
	error = SET_ERROR(EINVAL);
	break;
	}

	*hashp = '\0';
	error = zfs_secpolicy_write_perms(name,
	ZFS_DELEG_PERM_DESTROY, cr);
	*hashp = '#';
	if (error == ENOENT) {
	/*
	* Ignore any filesystems that don't exist (we consider
	* their bookmarks "already destroyed"). Remove
	* the name from the nvl here in case the filesystem
	* is created between now and when we try to destroy
	* the bookmark (in which case we don't want to
	* destroy it since we haven't checked for permission).
	*/
	fnvlist_remove_nvpair(innvl, pair);
	error = 0;
	}
	if (error != 0)
	break;
	}

	return (error);
	}

	static int
	zfs_secpolicy_log_history(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc, (void) innvl, (void) cr;
	/*
	* Even root must have a proper TSD so that we know what pool
	* to log to.
	*/
	if (tsd_get(zfs_allow_log_key) == NULL)
	return (SET_ERROR(EPERM));
	return (0);
	}

	static int
	zfs_secpolicy_create_clone(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	char parentname[ZFS_MAX_DATASET_NAME_LEN];
	int error;
	const char *origin;

	if ((error = zfs_get_parent(zc->zc_name, parentname,
	sizeof (parentname))) != 0)
	return (error);

	if (nvlist_lookup_string(innvl, "origin", &origin) == 0 &&
	(error = zfs_secpolicy_write_perms(origin,
	ZFS_DELEG_PERM_CLONE, cr)) != 0)
	return (error);

	if ((error = zfs_secpolicy_write_perms(parentname,
	ZFS_DELEG_PERM_CREATE, cr)) != 0)
	return (error);

	return (zfs_secpolicy_write_perms(parentname,
	ZFS_DELEG_PERM_MOUNT, cr));
	}

	/*
	* Policy for pool operations - create/destroy pools, add vdevs, etc. Requires
	* SYS_CONFIG privilege, which is not available in a local zone.
	*/
	int
	zfs_secpolicy_config(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc, (void) innvl;

	if (secpolicy_sys_config(cr, B_FALSE) != 0)
	return (SET_ERROR(EPERM));

	return (0);
	}

	/*
	* Policy for object to name lookups.
	*/
	static int
	zfs_secpolicy_diff(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) innvl;
	int error;

	if (secpolicy_sys_config(cr, B_FALSE) == 0)
	return (0);

	error = zfs_secpolicy_write_perms(zc->zc_name, ZFS_DELEG_PERM_DIFF, cr);
	return (error);
	}

	/*
	* Policy for fault injection. Requires all privileges.
	*/
	static int
	zfs_secpolicy_inject(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc, (void) innvl;
	return (secpolicy_zinject(cr));
	}

	static int
	zfs_secpolicy_inherit_prop(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) innvl;
	zfs_prop_t prop = zfs_name_to_prop(zc->zc_value);

	if (prop == ZPROP_USERPROP) {
	if (!zfs_prop_user(zc->zc_value))
	return (SET_ERROR(EINVAL));
	return (zfs_secpolicy_write_perms(zc->zc_name,
	ZFS_DELEG_PERM_USERPROP, cr));
	} else {
	return (zfs_secpolicy_setprop(zc->zc_name, prop,
	NULL, cr));
	}
	}

	static int
	zfs_secpolicy_userspace_one(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	int err = zfs_secpolicy_read(zc, innvl, cr);
	if (err)
	return (err);

	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
	return (SET_ERROR(EINVAL));

	if (zc->zc_value[0] == 0) {
	/*
	* They are asking about a posix uid/gid. If it's
	* themself, allow it.
	*/
	if (zc->zc_objset_type == ZFS_PROP_USERUSED \|\|
	zc->zc_objset_type == ZFS_PROP_USERQUOTA \|\|
	zc->zc_objset_type == ZFS_PROP_USEROBJUSED \|\|
	zc->zc_objset_type == ZFS_PROP_USEROBJQUOTA) {
	if (zc->zc_guid == crgetuid(cr))
	return (0);
	} else if (zc->zc_objset_type == ZFS_PROP_GROUPUSED \|\|
	zc->zc_objset_type == ZFS_PROP_GROUPQUOTA \|\|
	zc->zc_objset_type == ZFS_PROP_GROUPOBJUSED \|\|
	zc->zc_objset_type == ZFS_PROP_GROUPOBJQUOTA) {
	if (groupmember(zc->zc_guid, cr))
	return (0);
	}
	/* else is for project quota/used */
	}

	return (zfs_secpolicy_write_perms(zc->zc_name,
	userquota_perms[zc->zc_objset_type], cr));
	}

	static int
	zfs_secpolicy_userspace_many(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	int err = zfs_secpolicy_read(zc, innvl, cr);
	if (err)
	return (err);

	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
	return (SET_ERROR(EINVAL));

	return (zfs_secpolicy_write_perms(zc->zc_name,
	userquota_perms[zc->zc_objset_type], cr));
	}

	static int
	zfs_secpolicy_userspace_upgrade(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) innvl;
	return (zfs_secpolicy_setprop(zc->zc_name, ZFS_PROP_VERSION,
	NULL, cr));
	}

	static int
	zfs_secpolicy_hold(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc;
	nvpair_t *pair;
	nvlist_t *holds;
	int error;

	holds = fnvlist_lookup_nvlist(innvl, "holds");

	for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
	pair = nvlist_next_nvpair(holds, pair)) {
	char fsname[ZFS_MAX_DATASET_NAME_LEN];
	error = dmu_fsname(nvpair_name(pair), fsname);
	if (error != 0)
	return (error);
	error = zfs_secpolicy_write_perms(fsname,
	ZFS_DELEG_PERM_HOLD, cr);
	if (error != 0)
	return (error);
	}
	return (0);
	}

	static int
	zfs_secpolicy_release(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	(void) zc;
	nvpair_t *pair;
	int error;

	for (pair = nvlist_next_nvpair(innvl, NULL); pair != NULL;
	pair = nvlist_next_nvpair(innvl, pair)) {
	char fsname[ZFS_MAX_DATASET_NAME_LEN];
	error = dmu_fsname(nvpair_name(pair), fsname);
	if (error != 0)
	return (error);
	error = zfs_secpolicy_write_perms(fsname,
	ZFS_DELEG_PERM_RELEASE, cr);
	if (error != 0)
	return (error);
	}
	return (0);
	}

	/*
	* Policy for allowing temporary snapshots to be taken or released
	*/
	static int
	zfs_secpolicy_tmp_snapshot(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	/*
	* A temporary snapshot is the same as a snapshot,
	* hold, destroy and release all rolled into one.
	* Delegated diff alone is sufficient that we allow this.
	*/
	int error;

	if (zfs_secpolicy_write_perms(zc->zc_name,
	ZFS_DELEG_PERM_DIFF, cr) == 0)
	return (0);

	error = zfs_secpolicy_snapshot_perms(zc->zc_name, cr);

	if (innvl != NULL) {
	if (error == 0)
	error = zfs_secpolicy_hold(zc, innvl, cr);
	if (error == 0)
	error = zfs_secpolicy_release(zc, innvl, cr);
	if (error == 0)
	error = zfs_secpolicy_destroy(zc, innvl, cr);
	}
	return (error);
	}

	static int
	zfs_secpolicy_load_key(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	return (zfs_secpolicy_write_perms(zc->zc_name,
	ZFS_DELEG_PERM_LOAD_KEY, cr));
	}

	static int
	zfs_secpolicy_change_key(zfs_cmd_t zc, nvlist_t innvl, cred_t *cr)
	{
	return (zfs_secpolicy_write_perms(zc->zc_name,
	ZFS_DELEG_PERM_CHANGE_KEY, cr));
	}

	/*
	* Returns the nvlist as specified by the user in the zfs_cmd_t.
	*/
	static int
	get_nvlist(uint64_t nvl, uint64_t size, int iflag, nvlist_t **nvp)
	{
	char *packed;
	int error;
	nvlist_t *list = NULL;

	/*
	* Read in and unpack the user-supplied nvlist.
	*/
	if (size == 0)
	return (SET_ERROR(EINVAL));

	packed = vmem_alloc(size, KM_SLEEP);

	if (ddi_copyin((void *)(uintptr_t)nvl, packed, size, iflag) != 0) {
	vmem_free(packed, size);
	return (SET_ERROR(EFAULT));
	}

	if ((error = nvlist_unpack(packed, size, &list, 0)) != 0) {
	vmem_free(packed, size);
	return (error);
	}

	vmem_free(packed, size);

	*nvp = list;
	return (0);
	}

	/*
	* Reduce the size of this nvlist until it can be serialized in 'max' bytes.
	* Entries will be removed from the end of the nvlist, and one int32 entry
	* named "N_MORE_ERRORS" will be added indicating how many entries were
	* removed.
	*/
	static int
	nvlist_smush(nvlist_t *errors, size_t max)
	{
	size_t size;

	size = fnvlist_size(errors);

	if (size > max) {
	nvpair_t *more_errors;
	int n = 0;

	if (max < 1024)
	return (SET_ERROR(ENOMEM));

	fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, 0);
	more_errors = nvlist_prev_nvpair(errors, NULL);

	do {
	nvpair_t *pair = nvlist_prev_nvpair(errors,
	more_errors);
	fnvlist_remove_nvpair(errors, pair);
	n++;
	size = fnvlist_size(errors);
	} while (size > max);

	fnvlist_remove_nvpair(errors, more_errors);
	fnvlist_add_int32(errors, ZPROP_N_MORE_ERRORS, n);
	ASSERT3U(fnvlist_size(errors), <=, max);
	}

	return (0);
	}

	static int
	put_nvlist(zfs_cmd_t zc, nvlist_t nvl)
	{
	char *packed = NULL;
	int error = 0;
	size_t size;

	size = fnvlist_size(nvl);

	if (size > zc->zc_nvlist_dst_size) {
	error = SET_ERROR(ENOMEM);
	} else {
	packed = fnvlist_pack(nvl, &size);
	if (ddi_copyout(packed, (void *)(uintptr_t)zc->zc_nvlist_dst,
	size, zc->zc_iflags) != 0)
	error = SET_ERROR(EFAULT);
	fnvlist_pack_free(packed, size);
	}

	zc->zc_nvlist_dst_size = size;
	zc->zc_nvlist_dst_filled = B_TRUE;
	return (error);
	}

	int
	getzfsvfs_impl(objset_t os, zfsvfs_t *zfvp)
	{
	int error = 0;
	if (dmu_objset_type(os) != DMU_OST_ZFS) {
	return (SET_ERROR(EINVAL));
	}

	mutex_enter(&os->os_user_ptr_lock);
	*zfvp = dmu_objset_get_user(os);
	/* bump s_active only when non-zero to prevent umount race */
	error = zfs_vfs_ref(zfvp);
	mutex_exit(&os->os_user_ptr_lock);
	return (error);
	}

	int
	getzfsvfs(const char dsname, zfsvfs_t *zfvp)
	{
	objset_t *os;
	int error;

	error = dmu_objset_hold(dsname, FTAG, &os);
	if (error != 0)
	return (error);

	error = getzfsvfs_impl(os, zfvp);
	dmu_objset_rele(os, FTAG);
	return (error);
	}

	/*
	* Find a zfsvfs_t for a mounted filesystem, or create our own, in which
	* case its z_sb will be NULL, and it will be opened as the owner.
	* If 'writer' is set, the z_teardown_lock will be held for RW_WRITER,
	* which prevents all inode ops from running.
	*/
	static int
	zfsvfs_hold(const char name, const void tag, zfsvfs_t **zfvp,
	boolean_t writer)
	{
	int error = 0;

	if (getzfsvfs(name, zfvp) != 0)
	error = zfsvfs_create(name, B_FALSE, zfvp);
	if (error == 0) {
	if (writer)
	ZFS_TEARDOWN_ENTER_WRITE(*zfvp, tag);
	else
	ZFS_TEARDOWN_ENTER_READ(*zfvp, tag);
	if ((*zfvp)->z_unmounted) {
	/*
	* XXX we could probably try again, since the unmounting
	* thread should be just about to disassociate the
	* objset from the zfsvfs.
	*/
	ZFS_TEARDOWN_EXIT(*zfvp, tag);
	return (SET_ERROR(EBUSY));
	}
	}
	return (error);
	}

	static void
	zfsvfs_rele(zfsvfs_t zfsvfs, const void tag)
	{
	ZFS_TEARDOWN_EXIT(zfsvfs, tag);

	if (zfs_vfs_held(zfsvfs)) {
	zfs_vfs_rele(zfsvfs);
	} else {
	dmu_objset_disown(zfsvfs->z_os, B_TRUE, zfsvfs);
	zfsvfs_free(zfsvfs);
	}
	}

	static int
	zfs_ioc_pool_create(zfs_cmd_t *zc)
	{
	int error;
	nvlist_t config, props = NULL;
	nvlist_t *rootprops = NULL;
	nvlist_t *zplprops = NULL;
	dsl_crypto_params_t *dcp = NULL;
	const char *spa_name = zc->zc_name;
	boolean_t unload_wkey = B_TRUE;

	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
	zc->zc_iflags, &config)))
	return (error);

	if (zc->zc_nvlist_src_size != 0 && (error =
	get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
	zc->zc_iflags, &props))) {
	nvlist_free(config);
	return (error);
	}

	if (props) {
	nvlist_t *nvl = NULL;
	nvlist_t *hidden_args = NULL;
	uint64_t version = SPA_VERSION;
	const char *tname;

	(void) nvlist_lookup_uint64(props,
	zpool_prop_to_name(ZPOOL_PROP_VERSION), &version);
	if (!SPA_VERSION_IS_SUPPORTED(version)) {
	error = SET_ERROR(EINVAL);
	goto pool_props_bad;
	}
	(void) nvlist_lookup_nvlist(props, ZPOOL_ROOTFS_PROPS, &nvl);
	if (nvl) {
	error = nvlist_dup(nvl, &rootprops, KM_SLEEP);
	if (error != 0)
	goto pool_props_bad;
	(void) nvlist_remove_all(props, ZPOOL_ROOTFS_PROPS);
	}

	(void) nvlist_lookup_nvlist(props, ZPOOL_HIDDEN_ARGS,
	&hidden_args);
	error = dsl_crypto_params_create_nvlist(DCP_CMD_NONE,
	rootprops, hidden_args, &dcp);
	if (error != 0)
	goto pool_props_bad;
	(void) nvlist_remove_all(props, ZPOOL_HIDDEN_ARGS);

	VERIFY(nvlist_alloc(&zplprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
	error = zfs_fill_zplprops_root(version, rootprops,
	zplprops, NULL);
	if (error != 0)
	goto pool_props_bad;

	if (nvlist_lookup_string(props,
	zpool_prop_to_name(ZPOOL_PROP_TNAME), &tname) == 0)
	spa_name = tname;
	}

	error = spa_create(zc->zc_name, config, props, zplprops, dcp);

	/*
	* Set the remaining root properties
	*/
	if (!error && (error = zfs_set_prop_nvlist(spa_name,
	ZPROP_SRC_LOCAL, rootprops, NULL)) != 0) {
	(void) spa_destroy(spa_name);
	unload_wkey = B_FALSE; /* spa_destroy() unloads wrapping keys */
	}

	pool_props_bad:
	nvlist_free(rootprops);
	nvlist_free(zplprops);
	nvlist_free(config);
	nvlist_free(props);
	dsl_crypto_params_free(dcp, unload_wkey && !!error);

	return (error);
	}

	static int
	zfs_ioc_pool_destroy(zfs_cmd_t *zc)
	{
	int error;
	zfs_log_history(zc);
	error = spa_destroy(zc->zc_name);

	return (error);
	}

	static int
	zfs_ioc_pool_import(zfs_cmd_t *zc)
	{
	nvlist_t config, props = NULL;
	uint64_t guid;
	int error;

	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
	zc->zc_iflags, &config)) != 0)
	return (error);

	if (zc->zc_nvlist_src_size != 0 && (error =
	get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
	zc->zc_iflags, &props))) {
	nvlist_free(config);
	return (error);
	}

	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, &guid) != 0 \|\|
	guid != zc->zc_guid)
	error = SET_ERROR(EINVAL);
	else
	error = spa_import(zc->zc_name, config, props, zc->zc_cookie);

	if (zc->zc_nvlist_dst != 0) {
	int err;

	if ((err = put_nvlist(zc, config)) != 0)
	error = err;
	}

	nvlist_free(config);
	nvlist_free(props);

	return (error);
	}

	static int
	zfs_ioc_pool_export(zfs_cmd_t *zc)
	{
	int error;
	boolean_t force = (boolean_t)zc->zc_cookie;
	boolean_t hardforce = (boolean_t)zc->zc_guid;

	zfs_log_history(zc);
	error = spa_export(zc->zc_name, NULL, force, hardforce);

	return (error);
	}

	static int
	zfs_ioc_pool_configs(zfs_cmd_t *zc)
	{
	nvlist_t *configs;
	int error;

	error = spa_all_configs(&zc->zc_cookie, &configs);
	if (error)
	return (error);

	error = put_nvlist(zc, configs);

	nvlist_free(configs);

	return (error);
	}

	/*
	* inputs:
	* zc_name name of the pool
	*
	* outputs:
	* zc_cookie real errno
	* zc_nvlist_dst config nvlist
	* zc_nvlist_dst_size size of config nvlist
	*/
	static int
	zfs_ioc_pool_stats(zfs_cmd_t *zc)
	{
	nvlist_t *config;
	int error;
	int ret = 0;

	error = spa_get_stats(zc->zc_name, &config, zc->zc_value,
	sizeof (zc->zc_value));

	if (config != NULL) {
	ret = put_nvlist(zc, config);
	nvlist_free(config);

	/*
	* The config may be present even if 'error' is non-zero.
	* In this case we return success, and preserve the real errno
	* in 'zc_cookie'.
	*/
	zc->zc_cookie = error;
	} else {
	ret = error;
	}

	return (ret);
	}

	/*
	* Try to import the given pool, returning pool stats as appropriate so that
	* user land knows which devices are available and overall pool health.
	*/
	static int
	zfs_ioc_pool_tryimport(zfs_cmd_t *zc)
	{
	nvlist_t tryconfig, config = NULL;
	int error;

	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
	zc->zc_iflags, &tryconfig)) != 0)
	return (error);

	config = spa_tryimport(tryconfig);

	nvlist_free(tryconfig);

	if (config == NULL)
	return (SET_ERROR(EINVAL));

	error = put_nvlist(zc, config);
	nvlist_free(config);

	return (error);
	}

	/*
	* inputs:
	* zc_name name of the pool
	* zc_cookie scan func (pool_scan_func_t)
	* zc_flags scrub pause/resume flag (pool_scrub_cmd_t)
	*/
	static int
	zfs_ioc_pool_scan(zfs_cmd_t *zc)
	{
	spa_t *spa;
	int error;

	if (zc->zc_flags >= POOL_SCRUB_FLAGS_END)
	return (SET_ERROR(EINVAL));

	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
	return (error);

	if (zc->zc_flags == POOL_SCRUB_PAUSE)
	error = spa_scrub_pause_resume(spa, POOL_SCRUB_PAUSE);
	else if (zc->zc_cookie == POOL_SCAN_NONE)
	error = spa_scan_stop(spa);
	else
	error = spa_scan(spa, zc->zc_cookie);

	spa_close(spa, FTAG);

	return (error);
	}

	/*
	* inputs:
	* poolname name of the pool
	* scan_type scan func (pool_scan_func_t)
	* scan_command scrub pause/resume flag (pool_scrub_cmd_t)
	*/
	static const zfs_ioc_key_t zfs_keys_pool_scrub[] = {
	{"scan_type", DATA_TYPE_UINT64, 0},
	{"scan_command", DATA_TYPE_UINT64, 0},
	};

	static int
	zfs_ioc_pool_scrub(const char poolname, nvlist_t innvl, nvlist_t *outnvl)
	{
	spa_t *spa;
	int error;
	uint64_t scan_type, scan_cmd;

	if (nvlist_lookup_uint64(innvl, "scan_type", &scan_type) != 0)
	return (SET_ERROR(EINVAL));
	if (nvlist_lookup_uint64(innvl, "scan_command", &scan_cmd) != 0)
	return (SET_ERROR(EINVAL));

	if (scan_cmd >= POOL_SCRUB_FLAGS_END)
	return (SET_ERROR(EINVAL));

	if ((error = spa_open(poolname, &spa, FTAG)) != 0)
	return (error);

	if (scan_cmd == POOL_SCRUB_PAUSE) {
	error = spa_scrub_pause_resume(spa, POOL_SCRUB_PAUSE);
	} else if (scan_type == POOL_SCAN_NONE) {
	error = spa_scan_stop(spa);
	} else {
	error = spa_scan(spa, scan_type);
	}

	spa_close(spa, FTAG);
	return (error);
	}

	static int
	zfs_ioc_pool_freeze(zfs_cmd_t *zc)
	{
	spa_t *spa;
	int error;

	error = spa_open(zc->zc_name, &spa, FTAG);
	if (error == 0) {
	spa_freeze(spa);
	spa_close(spa, FTAG);
	}
	return (error);
	}

	static int
	zfs_ioc_pool_upgrade(zfs_cmd_t *zc)
	{
	spa_t *spa;
	int error;

	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
	return (error);

	if (zc->zc_cookie < spa_version(spa) \|\|
	!SPA_VERSION_IS_SUPPORTED(zc->zc_cookie)) {
	spa_close(spa, FTAG);
	return (SET_ERROR(EINVAL));
	}

	spa_upgrade(spa, zc->zc_cookie);
	spa_close(spa, FTAG);

	return (error);
	}

	static int
	zfs_ioc_pool_get_history(zfs_cmd_t *zc)
	{
	spa_t *spa;
	char *hist_buf;
	uint64_t size;
	int error;

	if ((size = zc->zc_history_len) == 0)
	return (SET_ERROR(EINVAL));

	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
	return (error);

	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
	spa_close(spa, FTAG);
	return (SET_ERROR(ENOTSUP));
	}

	hist_buf = vmem_alloc(size, KM_SLEEP);
	if ((error = spa_history_get(spa, &zc->zc_history_offset,
	&zc->zc_history_len, hist_buf)) == 0) {
	error = ddi_copyout(hist_buf,
	(void *)(uintptr_t)zc->zc_history,
	zc->zc_history_len, zc->zc_iflags);
	}

	spa_close(spa, FTAG);
	vmem_free(hist_buf, size);
	return (error);
	}

	static int
	zfs_ioc_pool_reguid(zfs_cmd_t *zc)
	{
	spa_t *spa;
	int error;

	error = spa_open(zc->zc_name, &spa, FTAG);
	if (error == 0) {
	error = spa_change_guid(spa);
	spa_close(spa, FTAG);
	}
	return (error);
	}

	static int
	zfs_ioc_dsobj_to_dsname(zfs_cmd_t *zc)
	{
	return (dsl_dsobj_to_dsname(zc->zc_name, zc->zc_obj, zc->zc_value));
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_obj object to find
	*
	* outputs:
	* zc_value name of object
	*/
	static int
	zfs_ioc_obj_to_path(zfs_cmd_t *zc)
	{
	objset_t *os;
	int error;

	/* XXX reading from objset not owned */
	if ((error = dmu_objset_hold_flags(zc->zc_name, B_TRUE,
	FTAG, &os)) != 0)
	return (error);
	if (dmu_objset_type(os) != DMU_OST_ZFS) {
	dmu_objset_rele_flags(os, B_TRUE, FTAG);
	return (SET_ERROR(EINVAL));
	}
	error = zfs_obj_to_path(os, zc->zc_obj, zc->zc_value,
	sizeof (zc->zc_value));
	dmu_objset_rele_flags(os, B_TRUE, FTAG);

	return (error);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_obj object to find
	*
	* outputs:
	* zc_stat stats on object
	* zc_value path to object
	*/
	static int
	zfs_ioc_obj_to_stats(zfs_cmd_t *zc)
	{
	objset_t *os;
	int error;

	/* XXX reading from objset not owned */
	if ((error = dmu_objset_hold_flags(zc->zc_name, B_TRUE,
	FTAG, &os)) != 0)
	return (error);
	if (dmu_objset_type(os) != DMU_OST_ZFS) {
	dmu_objset_rele_flags(os, B_TRUE, FTAG);
	return (SET_ERROR(EINVAL));
	}
	error = zfs_obj_to_stats(os, zc->zc_obj, &zc->zc_stat, zc->zc_value,
	sizeof (zc->zc_value));
	dmu_objset_rele_flags(os, B_TRUE, FTAG);

	return (error);
	}

	static int
	zfs_ioc_vdev_add(zfs_cmd_t *zc)
	{
	spa_t *spa;
	int error;
	nvlist_t *config;

	error = spa_open(zc->zc_name, &spa, FTAG);
	if (error != 0)
	return (error);

	error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
	zc->zc_iflags, &config);
	if (error == 0) {
	error = spa_vdev_add(spa, config, zc->zc_flags);
	nvlist_free(config);
	}
	spa_close(spa, FTAG);
	return (error);
	}

	/*
	* inputs:
	* zc_name name of the pool
	* zc_guid guid of vdev to remove
	* zc_cookie cancel removal
	*/
	static int
	zfs_ioc_vdev_remove(zfs_cmd_t *zc)
	{
	spa_t *spa;
	int error;

	error = spa_open(zc->zc_name, &spa, FTAG);
	if (error != 0)
	return (error);
	if (zc->zc_cookie != 0) {
	error = spa_vdev_remove_cancel(spa);
	} else {
	error = spa_vdev_remove(spa, zc->zc_guid, B_FALSE);
	}
	spa_close(spa, FTAG);
	return (error);
	}

	static int
	zfs_ioc_vdev_set_state(zfs_cmd_t *zc)
	{
	spa_t *spa;
	int error;
	vdev_state_t newstate = VDEV_STATE_UNKNOWN;

	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
	return (error);
	switch (zc->zc_cookie) {
	case VDEV_STATE_ONLINE:
	error = vdev_online(spa, zc->zc_guid, zc->zc_obj, &newstate);
	break;

	case VDEV_STATE_OFFLINE:
	error = vdev_offline(spa, zc->zc_guid, zc->zc_obj);
	break;

	case VDEV_STATE_FAULTED:
	if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
	zc->zc_obj != VDEV_AUX_EXTERNAL &&
	zc->zc_obj != VDEV_AUX_EXTERNAL_PERSIST)
	zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;

	error = vdev_fault(spa, zc->zc_guid, zc->zc_obj);
	break;

	case VDEV_STATE_DEGRADED:
	if (zc->zc_obj != VDEV_AUX_ERR_EXCEEDED &&
	zc->zc_obj != VDEV_AUX_EXTERNAL)
	zc->zc_obj = VDEV_AUX_ERR_EXCEEDED;

	error = vdev_degrade(spa, zc->zc_guid, zc->zc_obj);
	break;

	case VDEV_STATE_REMOVED:
	error = vdev_remove_wanted(spa, zc->zc_guid);
	break;

	default:
	error = SET_ERROR(EINVAL);
	}
	zc->zc_cookie = newstate;
	spa_close(spa, FTAG);
	return (error);
	}

	static int
	zfs_ioc_vdev_attach(zfs_cmd_t *zc)
	{
	spa_t *spa;
	nvlist_t *config;
	int replacing = zc->zc_cookie;
	int rebuild = zc->zc_simple;
	int error;

	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
	return (error);

	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
	zc->zc_iflags, &config)) == 0) {
	error = spa_vdev_attach(spa, zc->zc_guid, config, replacing,
	rebuild);
	nvlist_free(config);
	}

	spa_close(spa, FTAG);
	return (error);
	}

	static int
	zfs_ioc_vdev_detach(zfs_cmd_t *zc)
	{
	spa_t *spa;
	int error;

	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
	return (error);

	error = spa_vdev_detach(spa, zc->zc_guid, 0, B_FALSE);

	spa_close(spa, FTAG);
	return (error);
	}

	static int
	zfs_ioc_vdev_split(zfs_cmd_t *zc)
	{
	spa_t *spa;
	nvlist_t config, props = NULL;
	int error;
	boolean_t exp = !!(zc->zc_cookie & ZPOOL_EXPORT_AFTER_SPLIT);

	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
	return (error);

	if ((error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
	zc->zc_iflags, &config))) {
	spa_close(spa, FTAG);
	return (error);
	}

	if (zc->zc_nvlist_src_size != 0 && (error =
	get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
	zc->zc_iflags, &props))) {
	spa_close(spa, FTAG);
	nvlist_free(config);
	return (error);
	}

	error = spa_vdev_split_mirror(spa, zc->zc_string, config, props, exp);

	spa_close(spa, FTAG);

	nvlist_free(config);
	nvlist_free(props);

	return (error);
	}

	static int
	zfs_ioc_vdev_setpath(zfs_cmd_t *zc)
	{
	spa_t *spa;
	const char *path = zc->zc_value;
	uint64_t guid = zc->zc_guid;
	int error;

	error = spa_open(zc->zc_name, &spa, FTAG);
	if (error != 0)
	return (error);

	error = spa_vdev_setpath(spa, guid, path);
	spa_close(spa, FTAG);
	return (error);
	}

	static int
	zfs_ioc_vdev_setfru(zfs_cmd_t *zc)
	{
	spa_t *spa;
	const char *fru = zc->zc_value;
	uint64_t guid = zc->zc_guid;
	int error;

	error = spa_open(zc->zc_name, &spa, FTAG);
	if (error != 0)
	return (error);

	error = spa_vdev_setfru(spa, guid, fru);
	spa_close(spa, FTAG);
	return (error);
	}

	static int
	zfs_ioc_objset_stats_impl(zfs_cmd_t zc, objset_t os)
	{
	int error = 0;
	nvlist_t *nv;

	dmu_objset_fast_stat(os, &zc->zc_objset_stats);

	if (!zc->zc_simple && zc->zc_nvlist_dst != 0 &&
	(error = dsl_prop_get_all(os, &nv)) == 0) {
	dmu_objset_stats(os, nv);
	/*
	* NB: zvol_get_stats() will read the objset contents,
	* which we aren't supposed to do with a
	* DS_MODE_USER hold, because it could be
	* inconsistent. So this is a bit of a workaround...
	* XXX reading without owning
	*/
	if (!zc->zc_objset_stats.dds_inconsistent &&
	dmu_objset_type(os) == DMU_OST_ZVOL) {
	error = zvol_get_stats(os, nv);
	if (error == EIO) {
	nvlist_free(nv);
	return (error);
	}
	VERIFY0(error);
	}
	if (error == 0)
	error = put_nvlist(zc, nv);
	nvlist_free(nv);
	}

	return (error);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_nvlist_dst_size size of buffer for property nvlist
	*
	* outputs:
	* zc_objset_stats stats
	* zc_nvlist_dst property nvlist
	* zc_nvlist_dst_size size of property nvlist
	*/
	static int
	zfs_ioc_objset_stats(zfs_cmd_t *zc)
	{
	objset_t *os;
	int error;

	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
	if (error == 0) {
	error = zfs_ioc_objset_stats_impl(zc, os);
	dmu_objset_rele(os, FTAG);
	}

	return (error);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_nvlist_dst_size size of buffer for property nvlist
	*
	* outputs:
	* zc_nvlist_dst received property nvlist
	* zc_nvlist_dst_size size of received property nvlist
	*
	* Gets received properties (distinct from local properties on or after
	* SPA_VERSION_RECVD_PROPS) for callers who want to differentiate received from
	* local property values.
	*/
	static int
	zfs_ioc_objset_recvd_props(zfs_cmd_t *zc)
	{
	int error = 0;
	nvlist_t *nv;

	/*
	* Without this check, we would return local property values if the
	* caller has not already received properties on or after
	* SPA_VERSION_RECVD_PROPS.
	*/
	if (!dsl_prop_get_hasrecvd(zc->zc_name))
	return (SET_ERROR(ENOTSUP));

	if (zc->zc_nvlist_dst != 0 &&
	(error = dsl_prop_get_received(zc->zc_name, &nv)) == 0) {
	error = put_nvlist(zc, nv);
	nvlist_free(nv);
	}

	return (error);
	}

	static int
	nvl_add_zplprop(objset_t os, nvlist_t props, zfs_prop_t prop)
	{
	uint64_t value;
	int error;

	/*
	* zfs_get_zplprop() will either find a value or give us
	* the default value (if there is one).
	*/
	if ((error = zfs_get_zplprop(os, prop, &value)) != 0)
	return (error);
	VERIFY(nvlist_add_uint64(props, zfs_prop_to_name(prop), value) == 0);
	return (0);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_nvlist_dst_size size of buffer for zpl property nvlist
	*
	* outputs:
	* zc_nvlist_dst zpl property nvlist
	* zc_nvlist_dst_size size of zpl property nvlist
	*/
	static int
	zfs_ioc_objset_zplprops(zfs_cmd_t *zc)
	{
	objset_t *os;
	int err;

	/* XXX reading without owning */
	if ((err = dmu_objset_hold(zc->zc_name, FTAG, &os)))
	return (err);

	dmu_objset_fast_stat(os, &zc->zc_objset_stats);

	/*
	* NB: nvl_add_zplprop() will read the objset contents,
	* which we aren't supposed to do with a DS_MODE_USER
	* hold, because it could be inconsistent.
	*/
	if (zc->zc_nvlist_dst != 0 &&
	!zc->zc_objset_stats.dds_inconsistent &&
	dmu_objset_type(os) == DMU_OST_ZFS) {
	nvlist_t *nv;

	VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0);
	if ((err = nvl_add_zplprop(os, nv, ZFS_PROP_VERSION)) == 0 &&
	(err = nvl_add_zplprop(os, nv, ZFS_PROP_NORMALIZE)) == 0 &&
	(err = nvl_add_zplprop(os, nv, ZFS_PROP_UTF8ONLY)) == 0 &&
	(err = nvl_add_zplprop(os, nv, ZFS_PROP_CASE)) == 0)
	err = put_nvlist(zc, nv);
	nvlist_free(nv);
	} else {
	err = SET_ERROR(ENOENT);
	}
	dmu_objset_rele(os, FTAG);
	return (err);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_cookie zap cursor
	* zc_nvlist_dst_size size of buffer for property nvlist
	*
	* outputs:
	* zc_name name of next filesystem
	* zc_cookie zap cursor
	* zc_objset_stats stats
	* zc_nvlist_dst property nvlist
	* zc_nvlist_dst_size size of property nvlist
	*/
	static int
	zfs_ioc_dataset_list_next(zfs_cmd_t *zc)
	{
	objset_t *os;
	int error;
	char *p;
	size_t orig_len = strlen(zc->zc_name);

	top:
	if ((error = dmu_objset_hold(zc->zc_name, FTAG, &os))) {
	if (error == ENOENT)
	error = SET_ERROR(ESRCH);
	return (error);
	}

	p = strrchr(zc->zc_name, '/');
	if (p == NULL \|\| p[1] != '\0')
	(void) strlcat(zc->zc_name, "/", sizeof (zc->zc_name));
	p = zc->zc_name + strlen(zc->zc_name);

	do {
	error = dmu_dir_list_next(os,
	sizeof (zc->zc_name) - (p - zc->zc_name), p,
	NULL, &zc->zc_cookie);
	if (error == ENOENT)
	error = SET_ERROR(ESRCH);
	} while (error == 0 && zfs_dataset_name_hidden(zc->zc_name));
	dmu_objset_rele(os, FTAG);

	/*
	* If it's an internal dataset (ie. with a '$' in its name),
	* don't try to get stats for it, otherwise we'll return ENOENT.
	*/
	if (error == 0 && strchr(zc->zc_name, '$') == NULL) {
	error = zfs_ioc_objset_stats(zc); /* fill in the stats */
	if (error == ENOENT) {
	/* We lost a race with destroy, get the next one. */
	zc->zc_name[orig_len] = '\0';
	goto top;
	}
	}
	return (error);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_cookie zap cursor
	* zc_nvlist_src iteration range nvlist
	* zc_nvlist_src_size size of iteration range nvlist
	*
	* outputs:
	* zc_name name of next snapshot
	* zc_objset_stats stats
	* zc_nvlist_dst property nvlist
	* zc_nvlist_dst_size size of property nvlist
	*/
	static int
	zfs_ioc_snapshot_list_next(zfs_cmd_t *zc)
	{
	int error;
	objset_t os, ossnap;
	dsl_dataset_t *ds;
	uint64_t min_txg = 0, max_txg = 0;

	if (zc->zc_nvlist_src_size != 0) {
	nvlist_t *props = NULL;
	error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
	zc->zc_iflags, &props);
	if (error != 0)
	return (error);
	(void) nvlist_lookup_uint64(props, SNAP_ITER_MIN_TXG,
	&min_txg);
	(void) nvlist_lookup_uint64(props, SNAP_ITER_MAX_TXG,
	&max_txg);
	nvlist_free(props);
	}

	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
	if (error != 0) {
	return (error == ENOENT ? SET_ERROR(ESRCH) : error);
	}

	/*
	* A dataset name of maximum length cannot have any snapshots,
	* so exit immediately.
	*/
	if (strlcat(zc->zc_name, "@", sizeof (zc->zc_name)) >=
	ZFS_MAX_DATASET_NAME_LEN) {
	dmu_objset_rele(os, FTAG);
	return (SET_ERROR(ESRCH));
	}

	while (error == 0) {
	- if (issig(JUSTLOOKING) && issig(FORREAL)) {
	+ if (issig()) {
	error = SET_ERROR(EINTR);
	break;
	}

	error = dmu_snapshot_list_next(os,
	sizeof (zc->zc_name) - strlen(zc->zc_name),
	zc->zc_name + strlen(zc->zc_name), &zc->zc_obj,
	&zc->zc_cookie, NULL);
	if (error == ENOENT) {
	error = SET_ERROR(ESRCH);
	break;
	} else if (error != 0) {
	break;
	}

	error = dsl_dataset_hold_obj(dmu_objset_pool(os), zc->zc_obj,
	FTAG, &ds);
	if (error != 0)
	break;

	if ((min_txg != 0 && dsl_get_creationtxg(ds) < min_txg) \|\|
	(max_txg != 0 && dsl_get_creationtxg(ds) > max_txg)) {
	dsl_dataset_rele(ds, FTAG);
	/* undo snapshot name append */
	*(strchr(zc->zc_name, '@') + 1) = '\0';
	/* skip snapshot */
	continue;
	}

	if (zc->zc_simple) {
	dsl_dataset_fast_stat(ds, &zc->zc_objset_stats);
	dsl_dataset_rele(ds, FTAG);
	break;
	}

	if ((error = dmu_objset_from_ds(ds, &ossnap)) != 0) {
	dsl_dataset_rele(ds, FTAG);
	break;
	}
	if ((error = zfs_ioc_objset_stats_impl(zc, ossnap)) != 0) {
	dsl_dataset_rele(ds, FTAG);
	break;
	}
	dsl_dataset_rele(ds, FTAG);
	break;
	}

	dmu_objset_rele(os, FTAG);
	/* if we failed, undo the @ that we tacked on to zc_name */
	if (error != 0)
	*strchr(zc->zc_name, '@') = '\0';
	return (error);
	}

	static int
	zfs_prop_set_userquota(const char dsname, nvpair_t pair)
	{
	const char *propname = nvpair_name(pair);
	uint64_t *valary;
	unsigned int vallen;
	const char dash, domain;
	zfs_userquota_prop_t type;
	uint64_t rid;
	uint64_t quota;
	zfsvfs_t *zfsvfs;
	int err;

	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
	nvlist_t *attrs;
	VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
	if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
	&pair) != 0)
	return (SET_ERROR(EINVAL));
	}

	/*
	* A correctly constructed propname is encoded as
	* userquota@<rid>-<domain>.
	*/
	if ((dash = strchr(propname, '-')) == NULL \|\|
	nvpair_value_uint64_array(pair, &valary, &vallen) != 0 \|\|
	vallen != 3)
	return (SET_ERROR(EINVAL));

	domain = dash + 1;
	type = valary[0];
	rid = valary[1];
	quota = valary[2];

	err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE);
	if (err == 0) {
	err = zfs_set_userquota(zfsvfs, type, domain, rid, quota);
	zfsvfs_rele(zfsvfs, FTAG);
	}

	return (err);
	}

	/*
	* If the named property is one that has a special function to set its value,
	* return 0 on success and a positive error code on failure; otherwise if it is
	* not one of the special properties handled by this function, return -1.
	*
	* XXX: It would be better for callers of the property interface if we handled
	* these special cases in dsl_prop.c (in the dsl layer).
	*/
	static int
	zfs_prop_set_special(const char *dsname, zprop_source_t source,
	nvpair_t *pair)
	{
	const char *propname = nvpair_name(pair);
	zfs_prop_t prop = zfs_name_to_prop(propname);
	uint64_t intval = 0;
	const char *strval = NULL;
	int err = -1;

	if (prop == ZPROP_USERPROP) {
	if (zfs_prop_userquota(propname))
	return (zfs_prop_set_userquota(dsname, pair));
	return (-1);
	}

	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
	nvlist_t *attrs;
	VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
	VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
	&pair) == 0);
	}

	/* all special properties are numeric except for keylocation */
	if (zfs_prop_get_type(prop) == PROP_TYPE_STRING) {
	strval = fnvpair_value_string(pair);
	} else {
	intval = fnvpair_value_uint64(pair);
	}

	switch (prop) {
	case ZFS_PROP_QUOTA:
	err = dsl_dir_set_quota(dsname, source, intval);
	break;
	case ZFS_PROP_REFQUOTA:
	err = dsl_dataset_set_refquota(dsname, source, intval);
	break;
	case ZFS_PROP_FILESYSTEM_LIMIT:
	case ZFS_PROP_SNAPSHOT_LIMIT:
	if (intval == UINT64_MAX) {
	/* clearing the limit, just do it */
	err = 0;
	} else {
	err = dsl_dir_activate_fs_ss_limit(dsname);
	}
	/*
	* Set err to -1 to force the zfs_set_prop_nvlist code down the
	* default path to set the value in the nvlist.
	*/
	if (err == 0)
	err = -1;
	break;
	case ZFS_PROP_KEYLOCATION:
	err = dsl_crypto_can_set_keylocation(dsname, strval);

	/*
	* Set err to -1 to force the zfs_set_prop_nvlist code down the
	* default path to set the value in the nvlist.
	*/
	if (err == 0)
	err = -1;
	break;
	case ZFS_PROP_RESERVATION:
	err = dsl_dir_set_reservation(dsname, source, intval);
	break;
	case ZFS_PROP_REFRESERVATION:
	err = dsl_dataset_set_refreservation(dsname, source, intval);
	break;
	case ZFS_PROP_COMPRESSION:
	err = dsl_dataset_set_compression(dsname, source, intval);
	/*
	* Set err to -1 to force the zfs_set_prop_nvlist code down the
	* default path to set the value in the nvlist.
	*/
	if (err == 0)
	err = -1;
	break;
	case ZFS_PROP_VOLSIZE:
	err = zvol_set_volsize(dsname, intval);
	break;
	case ZFS_PROP_VOLTHREADING:
	err = zvol_set_volthreading(dsname, intval);
	/*
	* Set err to -1 to force the zfs_set_prop_nvlist code down the
	* default path to set the value in the nvlist.
	*/
	if (err == 0)
	err = -1;
	break;
	case ZFS_PROP_SNAPDEV:
	case ZFS_PROP_VOLMODE:
	err = zvol_set_common(dsname, prop, source, intval);
	break;
	case ZFS_PROP_READONLY:
	err = zvol_set_ro(dsname, intval);
	/*
	* Set err to -1 to force the zfs_set_prop_nvlist code down the
	* default path to set the value in the nvlist.
	*/
	if (err == 0)
	err = -1;
	break;
	case ZFS_PROP_VERSION:
	{
	zfsvfs_t *zfsvfs;

	if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_TRUE)) != 0)
	break;

	err = zfs_set_version(zfsvfs, intval);
	zfsvfs_rele(zfsvfs, FTAG);

	if (err == 0 && intval >= ZPL_VERSION_USERSPACE) {
	zfs_cmd_t *zc;

	zc = kmem_zalloc(sizeof (zfs_cmd_t), KM_SLEEP);
	(void) strlcpy(zc->zc_name, dsname,
	sizeof (zc->zc_name));
	(void) zfs_ioc_userspace_upgrade(zc);
	(void) zfs_ioc_id_quota_upgrade(zc);
	kmem_free(zc, sizeof (zfs_cmd_t));
	}
	break;
	}
	default:
	err = -1;
	}

	return (err);
	}

	static boolean_t
	zfs_is_namespace_prop(zfs_prop_t prop)
	{
	switch (prop) {

	case ZFS_PROP_ATIME:
	case ZFS_PROP_RELATIME:
	case ZFS_PROP_DEVICES:
	case ZFS_PROP_EXEC:
	case ZFS_PROP_SETUID:
	case ZFS_PROP_READONLY:
	case ZFS_PROP_XATTR:
	case ZFS_PROP_NBMAND:
	return (B_TRUE);

	default:
	return (B_FALSE);
	}
	}

	/*
	* This function is best effort. If it fails to set any of the given properties,
	* it continues to set as many as it can and returns the last error
	* encountered. If the caller provides a non-NULL errlist, it will be filled in
	* with the list of names of all the properties that failed along with the
	* corresponding error numbers.
	*
	* If every property is set successfully, zero is returned and errlist is not
	* modified.
	*/
	int
	zfs_set_prop_nvlist(const char dsname, zprop_source_t source, nvlist_t nvl,
	nvlist_t *errlist)
	{
	nvpair_t *pair;
	nvpair_t *propval;
	int rv = 0;
	int err;
	uint64_t intval;
	const char *strval;
	boolean_t should_update_mount_cache = B_FALSE;

	nvlist_t *genericnvl = fnvlist_alloc();
	nvlist_t *retrynvl = fnvlist_alloc();
	retry:
	pair = NULL;
	while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
	const char *propname = nvpair_name(pair);
	zfs_prop_t prop = zfs_name_to_prop(propname);
	err = 0;

	/* decode the property value */
	propval = pair;
	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
	nvlist_t *attrs;
	attrs = fnvpair_value_nvlist(pair);
	if (nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
	&propval) != 0)
	err = SET_ERROR(EINVAL);
	}

	/* Validate value type */
	if (err == 0 && source == ZPROP_SRC_INHERITED) {
	/* inherited properties are expected to be booleans */
	if (nvpair_type(propval) != DATA_TYPE_BOOLEAN)
	err = SET_ERROR(EINVAL);
	} else if (err == 0 && prop == ZPROP_USERPROP) {
	if (zfs_prop_user(propname)) {
	if (nvpair_type(propval) != DATA_TYPE_STRING)
	err = SET_ERROR(EINVAL);
	} else if (zfs_prop_userquota(propname)) {
	if (nvpair_type(propval) !=
	DATA_TYPE_UINT64_ARRAY)
	err = SET_ERROR(EINVAL);
	} else {
	err = SET_ERROR(EINVAL);
	}
	} else if (err == 0) {
	if (nvpair_type(propval) == DATA_TYPE_STRING) {
	if (zfs_prop_get_type(prop) != PROP_TYPE_STRING)
	err = SET_ERROR(EINVAL);
	} else if (nvpair_type(propval) == DATA_TYPE_UINT64) {
	const char *unused;

	intval = fnvpair_value_uint64(propval);

	switch (zfs_prop_get_type(prop)) {
	case PROP_TYPE_NUMBER:
	break;
	case PROP_TYPE_STRING:
	err = SET_ERROR(EINVAL);
	break;
	case PROP_TYPE_INDEX:
	if (zfs_prop_index_to_string(prop,
	intval, &unused) != 0)
	err =
	SET_ERROR(ZFS_ERR_BADPROP);
	break;
	default:
	cmn_err(CE_PANIC,
	"unknown property type");
	}
	} else {
	err = SET_ERROR(EINVAL);
	}
	}

	/* Validate permissions */
	if (err == 0)
	err = zfs_check_settable(dsname, pair, CRED());

	if (err == 0) {
	if (source == ZPROP_SRC_INHERITED)
	err = -1; /* does not need special handling */
	else
	err = zfs_prop_set_special(dsname, source,
	pair);
	if (err == -1) {
	/*
	* For better performance we build up a list of
	* properties to set in a single transaction.
	*/
	err = nvlist_add_nvpair(genericnvl, pair);
	} else if (err != 0 && nvl != retrynvl) {
	/*
	* This may be a spurious error caused by
	* receiving quota and reservation out of order.
	* Try again in a second pass.
	*/
	err = nvlist_add_nvpair(retrynvl, pair);
	}
	}

	if (err != 0) {
	if (errlist != NULL)
	fnvlist_add_int32(errlist, propname, err);
	rv = err;
	}

	if (zfs_is_namespace_prop(prop))
	should_update_mount_cache = B_TRUE;
	}

	if (nvl != retrynvl && !nvlist_empty(retrynvl)) {
	nvl = retrynvl;
	goto retry;
	}

	if (nvlist_empty(genericnvl))
	goto out;

	/*
	* Try to set them all in one batch.
	*/
	err = dsl_props_set(dsname, source, genericnvl);
	if (err == 0)
	goto out;

	/*
	* If batching fails, we still want to set as many properties as we
	* can, so try setting them individually.
	*/
	pair = NULL;
	while ((pair = nvlist_next_nvpair(genericnvl, pair)) != NULL) {
	const char *propname = nvpair_name(pair);

	propval = pair;
	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
	nvlist_t *attrs;
	attrs = fnvpair_value_nvlist(pair);
	propval = fnvlist_lookup_nvpair(attrs, ZPROP_VALUE);
	}

	if (nvpair_type(propval) == DATA_TYPE_STRING) {
	strval = fnvpair_value_string(propval);
	err = dsl_prop_set_string(dsname, propname,
	source, strval);
	} else if (nvpair_type(propval) == DATA_TYPE_BOOLEAN) {
	err = dsl_prop_inherit(dsname, propname, source);
	} else {
	intval = fnvpair_value_uint64(propval);
	err = dsl_prop_set_int(dsname, propname, source,
	intval);
	}

	if (err != 0) {
	if (errlist != NULL) {
	fnvlist_add_int32(errlist, propname, err);
	}
	rv = err;
	}
	}

	out:
	if (should_update_mount_cache)
	zfs_ioctl_update_mount_cache(dsname);

	nvlist_free(genericnvl);
	nvlist_free(retrynvl);

	return (rv);
	}

	/*
	* Check that all the properties are valid user properties.
	*/
	static int
	zfs_check_userprops(nvlist_t *nvl)
	{
	nvpair_t *pair = NULL;

	while ((pair = nvlist_next_nvpair(nvl, pair)) != NULL) {
	const char *propname = nvpair_name(pair);

	if (!zfs_prop_user(propname) \|\|
	nvpair_type(pair) != DATA_TYPE_STRING)
	return (SET_ERROR(EINVAL));

	if (strlen(propname) >= ZAP_MAXNAMELEN)
	return (SET_ERROR(ENAMETOOLONG));

	if (strlen(fnvpair_value_string(pair)) >= ZAP_MAXVALUELEN)
	return (SET_ERROR(E2BIG));
	}
	return (0);
	}

	static void
	props_skip(nvlist_t props, nvlist_t skipped, nvlist_t **newprops)
	{
	nvpair_t *pair;

	VERIFY(nvlist_alloc(newprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);

	pair = NULL;
	while ((pair = nvlist_next_nvpair(props, pair)) != NULL) {
	if (nvlist_exists(skipped, nvpair_name(pair)))
	continue;

	VERIFY(nvlist_add_nvpair(*newprops, pair) == 0);
	}
	}

	static int
	clear_received_props(const char dsname, nvlist_t props,
	nvlist_t *skipped)
	{
	int err = 0;
	nvlist_t *cleared_props = NULL;
	props_skip(props, skipped, &cleared_props);
	if (!nvlist_empty(cleared_props)) {
	/*
	* Acts on local properties until the dataset has received
	* properties at least once on or after SPA_VERSION_RECVD_PROPS.
	*/
	zprop_source_t flags = (ZPROP_SRC_NONE \|
	(dsl_prop_get_hasrecvd(dsname) ? ZPROP_SRC_RECEIVED : 0));
	err = zfs_set_prop_nvlist(dsname, flags, cleared_props, NULL);
	}
	nvlist_free(cleared_props);
	return (err);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_value name of property to set
	* zc_nvlist_src{_size} nvlist of properties to apply
	* zc_cookie received properties flag
	*
	* outputs:
	* zc_nvlist_dst{_size} error for each unapplied received property
	*/
	static int
	zfs_ioc_set_prop(zfs_cmd_t *zc)
	{
	nvlist_t *nvl;
	boolean_t received = zc->zc_cookie;
	zprop_source_t source = (received ? ZPROP_SRC_RECEIVED :
	ZPROP_SRC_LOCAL);
	nvlist_t *errors;
	int error;

	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
	zc->zc_iflags, &nvl)) != 0)
	return (error);

	if (received) {
	nvlist_t *origprops;

	if (dsl_prop_get_received(zc->zc_name, &origprops) == 0) {
	(void) clear_received_props(zc->zc_name,
	origprops, nvl);
	nvlist_free(origprops);
	}

	error = dsl_prop_set_hasrecvd(zc->zc_name);
	}

	errors = fnvlist_alloc();
	if (error == 0)
	error = zfs_set_prop_nvlist(zc->zc_name, source, nvl, errors);

	if (zc->zc_nvlist_dst != 0 && errors != NULL) {
	(void) put_nvlist(zc, errors);
	}

	nvlist_free(errors);
	nvlist_free(nvl);
	return (error);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_value name of property to inherit
	* zc_cookie revert to received value if TRUE
	*
	* outputs: none
	*/
	static int
	zfs_ioc_inherit_prop(zfs_cmd_t *zc)
	{
	const char *propname = zc->zc_value;
	zfs_prop_t prop = zfs_name_to_prop(propname);
	boolean_t received = zc->zc_cookie;
	zprop_source_t source = (received
	? ZPROP_SRC_NONE /* revert to received value, if any */
	: ZPROP_SRC_INHERITED); /* explicitly inherit */
	nvlist_t *dummy;
	nvpair_t *pair;
	zprop_type_t type;
	int err;

	if (!received) {
	/*
	* Only check this in the non-received case. We want to allow
	* 'inherit -S' to revert non-inheritable properties like quota
	* and reservation to the received or default values even though
	* they are not considered inheritable.
	*/
	if (prop != ZPROP_USERPROP && !zfs_prop_inheritable(prop))
	return (SET_ERROR(EINVAL));
	}

	if (prop == ZPROP_USERPROP) {
	if (!zfs_prop_user(propname))
	return (SET_ERROR(EINVAL));

	type = PROP_TYPE_STRING;
	} else if (prop == ZFS_PROP_VOLSIZE \|\| prop == ZFS_PROP_VERSION) {
	return (SET_ERROR(EINVAL));
	} else {
	type = zfs_prop_get_type(prop);
	}

	/*
	* zfs_prop_set_special() expects properties in the form of an
	* nvpair with type info.
	*/
	dummy = fnvlist_alloc();

	switch (type) {
	case PROP_TYPE_STRING:
	VERIFY(0 == nvlist_add_string(dummy, propname, ""));
	break;
	case PROP_TYPE_NUMBER:
	case PROP_TYPE_INDEX:
	VERIFY(0 == nvlist_add_uint64(dummy, propname, 0));
	break;
	default:
	err = SET_ERROR(EINVAL);
	goto errout;
	}

	pair = nvlist_next_nvpair(dummy, NULL);
	if (pair == NULL) {
	err = SET_ERROR(EINVAL);
	} else {
	err = zfs_prop_set_special(zc->zc_name, source, pair);
	if (err == -1) /* property is not "special", needs handling */
	err = dsl_prop_inherit(zc->zc_name, zc->zc_value,
	source);
	}

	errout:
	nvlist_free(dummy);
	return (err);
	}

	static int
	zfs_ioc_pool_set_props(zfs_cmd_t *zc)
	{
	nvlist_t *props;
	spa_t *spa;
	int error;
	nvpair_t *pair;

	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
	zc->zc_iflags, &props)))
	return (error);

	/*
	* If the only property is the configfile, then just do a spa_lookup()
	* to handle the faulted case.
	*/
	pair = nvlist_next_nvpair(props, NULL);
	if (pair != NULL && strcmp(nvpair_name(pair),
	zpool_prop_to_name(ZPOOL_PROP_CACHEFILE)) == 0 &&
	nvlist_next_nvpair(props, pair) == NULL) {
	mutex_enter(&spa_namespace_lock);
	if ((spa = spa_lookup(zc->zc_name)) != NULL) {
	spa_configfile_set(spa, props, B_FALSE);
	spa_write_cachefile(spa, B_FALSE, B_TRUE, B_FALSE);
	}
	mutex_exit(&spa_namespace_lock);
	if (spa != NULL) {
	nvlist_free(props);
	return (0);
	}
	}

	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
	nvlist_free(props);
	return (error);
	}

	error = spa_prop_set(spa, props);

	nvlist_free(props);
	spa_close(spa, FTAG);

	return (error);
	}

	static int
	zfs_ioc_pool_get_props(zfs_cmd_t *zc)
	{
	spa_t *spa;
	int error;
	nvlist_t *nvp = NULL;

	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0) {
	/*
	* If the pool is faulted, there may be properties we can still
	* get (such as altroot and cachefile), so attempt to get them
	* anyway.
	*/
	mutex_enter(&spa_namespace_lock);
	if ((spa = spa_lookup(zc->zc_name)) != NULL)
	error = spa_prop_get(spa, &nvp);
	mutex_exit(&spa_namespace_lock);
	} else {
	error = spa_prop_get(spa, &nvp);
	spa_close(spa, FTAG);
	}

	if (error == 0 && zc->zc_nvlist_dst != 0)
	error = put_nvlist(zc, nvp);
	else
	error = SET_ERROR(EFAULT);

	nvlist_free(nvp);
	return (error);
	}

	/*
	* innvl: {
	* "vdevprops_set_vdev" -> guid
	* "vdevprops_set_props" -> { prop -> value }
	* }
	*
	* outnvl: propname -> error code (int32)
	*/
	static const zfs_ioc_key_t zfs_keys_vdev_set_props[] = {
	{ZPOOL_VDEV_PROPS_SET_VDEV, DATA_TYPE_UINT64, 0},
	{ZPOOL_VDEV_PROPS_SET_PROPS, DATA_TYPE_NVLIST, 0}
	};

	static int
	zfs_ioc_vdev_set_props(const char poolname, nvlist_t innvl, nvlist_t *outnvl)
	{
	spa_t *spa;
	int error;
	vdev_t *vd;
	uint64_t vdev_guid;

	/* Early validation */
	if (nvlist_lookup_uint64(innvl, ZPOOL_VDEV_PROPS_SET_VDEV,
	&vdev_guid) != 0)
	return (SET_ERROR(EINVAL));

	if (outnvl == NULL)
	return (SET_ERROR(EINVAL));

	if ((error = spa_open(poolname, &spa, FTAG)) != 0)
	return (error);

	ASSERT(spa_writeable(spa));

	if ((vd = spa_lookup_by_guid(spa, vdev_guid, B_TRUE)) == NULL) {
	spa_close(spa, FTAG);
	return (SET_ERROR(ENOENT));
	}

	error = vdev_prop_set(vd, innvl, outnvl);

	spa_close(spa, FTAG);

	return (error);
	}

	/*
	* innvl: {
	* "vdevprops_get_vdev" -> guid
	* (optional) "vdevprops_get_props" -> { propname -> propid }
	* }
	*
	* outnvl: propname -> value
	*/
	static const zfs_ioc_key_t zfs_keys_vdev_get_props[] = {
	{ZPOOL_VDEV_PROPS_GET_VDEV, DATA_TYPE_UINT64, 0},
	{ZPOOL_VDEV_PROPS_GET_PROPS, DATA_TYPE_NVLIST, ZK_OPTIONAL}
	};

	static int
	zfs_ioc_vdev_get_props(const char poolname, nvlist_t innvl, nvlist_t *outnvl)
	{
	spa_t *spa;
	int error;
	vdev_t *vd;
	uint64_t vdev_guid;

	/* Early validation */
	if (nvlist_lookup_uint64(innvl, ZPOOL_VDEV_PROPS_GET_VDEV,
	&vdev_guid) != 0)
	return (SET_ERROR(EINVAL));

	if (outnvl == NULL)
	return (SET_ERROR(EINVAL));

	if ((error = spa_open(poolname, &spa, FTAG)) != 0)
	return (error);

	if ((vd = spa_lookup_by_guid(spa, vdev_guid, B_TRUE)) == NULL) {
	spa_close(spa, FTAG);
	return (SET_ERROR(ENOENT));
	}

	error = vdev_prop_get(vd, innvl, outnvl);

	spa_close(spa, FTAG);

	return (error);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_nvlist_src{_size} nvlist of delegated permissions
	* zc_perm_action allow/unallow flag
	*
	* outputs: none
	*/
	static int
	zfs_ioc_set_fsacl(zfs_cmd_t *zc)
	{
	int error;
	nvlist_t *fsaclnv = NULL;

	if ((error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
	zc->zc_iflags, &fsaclnv)) != 0)
	return (error);

	/*
	* Verify nvlist is constructed correctly
	*/
	if (zfs_deleg_verify_nvlist(fsaclnv) != 0) {
	nvlist_free(fsaclnv);
	return (SET_ERROR(EINVAL));
	}

	/*
	* If we don't have PRIV_SYS_MOUNT, then validate
	* that user is allowed to hand out each permission in
	* the nvlist(s)
	*/

	error = secpolicy_zfs(CRED());
	if (error != 0) {
	if (zc->zc_perm_action == B_FALSE) {
	error = dsl_deleg_can_allow(zc->zc_name,
	fsaclnv, CRED());
	} else {
	error = dsl_deleg_can_unallow(zc->zc_name,
	fsaclnv, CRED());
	}
	}

	if (error == 0)
	error = dsl_deleg_set(zc->zc_name, fsaclnv, zc->zc_perm_action);

	nvlist_free(fsaclnv);
	return (error);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	*
	* outputs:
	* zc_nvlist_src{_size} nvlist of delegated permissions
	*/
	static int
	zfs_ioc_get_fsacl(zfs_cmd_t *zc)
	{
	nvlist_t *nvp;
	int error;

	if ((error = dsl_deleg_get(zc->zc_name, &nvp)) == 0) {
	error = put_nvlist(zc, nvp);
	nvlist_free(nvp);
	}

	return (error);
	}

	static void
	zfs_create_cb(objset_t os, void arg, cred_t cr, dmu_tx_t tx)
	{
	zfs_creat_t *zct = arg;

	zfs_create_fs(os, cr, zct->zct_zplprops, tx);
	}

	#define ZFS_PROP_UNDEFINED ((uint64_t)-1)

	/*
	* inputs:
	* os parent objset pointer (NULL if root fs)
	* fuids_ok fuids allowed in this version of the spa?
	* sa_ok SAs allowed in this version of the spa?
	* createprops list of properties requested by creator
	*
	* outputs:
	* zplprops values for the zplprops we attach to the master node object
	* is_ci true if requested file system will be purely case-insensitive
	*
	* Determine the settings for utf8only, normalization and
	* casesensitivity. Specific values may have been requested by the
	* creator and/or we can inherit values from the parent dataset. If
	* the file system is of too early a vintage, a creator can not
	* request settings for these properties, even if the requested
	* setting is the default value. We don't actually want to create dsl
	* properties for these, so remove them from the source nvlist after
	* processing.
	*/
	static int
	zfs_fill_zplprops_impl(objset_t *os, uint64_t zplver,
	boolean_t fuids_ok, boolean_t sa_ok, nvlist_t *createprops,
	nvlist_t zplprops, boolean_t is_ci)
	{
	uint64_t sense = ZFS_PROP_UNDEFINED;
	uint64_t norm = ZFS_PROP_UNDEFINED;
	uint64_t u8 = ZFS_PROP_UNDEFINED;
	int error;

	ASSERT(zplprops != NULL);

	/* parent dataset must be a filesystem */
	if (os != NULL && os->os_phys->os_type != DMU_OST_ZFS)
	return (SET_ERROR(ZFS_ERR_WRONG_PARENT));

	/*
	* Pull out creator prop choices, if any.
	*/
	if (createprops) {
	(void) nvlist_lookup_uint64(createprops,
	zfs_prop_to_name(ZFS_PROP_VERSION), &zplver);
	(void) nvlist_lookup_uint64(createprops,
	zfs_prop_to_name(ZFS_PROP_NORMALIZE), &norm);
	(void) nvlist_remove_all(createprops,
	zfs_prop_to_name(ZFS_PROP_NORMALIZE));
	(void) nvlist_lookup_uint64(createprops,
	zfs_prop_to_name(ZFS_PROP_UTF8ONLY), &u8);
	(void) nvlist_remove_all(createprops,
	zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
	(void) nvlist_lookup_uint64(createprops,
	zfs_prop_to_name(ZFS_PROP_CASE), &sense);
	(void) nvlist_remove_all(createprops,
	zfs_prop_to_name(ZFS_PROP_CASE));
	}

	/*
	* If the zpl version requested is whacky or the file system
	* or pool is version is too "young" to support normalization
	* and the creator tried to set a value for one of the props,
	* error out.
	*/
	if ((zplver < ZPL_VERSION_INITIAL \|\| zplver > ZPL_VERSION) \|\|
	(zplver >= ZPL_VERSION_FUID && !fuids_ok) \|\|
	(zplver >= ZPL_VERSION_SA && !sa_ok) \|\|
	(zplver < ZPL_VERSION_NORMALIZATION &&
	(norm != ZFS_PROP_UNDEFINED \|\| u8 != ZFS_PROP_UNDEFINED \|\|
	sense != ZFS_PROP_UNDEFINED)))
	return (SET_ERROR(ENOTSUP));

	/*
	* Put the version in the zplprops
	*/
	VERIFY(nvlist_add_uint64(zplprops,
	zfs_prop_to_name(ZFS_PROP_VERSION), zplver) == 0);

	if (norm == ZFS_PROP_UNDEFINED &&
	(error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &norm)) != 0)
	return (error);
	VERIFY(nvlist_add_uint64(zplprops,
	zfs_prop_to_name(ZFS_PROP_NORMALIZE), norm) == 0);

	/*
	* If we're normalizing, names must always be valid UTF-8 strings.
	*/
	if (norm)
	u8 = 1;
	if (u8 == ZFS_PROP_UNDEFINED &&
	(error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &u8)) != 0)
	return (error);
	VERIFY(nvlist_add_uint64(zplprops,
	zfs_prop_to_name(ZFS_PROP_UTF8ONLY), u8) == 0);

	if (sense == ZFS_PROP_UNDEFINED &&
	(error = zfs_get_zplprop(os, ZFS_PROP_CASE, &sense)) != 0)
	return (error);
	VERIFY(nvlist_add_uint64(zplprops,
	zfs_prop_to_name(ZFS_PROP_CASE), sense) == 0);

	if (is_ci)
	*is_ci = (sense == ZFS_CASE_INSENSITIVE);

	return (0);
	}

	static int
	zfs_fill_zplprops(const char dataset, nvlist_t createprops,
	nvlist_t zplprops, boolean_t is_ci)
	{
	boolean_t fuids_ok, sa_ok;
	uint64_t zplver = ZPL_VERSION;
	objset_t *os = NULL;
	char parentname[ZFS_MAX_DATASET_NAME_LEN];
	spa_t *spa;
	uint64_t spa_vers;
	int error;

	zfs_get_parent(dataset, parentname, sizeof (parentname));

	if ((error = spa_open(dataset, &spa, FTAG)) != 0)
	return (error);

	spa_vers = spa_version(spa);
	spa_close(spa, FTAG);

	zplver = zfs_zpl_version_map(spa_vers);
	fuids_ok = (zplver >= ZPL_VERSION_FUID);
	sa_ok = (zplver >= ZPL_VERSION_SA);

	/*
	* Open parent object set so we can inherit zplprop values.
	*/
	if ((error = dmu_objset_hold(parentname, FTAG, &os)) != 0)
	return (error);

	error = zfs_fill_zplprops_impl(os, zplver, fuids_ok, sa_ok, createprops,
	zplprops, is_ci);
	dmu_objset_rele(os, FTAG);
	return (error);
	}

	static int
	zfs_fill_zplprops_root(uint64_t spa_vers, nvlist_t *createprops,
	nvlist_t zplprops, boolean_t is_ci)
	{
	boolean_t fuids_ok;
	boolean_t sa_ok;
	uint64_t zplver = ZPL_VERSION;
	int error;

	zplver = zfs_zpl_version_map(spa_vers);
	fuids_ok = (zplver >= ZPL_VERSION_FUID);
	sa_ok = (zplver >= ZPL_VERSION_SA);

	error = zfs_fill_zplprops_impl(NULL, zplver, fuids_ok, sa_ok,
	createprops, zplprops, is_ci);
	return (error);
	}

	/*
	* innvl: {
	* "type" -> dmu_objset_type_t (int32)
	* (optional) "props" -> { prop -> value }
	* (optional) "hidden_args" -> { "wkeydata" -> value }
	* raw uint8_t array of encryption wrapping key data (32 bytes)
	* }
	*
	* outnvl: propname -> error code (int32)
	*/

	static const zfs_ioc_key_t zfs_keys_create[] = {
	{"type", DATA_TYPE_INT32, 0},
	{"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
	{"hidden_args", DATA_TYPE_NVLIST, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_create(const char fsname, nvlist_t innvl, nvlist_t *outnvl)
	{
	int error = 0;
	zfs_creat_t zct = { 0 };
	nvlist_t *nvprops = NULL;
	nvlist_t *hidden_args = NULL;
	void (cbfunc)(objset_t os, void arg, cred_t cr, dmu_tx_t *tx);
	dmu_objset_type_t type;
	boolean_t is_insensitive = B_FALSE;
	dsl_crypto_params_t *dcp = NULL;

	type = (dmu_objset_type_t)fnvlist_lookup_int32(innvl, "type");
	(void) nvlist_lookup_nvlist(innvl, "props", &nvprops);
	(void) nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);

	switch (type) {
	case DMU_OST_ZFS:
	cbfunc = zfs_create_cb;
	break;

	case DMU_OST_ZVOL:
	cbfunc = zvol_create_cb;
	break;

	default:
	cbfunc = NULL;
	break;
	}
	if (strchr(fsname, '@') \|\|
	strchr(fsname, '%'))
	return (SET_ERROR(EINVAL));

	zct.zct_props = nvprops;

	if (cbfunc == NULL)
	return (SET_ERROR(EINVAL));

	if (type == DMU_OST_ZVOL) {
	uint64_t volsize, volblocksize;

	if (nvprops == NULL)
	return (SET_ERROR(EINVAL));
	if (nvlist_lookup_uint64(nvprops,
	zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) != 0)
	return (SET_ERROR(EINVAL));

	if ((error = nvlist_lookup_uint64(nvprops,
	zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
	&volblocksize)) != 0 && error != ENOENT)
	return (SET_ERROR(EINVAL));

	if (error != 0)
	volblocksize = zfs_prop_default_numeric(
	ZFS_PROP_VOLBLOCKSIZE);

	if ((error = zvol_check_volblocksize(fsname,
	volblocksize)) != 0 \|\|
	(error = zvol_check_volsize(volsize,
	volblocksize)) != 0)
	return (error);
	} else if (type == DMU_OST_ZFS) {
	int error;

	/*
	* We have to have normalization and
	* case-folding flags correct when we do the
	* file system creation, so go figure them out
	* now.
	*/
	VERIFY(nvlist_alloc(&zct.zct_zplprops,
	NV_UNIQUE_NAME, KM_SLEEP) == 0);
	error = zfs_fill_zplprops(fsname, nvprops,
	zct.zct_zplprops, &is_insensitive);
	if (error != 0) {
	nvlist_free(zct.zct_zplprops);
	return (error);
	}
	}

	error = dsl_crypto_params_create_nvlist(DCP_CMD_NONE, nvprops,
	hidden_args, &dcp);
	if (error != 0) {
	nvlist_free(zct.zct_zplprops);
	return (error);
	}

	error = dmu_objset_create(fsname, type,
	is_insensitive ? DS_FLAG_CI_DATASET : 0, dcp, cbfunc, &zct);

	nvlist_free(zct.zct_zplprops);
	dsl_crypto_params_free(dcp, !!error);

	/*
	* It would be nice to do this atomically.
	*/
	if (error == 0) {
	error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
	nvprops, outnvl);
	if (error != 0) {
	spa_t *spa;
	int error2;

	/*
	* Volumes will return EBUSY and cannot be destroyed
	* until all asynchronous minor handling (e.g. from
	* setting the volmode property) has completed. Wait for
	* the spa_zvol_taskq to drain then retry.
	*/
	error2 = dsl_destroy_head(fsname);
	while ((error2 == EBUSY) && (type == DMU_OST_ZVOL)) {
	error2 = spa_open(fsname, &spa, FTAG);
	if (error2 == 0) {
	taskq_wait(spa->spa_zvol_taskq);
	spa_close(spa, FTAG);
	}
	error2 = dsl_destroy_head(fsname);
	}
	}
	}
	return (error);
	}

	/*
	* innvl: {
	* "origin" -> name of origin snapshot
	* (optional) "props" -> { prop -> value }
	* (optional) "hidden_args" -> { "wkeydata" -> value }
	* raw uint8_t array of encryption wrapping key data (32 bytes)
	* }
	*
	* outputs:
	* outnvl: propname -> error code (int32)
	*/
	static const zfs_ioc_key_t zfs_keys_clone[] = {
	{"origin", DATA_TYPE_STRING, 0},
	{"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
	{"hidden_args", DATA_TYPE_NVLIST, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_clone(const char fsname, nvlist_t innvl, nvlist_t *outnvl)
	{
	int error = 0;
	nvlist_t *nvprops = NULL;
	const char *origin_name;

	origin_name = fnvlist_lookup_string(innvl, "origin");
	(void) nvlist_lookup_nvlist(innvl, "props", &nvprops);

	if (strchr(fsname, '@') \|\|
	strchr(fsname, '%'))
	return (SET_ERROR(EINVAL));

	if (dataset_namecheck(origin_name, NULL, NULL) != 0)
	return (SET_ERROR(EINVAL));

	error = dmu_objset_clone(fsname, origin_name);

	/*
	* It would be nice to do this atomically.
	*/
	if (error == 0) {
	error = zfs_set_prop_nvlist(fsname, ZPROP_SRC_LOCAL,
	nvprops, outnvl);
	if (error != 0)
	(void) dsl_destroy_head(fsname);
	}
	return (error);
	}

	static const zfs_ioc_key_t zfs_keys_remap[] = {
	/* no nvl keys */
	};

	static int
	zfs_ioc_remap(const char fsname, nvlist_t innvl, nvlist_t *outnvl)
	{
	/* This IOCTL is no longer supported. */
	(void) fsname, (void) innvl, (void) outnvl;
	return (0);
	}

	/*
	* innvl: {
	* "snaps" -> { snapshot1, snapshot2 }
	* (optional) "props" -> { prop -> value (string) }
	* }
	*
	* outnvl: snapshot -> error code (int32)
	*/
	static const zfs_ioc_key_t zfs_keys_snapshot[] = {
	{"snaps", DATA_TYPE_NVLIST, 0},
	{"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_snapshot(const char poolname, nvlist_t innvl, nvlist_t *outnvl)
	{
	nvlist_t *snaps;
	nvlist_t *props = NULL;
	int error, poollen;
	nvpair_t *pair;

	(void) nvlist_lookup_nvlist(innvl, "props", &props);
	if (!nvlist_empty(props) &&
	zfs_earlier_version(poolname, SPA_VERSION_SNAP_PROPS))
	return (SET_ERROR(ENOTSUP));
	if ((error = zfs_check_userprops(props)) != 0)
	return (error);

	snaps = fnvlist_lookup_nvlist(innvl, "snaps");
	poollen = strlen(poolname);
	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
	pair = nvlist_next_nvpair(snaps, pair)) {
	const char *name = nvpair_name(pair);
	char *cp = strchr(name, '@');

	/*
	* The snap name must contain an @, and the part after it must
	* contain only valid characters.
	*/
	if (cp == NULL \|\|
	zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
	return (SET_ERROR(EINVAL));

	/*
	* The snap must be in the specified pool.
	*/
	if (strncmp(name, poolname, poollen) != 0 \|\|
	(name[poollen] != '/' && name[poollen] != '@'))
	return (SET_ERROR(EXDEV));

	/*
	* Check for permission to set the properties on the fs.
	*/
	if (!nvlist_empty(props)) {
	*cp = '\0';
	error = zfs_secpolicy_write_perms(name,
	ZFS_DELEG_PERM_USERPROP, CRED());
	*cp = '@';
	if (error != 0)
	return (error);
	}

	/* This must be the only snap of this fs. */
	for (nvpair_t *pair2 = nvlist_next_nvpair(snaps, pair);
	pair2 != NULL; pair2 = nvlist_next_nvpair(snaps, pair2)) {
	if (strncmp(name, nvpair_name(pair2), cp - name + 1)
	== 0) {
	return (SET_ERROR(EXDEV));
	}
	}
	}

	error = dsl_dataset_snapshot(snaps, props, outnvl);

	return (error);
	}

	/*
	* innvl: "message" -> string
	*/
	static const zfs_ioc_key_t zfs_keys_log_history[] = {
	{"message", DATA_TYPE_STRING, 0},
	};

	static int
	zfs_ioc_log_history(const char unused, nvlist_t innvl, nvlist_t *outnvl)
	{
	(void) unused, (void) outnvl;
	const char *message;
	char *poolname;
	spa_t *spa;
	int error;

	/*
	* The poolname in the ioctl is not set, we get it from the TSD,
	* which was set at the end of the last successful ioctl that allows
	* logging. The secpolicy func already checked that it is set.
	* Only one log ioctl is allowed after each successful ioctl, so
	* we clear the TSD here.
	*/
	poolname = tsd_get(zfs_allow_log_key);
	if (poolname == NULL)
	return (SET_ERROR(EINVAL));
	(void) tsd_set(zfs_allow_log_key, NULL);
	error = spa_open(poolname, &spa, FTAG);
	kmem_strfree(poolname);
	if (error != 0)
	return (error);

	message = fnvlist_lookup_string(innvl, "message");

	if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY) {
	spa_close(spa, FTAG);
	return (SET_ERROR(ENOTSUP));
	}

	error = spa_history_log(spa, message);
	spa_close(spa, FTAG);
	return (error);
	}

	/*
	* This ioctl is used to set the bootenv configuration on the current
	* pool. This configuration is stored in the second padding area of the label,
	* and it is used by the bootloader(s) to store the bootloader and/or system
	* specific data.
	* The data is stored as nvlist data stream, and is protected by
	* an embedded checksum.
	* The version can have two possible values:
	* VB_RAW: nvlist should have key GRUB_ENVMAP, value DATA_TYPE_STRING.
	* VB_NVLIST: nvlist with arbitrary <key, value> pairs.
	*/
	static const zfs_ioc_key_t zfs_keys_set_bootenv[] = {
	{"version", DATA_TYPE_UINT64, 0},
	{"<keys>", DATA_TYPE_ANY, ZK_OPTIONAL \| ZK_WILDCARDLIST},
	};

	static int
	zfs_ioc_set_bootenv(const char name, nvlist_t innvl, nvlist_t *outnvl)
	{
	int error;
	spa_t *spa;

	if ((error = spa_open(name, &spa, FTAG)) != 0)
	return (error);
	spa_vdev_state_enter(spa, SCL_ALL);
	error = vdev_label_write_bootenv(spa->spa_root_vdev, innvl);
	(void) spa_vdev_state_exit(spa, NULL, 0);
	spa_close(spa, FTAG);
	return (error);
	}

	static const zfs_ioc_key_t zfs_keys_get_bootenv[] = {
	/* no nvl keys */
	};

	static int
	zfs_ioc_get_bootenv(const char name, nvlist_t innvl, nvlist_t *outnvl)
	{
	spa_t *spa;
	int error;

	if ((error = spa_open(name, &spa, FTAG)) != 0)
	return (error);
	spa_vdev_state_enter(spa, SCL_ALL);
	error = vdev_label_read_bootenv(spa->spa_root_vdev, outnvl);
	(void) spa_vdev_state_exit(spa, NULL, 0);
	spa_close(spa, FTAG);
	return (error);
	}

	/*
	* The dp_config_rwlock must not be held when calling this, because the
	* unmount may need to write out data.
	*
	* This function is best-effort. Callers must deal gracefully if it
	* remains mounted (or is remounted after this call).
	*
	* Returns 0 if the argument is not a snapshot, or it is not currently a
	* filesystem, or we were able to unmount it. Returns error code otherwise.
	*/
	void
	zfs_unmount_snap(const char *snapname)
	{
	if (strchr(snapname, '@') == NULL)
	return;

	(void) zfsctl_snapshot_unmount(snapname, MNT_FORCE);
	}

	static int
	zfs_unmount_snap_cb(const char snapname, void arg)
	{
	(void) arg;
	zfs_unmount_snap(snapname);
	return (0);
	}

	/*
	* When a clone is destroyed, its origin may also need to be destroyed,
	* in which case it must be unmounted. This routine will do that unmount
	* if necessary.
	*/
	void
	zfs_destroy_unmount_origin(const char *fsname)
	{
	int error;
	objset_t *os;
	dsl_dataset_t *ds;

	error = dmu_objset_hold(fsname, FTAG, &os);
	if (error != 0)
	return;
	ds = dmu_objset_ds(os);
	if (dsl_dir_is_clone(ds->ds_dir) && DS_IS_DEFER_DESTROY(ds->ds_prev)) {
	char originname[ZFS_MAX_DATASET_NAME_LEN];
	dsl_dataset_name(ds->ds_prev, originname);
	dmu_objset_rele(os, FTAG);
	zfs_unmount_snap(originname);
	} else {
	dmu_objset_rele(os, FTAG);
	}
	}

	/*
	* innvl: {
	* "snaps" -> { snapshot1, snapshot2 }
	* (optional boolean) "defer"
	* }
	*
	* outnvl: snapshot -> error code (int32)
	*/
	static const zfs_ioc_key_t zfs_keys_destroy_snaps[] = {
	{"snaps", DATA_TYPE_NVLIST, 0},
	{"defer", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_destroy_snaps(const char poolname, nvlist_t innvl, nvlist_t *outnvl)
	{
	int poollen;
	nvlist_t *snaps;
	nvpair_t *pair;
	boolean_t defer;
	spa_t *spa;

	snaps = fnvlist_lookup_nvlist(innvl, "snaps");
	defer = nvlist_exists(innvl, "defer");

	poollen = strlen(poolname);
	for (pair = nvlist_next_nvpair(snaps, NULL); pair != NULL;
	pair = nvlist_next_nvpair(snaps, pair)) {
	const char *name = nvpair_name(pair);

	/*
	* The snap must be in the specified pool to prevent the
	* invalid removal of zvol minors below.
	*/
	if (strncmp(name, poolname, poollen) != 0 \|\|
	(name[poollen] != '/' && name[poollen] != '@'))
	return (SET_ERROR(EXDEV));

	zfs_unmount_snap(nvpair_name(pair));
	if (spa_open(name, &spa, FTAG) == 0) {
	zvol_remove_minors(spa, name, B_TRUE);
	spa_close(spa, FTAG);
	}
	}

	return (dsl_destroy_snapshots_nvl(snaps, defer, outnvl));
	}

	/*
	* Create bookmarks. The bookmark names are of the form <fs>#<bmark>.
	* All bookmarks and snapshots must be in the same pool.
	* dsl_bookmark_create_nvl_validate describes the nvlist schema in more detail.
	*
	* innvl: {
	* new_bookmark1 -> existing_snapshot,
	* new_bookmark2 -> existing_bookmark,
	* }
	*
	* outnvl: bookmark -> error code (int32)
	*
	*/
	static const zfs_ioc_key_t zfs_keys_bookmark[] = {
	{"<bookmark>...", DATA_TYPE_STRING, ZK_WILDCARDLIST},
	};

	static int
	zfs_ioc_bookmark(const char poolname, nvlist_t innvl, nvlist_t *outnvl)
	{
	(void) poolname;
	return (dsl_bookmark_create(innvl, outnvl));
	}

	/*
	* innvl: {
	* property 1, property 2, ...
	* }
	*
	* outnvl: {
	* bookmark name 1 -> { property 1, property 2, ... },
	* bookmark name 2 -> { property 1, property 2, ... }
	* }
	*
	*/
	static const zfs_ioc_key_t zfs_keys_get_bookmarks[] = {
	{"<property>...", DATA_TYPE_BOOLEAN, ZK_WILDCARDLIST \| ZK_OPTIONAL},
	};

	static int
	zfs_ioc_get_bookmarks(const char fsname, nvlist_t innvl, nvlist_t *outnvl)
	{
	return (dsl_get_bookmarks(fsname, innvl, outnvl));
	}

	/*
	* innvl is not used.
	*
	* outnvl: {
	* property 1, property 2, ...
	* }
	*
	*/
	static const zfs_ioc_key_t zfs_keys_get_bookmark_props[] = {
	/* no nvl keys */
	};

	static int
	zfs_ioc_get_bookmark_props(const char bookmark, nvlist_t innvl,
	nvlist_t *outnvl)
	{
	(void) innvl;
	char fsname[ZFS_MAX_DATASET_NAME_LEN];
	char *bmname;

	bmname = strchr(bookmark, '#');
	if (bmname == NULL)
	return (SET_ERROR(EINVAL));
	bmname++;

	(void) strlcpy(fsname, bookmark, sizeof (fsname));
	*(strchr(fsname, '#')) = '\0';

	return (dsl_get_bookmark_props(fsname, bmname, outnvl));
	}

	/*
	* innvl: {
	* bookmark name 1, bookmark name 2
	* }
	*
	* outnvl: bookmark -> error code (int32)
	*
	*/
	static const zfs_ioc_key_t zfs_keys_destroy_bookmarks[] = {
	{"<bookmark>...", DATA_TYPE_BOOLEAN, ZK_WILDCARDLIST},
	};

	static int
	zfs_ioc_destroy_bookmarks(const char poolname, nvlist_t innvl,
	nvlist_t *outnvl)
	{
	int error, poollen;

	poollen = strlen(poolname);
	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
	pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
	const char *name = nvpair_name(pair);
	const char *cp = strchr(name, '#');

	/*
	* The bookmark name must contain an #, and the part after it
	* must contain only valid characters.
	*/
	if (cp == NULL \|\|
	zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
	return (SET_ERROR(EINVAL));

	/*
	* The bookmark must be in the specified pool.
	*/
	if (strncmp(name, poolname, poollen) != 0 \|\|
	(name[poollen] != '/' && name[poollen] != '#'))
	return (SET_ERROR(EXDEV));
	}

	error = dsl_bookmark_destroy(innvl, outnvl);
	return (error);
	}

	static const zfs_ioc_key_t zfs_keys_channel_program[] = {
	{"program", DATA_TYPE_STRING, 0},
	{"arg", DATA_TYPE_ANY, 0},
	{"sync", DATA_TYPE_BOOLEAN_VALUE, ZK_OPTIONAL},
	{"instrlimit", DATA_TYPE_UINT64, ZK_OPTIONAL},
	{"memlimit", DATA_TYPE_UINT64, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_channel_program(const char poolname, nvlist_t innvl,
	nvlist_t *outnvl)
	{
	const char *program;
	uint64_t instrlimit, memlimit;
	boolean_t sync_flag;
	nvpair_t *nvarg = NULL;

	program = fnvlist_lookup_string(innvl, ZCP_ARG_PROGRAM);
	if (0 != nvlist_lookup_boolean_value(innvl, ZCP_ARG_SYNC, &sync_flag)) {
	sync_flag = B_TRUE;
	}
	if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_INSTRLIMIT, &instrlimit)) {
	instrlimit = ZCP_DEFAULT_INSTRLIMIT;
	}
	if (0 != nvlist_lookup_uint64(innvl, ZCP_ARG_MEMLIMIT, &memlimit)) {
	memlimit = ZCP_DEFAULT_MEMLIMIT;
	}
	nvarg = fnvlist_lookup_nvpair(innvl, ZCP_ARG_ARGLIST);

	if (instrlimit == 0 \|\| instrlimit > zfs_lua_max_instrlimit)
	return (SET_ERROR(EINVAL));
	if (memlimit == 0 \|\| memlimit > zfs_lua_max_memlimit)
	return (SET_ERROR(EINVAL));

	return (zcp_eval(poolname, program, sync_flag, instrlimit, memlimit,
	nvarg, outnvl));
	}

	/*
	* innvl: unused
	* outnvl: empty
	*/
	static const zfs_ioc_key_t zfs_keys_pool_checkpoint[] = {
	/* no nvl keys */
	};

	static int
	zfs_ioc_pool_checkpoint(const char poolname, nvlist_t innvl, nvlist_t *outnvl)
	{
	(void) innvl, (void) outnvl;
	return (spa_checkpoint(poolname));
	}

	/*
	* innvl: unused
	* outnvl: empty
	*/
	static const zfs_ioc_key_t zfs_keys_pool_discard_checkpoint[] = {
	/* no nvl keys */
	};

	static int
	zfs_ioc_pool_discard_checkpoint(const char poolname, nvlist_t innvl,
	nvlist_t *outnvl)
	{
	(void) innvl, (void) outnvl;
	return (spa_checkpoint_discard(poolname));
	}

	/*
	* inputs:
	* zc_name name of dataset to destroy
	* zc_defer_destroy mark for deferred destroy
	*
	* outputs: none
	*/
	static int
	zfs_ioc_destroy(zfs_cmd_t *zc)
	{
	objset_t *os;
	dmu_objset_type_t ost;
	int err;

	err = dmu_objset_hold(zc->zc_name, FTAG, &os);
	if (err != 0)
	return (err);
	ost = dmu_objset_type(os);
	dmu_objset_rele(os, FTAG);

	if (ost == DMU_OST_ZFS)
	zfs_unmount_snap(zc->zc_name);

	if (strchr(zc->zc_name, '@')) {
	err = dsl_destroy_snapshot(zc->zc_name, zc->zc_defer_destroy);
	} else {
	err = dsl_destroy_head(zc->zc_name);
	if (err == EEXIST) {
	/*
	* It is possible that the given DS may have
	* hidden child (%recv) datasets - "leftovers"
	* resulting from the previously interrupted
	* 'zfs receive'.
	*
	* 6 extra bytes for /%recv
	*/
	char namebuf[ZFS_MAX_DATASET_NAME_LEN + 6];

	if (snprintf(namebuf, sizeof (namebuf), "%s/%s",
	zc->zc_name, recv_clone_name) >=
	sizeof (namebuf))
	return (SET_ERROR(EINVAL));

	/*
	* Try to remove the hidden child (%recv) and after
	* that try to remove the target dataset.
	* If the hidden child (%recv) does not exist
	* the original error (EEXIST) will be returned
	*/
	err = dsl_destroy_head(namebuf);
	if (err == 0)
	err = dsl_destroy_head(zc->zc_name);
	else if (err == ENOENT)
	err = SET_ERROR(EEXIST);
	}
	}

	return (err);
	}

	/*
	* innvl: {
	* "initialize_command" -> POOL_INITIALIZE_{CANCEL\|START\|SUSPEND} (uint64)
	* "initialize_vdevs": { -> guids to initialize (nvlist)
	* "vdev_path_1": vdev_guid_1, (uint64),
	* "vdev_path_2": vdev_guid_2, (uint64),
	* ...
	* },
	* }
	*
	* outnvl: {
	* "initialize_vdevs": { -> initialization errors (nvlist)
	* "vdev_path_1": errno, see function body for possible errnos (uint64)
	* "vdev_path_2": errno, ... (uint64)
	* ...
	* }
	* }
	*
	* EINVAL is returned for an unknown commands or if any of the provided vdev
	* guids have be specified with a type other than uint64.
	*/
	static const zfs_ioc_key_t zfs_keys_pool_initialize[] = {
	{ZPOOL_INITIALIZE_COMMAND, DATA_TYPE_UINT64, 0},
	{ZPOOL_INITIALIZE_VDEVS, DATA_TYPE_NVLIST, 0}
	};

	static int
	zfs_ioc_pool_initialize(const char poolname, nvlist_t innvl, nvlist_t *outnvl)
	{
	uint64_t cmd_type;
	if (nvlist_lookup_uint64(innvl, ZPOOL_INITIALIZE_COMMAND,
	&cmd_type) != 0) {
	return (SET_ERROR(EINVAL));
	}

	if (!(cmd_type == POOL_INITIALIZE_CANCEL \|\|
	cmd_type == POOL_INITIALIZE_START \|\|
	cmd_type == POOL_INITIALIZE_SUSPEND \|\|
	cmd_type == POOL_INITIALIZE_UNINIT)) {
	return (SET_ERROR(EINVAL));
	}

	nvlist_t *vdev_guids;
	if (nvlist_lookup_nvlist(innvl, ZPOOL_INITIALIZE_VDEVS,
	&vdev_guids) != 0) {
	return (SET_ERROR(EINVAL));
	}

	for (nvpair_t *pair = nvlist_next_nvpair(vdev_guids, NULL);
	pair != NULL; pair = nvlist_next_nvpair(vdev_guids, pair)) {
	uint64_t vdev_guid;
	if (nvpair_value_uint64(pair, &vdev_guid) != 0) {
	return (SET_ERROR(EINVAL));
	}
	}

	spa_t *spa;
	int error = spa_open(poolname, &spa, FTAG);
	if (error != 0)
	return (error);

	nvlist_t *vdev_errlist = fnvlist_alloc();
	int total_errors = spa_vdev_initialize(spa, vdev_guids, cmd_type,
	vdev_errlist);

	if (fnvlist_size(vdev_errlist) > 0) {
	fnvlist_add_nvlist(outnvl, ZPOOL_INITIALIZE_VDEVS,
	vdev_errlist);
	}
	fnvlist_free(vdev_errlist);

	spa_close(spa, FTAG);
	return (total_errors > 0 ? SET_ERROR(EINVAL) : 0);
	}

	/*
	* innvl: {
	* "trim_command" -> POOL_TRIM_{CANCEL\|START\|SUSPEND} (uint64)
	* "trim_vdevs": { -> guids to TRIM (nvlist)
	* "vdev_path_1": vdev_guid_1, (uint64),
	* "vdev_path_2": vdev_guid_2, (uint64),
	* ...
	* },
	* "trim_rate" -> Target TRIM rate in bytes/sec.
	* "trim_secure" -> Set to request a secure TRIM.
	* }
	*
	* outnvl: {
	* "trim_vdevs": { -> TRIM errors (nvlist)
	* "vdev_path_1": errno, see function body for possible errnos (uint64)
	* "vdev_path_2": errno, ... (uint64)
	* ...
	* }
	* }
	*
	* EINVAL is returned for an unknown commands or if any of the provided vdev
	* guids have be specified with a type other than uint64.
	*/
	static const zfs_ioc_key_t zfs_keys_pool_trim[] = {
	{ZPOOL_TRIM_COMMAND, DATA_TYPE_UINT64, 0},
	{ZPOOL_TRIM_VDEVS, DATA_TYPE_NVLIST, 0},
	{ZPOOL_TRIM_RATE, DATA_TYPE_UINT64, ZK_OPTIONAL},
	{ZPOOL_TRIM_SECURE, DATA_TYPE_BOOLEAN_VALUE, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_pool_trim(const char poolname, nvlist_t innvl, nvlist_t *outnvl)
	{
	uint64_t cmd_type;
	if (nvlist_lookup_uint64(innvl, ZPOOL_TRIM_COMMAND, &cmd_type) != 0)
	return (SET_ERROR(EINVAL));

	if (!(cmd_type == POOL_TRIM_CANCEL \|\|
	cmd_type == POOL_TRIM_START \|\|
	cmd_type == POOL_TRIM_SUSPEND)) {
	return (SET_ERROR(EINVAL));
	}

	nvlist_t *vdev_guids;
	if (nvlist_lookup_nvlist(innvl, ZPOOL_TRIM_VDEVS, &vdev_guids) != 0)
	return (SET_ERROR(EINVAL));

	for (nvpair_t *pair = nvlist_next_nvpair(vdev_guids, NULL);
	pair != NULL; pair = nvlist_next_nvpair(vdev_guids, pair)) {
	uint64_t vdev_guid;
	if (nvpair_value_uint64(pair, &vdev_guid) != 0) {
	return (SET_ERROR(EINVAL));
	}
	}

	/* Optional, defaults to maximum rate when not provided */
	uint64_t rate;
	if (nvlist_lookup_uint64(innvl, ZPOOL_TRIM_RATE, &rate) != 0)
	rate = 0;

	/* Optional, defaults to standard TRIM when not provided */
	boolean_t secure;
	if (nvlist_lookup_boolean_value(innvl, ZPOOL_TRIM_SECURE,
	&secure) != 0) {
	secure = B_FALSE;
	}

	spa_t *spa;
	int error = spa_open(poolname, &spa, FTAG);
	if (error != 0)
	return (error);

	nvlist_t *vdev_errlist = fnvlist_alloc();
	int total_errors = spa_vdev_trim(spa, vdev_guids, cmd_type,
	rate, !!zfs_trim_metaslab_skip, secure, vdev_errlist);

	if (fnvlist_size(vdev_errlist) > 0)
	fnvlist_add_nvlist(outnvl, ZPOOL_TRIM_VDEVS, vdev_errlist);

	fnvlist_free(vdev_errlist);

	spa_close(spa, FTAG);
	return (total_errors > 0 ? SET_ERROR(EINVAL) : 0);
	}

	/*
	* This ioctl waits for activity of a particular type to complete. If there is
	* no activity of that type in progress, it returns immediately, and the
	* returned value "waited" is false. If there is activity in progress, and no
	* tag is passed in, the ioctl blocks until all activity of that type is
	* complete, and then returns with "waited" set to true.
	*
	* If a tag is provided, it identifies a particular instance of an activity to
	* wait for. Currently, this is only valid for use with 'initialize', because
	* that is the only activity for which there can be multiple instances running
	* concurrently. In the case of 'initialize', the tag corresponds to the guid of
	* the vdev on which to wait.
	*
	* If a thread waiting in the ioctl receives a signal, the call will return
	* immediately, and the return value will be EINTR.
	*
	* innvl: {
	* "wait_activity" -> int32_t
	* (optional) "wait_tag" -> uint64_t
	* }
	*
	* outnvl: "waited" -> boolean_t
	*/
	static const zfs_ioc_key_t zfs_keys_pool_wait[] = {
	{ZPOOL_WAIT_ACTIVITY, DATA_TYPE_INT32, 0},
	{ZPOOL_WAIT_TAG, DATA_TYPE_UINT64, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_wait(const char name, nvlist_t innvl, nvlist_t *outnvl)
	{
	int32_t activity;
	uint64_t tag;
	boolean_t waited;
	int error;

	if (nvlist_lookup_int32(innvl, ZPOOL_WAIT_ACTIVITY, &activity) != 0)
	return (EINVAL);

	if (nvlist_lookup_uint64(innvl, ZPOOL_WAIT_TAG, &tag) == 0)
	error = spa_wait_tag(name, activity, tag, &waited);
	else
	error = spa_wait(name, activity, &waited);

	if (error == 0)
	fnvlist_add_boolean_value(outnvl, ZPOOL_WAIT_WAITED, waited);

	return (error);
	}

	/*
	* This ioctl waits for activity of a particular type to complete. If there is
	* no activity of that type in progress, it returns immediately, and the
	* returned value "waited" is false. If there is activity in progress, and no
	* tag is passed in, the ioctl blocks until all activity of that type is
	* complete, and then returns with "waited" set to true.
	*
	* If a thread waiting in the ioctl receives a signal, the call will return
	* immediately, and the return value will be EINTR.
	*
	* innvl: {
	* "wait_activity" -> int32_t
	* }
	*
	* outnvl: "waited" -> boolean_t
	*/
	static const zfs_ioc_key_t zfs_keys_fs_wait[] = {
	{ZFS_WAIT_ACTIVITY, DATA_TYPE_INT32, 0},
	};

	static int
	zfs_ioc_wait_fs(const char name, nvlist_t innvl, nvlist_t *outnvl)
	{
	int32_t activity;
	boolean_t waited = B_FALSE;
	int error;
	dsl_pool_t *dp;
	dsl_dir_t *dd;
	dsl_dataset_t *ds;

	if (nvlist_lookup_int32(innvl, ZFS_WAIT_ACTIVITY, &activity) != 0)
	return (SET_ERROR(EINVAL));

	if (activity >= ZFS_WAIT_NUM_ACTIVITIES \|\| activity < 0)
	return (SET_ERROR(EINVAL));

	if ((error = dsl_pool_hold(name, FTAG, &dp)) != 0)
	return (error);

	if ((error = dsl_dataset_hold(dp, name, FTAG, &ds)) != 0) {
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	dd = ds->ds_dir;
	mutex_enter(&dd->dd_activity_lock);
	dd->dd_activity_waiters++;

	/*
	* We get a long-hold here so that the dsl_dataset_t and dsl_dir_t
	* aren't evicted while we're waiting. Normally this is prevented by
	* holding the pool, but we can't do that while we're waiting since
	* that would prevent TXGs from syncing out. Some of the functionality
	* of long-holds (e.g. preventing deletion) is unnecessary for this
	* case, since we would cancel the waiters before proceeding with a
	* deletion. An alternative mechanism for keeping the dataset around
	* could be developed but this is simpler.
	*/
	dsl_dataset_long_hold(ds, FTAG);
	dsl_pool_rele(dp, FTAG);

	error = dsl_dir_wait(dd, ds, activity, &waited);

	dsl_dataset_long_rele(ds, FTAG);
	dd->dd_activity_waiters--;
	if (dd->dd_activity_waiters == 0)
	cv_signal(&dd->dd_activity_cv);
	mutex_exit(&dd->dd_activity_lock);

	dsl_dataset_rele(ds, FTAG);

	if (error == 0)
	fnvlist_add_boolean_value(outnvl, ZFS_WAIT_WAITED, waited);

	return (error);
	}

	/*
	* fsname is name of dataset to rollback (to most recent snapshot)
	*
	* innvl may contain name of expected target snapshot
	*
	* outnvl: "target" -> name of most recent snapshot
	* }
	*/
	static const zfs_ioc_key_t zfs_keys_rollback[] = {
	{"target", DATA_TYPE_STRING, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_rollback(const char fsname, nvlist_t innvl, nvlist_t *outnvl)
	{
	zfsvfs_t *zfsvfs;
	zvol_state_handle_t *zv;
	const char *target = NULL;
	int error;

	(void) nvlist_lookup_string(innvl, "target", &target);
	if (target != NULL) {
	const char *cp = strchr(target, '@');

	/*
	* The snap name must contain an @, and the part after it must
	* contain only valid characters.
	*/
	if (cp == NULL \|\|
	zfs_component_namecheck(cp + 1, NULL, NULL) != 0)
	return (SET_ERROR(EINVAL));
	}

	if (getzfsvfs(fsname, &zfsvfs) == 0) {
	dsl_dataset_t *ds;

	ds = dmu_objset_ds(zfsvfs->z_os);
	error = zfs_suspend_fs(zfsvfs);
	if (error == 0) {
	int resume_err;

	error = dsl_dataset_rollback(fsname, target, zfsvfs,
	outnvl);
	resume_err = zfs_resume_fs(zfsvfs, ds);
	error = error ? error : resume_err;
	}
	zfs_vfs_rele(zfsvfs);
	} else if ((zv = zvol_suspend(fsname)) != NULL) {
	error = dsl_dataset_rollback(fsname, target, zvol_tag(zv),
	outnvl);
	zvol_resume(zv);
	} else {
	error = dsl_dataset_rollback(fsname, target, NULL, outnvl);
	}
	return (error);
	}

	static int
	recursive_unmount(const char fsname, void arg)
	{
	const char *snapname = arg;
	char *fullname;

	fullname = kmem_asprintf("%s@%s", fsname, snapname);
	zfs_unmount_snap(fullname);
	kmem_strfree(fullname);

	return (0);
	}

	/*
	*
	* snapname is the snapshot to redact.
	* innvl: {
	* "bookname" -> (string)
	* shortname of the redaction bookmark to generate
	* "snapnv" -> (nvlist, values ignored)
	* snapshots to redact snapname with respect to
	* }
	*
	* outnvl is unused
	*/

	static const zfs_ioc_key_t zfs_keys_redact[] = {
	{"bookname", DATA_TYPE_STRING, 0},
	{"snapnv", DATA_TYPE_NVLIST, 0},
	};

	static int
	zfs_ioc_redact(const char snapname, nvlist_t innvl, nvlist_t *outnvl)
	{
	(void) outnvl;
	nvlist_t *redactnvl = NULL;
	const char *redactbook = NULL;

	if (nvlist_lookup_nvlist(innvl, "snapnv", &redactnvl) != 0)
	return (SET_ERROR(EINVAL));
	if (fnvlist_num_pairs(redactnvl) == 0)
	return (SET_ERROR(ENXIO));
	if (nvlist_lookup_string(innvl, "bookname", &redactbook) != 0)
	return (SET_ERROR(EINVAL));

	return (dmu_redact_snap(snapname, redactnvl, redactbook));
	}

	/*
	* inputs:
	* zc_name old name of dataset
	* zc_value new name of dataset
	* zc_cookie recursive flag (only valid for snapshots)
	*
	* outputs: none
	*/
	static int
	zfs_ioc_rename(zfs_cmd_t *zc)
	{
	objset_t *os;
	dmu_objset_type_t ost;
	boolean_t recursive = zc->zc_cookie & 1;
	boolean_t nounmount = !!(zc->zc_cookie & 2);
	char *at;
	int err;

	/* "zfs rename" from and to ...%recv datasets should both fail */
	zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
	zc->zc_value[sizeof (zc->zc_value) - 1] = '\0';
	if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 \|\|
	dataset_namecheck(zc->zc_value, NULL, NULL) != 0 \|\|
	strchr(zc->zc_name, '%') \|\| strchr(zc->zc_value, '%'))
	return (SET_ERROR(EINVAL));

	err = dmu_objset_hold(zc->zc_name, FTAG, &os);
	if (err != 0)
	return (err);
	ost = dmu_objset_type(os);
	dmu_objset_rele(os, FTAG);

	at = strchr(zc->zc_name, '@');
	if (at != NULL) {
	/* snaps must be in same fs */
	int error;

	if (strncmp(zc->zc_name, zc->zc_value, at - zc->zc_name + 1))
	return (SET_ERROR(EXDEV));
	*at = '\0';
	if (ost == DMU_OST_ZFS && !nounmount) {
	error = dmu_objset_find(zc->zc_name,
	recursive_unmount, at + 1,
	recursive ? DS_FIND_CHILDREN : 0);
	if (error != 0) {
	*at = '@';
	return (error);
	}
	}
	error = dsl_dataset_rename_snapshot(zc->zc_name,
	at + 1, strchr(zc->zc_value, '@') + 1, recursive);
	*at = '@';

	return (error);
	} else {
	return (dsl_dir_rename(zc->zc_name, zc->zc_value));
	}
	}

	static int
	zfs_check_settable(const char dsname, nvpair_t pair, cred_t *cr)
	{
	const char *propname = nvpair_name(pair);
	boolean_t issnap = (strchr(dsname, '@') != NULL);
	zfs_prop_t prop = zfs_name_to_prop(propname);
	uint64_t intval, compval;
	int err;

	if (prop == ZPROP_USERPROP) {
	if (zfs_prop_user(propname)) {
	if ((err = zfs_secpolicy_write_perms(dsname,
	ZFS_DELEG_PERM_USERPROP, cr)))
	return (err);
	return (0);
	}

	if (!issnap && zfs_prop_userquota(propname)) {
	const char *perm = NULL;
	const char *uq_prefix =
	zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA];
	const char *gq_prefix =
	zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA];
	const char *uiq_prefix =
	zfs_userquota_prop_prefixes[ZFS_PROP_USEROBJQUOTA];
	const char *giq_prefix =
	zfs_userquota_prop_prefixes[ZFS_PROP_GROUPOBJQUOTA];
	const char *pq_prefix =
	zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTQUOTA];
	const char *piq_prefix = zfs_userquota_prop_prefixes[\
	ZFS_PROP_PROJECTOBJQUOTA];

	if (strncmp(propname, uq_prefix,
	strlen(uq_prefix)) == 0) {
	perm = ZFS_DELEG_PERM_USERQUOTA;
	} else if (strncmp(propname, uiq_prefix,
	strlen(uiq_prefix)) == 0) {
	perm = ZFS_DELEG_PERM_USEROBJQUOTA;
	} else if (strncmp(propname, gq_prefix,
	strlen(gq_prefix)) == 0) {
	perm = ZFS_DELEG_PERM_GROUPQUOTA;
	} else if (strncmp(propname, giq_prefix,
	strlen(giq_prefix)) == 0) {
	perm = ZFS_DELEG_PERM_GROUPOBJQUOTA;
	} else if (strncmp(propname, pq_prefix,
	strlen(pq_prefix)) == 0) {
	perm = ZFS_DELEG_PERM_PROJECTQUOTA;
	} else if (strncmp(propname, piq_prefix,
	strlen(piq_prefix)) == 0) {
	perm = ZFS_DELEG_PERM_PROJECTOBJQUOTA;
	} else {
	/* {USER\|GROUP\|PROJECT}USED are read-only */
	return (SET_ERROR(EINVAL));
	}

	if ((err = zfs_secpolicy_write_perms(dsname, perm, cr)))
	return (err);
	return (0);
	}

	return (SET_ERROR(EINVAL));
	}

	if (issnap)
	return (SET_ERROR(EINVAL));

	if (nvpair_type(pair) == DATA_TYPE_NVLIST) {
	/*
	* dsl_prop_get_all_impl() returns properties in this
	* format.
	*/
	nvlist_t *attrs;
	VERIFY(nvpair_value_nvlist(pair, &attrs) == 0);
	VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
	&pair) == 0);
	}

	/*
	* Check that this value is valid for this pool version
	*/
	switch (prop) {
	case ZFS_PROP_COMPRESSION:
	/*
	* If the user specified gzip compression, make sure
	* the SPA supports it. We ignore any errors here since
	* we'll catch them later.
	*/
	if (nvpair_value_uint64(pair, &intval) == 0) {
	compval = ZIO_COMPRESS_ALGO(intval);
	if (compval >= ZIO_COMPRESS_GZIP_1 &&
	compval <= ZIO_COMPRESS_GZIP_9 &&
	zfs_earlier_version(dsname,
	SPA_VERSION_GZIP_COMPRESSION)) {
	return (SET_ERROR(ENOTSUP));
	}

	if (compval == ZIO_COMPRESS_ZLE &&
	zfs_earlier_version(dsname,
	SPA_VERSION_ZLE_COMPRESSION))
	return (SET_ERROR(ENOTSUP));

	if (compval == ZIO_COMPRESS_LZ4) {
	spa_t *spa;

	if ((err = spa_open(dsname, &spa, FTAG)) != 0)
	return (err);

	if (!spa_feature_is_enabled(spa,
	SPA_FEATURE_LZ4_COMPRESS)) {
	spa_close(spa, FTAG);
	return (SET_ERROR(ENOTSUP));
	}
	spa_close(spa, FTAG);
	}

	if (compval == ZIO_COMPRESS_ZSTD) {
	spa_t *spa;

	if ((err = spa_open(dsname, &spa, FTAG)) != 0)
	return (err);

	if (!spa_feature_is_enabled(spa,
	SPA_FEATURE_ZSTD_COMPRESS)) {
	spa_close(spa, FTAG);
	return (SET_ERROR(ENOTSUP));
	}
	spa_close(spa, FTAG);
	}
	}
	break;

	case ZFS_PROP_COPIES:
	if (zfs_earlier_version(dsname, SPA_VERSION_DITTO_BLOCKS))
	return (SET_ERROR(ENOTSUP));
	break;

	case ZFS_PROP_VOLBLOCKSIZE:
	case ZFS_PROP_RECORDSIZE:
	/* Record sizes above 128k need the feature to be enabled */
	if (nvpair_value_uint64(pair, &intval) == 0 &&
	intval > SPA_OLD_MAXBLOCKSIZE) {
	spa_t *spa;

	/*
	* We don't allow setting the property above 1MB,
	* unless the tunable has been changed.
	*/
	if (intval > zfs_max_recordsize \|\|
	intval > SPA_MAXBLOCKSIZE)
	return (SET_ERROR(ERANGE));

	if ((err = spa_open(dsname, &spa, FTAG)) != 0)
	return (err);

	if (!spa_feature_is_enabled(spa,
	SPA_FEATURE_LARGE_BLOCKS)) {
	spa_close(spa, FTAG);
	return (SET_ERROR(ENOTSUP));
	}
	spa_close(spa, FTAG);
	}
	break;

	case ZFS_PROP_DNODESIZE:
	/* Dnode sizes above 512 need the feature to be enabled */
	if (nvpair_value_uint64(pair, &intval) == 0 &&
	intval != ZFS_DNSIZE_LEGACY) {
	spa_t *spa;

	if ((err = spa_open(dsname, &spa, FTAG)) != 0)
	return (err);

	if (!spa_feature_is_enabled(spa,
	SPA_FEATURE_LARGE_DNODE)) {
	spa_close(spa, FTAG);
	return (SET_ERROR(ENOTSUP));
	}
	spa_close(spa, FTAG);
	}
	break;

	case ZFS_PROP_SPECIAL_SMALL_BLOCKS:
	/*
	* This property could require the allocation classes
	* feature to be active for setting, however we allow
	* it so that tests of settable properties succeed.
	* The CLI will issue a warning in this case.
	*/
	break;

	case ZFS_PROP_SHARESMB:
	if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
	return (SET_ERROR(ENOTSUP));
	break;

	case ZFS_PROP_ACLINHERIT:
	if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
	nvpair_value_uint64(pair, &intval) == 0) {
	if (intval == ZFS_ACL_PASSTHROUGH_X &&
	zfs_earlier_version(dsname,
	SPA_VERSION_PASSTHROUGH_X))
	return (SET_ERROR(ENOTSUP));
	}
	break;
	case ZFS_PROP_CHECKSUM:
	case ZFS_PROP_DEDUP:
	{
	spa_feature_t feature;
	spa_t *spa;
	int err;

	/* dedup feature version checks */
	if (prop == ZFS_PROP_DEDUP &&
	zfs_earlier_version(dsname, SPA_VERSION_DEDUP))
	return (SET_ERROR(ENOTSUP));

	if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
	nvpair_value_uint64(pair, &intval) == 0) {
	/* check prop value is enabled in features */
	feature = zio_checksum_to_feature(
	intval & ZIO_CHECKSUM_MASK);
	if (feature == SPA_FEATURE_NONE)
	break;

	if ((err = spa_open(dsname, &spa, FTAG)) != 0)
	return (err);

	if (!spa_feature_is_enabled(spa, feature)) {
	spa_close(spa, FTAG);
	return (SET_ERROR(ENOTSUP));
	}
	spa_close(spa, FTAG);
	}
	break;
	}

	default:
	break;
	}

	return (zfs_secpolicy_setprop(dsname, prop, pair, CRED()));
	}

	/*
	* Removes properties from the given props list that fail permission checks
	* needed to clear them and to restore them in case of a receive error. For each
	* property, make sure we have both set and inherit permissions.
	*
	* Returns the first error encountered if any permission checks fail. If the
	* caller provides a non-NULL errlist, it also gives the complete list of names
	* of all the properties that failed a permission check along with the
	* corresponding error numbers. The caller is responsible for freeing the
	* returned errlist.
	*
	* If every property checks out successfully, zero is returned and the list
	* pointed at by errlist is NULL.
	*/
	static int
	zfs_check_clearable(const char dataset, nvlist_t props, nvlist_t **errlist)
	{
	zfs_cmd_t *zc;
	nvpair_t pair, next_pair;
	nvlist_t *errors;
	int err, rv = 0;

	if (props == NULL)
	return (0);

	VERIFY(nvlist_alloc(&errors, NV_UNIQUE_NAME, KM_SLEEP) == 0);

	zc = kmem_alloc(sizeof (zfs_cmd_t), KM_SLEEP);
	(void) strlcpy(zc->zc_name, dataset, sizeof (zc->zc_name));
	pair = nvlist_next_nvpair(props, NULL);
	while (pair != NULL) {
	next_pair = nvlist_next_nvpair(props, pair);

	(void) strlcpy(zc->zc_value, nvpair_name(pair),
	sizeof (zc->zc_value));
	if ((err = zfs_check_settable(dataset, pair, CRED())) != 0 \|\|
	(err = zfs_secpolicy_inherit_prop(zc, NULL, CRED())) != 0) {
	VERIFY(nvlist_remove_nvpair(props, pair) == 0);
	VERIFY(nvlist_add_int32(errors,
	zc->zc_value, err) == 0);
	}
	pair = next_pair;
	}
	kmem_free(zc, sizeof (zfs_cmd_t));

	if ((pair = nvlist_next_nvpair(errors, NULL)) == NULL) {
	nvlist_free(errors);
	errors = NULL;
	} else {
	VERIFY(nvpair_value_int32(pair, &rv) == 0);
	}

	if (errlist == NULL)
	nvlist_free(errors);
	else
	*errlist = errors;

	return (rv);
	}

	static boolean_t
	propval_equals(nvpair_t p1, nvpair_t p2)
	{
	if (nvpair_type(p1) == DATA_TYPE_NVLIST) {
	/* dsl_prop_get_all_impl() format */
	nvlist_t *attrs;
	VERIFY(nvpair_value_nvlist(p1, &attrs) == 0);
	VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
	&p1) == 0);
	}

	if (nvpair_type(p2) == DATA_TYPE_NVLIST) {
	nvlist_t *attrs;
	VERIFY(nvpair_value_nvlist(p2, &attrs) == 0);
	VERIFY(nvlist_lookup_nvpair(attrs, ZPROP_VALUE,
	&p2) == 0);
	}

	if (nvpair_type(p1) != nvpair_type(p2))
	return (B_FALSE);

	if (nvpair_type(p1) == DATA_TYPE_STRING) {
	const char valstr1, valstr2;

	VERIFY(nvpair_value_string(p1, &valstr1) == 0);
	VERIFY(nvpair_value_string(p2, &valstr2) == 0);
	return (strcmp(valstr1, valstr2) == 0);
	} else {
	uint64_t intval1, intval2;

	VERIFY(nvpair_value_uint64(p1, &intval1) == 0);
	VERIFY(nvpair_value_uint64(p2, &intval2) == 0);
	return (intval1 == intval2);
	}
	}

	/*
	* Remove properties from props if they are not going to change (as determined
	* by comparison with origprops). Remove them from origprops as well, since we
	* do not need to clear or restore properties that won't change.
	*/
	static void
	props_reduce(nvlist_t props, nvlist_t origprops)
	{
	nvpair_t pair, next_pair;

	if (origprops == NULL)
	return; /* all props need to be received */

	pair = nvlist_next_nvpair(props, NULL);
	while (pair != NULL) {
	const char *propname = nvpair_name(pair);
	nvpair_t *match;

	next_pair = nvlist_next_nvpair(props, pair);

	if ((nvlist_lookup_nvpair(origprops, propname,
	&match) != 0) \|\| !propval_equals(pair, match))
	goto next; /* need to set received value */

	/* don't clear the existing received value */
	(void) nvlist_remove_nvpair(origprops, match);
	/* don't bother receiving the property */
	(void) nvlist_remove_nvpair(props, pair);
	next:
	pair = next_pair;
	}
	}

	/*
	* Extract properties that cannot be set PRIOR to the receipt of a dataset.
	* For example, refquota cannot be set until after the receipt of a dataset,
	* because in replication streams, an older/earlier snapshot may exceed the
	* refquota. We want to receive the older/earlier snapshot, but setting
	* refquota pre-receipt will set the dsl's ACTUAL quota, which will prevent
	* the older/earlier snapshot from being received (with EDQUOT).
	*
	* The ZFS test "zfs_receive_011_pos" demonstrates such a scenario.
	*
	* libzfs will need to be judicious handling errors encountered by props
	* extracted by this function.
	*/
	static nvlist_t *
	extract_delay_props(nvlist_t *props)
	{
	nvlist_t *delayprops;
	nvpair_t nvp, tmp;
	static const zfs_prop_t delayable[] = {
	ZFS_PROP_REFQUOTA,
	ZFS_PROP_KEYLOCATION,
	/*
	* Setting ZFS_PROP_SHARESMB requires the objset type to be
	* known, which is not possible prior to receipt of raw sends.
	*/
	ZFS_PROP_SHARESMB,
	0
	};
	int i;

	VERIFY(nvlist_alloc(&delayprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);

	for (nvp = nvlist_next_nvpair(props, NULL); nvp != NULL;
	nvp = nvlist_next_nvpair(props, nvp)) {
	/*
	* strcmp() is safe because zfs_prop_to_name() always returns
	* a bounded string.
	*/
	for (i = 0; delayable[i] != 0; i++) {
	if (strcmp(zfs_prop_to_name(delayable[i]),
	nvpair_name(nvp)) == 0) {
	break;
	}
	}
	if (delayable[i] != 0) {
	tmp = nvlist_prev_nvpair(props, nvp);
	VERIFY(nvlist_add_nvpair(delayprops, nvp) == 0);
	VERIFY(nvlist_remove_nvpair(props, nvp) == 0);
	nvp = tmp;
	}
	}

	if (nvlist_empty(delayprops)) {
	nvlist_free(delayprops);
	delayprops = NULL;
	}
	return (delayprops);
	}

	static void
	zfs_allow_log_destroy(void *arg)
	{
	char *poolname = arg;

	if (poolname != NULL)
	kmem_strfree(poolname);
	}

	#ifdef ZFS_DEBUG
	static boolean_t zfs_ioc_recv_inject_err;
	#endif

	/*
	* nvlist 'errors' is always allocated. It will contain descriptions of
	* encountered errors, if any. It's the callers responsibility to free.
	*/
	static int
	zfs_ioc_recv_impl(char tofs, char tosnap, const char *origin,
	nvlist_t recvprops, nvlist_t localprops, nvlist_t *hidden_args,
	boolean_t force, boolean_t heal, boolean_t resumable, int input_fd,
	dmu_replay_record_t begin_record, uint64_t read_bytes,
	uint64_t errflags, nvlist_t *errors)
	{
	dmu_recv_cookie_t drc;
	int error = 0;
	int props_error = 0;
	offset_t off, noff;
	nvlist_t *local_delayprops = NULL;
	nvlist_t *recv_delayprops = NULL;
	nvlist_t *inherited_delayprops = NULL;
	nvlist_t origprops = NULL; / existing properties */
	nvlist_t origrecvd = NULL; / existing received properties */
	boolean_t first_recvd_props = B_FALSE;
	boolean_t tofs_was_redacted;
	zfs_file_t *input_fp;

	*read_bytes = 0;
	*errflags = 0;
	*errors = fnvlist_alloc();
	off = 0;

	if ((input_fp = zfs_file_get(input_fd)) == NULL)
	return (SET_ERROR(EBADF));

	noff = off = zfs_file_off(input_fp);
	error = dmu_recv_begin(tofs, tosnap, begin_record, force, heal,
	resumable, localprops, hidden_args, origin, &drc, input_fp,
	&off);
	if (error != 0)
	goto out;
	tofs_was_redacted = dsl_get_redacted(drc.drc_ds);

	/*
	* Set properties before we receive the stream so that they are applied
	* to the new data. Note that we must call dmu_recv_stream() if
	* dmu_recv_begin() succeeds.
	*/
	if (recvprops != NULL && !drc.drc_newfs) {
	if (spa_version(dsl_dataset_get_spa(drc.drc_ds)) >=
	SPA_VERSION_RECVD_PROPS &&
	!dsl_prop_get_hasrecvd(tofs))
	first_recvd_props = B_TRUE;

	/*
	* If new received properties are supplied, they are to
	* completely replace the existing received properties,
	* so stash away the existing ones.
	*/
	if (dsl_prop_get_received(tofs, &origrecvd) == 0) {
	nvlist_t *errlist = NULL;
	/*
	* Don't bother writing a property if its value won't
	* change (and avoid the unnecessary security checks).
	*
	* The first receive after SPA_VERSION_RECVD_PROPS is a
	* special case where we blow away all local properties
	* regardless.
	*/
	if (!first_recvd_props)
	props_reduce(recvprops, origrecvd);
	if (zfs_check_clearable(tofs, origrecvd, &errlist) != 0)
	(void) nvlist_merge(*errors, errlist, 0);
	nvlist_free(errlist);

	if (clear_received_props(tofs, origrecvd,
	first_recvd_props ? NULL : recvprops) != 0)
	*errflags \|= ZPROP_ERR_NOCLEAR;
	} else {
	*errflags \|= ZPROP_ERR_NOCLEAR;
	}
	}

	/*
	* Stash away existing properties so we can restore them on error unless
	* we're doing the first receive after SPA_VERSION_RECVD_PROPS, in which
	* case "origrecvd" will take care of that.
	*/
	if (localprops != NULL && !drc.drc_newfs && !first_recvd_props) {
	objset_t *os;
	if (dmu_objset_hold(tofs, FTAG, &os) == 0) {
	if (dsl_prop_get_all(os, &origprops) != 0) {
	*errflags \|= ZPROP_ERR_NOCLEAR;
	}
	dmu_objset_rele(os, FTAG);
	} else {
	*errflags \|= ZPROP_ERR_NOCLEAR;
	}
	}

	if (recvprops != NULL) {
	props_error = dsl_prop_set_hasrecvd(tofs);

	if (props_error == 0) {
	recv_delayprops = extract_delay_props(recvprops);
	(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
	recvprops, *errors);
	}
	}

	if (localprops != NULL) {
	nvlist_t *oprops = fnvlist_alloc();
	nvlist_t *xprops = fnvlist_alloc();
	nvpair_t *nvp = NULL;

	while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) {
	if (nvpair_type(nvp) == DATA_TYPE_BOOLEAN) {
	/* -x property */
	const char *name = nvpair_name(nvp);
	zfs_prop_t prop = zfs_name_to_prop(name);
	if (prop != ZPROP_USERPROP) {
	if (!zfs_prop_inheritable(prop))
	continue;
	} else if (!zfs_prop_user(name))
	continue;
	fnvlist_add_boolean(xprops, name);
	} else {
	/* -o property=value */
	fnvlist_add_nvpair(oprops, nvp);
	}
	}

	local_delayprops = extract_delay_props(oprops);
	(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL,
	oprops, *errors);
	inherited_delayprops = extract_delay_props(xprops);
	(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED,
	xprops, *errors);

	nvlist_free(oprops);
	nvlist_free(xprops);
	}

	error = dmu_recv_stream(&drc, &off);

	if (error == 0) {
	zfsvfs_t *zfsvfs = NULL;
	zvol_state_handle_t *zv = NULL;

	if (getzfsvfs(tofs, &zfsvfs) == 0) {
	/* online recv */
	dsl_dataset_t *ds;
	int end_err;
	boolean_t stream_is_redacted = DMU_GET_FEATUREFLAGS(
	begin_record->drr_u.drr_begin.
	drr_versioninfo) & DMU_BACKUP_FEATURE_REDACTED;

	ds = dmu_objset_ds(zfsvfs->z_os);
	error = zfs_suspend_fs(zfsvfs);
	/*
	* If the suspend fails, then the recv_end will
	* likely also fail, and clean up after itself.
	*/
	end_err = dmu_recv_end(&drc, zfsvfs);
	/*
	* If the dataset was not redacted, but we received a
	* redacted stream onto it, we need to unmount the
	* dataset. Otherwise, resume the filesystem.
	*/
	if (error == 0 && !drc.drc_newfs &&
	stream_is_redacted && !tofs_was_redacted) {
	error = zfs_end_fs(zfsvfs, ds);
	} else if (error == 0) {
	error = zfs_resume_fs(zfsvfs, ds);
	}
	error = error ? error : end_err;
	zfs_vfs_rele(zfsvfs);
	} else if ((zv = zvol_suspend(tofs)) != NULL) {
	error = dmu_recv_end(&drc, zvol_tag(zv));
	zvol_resume(zv);
	} else {
	error = dmu_recv_end(&drc, NULL);
	}

	/* Set delayed properties now, after we're done receiving. */
	if (recv_delayprops != NULL && error == 0) {
	(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_RECEIVED,
	recv_delayprops, *errors);
	}
	if (local_delayprops != NULL && error == 0) {
	(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL,
	local_delayprops, *errors);
	}
	if (inherited_delayprops != NULL && error == 0) {
	(void) zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED,
	inherited_delayprops, *errors);
	}
	}

	/*
	* Merge delayed props back in with initial props, in case
	* we're DEBUG and zfs_ioc_recv_inject_err is set (which means
	* we have to make sure clear_received_props() includes
	* the delayed properties).
	*
	* Since zfs_ioc_recv_inject_err is only in DEBUG kernels,
	* using ASSERT() will be just like a VERIFY.
	*/
	if (recv_delayprops != NULL) {
	ASSERT(nvlist_merge(recvprops, recv_delayprops, 0) == 0);
	nvlist_free(recv_delayprops);
	}
	if (local_delayprops != NULL) {
	ASSERT(nvlist_merge(localprops, local_delayprops, 0) == 0);
	nvlist_free(local_delayprops);
	}
	if (inherited_delayprops != NULL) {
	ASSERT(nvlist_merge(localprops, inherited_delayprops, 0) == 0);
	nvlist_free(inherited_delayprops);
	}
	*read_bytes = off - noff;

	#ifdef ZFS_DEBUG
	if (zfs_ioc_recv_inject_err) {
	zfs_ioc_recv_inject_err = B_FALSE;
	error = 1;
	}
	#endif

	/*
	* On error, restore the original props.
	*/
	if (error != 0 && recvprops != NULL && !drc.drc_newfs) {
	if (clear_received_props(tofs, recvprops, NULL) != 0) {
	/*
	* We failed to clear the received properties.
	* Since we may have left a $recvd value on the
	* system, we can't clear the $hasrecvd flag.
	*/
	*errflags \|= ZPROP_ERR_NORESTORE;
	} else if (first_recvd_props) {
	dsl_prop_unset_hasrecvd(tofs);
	}

	if (origrecvd == NULL && !drc.drc_newfs) {
	/* We failed to stash the original properties. */
	*errflags \|= ZPROP_ERR_NORESTORE;
	}

	/*
	* dsl_props_set() will not convert RECEIVED to LOCAL on or
	* after SPA_VERSION_RECVD_PROPS, so we need to specify LOCAL
	* explicitly if we're restoring local properties cleared in the
	* first new-style receive.
	*/
	if (origrecvd != NULL &&
	zfs_set_prop_nvlist(tofs, (first_recvd_props ?
	ZPROP_SRC_LOCAL : ZPROP_SRC_RECEIVED),
	origrecvd, NULL) != 0) {
	/*
	* We stashed the original properties but failed to
	* restore them.
	*/
	*errflags \|= ZPROP_ERR_NORESTORE;
	}
	}
	if (error != 0 && localprops != NULL && !drc.drc_newfs &&
	!first_recvd_props) {
	nvlist_t *setprops;
	nvlist_t *inheritprops;
	nvpair_t *nvp;

	if (origprops == NULL) {
	/* We failed to stash the original properties. */
	*errflags \|= ZPROP_ERR_NORESTORE;
	goto out;
	}

	/* Restore original props */
	setprops = fnvlist_alloc();
	inheritprops = fnvlist_alloc();
	nvp = NULL;
	while ((nvp = nvlist_next_nvpair(localprops, nvp)) != NULL) {
	const char *name = nvpair_name(nvp);
	const char *source;
	nvlist_t *attrs;

	if (!nvlist_exists(origprops, name)) {
	/*
	* Property was not present or was explicitly
	* inherited before the receive, restore this.
	*/
	fnvlist_add_boolean(inheritprops, name);
	continue;
	}
	attrs = fnvlist_lookup_nvlist(origprops, name);
	source = fnvlist_lookup_string(attrs, ZPROP_SOURCE);

	/* Skip received properties */
	if (strcmp(source, ZPROP_SOURCE_VAL_RECVD) == 0)
	continue;

	if (strcmp(source, tofs) == 0) {
	/* Property was locally set */
	fnvlist_add_nvlist(setprops, name, attrs);
	} else {
	/* Property was implicitly inherited */
	fnvlist_add_boolean(inheritprops, name);
	}
	}

	if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_LOCAL, setprops,
	NULL) != 0)
	*errflags \|= ZPROP_ERR_NORESTORE;
	if (zfs_set_prop_nvlist(tofs, ZPROP_SRC_INHERITED, inheritprops,
	NULL) != 0)
	*errflags \|= ZPROP_ERR_NORESTORE;

	nvlist_free(setprops);
	nvlist_free(inheritprops);
	}
	out:
	zfs_file_put(input_fp);
	nvlist_free(origrecvd);
	nvlist_free(origprops);

	if (error == 0)
	error = props_error;

	return (error);
	}

	/*
	* inputs:
	* zc_name name of containing filesystem (unused)
	* zc_nvlist_src{_size} nvlist of properties to apply
	* zc_nvlist_conf{_size} nvlist of properties to exclude
	* (DATA_TYPE_BOOLEAN) and override (everything else)
	* zc_value name of snapshot to create
	* zc_string name of clone origin (if DRR_FLAG_CLONE)
	* zc_cookie file descriptor to recv from
	* zc_begin_record the BEGIN record of the stream (not byteswapped)
	* zc_guid force flag
	*
	* outputs:
	* zc_cookie number of bytes read
	* zc_obj zprop_errflags_t
	* zc_nvlist_dst{_size} error for each unapplied received property
	*/
	static int
	zfs_ioc_recv(zfs_cmd_t *zc)
	{
	dmu_replay_record_t begin_record;
	nvlist_t *errors = NULL;
	nvlist_t *recvdprops = NULL;
	nvlist_t *localprops = NULL;
	const char *origin = NULL;
	char *tosnap;
	char tofs[ZFS_MAX_DATASET_NAME_LEN];
	int error = 0;

	if (dataset_namecheck(zc->zc_value, NULL, NULL) != 0 \|\|
	strchr(zc->zc_value, '@') == NULL \|\|
	strchr(zc->zc_value, '%') != NULL) {
	return (SET_ERROR(EINVAL));
	}

	(void) strlcpy(tofs, zc->zc_value, sizeof (tofs));
	tosnap = strchr(tofs, '@');
	*tosnap++ = '\0';

	if (zc->zc_nvlist_src != 0 &&
	(error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
	zc->zc_iflags, &recvdprops)) != 0) {
	goto out;
	}

	if (zc->zc_nvlist_conf != 0 &&
	(error = get_nvlist(zc->zc_nvlist_conf, zc->zc_nvlist_conf_size,
	zc->zc_iflags, &localprops)) != 0) {
	goto out;
	}

	if (zc->zc_string[0])
	origin = zc->zc_string;

	begin_record.drr_type = DRR_BEGIN;
	begin_record.drr_payloadlen = 0;
	begin_record.drr_u.drr_begin = zc->zc_begin_record;

	error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvdprops, localprops,
	NULL, zc->zc_guid, B_FALSE, B_FALSE, zc->zc_cookie, &begin_record,
	&zc->zc_cookie, &zc->zc_obj, &errors);

	/*
	* Now that all props, initial and delayed, are set, report the prop
	* errors to the caller.
	*/
	if (zc->zc_nvlist_dst_size != 0 && errors != NULL &&
	(nvlist_smush(errors, zc->zc_nvlist_dst_size) != 0 \|\|
	put_nvlist(zc, errors) != 0)) {
	/*
	* Caller made zc->zc_nvlist_dst less than the minimum expected
	* size or supplied an invalid address.
	*/
	error = SET_ERROR(EINVAL);
	}

	out:
	nvlist_free(errors);
	nvlist_free(recvdprops);
	nvlist_free(localprops);

	return (error);
	}

	/*
	* innvl: {
	* "snapname" -> full name of the snapshot to create
	* (optional) "props" -> received properties to set (nvlist)
	* (optional) "localprops" -> override and exclude properties (nvlist)
	* (optional) "origin" -> name of clone origin (DRR_FLAG_CLONE)
	* "begin_record" -> non-byteswapped dmu_replay_record_t
	* "input_fd" -> file descriptor to read stream from (int32)
	* (optional) "force" -> force flag (value ignored)
	* (optional) "heal" -> use send stream to heal data corruption
	* (optional) "resumable" -> resumable flag (value ignored)
	* (optional) "cleanup_fd" -> unused
	* (optional) "action_handle" -> unused
	* (optional) "hidden_args" -> { "wkeydata" -> value }
	* }
	*
	* outnvl: {
	* "read_bytes" -> number of bytes read
	* "error_flags" -> zprop_errflags_t
	* "errors" -> error for each unapplied received property (nvlist)
	* }
	*/
	static const zfs_ioc_key_t zfs_keys_recv_new[] = {
	{"snapname", DATA_TYPE_STRING, 0},
	{"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
	{"localprops", DATA_TYPE_NVLIST, ZK_OPTIONAL},
	{"origin", DATA_TYPE_STRING, ZK_OPTIONAL},
	{"begin_record", DATA_TYPE_BYTE_ARRAY, 0},
	{"input_fd", DATA_TYPE_INT32, 0},
	{"force", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"heal", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"resumable", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"cleanup_fd", DATA_TYPE_INT32, ZK_OPTIONAL},
	{"action_handle", DATA_TYPE_UINT64, ZK_OPTIONAL},
	{"hidden_args", DATA_TYPE_NVLIST, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_recv_new(const char fsname, nvlist_t innvl, nvlist_t *outnvl)
	{
	dmu_replay_record_t *begin_record;
	uint_t begin_record_size;
	nvlist_t *errors = NULL;
	nvlist_t *recvprops = NULL;
	nvlist_t *localprops = NULL;
	nvlist_t *hidden_args = NULL;
	const char *snapname;
	const char *origin = NULL;
	char *tosnap;
	char tofs[ZFS_MAX_DATASET_NAME_LEN];
	boolean_t force;
	boolean_t heal;
	boolean_t resumable;
	uint64_t read_bytes = 0;
	uint64_t errflags = 0;
	int input_fd = -1;
	int error;

	snapname = fnvlist_lookup_string(innvl, "snapname");

	if (dataset_namecheck(snapname, NULL, NULL) != 0 \|\|
	strchr(snapname, '@') == NULL \|\|
	strchr(snapname, '%') != NULL) {
	return (SET_ERROR(EINVAL));
	}

	(void) strlcpy(tofs, snapname, sizeof (tofs));
	tosnap = strchr(tofs, '@');
	*tosnap++ = '\0';

	error = nvlist_lookup_string(innvl, "origin", &origin);
	if (error && error != ENOENT)
	return (error);

	error = nvlist_lookup_byte_array(innvl, "begin_record",
	(uchar_t **)&begin_record, &begin_record_size);
	if (error != 0 \|\| begin_record_size != sizeof (*begin_record))
	return (SET_ERROR(EINVAL));

	input_fd = fnvlist_lookup_int32(innvl, "input_fd");

	force = nvlist_exists(innvl, "force");
	heal = nvlist_exists(innvl, "heal");
	resumable = nvlist_exists(innvl, "resumable");

	/* we still use "props" here for backwards compatibility */
	error = nvlist_lookup_nvlist(innvl, "props", &recvprops);
	if (error && error != ENOENT)
	goto out;

	error = nvlist_lookup_nvlist(innvl, "localprops", &localprops);
	if (error && error != ENOENT)
	goto out;

	error = nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);
	if (error && error != ENOENT)
	goto out;

	error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvprops, localprops,
	hidden_args, force, heal, resumable, input_fd, begin_record,
	&read_bytes, &errflags, &errors);

	fnvlist_add_uint64(outnvl, "read_bytes", read_bytes);
	fnvlist_add_uint64(outnvl, "error_flags", errflags);
	fnvlist_add_nvlist(outnvl, "errors", errors);

	out:
	nvlist_free(errors);
	nvlist_free(recvprops);
	nvlist_free(localprops);
	nvlist_free(hidden_args);

	return (error);
	}

	+/*
	+ * When stack space is limited, we write replication stream data to the target
	+ * on a separate taskq thread, to make sure there's enough stack space.
	+ */
	+#ifndef HAVE_LARGE_STACKS
	+#define USE_SEND_TASKQ 1
	+#endif
	+
	typedef struct dump_bytes_io {
	zfs_file_t *dbi_fp;
	caddr_t dbi_buf;
	int dbi_len;
	int dbi_err;
	} dump_bytes_io_t;

	static void
	dump_bytes_cb(void *arg)
	{
	dump_bytes_io_t dbi = (dump_bytes_io_t )arg;
	zfs_file_t *fp;
	caddr_t buf;

	fp = dbi->dbi_fp;
	buf = dbi->dbi_buf;

	dbi->dbi_err = zfs_file_write(fp, buf, dbi->dbi_len, NULL);
	}

	+typedef struct dump_bytes_arg {
	+ zfs_file_t *dba_fp;
	+#ifdef USE_SEND_TASKQ
	+ taskq_t *dba_tq;
	+ taskq_ent_t dba_tqent;
	+#endif
	+} dump_bytes_arg_t;
	+
	static int
	dump_bytes(objset_t os, void buf, int len, void *arg)
	{
	+ dump_bytes_arg_t dba = (dump_bytes_arg_t )arg;
	dump_bytes_io_t dbi;

	- dbi.dbi_fp = arg;
	+ dbi.dbi_fp = dba->dba_fp;
	dbi.dbi_buf = buf;
	dbi.dbi_len = len;

	-#if defined(HAVE_LARGE_STACKS)
	- dump_bytes_cb(&dbi);
	+#ifdef USE_SEND_TASKQ
	+ taskq_dispatch_ent(dba->dba_tq, dump_bytes_cb, &dbi, TQ_SLEEP,
	+ &dba->dba_tqent);
	+ taskq_wait(dba->dba_tq);
	#else
	- /*
	- * The vn_rdwr() call is performed in a taskq to ensure that there is
	- * always enough stack space to write safely to the target filesystem.
	- * The ZIO_TYPE_FREE threads are used because there can be a lot of
	- * them and they are used in vdev_file.c for a similar purpose.
	- */
	- spa_taskq_dispatch_sync(dmu_objset_spa(os), ZIO_TYPE_FREE,
	- ZIO_TASKQ_ISSUE, dump_bytes_cb, &dbi, TQ_SLEEP);
	-#endif /* HAVE_LARGE_STACKS */
	+ dump_bytes_cb(&dbi);
	+#endif

	return (dbi.dbi_err);
	}

	+static int
	+dump_bytes_init(dump_bytes_arg_t dba, int fd, dmu_send_outparams_t out)
	+{
	+ zfs_file_t *fp = zfs_file_get(fd);
	+ if (fp == NULL)
	+ return (SET_ERROR(EBADF));
	+
	+ dba->dba_fp = fp;
	+#ifdef USE_SEND_TASKQ
	+ dba->dba_tq = taskq_create("z_send", 1, defclsyspri, 0, 0, 0);
	+ taskq_init_ent(&dba->dba_tqent);
	+#endif
	+
	+ memset(out, 0, sizeof (dmu_send_outparams_t));
	+ out->dso_outfunc = dump_bytes;
	+ out->dso_arg = dba;
	+ out->dso_dryrun = B_FALSE;
	+
	+ return (0);
	+}
	+
	+static void
	+dump_bytes_fini(dump_bytes_arg_t *dba)
	+{
	+ zfs_file_put(dba->dba_fp);
	+#ifdef USE_SEND_TASKQ
	+ taskq_destroy(dba->dba_tq);
	+#endif
	+}
	+
	/*
	* inputs:
	* zc_name name of snapshot to send
	* zc_cookie file descriptor to send stream to
	* zc_obj fromorigin flag (mutually exclusive with zc_fromobj)
	* zc_sendobj objsetid of snapshot to send
	* zc_fromobj objsetid of incremental fromsnap (may be zero)
	* zc_guid if set, estimate size of stream only. zc_cookie is ignored.
	* output size in zc_objset_type.
	* zc_flags lzc_send_flags
	*
	* outputs:
	* zc_objset_type estimated size, if zc_guid is set
	*
	* NOTE: This is no longer the preferred interface, any new functionality
	* should be added to zfs_ioc_send_new() instead.
	*/
	static int
	zfs_ioc_send(zfs_cmd_t *zc)
	{
	int error;
	offset_t off;
	boolean_t estimate = (zc->zc_guid != 0);
	boolean_t embedok = (zc->zc_flags & 0x1);
	boolean_t large_block_ok = (zc->zc_flags & 0x2);
	boolean_t compressok = (zc->zc_flags & 0x4);
	boolean_t rawok = (zc->zc_flags & 0x8);
	boolean_t savedok = (zc->zc_flags & 0x10);

	if (zc->zc_obj != 0) {
	dsl_pool_t *dp;
	dsl_dataset_t *tosnap;

	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
	if (error != 0)
	return (error);

	error = dsl_dataset_hold_obj(dp, zc->zc_sendobj, FTAG, &tosnap);
	if (error != 0) {
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	if (dsl_dir_is_clone(tosnap->ds_dir))
	zc->zc_fromobj =
	dsl_dir_phys(tosnap->ds_dir)->dd_origin_obj;
	dsl_dataset_rele(tosnap, FTAG);
	dsl_pool_rele(dp, FTAG);
	}

	if (estimate) {
	dsl_pool_t *dp;
	dsl_dataset_t *tosnap;
	dsl_dataset_t *fromsnap = NULL;

	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
	if (error != 0)
	return (error);

	error = dsl_dataset_hold_obj(dp, zc->zc_sendobj,
	FTAG, &tosnap);
	if (error != 0) {
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	if (zc->zc_fromobj != 0) {
	error = dsl_dataset_hold_obj(dp, zc->zc_fromobj,
	FTAG, &fromsnap);
	if (error != 0) {
	dsl_dataset_rele(tosnap, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (error);
	}
	}

	error = dmu_send_estimate_fast(tosnap, fromsnap, NULL,
	compressok \|\| rawok, savedok, &zc->zc_objset_type);

	if (fromsnap != NULL)
	dsl_dataset_rele(fromsnap, FTAG);
	dsl_dataset_rele(tosnap, FTAG);
	dsl_pool_rele(dp, FTAG);
	} else {
	- zfs_file_t *fp;
	- dmu_send_outparams_t out = {0};
	-
	- if ((fp = zfs_file_get(zc->zc_cookie)) == NULL)
	- return (SET_ERROR(EBADF));
	+ dump_bytes_arg_t dba;
	+ dmu_send_outparams_t out;
	+ error = dump_bytes_init(&dba, zc->zc_cookie, &out);
	+ if (error)
	+ return (error);

	- off = zfs_file_off(fp);
	- out.dso_outfunc = dump_bytes;
	- out.dso_arg = fp;
	- out.dso_dryrun = B_FALSE;
	+ off = zfs_file_off(dba.dba_fp);
	error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
	zc->zc_fromobj, embedok, large_block_ok, compressok,
	rawok, savedok, zc->zc_cookie, &off, &out);

	- zfs_file_put(fp);
	+ dump_bytes_fini(&dba);
	}
	return (error);
	}

	/*
	* inputs:
	* zc_name name of snapshot on which to report progress
	* zc_cookie file descriptor of send stream
	*
	* outputs:
	* zc_cookie number of bytes written in send stream thus far
	* zc_objset_type logical size of data traversed by send thus far
	*/
	static int
	zfs_ioc_send_progress(zfs_cmd_t *zc)
	{
	dsl_pool_t *dp;
	dsl_dataset_t *ds;
	dmu_sendstatus_t *dsp = NULL;
	int error;

	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
	if (error != 0)
	return (error);

	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
	if (error != 0) {
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	mutex_enter(&ds->ds_sendstream_lock);

	/*
	* Iterate over all the send streams currently active on this dataset.
	* If there's one which matches the specified file descriptor _and_ the
	* stream was started by the current process, return the progress of
	* that stream.
	*/

	for (dsp = list_head(&ds->ds_sendstreams); dsp != NULL;
	dsp = list_next(&ds->ds_sendstreams, dsp)) {
	if (dsp->dss_outfd == zc->zc_cookie &&
	zfs_proc_is_caller(dsp->dss_proc))
	break;
	}

	if (dsp != NULL) {
	zc->zc_cookie = atomic_cas_64((volatile uint64_t *)dsp->dss_off,
	0, 0);
	/* This is the closest thing we have to atomic_read_64. */
	zc->zc_objset_type = atomic_cas_64(&dsp->dss_blocks, 0, 0);
	} else {
	error = SET_ERROR(ENOENT);
	}

	mutex_exit(&ds->ds_sendstream_lock);
	dsl_dataset_rele(ds, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	static int
	zfs_ioc_inject_fault(zfs_cmd_t *zc)
	{
	int id, error;

	error = zio_inject_fault(zc->zc_name, (int)zc->zc_guid, &id,
	&zc->zc_inject_record);

	if (error == 0)
	zc->zc_guid = (uint64_t)id;

	return (error);
	}

	static int
	zfs_ioc_clear_fault(zfs_cmd_t *zc)
	{
	return (zio_clear_fault((int)zc->zc_guid));
	}

	static int
	zfs_ioc_inject_list_next(zfs_cmd_t *zc)
	{
	int id = (int)zc->zc_guid;
	int error;

	error = zio_inject_list_next(&id, zc->zc_name, sizeof (zc->zc_name),
	&zc->zc_inject_record);

	zc->zc_guid = id;

	return (error);
	}

	static int
	zfs_ioc_error_log(zfs_cmd_t *zc)
	{
	spa_t *spa;
	int error;

	if ((error = spa_open(zc->zc_name, &spa, FTAG)) != 0)
	return (error);

	error = spa_get_errlog(spa, (void *)(uintptr_t)zc->zc_nvlist_dst,
	&zc->zc_nvlist_dst_size);

	spa_close(spa, FTAG);

	return (error);
	}

	static int
	zfs_ioc_clear(zfs_cmd_t *zc)
	{
	spa_t *spa;
	vdev_t *vd;
	int error;

	/*
	* On zpool clear we also fix up missing slogs
	*/
	mutex_enter(&spa_namespace_lock);
	spa = spa_lookup(zc->zc_name);
	if (spa == NULL) {
	mutex_exit(&spa_namespace_lock);
	return (SET_ERROR(EIO));
	}
	if (spa_get_log_state(spa) == SPA_LOG_MISSING) {
	/* we need to let spa_open/spa_load clear the chains */
	spa_set_log_state(spa, SPA_LOG_CLEAR);
	}
	spa->spa_last_open_failed = 0;
	mutex_exit(&spa_namespace_lock);

	if (zc->zc_cookie & ZPOOL_NO_REWIND) {
	error = spa_open(zc->zc_name, &spa, FTAG);
	} else {
	nvlist_t *policy;
	nvlist_t *config = NULL;

	if (zc->zc_nvlist_src == 0)
	return (SET_ERROR(EINVAL));

	if ((error = get_nvlist(zc->zc_nvlist_src,
	zc->zc_nvlist_src_size, zc->zc_iflags, &policy)) == 0) {
	error = spa_open_rewind(zc->zc_name, &spa, FTAG,
	policy, &config);
	if (config != NULL) {
	int err;

	if ((err = put_nvlist(zc, config)) != 0)
	error = err;
	nvlist_free(config);
	}
	nvlist_free(policy);
	}
	}

	if (error != 0)
	return (error);

	/*
	* If multihost is enabled, resuming I/O is unsafe as another
	* host may have imported the pool. Check for remote activity.
	*/
	if (spa_multihost(spa) && spa_suspended(spa) &&
	spa_mmp_remote_host_activity(spa)) {
	spa_close(spa, FTAG);
	return (SET_ERROR(EREMOTEIO));
	}

	spa_vdev_state_enter(spa, SCL_NONE);

	if (zc->zc_guid == 0) {
	vd = NULL;
	} else {
	vd = spa_lookup_by_guid(spa, zc->zc_guid, B_TRUE);
	if (vd == NULL) {
	error = SET_ERROR(ENODEV);
	(void) spa_vdev_state_exit(spa, NULL, error);
	spa_close(spa, FTAG);
	return (error);
	}
	}

	vdev_clear(spa, vd);

	(void) spa_vdev_state_exit(spa, spa_suspended(spa) ?
	NULL : spa->spa_root_vdev, 0);

	/*
	* Resume any suspended I/Os.
	*/
	if (zio_resume(spa) != 0)
	error = SET_ERROR(EIO);

	spa_close(spa, FTAG);

	return (error);
	}

	/*
	* Reopen all the vdevs associated with the pool.
	*
	* innvl: {
	* "scrub_restart" -> when true and scrub is running, allow to restart
	* scrub as the side effect of the reopen (boolean).
	* }
	*
	* outnvl is unused
	*/
	static const zfs_ioc_key_t zfs_keys_pool_reopen[] = {
	{"scrub_restart", DATA_TYPE_BOOLEAN_VALUE, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_pool_reopen(const char pool, nvlist_t innvl, nvlist_t *outnvl)
	{
	(void) outnvl;
	spa_t *spa;
	int error;
	boolean_t rc, scrub_restart = B_TRUE;

	if (innvl) {
	error = nvlist_lookup_boolean_value(innvl,
	"scrub_restart", &rc);
	if (error == 0)
	scrub_restart = rc;
	}

	error = spa_open(pool, &spa, FTAG);
	if (error != 0)
	return (error);

	spa_vdev_state_enter(spa, SCL_NONE);

	/*
	* If the scrub_restart flag is B_FALSE and a scrub is already
	* in progress then set spa_scrub_reopen flag to B_TRUE so that
	* we don't restart the scrub as a side effect of the reopen.
	* Otherwise, let vdev_open() decided if a resilver is required.
	*/

	spa->spa_scrub_reopen = (!scrub_restart &&
	dsl_scan_scrubbing(spa->spa_dsl_pool));
	vdev_reopen(spa->spa_root_vdev);
	spa->spa_scrub_reopen = B_FALSE;

	(void) spa_vdev_state_exit(spa, NULL, 0);
	spa_close(spa, FTAG);
	return (0);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	*
	* outputs:
	* zc_string name of conflicting snapshot, if there is one
	*/
	static int
	zfs_ioc_promote(zfs_cmd_t *zc)
	{
	dsl_pool_t *dp;
	dsl_dataset_t ds, ods;
	char origin[ZFS_MAX_DATASET_NAME_LEN];
	char *cp;
	int error;

	zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
	if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0 \|\|
	strchr(zc->zc_name, '%'))
	return (SET_ERROR(EINVAL));

	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
	if (error != 0)
	return (error);

	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &ds);
	if (error != 0) {
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	if (!dsl_dir_is_clone(ds->ds_dir)) {
	dsl_dataset_rele(ds, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (SET_ERROR(EINVAL));
	}

	error = dsl_dataset_hold_obj(dp,
	dsl_dir_phys(ds->ds_dir)->dd_origin_obj, FTAG, &ods);
	if (error != 0) {
	dsl_dataset_rele(ds, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	dsl_dataset_name(ods, origin);
	dsl_dataset_rele(ods, FTAG);
	dsl_dataset_rele(ds, FTAG);
	dsl_pool_rele(dp, FTAG);

	/*
	* We don't need to unmount all the origin fs's snapshots, but
	* it's easier.
	*/
	cp = strchr(origin, '@');
	if (cp)
	*cp = '\0';
	(void) dmu_objset_find(origin,
	zfs_unmount_snap_cb, NULL, DS_FIND_SNAPSHOTS);
	return (dsl_dataset_promote(zc->zc_name, zc->zc_string));
	}

	/*
	* Retrieve a single {user\|group\|project}{used\|quota}@... property.
	*
	* inputs:
	* zc_name name of filesystem
	* zc_objset_type zfs_userquota_prop_t
	* zc_value domain name (eg. "S-1-234-567-89")
	* zc_guid RID/UID/GID
	*
	* outputs:
	* zc_cookie property value
	*/
	static int
	zfs_ioc_userspace_one(zfs_cmd_t *zc)
	{
	zfsvfs_t *zfsvfs;
	int error;

	if (zc->zc_objset_type >= ZFS_NUM_USERQUOTA_PROPS)
	return (SET_ERROR(EINVAL));

	error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
	if (error != 0)
	return (error);

	error = zfs_userspace_one(zfsvfs,
	zc->zc_objset_type, zc->zc_value, zc->zc_guid, &zc->zc_cookie);
	zfsvfs_rele(zfsvfs, FTAG);

	return (error);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_cookie zap cursor
	* zc_objset_type zfs_userquota_prop_t
	* zc_nvlist_dst[_size] buffer to fill (not really an nvlist)
	*
	* outputs:
	* zc_nvlist_dst[_size] data buffer (array of zfs_useracct_t)
	* zc_cookie zap cursor
	*/
	static int
	zfs_ioc_userspace_many(zfs_cmd_t *zc)
	{
	zfsvfs_t *zfsvfs;
	int bufsize = zc->zc_nvlist_dst_size;

	if (bufsize <= 0)
	return (SET_ERROR(ENOMEM));

	int error = zfsvfs_hold(zc->zc_name, FTAG, &zfsvfs, B_FALSE);
	if (error != 0)
	return (error);

	void *buf = vmem_alloc(bufsize, KM_SLEEP);

	error = zfs_userspace_many(zfsvfs, zc->zc_objset_type, &zc->zc_cookie,
	buf, &zc->zc_nvlist_dst_size);

	if (error == 0) {
	error = xcopyout(buf,
	(void *)(uintptr_t)zc->zc_nvlist_dst,
	zc->zc_nvlist_dst_size);
	}
	vmem_free(buf, bufsize);
	zfsvfs_rele(zfsvfs, FTAG);

	return (error);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	*
	* outputs:
	* none
	*/
	static int
	zfs_ioc_userspace_upgrade(zfs_cmd_t *zc)
	{
	int error = 0;
	zfsvfs_t *zfsvfs;

	if (getzfsvfs(zc->zc_name, &zfsvfs) == 0) {
	if (!dmu_objset_userused_enabled(zfsvfs->z_os)) {
	/*
	* If userused is not enabled, it may be because the
	* objset needs to be closed & reopened (to grow the
	* objset_phys_t). Suspend/resume the fs will do that.
	*/
	dsl_dataset_t ds, newds;

	ds = dmu_objset_ds(zfsvfs->z_os);
	error = zfs_suspend_fs(zfsvfs);
	if (error == 0) {
	dmu_objset_refresh_ownership(ds, &newds,
	B_TRUE, zfsvfs);
	error = zfs_resume_fs(zfsvfs, newds);
	}
	}
	if (error == 0) {
	mutex_enter(&zfsvfs->z_os->os_upgrade_lock);
	if (zfsvfs->z_os->os_upgrade_id == 0) {
	/* clear potential error code and retry */
	zfsvfs->z_os->os_upgrade_status = 0;
	mutex_exit(&zfsvfs->z_os->os_upgrade_lock);

	dsl_pool_config_enter(
	dmu_objset_pool(zfsvfs->z_os), FTAG);
	dmu_objset_userspace_upgrade(zfsvfs->z_os);
	dsl_pool_config_exit(
	dmu_objset_pool(zfsvfs->z_os), FTAG);
	} else {
	mutex_exit(&zfsvfs->z_os->os_upgrade_lock);
	}

	taskq_wait_id(zfsvfs->z_os->os_spa->spa_upgrade_taskq,
	zfsvfs->z_os->os_upgrade_id);
	error = zfsvfs->z_os->os_upgrade_status;
	}
	zfs_vfs_rele(zfsvfs);
	} else {
	objset_t *os;

	/* XXX kind of reading contents without owning */
	error = dmu_objset_hold_flags(zc->zc_name, B_TRUE, FTAG, &os);
	if (error != 0)
	return (error);

	mutex_enter(&os->os_upgrade_lock);
	if (os->os_upgrade_id == 0) {
	/* clear potential error code and retry */
	os->os_upgrade_status = 0;
	mutex_exit(&os->os_upgrade_lock);

	dmu_objset_userspace_upgrade(os);
	} else {
	mutex_exit(&os->os_upgrade_lock);
	}

	dsl_pool_rele(dmu_objset_pool(os), FTAG);

	taskq_wait_id(os->os_spa->spa_upgrade_taskq, os->os_upgrade_id);
	error = os->os_upgrade_status;

	dsl_dataset_rele_flags(dmu_objset_ds(os), DS_HOLD_FLAG_DECRYPT,
	FTAG);
	}
	return (error);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	*
	* outputs:
	* none
	*/
	static int
	zfs_ioc_id_quota_upgrade(zfs_cmd_t *zc)
	{
	objset_t *os;
	int error;

	error = dmu_objset_hold_flags(zc->zc_name, B_TRUE, FTAG, &os);
	if (error != 0)
	return (error);

	if (dmu_objset_userobjspace_upgradable(os) \|\|
	dmu_objset_projectquota_upgradable(os)) {
	mutex_enter(&os->os_upgrade_lock);
	if (os->os_upgrade_id == 0) {
	/* clear potential error code and retry */
	os->os_upgrade_status = 0;
	mutex_exit(&os->os_upgrade_lock);

	dmu_objset_id_quota_upgrade(os);
	} else {
	mutex_exit(&os->os_upgrade_lock);
	}

	dsl_pool_rele(dmu_objset_pool(os), FTAG);

	taskq_wait_id(os->os_spa->spa_upgrade_taskq, os->os_upgrade_id);
	error = os->os_upgrade_status;
	} else {
	dsl_pool_rele(dmu_objset_pool(os), FTAG);
	}

	dsl_dataset_rele_flags(dmu_objset_ds(os), DS_HOLD_FLAG_DECRYPT, FTAG);

	return (error);
	}

	static int
	zfs_ioc_share(zfs_cmd_t *zc)
	{
	return (SET_ERROR(ENOSYS));
	}

	/*
	* inputs:
	* zc_name name of containing filesystem
	* zc_obj object # beyond which we want next in-use object #
	*
	* outputs:
	* zc_obj next in-use object #
	*/
	static int
	zfs_ioc_next_obj(zfs_cmd_t *zc)
	{
	objset_t *os = NULL;
	int error;

	error = dmu_objset_hold(zc->zc_name, FTAG, &os);
	if (error != 0)
	return (error);

	error = dmu_object_next(os, &zc->zc_obj, B_FALSE, 0);

	dmu_objset_rele(os, FTAG);
	return (error);
	}

	/*
	* inputs:
	* zc_name name of filesystem
	* zc_value prefix name for snapshot
	* zc_cleanup_fd cleanup-on-exit file descriptor for calling process
	*
	* outputs:
	* zc_value short name of new snapshot
	*/
	static int
	zfs_ioc_tmp_snapshot(zfs_cmd_t *zc)
	{
	char *snap_name;
	char *hold_name;
	minor_t minor;

	zfs_file_t *fp = zfs_onexit_fd_hold(zc->zc_cleanup_fd, &minor);
	if (fp == NULL)
	return (SET_ERROR(EBADF));

	snap_name = kmem_asprintf("%s-%016llx", zc->zc_value,
	(u_longlong_t)ddi_get_lbolt64());
	hold_name = kmem_asprintf("%%%s", zc->zc_value);

	int error = dsl_dataset_snapshot_tmp(zc->zc_name, snap_name, minor,
	hold_name);
	if (error == 0)
	(void) strlcpy(zc->zc_value, snap_name,
	sizeof (zc->zc_value));
	kmem_strfree(snap_name);
	kmem_strfree(hold_name);
	zfs_onexit_fd_rele(fp);
	return (error);
	}

	/*
	* inputs:
	* zc_name name of "to" snapshot
	* zc_value name of "from" snapshot
	* zc_cookie file descriptor to write diff data on
	*
	* outputs:
	* dmu_diff_record_t's to the file descriptor
	*/
	static int
	zfs_ioc_diff(zfs_cmd_t *zc)
	{
	zfs_file_t *fp;
	offset_t off;
	int error;

	if ((fp = zfs_file_get(zc->zc_cookie)) == NULL)
	return (SET_ERROR(EBADF));

	off = zfs_file_off(fp);
	error = dmu_diff(zc->zc_name, zc->zc_value, fp, &off);

	zfs_file_put(fp);

	return (error);
	}

	static int
	zfs_ioc_smb_acl(zfs_cmd_t *zc)
	{
	return (SET_ERROR(ENOTSUP));
	}

	/*
	* innvl: {
	* "holds" -> { snapname -> holdname (string), ... }
	* (optional) "cleanup_fd" -> fd (int32)
	* }
	*
	* outnvl: {
	* snapname -> error value (int32)
	* ...
	* }
	*/
	static const zfs_ioc_key_t zfs_keys_hold[] = {
	{"holds", DATA_TYPE_NVLIST, 0},
	{"cleanup_fd", DATA_TYPE_INT32, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_hold(const char pool, nvlist_t args, nvlist_t *errlist)
	{
	(void) pool;
	nvpair_t *pair;
	nvlist_t *holds;
	int cleanup_fd = -1;
	int error;
	minor_t minor = 0;
	zfs_file_t *fp = NULL;

	holds = fnvlist_lookup_nvlist(args, "holds");

	/* make sure the user didn't pass us any invalid (empty) tags */
	for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
	pair = nvlist_next_nvpair(holds, pair)) {
	const char *htag;

	error = nvpair_value_string(pair, &htag);
	if (error != 0)
	return (SET_ERROR(error));

	if (strlen(htag) == 0)
	return (SET_ERROR(EINVAL));
	}

	if (nvlist_lookup_int32(args, "cleanup_fd", &cleanup_fd) == 0) {
	fp = zfs_onexit_fd_hold(cleanup_fd, &minor);
	if (fp == NULL)
	return (SET_ERROR(EBADF));
	}

	error = dsl_dataset_user_hold(holds, minor, errlist);
	if (fp != NULL) {
	ASSERT3U(minor, !=, 0);
	zfs_onexit_fd_rele(fp);
	}
	return (SET_ERROR(error));
	}

	/*
	* innvl is not used.
	*
	* outnvl: {
	* holdname -> time added (uint64 seconds since epoch)
	* ...
	* }
	*/
	static const zfs_ioc_key_t zfs_keys_get_holds[] = {
	/* no nvl keys */
	};

	static int
	zfs_ioc_get_holds(const char snapname, nvlist_t args, nvlist_t *outnvl)
	{
	(void) args;
	return (dsl_dataset_get_holds(snapname, outnvl));
	}

	/*
	* innvl: {
	* snapname -> { holdname, ... }
	* ...
	* }
	*
	* outnvl: {
	* snapname -> error value (int32)
	* ...
	* }
	*/
	static const zfs_ioc_key_t zfs_keys_release[] = {
	{"<snapname>...", DATA_TYPE_NVLIST, ZK_WILDCARDLIST},
	};

	static int
	zfs_ioc_release(const char pool, nvlist_t holds, nvlist_t *errlist)
	{
	(void) pool;
	return (dsl_dataset_user_release(holds, errlist));
	}

	/*
	* inputs:
	* zc_guid flags (ZEVENT_NONBLOCK)
	* zc_cleanup_fd zevent file descriptor
	*
	* outputs:
	* zc_nvlist_dst next nvlist event
	* zc_cookie dropped events since last get
	*/
	static int
	zfs_ioc_events_next(zfs_cmd_t *zc)
	{
	zfs_zevent_t *ze;
	nvlist_t *event = NULL;
	minor_t minor;
	uint64_t dropped = 0;
	int error;

	zfs_file_t *fp = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
	if (fp == NULL)
	return (SET_ERROR(EBADF));

	do {
	error = zfs_zevent_next(ze, &event,
	&zc->zc_nvlist_dst_size, &dropped);
	if (event != NULL) {
	zc->zc_cookie = dropped;
	error = put_nvlist(zc, event);
	nvlist_free(event);
	}

	if (zc->zc_guid & ZEVENT_NONBLOCK)
	break;

	if ((error == 0) \|\| (error != ENOENT))
	break;

	error = zfs_zevent_wait(ze);
	if (error != 0)
	break;
	} while (1);

	zfs_zevent_fd_rele(fp);

	return (error);
	}

	/*
	* outputs:
	* zc_cookie cleared events count
	*/
	static int
	zfs_ioc_events_clear(zfs_cmd_t *zc)
	{
	uint_t count;

	zfs_zevent_drain_all(&count);
	zc->zc_cookie = count;

	return (0);
	}

	/*
	* inputs:
	* zc_guid eid \| ZEVENT_SEEK_START \| ZEVENT_SEEK_END
	* zc_cleanup zevent file descriptor
	*/
	static int
	zfs_ioc_events_seek(zfs_cmd_t *zc)
	{
	zfs_zevent_t *ze;
	minor_t minor;
	int error;

	zfs_file_t *fp = zfs_zevent_fd_hold(zc->zc_cleanup_fd, &minor, &ze);
	if (fp == NULL)
	return (SET_ERROR(EBADF));

	error = zfs_zevent_seek(ze, zc->zc_guid);
	zfs_zevent_fd_rele(fp);

	return (error);
	}

	/*
	* inputs:
	* zc_name name of later filesystem or snapshot
	* zc_value full name of old snapshot or bookmark
	*
	* outputs:
	* zc_cookie space in bytes
	* zc_objset_type compressed space in bytes
	* zc_perm_action uncompressed space in bytes
	*/
	static int
	zfs_ioc_space_written(zfs_cmd_t *zc)
	{
	int error;
	dsl_pool_t *dp;
	dsl_dataset_t *new;

	error = dsl_pool_hold(zc->zc_name, FTAG, &dp);
	if (error != 0)
	return (error);
	error = dsl_dataset_hold(dp, zc->zc_name, FTAG, &new);
	if (error != 0) {
	dsl_pool_rele(dp, FTAG);
	return (error);
	}
	if (strchr(zc->zc_value, '#') != NULL) {
	zfs_bookmark_phys_t bmp;
	error = dsl_bookmark_lookup(dp, zc->zc_value,
	new, &bmp);
	if (error == 0) {
	error = dsl_dataset_space_written_bookmark(&bmp, new,
	&zc->zc_cookie,
	&zc->zc_objset_type, &zc->zc_perm_action);
	}
	} else {
	dsl_dataset_t *old;
	error = dsl_dataset_hold(dp, zc->zc_value, FTAG, &old);

	if (error == 0) {
	error = dsl_dataset_space_written(old, new,
	&zc->zc_cookie,
	&zc->zc_objset_type, &zc->zc_perm_action);
	dsl_dataset_rele(old, FTAG);
	}
	}
	dsl_dataset_rele(new, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	/*
	* innvl: {
	* "firstsnap" -> snapshot name
	* }
	*
	* outnvl: {
	* "used" -> space in bytes
	* "compressed" -> compressed space in bytes
	* "uncompressed" -> uncompressed space in bytes
	* }
	*/
	static const zfs_ioc_key_t zfs_keys_space_snaps[] = {
	{"firstsnap", DATA_TYPE_STRING, 0},
	};

	static int
	zfs_ioc_space_snaps(const char lastsnap, nvlist_t innvl, nvlist_t *outnvl)
	{
	int error;
	dsl_pool_t *dp;
	dsl_dataset_t new, old;
	const char *firstsnap;
	uint64_t used, comp, uncomp;

	firstsnap = fnvlist_lookup_string(innvl, "firstsnap");

	error = dsl_pool_hold(lastsnap, FTAG, &dp);
	if (error != 0)
	return (error);

	error = dsl_dataset_hold(dp, lastsnap, FTAG, &new);
	if (error == 0 && !new->ds_is_snapshot) {
	dsl_dataset_rele(new, FTAG);
	error = SET_ERROR(EINVAL);
	}
	if (error != 0) {
	dsl_pool_rele(dp, FTAG);
	return (error);
	}
	error = dsl_dataset_hold(dp, firstsnap, FTAG, &old);
	if (error == 0 && !old->ds_is_snapshot) {
	dsl_dataset_rele(old, FTAG);
	error = SET_ERROR(EINVAL);
	}
	if (error != 0) {
	dsl_dataset_rele(new, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	error = dsl_dataset_space_wouldfree(old, new, &used, &comp, &uncomp);
	dsl_dataset_rele(old, FTAG);
	dsl_dataset_rele(new, FTAG);
	dsl_pool_rele(dp, FTAG);
	fnvlist_add_uint64(outnvl, "used", used);
	fnvlist_add_uint64(outnvl, "compressed", comp);
	fnvlist_add_uint64(outnvl, "uncompressed", uncomp);
	return (error);
	}

	/*
	* innvl: {
	* "fd" -> file descriptor to write stream to (int32)
	* (optional) "fromsnap" -> full snap name to send an incremental from
	* (optional) "largeblockok" -> (value ignored)
	* indicates that blocks > 128KB are permitted
	* (optional) "embedok" -> (value ignored)
	* presence indicates DRR_WRITE_EMBEDDED records are permitted
	* (optional) "compressok" -> (value ignored)
	* presence indicates compressed DRR_WRITE records are permitted
	* (optional) "rawok" -> (value ignored)
	* presence indicates raw encrypted records should be used.
	* (optional) "savedok" -> (value ignored)
	* presence indicates we should send a partially received snapshot
	* (optional) "resume_object" and "resume_offset" -> (uint64)
	* if present, resume send stream from specified object and offset.
	* (optional) "redactbook" -> (string)
	* if present, use this bookmark's redaction list to generate a redacted
	* send stream
	* }
	*
	* outnvl is unused
	*/
	static const zfs_ioc_key_t zfs_keys_send_new[] = {
	{"fd", DATA_TYPE_INT32, 0},
	{"fromsnap", DATA_TYPE_STRING, ZK_OPTIONAL},
	{"largeblockok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"embedok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"compressok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"rawok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"savedok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"resume_object", DATA_TYPE_UINT64, ZK_OPTIONAL},
	{"resume_offset", DATA_TYPE_UINT64, ZK_OPTIONAL},
	{"redactbook", DATA_TYPE_STRING, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_send_new(const char snapname, nvlist_t innvl, nvlist_t *outnvl)
	{
	(void) outnvl;
	int error;
	offset_t off;
	const char *fromname = NULL;
	int fd;
	- zfs_file_t *fp;
	boolean_t largeblockok;
	boolean_t embedok;
	boolean_t compressok;
	boolean_t rawok;
	boolean_t savedok;
	uint64_t resumeobj = 0;
	uint64_t resumeoff = 0;
	const char *redactbook = NULL;

	fd = fnvlist_lookup_int32(innvl, "fd");

	(void) nvlist_lookup_string(innvl, "fromsnap", &fromname);

	largeblockok = nvlist_exists(innvl, "largeblockok");
	embedok = nvlist_exists(innvl, "embedok");
	compressok = nvlist_exists(innvl, "compressok");
	rawok = nvlist_exists(innvl, "rawok");
	savedok = nvlist_exists(innvl, "savedok");

	(void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
	(void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);

	(void) nvlist_lookup_string(innvl, "redactbook", &redactbook);

	- if ((fp = zfs_file_get(fd)) == NULL)
	- return (SET_ERROR(EBADF));
	-
	- off = zfs_file_off(fp);
	+ dump_bytes_arg_t dba;
	+ dmu_send_outparams_t out;
	+ error = dump_bytes_init(&dba, fd, &out);
	+ if (error)
	+ return (error);

	- dmu_send_outparams_t out = {0};
	- out.dso_outfunc = dump_bytes;
	- out.dso_arg = fp;
	- out.dso_dryrun = B_FALSE;
	+ off = zfs_file_off(dba.dba_fp);
	error = dmu_send(snapname, fromname, embedok, largeblockok,
	compressok, rawok, savedok, resumeobj, resumeoff,
	redactbook, fd, &off, &out);

	- zfs_file_put(fp);
	+ dump_bytes_fini(&dba);
	+
	return (error);
	}

	static int
	send_space_sum(objset_t os, void buf, int len, void *arg)
	{
	(void) os, (void) buf;
	uint64_t *size = arg;

	*size += len;
	return (0);
	}

	/*
	* Determine approximately how large a zfs send stream will be -- the number
	* of bytes that will be written to the fd supplied to zfs_ioc_send_new().
	*
	* innvl: {
	* (optional) "from" -> full snap or bookmark name to send an incremental
	* from
	* (optional) "largeblockok" -> (value ignored)
	* indicates that blocks > 128KB are permitted
	* (optional) "embedok" -> (value ignored)
	* presence indicates DRR_WRITE_EMBEDDED records are permitted
	* (optional) "compressok" -> (value ignored)
	* presence indicates compressed DRR_WRITE records are permitted
	* (optional) "rawok" -> (value ignored)
	* presence indicates raw encrypted records should be used.
	* (optional) "resume_object" and "resume_offset" -> (uint64)
	* if present, resume send stream from specified object and offset.
	* (optional) "fd" -> file descriptor to use as a cookie for progress
	* tracking (int32)
	* }
	*
	* outnvl: {
	* "space" -> bytes of space (uint64)
	* }
	*/
	static const zfs_ioc_key_t zfs_keys_send_space[] = {
	{"from", DATA_TYPE_STRING, ZK_OPTIONAL},
	{"fromsnap", DATA_TYPE_STRING, ZK_OPTIONAL},
	{"largeblockok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"embedok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"compressok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"rawok", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	{"fd", DATA_TYPE_INT32, ZK_OPTIONAL},
	{"redactbook", DATA_TYPE_STRING, ZK_OPTIONAL},
	{"resume_object", DATA_TYPE_UINT64, ZK_OPTIONAL},
	{"resume_offset", DATA_TYPE_UINT64, ZK_OPTIONAL},
	{"bytes", DATA_TYPE_UINT64, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_send_space(const char snapname, nvlist_t innvl, nvlist_t *outnvl)
	{
	dsl_pool_t *dp;
	dsl_dataset_t *tosnap;
	dsl_dataset_t *fromsnap = NULL;
	int error;
	const char *fromname = NULL;
	const char *redactlist_book = NULL;
	boolean_t largeblockok;
	boolean_t embedok;
	boolean_t compressok;
	boolean_t rawok;
	boolean_t savedok;
	uint64_t space = 0;
	boolean_t full_estimate = B_FALSE;
	uint64_t resumeobj = 0;
	uint64_t resumeoff = 0;
	uint64_t resume_bytes = 0;
	int32_t fd = -1;
	zfs_bookmark_phys_t zbm = {0};

	error = dsl_pool_hold(snapname, FTAG, &dp);
	if (error != 0)
	return (error);

	error = dsl_dataset_hold(dp, snapname, FTAG, &tosnap);
	if (error != 0) {
	dsl_pool_rele(dp, FTAG);
	return (error);
	}
	(void) nvlist_lookup_int32(innvl, "fd", &fd);

	largeblockok = nvlist_exists(innvl, "largeblockok");
	embedok = nvlist_exists(innvl, "embedok");
	compressok = nvlist_exists(innvl, "compressok");
	rawok = nvlist_exists(innvl, "rawok");
	savedok = nvlist_exists(innvl, "savedok");
	boolean_t from = (nvlist_lookup_string(innvl, "from", &fromname) == 0);
	boolean_t altbook = (nvlist_lookup_string(innvl, "redactbook",
	&redactlist_book) == 0);

	(void) nvlist_lookup_uint64(innvl, "resume_object", &resumeobj);
	(void) nvlist_lookup_uint64(innvl, "resume_offset", &resumeoff);
	(void) nvlist_lookup_uint64(innvl, "bytes", &resume_bytes);

	if (altbook) {
	full_estimate = B_TRUE;
	} else if (from) {
	if (strchr(fromname, '#')) {
	error = dsl_bookmark_lookup(dp, fromname, tosnap, &zbm);

	/*
	* dsl_bookmark_lookup() will fail with EXDEV if
	* the from-bookmark and tosnap are at the same txg.
	* However, it's valid to do a send (and therefore,
	* a send estimate) from and to the same time point,
	* if the bookmark is redacted (the incremental send
	* can change what's redacted on the target). In
	* this case, dsl_bookmark_lookup() fills in zbm
	* but returns EXDEV. Ignore this error.
	*/
	if (error == EXDEV && zbm.zbm_redaction_obj != 0 &&
	zbm.zbm_guid ==
	dsl_dataset_phys(tosnap)->ds_guid)
	error = 0;

	if (error != 0) {
	dsl_dataset_rele(tosnap, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (error);
	}
	if (zbm.zbm_redaction_obj != 0 \|\| !(zbm.zbm_flags &
	ZBM_FLAG_HAS_FBN)) {
	full_estimate = B_TRUE;
	}
	} else if (strchr(fromname, '@')) {
	error = dsl_dataset_hold(dp, fromname, FTAG, &fromsnap);
	if (error != 0) {
	dsl_dataset_rele(tosnap, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (error);
	}

	if (!dsl_dataset_is_before(tosnap, fromsnap, 0)) {
	full_estimate = B_TRUE;
	dsl_dataset_rele(fromsnap, FTAG);
	}
	} else {
	/*
	* from is not properly formatted as a snapshot or
	* bookmark
	*/
	dsl_dataset_rele(tosnap, FTAG);
	dsl_pool_rele(dp, FTAG);
	return (SET_ERROR(EINVAL));
	}
	}

	if (full_estimate) {
	dmu_send_outparams_t out = {0};
	offset_t off = 0;
	out.dso_outfunc = send_space_sum;
	out.dso_arg = &space;
	out.dso_dryrun = B_TRUE;
	/*
	* We have to release these holds so dmu_send can take them. It
	* will do all the error checking we need.
	*/
	dsl_dataset_rele(tosnap, FTAG);
	dsl_pool_rele(dp, FTAG);
	error = dmu_send(snapname, fromname, embedok, largeblockok,
	compressok, rawok, savedok, resumeobj, resumeoff,
	redactlist_book, fd, &off, &out);
	} else {
	error = dmu_send_estimate_fast(tosnap, fromsnap,
	(from && strchr(fromname, '#') != NULL ? &zbm : NULL),
	compressok \|\| rawok, savedok, &space);
	space -= resume_bytes;
	if (fromsnap != NULL)
	dsl_dataset_rele(fromsnap, FTAG);
	dsl_dataset_rele(tosnap, FTAG);
	dsl_pool_rele(dp, FTAG);
	}

	fnvlist_add_uint64(outnvl, "space", space);

	return (error);
	}

	/*
	* Sync the currently open TXG to disk for the specified pool.
	* This is somewhat similar to 'zfs_sync()'.
	* For cases that do not result in error this ioctl will wait for
	* the currently open TXG to commit before returning back to the caller.
	*
	* innvl: {
	* "force" -> when true, force uberblock update even if there is no dirty data.
	* In addition this will cause the vdev configuration to be written
	* out including updating the zpool cache file. (boolean_t)
	* }
	*
	* onvl is unused
	*/
	static const zfs_ioc_key_t zfs_keys_pool_sync[] = {
	{"force", DATA_TYPE_BOOLEAN_VALUE, 0},
	};

	static int
	zfs_ioc_pool_sync(const char pool, nvlist_t innvl, nvlist_t *onvl)
	{
	(void) onvl;
	int err;
	boolean_t rc, force = B_FALSE;
	spa_t *spa;

	if ((err = spa_open(pool, &spa, FTAG)) != 0)
	return (err);

	if (innvl) {
	err = nvlist_lookup_boolean_value(innvl, "force", &rc);
	if (err == 0)
	force = rc;
	}

	if (force) {
	spa_config_enter(spa, SCL_CONFIG, FTAG, RW_WRITER);
	vdev_config_dirty(spa->spa_root_vdev);
	spa_config_exit(spa, SCL_CONFIG, FTAG);
	}
	txg_wait_synced(spa_get_dsl(spa), 0);

	spa_close(spa, FTAG);

	return (0);
	}

	/*
	* Load a user's wrapping key into the kernel.
	* innvl: {
	* "hidden_args" -> { "wkeydata" -> value }
	* raw uint8_t array of encryption wrapping key data (32 bytes)
	* (optional) "noop" -> (value ignored)
	* presence indicated key should only be verified, not loaded
	* }
	*/
	static const zfs_ioc_key_t zfs_keys_load_key[] = {
	{"hidden_args", DATA_TYPE_NVLIST, 0},
	{"noop", DATA_TYPE_BOOLEAN, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_load_key(const char dsname, nvlist_t innvl, nvlist_t *outnvl)
	{
	(void) outnvl;
	int ret;
	dsl_crypto_params_t *dcp = NULL;
	nvlist_t *hidden_args;
	boolean_t noop = nvlist_exists(innvl, "noop");

	if (strchr(dsname, '@') != NULL \|\| strchr(dsname, '%') != NULL) {
	ret = SET_ERROR(EINVAL);
	goto error;
	}

	hidden_args = fnvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS);

	ret = dsl_crypto_params_create_nvlist(DCP_CMD_NONE, NULL,
	hidden_args, &dcp);
	if (ret != 0)
	goto error;

	ret = spa_keystore_load_wkey(dsname, dcp, noop);
	if (ret != 0)
	goto error;

	dsl_crypto_params_free(dcp, noop);

	return (0);

	error:
	dsl_crypto_params_free(dcp, B_TRUE);
	return (ret);
	}

	/*
	* Unload a user's wrapping key from the kernel.
	* Both innvl and outnvl are unused.
	*/
	static const zfs_ioc_key_t zfs_keys_unload_key[] = {
	/* no nvl keys */
	};

	static int
	zfs_ioc_unload_key(const char dsname, nvlist_t innvl, nvlist_t *outnvl)
	{
	(void) innvl, (void) outnvl;
	int ret = 0;

	if (strchr(dsname, '@') != NULL \|\| strchr(dsname, '%') != NULL) {
	ret = (SET_ERROR(EINVAL));
	goto out;
	}

	ret = spa_keystore_unload_wkey(dsname);
	if (ret != 0)
	goto out;

	out:
	return (ret);
	}

	/*
	* Changes a user's wrapping key used to decrypt a dataset. The keyformat,
	* keylocation, pbkdf2salt, and pbkdf2iters properties can also be specified
	* here to change how the key is derived in userspace.
	*
	* innvl: {
	* "hidden_args" (optional) -> { "wkeydata" -> value }
	* raw uint8_t array of new encryption wrapping key data (32 bytes)
	* "props" (optional) -> { prop -> value }
	* }
	*
	* outnvl is unused
	*/
	static const zfs_ioc_key_t zfs_keys_change_key[] = {
	{"crypt_cmd", DATA_TYPE_UINT64, ZK_OPTIONAL},
	{"hidden_args", DATA_TYPE_NVLIST, ZK_OPTIONAL},
	{"props", DATA_TYPE_NVLIST, ZK_OPTIONAL},
	};

	static int
	zfs_ioc_change_key(const char dsname, nvlist_t innvl, nvlist_t *outnvl)
	{
	(void) outnvl;
	int ret;
	uint64_t cmd = DCP_CMD_NONE;
	dsl_crypto_params_t *dcp = NULL;
	nvlist_t args = NULL, hidden_args = NULL;

	if (strchr(dsname, '@') != NULL \|\| strchr(dsname, '%') != NULL) {
	ret = (SET_ERROR(EINVAL));
	goto error;
	}

	(void) nvlist_lookup_uint64(innvl, "crypt_cmd", &cmd);
	(void) nvlist_lookup_nvlist(innvl, "props", &args);
	(void) nvlist_lookup_nvlist(innvl, ZPOOL_HIDDEN_ARGS, &hidden_args);

	ret = dsl_crypto_params_create_nvlist(cmd, args, hidden_args, &dcp);
	if (ret != 0)
	goto error;

	ret = spa_keystore_change_key(dsname, dcp);
	if (ret != 0)
	goto error;

	dsl_crypto_params_free(dcp, B_FALSE);

	return (0);

	error:
	dsl_crypto_params_free(dcp, B_TRUE);
	return (ret);
	}

	static zfs_ioc_vec_t zfs_ioc_vec[ZFS_IOC_LAST - ZFS_IOC_FIRST];

	static void
	zfs_ioctl_register_legacy(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
	zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
	boolean_t log_history, zfs_ioc_poolcheck_t pool_check)
	{
	zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];

	ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
	ASSERT3U(ioc, <, ZFS_IOC_LAST);
	ASSERT3P(vec->zvec_legacy_func, ==, NULL);
	ASSERT3P(vec->zvec_func, ==, NULL);

	vec->zvec_legacy_func = func;
	vec->zvec_secpolicy = secpolicy;
	vec->zvec_namecheck = namecheck;
	vec->zvec_allow_log = log_history;
	vec->zvec_pool_check = pool_check;
	}

	/*
	* See the block comment at the beginning of this file for details on
	* each argument to this function.
	*/
	void
	zfs_ioctl_register(const char name, zfs_ioc_t ioc, zfs_ioc_func_t func,
	zfs_secpolicy_func_t *secpolicy, zfs_ioc_namecheck_t namecheck,
	zfs_ioc_poolcheck_t pool_check, boolean_t smush_outnvlist,
	boolean_t allow_log, const zfs_ioc_key_t *nvl_keys, size_t num_keys)
	{
	zfs_ioc_vec_t *vec = &zfs_ioc_vec[ioc - ZFS_IOC_FIRST];

	ASSERT3U(ioc, >=, ZFS_IOC_FIRST);
	ASSERT3U(ioc, <, ZFS_IOC_LAST);
	ASSERT3P(vec->zvec_legacy_func, ==, NULL);
	ASSERT3P(vec->zvec_func, ==, NULL);

	/* if we are logging, the name must be valid */
	ASSERT(!allow_log \|\| namecheck != NO_NAME);

	vec->zvec_name = name;
	vec->zvec_func = func;
	vec->zvec_secpolicy = secpolicy;
	vec->zvec_namecheck = namecheck;
	vec->zvec_pool_check = pool_check;
	vec->zvec_smush_outnvlist = smush_outnvlist;
	vec->zvec_allow_log = allow_log;
	vec->zvec_nvl_keys = nvl_keys;
	vec->zvec_nvl_key_count = num_keys;
	}

	static void
	zfs_ioctl_register_pool(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
	zfs_secpolicy_func_t *secpolicy, boolean_t log_history,
	zfs_ioc_poolcheck_t pool_check)
	{
	zfs_ioctl_register_legacy(ioc, func, secpolicy,
	POOL_NAME, log_history, pool_check);
	}

	void
	zfs_ioctl_register_dataset_nolog(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
	zfs_secpolicy_func_t *secpolicy, zfs_ioc_poolcheck_t pool_check)
	{
	zfs_ioctl_register_legacy(ioc, func, secpolicy,
	DATASET_NAME, B_FALSE, pool_check);
	}

	static void
	zfs_ioctl_register_pool_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
	{
	zfs_ioctl_register_legacy(ioc, func, zfs_secpolicy_config,
	POOL_NAME, B_TRUE, POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY);
	}

	static void
	zfs_ioctl_register_pool_meta(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
	zfs_secpolicy_func_t *secpolicy)
	{
	zfs_ioctl_register_legacy(ioc, func, secpolicy,
	NO_NAME, B_FALSE, POOL_CHECK_NONE);
	}

	static void
	zfs_ioctl_register_dataset_read_secpolicy(zfs_ioc_t ioc,
	zfs_ioc_legacy_func_t func, zfs_secpolicy_func_t secpolicy)
	{
	zfs_ioctl_register_legacy(ioc, func, secpolicy,
	DATASET_NAME, B_FALSE, POOL_CHECK_SUSPENDED);
	}

	static void
	zfs_ioctl_register_dataset_read(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func)
	{
	zfs_ioctl_register_dataset_read_secpolicy(ioc, func,
	zfs_secpolicy_read);
	}

	static void
	zfs_ioctl_register_dataset_modify(zfs_ioc_t ioc, zfs_ioc_legacy_func_t *func,
	zfs_secpolicy_func_t *secpolicy)
	{
	zfs_ioctl_register_legacy(ioc, func, secpolicy,
	DATASET_NAME, B_TRUE, POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY);
	}

	static void
	zfs_ioctl_init(void)
	{
	zfs_ioctl_register("snapshot", ZFS_IOC_SNAPSHOT,
	zfs_ioc_snapshot, zfs_secpolicy_snapshot, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_snapshot, ARRAY_SIZE(zfs_keys_snapshot));

	zfs_ioctl_register("log_history", ZFS_IOC_LOG_HISTORY,
	zfs_ioc_log_history, zfs_secpolicy_log_history, NO_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_FALSE, B_FALSE,
	zfs_keys_log_history, ARRAY_SIZE(zfs_keys_log_history));

	zfs_ioctl_register("space_snaps", ZFS_IOC_SPACE_SNAPS,
	zfs_ioc_space_snaps, zfs_secpolicy_read, DATASET_NAME,
	POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
	zfs_keys_space_snaps, ARRAY_SIZE(zfs_keys_space_snaps));

	zfs_ioctl_register("send", ZFS_IOC_SEND_NEW,
	zfs_ioc_send_new, zfs_secpolicy_send_new, DATASET_NAME,
	POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
	zfs_keys_send_new, ARRAY_SIZE(zfs_keys_send_new));

	zfs_ioctl_register("send_space", ZFS_IOC_SEND_SPACE,
	zfs_ioc_send_space, zfs_secpolicy_read, DATASET_NAME,
	POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
	zfs_keys_send_space, ARRAY_SIZE(zfs_keys_send_space));

	zfs_ioctl_register("create", ZFS_IOC_CREATE,
	zfs_ioc_create, zfs_secpolicy_create_clone, DATASET_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_create, ARRAY_SIZE(zfs_keys_create));

	zfs_ioctl_register("clone", ZFS_IOC_CLONE,
	zfs_ioc_clone, zfs_secpolicy_create_clone, DATASET_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_clone, ARRAY_SIZE(zfs_keys_clone));

	zfs_ioctl_register("remap", ZFS_IOC_REMAP,
	zfs_ioc_remap, zfs_secpolicy_none, DATASET_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_FALSE, B_TRUE,
	zfs_keys_remap, ARRAY_SIZE(zfs_keys_remap));

	zfs_ioctl_register("destroy_snaps", ZFS_IOC_DESTROY_SNAPS,
	zfs_ioc_destroy_snaps, zfs_secpolicy_destroy_snaps, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_destroy_snaps, ARRAY_SIZE(zfs_keys_destroy_snaps));

	zfs_ioctl_register("hold", ZFS_IOC_HOLD,
	zfs_ioc_hold, zfs_secpolicy_hold, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_hold, ARRAY_SIZE(zfs_keys_hold));
	zfs_ioctl_register("release", ZFS_IOC_RELEASE,
	zfs_ioc_release, zfs_secpolicy_release, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_release, ARRAY_SIZE(zfs_keys_release));

	zfs_ioctl_register("get_holds", ZFS_IOC_GET_HOLDS,
	zfs_ioc_get_holds, zfs_secpolicy_read, DATASET_NAME,
	POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
	zfs_keys_get_holds, ARRAY_SIZE(zfs_keys_get_holds));

	zfs_ioctl_register("rollback", ZFS_IOC_ROLLBACK,
	zfs_ioc_rollback, zfs_secpolicy_rollback, DATASET_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_FALSE, B_TRUE,
	zfs_keys_rollback, ARRAY_SIZE(zfs_keys_rollback));

	zfs_ioctl_register("bookmark", ZFS_IOC_BOOKMARK,
	zfs_ioc_bookmark, zfs_secpolicy_bookmark, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_bookmark, ARRAY_SIZE(zfs_keys_bookmark));

	zfs_ioctl_register("get_bookmarks", ZFS_IOC_GET_BOOKMARKS,
	zfs_ioc_get_bookmarks, zfs_secpolicy_read, DATASET_NAME,
	POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE,
	zfs_keys_get_bookmarks, ARRAY_SIZE(zfs_keys_get_bookmarks));

	zfs_ioctl_register("get_bookmark_props", ZFS_IOC_GET_BOOKMARK_PROPS,
	zfs_ioc_get_bookmark_props, zfs_secpolicy_read, ENTITY_NAME,
	POOL_CHECK_SUSPENDED, B_FALSE, B_FALSE, zfs_keys_get_bookmark_props,
	ARRAY_SIZE(zfs_keys_get_bookmark_props));

	zfs_ioctl_register("destroy_bookmarks", ZFS_IOC_DESTROY_BOOKMARKS,
	zfs_ioc_destroy_bookmarks, zfs_secpolicy_destroy_bookmarks,
	POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_destroy_bookmarks,
	ARRAY_SIZE(zfs_keys_destroy_bookmarks));

	zfs_ioctl_register("receive", ZFS_IOC_RECV_NEW,
	zfs_ioc_recv_new, zfs_secpolicy_recv, DATASET_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_recv_new, ARRAY_SIZE(zfs_keys_recv_new));
	zfs_ioctl_register("load-key", ZFS_IOC_LOAD_KEY,
	zfs_ioc_load_key, zfs_secpolicy_load_key,
	DATASET_NAME, POOL_CHECK_SUSPENDED, B_TRUE, B_TRUE,
	zfs_keys_load_key, ARRAY_SIZE(zfs_keys_load_key));
	zfs_ioctl_register("unload-key", ZFS_IOC_UNLOAD_KEY,
	zfs_ioc_unload_key, zfs_secpolicy_load_key,
	DATASET_NAME, POOL_CHECK_SUSPENDED, B_TRUE, B_TRUE,
	zfs_keys_unload_key, ARRAY_SIZE(zfs_keys_unload_key));
	zfs_ioctl_register("change-key", ZFS_IOC_CHANGE_KEY,
	zfs_ioc_change_key, zfs_secpolicy_change_key,
	DATASET_NAME, POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY,
	B_TRUE, B_TRUE, zfs_keys_change_key,
	ARRAY_SIZE(zfs_keys_change_key));

	zfs_ioctl_register("sync", ZFS_IOC_POOL_SYNC,
	zfs_ioc_pool_sync, zfs_secpolicy_none, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_FALSE, B_FALSE,
	zfs_keys_pool_sync, ARRAY_SIZE(zfs_keys_pool_sync));
	zfs_ioctl_register("reopen", ZFS_IOC_POOL_REOPEN, zfs_ioc_pool_reopen,
	zfs_secpolicy_config, POOL_NAME, POOL_CHECK_SUSPENDED, B_TRUE,
	B_TRUE, zfs_keys_pool_reopen, ARRAY_SIZE(zfs_keys_pool_reopen));

	zfs_ioctl_register("channel_program", ZFS_IOC_CHANNEL_PROGRAM,
	zfs_ioc_channel_program, zfs_secpolicy_config,
	POOL_NAME, POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE,
	B_TRUE, zfs_keys_channel_program,
	ARRAY_SIZE(zfs_keys_channel_program));

	zfs_ioctl_register("redact", ZFS_IOC_REDACT,
	zfs_ioc_redact, zfs_secpolicy_config, DATASET_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_redact, ARRAY_SIZE(zfs_keys_redact));

	zfs_ioctl_register("zpool_checkpoint", ZFS_IOC_POOL_CHECKPOINT,
	zfs_ioc_pool_checkpoint, zfs_secpolicy_config, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_pool_checkpoint, ARRAY_SIZE(zfs_keys_pool_checkpoint));

	zfs_ioctl_register("zpool_discard_checkpoint",
	ZFS_IOC_POOL_DISCARD_CHECKPOINT, zfs_ioc_pool_discard_checkpoint,
	zfs_secpolicy_config, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_pool_discard_checkpoint,
	ARRAY_SIZE(zfs_keys_pool_discard_checkpoint));

	zfs_ioctl_register("initialize", ZFS_IOC_POOL_INITIALIZE,
	zfs_ioc_pool_initialize, zfs_secpolicy_config, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_pool_initialize, ARRAY_SIZE(zfs_keys_pool_initialize));

	zfs_ioctl_register("trim", ZFS_IOC_POOL_TRIM,
	zfs_ioc_pool_trim, zfs_secpolicy_config, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_TRUE, B_TRUE,
	zfs_keys_pool_trim, ARRAY_SIZE(zfs_keys_pool_trim));

	zfs_ioctl_register("wait", ZFS_IOC_WAIT,
	zfs_ioc_wait, zfs_secpolicy_none, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_FALSE, B_FALSE,
	zfs_keys_pool_wait, ARRAY_SIZE(zfs_keys_pool_wait));

	zfs_ioctl_register("wait_fs", ZFS_IOC_WAIT_FS,
	zfs_ioc_wait_fs, zfs_secpolicy_none, DATASET_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_FALSE, B_FALSE,
	zfs_keys_fs_wait, ARRAY_SIZE(zfs_keys_fs_wait));

	zfs_ioctl_register("set_bootenv", ZFS_IOC_SET_BOOTENV,
	zfs_ioc_set_bootenv, zfs_secpolicy_config, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_FALSE, B_TRUE,
	zfs_keys_set_bootenv, ARRAY_SIZE(zfs_keys_set_bootenv));

	zfs_ioctl_register("get_bootenv", ZFS_IOC_GET_BOOTENV,
	zfs_ioc_get_bootenv, zfs_secpolicy_none, POOL_NAME,
	POOL_CHECK_SUSPENDED, B_FALSE, B_TRUE,
	zfs_keys_get_bootenv, ARRAY_SIZE(zfs_keys_get_bootenv));

	zfs_ioctl_register("zpool_vdev_get_props", ZFS_IOC_VDEV_GET_PROPS,
	zfs_ioc_vdev_get_props, zfs_secpolicy_read, POOL_NAME,
	POOL_CHECK_NONE, B_FALSE, B_FALSE, zfs_keys_vdev_get_props,
	ARRAY_SIZE(zfs_keys_vdev_get_props));

	zfs_ioctl_register("zpool_vdev_set_props", ZFS_IOC_VDEV_SET_PROPS,
	zfs_ioc_vdev_set_props, zfs_secpolicy_config, POOL_NAME,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY, B_FALSE, B_FALSE,
	zfs_keys_vdev_set_props, ARRAY_SIZE(zfs_keys_vdev_set_props));

	zfs_ioctl_register("scrub", ZFS_IOC_POOL_SCRUB,
	zfs_ioc_pool_scrub, zfs_secpolicy_config, POOL_NAME,
	POOL_CHECK_NONE, B_TRUE, B_TRUE,
	zfs_keys_pool_scrub, ARRAY_SIZE(zfs_keys_pool_scrub));

	/* IOCTLS that use the legacy function signature */

	zfs_ioctl_register_legacy(ZFS_IOC_POOL_FREEZE, zfs_ioc_pool_freeze,
	zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_READONLY);

	zfs_ioctl_register_pool(ZFS_IOC_POOL_CREATE, zfs_ioc_pool_create,
	zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);
	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SCAN,
	zfs_ioc_pool_scan);
	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_UPGRADE,
	zfs_ioc_pool_upgrade);
	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ADD,
	zfs_ioc_vdev_add);
	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_REMOVE,
	zfs_ioc_vdev_remove);
	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SET_STATE,
	zfs_ioc_vdev_set_state);
	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_ATTACH,
	zfs_ioc_vdev_attach);
	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_DETACH,
	zfs_ioc_vdev_detach);
	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETPATH,
	zfs_ioc_vdev_setpath);
	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SETFRU,
	zfs_ioc_vdev_setfru);
	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_SET_PROPS,
	zfs_ioc_pool_set_props);
	zfs_ioctl_register_pool_modify(ZFS_IOC_VDEV_SPLIT,
	zfs_ioc_vdev_split);
	zfs_ioctl_register_pool_modify(ZFS_IOC_POOL_REGUID,
	zfs_ioc_pool_reguid);

	zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_CONFIGS,
	zfs_ioc_pool_configs, zfs_secpolicy_none);
	zfs_ioctl_register_pool_meta(ZFS_IOC_POOL_TRYIMPORT,
	zfs_ioc_pool_tryimport, zfs_secpolicy_config);
	zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_FAULT,
	zfs_ioc_inject_fault, zfs_secpolicy_inject);
	zfs_ioctl_register_pool_meta(ZFS_IOC_CLEAR_FAULT,
	zfs_ioc_clear_fault, zfs_secpolicy_inject);
	zfs_ioctl_register_pool_meta(ZFS_IOC_INJECT_LIST_NEXT,
	zfs_ioc_inject_list_next, zfs_secpolicy_inject);

	/*
	* pool destroy, and export don't log the history as part of
	* zfsdev_ioctl, but rather zfs_ioc_pool_export
	* does the logging of those commands.
	*/
	zfs_ioctl_register_pool(ZFS_IOC_POOL_DESTROY, zfs_ioc_pool_destroy,
	zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);
	zfs_ioctl_register_pool(ZFS_IOC_POOL_EXPORT, zfs_ioc_pool_export,
	zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);

	zfs_ioctl_register_pool(ZFS_IOC_POOL_STATS, zfs_ioc_pool_stats,
	zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);
	zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_PROPS, zfs_ioc_pool_get_props,
	zfs_secpolicy_read, B_FALSE, POOL_CHECK_NONE);

	zfs_ioctl_register_pool(ZFS_IOC_ERROR_LOG, zfs_ioc_error_log,
	zfs_secpolicy_inject, B_FALSE, POOL_CHECK_SUSPENDED);
	zfs_ioctl_register_pool(ZFS_IOC_DSOBJ_TO_DSNAME,
	zfs_ioc_dsobj_to_dsname,
	zfs_secpolicy_diff, B_FALSE, POOL_CHECK_SUSPENDED);
	zfs_ioctl_register_pool(ZFS_IOC_POOL_GET_HISTORY,
	zfs_ioc_pool_get_history,
	zfs_secpolicy_config, B_FALSE, POOL_CHECK_SUSPENDED);

	zfs_ioctl_register_pool(ZFS_IOC_POOL_IMPORT, zfs_ioc_pool_import,
	zfs_secpolicy_config, B_TRUE, POOL_CHECK_NONE);

	zfs_ioctl_register_pool(ZFS_IOC_CLEAR, zfs_ioc_clear,
	zfs_secpolicy_config, B_TRUE, POOL_CHECK_READONLY);

	zfs_ioctl_register_dataset_read(ZFS_IOC_SPACE_WRITTEN,
	zfs_ioc_space_written);
	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_RECVD_PROPS,
	zfs_ioc_objset_recvd_props);
	zfs_ioctl_register_dataset_read(ZFS_IOC_NEXT_OBJ,
	zfs_ioc_next_obj);
	zfs_ioctl_register_dataset_read(ZFS_IOC_GET_FSACL,
	zfs_ioc_get_fsacl);
	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_STATS,
	zfs_ioc_objset_stats);
	zfs_ioctl_register_dataset_read(ZFS_IOC_OBJSET_ZPLPROPS,
	zfs_ioc_objset_zplprops);
	zfs_ioctl_register_dataset_read(ZFS_IOC_DATASET_LIST_NEXT,
	zfs_ioc_dataset_list_next);
	zfs_ioctl_register_dataset_read(ZFS_IOC_SNAPSHOT_LIST_NEXT,
	zfs_ioc_snapshot_list_next);
	zfs_ioctl_register_dataset_read(ZFS_IOC_SEND_PROGRESS,
	zfs_ioc_send_progress);

	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_DIFF,
	zfs_ioc_diff, zfs_secpolicy_diff);
	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_STATS,
	zfs_ioc_obj_to_stats, zfs_secpolicy_diff);
	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_OBJ_TO_PATH,
	zfs_ioc_obj_to_path, zfs_secpolicy_diff);
	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_ONE,
	zfs_ioc_userspace_one, zfs_secpolicy_userspace_one);
	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_USERSPACE_MANY,
	zfs_ioc_userspace_many, zfs_secpolicy_userspace_many);
	zfs_ioctl_register_dataset_read_secpolicy(ZFS_IOC_SEND,
	zfs_ioc_send, zfs_secpolicy_send);

	zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_PROP, zfs_ioc_set_prop,
	zfs_secpolicy_none);
	zfs_ioctl_register_dataset_modify(ZFS_IOC_DESTROY, zfs_ioc_destroy,
	zfs_secpolicy_destroy);
	zfs_ioctl_register_dataset_modify(ZFS_IOC_RENAME, zfs_ioc_rename,
	zfs_secpolicy_rename);
	zfs_ioctl_register_dataset_modify(ZFS_IOC_RECV, zfs_ioc_recv,
	zfs_secpolicy_recv);
	zfs_ioctl_register_dataset_modify(ZFS_IOC_PROMOTE, zfs_ioc_promote,
	zfs_secpolicy_promote);
	zfs_ioctl_register_dataset_modify(ZFS_IOC_INHERIT_PROP,
	zfs_ioc_inherit_prop, zfs_secpolicy_inherit_prop);
	zfs_ioctl_register_dataset_modify(ZFS_IOC_SET_FSACL, zfs_ioc_set_fsacl,
	zfs_secpolicy_set_fsacl);

	zfs_ioctl_register_dataset_nolog(ZFS_IOC_SHARE, zfs_ioc_share,
	zfs_secpolicy_share, POOL_CHECK_NONE);
	zfs_ioctl_register_dataset_nolog(ZFS_IOC_SMB_ACL, zfs_ioc_smb_acl,
	zfs_secpolicy_smb_acl, POOL_CHECK_NONE);
	zfs_ioctl_register_dataset_nolog(ZFS_IOC_USERSPACE_UPGRADE,
	zfs_ioc_userspace_upgrade, zfs_secpolicy_userspace_upgrade,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY);
	zfs_ioctl_register_dataset_nolog(ZFS_IOC_TMP_SNAPSHOT,
	zfs_ioc_tmp_snapshot, zfs_secpolicy_tmp_snapshot,
	POOL_CHECK_SUSPENDED \| POOL_CHECK_READONLY);

	zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_NEXT, zfs_ioc_events_next,
	zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
	zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_CLEAR, zfs_ioc_events_clear,
	zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);
	zfs_ioctl_register_legacy(ZFS_IOC_EVENTS_SEEK, zfs_ioc_events_seek,
	zfs_secpolicy_config, NO_NAME, B_FALSE, POOL_CHECK_NONE);

	zfs_ioctl_init_os();
	}

	/*
	* Verify that for non-legacy ioctls the input nvlist
	* pairs match against the expected input.
	*
	* Possible errors are:
	* ZFS_ERR_IOC_ARG_UNAVAIL An unrecognized nvpair was encountered
	* ZFS_ERR_IOC_ARG_REQUIRED A required nvpair is missing
	* ZFS_ERR_IOC_ARG_BADTYPE Invalid type for nvpair
	*/
	static int
	zfs_check_input_nvpairs(nvlist_t innvl, const zfs_ioc_vec_t vec)
	{
	const zfs_ioc_key_t *nvl_keys = vec->zvec_nvl_keys;
	boolean_t required_keys_found = B_FALSE;

	/*
	* examine each input pair
	*/
	for (nvpair_t *pair = nvlist_next_nvpair(innvl, NULL);
	pair != NULL; pair = nvlist_next_nvpair(innvl, pair)) {
	const char *name = nvpair_name(pair);
	data_type_t type = nvpair_type(pair);
	boolean_t identified = B_FALSE;

	/*
	* check pair against the documented names and type
	*/
	for (int k = 0; k < vec->zvec_nvl_key_count; k++) {
	/* if not a wild card name, check for an exact match */
	if ((nvl_keys[k].zkey_flags & ZK_WILDCARDLIST) == 0 &&
	strcmp(nvl_keys[k].zkey_name, name) != 0)
	continue;

	identified = B_TRUE;

	if (nvl_keys[k].zkey_type != DATA_TYPE_ANY &&
	nvl_keys[k].zkey_type != type) {
	return (SET_ERROR(ZFS_ERR_IOC_ARG_BADTYPE));
	}

	if (nvl_keys[k].zkey_flags & ZK_OPTIONAL)
	continue;

	required_keys_found = B_TRUE;
	break;
	}

	/* allow an 'optional' key, everything else is invalid */
	if (!identified &&
	(strcmp(name, "optional") != 0 \|\|
	type != DATA_TYPE_NVLIST)) {
	return (SET_ERROR(ZFS_ERR_IOC_ARG_UNAVAIL));
	}
	}

	/* verify that all required keys were found */
	for (int k = 0; k < vec->zvec_nvl_key_count; k++) {
	if (nvl_keys[k].zkey_flags & ZK_OPTIONAL)
	continue;

	if (nvl_keys[k].zkey_flags & ZK_WILDCARDLIST) {
	/* at least one non-optional key is expected here */
	if (!required_keys_found)
	return (SET_ERROR(ZFS_ERR_IOC_ARG_REQUIRED));
	continue;
	}

	if (!nvlist_exists(innvl, nvl_keys[k].zkey_name))
	return (SET_ERROR(ZFS_ERR_IOC_ARG_REQUIRED));
	}

	return (0);
	}

	static int
	pool_status_check(const char *name, zfs_ioc_namecheck_t type,
	zfs_ioc_poolcheck_t check)
	{
	spa_t *spa;
	int error;

	ASSERT(type == POOL_NAME \|\| type == DATASET_NAME \|\|
	type == ENTITY_NAME);

	if (check & POOL_CHECK_NONE)
	return (0);

	error = spa_open(name, &spa, FTAG);
	if (error == 0) {
	if ((check & POOL_CHECK_SUSPENDED) && spa_suspended(spa))
	error = SET_ERROR(EAGAIN);
	else if ((check & POOL_CHECK_READONLY) && !spa_writeable(spa))
	error = SET_ERROR(EROFS);
	spa_close(spa, FTAG);
	}
	return (error);
	}

	int
	zfsdev_getminor(zfs_file_t fp, minor_t minorp)
	{
	zfsdev_state_t zs, fpd;

	ASSERT(!MUTEX_HELD(&zfsdev_state_lock));

	fpd = zfs_file_private(fp);
	if (fpd == NULL)
	return (SET_ERROR(EBADF));

	mutex_enter(&zfsdev_state_lock);

	for (zs = &zfsdev_state_listhead; zs != NULL; zs = zs->zs_next) {

	if (zs->zs_minor == -1)
	continue;

	if (fpd == zs) {
	*minorp = fpd->zs_minor;
	mutex_exit(&zfsdev_state_lock);
	return (0);
	}
	}

	mutex_exit(&zfsdev_state_lock);

	return (SET_ERROR(EBADF));
	}

	void *
	zfsdev_get_state(minor_t minor, enum zfsdev_state_type which)
	{
	zfsdev_state_t *zs;

	for (zs = &zfsdev_state_listhead; zs != NULL; zs = zs->zs_next) {
	if (zs->zs_minor == minor) {
	membar_consumer();
	switch (which) {
	case ZST_ONEXIT:
	return (zs->zs_onexit);
	case ZST_ZEVENT:
	return (zs->zs_zevent);
	case ZST_ALL:
	return (zs);
	}
	}
	}

	return (NULL);
	}

	/*
	* Find a free minor number. The zfsdev_state_list is expected to
	* be short since it is only a list of currently open file handles.
	*/
	static minor_t
	zfsdev_minor_alloc(void)
	{
	static minor_t last_minor = 0;
	minor_t m;

	ASSERT(MUTEX_HELD(&zfsdev_state_lock));

	for (m = last_minor + 1; m != last_minor; m++) {
	if (m > ZFSDEV_MAX_MINOR)
	m = 1;
	if (zfsdev_get_state(m, ZST_ALL) == NULL) {
	last_minor = m;
	return (m);
	}
	}

	return (0);
	}

	int
	zfsdev_state_init(void *priv)
	{
	zfsdev_state_t zs, zsprev = NULL;
	minor_t minor;
	boolean_t newzs = B_FALSE;

	ASSERT(MUTEX_HELD(&zfsdev_state_lock));

	minor = zfsdev_minor_alloc();
	if (minor == 0)
	return (SET_ERROR(ENXIO));

	for (zs = &zfsdev_state_listhead; zs != NULL; zs = zs->zs_next) {
	if (zs->zs_minor == -1)
	break;
	zsprev = zs;
	}

	if (!zs) {
	zs = kmem_zalloc(sizeof (zfsdev_state_t), KM_SLEEP);
	newzs = B_TRUE;
	}

	zfsdev_private_set_state(priv, zs);

	zfs_onexit_init((zfs_onexit_t **)&zs->zs_onexit);
	zfs_zevent_init((zfs_zevent_t **)&zs->zs_zevent);

	/*
	* In order to provide for lock-free concurrent read access
	* to the minor list in zfsdev_get_state(), new entries
	* must be completely written before linking them into the
	* list whereas existing entries are already linked; the last
	* operation must be updating zs_minor (from -1 to the new
	* value).
	*/
	if (newzs) {
	zs->zs_minor = minor;
	membar_producer();
	zsprev->zs_next = zs;
	} else {
	membar_producer();
	zs->zs_minor = minor;
	}

	return (0);
	}

	void
	zfsdev_state_destroy(void *priv)
	{
	zfsdev_state_t *zs = zfsdev_private_get_state(priv);

	ASSERT(zs != NULL);
	ASSERT3S(zs->zs_minor, >, 0);

	/*
	* The last reference to this zfsdev file descriptor is being dropped.
	* We don't have to worry about lookup grabbing this state object, and
	* zfsdev_state_init() will not try to reuse this object until it is
	* invalidated by setting zs_minor to -1. Invalidation must be done
	* last, with a memory barrier to ensure ordering. This lets us avoid
	* taking the global zfsdev state lock around destruction.
	*/
	zfs_onexit_destroy(zs->zs_onexit);
	zfs_zevent_destroy(zs->zs_zevent);
	zs->zs_onexit = NULL;
	zs->zs_zevent = NULL;
	membar_producer();
	zs->zs_minor = -1;
	}

	long
	zfsdev_ioctl_common(uint_t vecnum, zfs_cmd_t *zc, int flag)
	{
	int error, cmd;
	const zfs_ioc_vec_t *vec;
	char *saved_poolname = NULL;
	uint64_t max_nvlist_src_size;
	size_t saved_poolname_len = 0;
	nvlist_t *innvl = NULL;
	fstrans_cookie_t cookie;
	hrtime_t start_time = gethrtime();

	cmd = vecnum;
	error = 0;
	if (vecnum >= sizeof (zfs_ioc_vec) / sizeof (zfs_ioc_vec[0]))
	return (SET_ERROR(ZFS_ERR_IOC_CMD_UNAVAIL));

	vec = &zfs_ioc_vec[vecnum];

	/*
	* The registered ioctl list may be sparse, verify that either
	* a normal or legacy handler are registered.
	*/
	if (vec->zvec_func == NULL && vec->zvec_legacy_func == NULL)
	return (SET_ERROR(ZFS_ERR_IOC_CMD_UNAVAIL));

	zc->zc_iflags = flag & FKIOCTL;
	max_nvlist_src_size = zfs_max_nvlist_src_size_os();
	if (zc->zc_nvlist_src_size > max_nvlist_src_size) {
	/*
	* Make sure the user doesn't pass in an insane value for
	* zc_nvlist_src_size. We have to check, since we will end
	* up allocating that much memory inside of get_nvlist(). This
	* prevents a nefarious user from allocating tons of kernel
	* memory.
	*
	* Also, we return EINVAL instead of ENOMEM here. The reason
	* being that returning ENOMEM from an ioctl() has a special
	* connotation; that the user's size value is too small and
	* needs to be expanded to hold the nvlist. See
	* zcmd_expand_dst_nvlist() for details.
	*/
	error = SET_ERROR(EINVAL); /* User's size too big */

	} else if (zc->zc_nvlist_src_size != 0) {
	error = get_nvlist(zc->zc_nvlist_src, zc->zc_nvlist_src_size,
	zc->zc_iflags, &innvl);
	if (error != 0)
	goto out;
	}

	/*
	* Ensure that all pool/dataset names are valid before we pass down to
	* the lower layers.
	*/
	zc->zc_name[sizeof (zc->zc_name) - 1] = '\0';
	switch (vec->zvec_namecheck) {
	case POOL_NAME:
	if (pool_namecheck(zc->zc_name, NULL, NULL) != 0)
	error = SET_ERROR(EINVAL);
	else
	error = pool_status_check(zc->zc_name,
	vec->zvec_namecheck, vec->zvec_pool_check);
	break;

	case DATASET_NAME:
	if (dataset_namecheck(zc->zc_name, NULL, NULL) != 0)
	error = SET_ERROR(EINVAL);
	else
	error = pool_status_check(zc->zc_name,
	vec->zvec_namecheck, vec->zvec_pool_check);
	break;

	case ENTITY_NAME:
	if (entity_namecheck(zc->zc_name, NULL, NULL) != 0) {
	error = SET_ERROR(EINVAL);
	} else {
	error = pool_status_check(zc->zc_name,
	vec->zvec_namecheck, vec->zvec_pool_check);
	}
	break;

	case NO_NAME:
	break;
	}
	/*
	* Ensure that all input pairs are valid before we pass them down
	* to the lower layers.
	*
	* The vectored functions can use fnvlist_lookup_{type} for any
	* required pairs since zfs_check_input_nvpairs() confirmed that
	* they exist and are of the correct type.
	*/
	if (error == 0 && vec->zvec_func != NULL) {
	error = zfs_check_input_nvpairs(innvl, vec);
	if (error != 0)
	goto out;
	}

	if (error == 0) {
	cookie = spl_fstrans_mark();
	error = vec->zvec_secpolicy(zc, innvl, CRED());
	spl_fstrans_unmark(cookie);
	}

	if (error != 0)
	goto out;

	/* legacy ioctls can modify zc_name */
	/*
	* Can't use kmem_strdup() as we might truncate the string and
	* kmem_strfree() would then free with incorrect size.
	*/
	saved_poolname_len = strlen(zc->zc_name) + 1;
	saved_poolname = kmem_alloc(saved_poolname_len, KM_SLEEP);

	strlcpy(saved_poolname, zc->zc_name, saved_poolname_len);
	saved_poolname[strcspn(saved_poolname, "/@#")] = '\0';

	if (vec->zvec_func != NULL) {
	nvlist_t *outnvl;
	int puterror = 0;
	spa_t *spa;
	nvlist_t *lognv = NULL;

	ASSERT(vec->zvec_legacy_func == NULL);

	/*
	* Add the innvl to the lognv before calling the func,
	* in case the func changes the innvl.
	*/
	if (vec->zvec_allow_log) {
	lognv = fnvlist_alloc();
	fnvlist_add_string(lognv, ZPOOL_HIST_IOCTL,
	vec->zvec_name);
	if (!nvlist_empty(innvl)) {
	fnvlist_add_nvlist(lognv, ZPOOL_HIST_INPUT_NVL,
	innvl);
	}
	}

	outnvl = fnvlist_alloc();
	cookie = spl_fstrans_mark();
	error = vec->zvec_func(zc->zc_name, innvl, outnvl);
	spl_fstrans_unmark(cookie);

	/*
	* Some commands can partially execute, modify state, and still
	* return an error. In these cases, attempt to record what
	* was modified.
	*/
	if ((error == 0 \|\|
	(cmd == ZFS_IOC_CHANNEL_PROGRAM && error != EINVAL)) &&
	vec->zvec_allow_log &&
	spa_open(zc->zc_name, &spa, FTAG) == 0) {
	if (!nvlist_empty(outnvl)) {
	size_t out_size = fnvlist_size(outnvl);
	if (out_size > zfs_history_output_max) {
	fnvlist_add_int64(lognv,
	ZPOOL_HIST_OUTPUT_SIZE, out_size);
	} else {
	fnvlist_add_nvlist(lognv,
	ZPOOL_HIST_OUTPUT_NVL, outnvl);
	}
	}
	if (error != 0) {
	fnvlist_add_int64(lognv, ZPOOL_HIST_ERRNO,
	error);
	}
	fnvlist_add_int64(lognv, ZPOOL_HIST_ELAPSED_NS,
	gethrtime() - start_time);
	(void) spa_history_log_nvl(spa, lognv);
	spa_close(spa, FTAG);
	}
	fnvlist_free(lognv);

	if (!nvlist_empty(outnvl) \|\| zc->zc_nvlist_dst_size != 0) {
	int smusherror = 0;
	if (vec->zvec_smush_outnvlist) {
	smusherror = nvlist_smush(outnvl,
	zc->zc_nvlist_dst_size);
	}
	if (smusherror == 0)
	puterror = put_nvlist(zc, outnvl);
	}

	if (puterror != 0)
	error = puterror;

	nvlist_free(outnvl);
	} else {
	cookie = spl_fstrans_mark();
	error = vec->zvec_legacy_func(zc);
	spl_fstrans_unmark(cookie);
	}

	out:
	nvlist_free(innvl);
	if (error == 0 && vec->zvec_allow_log) {
	char *s = tsd_get(zfs_allow_log_key);
	if (s != NULL)
	kmem_strfree(s);
	(void) tsd_set(zfs_allow_log_key, kmem_strdup(saved_poolname));
	}
	if (saved_poolname != NULL)
	kmem_free(saved_poolname, saved_poolname_len);

	return (error);
	}

	int
	zfs_kmod_init(void)
	{
	int error;

	if ((error = zvol_init()) != 0)
	return (error);

	spa_init(SPA_MODE_READ \| SPA_MODE_WRITE);
	zfs_init();

	zfs_ioctl_init();

	mutex_init(&zfsdev_state_lock, NULL, MUTEX_DEFAULT, NULL);
	zfsdev_state_listhead.zs_minor = -1;

	if ((error = zfsdev_attach()) != 0)
	goto out;

	tsd_create(&rrw_tsd_key, rrw_tsd_destroy);
	tsd_create(&zfs_allow_log_key, zfs_allow_log_destroy);

	return (0);
	out:
	zfs_fini();
	spa_fini();
	zvol_fini();

	return (error);
	}

	void
	zfs_kmod_fini(void)
	{
	zfsdev_state_t zs, zsnext = NULL;

	zfsdev_detach();

	mutex_destroy(&zfsdev_state_lock);

	for (zs = &zfsdev_state_listhead; zs != NULL; zs = zsnext) {
	zsnext = zs->zs_next;
	if (zs->zs_onexit)
	zfs_onexit_destroy(zs->zs_onexit);
	if (zs->zs_zevent)
	zfs_zevent_destroy(zs->zs_zevent);
	if (zs != &zfsdev_state_listhead)
	kmem_free(zs, sizeof (zfsdev_state_t));
	}

	zfs_ereport_taskq_fini(); /* run before zfs_fini() on Linux */
	zfs_fini();
	spa_fini();
	zvol_fini();

	tsd_destroy(&rrw_tsd_key);
	tsd_destroy(&zfs_allow_log_key);
	}

	ZFS_MODULE_PARAM(zfs, zfs_, max_nvlist_src_size, U64, ZMOD_RW,
	"Maximum size in bytes allowed for src nvlist passed with ZFS ioctls");

	ZFS_MODULE_PARAM(zfs, zfs_, history_output_max, U64, ZMOD_RW,
	"Maximum size in bytes of ZFS ioctl output that will be logged");
	diff --git a/sys/contrib/openzfs/module/zfs/zfs_log.c b/sys/contrib/openzfs/module/zfs/zfs_log.c
	index 433a653e5500..fa4e7093ca46 100644
	--- a/sys/contrib/openzfs/module/zfs/zfs_log.c
	+++ b/sys/contrib/openzfs/module/zfs/zfs_log.c
	@@ -1,937 +1,935 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2015, 2018 by Delphix. All rights reserved.
	* Copyright (c) 2022 by Pawel Jakub Dawidek
	*/


	#include <sys/types.h>
	#include <sys/param.h>
	#include <sys/sysmacros.h>
	#include <sys/cmn_err.h>
	#include <sys/kmem.h>
	#include <sys/thread.h>
	#include <sys/file.h>
	#include <sys/vfs.h>
	#include <sys/zfs_znode.h>
	#include <sys/zfs_dir.h>
	#include <sys/zil.h>
	#include <sys/zil_impl.h>
	#include <sys/byteorder.h>
	#include <sys/policy.h>
	#include <sys/stat.h>
	#include <sys/acl.h>
	#include <sys/dmu.h>
	#include <sys/dbuf.h>
	#include <sys/spa.h>
	#include <sys/zfs_fuid.h>
	#include <sys/dsl_dataset.h>

	/*
	* These zfs_log_* functions must be called within a dmu tx, in one
	* of 2 contexts depending on zilog->z_replay:
	*
	* Non replay mode
	* ---------------
	* We need to record the transaction so that if it is committed to
	* the Intent Log then it can be replayed. An intent log transaction
	* structure (itx_t) is allocated and all the information necessary to
	* possibly replay the transaction is saved in it. The itx is then assigned
	* a sequence number and inserted in the in-memory list anchored in the zilog.
	*
	* Replay mode
	* -----------
	* We need to mark the intent log record as replayed in the log header.
	* This is done in the same transaction as the replay so that they
	* commit atomically.
	*/

	int
	zfs_log_create_txtype(zil_create_t type, vsecattr_t vsecp, vattr_t vap)
	{
	int isxvattr = (vap->va_mask & ATTR_XVATTR);
	switch (type) {
	case Z_FILE:
	if (vsecp == NULL && !isxvattr)
	return (TX_CREATE);
	if (vsecp && isxvattr)
	return (TX_CREATE_ACL_ATTR);
	if (vsecp)
	return (TX_CREATE_ACL);
	else
	return (TX_CREATE_ATTR);
	case Z_DIR:
	if (vsecp == NULL && !isxvattr)
	return (TX_MKDIR);
	if (vsecp && isxvattr)
	return (TX_MKDIR_ACL_ATTR);
	if (vsecp)
	return (TX_MKDIR_ACL);
	else
	return (TX_MKDIR_ATTR);
	case Z_XATTRDIR:
	return (TX_MKXATTR);
	}
	ASSERT(0);
	return (TX_MAX_TYPE);
	}

	/*
	* build up the log data necessary for logging xvattr_t
	* First lr_attr_t is initialized. following the lr_attr_t
	* is the mapsize and attribute bitmap copied from the xvattr_t.
	* Following the bitmap and bitmapsize two 64 bit words are reserved
	* for the create time which may be set. Following the create time
	* records a single 64 bit integer which has the bits to set on
	* replay for the xvattr.
	*/
	static void
	zfs_log_xvattr(lr_attr_t lrattr, xvattr_t xvap)
	{
	xoptattr_t *xoap;

	xoap = xva_getxoptattr(xvap);
	ASSERT(xoap);

	lrattr->lr_attr_masksize = xvap->xva_mapsize;
	uint32_t *bitmap = &lrattr->lr_attr_bitmap;
	for (int i = 0; i != xvap->xva_mapsize; i++, bitmap++)
	*bitmap = xvap->xva_reqattrmap[i];

	lr_attr_end_t end = (lr_attr_end_t )bitmap;
	end->lr_attr_attrs = 0;
	end->lr_attr_crtime[0] = 0;
	end->lr_attr_crtime[1] = 0;
	memset(end->lr_attr_scanstamp, 0, AV_SCANSTAMP_SZ);

	if (XVA_ISSET_REQ(xvap, XAT_READONLY))
	end->lr_attr_attrs \|= (xoap->xoa_readonly == 0) ? 0 :
	XAT0_READONLY;
	if (XVA_ISSET_REQ(xvap, XAT_HIDDEN))
	end->lr_attr_attrs \|= (xoap->xoa_hidden == 0) ? 0 :
	XAT0_HIDDEN;
	if (XVA_ISSET_REQ(xvap, XAT_SYSTEM))
	end->lr_attr_attrs \|= (xoap->xoa_system == 0) ? 0 :
	XAT0_SYSTEM;
	if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE))
	end->lr_attr_attrs \|= (xoap->xoa_archive == 0) ? 0 :
	XAT0_ARCHIVE;
	if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE))
	end->lr_attr_attrs \|= (xoap->xoa_immutable == 0) ? 0 :
	XAT0_IMMUTABLE;
	if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK))
	end->lr_attr_attrs \|= (xoap->xoa_nounlink == 0) ? 0 :
	XAT0_NOUNLINK;
	if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY))
	end->lr_attr_attrs \|= (xoap->xoa_appendonly == 0) ? 0 :
	XAT0_APPENDONLY;
	if (XVA_ISSET_REQ(xvap, XAT_OPAQUE))
	end->lr_attr_attrs \|= (xoap->xoa_opaque == 0) ? 0 :
	XAT0_APPENDONLY;
	if (XVA_ISSET_REQ(xvap, XAT_NODUMP))
	end->lr_attr_attrs \|= (xoap->xoa_nodump == 0) ? 0 :
	XAT0_NODUMP;
	if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED))
	end->lr_attr_attrs \|= (xoap->xoa_av_quarantined == 0) ? 0 :
	XAT0_AV_QUARANTINED;
	if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED))
	end->lr_attr_attrs \|= (xoap->xoa_av_modified == 0) ? 0 :
	XAT0_AV_MODIFIED;
	if (XVA_ISSET_REQ(xvap, XAT_CREATETIME))
	ZFS_TIME_ENCODE(&xoap->xoa_createtime, end->lr_attr_crtime);
	if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
	ASSERT(!XVA_ISSET_REQ(xvap, XAT_PROJID));

	memcpy(end->lr_attr_scanstamp, xoap->xoa_av_scanstamp,
	AV_SCANSTAMP_SZ);
	} else if (XVA_ISSET_REQ(xvap, XAT_PROJID)) {
	/*
	* XAT_PROJID and XAT_AV_SCANSTAMP will never be valid
	* at the same time, so we can share the same space.
	*/
	memcpy(end->lr_attr_scanstamp, &xoap->xoa_projid,
	sizeof (uint64_t));
	}
	if (XVA_ISSET_REQ(xvap, XAT_REPARSE))
	end->lr_attr_attrs \|= (xoap->xoa_reparse == 0) ? 0 :
	XAT0_REPARSE;
	if (XVA_ISSET_REQ(xvap, XAT_OFFLINE))
	end->lr_attr_attrs \|= (xoap->xoa_offline == 0) ? 0 :
	XAT0_OFFLINE;
	if (XVA_ISSET_REQ(xvap, XAT_SPARSE))
	end->lr_attr_attrs \|= (xoap->xoa_sparse == 0) ? 0 :
	XAT0_SPARSE;
	if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT))
	end->lr_attr_attrs \|= (xoap->xoa_projinherit == 0) ? 0 :
	XAT0_PROJINHERIT;
	}

	static void *
	zfs_log_fuid_ids(zfs_fuid_info_t fuidp, void start)
	{
	zfs_fuid_t *zfuid;
	uint64_t *fuidloc = start;

	/* First copy in the ACE FUIDs */
	for (zfuid = list_head(&fuidp->z_fuids); zfuid;
	zfuid = list_next(&fuidp->z_fuids, zfuid)) {
	*fuidloc++ = zfuid->z_logfuid;
	}
	return (fuidloc);
	}


	static void *
	zfs_log_fuid_domains(zfs_fuid_info_t fuidp, void start)
	{
	zfs_fuid_domain_t *zdomain;

	/* now copy in the domain info, if any */
	if (fuidp->z_domain_str_sz != 0) {
	for (zdomain = list_head(&fuidp->z_domains); zdomain;
	zdomain = list_next(&fuidp->z_domains, zdomain)) {
	memcpy(start, zdomain->z_domain,
	strlen(zdomain->z_domain) + 1);
	start = (caddr_t)start +
	strlen(zdomain->z_domain) + 1;
	}
	}
	return (start);
	}

	/*
	* If zp is an xattr node, check whether the xattr owner is unlinked.
	* We don't want to log anything if the owner is unlinked.
	*/
	static int
	zfs_xattr_owner_unlinked(znode_t *zp)
	{
	int unlinked = 0;
	znode_t *dzp;
	#ifdef __FreeBSD__
	znode_t *tzp = zp;

	/*
	* zrele drops the vnode lock which violates the VOP locking contract
	* on FreeBSD. See comment at the top of zfs_replay.c for more detail.
	*/
	/*
	* if zp is XATTR node, keep walking up via z_xattr_parent until we
	* get the owner
	*/
	while (tzp->z_pflags & ZFS_XATTR) {
	ASSERT3U(zp->z_xattr_parent, !=, 0);
	if (zfs_zget(ZTOZSB(tzp), tzp->z_xattr_parent, &dzp) != 0) {
	unlinked = 1;
	break;
	}

	if (tzp != zp)
	zrele(tzp);
	tzp = dzp;
	unlinked = tzp->z_unlinked;
	}
	if (tzp != zp)
	zrele(tzp);
	#else
	zhold(zp);
	/*
	* if zp is XATTR node, keep walking up via z_xattr_parent until we
	* get the owner
	*/
	while (zp->z_pflags & ZFS_XATTR) {
	ASSERT3U(zp->z_xattr_parent, !=, 0);
	if (zfs_zget(ZTOZSB(zp), zp->z_xattr_parent, &dzp) != 0) {
	unlinked = 1;
	break;
	}

	zrele(zp);
	zp = dzp;
	unlinked = zp->z_unlinked;
	}
	zrele(zp);
	#endif
	return (unlinked);
	}

	/*
	* Handles TX_CREATE, TX_CREATE_ATTR, TX_MKDIR, TX_MKDIR_ATTR and
	* TK_MKXATTR transactions.
	*
	* TX_CREATE and TX_MKDIR are standard creates, but they may have FUID
	* domain information appended prior to the name. In this case the
	* uid/gid in the log record will be a log centric FUID.
	*
	* TX_CREATE_ACL_ATTR and TX_MKDIR_ACL_ATTR handle special creates that
	* may contain attributes, ACL and optional fuid information.
	*
	* TX_CREATE_ACL and TX_MKDIR_ACL handle special creates that specify
	* and ACL and normal users/groups in the ACEs.
	*
	* There may be an optional xvattr attribute information similar
	* to zfs_log_setattr.
	*
	* Also, after the file name "domain" strings may be appended.
	*/
	void
	zfs_log_create(zilog_t zilog, dmu_tx_t tx, uint64_t txtype,
	znode_t dzp, znode_t zp, const char name, vsecattr_t vsecp,
	zfs_fuid_info_t fuidp, vattr_t vap)
	{
	itx_t *itx;
	lr_create_t *lr;
	lr_acl_create_t *lracl;
	size_t aclsize = 0;
	size_t xvatsize = 0;
	size_t txsize;
	xvattr_t xvap = (xvattr_t )vap;
	void *end;
	size_t lrsize;
	size_t namesize = strlen(name) + 1;
	size_t fuidsz = 0;

	if (zil_replaying(zilog, tx) \|\| zfs_xattr_owner_unlinked(dzp))
	return;

	/*
	* If we have FUIDs present then add in space for
	* domains and ACE fuid's if any.
	*/
	if (fuidp) {
	fuidsz += fuidp->z_domain_str_sz;
	fuidsz += fuidp->z_fuid_cnt * sizeof (uint64_t);
	}

	if (vap->va_mask & ATTR_XVATTR)
	xvatsize = ZIL_XVAT_SIZE(xvap->xva_mapsize);

	if ((int)txtype == TX_CREATE_ATTR \|\| (int)txtype == TX_MKDIR_ATTR \|\|
	(int)txtype == TX_CREATE \|\| (int)txtype == TX_MKDIR \|\|
	(int)txtype == TX_MKXATTR) {
	txsize = sizeof (*lr) + namesize + fuidsz + xvatsize;
	lrsize = sizeof (*lr);
	} else {
	txsize =
	sizeof (lr_acl_create_t) + namesize + fuidsz +
	ZIL_ACE_LENGTH(aclsize) + xvatsize;
	lrsize = sizeof (lr_acl_create_t);
	}

	itx = zil_itx_create(txtype, txsize);

	lr = (lr_create_t *)&itx->itx_lr;
	lr->lr_doid = dzp->z_id;
	lr->lr_foid = zp->z_id;
	/* Store dnode slot count in 8 bits above object id. */
	LR_FOID_SET_SLOTS(lr->lr_foid, zp->z_dnodesize >> DNODE_SHIFT);
	lr->lr_mode = zp->z_mode;
	if (!IS_EPHEMERAL(KUID_TO_SUID(ZTOUID(zp)))) {
	lr->lr_uid = (uint64_t)KUID_TO_SUID(ZTOUID(zp));
	} else {
	lr->lr_uid = fuidp->z_fuid_owner;
	}
	if (!IS_EPHEMERAL(KGID_TO_SGID(ZTOGID(zp)))) {
	lr->lr_gid = (uint64_t)KGID_TO_SGID(ZTOGID(zp));
	} else {
	lr->lr_gid = fuidp->z_fuid_group;
	}
	(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(ZTOZSB(zp)), &lr->lr_gen,
	sizeof (uint64_t));
	(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)),
	lr->lr_crtime, sizeof (uint64_t) * 2);

	if (sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(ZTOZSB(zp)), &lr->lr_rdev,
	sizeof (lr->lr_rdev)) != 0)
	lr->lr_rdev = 0;

	/*
	* Fill in xvattr info if any
	*/
	if (vap->va_mask & ATTR_XVATTR) {
	zfs_log_xvattr((lr_attr_t *)((caddr_t)lr + lrsize), xvap);
	end = (caddr_t)lr + lrsize + xvatsize;
	} else {
	end = (caddr_t)lr + lrsize;
	}

	/* Now fill in any ACL info */

	if (vsecp) {
	lracl = (lr_acl_create_t *)&itx->itx_lr;
	lracl->lr_aclcnt = vsecp->vsa_aclcnt;
	lracl->lr_acl_bytes = aclsize;
	lracl->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
	lracl->lr_fuidcnt = fuidp ? fuidp->z_fuid_cnt : 0;
	if (vsecp->vsa_aclflags & VSA_ACE_ACLFLAGS)
	lracl->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
	else
	lracl->lr_acl_flags = 0;

	memcpy(end, vsecp->vsa_aclentp, aclsize);
	end = (caddr_t)end + ZIL_ACE_LENGTH(aclsize);
	}

	/* drop in FUID info */
	if (fuidp) {
	end = zfs_log_fuid_ids(fuidp, end);
	end = zfs_log_fuid_domains(fuidp, end);
	}
	/*
	* Now place file name in log record
	*/
	memcpy(end, name, namesize);

	zil_itx_assign(zilog, itx, tx);
	}

	/*
	* Handles both TX_REMOVE and TX_RMDIR transactions.
	*/
	void
	zfs_log_remove(zilog_t zilog, dmu_tx_t tx, uint64_t txtype,
	znode_t dzp, const char name, uint64_t foid, boolean_t unlinked)
	{
	itx_t *itx;
	lr_remove_t *lr;
	size_t namesize = strlen(name) + 1;

	if (zil_replaying(zilog, tx) \|\| zfs_xattr_owner_unlinked(dzp))
	return;

	itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
	lr = (lr_remove_t *)&itx->itx_lr;
	lr->lr_doid = dzp->z_id;
	memcpy(lr + 1, name, namesize);

	itx->itx_oid = foid;

	/*
	* Object ids can be re-instantiated in the next txg so
	* remove any async transactions to avoid future leaks.
	* This can happen if a fsync occurs on the re-instantiated
	* object for a WR_INDIRECT or WR_NEED_COPY write, which gets
	* the new file data and flushes a write record for the old object.
	*/
	if (unlinked) {
	ASSERT((txtype & ~TX_CI) == TX_REMOVE);
	zil_remove_async(zilog, foid);
	}
	zil_itx_assign(zilog, itx, tx);
	}

	/*
	* Handles TX_LINK transactions.
	*/
	void
	zfs_log_link(zilog_t zilog, dmu_tx_t tx, uint64_t txtype,
	znode_t dzp, znode_t zp, const char *name)
	{
	itx_t *itx;
	lr_link_t *lr;
	size_t namesize = strlen(name) + 1;

	if (zil_replaying(zilog, tx))
	return;

	itx = zil_itx_create(txtype, sizeof (*lr) + namesize);
	lr = (lr_link_t *)&itx->itx_lr;
	lr->lr_doid = dzp->z_id;
	lr->lr_link_obj = zp->z_id;
	memcpy(lr + 1, name, namesize);

	zil_itx_assign(zilog, itx, tx);
	}

	/*
	* Handles TX_SYMLINK transactions.
	*/
	void
	zfs_log_symlink(zilog_t zilog, dmu_tx_t tx, uint64_t txtype,
	znode_t dzp, znode_t zp, const char name, const char link)
	{
	itx_t *itx;
	lr_create_t *lr;
	size_t namesize = strlen(name) + 1;
	size_t linksize = strlen(link) + 1;

	if (zil_replaying(zilog, tx))
	return;

	itx = zil_itx_create(txtype, sizeof (*lr) + namesize + linksize);
	lr = (lr_create_t *)&itx->itx_lr;
	lr->lr_doid = dzp->z_id;
	lr->lr_foid = zp->z_id;
	lr->lr_uid = KUID_TO_SUID(ZTOUID(zp));
	lr->lr_gid = KGID_TO_SGID(ZTOGID(zp));
	lr->lr_mode = zp->z_mode;
	(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(ZTOZSB(zp)), &lr->lr_gen,
	sizeof (uint64_t));
	(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)),
	lr->lr_crtime, sizeof (uint64_t) * 2);
	memcpy((char *)(lr + 1), name, namesize);
	memcpy((char *)(lr + 1) + namesize, link, linksize);

	zil_itx_assign(zilog, itx, tx);
	}

	static void
	do_zfs_log_rename(zilog_t zilog, dmu_tx_t tx, uint64_t txtype, znode_t *sdzp,
	const char sname, znode_t tdzp, const char dname, znode_t szp)
	{
	itx_t *itx;
	lr_rename_t *lr;
	size_t snamesize = strlen(sname) + 1;
	size_t dnamesize = strlen(dname) + 1;

	if (zil_replaying(zilog, tx))
	return;

	itx = zil_itx_create(txtype, sizeof (*lr) + snamesize + dnamesize);
	lr = (lr_rename_t *)&itx->itx_lr;
	lr->lr_sdoid = sdzp->z_id;
	lr->lr_tdoid = tdzp->z_id;
	memcpy((char *)(lr + 1), sname, snamesize);
	memcpy((char *)(lr + 1) + snamesize, dname, dnamesize);
	itx->itx_oid = szp->z_id;

	zil_itx_assign(zilog, itx, tx);
	}

	/*
	* Handles TX_RENAME transactions.
	*/
	void
	zfs_log_rename(zilog_t zilog, dmu_tx_t tx, uint64_t txtype, znode_t *sdzp,
	const char sname, znode_t tdzp, const char dname, znode_t szp)
	{
	txtype \|= TX_RENAME;
	do_zfs_log_rename(zilog, tx, txtype, sdzp, sname, tdzp, dname, szp);
	}

	/*
	* Handles TX_RENAME_EXCHANGE transactions.
	*/
	void
	zfs_log_rename_exchange(zilog_t zilog, dmu_tx_t tx, uint64_t txtype,
	znode_t sdzp, const char sname, znode_t tdzp, const char dname,
	znode_t *szp)
	{
	txtype \|= TX_RENAME_EXCHANGE;
	do_zfs_log_rename(zilog, tx, txtype, sdzp, sname, tdzp, dname, szp);
	}

	/*
	* Handles TX_RENAME_WHITEOUT transactions.
	*
	* Unfortunately we cannot reuse do_zfs_log_rename because we we need to call
	* zfs_mknode() on replay which requires stashing bits as with TX_CREATE.
	*/
	void
	zfs_log_rename_whiteout(zilog_t zilog, dmu_tx_t tx, uint64_t txtype,
	znode_t sdzp, const char sname, znode_t tdzp, const char dname,
	znode_t szp, znode_t wzp)
	{
	itx_t *itx;
	lr_rename_whiteout_t *lr;
	size_t snamesize = strlen(sname) + 1;
	size_t dnamesize = strlen(dname) + 1;

	if (zil_replaying(zilog, tx))
	return;

	txtype \|= TX_RENAME_WHITEOUT;
	itx = zil_itx_create(txtype, sizeof (*lr) + snamesize + dnamesize);
	lr = (lr_rename_whiteout_t *)&itx->itx_lr;
	lr->lr_rename.lr_sdoid = sdzp->z_id;
	lr->lr_rename.lr_tdoid = tdzp->z_id;

	/*
	* RENAME_WHITEOUT will create an entry at the source znode, so we need
	* to store the same data that the equivalent call to zfs_log_create()
	* would.
	*/
	lr->lr_wfoid = wzp->z_id;
	LR_FOID_SET_SLOTS(lr->lr_wfoid, wzp->z_dnodesize >> DNODE_SHIFT);
	(void) sa_lookup(wzp->z_sa_hdl, SA_ZPL_GEN(ZTOZSB(wzp)), &lr->lr_wgen,
	sizeof (uint64_t));
	(void) sa_lookup(wzp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(wzp)),
	lr->lr_wcrtime, sizeof (uint64_t) * 2);
	lr->lr_wmode = wzp->z_mode;
	lr->lr_wuid = (uint64_t)KUID_TO_SUID(ZTOUID(wzp));
	lr->lr_wgid = (uint64_t)KGID_TO_SGID(ZTOGID(wzp));

	/*
	* This rdev will always be makdevice(0, 0) but because the ZIL log and
	* replay code needs to be platform independent (and there is no
	* platform independent makdev()) we need to copy the one created
	* during the rename operation.
	*/
	(void) sa_lookup(wzp->z_sa_hdl, SA_ZPL_RDEV(ZTOZSB(wzp)), &lr->lr_wrdev,
	sizeof (lr->lr_wrdev));

	memcpy((char *)(lr + 1), sname, snamesize);
	memcpy((char *)(lr + 1) + snamesize, dname, dnamesize);
	itx->itx_oid = szp->z_id;

	zil_itx_assign(zilog, itx, tx);
	}

	/*
	* zfs_log_write() handles TX_WRITE transactions. The specified callback is
	* called as soon as the write is on stable storage (be it via a DMU sync or a
	* ZIL commit).
	*/
	static int64_t zfs_immediate_write_sz = 32768;

	void
	zfs_log_write(zilog_t zilog, dmu_tx_t tx, int txtype,
	znode_t *zp, offset_t off, ssize_t resid, boolean_t commit,
	zil_callback_t callback, void *callback_data)
	{
	dmu_buf_impl_t db = (dmu_buf_impl_t )sa_get_db(zp->z_sa_hdl);
	uint32_t blocksize = zp->z_blksz;
	itx_wr_state_t write_state;
	uint64_t gen = 0;
	ssize_t size = resid;

	if (zil_replaying(zilog, tx) \|\| zp->z_unlinked \|\|
	zfs_xattr_owner_unlinked(zp)) {
	if (callback != NULL)
	callback(callback_data);
	return;
	}

	if (zilog->zl_logbias == ZFS_LOGBIAS_THROUGHPUT)
	write_state = WR_INDIRECT;
	else if (!spa_has_slogs(zilog->zl_spa) &&
	resid >= zfs_immediate_write_sz)
	write_state = WR_INDIRECT;
	else if (commit)
	write_state = WR_COPIED;
	else
	write_state = WR_NEED_COPY;

	(void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(ZTOZSB(zp)), &gen,
	sizeof (gen));

	while (resid) {
	itx_t *itx;
	lr_write_t *lr;
	itx_wr_state_t wr_state = write_state;
	ssize_t len = resid;

	/*
	* A WR_COPIED record must fit entirely in one log block.
	* Large writes can use WR_NEED_COPY, which the ZIL will
	* split into multiple records across several log blocks
	* if necessary.
	*/
	if (wr_state == WR_COPIED &&
	resid > zil_max_copied_data(zilog))
	wr_state = WR_NEED_COPY;
	else if (wr_state == WR_INDIRECT)
	len = MIN(blocksize - P2PHASE(off, blocksize), resid);

	itx = zil_itx_create(txtype, sizeof (*lr) +
	(wr_state == WR_COPIED ? len : 0));
	lr = (lr_write_t *)&itx->itx_lr;

	/*
	* For WR_COPIED records, copy the data into the lr_write_t.
	*/
	if (wr_state == WR_COPIED) {
	int err;
	DB_DNODE_ENTER(db);
	err = dmu_read_by_dnode(DB_DNODE(db), off, len, lr + 1,
	DMU_READ_NO_PREFETCH);
	if (err != 0) {
	zil_itx_destroy(itx);
	itx = zil_itx_create(txtype, sizeof (*lr));
	lr = (lr_write_t *)&itx->itx_lr;
	wr_state = WR_NEED_COPY;
	}
	DB_DNODE_EXIT(db);
	}

	itx->itx_wr_state = wr_state;
	lr->lr_foid = zp->z_id;
	lr->lr_offset = off;
	lr->lr_length = len;
	lr->lr_blkoff = 0;
	BP_ZERO(&lr->lr_blkptr);

	itx->itx_private = ZTOZSB(zp);
	itx->itx_sync = (zp->z_sync_cnt != 0);
	itx->itx_gen = gen;

	itx->itx_callback = callback;
	itx->itx_callback_data = callback_data;
	zil_itx_assign(zilog, itx, tx);

	off += len;
	resid -= len;
	}

	if (write_state == WR_COPIED \|\| write_state == WR_NEED_COPY) {
	dsl_pool_wrlog_count(zilog->zl_dmu_pool, size, tx->tx_txg);
	}
	}

	/*
	* Handles TX_TRUNCATE transactions.
	*/
	void
	zfs_log_truncate(zilog_t zilog, dmu_tx_t tx, int txtype,
	znode_t *zp, uint64_t off, uint64_t len)
	{
	itx_t *itx;
	lr_truncate_t *lr;

	if (zil_replaying(zilog, tx) \|\| zp->z_unlinked \|\|
	zfs_xattr_owner_unlinked(zp))
	return;

	itx = zil_itx_create(txtype, sizeof (*lr));
	lr = (lr_truncate_t *)&itx->itx_lr;
	lr->lr_foid = zp->z_id;
	lr->lr_offset = off;
	lr->lr_length = len;

	itx->itx_sync = (zp->z_sync_cnt != 0);
	zil_itx_assign(zilog, itx, tx);
	}

	/*
	* Handles TX_SETATTR transactions.
	*/
	void
	zfs_log_setattr(zilog_t zilog, dmu_tx_t tx, int txtype,
	znode_t zp, vattr_t vap, uint_t mask_applied, zfs_fuid_info_t *fuidp)
	{
	itx_t *itx;
	lr_setattr_t *lr;
	xvattr_t xvap = (xvattr_t )vap;
	size_t recsize = sizeof (lr_setattr_t);
	void *start;

	if (zil_replaying(zilog, tx) \|\| zp->z_unlinked)
	return;

	/*
	* If XVATTR set, then log record size needs to allow
	* for lr_attr_t + xvattr mask, mapsize and create time
	* plus actual attribute values
	*/
	if (vap->va_mask & ATTR_XVATTR)
	recsize = sizeof (*lr) + ZIL_XVAT_SIZE(xvap->xva_mapsize);

	if (fuidp)
	recsize += fuidp->z_domain_str_sz;

	itx = zil_itx_create(txtype, recsize);
	lr = (lr_setattr_t *)&itx->itx_lr;
	lr->lr_foid = zp->z_id;
	lr->lr_mask = (uint64_t)mask_applied;
	lr->lr_mode = (uint64_t)vap->va_mode;
	if ((mask_applied & ATTR_UID) && IS_EPHEMERAL(vap->va_uid))
	lr->lr_uid = fuidp->z_fuid_owner;
	else
	lr->lr_uid = (uint64_t)vap->va_uid;

	if ((mask_applied & ATTR_GID) && IS_EPHEMERAL(vap->va_gid))
	lr->lr_gid = fuidp->z_fuid_group;
	else
	lr->lr_gid = (uint64_t)vap->va_gid;

	lr->lr_size = (uint64_t)vap->va_size;
	ZFS_TIME_ENCODE(&vap->va_atime, lr->lr_atime);
	ZFS_TIME_ENCODE(&vap->va_mtime, lr->lr_mtime);
	start = (lr_setattr_t *)(lr + 1);
	if (vap->va_mask & ATTR_XVATTR) {
	zfs_log_xvattr((lr_attr_t *)start, xvap);
	start = (caddr_t)start + ZIL_XVAT_SIZE(xvap->xva_mapsize);
	}

	/*
	* Now stick on domain information if any on end
	*/

	if (fuidp)
	(void) zfs_log_fuid_domains(fuidp, start);

	itx->itx_sync = (zp->z_sync_cnt != 0);
	zil_itx_assign(zilog, itx, tx);
	}

	/*
	* Handles TX_SETSAXATTR transactions.
	*/
	void
	zfs_log_setsaxattr(zilog_t zilog, dmu_tx_t tx, int txtype,
	znode_t zp, const char name, const void *value, size_t size)
	{
	itx_t *itx;
	lr_setsaxattr_t *lr;
	size_t recsize = sizeof (lr_setsaxattr_t);
	void *xattrstart;
	int namelen;

	if (zil_replaying(zilog, tx) \|\| zp->z_unlinked)
	return;

	namelen = strlen(name) + 1;
	recsize += (namelen + size);
	itx = zil_itx_create(txtype, recsize);
	lr = (lr_setsaxattr_t *)&itx->itx_lr;
	lr->lr_foid = zp->z_id;
	xattrstart = (char *)(lr + 1);
	memcpy(xattrstart, name, namelen);
	if (value != NULL) {
	memcpy((char *)xattrstart + namelen, value, size);
	lr->lr_size = size;
	} else {
	lr->lr_size = 0;
	}

	itx->itx_sync = (zp->z_sync_cnt != 0);
	zil_itx_assign(zilog, itx, tx);
	}

	/*
	* Handles TX_ACL transactions.
	*/
	void
	zfs_log_acl(zilog_t zilog, dmu_tx_t tx, znode_t *zp,
	vsecattr_t vsecp, zfs_fuid_info_t fuidp)
	{
	itx_t *itx;
	lr_acl_v0_t *lrv0;
	lr_acl_t *lr;
	int txtype;
	int lrsize;
	size_t txsize;
	size_t aclbytes = vsecp->vsa_aclentsz;

	if (zil_replaying(zilog, tx) \|\| zp->z_unlinked)
	return;

	txtype = (ZTOZSB(zp)->z_version < ZPL_VERSION_FUID) ?
	TX_ACL_V0 : TX_ACL;

	if (txtype == TX_ACL)
	lrsize = sizeof (*lr);
	else
	lrsize = sizeof (*lrv0);

	txsize = lrsize +
	((txtype == TX_ACL) ? ZIL_ACE_LENGTH(aclbytes) : aclbytes) +
	(fuidp ? fuidp->z_domain_str_sz : 0) +
	sizeof (uint64_t) * (fuidp ? fuidp->z_fuid_cnt : 0);

	itx = zil_itx_create(txtype, txsize);

	lr = (lr_acl_t *)&itx->itx_lr;
	lr->lr_foid = zp->z_id;
	if (txtype == TX_ACL) {
	lr->lr_acl_bytes = aclbytes;
	lr->lr_domcnt = fuidp ? fuidp->z_domain_cnt : 0;
	lr->lr_fuidcnt = fuidp ? fuidp->z_fuid_cnt : 0;
	if (vsecp->vsa_mask & VSA_ACE_ACLFLAGS)
	lr->lr_acl_flags = (uint64_t)vsecp->vsa_aclflags;
	else
	lr->lr_acl_flags = 0;
	}
	lr->lr_aclcnt = (uint64_t)vsecp->vsa_aclcnt;

	if (txtype == TX_ACL_V0) {
	lrv0 = (lr_acl_v0_t *)lr;
	memcpy(lrv0 + 1, vsecp->vsa_aclentp, aclbytes);
	} else {
	void start = (ace_t )(lr + 1);

	memcpy(start, vsecp->vsa_aclentp, aclbytes);

	start = (caddr_t)start + ZIL_ACE_LENGTH(aclbytes);

	if (fuidp) {
	start = zfs_log_fuid_ids(fuidp, start);
	(void) zfs_log_fuid_domains(fuidp, start);
	}
	}

	itx->itx_sync = (zp->z_sync_cnt != 0);
	zil_itx_assign(zilog, itx, tx);
	}

	/*
	* Handles TX_CLONE_RANGE transactions.
	*/
	void
	zfs_log_clone_range(zilog_t zilog, dmu_tx_t tx, int txtype, znode_t *zp,
	uint64_t off, uint64_t len, uint64_t blksz, const blkptr_t *bps,
	size_t nbps)
	{
	itx_t *itx;
	lr_clone_range_t *lr;
	uint64_t partlen, max_log_data;
	- size_t i, partnbps;
	+ size_t partnbps;

	if (zil_replaying(zilog, tx) \|\| zp->z_unlinked)
	return;

	max_log_data = zil_max_log_data(zilog, sizeof (lr_clone_range_t));

	while (nbps > 0) {
	partnbps = MIN(nbps, max_log_data / sizeof (bps[0]));
	- partlen = 0;
	- for (i = 0; i < partnbps; i++) {
	- partlen += BP_GET_LSIZE(&bps[i]);
	- }
	+ partlen = partnbps * blksz;
	+ ASSERT3U(partlen, <, len + blksz);
	partlen = MIN(partlen, len);

	itx = zil_itx_create(txtype,
	sizeof (lr) + sizeof (bps[0]) partnbps);
	lr = (lr_clone_range_t *)&itx->itx_lr;
	lr->lr_foid = zp->z_id;
	lr->lr_offset = off;
	lr->lr_length = partlen;
	lr->lr_blksz = blksz;
	lr->lr_nbps = partnbps;
	memcpy(lr->lr_bps, bps, sizeof (bps[0]) * partnbps);

	itx->itx_sync = (zp->z_sync_cnt != 0);

	zil_itx_assign(zilog, itx, tx);

	bps += partnbps;
	ASSERT3U(nbps, >=, partnbps);
	nbps -= partnbps;
	off += partlen;
	ASSERT3U(len, >=, partlen);
	len -= partlen;
	}
	}

	ZFS_MODULE_PARAM(zfs, zfs_, immediate_write_sz, S64, ZMOD_RW,
	"Largest data block to write to zil");
	diff --git a/sys/contrib/openzfs/module/zfs/zfs_vnops.c b/sys/contrib/openzfs/module/zfs/zfs_vnops.c
	index babb07ca25a9..f3db953eab46 100644
	--- a/sys/contrib/openzfs/module/zfs/zfs_vnops.c
	+++ b/sys/contrib/openzfs/module/zfs/zfs_vnops.c
	@@ -1,1556 +1,1561 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2012, 2018 by Delphix. All rights reserved.
	* Copyright (c) 2015 by Chunwei Chen. All rights reserved.
	* Copyright 2017 Nexenta Systems, Inc.
	* Copyright (c) 2021, 2022 by Pawel Jakub Dawidek
	*/

	/* Portions Copyright 2007 Jeremy Teo */
	/* Portions Copyright 2010 Robert Milkowski */

	#include <sys/types.h>
	#include <sys/param.h>
	#include <sys/time.h>
	#include <sys/sysmacros.h>
	#include <sys/vfs.h>
	#include <sys/uio_impl.h>
	#include <sys/file.h>
	#include <sys/stat.h>
	#include <sys/kmem.h>
	#include <sys/cmn_err.h>
	#include <sys/errno.h>
	#include <sys/zfs_dir.h>
	#include <sys/zfs_acl.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/fs/zfs.h>
	#include <sys/dmu.h>
	#include <sys/dmu_objset.h>
	#include <sys/dsl_crypt.h>
	#include <sys/spa.h>
	#include <sys/txg.h>
	#include <sys/dbuf.h>
	#include <sys/policy.h>
	#include <sys/zfeature.h>
	#include <sys/zfs_vnops.h>
	#include <sys/zfs_quota.h>
	#include <sys/zfs_vfsops.h>
	#include <sys/zfs_znode.h>

	/*
	* Enable the experimental block cloning feature. If this setting is 0, then
	* even if feature@block_cloning is enabled, attempts to clone blocks will act
	* as though the feature is disabled.
	*/
	int zfs_bclone_enabled = 1;

	/*
	* When set zfs_clone_range() waits for dirty data to be written to disk.
	* This allows the clone operation to reliably succeed when a file is modified
	* and then immediately cloned. For small files this may be slower than making
	* a copy of the file and is therefore not the default. However, in certain
	* scenarios this behavior may be desirable so a tunable is provided.
	*/
	static int zfs_bclone_wait_dirty = 0;

	/*
	* Maximum bytes to read per chunk in zfs_read().
	*/
	static uint64_t zfs_vnops_read_chunk_size = 1024 * 1024;

	int
	zfs_fsync(znode_t zp, int syncflag, cred_t cr)
	{
	int error = 0;
	zfsvfs_t *zfsvfs = ZTOZSB(zp);

	if (zfsvfs->z_os->os_sync != ZFS_SYNC_DISABLED) {
	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (error);
	atomic_inc_32(&zp->z_sync_writes_cnt);
	zil_commit(zfsvfs->z_log, zp->z_id);
	atomic_dec_32(&zp->z_sync_writes_cnt);
	zfs_exit(zfsvfs, FTAG);
	}
	return (error);
	}


	#if defined(SEEK_HOLE) && defined(SEEK_DATA)
	/*
	* Lseek support for finding holes (cmd == SEEK_HOLE) and
	* data (cmd == SEEK_DATA). "off" is an in/out parameter.
	*/
	static int
	zfs_holey_common(znode_t zp, ulong_t cmd, loff_t off)
	{
	zfs_locked_range_t *lr;
	uint64_t noff = (uint64_t)off; / new offset */
	uint64_t file_sz;
	int error;
	boolean_t hole;

	file_sz = zp->z_size;
	if (noff >= file_sz) {
	return (SET_ERROR(ENXIO));
	}

	if (cmd == F_SEEK_HOLE)
	hole = B_TRUE;
	else
	hole = B_FALSE;

	/* Flush any mmap()'d data to disk */
	if (zn_has_cached_data(zp, 0, file_sz - 1))
	zn_flush_cached_data(zp, B_TRUE);

	lr = zfs_rangelock_enter(&zp->z_rangelock, 0, UINT64_MAX, RL_READER);
	error = dmu_offset_next(ZTOZSB(zp)->z_os, zp->z_id, hole, &noff);
	zfs_rangelock_exit(lr);

	if (error == ESRCH)
	return (SET_ERROR(ENXIO));

	/* File was dirty, so fall back to using generic logic */
	if (error == EBUSY) {
	if (hole)
	*off = file_sz;

	return (0);
	}

	/*
	* We could find a hole that begins after the logical end-of-file,
	* because dmu_offset_next() only works on whole blocks. If the
	* EOF falls mid-block, then indicate that the "virtual hole"
	* at the end of the file begins at the logical EOF, rather than
	* at the end of the last block.
	*/
	if (noff > file_sz) {
	ASSERT(hole);
	noff = file_sz;
	}

	if (noff < *off)
	return (error);
	*off = noff;
	return (error);
	}

	int
	zfs_holey(znode_t zp, ulong_t cmd, loff_t off)
	{
	zfsvfs_t *zfsvfs = ZTOZSB(zp);
	int error;

	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (error);

	error = zfs_holey_common(zp, cmd, off);

	zfs_exit(zfsvfs, FTAG);
	return (error);
	}
	#endif /* SEEK_HOLE && SEEK_DATA */

	int
	zfs_access(znode_t zp, int mode, int flag, cred_t cr)
	{
	zfsvfs_t *zfsvfs = ZTOZSB(zp);
	int error;

	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (error);

	if (flag & V_ACE_MASK)
	#if defined(__linux__)
	error = zfs_zaccess(zp, mode, flag, B_FALSE, cr,
	zfs_init_idmap);
	#else
	error = zfs_zaccess(zp, mode, flag, B_FALSE, cr,
	NULL);
	#endif
	else
	#if defined(__linux__)
	error = zfs_zaccess_rwx(zp, mode, flag, cr, zfs_init_idmap);
	#else
	error = zfs_zaccess_rwx(zp, mode, flag, cr, NULL);
	#endif

	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	/*
	* Read bytes from specified file into supplied buffer.
	*
	* IN: zp - inode of file to be read from.
	* uio - structure supplying read location, range info,
	* and return buffer.
	* ioflag - O_SYNC flags; used to provide FRSYNC semantics.
	* O_DIRECT flag; used to bypass page cache.
	* cr - credentials of caller.
	*
	* OUT: uio - updated offset and range, buffer filled.
	*
	* RETURN: 0 on success, error code on failure.
	*
	* Side Effects:
	* inode - atime updated if byte count > 0
	*/
	int
	zfs_read(struct znode zp, zfs_uio_t uio, int ioflag, cred_t *cr)
	{
	(void) cr;
	int error = 0;
	boolean_t frsync = B_FALSE;

	zfsvfs_t *zfsvfs = ZTOZSB(zp);
	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (error);

	if (zp->z_pflags & ZFS_AV_QUARANTINED) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EACCES));
	}

	/* We don't copy out anything useful for directories. */
	if (Z_ISDIR(ZTOTYPE(zp))) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EISDIR));
	}

	/*
	* Validate file offset
	*/
	if (zfs_uio_offset(uio) < (offset_t)0) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EINVAL));
	}

	/*
	* Fasttrack empty reads
	*/
	if (zfs_uio_resid(uio) == 0) {
	zfs_exit(zfsvfs, FTAG);
	return (0);
	}

	#ifdef FRSYNC
	/*
	* If we're in FRSYNC mode, sync out this znode before reading it.
	* Only do this for non-snapshots.
	*
	* Some platforms do not support FRSYNC and instead map it
	* to O_SYNC, which results in unnecessary calls to zil_commit. We
	* only honor FRSYNC requests on platforms which support it.
	*/
	frsync = !!(ioflag & FRSYNC);
	#endif
	if (zfsvfs->z_log &&
	(frsync \|\| zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS))
	zil_commit(zfsvfs->z_log, zp->z_id);

	/*
	* Lock the range against changes.
	*/
	zfs_locked_range_t *lr = zfs_rangelock_enter(&zp->z_rangelock,
	zfs_uio_offset(uio), zfs_uio_resid(uio), RL_READER);

	/*
	* If we are reading past end-of-file we can skip
	* to the end; but we might still need to set atime.
	*/
	if (zfs_uio_offset(uio) >= zp->z_size) {
	error = 0;
	goto out;
	}

	ASSERT(zfs_uio_offset(uio) < zp->z_size);
	#if defined(__linux__)
	ssize_t start_offset = zfs_uio_offset(uio);
	#endif
	ssize_t n = MIN(zfs_uio_resid(uio), zp->z_size - zfs_uio_offset(uio));
	ssize_t start_resid = n;

	while (n > 0) {
	ssize_t nbytes = MIN(n, zfs_vnops_read_chunk_size -
	P2PHASE(zfs_uio_offset(uio), zfs_vnops_read_chunk_size));
	#ifdef UIO_NOCOPY
	if (zfs_uio_segflg(uio) == UIO_NOCOPY)
	error = mappedread_sf(zp, nbytes, uio);
	else
	#endif
	if (zn_has_cached_data(zp, zfs_uio_offset(uio),
	zfs_uio_offset(uio) + nbytes - 1) && !(ioflag & O_DIRECT)) {
	error = mappedread(zp, nbytes, uio);
	} else {
	error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl),
	uio, nbytes);
	}

	if (error) {
	/* convert checksum errors into IO errors */
	if (error == ECKSUM)
	error = SET_ERROR(EIO);

	#if defined(__linux__)
	/*
	* if we actually read some bytes, bubbling EFAULT
	* up to become EAGAIN isn't what we want here...
	*
	* ...on Linux, at least. On FBSD, doing this breaks.
	*/
	if (error == EFAULT &&
	(zfs_uio_offset(uio) - start_offset) != 0)
	error = 0;
	#endif
	break;
	}

	n -= nbytes;
	}

	int64_t nread = start_resid - n;
	dataset_kstats_update_read_kstats(&zfsvfs->z_kstat, nread);
	task_io_account_read(nread);
	out:
	zfs_rangelock_exit(lr);

	ZFS_ACCESSTIME_STAMP(zfsvfs, zp);
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	static void
	zfs_clear_setid_bits_if_necessary(zfsvfs_t zfsvfs, znode_t zp, cred_t *cr,
	uint64_t clear_setid_bits_txgp, dmu_tx_t tx)
	{
	zilog_t *zilog = zfsvfs->z_log;
	const uint64_t uid = KUID_TO_SUID(ZTOUID(zp));

	ASSERT(clear_setid_bits_txgp != NULL);
	ASSERT(tx != NULL);

	/*
	* Clear Set-UID/Set-GID bits on successful write if not
	* privileged and at least one of the execute bits is set.
	*
	* It would be nice to do this after all writes have
	* been done, but that would still expose the ISUID/ISGID
	* to another app after the partial write is committed.
	*
	* Note: we don't call zfs_fuid_map_id() here because
	* user 0 is not an ephemeral uid.
	*/
	mutex_enter(&zp->z_acl_lock);
	if ((zp->z_mode & (S_IXUSR \| (S_IXUSR >> 3) \| (S_IXUSR >> 6))) != 0 &&
	(zp->z_mode & (S_ISUID \| S_ISGID)) != 0 &&
	secpolicy_vnode_setid_retain(zp, cr,
	((zp->z_mode & S_ISUID) != 0 && uid == 0)) != 0) {
	uint64_t newmode;

	zp->z_mode &= ~(S_ISUID \| S_ISGID);
	newmode = zp->z_mode;
	(void) sa_update(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs),
	(void *)&newmode, sizeof (uint64_t), tx);

	mutex_exit(&zp->z_acl_lock);

	/*
	* Make sure SUID/SGID bits will be removed when we replay the
	* log. If the setid bits are keep coming back, don't log more
	* than one TX_SETATTR per transaction group.
	*/
	if (*clear_setid_bits_txgp != dmu_tx_get_txg(tx)) {
	vattr_t va = {0};

	va.va_mask = ATTR_MODE;
	va.va_nodeid = zp->z_id;
	va.va_mode = newmode;
	zfs_log_setattr(zilog, tx, TX_SETATTR, zp, &va,
	ATTR_MODE, NULL);
	*clear_setid_bits_txgp = dmu_tx_get_txg(tx);
	}
	} else {
	mutex_exit(&zp->z_acl_lock);
	}
	}

	/*
	* Write the bytes to a file.
	*
	* IN: zp - znode of file to be written to.
	* uio - structure supplying write location, range info,
	* and data buffer.
	* ioflag - O_APPEND flag set if in append mode.
	* O_DIRECT flag; used to bypass page cache.
	* cr - credentials of caller.
	*
	* OUT: uio - updated offset and range.
	*
	* RETURN: 0 if success
	* error code if failure
	*
	* Timestamps:
	* ip - ctime\|mtime updated if byte count > 0
	*/
	int
	zfs_write(znode_t zp, zfs_uio_t uio, int ioflag, cred_t *cr)
	{
	int error = 0, error1;
	ssize_t start_resid = zfs_uio_resid(uio);
	uint64_t clear_setid_bits_txg = 0;

	/*
	* Fasttrack empty write
	*/
	ssize_t n = start_resid;
	if (n == 0)
	return (0);

	zfsvfs_t *zfsvfs = ZTOZSB(zp);
	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (error);

	sa_bulk_attr_t bulk[4];
	int count = 0;
	uint64_t mtime[2], ctime[2];
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
	&zp->z_size, 8);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
	&zp->z_pflags, 8);

	/*
	* Callers might not be able to detect properly that we are read-only,
	* so check it explicitly here.
	*/
	if (zfs_is_readonly(zfsvfs)) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EROFS));
	}

	/*
	* If immutable or not appending then return EPERM.
	* Intentionally allow ZFS_READONLY through here.
	* See zfs_zaccess_common()
	*/
	if ((zp->z_pflags & ZFS_IMMUTABLE) \|\|
	((zp->z_pflags & ZFS_APPENDONLY) && !(ioflag & O_APPEND) &&
	(zfs_uio_offset(uio) < zp->z_size))) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EPERM));
	}

	/*
	* Validate file offset
	*/
	offset_t woff = ioflag & O_APPEND ? zp->z_size : zfs_uio_offset(uio);
	if (woff < 0) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EINVAL));
	}

	/*
	* Pre-fault the pages to ensure slow (eg NFS) pages
	* don't hold up txg.
	*/
	ssize_t pfbytes = MIN(n, DMU_MAX_ACCESS >> 1);
	if (zfs_uio_prefaultpages(pfbytes, uio)) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EFAULT));
	}

	/*
	* If in append mode, set the io offset pointer to eof.
	*/
	zfs_locked_range_t *lr;
	if (ioflag & O_APPEND) {
	/*
	* Obtain an appending range lock to guarantee file append
	* semantics. We reset the write offset once we have the lock.
	*/
	lr = zfs_rangelock_enter(&zp->z_rangelock, 0, n, RL_APPEND);
	woff = lr->lr_offset;
	if (lr->lr_length == UINT64_MAX) {
	/*
	* We overlocked the file because this write will cause
	* the file block size to increase.
	* Note that zp_size cannot change with this lock held.
	*/
	woff = zp->z_size;
	}
	zfs_uio_setoffset(uio, woff);
	} else {
	/*
	* Note that if the file block size will change as a result of
	* this write, then this range lock will lock the entire file
	* so that we can re-write the block safely.
	*/
	lr = zfs_rangelock_enter(&zp->z_rangelock, woff, n, RL_WRITER);
	}

	if (zn_rlimit_fsize_uio(zp, uio)) {
	zfs_rangelock_exit(lr);
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EFBIG));
	}

	const rlim64_t limit = MAXOFFSET_T;

	if (woff >= limit) {
	zfs_rangelock_exit(lr);
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EFBIG));
	}

	if (n > limit - woff)
	n = limit - woff;

	uint64_t end_size = MAX(zp->z_size, woff + n);
	zilog_t *zilog = zfsvfs->z_log;
	boolean_t commit = (ioflag & (O_SYNC \| O_DSYNC)) \|\|
	(zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS);

	const uint64_t uid = KUID_TO_SUID(ZTOUID(zp));
	const uint64_t gid = KGID_TO_SGID(ZTOGID(zp));
	const uint64_t projid = zp->z_projid;

	/*
	* Write the file in reasonable size chunks. Each chunk is written
	* in a separate transaction; this keeps the intent log records small
	* and allows us to do more fine-grained space accounting.
	*/
	while (n > 0) {
	woff = zfs_uio_offset(uio);

	if (zfs_id_overblockquota(zfsvfs, DMU_USERUSED_OBJECT, uid) \|\|
	zfs_id_overblockquota(zfsvfs, DMU_GROUPUSED_OBJECT, gid) \|\|
	(projid != ZFS_DEFAULT_PROJID &&
	zfs_id_overblockquota(zfsvfs, DMU_PROJECTUSED_OBJECT,
	projid))) {
	error = SET_ERROR(EDQUOT);
	break;
	}

	uint64_t blksz;
	if (lr->lr_length == UINT64_MAX && zp->z_size <= zp->z_blksz) {
	if (zp->z_blksz > zfsvfs->z_max_blksz &&
	!ISP2(zp->z_blksz)) {
	/*
	* File's blocksize is already larger than the
	* "recordsize" property. Only let it grow to
	* the next power of 2.
	*/
	blksz = 1 << highbit64(zp->z_blksz);
	} else {
	blksz = zfsvfs->z_max_blksz;
	}
	blksz = MIN(blksz, P2ROUNDUP(end_size,
	SPA_MINBLOCKSIZE));
	blksz = MAX(blksz, zp->z_blksz);
	} else {
	blksz = zp->z_blksz;
	}

	arc_buf_t *abuf = NULL;
	ssize_t nbytes = n;
	if (n >= blksz && woff >= zp->z_size &&
	P2PHASE(woff, blksz) == 0 &&
	(blksz >= SPA_OLD_MAXBLOCKSIZE \|\| n < 4 * blksz)) {
	/*
	* This write covers a full block. "Borrow" a buffer
	* from the dmu so that we can fill it before we enter
	* a transaction. This avoids the possibility of
	* holding up the transaction if the data copy hangs
	* up on a pagefault (e.g., from an NFS server mapping).
	*/
	abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl),
	blksz);
	ASSERT(abuf != NULL);
	ASSERT(arc_buf_size(abuf) == blksz);
	if ((error = zfs_uiocopy(abuf->b_data, blksz,
	UIO_WRITE, uio, &nbytes))) {
	dmu_return_arcbuf(abuf);
	break;
	}
	ASSERT3S(nbytes, ==, blksz);
	} else {
	nbytes = MIN(n, (DMU_MAX_ACCESS >> 1) -
	P2PHASE(woff, blksz));
	if (pfbytes < nbytes) {
	if (zfs_uio_prefaultpages(nbytes, uio)) {
	error = SET_ERROR(EFAULT);
	break;
	}
	pfbytes = nbytes;
	}
	}

	/*
	* Start a transaction.
	*/
	dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os);
	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
	dmu_buf_impl_t db = (dmu_buf_impl_t )sa_get_db(zp->z_sa_hdl);
	DB_DNODE_ENTER(db);
	dmu_tx_hold_write_by_dnode(tx, DB_DNODE(db), woff, nbytes);
	DB_DNODE_EXIT(db);
	zfs_sa_upgrade_txholds(tx, zp);
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error) {
	dmu_tx_abort(tx);
	if (abuf != NULL)
	dmu_return_arcbuf(abuf);
	break;
	}

	/*
	* NB: We must call zfs_clear_setid_bits_if_necessary before
	* committing the transaction!
	*/

	/*
	* If rangelock_enter() over-locked we grow the blocksize
	* and then reduce the lock range. This will only happen
	* on the first iteration since rangelock_reduce() will
	* shrink down lr_length to the appropriate size.
	*/
	if (lr->lr_length == UINT64_MAX) {
	zfs_grow_blocksize(zp, blksz, tx);
	zfs_rangelock_reduce(lr, woff, n);
	}

	ssize_t tx_bytes;
	if (abuf == NULL) {
	tx_bytes = zfs_uio_resid(uio);
	zfs_uio_fault_disable(uio, B_TRUE);
	error = dmu_write_uio_dbuf(sa_get_db(zp->z_sa_hdl),
	uio, nbytes, tx);
	zfs_uio_fault_disable(uio, B_FALSE);
	#ifdef __linux__
	if (error == EFAULT) {
	zfs_clear_setid_bits_if_necessary(zfsvfs, zp,
	cr, &clear_setid_bits_txg, tx);
	dmu_tx_commit(tx);
	/*
	* Account for partial writes before
	* continuing the loop.
	* Update needs to occur before the next
	* zfs_uio_prefaultpages, or prefaultpages may
	* error, and we may break the loop early.
	*/
	n -= tx_bytes - zfs_uio_resid(uio);
	pfbytes -= tx_bytes - zfs_uio_resid(uio);
	continue;
	}
	#endif
	/*
	* On FreeBSD, EFAULT should be propagated back to the
	* VFS, which will handle faulting and will retry.
	*/
	if (error != 0 && error != EFAULT) {
	zfs_clear_setid_bits_if_necessary(zfsvfs, zp,
	cr, &clear_setid_bits_txg, tx);
	dmu_tx_commit(tx);
	break;
	}
	tx_bytes -= zfs_uio_resid(uio);
	} else {
	/*
	* Thus, we're writing a full block at a block-aligned
	* offset and extending the file past EOF.
	*
	* dmu_assign_arcbuf_by_dbuf() will directly assign the
	* arc buffer to a dbuf.
	*/
	error = dmu_assign_arcbuf_by_dbuf(
	sa_get_db(zp->z_sa_hdl), woff, abuf, tx);
	if (error != 0) {
	/*
	* XXX This might not be necessary if
	* dmu_assign_arcbuf_by_dbuf is guaranteed
	* to be atomic.
	*/
	zfs_clear_setid_bits_if_necessary(zfsvfs, zp,
	cr, &clear_setid_bits_txg, tx);
	dmu_return_arcbuf(abuf);
	dmu_tx_commit(tx);
	break;
	}
	ASSERT3S(nbytes, <=, zfs_uio_resid(uio));
	zfs_uioskip(uio, nbytes);
	tx_bytes = nbytes;
	}
	if (tx_bytes &&
	zn_has_cached_data(zp, woff, woff + tx_bytes - 1) &&
	!(ioflag & O_DIRECT)) {
	update_pages(zp, woff, tx_bytes, zfsvfs->z_os);
	}

	/*
	* If we made no progress, we're done. If we made even
	* partial progress, update the znode and ZIL accordingly.
	*/
	if (tx_bytes == 0) {
	(void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs),
	(void *)&zp->z_size, sizeof (uint64_t), tx);
	dmu_tx_commit(tx);
	ASSERT(error != 0);
	break;
	}

	zfs_clear_setid_bits_if_necessary(zfsvfs, zp, cr,
	&clear_setid_bits_txg, tx);

	zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime);

	/*
	* Update the file size (zp_size) if it has changed;
	* account for possible concurrent updates.
	*/
	while ((end_size = zp->z_size) < zfs_uio_offset(uio)) {
	(void) atomic_cas_64(&zp->z_size, end_size,
	zfs_uio_offset(uio));
	ASSERT(error == 0 \|\| error == EFAULT);
	}
	/*
	* If we are replaying and eof is non zero then force
	* the file size to the specified eof. Note, there's no
	* concurrency during replay.
	*/
	if (zfsvfs->z_replay && zfsvfs->z_replay_eof != 0)
	zp->z_size = zfsvfs->z_replay_eof;

	error1 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
	if (error1 != 0)
	/* Avoid clobbering EFAULT. */
	error = error1;

	/*
	* NB: During replay, the TX_SETATTR record logged by
	* zfs_clear_setid_bits_if_necessary must precede any of
	* the TX_WRITE records logged here.
	*/
	zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, commit,
	NULL, NULL);

	dmu_tx_commit(tx);

	if (error != 0)
	break;
	ASSERT3S(tx_bytes, ==, nbytes);
	n -= nbytes;
	pfbytes -= nbytes;
	}

	zfs_znode_update_vfs(zp);
	zfs_rangelock_exit(lr);

	/*
	* If we're in replay mode, or we made no progress, or the
	* uio data is inaccessible return an error. Otherwise, it's
	* at least a partial write, so it's successful.
	*/
	if (zfsvfs->z_replay \|\| zfs_uio_resid(uio) == start_resid \|\|
	error == EFAULT) {
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	if (commit)
	zil_commit(zilog, zp->z_id);

	const int64_t nwritten = start_resid - zfs_uio_resid(uio);
	dataset_kstats_update_write_kstats(&zfsvfs->z_kstat, nwritten);
	task_io_account_write(nwritten);

	zfs_exit(zfsvfs, FTAG);
	return (0);
	}

	int
	zfs_getsecattr(znode_t zp, vsecattr_t vsecp, int flag, cred_t *cr)
	{
	zfsvfs_t *zfsvfs = ZTOZSB(zp);
	int error;
	boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;

	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (error);
	error = zfs_getacl(zp, vsecp, skipaclchk, cr);
	zfs_exit(zfsvfs, FTAG);

	return (error);
	}

	int
	zfs_setsecattr(znode_t zp, vsecattr_t vsecp, int flag, cred_t *cr)
	{
	zfsvfs_t *zfsvfs = ZTOZSB(zp);
	int error;
	boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE;
	zilog_t *zilog;

	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (error);
	zilog = zfsvfs->z_log;
	error = zfs_setacl(zp, vsecp, skipaclchk, cr);

	if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
	zil_commit(zilog, 0);

	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	#ifdef ZFS_DEBUG
	static int zil_fault_io = 0;
	#endif

	static void zfs_get_done(zgd_t *zgd, int error);

	/*
	* Get data to generate a TX_WRITE intent log record.
	*/
	int
	zfs_get_data(void arg, uint64_t gen, lr_write_t lr, char *buf,
	struct lwb lwb, zio_t zio)
	{
	zfsvfs_t *zfsvfs = arg;
	objset_t *os = zfsvfs->z_os;
	znode_t *zp;
	uint64_t object = lr->lr_foid;
	uint64_t offset = lr->lr_offset;
	uint64_t size = lr->lr_length;
	dmu_buf_t *db;
	zgd_t *zgd;
	int error = 0;
	uint64_t zp_gen;

	ASSERT3P(lwb, !=, NULL);
	ASSERT3U(size, !=, 0);

	/*
	* Nothing to do if the file has been removed
	*/
	if (zfs_zget(zfsvfs, object, &zp) != 0)
	return (SET_ERROR(ENOENT));
	if (zp->z_unlinked) {
	/*
	* Release the vnode asynchronously as we currently have the
	* txg stopped from syncing.
	*/
	zfs_zrele_async(zp);
	return (SET_ERROR(ENOENT));
	}
	/* check if generation number matches */
	if (sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), &zp_gen,
	sizeof (zp_gen)) != 0) {
	zfs_zrele_async(zp);
	return (SET_ERROR(EIO));
	}
	if (zp_gen != gen) {
	zfs_zrele_async(zp);
	return (SET_ERROR(ENOENT));
	}

	zgd = kmem_zalloc(sizeof (zgd_t), KM_SLEEP);
	zgd->zgd_lwb = lwb;
	zgd->zgd_private = zp;

	/*
	* Write records come in two flavors: immediate and indirect.
	* For small writes it's cheaper to store the data with the
	* log record (immediate); for large writes it's cheaper to
	* sync the data and get a pointer to it (indirect) so that
	* we don't have to write the data twice.
	*/
	if (buf != NULL) { /* immediate write */
	zgd->zgd_lr = zfs_rangelock_enter(&zp->z_rangelock,
	offset, size, RL_READER);
	/* test for truncation needs to be done while range locked */
	if (offset >= zp->z_size) {
	error = SET_ERROR(ENOENT);
	} else {
	error = dmu_read(os, object, offset, size, buf,
	DMU_READ_NO_PREFETCH);
	}
	ASSERT(error == 0 \|\| error == ENOENT);
	} else { /* indirect write */
	ASSERT3P(zio, !=, NULL);
	/*
	* Have to lock the whole block to ensure when it's
	* written out and its checksum is being calculated
	* that no one can change the data. We need to re-check
	* blocksize after we get the lock in case it's changed!
	*/
	for (;;) {
	uint64_t blkoff;
	size = zp->z_blksz;
	blkoff = ISP2(size) ? P2PHASE(offset, size) : offset;
	offset -= blkoff;
	zgd->zgd_lr = zfs_rangelock_enter(&zp->z_rangelock,
	offset, size, RL_READER);
	if (zp->z_blksz == size)
	break;
	offset += blkoff;
	zfs_rangelock_exit(zgd->zgd_lr);
	}
	/* test for truncation needs to be done while range locked */
	if (lr->lr_offset >= zp->z_size)
	error = SET_ERROR(ENOENT);
	#ifdef ZFS_DEBUG
	if (zil_fault_io) {
	error = SET_ERROR(EIO);
	zil_fault_io = 0;
	}
	#endif
	if (error == 0)
	error = dmu_buf_hold_noread(os, object, offset, zgd,
	&db);

	if (error == 0) {
	blkptr_t *bp = &lr->lr_blkptr;

	zgd->zgd_db = db;
	zgd->zgd_bp = bp;

	ASSERT(db->db_offset == offset);
	ASSERT(db->db_size == size);

	error = dmu_sync(zio, lr->lr_common.lrc_txg,
	zfs_get_done, zgd);
	ASSERT(error \|\| lr->lr_length <= size);

	/*
	* On success, we need to wait for the write I/O
	* initiated by dmu_sync() to complete before we can
	* release this dbuf. We will finish everything up
	* in the zfs_get_done() callback.
	*/
	if (error == 0)
	return (0);

	if (error == EALREADY) {
	lr->lr_common.lrc_txtype = TX_WRITE2;
	/*
	* TX_WRITE2 relies on the data previously
	* written by the TX_WRITE that caused
	* EALREADY. We zero out the BP because
	* it is the old, currently-on-disk BP.
	*/
	zgd->zgd_bp = NULL;
	BP_ZERO(bp);
	error = 0;
	}
	}
	}

	zfs_get_done(zgd, error);

	return (error);
	}


	static void
	zfs_get_done(zgd_t *zgd, int error)
	{
	(void) error;
	znode_t *zp = zgd->zgd_private;

	if (zgd->zgd_db)
	dmu_buf_rele(zgd->zgd_db, zgd);

	zfs_rangelock_exit(zgd->zgd_lr);

	/*
	* Release the vnode asynchronously as we currently have the
	* txg stopped from syncing.
	*/
	zfs_zrele_async(zp);

	kmem_free(zgd, sizeof (zgd_t));
	}

	static int
	zfs_enter_two(zfsvfs_t zfsvfs1, zfsvfs_t zfsvfs2, const char *tag)
	{
	int error;

	/* Swap. Not sure if the order of zfs_enter()s is important. */
	if (zfsvfs1 > zfsvfs2) {
	zfsvfs_t *tmpzfsvfs;

	tmpzfsvfs = zfsvfs2;
	zfsvfs2 = zfsvfs1;
	zfsvfs1 = tmpzfsvfs;
	}

	error = zfs_enter(zfsvfs1, tag);
	if (error != 0)
	return (error);
	if (zfsvfs1 != zfsvfs2) {
	error = zfs_enter(zfsvfs2, tag);
	if (error != 0) {
	zfs_exit(zfsvfs1, tag);
	return (error);
	}
	}

	return (0);
	}

	static void
	zfs_exit_two(zfsvfs_t zfsvfs1, zfsvfs_t zfsvfs2, const char *tag)
	{

	zfs_exit(zfsvfs1, tag);
	if (zfsvfs1 != zfsvfs2)
	zfs_exit(zfsvfs2, tag);
	}

	/*
	* We split each clone request in chunks that can fit into a single ZIL
	* log entry. Each ZIL log entry can fit 130816 bytes for a block cloning
	* operation (see zil_max_log_data() and zfs_log_clone_range()). This gives
	* us room for storing 1022 block pointers.
	*
	* On success, the function return the number of bytes copied in *lenp.
	* Note, it doesn't return how much bytes are left to be copied.
	* On errors which are caused by any file system limitations or
	* brt limitations `EINVAL` is returned. In the most cases a user
	* requested bad parameters, it could be possible to clone the file but
	* some parameters don't match the requirements.
	*/
	int
	zfs_clone_range(znode_t inzp, uint64_t inoffp, znode_t *outzp,
	uint64_t outoffp, uint64_t lenp, cred_t *cr)
	{
	zfsvfs_t inzfsvfs, outzfsvfs;
	objset_t inos, outos;
	zfs_locked_range_t inlr, outlr;
	dmu_buf_impl_t *db;
	dmu_tx_t *tx;
	zilog_t *zilog;
	uint64_t inoff, outoff, len, done;
	uint64_t outsize, size;
	int error;
	int count = 0;
	sa_bulk_attr_t bulk[3];
	uint64_t mtime[2], ctime[2];
	uint64_t uid, gid, projid;
	blkptr_t *bps;
	size_t maxblocks, nbps;
	uint_t inblksz;
	uint64_t clear_setid_bits_txg = 0;
	uint64_t last_synced_txg = 0;

	inoff = *inoffp;
	outoff = *outoffp;
	len = *lenp;
	done = 0;

	inzfsvfs = ZTOZSB(inzp);
	outzfsvfs = ZTOZSB(outzp);

	/*
	* We need to call zfs_enter() potentially on two different datasets,
	* so we need a dedicated function for that.
	*/
	error = zfs_enter_two(inzfsvfs, outzfsvfs, FTAG);
	if (error != 0)
	return (error);

	inos = inzfsvfs->z_os;
	outos = outzfsvfs->z_os;

	/*
	* Both source and destination have to belong to the same storage pool.
	*/
	if (dmu_objset_spa(inos) != dmu_objset_spa(outos)) {
	zfs_exit_two(inzfsvfs, outzfsvfs, FTAG);
	return (SET_ERROR(EXDEV));
	}

	/*
	* outos and inos belongs to the same storage pool.
	* see a few lines above, only one check.
	*/
	if (!spa_feature_is_enabled(dmu_objset_spa(outos),
	SPA_FEATURE_BLOCK_CLONING)) {
	zfs_exit_two(inzfsvfs, outzfsvfs, FTAG);
	return (SET_ERROR(EOPNOTSUPP));
	}

	ASSERT(!outzfsvfs->z_replay);

	/*
	* Block cloning from an unencrypted dataset into an encrypted
	* dataset and vice versa is not supported.
	*/
	if (inos->os_encrypted != outos->os_encrypted) {
	zfs_exit_two(inzfsvfs, outzfsvfs, FTAG);
	return (SET_ERROR(EXDEV));
	}

	/*
	* Cloning across encrypted datasets is possible only if they
	* share the same master key.
	*/
	if (inos != outos && inos->os_encrypted &&
	!dmu_objset_crypto_key_equal(inos, outos)) {
	zfs_exit_two(inzfsvfs, outzfsvfs, FTAG);
	return (SET_ERROR(EXDEV));
	}

	error = zfs_verify_zp(inzp);
	if (error == 0)
	error = zfs_verify_zp(outzp);
	if (error != 0) {
	zfs_exit_two(inzfsvfs, outzfsvfs, FTAG);
	return (error);
	}

	/*
	* We don't copy source file's flags that's why we don't allow to clone
	* files that are in quarantine.
	*/
	if (inzp->z_pflags & ZFS_AV_QUARANTINED) {
	zfs_exit_two(inzfsvfs, outzfsvfs, FTAG);
	return (SET_ERROR(EACCES));
	}

	if (inoff >= inzp->z_size) {
	*lenp = 0;
	zfs_exit_two(inzfsvfs, outzfsvfs, FTAG);
	return (0);
	}
	if (len > inzp->z_size - inoff) {
	len = inzp->z_size - inoff;
	}
	if (len == 0) {
	*lenp = 0;
	zfs_exit_two(inzfsvfs, outzfsvfs, FTAG);
	return (0);
	}

	/*
	* Callers might not be able to detect properly that we are read-only,
	* so check it explicitly here.
	*/
	if (zfs_is_readonly(outzfsvfs)) {
	zfs_exit_two(inzfsvfs, outzfsvfs, FTAG);
	return (SET_ERROR(EROFS));
	}

	/*
	* If immutable or not appending then return EPERM.
	* Intentionally allow ZFS_READONLY through here.
	* See zfs_zaccess_common()
	*/
	if ((outzp->z_pflags & ZFS_IMMUTABLE) != 0) {
	zfs_exit_two(inzfsvfs, outzfsvfs, FTAG);
	return (SET_ERROR(EPERM));
	}

	/*
	* No overlapping if we are cloning within the same file.
	*/
	if (inzp == outzp) {
	if (inoff < outoff + len && outoff < inoff + len) {
	zfs_exit_two(inzfsvfs, outzfsvfs, FTAG);
	return (SET_ERROR(EINVAL));
	}
	}

	/* Flush any mmap()'d data to disk */
	if (zn_has_cached_data(inzp, inoff, inoff + len - 1))
	zn_flush_cached_data(inzp, B_TRUE);

	/*
	* Maintain predictable lock order.
	*/
	if (inzp < outzp \|\| (inzp == outzp && inoff < outoff)) {
	inlr = zfs_rangelock_enter(&inzp->z_rangelock, inoff, len,
	RL_READER);
	outlr = zfs_rangelock_enter(&outzp->z_rangelock, outoff, len,
	RL_WRITER);
	} else {
	outlr = zfs_rangelock_enter(&outzp->z_rangelock, outoff, len,
	RL_WRITER);
	inlr = zfs_rangelock_enter(&inzp->z_rangelock, inoff, len,
	RL_READER);
	}

	inblksz = inzp->z_blksz;

	/*
	* We cannot clone into a file with different block size if we can't
	* grow it (block size is already bigger, has more than one block, or
	* not locked for growth). There are other possible reasons for the
	* grow to fail, but we cover what we can before opening transaction
	* and the rest detect after we try to do it.
	*/
	if (inblksz < outzp->z_blksz) {
	error = SET_ERROR(EINVAL);
	goto unlock;
	}
	if (inblksz != outzp->z_blksz && (outzp->z_size > outzp->z_blksz \|\|
	outlr->lr_length != UINT64_MAX)) {
	error = SET_ERROR(EINVAL);
	goto unlock;
	}

	/*
	* Block size must be power-of-2 if destination offset != 0.
	* There can be no multiple blocks of non-power-of-2 size.
	*/
	if (outoff != 0 && !ISP2(inblksz)) {
	error = SET_ERROR(EINVAL);
	goto unlock;
	}

	/*
	* Offsets and len must be at block boundries.
	*/
	if ((inoff % inblksz) != 0 \|\| (outoff % inblksz) != 0) {
	error = SET_ERROR(EINVAL);
	goto unlock;
	}
	/*
	* Length must be multipe of blksz, except for the end of the file.
	*/
	if ((len % inblksz) != 0 &&
	(len < inzp->z_size - inoff \|\| len < outzp->z_size - outoff)) {
	error = SET_ERROR(EINVAL);
	goto unlock;
	}

	/*
	* If we are copying only one block and it is smaller than recordsize
	* property, do not allow destination to grow beyond one block if it
	* is not there yet. Otherwise the destination will get stuck with
	* that block size forever, that can be as small as 512 bytes, no
	* matter how big the destination grow later.
	*/
	if (len <= inblksz && inblksz < outzfsvfs->z_max_blksz &&
	outzp->z_size <= inblksz && outoff + len > inblksz) {
	error = SET_ERROR(EINVAL);
	goto unlock;
	}

	error = zn_rlimit_fsize(outoff + len);
	if (error != 0) {
	goto unlock;
	}

	if (inoff >= MAXOFFSET_T \|\| outoff >= MAXOFFSET_T) {
	error = SET_ERROR(EFBIG);
	goto unlock;
	}

	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(outzfsvfs), NULL,
	&mtime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(outzfsvfs), NULL,
	&ctime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(outzfsvfs), NULL,
	&outzp->z_size, 8);

	zilog = outzfsvfs->z_log;
	maxblocks = zil_max_log_data(zilog, sizeof (lr_clone_range_t)) /
	sizeof (bps[0]);

	uid = KUID_TO_SUID(ZTOUID(outzp));
	gid = KGID_TO_SGID(ZTOGID(outzp));
	projid = outzp->z_projid;

	bps = vmem_alloc(sizeof (bps[0]) * maxblocks, KM_SLEEP);

	/*
	* Clone the file in reasonable size chunks. Each chunk is cloned
	* in a separate transaction; this keeps the intent log records small
	* and allows us to do more fine-grained space accounting.
	*/
	while (len > 0) {
	size = MIN(inblksz * maxblocks, len);

	if (zfs_id_overblockquota(outzfsvfs, DMU_USERUSED_OBJECT,
	uid) \|\|
	zfs_id_overblockquota(outzfsvfs, DMU_GROUPUSED_OBJECT,
	gid) \|\|
	(projid != ZFS_DEFAULT_PROJID &&
	zfs_id_overblockquota(outzfsvfs, DMU_PROJECTUSED_OBJECT,
	projid))) {
	error = SET_ERROR(EDQUOT);
	break;
	}

	nbps = maxblocks;
	last_synced_txg = spa_last_synced_txg(dmu_objset_spa(inos));
	error = dmu_read_l0_bps(inos, inzp->z_id, inoff, size, bps,
	&nbps);
	if (error != 0) {
	/*
	* If we are trying to clone a block that was created
	* in the current transaction group, the error will be
	* EAGAIN here. Based on zfs_bclone_wait_dirty either
	* return a shortened range to the caller so it can
	* fallback, or wait for the next TXG and check again.
	*/
	if (error == EAGAIN && zfs_bclone_wait_dirty) {
	txg_wait_synced(dmu_objset_pool(inos),
	last_synced_txg + 1);
	continue;
	}

	break;
	}

	/*
	* Start a transaction.
	*/
	tx = dmu_tx_create(outos);
	dmu_tx_hold_sa(tx, outzp->z_sa_hdl, B_FALSE);
	db = (dmu_buf_impl_t *)sa_get_db(outzp->z_sa_hdl);
	DB_DNODE_ENTER(db);
	dmu_tx_hold_clone_by_dnode(tx, DB_DNODE(db), outoff, size);
	DB_DNODE_EXIT(db);
	zfs_sa_upgrade_txholds(tx, outzp);
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error != 0) {
	dmu_tx_abort(tx);
	break;
	}

	/*
	* Copy source znode's block size. This is done only if the
	* whole znode is locked (see zfs_rangelock_cb()) and only
	* on the first iteration since zfs_rangelock_reduce() will
	* shrink down lr_length to the appropriate size.
	*/
	if (outlr->lr_length == UINT64_MAX) {
	zfs_grow_blocksize(outzp, inblksz, tx);

	/*
	* Block growth may fail for many reasons we can not
	* predict here. If it happen the cloning is doomed.
	*/
	if (inblksz != outzp->z_blksz) {
	error = SET_ERROR(EINVAL);
	dmu_tx_abort(tx);
	break;
	}

	/*
	* Round range lock up to the block boundary, so we
	* prevent appends until we are done.
	*/
	zfs_rangelock_reduce(outlr, outoff,
	((len - 1) / inblksz + 1) * inblksz);
	}

	error = dmu_brt_clone(outos, outzp->z_id, outoff, size, tx,
	bps, nbps);
	if (error != 0) {
	dmu_tx_commit(tx);
	break;
	}

	if (zn_has_cached_data(outzp, outoff, outoff + size - 1)) {
	update_pages(outzp, outoff, size, outos);
	}

	zfs_clear_setid_bits_if_necessary(outzfsvfs, outzp, cr,
	&clear_setid_bits_txg, tx);

	zfs_tstamp_update_setup(outzp, CONTENT_MODIFIED, mtime, ctime);

	/*
	* Update the file size (zp_size) if it has changed;
	* account for possible concurrent updates.
	*/
	while ((outsize = outzp->z_size) < outoff + size) {
	(void) atomic_cas_64(&outzp->z_size, outsize,
	outoff + size);
	}

	error = sa_bulk_update(outzp->z_sa_hdl, bulk, count, tx);

	zfs_log_clone_range(zilog, tx, TX_CLONE_RANGE, outzp, outoff,
	size, inblksz, bps, nbps);

	dmu_tx_commit(tx);

	if (error != 0)
	break;

	inoff += size;
	outoff += size;
	len -= size;
	done += size;
	+
	+ if (issig()) {
	+ error = SET_ERROR(EINTR);
	+ break;
	+ }
	}

	vmem_free(bps, sizeof (bps[0]) * maxblocks);
	zfs_znode_update_vfs(outzp);

	unlock:
	zfs_rangelock_exit(outlr);
	zfs_rangelock_exit(inlr);

	if (done > 0) {
	/*
	* If we have made at least partial progress, reset the error.
	*/
	error = 0;

	ZFS_ACCESSTIME_STAMP(inzfsvfs, inzp);

	if (outos->os_sync == ZFS_SYNC_ALWAYS) {
	zil_commit(zilog, outzp->z_id);
	}

	*inoffp += done;
	*outoffp += done;
	*lenp = done;
	} else {
	/*
	* If we made no progress, there must be a good reason.
	* EOF is handled explicitly above, before the loop.
	*/
	ASSERT3S(error, !=, 0);
	}

	zfs_exit_two(inzfsvfs, outzfsvfs, FTAG);

	return (error);
	}

	/*
	* Usual pattern would be to call zfs_clone_range() from zfs_replay_clone(),
	* but we cannot do that, because when replaying we don't have source znode
	* available. This is why we need a dedicated replay function.
	*/
	int
	zfs_clone_range_replay(znode_t *zp, uint64_t off, uint64_t len, uint64_t blksz,
	const blkptr_t *bps, size_t nbps)
	{
	zfsvfs_t *zfsvfs;
	dmu_buf_impl_t *db;
	dmu_tx_t *tx;
	int error;
	int count = 0;
	sa_bulk_attr_t bulk[3];
	uint64_t mtime[2], ctime[2];

	ASSERT3U(off, <, MAXOFFSET_T);
	ASSERT3U(len, >, 0);
	ASSERT3U(nbps, >, 0);

	zfsvfs = ZTOZSB(zp);

	ASSERT(spa_feature_is_enabled(dmu_objset_spa(zfsvfs->z_os),
	SPA_FEATURE_BLOCK_CLONING));

	if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
	return (error);

	ASSERT(zfsvfs->z_replay);
	ASSERT(!zfs_is_readonly(zfsvfs));

	if ((off % blksz) != 0) {
	zfs_exit(zfsvfs, FTAG);
	return (SET_ERROR(EINVAL));
	}

	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16);
	SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
	&zp->z_size, 8);

	/*
	* Start a transaction.
	*/
	tx = dmu_tx_create(zfsvfs->z_os);

	dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
	db = (dmu_buf_impl_t *)sa_get_db(zp->z_sa_hdl);
	DB_DNODE_ENTER(db);
	dmu_tx_hold_clone_by_dnode(tx, DB_DNODE(db), off, len);
	DB_DNODE_EXIT(db);
	zfs_sa_upgrade_txholds(tx, zp);
	error = dmu_tx_assign(tx, TXG_WAIT);
	if (error != 0) {
	dmu_tx_abort(tx);
	zfs_exit(zfsvfs, FTAG);
	return (error);
	}

	if (zp->z_blksz < blksz)
	zfs_grow_blocksize(zp, blksz, tx);

	dmu_brt_clone(zfsvfs->z_os, zp->z_id, off, len, tx, bps, nbps);

	zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime);

	if (zp->z_size < off + len)
	zp->z_size = off + len;

	error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);

	/*
	* zil_replaying() not only check if we are replaying ZIL, but also
	* updates the ZIL header to record replay progress.
	*/
	VERIFY(zil_replaying(zfsvfs->z_log, tx));

	dmu_tx_commit(tx);

	zfs_znode_update_vfs(zp);

	zfs_exit(zfsvfs, FTAG);

	return (error);
	}

	EXPORT_SYMBOL(zfs_access);
	EXPORT_SYMBOL(zfs_fsync);
	EXPORT_SYMBOL(zfs_holey);
	EXPORT_SYMBOL(zfs_read);
	EXPORT_SYMBOL(zfs_write);
	EXPORT_SYMBOL(zfs_getsecattr);
	EXPORT_SYMBOL(zfs_setsecattr);
	EXPORT_SYMBOL(zfs_clone_range);
	EXPORT_SYMBOL(zfs_clone_range_replay);

	ZFS_MODULE_PARAM(zfs_vnops, zfs_vnops_, read_chunk_size, U64, ZMOD_RW,
	"Bytes to read per chunk");

	ZFS_MODULE_PARAM(zfs, zfs_, bclone_enabled, INT, ZMOD_RW,
	"Enable block cloning");

	ZFS_MODULE_PARAM(zfs, zfs_, bclone_wait_dirty, INT, ZMOD_RW,
	"Wait for dirty blocks when cloning");
	diff --git a/sys/contrib/openzfs/module/zfs/zio.c b/sys/contrib/openzfs/module/zfs/zio.c
	index 870343bf4fa3..d68d5ababe79 100644
	--- a/sys/contrib/openzfs/module/zfs/zio.c
	+++ b/sys/contrib/openzfs/module/zfs/zio.c
	@@ -1,5271 +1,5264 @@
	/*
	* CDDL HEADER START
	*
	* The contents of this file are subject to the terms of the
	* Common Development and Distribution License (the "License").
	* You may not use this file except in compliance with the License.
	*
	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	* or https://opensource.org/licenses/CDDL-1.0.
	* See the License for the specific language governing permissions
	* and limitations under the License.
	*
	* When distributing Covered Code, include this CDDL HEADER in each
	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	* If applicable, add the following below this CDDL HEADER, with the
	* fields enclosed by brackets "[]" replaced with your own identifying
	* information: Portions Copyright [yyyy] [name of copyright owner]
	*
	* CDDL HEADER END
	*/
	/*
	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
	* Copyright (c) 2011, 2022 by Delphix. All rights reserved.
	* Copyright (c) 2011 Nexenta Systems, Inc. All rights reserved.
	* Copyright (c) 2017, Intel Corporation.
	* Copyright (c) 2019, 2023, 2024, Klara Inc.
	* Copyright (c) 2019, Allan Jude
	* Copyright (c) 2021, Datto, Inc.
	*/

	#include <sys/sysmacros.h>
	#include <sys/zfs_context.h>
	#include <sys/fm/fs/zfs.h>
	#include <sys/spa.h>
	#include <sys/txg.h>
	#include <sys/spa_impl.h>
	#include <sys/vdev_impl.h>
	#include <sys/vdev_trim.h>
	#include <sys/zio_impl.h>
	#include <sys/zio_compress.h>
	#include <sys/zio_checksum.h>
	#include <sys/dmu_objset.h>
	#include <sys/arc.h>
	#include <sys/brt.h>
	#include <sys/ddt.h>
	#include <sys/blkptr.h>
	#include <sys/zfeature.h>
	#include <sys/dsl_scan.h>
	#include <sys/metaslab_impl.h>
	#include <sys/time.h>
	#include <sys/trace_zfs.h>
	#include <sys/abd.h>
	#include <sys/dsl_crypt.h>
	#include <cityhash.h>

	/*
	* ==========================================================================
	* I/O type descriptions
	* ==========================================================================
	*/
	const char *const zio_type_name[ZIO_TYPES] = {
	/*
	* Note: Linux kernel thread name length is limited
	* so these names will differ from upstream open zfs.
	*/
	"z_null", "z_rd", "z_wr", "z_fr", "z_cl", "z_flush", "z_trim"
	};

	int zio_dva_throttle_enabled = B_TRUE;
	static int zio_deadman_log_all = B_FALSE;

	/*
	* ==========================================================================
	* I/O kmem caches
	* ==========================================================================
	*/
	static kmem_cache_t *zio_cache;
	static kmem_cache_t *zio_link_cache;
	kmem_cache_t *zio_buf_cache[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT];
	kmem_cache_t *zio_data_buf_cache[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT];
	#if defined(ZFS_DEBUG) && !defined(_KERNEL)
	static uint64_t zio_buf_cache_allocs[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT];
	static uint64_t zio_buf_cache_frees[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT];
	#endif

	/* Mark IOs as "slow" if they take longer than 30 seconds */
	static uint_t zio_slow_io_ms = (30 * MILLISEC);

	#define BP_SPANB(indblkshift, level) \
	(((uint64_t)1) << ((level) * ((indblkshift) - SPA_BLKPTRSHIFT)))
	#define COMPARE_META_LEVEL 0x80000000ul
	/*
	* The following actions directly effect the spa's sync-to-convergence logic.
	* The values below define the sync pass when we start performing the action.
	* Care should be taken when changing these values as they directly impact
	* spa_sync() performance. Tuning these values may introduce subtle performance
	* pathologies and should only be done in the context of performance analysis.
	* These tunables will eventually be removed and replaced with #defines once
	* enough analysis has been done to determine optimal values.
	*
	* The 'zfs_sync_pass_deferred_free' pass must be greater than 1 to ensure that
	* regular blocks are not deferred.
	*
	* Starting in sync pass 8 (zfs_sync_pass_dont_compress), we disable
	* compression (including of metadata). In practice, we don't have this
	* many sync passes, so this has no effect.
	*
	* The original intent was that disabling compression would help the sync
	* passes to converge. However, in practice disabling compression increases
	* the average number of sync passes, because when we turn compression off, a
	* lot of block's size will change and thus we have to re-allocate (not
	* overwrite) them. It also increases the number of 128KB allocations (e.g.
	* for indirect blocks and spacemaps) because these will not be compressed.
	* The 128K allocations are especially detrimental to performance on highly
	* fragmented systems, which may have very few free segments of this size,
	* and may need to load new metaslabs to satisfy 128K allocations.
	*/

	/* defer frees starting in this pass */
	uint_t zfs_sync_pass_deferred_free = 2;

	/* don't compress starting in this pass */
	static uint_t zfs_sync_pass_dont_compress = 8;

	/* rewrite new bps starting in this pass */
	static uint_t zfs_sync_pass_rewrite = 2;

	/*
	* An allocating zio is one that either currently has the DVA allocate
	* stage set or will have it later in its lifetime.
	*/
	#define IO_IS_ALLOCATING(zio) ((zio)->io_orig_pipeline & ZIO_STAGE_DVA_ALLOCATE)

	/*
	* Enable smaller cores by excluding metadata
	* allocations as well.
	*/
	int zio_exclude_metadata = 0;
	static int zio_requeue_io_start_cut_in_line = 1;

	#ifdef ZFS_DEBUG
	static const int zio_buf_debug_limit = 16384;
	#else
	static const int zio_buf_debug_limit = 0;
	#endif

	static inline void __zio_execute(zio_t *zio);

	static void zio_taskq_dispatch(zio_t *, zio_taskq_type_t, boolean_t);

	void
	zio_init(void)
	{
	size_t c;

	zio_cache = kmem_cache_create("zio_cache",
	sizeof (zio_t), 0, NULL, NULL, NULL, NULL, NULL, 0);
	zio_link_cache = kmem_cache_create("zio_link_cache",
	sizeof (zio_link_t), 0, NULL, NULL, NULL, NULL, NULL, 0);

	for (c = 0; c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; c++) {
	size_t size = (c + 1) << SPA_MINBLOCKSHIFT;
	size_t align, cflags, data_cflags;
	char name[32];

	/*
	* Create cache for each half-power of 2 size, starting from
	* SPA_MINBLOCKSIZE. It should give us memory space efficiency
	* of ~7/8, sufficient for transient allocations mostly using
	* these caches.
	*/
	size_t p2 = size;
	while (!ISP2(p2))
	p2 &= p2 - 1;
	if (!IS_P2ALIGNED(size, p2 / 2))
	continue;

	#ifndef _KERNEL
	/*
	* If we are using watchpoints, put each buffer on its own page,
	* to eliminate the performance overhead of trapping to the
	* kernel when modifying a non-watched buffer that shares the
	* page with a watched buffer.
	*/
	if (arc_watch && !IS_P2ALIGNED(size, PAGESIZE))
	continue;
	#endif

	if (IS_P2ALIGNED(size, PAGESIZE))
	align = PAGESIZE;
	else
	align = 1 << (highbit64(size ^ (size - 1)) - 1);

	cflags = (zio_exclude_metadata \|\| size > zio_buf_debug_limit) ?
	KMC_NODEBUG : 0;
	data_cflags = KMC_NODEBUG;
	if (cflags == data_cflags) {
	/*
	* Resulting kmem caches would be identical.
	* Save memory by creating only one.
	*/
	(void) snprintf(name, sizeof (name),
	"zio_buf_comb_%lu", (ulong_t)size);
	zio_buf_cache[c] = kmem_cache_create(name, size, align,
	NULL, NULL, NULL, NULL, NULL, cflags);
	zio_data_buf_cache[c] = zio_buf_cache[c];
	continue;
	}
	(void) snprintf(name, sizeof (name), "zio_buf_%lu",
	(ulong_t)size);
	zio_buf_cache[c] = kmem_cache_create(name, size, align,
	NULL, NULL, NULL, NULL, NULL, cflags);

	(void) snprintf(name, sizeof (name), "zio_data_buf_%lu",
	(ulong_t)size);
	zio_data_buf_cache[c] = kmem_cache_create(name, size, align,
	NULL, NULL, NULL, NULL, NULL, data_cflags);
	}

	while (--c != 0) {
	ASSERT(zio_buf_cache[c] != NULL);
	if (zio_buf_cache[c - 1] == NULL)
	zio_buf_cache[c - 1] = zio_buf_cache[c];

	ASSERT(zio_data_buf_cache[c] != NULL);
	if (zio_data_buf_cache[c - 1] == NULL)
	zio_data_buf_cache[c - 1] = zio_data_buf_cache[c];
	}

	zio_inject_init();

	lz4_init();
	}

	void
	zio_fini(void)
	{
	size_t n = SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT;

	#if defined(ZFS_DEBUG) && !defined(_KERNEL)
	for (size_t i = 0; i < n; i++) {
	if (zio_buf_cache_allocs[i] != zio_buf_cache_frees[i])
	(void) printf("zio_fini: [%d] %llu != %llu\n",
	(int)((i + 1) << SPA_MINBLOCKSHIFT),
	(long long unsigned)zio_buf_cache_allocs[i],
	(long long unsigned)zio_buf_cache_frees[i]);
	}
	#endif

	/*
	* The same kmem cache can show up multiple times in both zio_buf_cache
	* and zio_data_buf_cache. Do a wasteful but trivially correct scan to
	* sort it out.
	*/
	for (size_t i = 0; i < n; i++) {
	kmem_cache_t *cache = zio_buf_cache[i];
	if (cache == NULL)
	continue;
	for (size_t j = i; j < n; j++) {
	if (cache == zio_buf_cache[j])
	zio_buf_cache[j] = NULL;
	if (cache == zio_data_buf_cache[j])
	zio_data_buf_cache[j] = NULL;
	}
	kmem_cache_destroy(cache);
	}

	for (size_t i = 0; i < n; i++) {
	kmem_cache_t *cache = zio_data_buf_cache[i];
	if (cache == NULL)
	continue;
	for (size_t j = i; j < n; j++) {
	if (cache == zio_data_buf_cache[j])
	zio_data_buf_cache[j] = NULL;
	}
	kmem_cache_destroy(cache);
	}

	for (size_t i = 0; i < n; i++) {
	VERIFY3P(zio_buf_cache[i], ==, NULL);
	VERIFY3P(zio_data_buf_cache[i], ==, NULL);
	}

	kmem_cache_destroy(zio_link_cache);
	kmem_cache_destroy(zio_cache);

	zio_inject_fini();

	lz4_fini();
	}

	/*
	* ==========================================================================
	* Allocate and free I/O buffers
	* ==========================================================================
	*/

	#ifdef ZFS_DEBUG
	static const ulong_t zio_buf_canary = (ulong_t)0xdeadc0dedead210b;
	#endif

	/*
	* Use empty space after the buffer to detect overflows.
	*
	* Since zio_init() creates kmem caches only for certain set of buffer sizes,
	* allocations of different sizes may have some unused space after the data.
	* Filling part of that space with a known pattern on allocation and checking
	* it on free should allow us to detect some buffer overflows.
	*/
	static void
	zio_buf_put_canary(ulong_t p, size_t size, kmem_cache_t *cache, size_t c)
	{
	#ifdef ZFS_DEBUG
	size_t off = P2ROUNDUP(size, sizeof (ulong_t));
	ulong_t *canary = p + off / sizeof (ulong_t);
	size_t asize = (c + 1) << SPA_MINBLOCKSHIFT;
	if (c + 1 < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT &&
	cache[c] == cache[c + 1])
	asize = (c + 2) << SPA_MINBLOCKSHIFT;
	for (; off < asize; canary++, off += sizeof (ulong_t))
	*canary = zio_buf_canary;
	#endif
	}

	static void
	zio_buf_check_canary(ulong_t p, size_t size, kmem_cache_t *cache, size_t c)
	{
	#ifdef ZFS_DEBUG
	size_t off = P2ROUNDUP(size, sizeof (ulong_t));
	ulong_t *canary = p + off / sizeof (ulong_t);
	size_t asize = (c + 1) << SPA_MINBLOCKSHIFT;
	if (c + 1 < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT &&
	cache[c] == cache[c + 1])
	asize = (c + 2) << SPA_MINBLOCKSHIFT;
	for (; off < asize; canary++, off += sizeof (ulong_t)) {
	if (unlikely(*canary != zio_buf_canary)) {
	PANIC("ZIO buffer overflow %p (%zu) + %zu %#lx != %#lx",
	p, size, (canary - p) * sizeof (ulong_t),
	*canary, zio_buf_canary);
	}
	}
	#endif
	}

	/*
	* Use zio_buf_alloc to allocate ZFS metadata. This data will appear in a
	* crashdump if the kernel panics, so use it judiciously. Obviously, it's
	* useful to inspect ZFS metadata, but if possible, we should avoid keeping
	* excess / transient data in-core during a crashdump.
	*/
	void *
	zio_buf_alloc(size_t size)
	{
	size_t c = (size - 1) >> SPA_MINBLOCKSHIFT;

	VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);
	#if defined(ZFS_DEBUG) && !defined(_KERNEL)
	atomic_add_64(&zio_buf_cache_allocs[c], 1);
	#endif

	void *p = kmem_cache_alloc(zio_buf_cache[c], KM_PUSHPAGE);
	zio_buf_put_canary(p, size, zio_buf_cache, c);
	return (p);
	}

	/*
	* Use zio_data_buf_alloc to allocate data. The data will not appear in a
	* crashdump if the kernel panics. This exists so that we will limit the amount
	* of ZFS data that shows up in a kernel crashdump. (Thus reducing the amount
	* of kernel heap dumped to disk when the kernel panics)
	*/
	void *
	zio_data_buf_alloc(size_t size)
	{
	size_t c = (size - 1) >> SPA_MINBLOCKSHIFT;

	VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);

	void *p = kmem_cache_alloc(zio_data_buf_cache[c], KM_PUSHPAGE);
	zio_buf_put_canary(p, size, zio_data_buf_cache, c);
	return (p);
	}

	void
	zio_buf_free(void *buf, size_t size)
	{
	size_t c = (size - 1) >> SPA_MINBLOCKSHIFT;

	VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);
	#if defined(ZFS_DEBUG) && !defined(_KERNEL)
	atomic_add_64(&zio_buf_cache_frees[c], 1);
	#endif

	zio_buf_check_canary(buf, size, zio_buf_cache, c);
	kmem_cache_free(zio_buf_cache[c], buf);
	}

	void
	zio_data_buf_free(void *buf, size_t size)
	{
	size_t c = (size - 1) >> SPA_MINBLOCKSHIFT;

	VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);

	zio_buf_check_canary(buf, size, zio_data_buf_cache, c);
	kmem_cache_free(zio_data_buf_cache[c], buf);
	}

	static void
	zio_abd_free(void *abd, size_t size)
	{
	(void) size;
	abd_free((abd_t *)abd);
	}

	/*
	* ==========================================================================
	* Push and pop I/O transform buffers
	* ==========================================================================
	*/
	void
	zio_push_transform(zio_t zio, abd_t data, uint64_t size, uint64_t bufsize,
	zio_transform_func_t *transform)
	{
	zio_transform_t *zt = kmem_alloc(sizeof (zio_transform_t), KM_SLEEP);

	zt->zt_orig_abd = zio->io_abd;
	zt->zt_orig_size = zio->io_size;
	zt->zt_bufsize = bufsize;
	zt->zt_transform = transform;

	zt->zt_next = zio->io_transform_stack;
	zio->io_transform_stack = zt;

	zio->io_abd = data;
	zio->io_size = size;
	}

	void
	zio_pop_transforms(zio_t *zio)
	{
	zio_transform_t *zt;

	while ((zt = zio->io_transform_stack) != NULL) {
	if (zt->zt_transform != NULL)
	zt->zt_transform(zio,
	zt->zt_orig_abd, zt->zt_orig_size);

	if (zt->zt_bufsize != 0)
	abd_free(zio->io_abd);

	zio->io_abd = zt->zt_orig_abd;
	zio->io_size = zt->zt_orig_size;
	zio->io_transform_stack = zt->zt_next;

	kmem_free(zt, sizeof (zio_transform_t));
	}
	}

	/*
	* ==========================================================================
	* I/O transform callbacks for subblocks, decompression, and decryption
	* ==========================================================================
	*/
	static void
	zio_subblock(zio_t zio, abd_t data, uint64_t size)
	{
	ASSERT(zio->io_size > size);

	if (zio->io_type == ZIO_TYPE_READ)
	abd_copy(data, zio->io_abd, size);
	}

	static void
	zio_decompress(zio_t zio, abd_t data, uint64_t size)
	{
	if (zio->io_error == 0) {
	void *tmp = abd_borrow_buf(data, size);
	int ret = zio_decompress_data(BP_GET_COMPRESS(zio->io_bp),
	zio->io_abd, tmp, zio->io_size, size,
	&zio->io_prop.zp_complevel);
	abd_return_buf_copy(data, tmp, size);

	if (zio_injection_enabled && ret == 0)
	ret = zio_handle_fault_injection(zio, EINVAL);

	if (ret != 0)
	zio->io_error = SET_ERROR(EIO);
	}
	}

	static void
	zio_decrypt(zio_t zio, abd_t data, uint64_t size)
	{
	int ret;
	void *tmp;
	blkptr_t *bp = zio->io_bp;
	spa_t *spa = zio->io_spa;
	uint64_t dsobj = zio->io_bookmark.zb_objset;
	uint64_t lsize = BP_GET_LSIZE(bp);
	dmu_object_type_t ot = BP_GET_TYPE(bp);
	uint8_t salt[ZIO_DATA_SALT_LEN];
	uint8_t iv[ZIO_DATA_IV_LEN];
	uint8_t mac[ZIO_DATA_MAC_LEN];
	boolean_t no_crypt = B_FALSE;

	ASSERT(BP_USES_CRYPT(bp));
	ASSERT3U(size, !=, 0);

	if (zio->io_error != 0)
	return;

	/*
	* Verify the cksum of MACs stored in an indirect bp. It will always
	* be possible to verify this since it does not require an encryption
	* key.
	*/
	if (BP_HAS_INDIRECT_MAC_CKSUM(bp)) {
	zio_crypt_decode_mac_bp(bp, mac);

	if (BP_GET_COMPRESS(bp) != ZIO_COMPRESS_OFF) {
	/*
	* We haven't decompressed the data yet, but
	* zio_crypt_do_indirect_mac_checksum() requires
	* decompressed data to be able to parse out the MACs
	* from the indirect block. We decompress it now and
	* throw away the result after we are finished.
	*/
	tmp = zio_buf_alloc(lsize);
	ret = zio_decompress_data(BP_GET_COMPRESS(bp),
	zio->io_abd, tmp, zio->io_size, lsize,
	&zio->io_prop.zp_complevel);
	if (ret != 0) {
	ret = SET_ERROR(EIO);
	goto error;
	}
	ret = zio_crypt_do_indirect_mac_checksum(B_FALSE,
	tmp, lsize, BP_SHOULD_BYTESWAP(bp), mac);
	zio_buf_free(tmp, lsize);
	} else {
	ret = zio_crypt_do_indirect_mac_checksum_abd(B_FALSE,
	zio->io_abd, size, BP_SHOULD_BYTESWAP(bp), mac);
	}
	abd_copy(data, zio->io_abd, size);

	if (zio_injection_enabled && ot != DMU_OT_DNODE && ret == 0) {
	ret = zio_handle_decrypt_injection(spa,
	&zio->io_bookmark, ot, ECKSUM);
	}
	if (ret != 0)
	goto error;

	return;
	}

	/*
	* If this is an authenticated block, just check the MAC. It would be
	* nice to separate this out into its own flag, but when this was done,
	* we had run out of bits in what is now zio_flag_t. Future cleanup
	* could make this a flag bit.
	*/
	if (BP_IS_AUTHENTICATED(bp)) {
	if (ot == DMU_OT_OBJSET) {
	ret = spa_do_crypt_objset_mac_abd(B_FALSE, spa,
	dsobj, zio->io_abd, size, BP_SHOULD_BYTESWAP(bp));
	} else {
	zio_crypt_decode_mac_bp(bp, mac);
	ret = spa_do_crypt_mac_abd(B_FALSE, spa, dsobj,
	zio->io_abd, size, mac);
	if (zio_injection_enabled && ret == 0) {
	ret = zio_handle_decrypt_injection(spa,
	&zio->io_bookmark, ot, ECKSUM);
	}
	}
	abd_copy(data, zio->io_abd, size);

	if (ret != 0)
	goto error;

	return;
	}

	zio_crypt_decode_params_bp(bp, salt, iv);

	if (ot == DMU_OT_INTENT_LOG) {
	tmp = abd_borrow_buf_copy(zio->io_abd, sizeof (zil_chain_t));
	zio_crypt_decode_mac_zil(tmp, mac);
	abd_return_buf(zio->io_abd, tmp, sizeof (zil_chain_t));
	} else {
	zio_crypt_decode_mac_bp(bp, mac);
	}

	ret = spa_do_crypt_abd(B_FALSE, spa, &zio->io_bookmark, BP_GET_TYPE(bp),
	BP_GET_DEDUP(bp), BP_SHOULD_BYTESWAP(bp), salt, iv, mac, size, data,
	zio->io_abd, &no_crypt);
	if (no_crypt)
	abd_copy(data, zio->io_abd, size);

	if (ret != 0)
	goto error;

	return;

	error:
	/* assert that the key was found unless this was speculative */
	ASSERT(ret != EACCES \|\| (zio->io_flags & ZIO_FLAG_SPECULATIVE));

	/*
	* If there was a decryption / authentication error return EIO as
	* the io_error. If this was not a speculative zio, create an ereport.
	*/
	if (ret == ECKSUM) {
	zio->io_error = SET_ERROR(EIO);
	if ((zio->io_flags & ZIO_FLAG_SPECULATIVE) == 0) {
	spa_log_error(spa, &zio->io_bookmark,
	BP_GET_LOGICAL_BIRTH(zio->io_bp));
	(void) zfs_ereport_post(FM_EREPORT_ZFS_AUTHENTICATION,
	spa, NULL, &zio->io_bookmark, zio, 0);
	}
	} else {
	zio->io_error = ret;
	}
	}

	/*
	* ==========================================================================
	* I/O parent/child relationships and pipeline interlocks
	* ==========================================================================
	*/
	zio_t *
	zio_walk_parents(zio_t cio, zio_link_t *zl)
	{
	list_t *pl = &cio->io_parent_list;

	zl = (zl == NULL) ? list_head(pl) : list_next(pl, *zl);
	if (*zl == NULL)
	return (NULL);

	ASSERT((*zl)->zl_child == cio);
	return ((*zl)->zl_parent);
	}

	zio_t *
	zio_walk_children(zio_t pio, zio_link_t *zl)
	{
	list_t *cl = &pio->io_child_list;

	ASSERT(MUTEX_HELD(&pio->io_lock));

	zl = (zl == NULL) ? list_head(cl) : list_next(cl, *zl);
	if (*zl == NULL)
	return (NULL);

	ASSERT((*zl)->zl_parent == pio);
	return ((*zl)->zl_child);
	}

	zio_t *
	zio_unique_parent(zio_t *cio)
	{
	zio_link_t *zl = NULL;
	zio_t *pio = zio_walk_parents(cio, &zl);

	VERIFY3P(zio_walk_parents(cio, &zl), ==, NULL);
	return (pio);
	}

	void
	zio_add_child(zio_t pio, zio_t cio)
	{
	/*
	* Logical I/Os can have logical, gang, or vdev children.
	* Gang I/Os can have gang or vdev children.
	* Vdev I/Os can only have vdev children.
	* The following ASSERT captures all of these constraints.
	*/
	ASSERT3S(cio->io_child_type, <=, pio->io_child_type);

	/* Parent should not have READY stage if child doesn't have it. */
	IMPLY((cio->io_pipeline & ZIO_STAGE_READY) == 0 &&
	(cio->io_child_type != ZIO_CHILD_VDEV),
	(pio->io_pipeline & ZIO_STAGE_READY) == 0);

	zio_link_t *zl = kmem_cache_alloc(zio_link_cache, KM_SLEEP);
	zl->zl_parent = pio;
	zl->zl_child = cio;

	mutex_enter(&pio->io_lock);
	mutex_enter(&cio->io_lock);

	ASSERT(pio->io_state[ZIO_WAIT_DONE] == 0);

	uint64_t *countp = pio->io_children[cio->io_child_type];
	for (int w = 0; w < ZIO_WAIT_TYPES; w++)
	countp[w] += !cio->io_state[w];

	list_insert_head(&pio->io_child_list, zl);
	list_insert_head(&cio->io_parent_list, zl);

	mutex_exit(&cio->io_lock);
	mutex_exit(&pio->io_lock);
	}

	void
	zio_add_child_first(zio_t pio, zio_t cio)
	{
	/*
	* Logical I/Os can have logical, gang, or vdev children.
	* Gang I/Os can have gang or vdev children.
	* Vdev I/Os can only have vdev children.
	* The following ASSERT captures all of these constraints.
	*/
	ASSERT3S(cio->io_child_type, <=, pio->io_child_type);

	/* Parent should not have READY stage if child doesn't have it. */
	IMPLY((cio->io_pipeline & ZIO_STAGE_READY) == 0 &&
	(cio->io_child_type != ZIO_CHILD_VDEV),
	(pio->io_pipeline & ZIO_STAGE_READY) == 0);

	zio_link_t *zl = kmem_cache_alloc(zio_link_cache, KM_SLEEP);
	zl->zl_parent = pio;
	zl->zl_child = cio;

	ASSERT(list_is_empty(&cio->io_parent_list));
	list_insert_head(&cio->io_parent_list, zl);

	mutex_enter(&pio->io_lock);

	ASSERT(pio->io_state[ZIO_WAIT_DONE] == 0);

	uint64_t *countp = pio->io_children[cio->io_child_type];
	for (int w = 0; w < ZIO_WAIT_TYPES; w++)
	countp[w] += !cio->io_state[w];

	list_insert_head(&pio->io_child_list, zl);

	mutex_exit(&pio->io_lock);
	}

	static void
	zio_remove_child(zio_t pio, zio_t cio, zio_link_t *zl)
	{
	ASSERT(zl->zl_parent == pio);
	ASSERT(zl->zl_child == cio);

	mutex_enter(&pio->io_lock);
	mutex_enter(&cio->io_lock);

	list_remove(&pio->io_child_list, zl);
	list_remove(&cio->io_parent_list, zl);

	mutex_exit(&cio->io_lock);
	mutex_exit(&pio->io_lock);
	kmem_cache_free(zio_link_cache, zl);
	}

	static boolean_t
	zio_wait_for_children(zio_t *zio, uint8_t childbits, enum zio_wait_type wait)
	{
	boolean_t waiting = B_FALSE;

	mutex_enter(&zio->io_lock);
	ASSERT(zio->io_stall == NULL);
	for (int c = 0; c < ZIO_CHILD_TYPES; c++) {
	if (!(ZIO_CHILD_BIT_IS_SET(childbits, c)))
	continue;

	uint64_t *countp = &zio->io_children[c][wait];
	if (*countp != 0) {
	zio->io_stage >>= 1;
	ASSERT3U(zio->io_stage, !=, ZIO_STAGE_OPEN);
	zio->io_stall = countp;
	waiting = B_TRUE;
	break;
	}
	}
	mutex_exit(&zio->io_lock);
	return (waiting);
	}

	__attribute__((always_inline))
	static inline void
	zio_notify_parent(zio_t pio, zio_t zio, enum zio_wait_type wait,
	zio_t **next_to_executep)
	{
	uint64_t *countp = &pio->io_children[zio->io_child_type][wait];
	int *errorp = &pio->io_child_error[zio->io_child_type];

	mutex_enter(&pio->io_lock);
	if (zio->io_error && !(zio->io_flags & ZIO_FLAG_DONT_PROPAGATE))
	errorp = zio_worst_error(errorp, zio->io_error);
	pio->io_reexecute \|= zio->io_reexecute;
	ASSERT3U(*countp, >, 0);

	(*countp)--;

	if (*countp == 0 && pio->io_stall == countp) {
	zio_taskq_type_t type =
	pio->io_stage < ZIO_STAGE_VDEV_IO_START ? ZIO_TASKQ_ISSUE :
	ZIO_TASKQ_INTERRUPT;
	pio->io_stall = NULL;
	mutex_exit(&pio->io_lock);

	/*
	* If we can tell the caller to execute this parent next, do
	* so. We do this if the parent's zio type matches the child's
	* type, or if it's a zio_null() with no done callback, and so
	* has no actual work to do. Otherwise dispatch the parent zio
	* in its own taskq.
	*
	* Having the caller execute the parent when possible reduces
	* locking on the zio taskq's, reduces context switch
	* overhead, and has no recursion penalty. Note that one
	* read from disk typically causes at least 3 zio's: a
	* zio_null(), the logical zio_read(), and then a physical
	* zio. When the physical ZIO completes, we are able to call
	* zio_done() on all 3 of these zio's from one invocation of
	* zio_execute() by returning the parent back to
	* zio_execute(). Since the parent isn't executed until this
	* thread returns back to zio_execute(), the caller should do
	* so promptly.
	*
	* In other cases, dispatching the parent prevents
	* overflowing the stack when we have deeply nested
	* parent-child relationships, as we do with the "mega zio"
	* of writes for spa_sync(), and the chain of ZIL blocks.
	*/
	if (next_to_executep != NULL && *next_to_executep == NULL &&
	(pio->io_type == zio->io_type \|\|
	(pio->io_type == ZIO_TYPE_NULL && !pio->io_done))) {
	*next_to_executep = pio;
	} else {
	zio_taskq_dispatch(pio, type, B_FALSE);
	}
	} else {
	mutex_exit(&pio->io_lock);
	}
	}

	static void
	zio_inherit_child_errors(zio_t *zio, enum zio_child c)
	{
	if (zio->io_child_error[c] != 0 && zio->io_error == 0)
	zio->io_error = zio->io_child_error[c];
	}

	int
	zio_bookmark_compare(const void x1, const void x2)
	{
	const zio_t *z1 = x1;
	const zio_t *z2 = x2;

	if (z1->io_bookmark.zb_objset < z2->io_bookmark.zb_objset)
	return (-1);
	if (z1->io_bookmark.zb_objset > z2->io_bookmark.zb_objset)
	return (1);

	if (z1->io_bookmark.zb_object < z2->io_bookmark.zb_object)
	return (-1);
	if (z1->io_bookmark.zb_object > z2->io_bookmark.zb_object)
	return (1);

	if (z1->io_bookmark.zb_level < z2->io_bookmark.zb_level)
	return (-1);
	if (z1->io_bookmark.zb_level > z2->io_bookmark.zb_level)
	return (1);

	if (z1->io_bookmark.zb_blkid < z2->io_bookmark.zb_blkid)
	return (-1);
	if (z1->io_bookmark.zb_blkid > z2->io_bookmark.zb_blkid)
	return (1);

	if (z1 < z2)
	return (-1);
	if (z1 > z2)
	return (1);

	return (0);
	}

	/*
	* ==========================================================================
	* Create the various types of I/O (read, write, free, etc)
	* ==========================================================================
	*/
	static zio_t *
	zio_create(zio_t pio, spa_t spa, uint64_t txg, const blkptr_t *bp,
	abd_t data, uint64_t lsize, uint64_t psize, zio_done_func_t done,
	void *private, zio_type_t type, zio_priority_t priority,
	zio_flag_t flags, vdev_t *vd, uint64_t offset,
	const zbookmark_phys_t *zb, enum zio_stage stage,
	enum zio_stage pipeline)
	{
	zio_t *zio;

	IMPLY(type != ZIO_TYPE_TRIM, psize <= SPA_MAXBLOCKSIZE);
	ASSERT(P2PHASE(psize, SPA_MINBLOCKSIZE) == 0);
	ASSERT(P2PHASE(offset, SPA_MINBLOCKSIZE) == 0);

	ASSERT(!vd \|\| spa_config_held(spa, SCL_STATE_ALL, RW_READER));
	ASSERT(!bp \|\| !(flags & ZIO_FLAG_CONFIG_WRITER));
	ASSERT(vd \|\| stage == ZIO_STAGE_OPEN);

	IMPLY(lsize != psize, (flags & ZIO_FLAG_RAW_COMPRESS) != 0);

	zio = kmem_cache_alloc(zio_cache, KM_SLEEP);
	memset(zio, 0, sizeof (zio_t));

	mutex_init(&zio->io_lock, NULL, MUTEX_NOLOCKDEP, NULL);
	cv_init(&zio->io_cv, NULL, CV_DEFAULT, NULL);

	list_create(&zio->io_parent_list, sizeof (zio_link_t),
	offsetof(zio_link_t, zl_parent_node));
	list_create(&zio->io_child_list, sizeof (zio_link_t),
	offsetof(zio_link_t, zl_child_node));
	metaslab_trace_init(&zio->io_alloc_list);

	if (vd != NULL)
	zio->io_child_type = ZIO_CHILD_VDEV;
	else if (flags & ZIO_FLAG_GANG_CHILD)
	zio->io_child_type = ZIO_CHILD_GANG;
	else if (flags & ZIO_FLAG_DDT_CHILD)
	zio->io_child_type = ZIO_CHILD_DDT;
	else
	zio->io_child_type = ZIO_CHILD_LOGICAL;

	if (bp != NULL) {
	if (type != ZIO_TYPE_WRITE \|\|
	zio->io_child_type == ZIO_CHILD_DDT) {
	zio->io_bp_copy = *bp;
	zio->io_bp = &zio->io_bp_copy; /* so caller can free */
	} else {
	zio->io_bp = (blkptr_t *)bp;
	}
	zio->io_bp_orig = *bp;
	if (zio->io_child_type == ZIO_CHILD_LOGICAL)
	zio->io_logical = zio;
	if (zio->io_child_type > ZIO_CHILD_GANG && BP_IS_GANG(bp))
	pipeline \|= ZIO_GANG_STAGES;
	}

	zio->io_spa = spa;
	zio->io_txg = txg;
	zio->io_done = done;
	zio->io_private = private;
	zio->io_type = type;
	zio->io_priority = priority;
	zio->io_vd = vd;
	zio->io_offset = offset;
	zio->io_orig_abd = zio->io_abd = data;
	zio->io_orig_size = zio->io_size = psize;
	zio->io_lsize = lsize;
	zio->io_orig_flags = zio->io_flags = flags;
	zio->io_orig_stage = zio->io_stage = stage;
	zio->io_orig_pipeline = zio->io_pipeline = pipeline;
	zio->io_pipeline_trace = ZIO_STAGE_OPEN;
	zio->io_allocator = ZIO_ALLOCATOR_NONE;

	zio->io_state[ZIO_WAIT_READY] = (stage >= ZIO_STAGE_READY) \|\|
	(pipeline & ZIO_STAGE_READY) == 0;
	zio->io_state[ZIO_WAIT_DONE] = (stage >= ZIO_STAGE_DONE);

	if (zb != NULL)
	zio->io_bookmark = *zb;

	if (pio != NULL) {
	zio->io_metaslab_class = pio->io_metaslab_class;
	if (zio->io_logical == NULL)
	zio->io_logical = pio->io_logical;
	if (zio->io_child_type == ZIO_CHILD_GANG)
	zio->io_gang_leader = pio->io_gang_leader;
	zio_add_child_first(pio, zio);
	}

	taskq_init_ent(&zio->io_tqent);

	return (zio);
	}

	void
	zio_destroy(zio_t *zio)
	{
	metaslab_trace_fini(&zio->io_alloc_list);
	list_destroy(&zio->io_parent_list);
	list_destroy(&zio->io_child_list);
	mutex_destroy(&zio->io_lock);
	cv_destroy(&zio->io_cv);
	kmem_cache_free(zio_cache, zio);
	}

	/*
	* ZIO intended to be between others. Provides synchronization at READY
	* and DONE pipeline stages and calls the respective callbacks.
	*/
	zio_t *
	zio_null(zio_t pio, spa_t spa, vdev_t vd, zio_done_func_t done,
	void *private, zio_flag_t flags)
	{
	zio_t *zio;

	zio = zio_create(pio, spa, 0, NULL, NULL, 0, 0, done, private,
	ZIO_TYPE_NULL, ZIO_PRIORITY_NOW, flags, vd, 0, NULL,
	ZIO_STAGE_OPEN, ZIO_INTERLOCK_PIPELINE);

	return (zio);
	}

	/*
	* ZIO intended to be a root of a tree. Unlike null ZIO does not have a
	* READY pipeline stage (is ready on creation), so it should not be used
	* as child of any ZIO that may need waiting for grandchildren READY stage
	* (any other ZIO type).
	*/
	zio_t *
	zio_root(spa_t spa, zio_done_func_t done, void *private, zio_flag_t flags)
	{
	zio_t *zio;

	zio = zio_create(NULL, spa, 0, NULL, NULL, 0, 0, done, private,
	ZIO_TYPE_NULL, ZIO_PRIORITY_NOW, flags, NULL, 0, NULL,
	ZIO_STAGE_OPEN, ZIO_ROOT_PIPELINE);

	return (zio);
	}

	static int
	zfs_blkptr_verify_log(spa_t spa, const blkptr_t bp,
	enum blk_verify_flag blk_verify, const char *fmt, ...)
	{
	va_list adx;
	char buf[256];

	va_start(adx, fmt);
	(void) vsnprintf(buf, sizeof (buf), fmt, adx);
	va_end(adx);

	zfs_dbgmsg("bad blkptr at %px: "
	"DVA[0]=%#llx/%#llx "
	"DVA[1]=%#llx/%#llx "
	"DVA[2]=%#llx/%#llx "
	"prop=%#llx "
	"pad=%#llx,%#llx "
	"phys_birth=%#llx "
	"birth=%#llx "
	"fill=%#llx "
	"cksum=%#llx/%#llx/%#llx/%#llx",
	bp,
	(long long)bp->blk_dva[0].dva_word[0],
	(long long)bp->blk_dva[0].dva_word[1],
	(long long)bp->blk_dva[1].dva_word[0],
	(long long)bp->blk_dva[1].dva_word[1],
	(long long)bp->blk_dva[2].dva_word[0],
	(long long)bp->blk_dva[2].dva_word[1],
	(long long)bp->blk_prop,
	(long long)bp->blk_pad[0],
	(long long)bp->blk_pad[1],
	(long long)BP_GET_PHYSICAL_BIRTH(bp),
	(long long)BP_GET_LOGICAL_BIRTH(bp),
	(long long)bp->blk_fill,
	(long long)bp->blk_cksum.zc_word[0],
	(long long)bp->blk_cksum.zc_word[1],
	(long long)bp->blk_cksum.zc_word[2],
	(long long)bp->blk_cksum.zc_word[3]);
	switch (blk_verify) {
	case BLK_VERIFY_HALT:
	zfs_panic_recover("%s: %s", spa_name(spa), buf);
	break;
	case BLK_VERIFY_LOG:
	zfs_dbgmsg("%s: %s", spa_name(spa), buf);
	break;
	case BLK_VERIFY_ONLY:
	break;
	}

	return (1);
	}

	/*
	* Verify the block pointer fields contain reasonable values. This means
	* it only contains known object types, checksum/compression identifiers,
	* block sizes within the maximum allowed limits, valid DVAs, etc.
	*
	* If everything checks out B_TRUE is returned. The zfs_blkptr_verify
	* argument controls the behavior when an invalid field is detected.
	*
	* Values for blk_verify_flag:
	* BLK_VERIFY_ONLY: evaluate the block
	* BLK_VERIFY_LOG: evaluate the block and log problems
	* BLK_VERIFY_HALT: call zfs_panic_recover on error
	*
	* Values for blk_config_flag:
	* BLK_CONFIG_HELD: caller holds SCL_VDEV for writer
	* BLK_CONFIG_NEEDED: caller holds no config lock, SCL_VDEV will be
	* obtained for reader
	* BLK_CONFIG_SKIP: skip checks which require SCL_VDEV, for better
	* performance
	*/
	boolean_t
	zfs_blkptr_verify(spa_t spa, const blkptr_t bp,
	enum blk_config_flag blk_config, enum blk_verify_flag blk_verify)
	{
	int errors = 0;

	if (!DMU_OT_IS_VALID(BP_GET_TYPE(bp))) {
	errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
	"blkptr at %px has invalid TYPE %llu",
	bp, (longlong_t)BP_GET_TYPE(bp));
	}
	if (BP_GET_CHECKSUM(bp) >= ZIO_CHECKSUM_FUNCTIONS) {
	errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
	"blkptr at %px has invalid CHECKSUM %llu",
	bp, (longlong_t)BP_GET_CHECKSUM(bp));
	}
	if (BP_GET_COMPRESS(bp) >= ZIO_COMPRESS_FUNCTIONS) {
	errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
	"blkptr at %px has invalid COMPRESS %llu",
	bp, (longlong_t)BP_GET_COMPRESS(bp));
	}
	if (BP_GET_LSIZE(bp) > SPA_MAXBLOCKSIZE) {
	errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
	"blkptr at %px has invalid LSIZE %llu",
	bp, (longlong_t)BP_GET_LSIZE(bp));
	}
	if (BP_GET_PSIZE(bp) > SPA_MAXBLOCKSIZE) {
	errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
	"blkptr at %px has invalid PSIZE %llu",
	bp, (longlong_t)BP_GET_PSIZE(bp));
	}

	if (BP_IS_EMBEDDED(bp)) {
	if (BPE_GET_ETYPE(bp) >= NUM_BP_EMBEDDED_TYPES) {
	errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
	"blkptr at %px has invalid ETYPE %llu",
	bp, (longlong_t)BPE_GET_ETYPE(bp));
	}
	}

	/*
	* Do not verify individual DVAs if the config is not trusted. This
	* will be done once the zio is executed in vdev_mirror_map_alloc.
	*/
	if (!spa->spa_trust_config)
	return (errors == 0);

	switch (blk_config) {
	case BLK_CONFIG_HELD:
	ASSERT(spa_config_held(spa, SCL_VDEV, RW_WRITER));
	break;
	case BLK_CONFIG_NEEDED:
	spa_config_enter(spa, SCL_VDEV, bp, RW_READER);
	break;
	case BLK_CONFIG_SKIP:
	return (errors == 0);
	default:
	panic("invalid blk_config %u", blk_config);
	}

	/*
	* Pool-specific checks.
	*
	* Note: it would be nice to verify that the logical birth
	* and physical birth are not too large. However,
	* spa_freeze() allows the birth time of log blocks (and
	* dmu_sync()-ed blocks that are in the log) to be arbitrarily
	* large.
	*/
	for (int i = 0; i < BP_GET_NDVAS(bp); i++) {
	const dva_t *dva = &bp->blk_dva[i];
	uint64_t vdevid = DVA_GET_VDEV(dva);

	if (vdevid >= spa->spa_root_vdev->vdev_children) {
	errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
	"blkptr at %px DVA %u has invalid VDEV %llu",
	bp, i, (longlong_t)vdevid);
	continue;
	}
	vdev_t *vd = spa->spa_root_vdev->vdev_child[vdevid];
	if (vd == NULL) {
	errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
	"blkptr at %px DVA %u has invalid VDEV %llu",
	bp, i, (longlong_t)vdevid);
	continue;
	}
	if (vd->vdev_ops == &vdev_hole_ops) {
	errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
	"blkptr at %px DVA %u has hole VDEV %llu",
	bp, i, (longlong_t)vdevid);
	continue;
	}
	if (vd->vdev_ops == &vdev_missing_ops) {
	/*
	* "missing" vdevs are valid during import, but we
	* don't have their detailed info (e.g. asize), so
	* we can't perform any more checks on them.
	*/
	continue;
	}
	uint64_t offset = DVA_GET_OFFSET(dva);
	uint64_t asize = DVA_GET_ASIZE(dva);
	if (DVA_GET_GANG(dva))
	asize = vdev_gang_header_asize(vd);
	if (offset + asize > vd->vdev_asize) {
	errors += zfs_blkptr_verify_log(spa, bp, blk_verify,
	"blkptr at %px DVA %u has invalid OFFSET %llu",
	bp, i, (longlong_t)offset);
	}
	}
	if (blk_config == BLK_CONFIG_NEEDED)
	spa_config_exit(spa, SCL_VDEV, bp);

	return (errors == 0);
	}

	boolean_t
	zfs_dva_valid(spa_t spa, const dva_t dva, const blkptr_t *bp)
	{
	(void) bp;
	uint64_t vdevid = DVA_GET_VDEV(dva);

	if (vdevid >= spa->spa_root_vdev->vdev_children)
	return (B_FALSE);

	vdev_t *vd = spa->spa_root_vdev->vdev_child[vdevid];
	if (vd == NULL)
	return (B_FALSE);

	if (vd->vdev_ops == &vdev_hole_ops)
	return (B_FALSE);

	if (vd->vdev_ops == &vdev_missing_ops) {
	return (B_FALSE);
	}

	uint64_t offset = DVA_GET_OFFSET(dva);
	uint64_t asize = DVA_GET_ASIZE(dva);

	if (DVA_GET_GANG(dva))
	asize = vdev_gang_header_asize(vd);
	if (offset + asize > vd->vdev_asize)
	return (B_FALSE);

	return (B_TRUE);
	}

	zio_t *
	zio_read(zio_t pio, spa_t spa, const blkptr_t *bp,
	abd_t data, uint64_t size, zio_done_func_t done, void *private,
	zio_priority_t priority, zio_flag_t flags, const zbookmark_phys_t *zb)
	{
	zio_t *zio;

	zio = zio_create(pio, spa, BP_GET_BIRTH(bp), bp,
	data, size, size, done, private,
	ZIO_TYPE_READ, priority, flags, NULL, 0, zb,
	ZIO_STAGE_OPEN, (flags & ZIO_FLAG_DDT_CHILD) ?
	ZIO_DDT_CHILD_READ_PIPELINE : ZIO_READ_PIPELINE);

	return (zio);
	}

	zio_t *
	zio_write(zio_t pio, spa_t spa, uint64_t txg, blkptr_t *bp,
	abd_t data, uint64_t lsize, uint64_t psize, const zio_prop_t zp,
	zio_done_func_t ready, zio_done_func_t children_ready,
	zio_done_func_t done, void private, zio_priority_t priority,
	zio_flag_t flags, const zbookmark_phys_t *zb)
	{
	zio_t *zio;

	ASSERT(zp->zp_checksum >= ZIO_CHECKSUM_OFF &&
	zp->zp_checksum < ZIO_CHECKSUM_FUNCTIONS &&
	zp->zp_compress >= ZIO_COMPRESS_OFF &&
	zp->zp_compress < ZIO_COMPRESS_FUNCTIONS &&
	DMU_OT_IS_VALID(zp->zp_type) &&
	zp->zp_level < 32 &&
	zp->zp_copies > 0 &&
	zp->zp_copies <= spa_max_replication(spa));

	zio = zio_create(pio, spa, txg, bp, data, lsize, psize, done, private,
	ZIO_TYPE_WRITE, priority, flags, NULL, 0, zb,
	ZIO_STAGE_OPEN, (flags & ZIO_FLAG_DDT_CHILD) ?
	ZIO_DDT_CHILD_WRITE_PIPELINE : ZIO_WRITE_PIPELINE);

	zio->io_ready = ready;
	zio->io_children_ready = children_ready;
	zio->io_prop = *zp;

	/*
	* Data can be NULL if we are going to call zio_write_override() to
	* provide the already-allocated BP. But we may need the data to
	* verify a dedup hit (if requested). In this case, don't try to
	* dedup (just take the already-allocated BP verbatim). Encrypted
	* dedup blocks need data as well so we also disable dedup in this
	* case.
	*/
	if (data == NULL &&
	(zio->io_prop.zp_dedup_verify \|\| zio->io_prop.zp_encrypt)) {
	zio->io_prop.zp_dedup = zio->io_prop.zp_dedup_verify = B_FALSE;
	}

	return (zio);
	}

	zio_t *
	zio_rewrite(zio_t pio, spa_t spa, uint64_t txg, blkptr_t bp, abd_t data,
	uint64_t size, zio_done_func_t done, void private,
	zio_priority_t priority, zio_flag_t flags, zbookmark_phys_t *zb)
	{
	zio_t *zio;

	zio = zio_create(pio, spa, txg, bp, data, size, size, done, private,
	ZIO_TYPE_WRITE, priority, flags \| ZIO_FLAG_IO_REWRITE, NULL, 0, zb,
	ZIO_STAGE_OPEN, ZIO_REWRITE_PIPELINE);

	return (zio);
	}

	void
	zio_write_override(zio_t zio, blkptr_t bp, int copies, boolean_t nopwrite,
	boolean_t brtwrite)
	{
	ASSERT(zio->io_type == ZIO_TYPE_WRITE);
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
	ASSERT(zio->io_stage == ZIO_STAGE_OPEN);
	ASSERT(zio->io_txg == spa_syncing_txg(zio->io_spa));
	ASSERT(!brtwrite \|\| !nopwrite);

	/*
	* We must reset the io_prop to match the values that existed
	* when the bp was first written by dmu_sync() keeping in mind
	* that nopwrite and dedup are mutually exclusive.
	*/
	zio->io_prop.zp_dedup = nopwrite ? B_FALSE : zio->io_prop.zp_dedup;
	zio->io_prop.zp_nopwrite = nopwrite;
	zio->io_prop.zp_brtwrite = brtwrite;
	zio->io_prop.zp_copies = copies;
	zio->io_bp_override = bp;
	}

	void
	zio_free(spa_t spa, uint64_t txg, const blkptr_t bp)
	{

	(void) zfs_blkptr_verify(spa, bp, BLK_CONFIG_NEEDED, BLK_VERIFY_HALT);

	/*
	* The check for EMBEDDED is a performance optimization. We
	* process the free here (by ignoring it) rather than
	* putting it on the list and then processing it in zio_free_sync().
	*/
	if (BP_IS_EMBEDDED(bp))
	return;

	/*
	* Frees that are for the currently-syncing txg, are not going to be
	* deferred, and which will not need to do a read (i.e. not GANG or
	* DEDUP), can be processed immediately. Otherwise, put them on the
	* in-memory list for later processing.
	*
	* Note that we only defer frees after zfs_sync_pass_deferred_free
	* when the log space map feature is disabled. [see relevant comment
	* in spa_sync_iterate_to_convergence()]
	*/
	if (BP_IS_GANG(bp) \|\|
	BP_GET_DEDUP(bp) \|\|
	txg != spa->spa_syncing_txg \|\|
	(spa_sync_pass(spa) >= zfs_sync_pass_deferred_free &&
	!spa_feature_is_active(spa, SPA_FEATURE_LOG_SPACEMAP)) \|\|
	brt_maybe_exists(spa, bp)) {
	metaslab_check_free(spa, bp);
	bplist_append(&spa->spa_free_bplist[txg & TXG_MASK], bp);
	} else {
	VERIFY3P(zio_free_sync(NULL, spa, txg, bp, 0), ==, NULL);
	}
	}

	/*
	* To improve performance, this function may return NULL if we were able
	* to do the free immediately. This avoids the cost of creating a zio
	* (and linking it to the parent, etc).
	*/
	zio_t *
	zio_free_sync(zio_t pio, spa_t spa, uint64_t txg, const blkptr_t *bp,
	zio_flag_t flags)
	{
	ASSERT(!BP_IS_HOLE(bp));
	ASSERT(spa_syncing_txg(spa) == txg);

	if (BP_IS_EMBEDDED(bp))
	return (NULL);

	metaslab_check_free(spa, bp);
	arc_freed(spa, bp);
	dsl_scan_freed(spa, bp);

	if (BP_IS_GANG(bp) \|\|
	BP_GET_DEDUP(bp) \|\|
	brt_maybe_exists(spa, bp)) {
	/*
	* GANG, DEDUP and BRT blocks can induce a read (for the gang
	* block header, the DDT or the BRT), so issue them
	* asynchronously so that this thread is not tied up.
	*/
	enum zio_stage stage =
	ZIO_FREE_PIPELINE \| ZIO_STAGE_ISSUE_ASYNC;

	return (zio_create(pio, spa, txg, bp, NULL, BP_GET_PSIZE(bp),
	BP_GET_PSIZE(bp), NULL, NULL,
	ZIO_TYPE_FREE, ZIO_PRIORITY_NOW,
	flags, NULL, 0, NULL, ZIO_STAGE_OPEN, stage));
	} else {
	metaslab_free(spa, bp, txg, B_FALSE);
	return (NULL);
	}
	}

	zio_t *
	zio_claim(zio_t pio, spa_t spa, uint64_t txg, const blkptr_t *bp,
	zio_done_func_t done, void private, zio_flag_t flags)
	{
	zio_t *zio;

	(void) zfs_blkptr_verify(spa, bp, (flags & ZIO_FLAG_CONFIG_WRITER) ?
	BLK_CONFIG_HELD : BLK_CONFIG_NEEDED, BLK_VERIFY_HALT);

	if (BP_IS_EMBEDDED(bp))
	return (zio_null(pio, spa, NULL, NULL, NULL, 0));

	/*
	* A claim is an allocation of a specific block. Claims are needed
	* to support immediate writes in the intent log. The issue is that
	* immediate writes contain committed data, but in a txg that was
	* not committed. Upon opening the pool after an unclean shutdown,
	* the intent log claims all blocks that contain immediate write data
	* so that the SPA knows they're in use.
	*
	* All claims must be resolved in the first txg -- before the SPA
	* starts allocating blocks -- so that nothing is allocated twice.
	* If txg == 0 we just verify that the block is claimable.
	*/
	ASSERT3U(BP_GET_LOGICAL_BIRTH(&spa->spa_uberblock.ub_rootbp), <,
	spa_min_claim_txg(spa));
	ASSERT(txg == spa_min_claim_txg(spa) \|\| txg == 0);
	ASSERT(!BP_GET_DEDUP(bp) \|\| !spa_writeable(spa)); /* zdb(8) */

	zio = zio_create(pio, spa, txg, bp, NULL, BP_GET_PSIZE(bp),
	BP_GET_PSIZE(bp), done, private, ZIO_TYPE_CLAIM, ZIO_PRIORITY_NOW,
	flags, NULL, 0, NULL, ZIO_STAGE_OPEN, ZIO_CLAIM_PIPELINE);
	ASSERT0(zio->io_queued_timestamp);

	return (zio);
	}

	zio_t *
	zio_trim(zio_t pio, vdev_t vd, uint64_t offset, uint64_t size,
	zio_done_func_t done, void private, zio_priority_t priority,
	zio_flag_t flags, enum trim_flag trim_flags)
	{
	zio_t *zio;

	ASSERT0(vd->vdev_children);
	ASSERT0(P2PHASE(offset, 1ULL << vd->vdev_ashift));
	ASSERT0(P2PHASE(size, 1ULL << vd->vdev_ashift));
	ASSERT3U(size, !=, 0);

	zio = zio_create(pio, vd->vdev_spa, 0, NULL, NULL, size, size, done,
	private, ZIO_TYPE_TRIM, priority, flags \| ZIO_FLAG_PHYSICAL,
	vd, offset, NULL, ZIO_STAGE_OPEN, ZIO_TRIM_PIPELINE);
	zio->io_trim_flags = trim_flags;

	return (zio);
	}

	zio_t *
	zio_read_phys(zio_t pio, vdev_t vd, uint64_t offset, uint64_t size,
	abd_t data, int checksum, zio_done_func_t done, void *private,
	zio_priority_t priority, zio_flag_t flags, boolean_t labels)
	{
	zio_t *zio;

	ASSERT(vd->vdev_children == 0);
	ASSERT(!labels \|\| offset + size <= VDEV_LABEL_START_SIZE \|\|
	offset >= vd->vdev_psize - VDEV_LABEL_END_SIZE);
	ASSERT3U(offset + size, <=, vd->vdev_psize);

	zio = zio_create(pio, vd->vdev_spa, 0, NULL, data, size, size, done,
	private, ZIO_TYPE_READ, priority, flags \| ZIO_FLAG_PHYSICAL, vd,
	offset, NULL, ZIO_STAGE_OPEN, ZIO_READ_PHYS_PIPELINE);

	zio->io_prop.zp_checksum = checksum;

	return (zio);
	}

	zio_t *
	zio_write_phys(zio_t pio, vdev_t vd, uint64_t offset, uint64_t size,
	abd_t data, int checksum, zio_done_func_t done, void *private,
	zio_priority_t priority, zio_flag_t flags, boolean_t labels)
	{
	zio_t *zio;

	ASSERT(vd->vdev_children == 0);
	ASSERT(!labels \|\| offset + size <= VDEV_LABEL_START_SIZE \|\|
	offset >= vd->vdev_psize - VDEV_LABEL_END_SIZE);
	ASSERT3U(offset + size, <=, vd->vdev_psize);

	zio = zio_create(pio, vd->vdev_spa, 0, NULL, data, size, size, done,
	private, ZIO_TYPE_WRITE, priority, flags \| ZIO_FLAG_PHYSICAL, vd,
	offset, NULL, ZIO_STAGE_OPEN, ZIO_WRITE_PHYS_PIPELINE);

	zio->io_prop.zp_checksum = checksum;

	if (zio_checksum_table[checksum].ci_flags & ZCHECKSUM_FLAG_EMBEDDED) {
	/*
	* zec checksums are necessarily destructive -- they modify
	* the end of the write buffer to hold the verifier/checksum.
	* Therefore, we must make a local copy in case the data is
	* being written to multiple places in parallel.
	*/
	abd_t *wbuf = abd_alloc_sametype(data, size);
	abd_copy(wbuf, data, size);

	zio_push_transform(zio, wbuf, size, size, NULL);
	}

	return (zio);
	}

	/*
	* Create a child I/O to do some work for us.
	*/
	zio_t *
	zio_vdev_child_io(zio_t pio, blkptr_t bp, vdev_t *vd, uint64_t offset,
	abd_t *data, uint64_t size, int type, zio_priority_t priority,
	zio_flag_t flags, zio_done_func_t done, void private)
	{
	enum zio_stage pipeline = ZIO_VDEV_CHILD_PIPELINE;
	zio_t *zio;

	/*
	* vdev child I/Os do not propagate their error to the parent.
	* Therefore, for correct operation the caller must check for
	* and handle the error in the child i/o's done callback.
	* The only exceptions are i/os that we don't care about
	* (OPTIONAL or REPAIR).
	*/
	ASSERT((flags & ZIO_FLAG_OPTIONAL) \|\| (flags & ZIO_FLAG_IO_REPAIR) \|\|
	done != NULL);

	if (type == ZIO_TYPE_READ && bp != NULL) {
	/*
	* If we have the bp, then the child should perform the
	* checksum and the parent need not. This pushes error
	* detection as close to the leaves as possible and
	* eliminates redundant checksums in the interior nodes.
	*/
	pipeline \|= ZIO_STAGE_CHECKSUM_VERIFY;
	pio->io_pipeline &= ~ZIO_STAGE_CHECKSUM_VERIFY;
	}

	if (vd->vdev_ops->vdev_op_leaf) {
	ASSERT0(vd->vdev_children);
	offset += VDEV_LABEL_START_SIZE;
	}

	flags \|= ZIO_VDEV_CHILD_FLAGS(pio);

	/*
	* If we've decided to do a repair, the write is not speculative --
	* even if the original read was.
	*/
	if (flags & ZIO_FLAG_IO_REPAIR)
	flags &= ~ZIO_FLAG_SPECULATIVE;

	/*
	* If we're creating a child I/O that is not associated with a
	* top-level vdev, then the child zio is not an allocating I/O.
	* If this is a retried I/O then we ignore it since we will
	* have already processed the original allocating I/O.
	*/
	if (flags & ZIO_FLAG_IO_ALLOCATING &&
	(vd != vd->vdev_top \|\| (flags & ZIO_FLAG_IO_RETRY))) {
	ASSERT(pio->io_metaslab_class != NULL);
	ASSERT(pio->io_metaslab_class->mc_alloc_throttle_enabled);
	ASSERT(type == ZIO_TYPE_WRITE);
	ASSERT(priority == ZIO_PRIORITY_ASYNC_WRITE);
	ASSERT(!(flags & ZIO_FLAG_IO_REPAIR));
	ASSERT(!(pio->io_flags & ZIO_FLAG_IO_REWRITE) \|\|
	pio->io_child_type == ZIO_CHILD_GANG);

	flags &= ~ZIO_FLAG_IO_ALLOCATING;
	}

	zio = zio_create(pio, pio->io_spa, pio->io_txg, bp, data, size, size,
	done, private, type, priority, flags, vd, offset, &pio->io_bookmark,
	ZIO_STAGE_VDEV_IO_START >> 1, pipeline);
	ASSERT3U(zio->io_child_type, ==, ZIO_CHILD_VDEV);

	return (zio);
	}

	zio_t *
	zio_vdev_delegated_io(vdev_t vd, uint64_t offset, abd_t data, uint64_t size,
	zio_type_t type, zio_priority_t priority, zio_flag_t flags,
	zio_done_func_t done, void private)
	{
	zio_t *zio;

	ASSERT(vd->vdev_ops->vdev_op_leaf);

	zio = zio_create(NULL, vd->vdev_spa, 0, NULL,
	data, size, size, done, private, type, priority,
	flags \| ZIO_FLAG_CANFAIL \| ZIO_FLAG_DONT_RETRY \| ZIO_FLAG_DELEGATED,
	vd, offset, NULL,
	ZIO_STAGE_VDEV_IO_START >> 1, ZIO_VDEV_CHILD_PIPELINE);

	return (zio);
	}


	/*
	* Send a flush command to the given vdev. Unlike most zio creation functions,
	* the flush zios are issued immediately. You can wait on pio to pause until
	* the flushes complete.
	*/
	void
	zio_flush(zio_t pio, vdev_t vd)
	{
	const zio_flag_t flags = ZIO_FLAG_CANFAIL \| ZIO_FLAG_DONT_PROPAGATE \|
	ZIO_FLAG_DONT_RETRY;

	if (vd->vdev_nowritecache)
	return;

	if (vd->vdev_children == 0) {
	zio_nowait(zio_create(pio, vd->vdev_spa, 0, NULL, NULL, 0, 0,
	NULL, NULL, ZIO_TYPE_FLUSH, ZIO_PRIORITY_NOW, flags, vd, 0,
	NULL, ZIO_STAGE_OPEN, ZIO_FLUSH_PIPELINE));
	} else {
	for (uint64_t c = 0; c < vd->vdev_children; c++)
	zio_flush(pio, vd->vdev_child[c]);
	}
	}

	void
	zio_shrink(zio_t *zio, uint64_t size)
	{
	ASSERT3P(zio->io_executor, ==, NULL);
	ASSERT3U(zio->io_orig_size, ==, zio->io_size);
	ASSERT3U(size, <=, zio->io_size);

	/*
	* We don't shrink for raidz because of problems with the
	* reconstruction when reading back less than the block size.
	* Note, BP_IS_RAIDZ() assumes no compression.
	*/
	ASSERT(BP_GET_COMPRESS(zio->io_bp) == ZIO_COMPRESS_OFF);
	if (!BP_IS_RAIDZ(zio->io_bp)) {
	/* we are not doing a raw write */
	ASSERT3U(zio->io_size, ==, zio->io_lsize);
	zio->io_orig_size = zio->io_size = zio->io_lsize = size;
	}
	}

	/*
	* Round provided allocation size up to a value that can be allocated
	* by at least some vdev(s) in the pool with minimum or no additional
	* padding and without extra space usage on others
	*/
	static uint64_t
	zio_roundup_alloc_size(spa_t *spa, uint64_t size)
	{
	if (size > spa->spa_min_alloc)
	return (roundup(size, spa->spa_gcd_alloc));
	return (spa->spa_min_alloc);
	}

	/*
	* ==========================================================================
	* Prepare to read and write logical blocks
	* ==========================================================================
	*/

	static zio_t *
	zio_read_bp_init(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;
	uint64_t psize =
	BP_IS_EMBEDDED(bp) ? BPE_GET_PSIZE(bp) : BP_GET_PSIZE(bp);

	ASSERT3P(zio->io_bp, ==, &zio->io_bp_copy);

	if (BP_GET_COMPRESS(bp) != ZIO_COMPRESS_OFF &&
	zio->io_child_type == ZIO_CHILD_LOGICAL &&
	!(zio->io_flags & ZIO_FLAG_RAW_COMPRESS)) {
	zio_push_transform(zio, abd_alloc_sametype(zio->io_abd, psize),
	psize, psize, zio_decompress);
	}

	if (((BP_IS_PROTECTED(bp) && !(zio->io_flags & ZIO_FLAG_RAW_ENCRYPT)) \|\|
	BP_HAS_INDIRECT_MAC_CKSUM(bp)) &&
	zio->io_child_type == ZIO_CHILD_LOGICAL) {
	zio_push_transform(zio, abd_alloc_sametype(zio->io_abd, psize),
	psize, psize, zio_decrypt);
	}

	if (BP_IS_EMBEDDED(bp) && BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA) {
	int psize = BPE_GET_PSIZE(bp);
	void *data = abd_borrow_buf(zio->io_abd, psize);

	zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
	decode_embedded_bp_compressed(bp, data);
	abd_return_buf_copy(zio->io_abd, data, psize);
	} else {
	ASSERT(!BP_IS_EMBEDDED(bp));
	}

	if (BP_GET_DEDUP(bp) && zio->io_child_type == ZIO_CHILD_LOGICAL)
	zio->io_pipeline = ZIO_DDT_READ_PIPELINE;

	return (zio);
	}

	static zio_t *
	zio_write_bp_init(zio_t *zio)
	{
	if (!IO_IS_ALLOCATING(zio))
	return (zio);

	ASSERT(zio->io_child_type != ZIO_CHILD_DDT);

	if (zio->io_bp_override) {
	blkptr_t *bp = zio->io_bp;
	zio_prop_t *zp = &zio->io_prop;

	ASSERT(BP_GET_LOGICAL_BIRTH(bp) != zio->io_txg);

	bp = zio->io_bp_override;
	zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;

	if (zp->zp_brtwrite)
	return (zio);

	ASSERT(!BP_GET_DEDUP(zio->io_bp_override));

	if (BP_IS_EMBEDDED(bp))
	return (zio);

	/*
	* If we've been overridden and nopwrite is set then
	* set the flag accordingly to indicate that a nopwrite
	* has already occurred.
	*/
	if (!BP_IS_HOLE(bp) && zp->zp_nopwrite) {
	ASSERT(!zp->zp_dedup);
	ASSERT3U(BP_GET_CHECKSUM(bp), ==, zp->zp_checksum);
	zio->io_flags \|= ZIO_FLAG_NOPWRITE;
	return (zio);
	}

	ASSERT(!zp->zp_nopwrite);

	if (BP_IS_HOLE(bp) \|\| !zp->zp_dedup)
	return (zio);

	ASSERT((zio_checksum_table[zp->zp_checksum].ci_flags &
	ZCHECKSUM_FLAG_DEDUP) \|\| zp->zp_dedup_verify);

	if (BP_GET_CHECKSUM(bp) == zp->zp_checksum &&
	!zp->zp_encrypt) {
	BP_SET_DEDUP(bp, 1);
	zio->io_pipeline \|= ZIO_STAGE_DDT_WRITE;
	return (zio);
	}

	/*
	* We were unable to handle this as an override bp, treat
	* it as a regular write I/O.
	*/
	zio->io_bp_override = NULL;
	*bp = zio->io_bp_orig;
	zio->io_pipeline = zio->io_orig_pipeline;
	}

	return (zio);
	}

	static zio_t *
	zio_write_compress(zio_t *zio)
	{
	spa_t *spa = zio->io_spa;
	zio_prop_t *zp = &zio->io_prop;
	enum zio_compress compress = zp->zp_compress;
	blkptr_t *bp = zio->io_bp;
	uint64_t lsize = zio->io_lsize;
	uint64_t psize = zio->io_size;
	uint32_t pass = 1;

	/*
	* If our children haven't all reached the ready stage,
	* wait for them and then repeat this pipeline stage.
	*/
	if (zio_wait_for_children(zio, ZIO_CHILD_LOGICAL_BIT \|
	ZIO_CHILD_GANG_BIT, ZIO_WAIT_READY)) {
	return (NULL);
	}

	if (!IO_IS_ALLOCATING(zio))
	return (zio);

	if (zio->io_children_ready != NULL) {
	/*
	* Now that all our children are ready, run the callback
	* associated with this zio in case it wants to modify the
	* data to be written.
	*/
	ASSERT3U(zp->zp_level, >, 0);
	zio->io_children_ready(zio);
	}

	ASSERT(zio->io_child_type != ZIO_CHILD_DDT);
	ASSERT(zio->io_bp_override == NULL);

	if (!BP_IS_HOLE(bp) && BP_GET_LOGICAL_BIRTH(bp) == zio->io_txg) {
	/*
	* We're rewriting an existing block, which means we're
	* working on behalf of spa_sync(). For spa_sync() to
	* converge, it must eventually be the case that we don't
	* have to allocate new blocks. But compression changes
	* the blocksize, which forces a reallocate, and makes
	* convergence take longer. Therefore, after the first
	* few passes, stop compressing to ensure convergence.
	*/
	pass = spa_sync_pass(spa);

	ASSERT(zio->io_txg == spa_syncing_txg(spa));
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
	ASSERT(!BP_GET_DEDUP(bp));

	if (pass >= zfs_sync_pass_dont_compress)
	compress = ZIO_COMPRESS_OFF;

	/* Make sure someone doesn't change their mind on overwrites */
	ASSERT(BP_IS_EMBEDDED(bp) \|\| BP_IS_GANG(bp) \|\|
	MIN(zp->zp_copies, spa_max_replication(spa))
	== BP_GET_NDVAS(bp));
	}

	/* If it's a compressed write that is not raw, compress the buffer. */
	if (compress != ZIO_COMPRESS_OFF &&
	!(zio->io_flags & ZIO_FLAG_RAW_COMPRESS)) {
	void *cbuf = NULL;
	psize = zio_compress_data(compress, zio->io_abd, &cbuf, lsize,
	zp->zp_complevel);
	if (psize == 0) {
	compress = ZIO_COMPRESS_OFF;
	} else if (psize >= lsize) {
	compress = ZIO_COMPRESS_OFF;
	if (cbuf != NULL)
	zio_buf_free(cbuf, lsize);
	} else if (!zp->zp_dedup && !zp->zp_encrypt &&
	psize <= BPE_PAYLOAD_SIZE &&
	zp->zp_level == 0 && !DMU_OT_HAS_FILL(zp->zp_type) &&
	spa_feature_is_enabled(spa, SPA_FEATURE_EMBEDDED_DATA)) {
	encode_embedded_bp_compressed(bp,
	cbuf, compress, lsize, psize);
	BPE_SET_ETYPE(bp, BP_EMBEDDED_TYPE_DATA);
	BP_SET_TYPE(bp, zio->io_prop.zp_type);
	BP_SET_LEVEL(bp, zio->io_prop.zp_level);
	zio_buf_free(cbuf, lsize);
	BP_SET_LOGICAL_BIRTH(bp, zio->io_txg);
	zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
	ASSERT(spa_feature_is_active(spa,
	SPA_FEATURE_EMBEDDED_DATA));
	return (zio);
	} else {
	/*
	* Round compressed size up to the minimum allocation
	* size of the smallest-ashift device, and zero the
	* tail. This ensures that the compressed size of the
	* BP (and thus compressratio property) are correct,
	* in that we charge for the padding used to fill out
	* the last sector.
	*/
	size_t rounded = (size_t)zio_roundup_alloc_size(spa,
	psize);
	if (rounded >= lsize) {
	compress = ZIO_COMPRESS_OFF;
	zio_buf_free(cbuf, lsize);
	psize = lsize;
	} else {
	abd_t *cdata = abd_get_from_buf(cbuf, lsize);
	abd_take_ownership_of_buf(cdata, B_TRUE);
	abd_zero_off(cdata, psize, rounded - psize);
	psize = rounded;
	zio_push_transform(zio, cdata,
	psize, lsize, NULL);
	}
	}

	/*
	* We were unable to handle this as an override bp, treat
	* it as a regular write I/O.
	*/
	zio->io_bp_override = NULL;
	*bp = zio->io_bp_orig;
	zio->io_pipeline = zio->io_orig_pipeline;

	} else if ((zio->io_flags & ZIO_FLAG_RAW_ENCRYPT) != 0 &&
	zp->zp_type == DMU_OT_DNODE) {
	/*
	* The DMU actually relies on the zio layer's compression
	* to free metadnode blocks that have had all contained
	* dnodes freed. As a result, even when doing a raw
	* receive, we must check whether the block can be compressed
	* to a hole.
	*/
	psize = zio_compress_data(ZIO_COMPRESS_EMPTY,
	zio->io_abd, NULL, lsize, zp->zp_complevel);
	if (psize == 0 \|\| psize >= lsize)
	compress = ZIO_COMPRESS_OFF;
	} else if (zio->io_flags & ZIO_FLAG_RAW_COMPRESS &&
	!(zio->io_flags & ZIO_FLAG_RAW_ENCRYPT)) {
	/*
	* If we are raw receiving an encrypted dataset we should not
	* take this codepath because it will change the on-disk block
	* and decryption will fail.
	*/
	size_t rounded = MIN((size_t)zio_roundup_alloc_size(spa, psize),
	lsize);

	if (rounded != psize) {
	abd_t *cdata = abd_alloc_linear(rounded, B_TRUE);
	abd_zero_off(cdata, psize, rounded - psize);
	abd_copy_off(cdata, zio->io_abd, 0, 0, psize);
	psize = rounded;
	zio_push_transform(zio, cdata,
	psize, rounded, NULL);
	}
	} else {
	ASSERT3U(psize, !=, 0);
	}

	/*
	* The final pass of spa_sync() must be all rewrites, but the first
	* few passes offer a trade-off: allocating blocks defers convergence,
	* but newly allocated blocks are sequential, so they can be written
	* to disk faster. Therefore, we allow the first few passes of
	* spa_sync() to allocate new blocks, but force rewrites after that.
	* There should only be a handful of blocks after pass 1 in any case.
	*/
	if (!BP_IS_HOLE(bp) && BP_GET_LOGICAL_BIRTH(bp) == zio->io_txg &&
	BP_GET_PSIZE(bp) == psize &&
	pass >= zfs_sync_pass_rewrite) {
	VERIFY3U(psize, !=, 0);
	enum zio_stage gang_stages = zio->io_pipeline & ZIO_GANG_STAGES;

	zio->io_pipeline = ZIO_REWRITE_PIPELINE \| gang_stages;
	zio->io_flags \|= ZIO_FLAG_IO_REWRITE;
	} else {
	BP_ZERO(bp);
	zio->io_pipeline = ZIO_WRITE_PIPELINE;
	}

	if (psize == 0) {
	if (BP_GET_LOGICAL_BIRTH(&zio->io_bp_orig) != 0 &&
	spa_feature_is_active(spa, SPA_FEATURE_HOLE_BIRTH)) {
	BP_SET_LSIZE(bp, lsize);
	BP_SET_TYPE(bp, zp->zp_type);
	BP_SET_LEVEL(bp, zp->zp_level);
	BP_SET_BIRTH(bp, zio->io_txg, 0);
	}
	zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
	} else {
	ASSERT(zp->zp_checksum != ZIO_CHECKSUM_GANG_HEADER);
	BP_SET_LSIZE(bp, lsize);
	BP_SET_TYPE(bp, zp->zp_type);
	BP_SET_LEVEL(bp, zp->zp_level);
	BP_SET_PSIZE(bp, psize);
	BP_SET_COMPRESS(bp, compress);
	BP_SET_CHECKSUM(bp, zp->zp_checksum);
	BP_SET_DEDUP(bp, zp->zp_dedup);
	BP_SET_BYTEORDER(bp, ZFS_HOST_BYTEORDER);
	if (zp->zp_dedup) {
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
	ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE));
	ASSERT(!zp->zp_encrypt \|\|
	DMU_OT_IS_ENCRYPTED(zp->zp_type));
	zio->io_pipeline = ZIO_DDT_WRITE_PIPELINE;
	}
	if (zp->zp_nopwrite) {
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
	ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE));
	zio->io_pipeline \|= ZIO_STAGE_NOP_WRITE;
	}
	}
	return (zio);
	}

	static zio_t *
	zio_free_bp_init(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;

	if (zio->io_child_type == ZIO_CHILD_LOGICAL) {
	if (BP_GET_DEDUP(bp))
	zio->io_pipeline = ZIO_DDT_FREE_PIPELINE;
	}

	ASSERT3P(zio->io_bp, ==, &zio->io_bp_copy);

	return (zio);
	}

	/*
	* ==========================================================================
	* Execute the I/O pipeline
	* ==========================================================================
	*/

	static void
	zio_taskq_dispatch(zio_t *zio, zio_taskq_type_t q, boolean_t cutinline)
	{
	spa_t *spa = zio->io_spa;
	zio_type_t t = zio->io_type;
	- int flags = (cutinline ? TQ_FRONT : 0);

	/*
	* If we're a config writer or a probe, the normal issue and
	* interrupt threads may all be blocked waiting for the config lock.
	* In this case, select the otherwise-unused taskq for ZIO_TYPE_NULL.
	*/
	if (zio->io_flags & (ZIO_FLAG_CONFIG_WRITER \| ZIO_FLAG_PROBE))
	t = ZIO_TYPE_NULL;

	/*
	* A similar issue exists for the L2ARC write thread until L2ARC 2.0.
	*/
	if (t == ZIO_TYPE_WRITE && zio->io_vd && zio->io_vd->vdev_aux)
	t = ZIO_TYPE_NULL;

	/*
	* If this is a high priority I/O, then use the high priority taskq if
	- * available.
	+ * available or cut the line otherwise.
	*/
	- if ((zio->io_priority == ZIO_PRIORITY_NOW \|\|
	- zio->io_priority == ZIO_PRIORITY_SYNC_WRITE) &&
	- spa->spa_zio_taskq[t][q + 1].stqs_count != 0)
	- q++;
	+ if (zio->io_priority == ZIO_PRIORITY_SYNC_WRITE) {
	+ if (spa->spa_zio_taskq[t][q + 1].stqs_count != 0)
	+ q++;
	+ else
	+ cutinline = B_TRUE;
	+ }

	ASSERT3U(q, <, ZIO_TASKQ_TYPES);

	- /*
	- * NB: We are assuming that the zio can only be dispatched
	- * to a single taskq at a time. It would be a grievous error
	- * to dispatch the zio to another taskq at the same time.
	- */
	- ASSERT(taskq_empty_ent(&zio->io_tqent));
	- spa_taskq_dispatch_ent(spa, t, q, zio_execute, zio, flags,
	- &zio->io_tqent, zio);
	+ spa_taskq_dispatch(spa, t, q, zio_execute, zio, cutinline);
	}

	static boolean_t
	zio_taskq_member(zio_t *zio, zio_taskq_type_t q)
	{
	spa_t *spa = zio->io_spa;

	taskq_t *tq = taskq_of_curthread();

	for (zio_type_t t = 0; t < ZIO_TYPES; t++) {
	spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];
	uint_t i;
	for (i = 0; i < tqs->stqs_count; i++) {
	if (tqs->stqs_taskq[i] == tq)
	return (B_TRUE);
	}
	}

	return (B_FALSE);
	}

	static zio_t *
	zio_issue_async(zio_t *zio)
	{
	ASSERT((zio->io_type != ZIO_TYPE_WRITE) \|\| ZIO_HAS_ALLOCATOR(zio));
	zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, B_FALSE);
	return (NULL);
	}

	void
	zio_interrupt(void *zio)
	{
	zio_taskq_dispatch(zio, ZIO_TASKQ_INTERRUPT, B_FALSE);
	}

	void
	zio_delay_interrupt(zio_t *zio)
	{
	/*
	* The timeout_generic() function isn't defined in userspace, so
	* rather than trying to implement the function, the zio delay
	* functionality has been disabled for userspace builds.
	*/

	#ifdef _KERNEL
	/*
	* If io_target_timestamp is zero, then no delay has been registered
	* for this IO, thus jump to the end of this function and "skip" the
	* delay; issuing it directly to the zio layer.
	*/
	if (zio->io_target_timestamp != 0) {
	hrtime_t now = gethrtime();

	if (now >= zio->io_target_timestamp) {
	/*
	* This IO has already taken longer than the target
	* delay to complete, so we don't want to delay it
	* any longer; we "miss" the delay and issue it
	* directly to the zio layer. This is likely due to
	* the target latency being set to a value less than
	* the underlying hardware can satisfy (e.g. delay
	* set to 1ms, but the disks take 10ms to complete an
	* IO request).
	*/

	DTRACE_PROBE2(zio__delay__miss, zio_t *, zio,
	hrtime_t, now);

	zio_interrupt(zio);
	} else {
	taskqid_t tid;
	hrtime_t diff = zio->io_target_timestamp - now;
	clock_t expire_at_tick = ddi_get_lbolt() +
	NSEC_TO_TICK(diff);

	DTRACE_PROBE3(zio__delay__hit, zio_t *, zio,
	hrtime_t, now, hrtime_t, diff);

	if (NSEC_TO_TICK(diff) == 0) {
	/* Our delay is less than a jiffy - just spin */
	zfs_sleep_until(zio->io_target_timestamp);
	zio_interrupt(zio);
	} else {
	/*
	* Use taskq_dispatch_delay() in the place of
	* OpenZFS's timeout_generic().
	*/
	tid = taskq_dispatch_delay(system_taskq,
	zio_interrupt, zio, TQ_NOSLEEP,
	expire_at_tick);
	if (tid == TASKQID_INVALID) {
	/*
	* Couldn't allocate a task. Just
	* finish the zio without a delay.
	*/
	zio_interrupt(zio);
	}
	}
	}
	return;
	}
	#endif
	DTRACE_PROBE1(zio__delay__skip, zio_t *, zio);
	zio_interrupt(zio);
	}

	static void
	zio_deadman_impl(zio_t *pio, int ziodepth)
	{
	zio_t cio, cio_next;
	zio_link_t *zl = NULL;
	vdev_t *vd = pio->io_vd;

	if (zio_deadman_log_all \|\| (vd != NULL && vd->vdev_ops->vdev_op_leaf)) {
	vdev_queue_t *vq = vd ? &vd->vdev_queue : NULL;
	zbookmark_phys_t *zb = &pio->io_bookmark;
	uint64_t delta = gethrtime() - pio->io_timestamp;
	uint64_t failmode = spa_get_deadman_failmode(pio->io_spa);

	zfs_dbgmsg("slow zio[%d]: zio=%px timestamp=%llu "
	"delta=%llu queued=%llu io=%llu "
	"path=%s "
	"last=%llu type=%d "
	"priority=%d flags=0x%llx stage=0x%x "
	"pipeline=0x%x pipeline-trace=0x%x "
	"objset=%llu object=%llu "
	"level=%llu blkid=%llu "
	"offset=%llu size=%llu "
	"error=%d",
	ziodepth, pio, pio->io_timestamp,
	(u_longlong_t)delta, pio->io_delta, pio->io_delay,
	vd ? vd->vdev_path : "NULL",
	vq ? vq->vq_io_complete_ts : 0, pio->io_type,
	pio->io_priority, (u_longlong_t)pio->io_flags,
	pio->io_stage, pio->io_pipeline, pio->io_pipeline_trace,
	(u_longlong_t)zb->zb_objset, (u_longlong_t)zb->zb_object,
	(u_longlong_t)zb->zb_level, (u_longlong_t)zb->zb_blkid,
	(u_longlong_t)pio->io_offset, (u_longlong_t)pio->io_size,
	pio->io_error);
	(void) zfs_ereport_post(FM_EREPORT_ZFS_DEADMAN,
	pio->io_spa, vd, zb, pio, 0);

	if (failmode == ZIO_FAILURE_MODE_CONTINUE &&
	taskq_empty_ent(&pio->io_tqent)) {
	zio_interrupt(pio);
	}
	}

	mutex_enter(&pio->io_lock);
	for (cio = zio_walk_children(pio, &zl); cio != NULL; cio = cio_next) {
	cio_next = zio_walk_children(pio, &zl);
	zio_deadman_impl(cio, ziodepth + 1);
	}
	mutex_exit(&pio->io_lock);
	}

	/*
	* Log the critical information describing this zio and all of its children
	* using the zfs_dbgmsg() interface then post deadman event for the ZED.
	*/
	void
	zio_deadman(zio_t pio, const char tag)
	{
	spa_t *spa = pio->io_spa;
	char *name = spa_name(spa);

	if (!zfs_deadman_enabled \|\| spa_suspended(spa))
	return;

	zio_deadman_impl(pio, 0);

	switch (spa_get_deadman_failmode(spa)) {
	case ZIO_FAILURE_MODE_WAIT:
	zfs_dbgmsg("%s waiting for hung I/O to pool '%s'", tag, name);
	break;

	case ZIO_FAILURE_MODE_CONTINUE:
	zfs_dbgmsg("%s restarting hung I/O for pool '%s'", tag, name);
	break;

	case ZIO_FAILURE_MODE_PANIC:
	fm_panic("%s determined I/O to pool '%s' is hung.", tag, name);
	break;
	}
	}

	/*
	* Execute the I/O pipeline until one of the following occurs:
	* (1) the I/O completes; (2) the pipeline stalls waiting for
	* dependent child I/Os; (3) the I/O issues, so we're waiting
	* for an I/O completion interrupt; (4) the I/O is delegated by
	* vdev-level caching or aggregation; (5) the I/O is deferred
	* due to vdev-level queueing; (6) the I/O is handed off to
	* another thread. In all cases, the pipeline stops whenever
	* there's no CPU work; it never burns a thread in cv_wait_io().
	*
	* There's no locking on io_stage because there's no legitimate way
	* for multiple threads to be attempting to process the same I/O.
	*/
	static zio_pipe_stage_t *zio_pipeline[];

	/*
	* zio_execute() is a wrapper around the static function
	* __zio_execute() so that we can force __zio_execute() to be
	* inlined. This reduces stack overhead which is important
	* because __zio_execute() is called recursively in several zio
	* code paths. zio_execute() itself cannot be inlined because
	* it is externally visible.
	*/
	void
	zio_execute(void *zio)
	{
	fstrans_cookie_t cookie;

	cookie = spl_fstrans_mark();
	__zio_execute(zio);
	spl_fstrans_unmark(cookie);
	}

	/*
	* Used to determine if in the current context the stack is sized large
	* enough to allow zio_execute() to be called recursively. A minimum
	* stack size of 16K is required to avoid needing to re-dispatch the zio.
	*/
	static boolean_t
	zio_execute_stack_check(zio_t *zio)
	{
	#if !defined(HAVE_LARGE_STACKS)
	dsl_pool_t *dp = spa_get_dsl(zio->io_spa);

	/* Executing in txg_sync_thread() context. */
	if (dp && curthread == dp->dp_tx.tx_sync_thread)
	return (B_TRUE);

	/* Pool initialization outside of zio_taskq context. */
	if (dp && spa_is_initializing(dp->dp_spa) &&
	!zio_taskq_member(zio, ZIO_TASKQ_ISSUE) &&
	!zio_taskq_member(zio, ZIO_TASKQ_ISSUE_HIGH))
	return (B_TRUE);
	#else
	(void) zio;
	#endif /* HAVE_LARGE_STACKS */

	return (B_FALSE);
	}

	__attribute__((always_inline))
	static inline void
	__zio_execute(zio_t *zio)
	{
	ASSERT3U(zio->io_queued_timestamp, >, 0);

	while (zio->io_stage < ZIO_STAGE_DONE) {
	enum zio_stage pipeline = zio->io_pipeline;
	enum zio_stage stage = zio->io_stage;

	zio->io_executor = curthread;

	ASSERT(!MUTEX_HELD(&zio->io_lock));
	ASSERT(ISP2(stage));
	ASSERT(zio->io_stall == NULL);

	do {
	stage <<= 1;
	} while ((stage & pipeline) == 0);

	ASSERT(stage <= ZIO_STAGE_DONE);

	/*
	* If we are in interrupt context and this pipeline stage
	* will grab a config lock that is held across I/O,
	* or may wait for an I/O that needs an interrupt thread
	* to complete, issue async to avoid deadlock.
	*
	* For VDEV_IO_START, we cut in line so that the io will
	* be sent to disk promptly.
	*/
	if ((stage & ZIO_BLOCKING_STAGES) && zio->io_vd == NULL &&
	zio_taskq_member(zio, ZIO_TASKQ_INTERRUPT)) {
	boolean_t cut = (stage == ZIO_STAGE_VDEV_IO_START) ?
	zio_requeue_io_start_cut_in_line : B_FALSE;
	zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, cut);
	return;
	}

	/*
	* If the current context doesn't have large enough stacks
	* the zio must be issued asynchronously to prevent overflow.
	*/
	if (zio_execute_stack_check(zio)) {
	boolean_t cut = (stage == ZIO_STAGE_VDEV_IO_START) ?
	zio_requeue_io_start_cut_in_line : B_FALSE;
	zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, cut);
	return;
	}

	zio->io_stage = stage;
	zio->io_pipeline_trace \|= zio->io_stage;

	/*
	* The zio pipeline stage returns the next zio to execute
	* (typically the same as this one), or NULL if we should
	* stop.
	*/
	zio = zio_pipeline[highbit64(stage) - 1](zio);

	if (zio == NULL)
	return;
	}
	}


	/*
	* ==========================================================================
	* Initiate I/O, either sync or async
	* ==========================================================================
	*/
	int
	zio_wait(zio_t *zio)
	{
	/*
	* Some routines, like zio_free_sync(), may return a NULL zio
	* to avoid the performance overhead of creating and then destroying
	* an unneeded zio. For the callers' simplicity, we accept a NULL
	* zio and ignore it.
	*/
	if (zio == NULL)
	return (0);

	long timeout = MSEC_TO_TICK(zfs_deadman_ziotime_ms);
	int error;

	ASSERT3S(zio->io_stage, ==, ZIO_STAGE_OPEN);
	ASSERT3P(zio->io_executor, ==, NULL);

	zio->io_waiter = curthread;
	ASSERT0(zio->io_queued_timestamp);
	zio->io_queued_timestamp = gethrtime();

	if (zio->io_type == ZIO_TYPE_WRITE) {
	spa_select_allocator(zio);
	}
	__zio_execute(zio);

	mutex_enter(&zio->io_lock);
	while (zio->io_executor != NULL) {
	error = cv_timedwait_io(&zio->io_cv, &zio->io_lock,
	ddi_get_lbolt() + timeout);

	if (zfs_deadman_enabled && error == -1 &&
	gethrtime() - zio->io_queued_timestamp >
	spa_deadman_ziotime(zio->io_spa)) {
	mutex_exit(&zio->io_lock);
	timeout = MSEC_TO_TICK(zfs_deadman_checktime_ms);
	zio_deadman(zio, FTAG);
	mutex_enter(&zio->io_lock);
	}
	}
	mutex_exit(&zio->io_lock);

	error = zio->io_error;
	zio_destroy(zio);

	return (error);
	}

	void
	zio_nowait(zio_t *zio)
	{
	/*
	* See comment in zio_wait().
	*/
	if (zio == NULL)
	return;

	ASSERT3P(zio->io_executor, ==, NULL);

	if (zio->io_child_type == ZIO_CHILD_LOGICAL &&
	list_is_empty(&zio->io_parent_list)) {
	zio_t *pio;

	/*
	* This is a logical async I/O with no parent to wait for it.
	* We add it to the spa_async_root_zio "Godfather" I/O which
	* will ensure they complete prior to unloading the pool.
	*/
	spa_t *spa = zio->io_spa;
	pio = spa->spa_async_zio_root[CPU_SEQID_UNSTABLE];

	zio_add_child(pio, zio);
	}

	ASSERT0(zio->io_queued_timestamp);
	zio->io_queued_timestamp = gethrtime();
	if (zio->io_type == ZIO_TYPE_WRITE) {
	spa_select_allocator(zio);
	}
	__zio_execute(zio);
	}

	/*
	* ==========================================================================
	* Reexecute, cancel, or suspend/resume failed I/O
	* ==========================================================================
	*/

	static void
	zio_reexecute(void *arg)
	{
	zio_t *pio = arg;
	zio_t cio, cio_next, *gio;

	ASSERT(pio->io_child_type == ZIO_CHILD_LOGICAL);
	ASSERT(pio->io_orig_stage == ZIO_STAGE_OPEN);
	ASSERT(pio->io_gang_leader == NULL);
	ASSERT(pio->io_gang_tree == NULL);

	mutex_enter(&pio->io_lock);
	pio->io_flags = pio->io_orig_flags;
	pio->io_stage = pio->io_orig_stage;
	pio->io_pipeline = pio->io_orig_pipeline;
	pio->io_reexecute = 0;
	pio->io_flags \|= ZIO_FLAG_REEXECUTED;
	pio->io_pipeline_trace = 0;
	pio->io_error = 0;
	pio->io_state[ZIO_WAIT_READY] = (pio->io_stage >= ZIO_STAGE_READY) \|\|
	(pio->io_pipeline & ZIO_STAGE_READY) == 0;
	pio->io_state[ZIO_WAIT_DONE] = (pio->io_stage >= ZIO_STAGE_DONE);
	zio_link_t *zl = NULL;
	while ((gio = zio_walk_parents(pio, &zl)) != NULL) {
	for (int w = 0; w < ZIO_WAIT_TYPES; w++) {
	gio->io_children[pio->io_child_type][w] +=
	!pio->io_state[w];
	}
	}
	for (int c = 0; c < ZIO_CHILD_TYPES; c++)
	pio->io_child_error[c] = 0;

	if (IO_IS_ALLOCATING(pio))
	BP_ZERO(pio->io_bp);

	/*
	* As we reexecute pio's children, new children could be created.
	* New children go to the head of pio's io_child_list, however,
	* so we will (correctly) not reexecute them. The key is that
	* the remainder of pio's io_child_list, from 'cio_next' onward,
	* cannot be affected by any side effects of reexecuting 'cio'.
	*/
	zl = NULL;
	for (cio = zio_walk_children(pio, &zl); cio != NULL; cio = cio_next) {
	cio_next = zio_walk_children(pio, &zl);
	mutex_exit(&pio->io_lock);
	zio_reexecute(cio);
	mutex_enter(&pio->io_lock);
	}
	mutex_exit(&pio->io_lock);

	/*
	* Now that all children have been reexecuted, execute the parent.
	* We don't reexecute "The Godfather" I/O here as it's the
	* responsibility of the caller to wait on it.
	*/
	if (!(pio->io_flags & ZIO_FLAG_GODFATHER)) {
	pio->io_queued_timestamp = gethrtime();
	__zio_execute(pio);
	}
	}

	void
	zio_suspend(spa_t spa, zio_t zio, zio_suspend_reason_t reason)
	{
	if (spa_get_failmode(spa) == ZIO_FAILURE_MODE_PANIC)
	fm_panic("Pool '%s' has encountered an uncorrectable I/O "
	"failure and the failure mode property for this pool "
	"is set to panic.", spa_name(spa));

	if (reason != ZIO_SUSPEND_MMP) {
	cmn_err(CE_WARN, "Pool '%s' has encountered an uncorrectable "
	"I/O failure and has been suspended.\n", spa_name(spa));
	}

	(void) zfs_ereport_post(FM_EREPORT_ZFS_IO_FAILURE, spa, NULL,
	NULL, NULL, 0);

	mutex_enter(&spa->spa_suspend_lock);

	if (spa->spa_suspend_zio_root == NULL)
	spa->spa_suspend_zio_root = zio_root(spa, NULL, NULL,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_SPECULATIVE \|
	ZIO_FLAG_GODFATHER);

	spa->spa_suspended = reason;

	if (zio != NULL) {
	ASSERT(!(zio->io_flags & ZIO_FLAG_GODFATHER));
	ASSERT(zio != spa->spa_suspend_zio_root);
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
	ASSERT(zio_unique_parent(zio) == NULL);
	ASSERT(zio->io_stage == ZIO_STAGE_DONE);
	zio_add_child(spa->spa_suspend_zio_root, zio);
	}

	mutex_exit(&spa->spa_suspend_lock);
	}

	int
	zio_resume(spa_t *spa)
	{
	zio_t *pio;

	/*
	* Reexecute all previously suspended i/o.
	*/
	mutex_enter(&spa->spa_suspend_lock);
	spa->spa_suspended = ZIO_SUSPEND_NONE;
	cv_broadcast(&spa->spa_suspend_cv);
	pio = spa->spa_suspend_zio_root;
	spa->spa_suspend_zio_root = NULL;
	mutex_exit(&spa->spa_suspend_lock);

	if (pio == NULL)
	return (0);

	zio_reexecute(pio);
	return (zio_wait(pio));
	}

	void
	zio_resume_wait(spa_t *spa)
	{
	mutex_enter(&spa->spa_suspend_lock);
	while (spa_suspended(spa))
	cv_wait(&spa->spa_suspend_cv, &spa->spa_suspend_lock);
	mutex_exit(&spa->spa_suspend_lock);
	}

	/*
	* ==========================================================================
	* Gang blocks.
	*
	* A gang block is a collection of small blocks that looks to the DMU
	* like one large block. When zio_dva_allocate() cannot find a block
	* of the requested size, due to either severe fragmentation or the pool
	* being nearly full, it calls zio_write_gang_block() to construct the
	* block from smaller fragments.
	*
	* A gang block consists of a gang header (zio_gbh_phys_t) and up to
	* three (SPA_GBH_NBLKPTRS) gang members. The gang header is just like
	* an indirect block: it's an array of block pointers. It consumes
	* only one sector and hence is allocatable regardless of fragmentation.
	* The gang header's bps point to its gang members, which hold the data.
	*
	* Gang blocks are self-checksumming, using the bp's <vdev, offset, txg>
	* as the verifier to ensure uniqueness of the SHA256 checksum.
	* Critically, the gang block bp's blk_cksum is the checksum of the data,
	* not the gang header. This ensures that data block signatures (needed for
	* deduplication) are independent of how the block is physically stored.
	*
	* Gang blocks can be nested: a gang member may itself be a gang block.
	* Thus every gang block is a tree in which root and all interior nodes are
	* gang headers, and the leaves are normal blocks that contain user data.
	* The root of the gang tree is called the gang leader.
	*
	* To perform any operation (read, rewrite, free, claim) on a gang block,
	* zio_gang_assemble() first assembles the gang tree (minus data leaves)
	* in the io_gang_tree field of the original logical i/o by recursively
	* reading the gang leader and all gang headers below it. This yields
	* an in-core tree containing the contents of every gang header and the
	* bps for every constituent of the gang block.
	*
	* With the gang tree now assembled, zio_gang_issue() just walks the gang tree
	* and invokes a callback on each bp. To free a gang block, zio_gang_issue()
	* calls zio_free_gang() -- a trivial wrapper around zio_free() -- for each bp.
	* zio_claim_gang() provides a similarly trivial wrapper for zio_claim().
	* zio_read_gang() is a wrapper around zio_read() that omits reading gang
	* headers, since we already have those in io_gang_tree. zio_rewrite_gang()
	* performs a zio_rewrite() of the data or, for gang headers, a zio_rewrite()
	* of the gang header plus zio_checksum_compute() of the data to update the
	* gang header's blk_cksum as described above.
	*
	* The two-phase assemble/issue model solves the problem of partial failure --
	* what if you'd freed part of a gang block but then couldn't read the
	* gang header for another part? Assembling the entire gang tree first
	* ensures that all the necessary gang header I/O has succeeded before
	* starting the actual work of free, claim, or write. Once the gang tree
	* is assembled, free and claim are in-memory operations that cannot fail.
	*
	* In the event that a gang write fails, zio_dva_unallocate() walks the
	* gang tree to immediately free (i.e. insert back into the space map)
	* everything we've allocated. This ensures that we don't get ENOSPC
	* errors during repeated suspend/resume cycles due to a flaky device.
	*
	* Gang rewrites only happen during sync-to-convergence. If we can't assemble
	* the gang tree, we won't modify the block, so we can safely defer the free
	* (knowing that the block is still intact). If we can assemble the gang
	* tree, then even if some of the rewrites fail, zio_dva_unallocate() will free
	* each constituent bp and we can allocate a new block on the next sync pass.
	*
	* In all cases, the gang tree allows complete recovery from partial failure.
	* ==========================================================================
	*/

	static void
	zio_gang_issue_func_done(zio_t *zio)
	{
	abd_free(zio->io_abd);
	}

	static zio_t *
	zio_read_gang(zio_t pio, blkptr_t bp, zio_gang_node_t gn, abd_t data,
	uint64_t offset)
	{
	if (gn != NULL)
	return (pio);

	return (zio_read(pio, pio->io_spa, bp, abd_get_offset(data, offset),
	BP_GET_PSIZE(bp), zio_gang_issue_func_done,
	NULL, pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio),
	&pio->io_bookmark));
	}

	static zio_t *
	zio_rewrite_gang(zio_t pio, blkptr_t bp, zio_gang_node_t gn, abd_t data,
	uint64_t offset)
	{
	zio_t *zio;

	if (gn != NULL) {
	abd_t *gbh_abd =
	abd_get_from_buf(gn->gn_gbh, SPA_GANGBLOCKSIZE);
	zio = zio_rewrite(pio, pio->io_spa, pio->io_txg, bp,
	gbh_abd, SPA_GANGBLOCKSIZE, zio_gang_issue_func_done, NULL,
	pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio),
	&pio->io_bookmark);
	/*
	* As we rewrite each gang header, the pipeline will compute
	* a new gang block header checksum for it; but no one will
	* compute a new data checksum, so we do that here. The one
	* exception is the gang leader: the pipeline already computed
	* its data checksum because that stage precedes gang assembly.
	* (Presently, nothing actually uses interior data checksums;
	* this is just good hygiene.)
	*/
	if (gn != pio->io_gang_leader->io_gang_tree) {
	abd_t *buf = abd_get_offset(data, offset);

	zio_checksum_compute(zio, BP_GET_CHECKSUM(bp),
	buf, BP_GET_PSIZE(bp));

	abd_free(buf);
	}
	/*
	* If we are here to damage data for testing purposes,
	* leave the GBH alone so that we can detect the damage.
	*/
	if (pio->io_gang_leader->io_flags & ZIO_FLAG_INDUCE_DAMAGE)
	zio->io_pipeline &= ~ZIO_VDEV_IO_STAGES;
	} else {
	zio = zio_rewrite(pio, pio->io_spa, pio->io_txg, bp,
	abd_get_offset(data, offset), BP_GET_PSIZE(bp),
	zio_gang_issue_func_done, NULL, pio->io_priority,
	ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark);
	}

	return (zio);
	}

	static zio_t *
	zio_free_gang(zio_t pio, blkptr_t bp, zio_gang_node_t gn, abd_t data,
	uint64_t offset)
	{
	(void) gn, (void) data, (void) offset;

	zio_t *zio = zio_free_sync(pio, pio->io_spa, pio->io_txg, bp,
	ZIO_GANG_CHILD_FLAGS(pio));
	if (zio == NULL) {
	zio = zio_null(pio, pio->io_spa,
	NULL, NULL, NULL, ZIO_GANG_CHILD_FLAGS(pio));
	}
	return (zio);
	}

	static zio_t *
	zio_claim_gang(zio_t pio, blkptr_t bp, zio_gang_node_t gn, abd_t data,
	uint64_t offset)
	{
	(void) gn, (void) data, (void) offset;
	return (zio_claim(pio, pio->io_spa, pio->io_txg, bp,
	NULL, NULL, ZIO_GANG_CHILD_FLAGS(pio)));
	}

	static zio_gang_issue_func_t *zio_gang_issue_func[ZIO_TYPES] = {
	NULL,
	zio_read_gang,
	zio_rewrite_gang,
	zio_free_gang,
	zio_claim_gang,
	NULL
	};

	static void zio_gang_tree_assemble_done(zio_t *zio);

	static zio_gang_node_t *
	zio_gang_node_alloc(zio_gang_node_t **gnpp)
	{
	zio_gang_node_t *gn;

	ASSERT(*gnpp == NULL);

	gn = kmem_zalloc(sizeof (*gn), KM_SLEEP);
	gn->gn_gbh = zio_buf_alloc(SPA_GANGBLOCKSIZE);
	*gnpp = gn;

	return (gn);
	}

	static void
	zio_gang_node_free(zio_gang_node_t **gnpp)
	{
	zio_gang_node_t gn = gnpp;

	for (int g = 0; g < SPA_GBH_NBLKPTRS; g++)
	ASSERT(gn->gn_child[g] == NULL);

	zio_buf_free(gn->gn_gbh, SPA_GANGBLOCKSIZE);
	kmem_free(gn, sizeof (*gn));
	*gnpp = NULL;
	}

	static void
	zio_gang_tree_free(zio_gang_node_t **gnpp)
	{
	zio_gang_node_t gn = gnpp;

	if (gn == NULL)
	return;

	for (int g = 0; g < SPA_GBH_NBLKPTRS; g++)
	zio_gang_tree_free(&gn->gn_child[g]);

	zio_gang_node_free(gnpp);
	}

	static void
	zio_gang_tree_assemble(zio_t gio, blkptr_t bp, zio_gang_node_t **gnpp)
	{
	zio_gang_node_t *gn = zio_gang_node_alloc(gnpp);
	abd_t *gbh_abd = abd_get_from_buf(gn->gn_gbh, SPA_GANGBLOCKSIZE);

	ASSERT(gio->io_gang_leader == gio);
	ASSERT(BP_IS_GANG(bp));

	zio_nowait(zio_read(gio, gio->io_spa, bp, gbh_abd, SPA_GANGBLOCKSIZE,
	zio_gang_tree_assemble_done, gn, gio->io_priority,
	ZIO_GANG_CHILD_FLAGS(gio), &gio->io_bookmark));
	}

	static void
	zio_gang_tree_assemble_done(zio_t *zio)
	{
	zio_t *gio = zio->io_gang_leader;
	zio_gang_node_t *gn = zio->io_private;
	blkptr_t *bp = zio->io_bp;

	ASSERT(gio == zio_unique_parent(zio));
	ASSERT(list_is_empty(&zio->io_child_list));

	if (zio->io_error)
	return;

	/* this ABD was created from a linear buf in zio_gang_tree_assemble */
	if (BP_SHOULD_BYTESWAP(bp))
	byteswap_uint64_array(abd_to_buf(zio->io_abd), zio->io_size);

	ASSERT3P(abd_to_buf(zio->io_abd), ==, gn->gn_gbh);
	ASSERT(zio->io_size == SPA_GANGBLOCKSIZE);
	ASSERT(gn->gn_gbh->zg_tail.zec_magic == ZEC_MAGIC);

	abd_free(zio->io_abd);

	for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) {
	blkptr_t *gbp = &gn->gn_gbh->zg_blkptr[g];
	if (!BP_IS_GANG(gbp))
	continue;
	zio_gang_tree_assemble(gio, gbp, &gn->gn_child[g]);
	}
	}

	static void
	zio_gang_tree_issue(zio_t pio, zio_gang_node_t gn, blkptr_t bp, abd_t data,
	uint64_t offset)
	{
	zio_t *gio = pio->io_gang_leader;
	zio_t *zio;

	ASSERT(BP_IS_GANG(bp) == !!gn);
	ASSERT(BP_GET_CHECKSUM(bp) == BP_GET_CHECKSUM(gio->io_bp));
	ASSERT(BP_GET_LSIZE(bp) == BP_GET_PSIZE(bp) \|\| gn == gio->io_gang_tree);

	/*
	* If you're a gang header, your data is in gn->gn_gbh.
	* If you're a gang member, your data is in 'data' and gn == NULL.
	*/
	zio = zio_gang_issue_func[gio->io_type](pio, bp, gn, data, offset);

	if (gn != NULL) {
	ASSERT(gn->gn_gbh->zg_tail.zec_magic == ZEC_MAGIC);

	for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) {
	blkptr_t *gbp = &gn->gn_gbh->zg_blkptr[g];
	if (BP_IS_HOLE(gbp))
	continue;
	zio_gang_tree_issue(zio, gn->gn_child[g], gbp, data,
	offset);
	offset += BP_GET_PSIZE(gbp);
	}
	}

	if (gn == gio->io_gang_tree)
	ASSERT3U(gio->io_size, ==, offset);

	if (zio != pio)
	zio_nowait(zio);
	}

	static zio_t *
	zio_gang_assemble(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;

	ASSERT(BP_IS_GANG(bp) && zio->io_gang_leader == NULL);
	ASSERT(zio->io_child_type > ZIO_CHILD_GANG);

	zio->io_gang_leader = zio;

	zio_gang_tree_assemble(zio, bp, &zio->io_gang_tree);

	return (zio);
	}

	static zio_t *
	zio_gang_issue(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;

	if (zio_wait_for_children(zio, ZIO_CHILD_GANG_BIT, ZIO_WAIT_DONE)) {
	return (NULL);
	}

	ASSERT(BP_IS_GANG(bp) && zio->io_gang_leader == zio);
	ASSERT(zio->io_child_type > ZIO_CHILD_GANG);

	if (zio->io_child_error[ZIO_CHILD_GANG] == 0)
	zio_gang_tree_issue(zio, zio->io_gang_tree, bp, zio->io_abd,
	0);
	else
	zio_gang_tree_free(&zio->io_gang_tree);

	zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;

	return (zio);
	}

	static void
	zio_gang_inherit_allocator(zio_t pio, zio_t cio)
	{
	cio->io_allocator = pio->io_allocator;
	}

	static void
	zio_write_gang_member_ready(zio_t *zio)
	{
	zio_t *pio = zio_unique_parent(zio);
	dva_t *cdva = zio->io_bp->blk_dva;
	dva_t *pdva = pio->io_bp->blk_dva;
	uint64_t asize;
	zio_t *gio __maybe_unused = zio->io_gang_leader;

	if (BP_IS_HOLE(zio->io_bp))
	return;

	ASSERT(BP_IS_HOLE(&zio->io_bp_orig));

	ASSERT(zio->io_child_type == ZIO_CHILD_GANG);
	ASSERT3U(zio->io_prop.zp_copies, ==, gio->io_prop.zp_copies);
	ASSERT3U(zio->io_prop.zp_copies, <=, BP_GET_NDVAS(zio->io_bp));
	ASSERT3U(pio->io_prop.zp_copies, <=, BP_GET_NDVAS(pio->io_bp));
	VERIFY3U(BP_GET_NDVAS(zio->io_bp), <=, BP_GET_NDVAS(pio->io_bp));

	mutex_enter(&pio->io_lock);
	for (int d = 0; d < BP_GET_NDVAS(zio->io_bp); d++) {
	ASSERT(DVA_GET_GANG(&pdva[d]));
	asize = DVA_GET_ASIZE(&pdva[d]);
	asize += DVA_GET_ASIZE(&cdva[d]);
	DVA_SET_ASIZE(&pdva[d], asize);
	}
	mutex_exit(&pio->io_lock);
	}

	static void
	zio_write_gang_done(zio_t *zio)
	{
	/*
	* The io_abd field will be NULL for a zio with no data. The io_flags
	* will initially have the ZIO_FLAG_NODATA bit flag set, but we can't
	* check for it here as it is cleared in zio_ready.
	*/
	if (zio->io_abd != NULL)
	abd_free(zio->io_abd);
	}

	static zio_t *
	zio_write_gang_block(zio_t pio, metaslab_class_t mc)
	{
	spa_t *spa = pio->io_spa;
	blkptr_t *bp = pio->io_bp;
	zio_t *gio = pio->io_gang_leader;
	zio_t *zio;
	zio_gang_node_t gn, *gnpp;
	zio_gbh_phys_t *gbh;
	abd_t *gbh_abd;
	uint64_t txg = pio->io_txg;
	uint64_t resid = pio->io_size;
	uint64_t lsize;
	int copies = gio->io_prop.zp_copies;
	zio_prop_t zp;
	int error;
	boolean_t has_data = !(pio->io_flags & ZIO_FLAG_NODATA);

	/*
	* If one copy was requested, store 2 copies of the GBH, so that we
	* can still traverse all the data (e.g. to free or scrub) even if a
	* block is damaged. Note that we can't store 3 copies of the GBH in
	* all cases, e.g. with encryption, which uses DVA[2] for the IV+salt.
	*/
	int gbh_copies = copies;
	if (gbh_copies == 1) {
	gbh_copies = MIN(2, spa_max_replication(spa));
	}

	ASSERT(ZIO_HAS_ALLOCATOR(pio));
	int flags = METASLAB_HINTBP_FAVOR \| METASLAB_GANG_HEADER;
	if (pio->io_flags & ZIO_FLAG_IO_ALLOCATING) {
	ASSERT(pio->io_priority == ZIO_PRIORITY_ASYNC_WRITE);
	ASSERT(has_data);

	flags \|= METASLAB_ASYNC_ALLOC;
	VERIFY(zfs_refcount_held(&mc->mc_allocator[pio->io_allocator].
	mca_alloc_slots, pio));

	/*
	* The logical zio has already placed a reservation for
	* 'copies' allocation slots but gang blocks may require
	* additional copies. These additional copies
	* (i.e. gbh_copies - copies) are guaranteed to succeed
	* since metaslab_class_throttle_reserve() always allows
	* additional reservations for gang blocks.
	*/
	VERIFY(metaslab_class_throttle_reserve(mc, gbh_copies - copies,
	pio->io_allocator, pio, flags));
	}

	error = metaslab_alloc(spa, mc, SPA_GANGBLOCKSIZE,
	bp, gbh_copies, txg, pio == gio ? NULL : gio->io_bp, flags,
	&pio->io_alloc_list, pio, pio->io_allocator);
	if (error) {
	if (pio->io_flags & ZIO_FLAG_IO_ALLOCATING) {
	ASSERT(pio->io_priority == ZIO_PRIORITY_ASYNC_WRITE);
	ASSERT(has_data);

	/*
	* If we failed to allocate the gang block header then
	* we remove any additional allocation reservations that
	* we placed here. The original reservation will
	* be removed when the logical I/O goes to the ready
	* stage.
	*/
	metaslab_class_throttle_unreserve(mc,
	gbh_copies - copies, pio->io_allocator, pio);
	}

	pio->io_error = error;
	return (pio);
	}

	if (pio == gio) {
	gnpp = &gio->io_gang_tree;
	} else {
	gnpp = pio->io_private;
	ASSERT(pio->io_ready == zio_write_gang_member_ready);
	}

	gn = zio_gang_node_alloc(gnpp);
	gbh = gn->gn_gbh;
	memset(gbh, 0, SPA_GANGBLOCKSIZE);
	gbh_abd = abd_get_from_buf(gbh, SPA_GANGBLOCKSIZE);

	/*
	* Create the gang header.
	*/
	zio = zio_rewrite(pio, spa, txg, bp, gbh_abd, SPA_GANGBLOCKSIZE,
	zio_write_gang_done, NULL, pio->io_priority,
	ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark);

	zio_gang_inherit_allocator(pio, zio);

	/*
	* Create and nowait the gang children.
	*/
	for (int g = 0; resid != 0; resid -= lsize, g++) {
	lsize = P2ROUNDUP(resid / (SPA_GBH_NBLKPTRS - g),
	SPA_MINBLOCKSIZE);
	ASSERT(lsize >= SPA_MINBLOCKSIZE && lsize <= resid);

	zp.zp_checksum = gio->io_prop.zp_checksum;
	zp.zp_compress = ZIO_COMPRESS_OFF;
	zp.zp_complevel = gio->io_prop.zp_complevel;
	zp.zp_type = DMU_OT_NONE;
	zp.zp_level = 0;
	zp.zp_copies = gio->io_prop.zp_copies;
	zp.zp_dedup = B_FALSE;
	zp.zp_dedup_verify = B_FALSE;
	zp.zp_nopwrite = B_FALSE;
	zp.zp_encrypt = gio->io_prop.zp_encrypt;
	zp.zp_byteorder = gio->io_prop.zp_byteorder;
	memset(zp.zp_salt, 0, ZIO_DATA_SALT_LEN);
	memset(zp.zp_iv, 0, ZIO_DATA_IV_LEN);
	memset(zp.zp_mac, 0, ZIO_DATA_MAC_LEN);

	zio_t *cio = zio_write(zio, spa, txg, &gbh->zg_blkptr[g],
	has_data ? abd_get_offset(pio->io_abd, pio->io_size -
	resid) : NULL, lsize, lsize, &zp,
	zio_write_gang_member_ready, NULL,
	zio_write_gang_done, &gn->gn_child[g], pio->io_priority,
	ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark);

	zio_gang_inherit_allocator(zio, cio);

	if (pio->io_flags & ZIO_FLAG_IO_ALLOCATING) {
	ASSERT(pio->io_priority == ZIO_PRIORITY_ASYNC_WRITE);
	ASSERT(has_data);

	/*
	* Gang children won't throttle but we should
	* account for their work, so reserve an allocation
	* slot for them here.
	*/
	VERIFY(metaslab_class_throttle_reserve(mc,
	zp.zp_copies, cio->io_allocator, cio, flags));
	}
	zio_nowait(cio);
	}

	/*
	* Set pio's pipeline to just wait for zio to finish.
	*/
	pio->io_pipeline = ZIO_INTERLOCK_PIPELINE;

	zio_nowait(zio);

	return (pio);
	}

	/*
	* The zio_nop_write stage in the pipeline determines if allocating a
	* new bp is necessary. The nopwrite feature can handle writes in
	* either syncing or open context (i.e. zil writes) and as a result is
	* mutually exclusive with dedup.
	*
	* By leveraging a cryptographically secure checksum, such as SHA256, we
	* can compare the checksums of the new data and the old to determine if
	* allocating a new block is required. Note that our requirements for
	* cryptographic strength are fairly weak: there can't be any accidental
	* hash collisions, but we don't need to be secure against intentional
	* (malicious) collisions. To trigger a nopwrite, you have to be able
	* to write the file to begin with, and triggering an incorrect (hash
	* collision) nopwrite is no worse than simply writing to the file.
	* That said, there are no known attacks against the checksum algorithms
	* used for nopwrite, assuming that the salt and the checksums
	* themselves remain secret.
	*/
	static zio_t *
	zio_nop_write(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;
	blkptr_t *bp_orig = &zio->io_bp_orig;
	zio_prop_t *zp = &zio->io_prop;

	ASSERT(BP_IS_HOLE(bp));
	ASSERT(BP_GET_LEVEL(bp) == 0);
	ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE));
	ASSERT(zp->zp_nopwrite);
	ASSERT(!zp->zp_dedup);
	ASSERT(zio->io_bp_override == NULL);
	ASSERT(IO_IS_ALLOCATING(zio));

	/*
	* Check to see if the original bp and the new bp have matching
	* characteristics (i.e. same checksum, compression algorithms, etc).
	* If they don't then just continue with the pipeline which will
	* allocate a new bp.
	*/
	if (BP_IS_HOLE(bp_orig) \|\|
	!(zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_flags &
	ZCHECKSUM_FLAG_NOPWRITE) \|\|
	BP_IS_ENCRYPTED(bp) \|\| BP_IS_ENCRYPTED(bp_orig) \|\|
	BP_GET_CHECKSUM(bp) != BP_GET_CHECKSUM(bp_orig) \|\|
	BP_GET_COMPRESS(bp) != BP_GET_COMPRESS(bp_orig) \|\|
	BP_GET_DEDUP(bp) != BP_GET_DEDUP(bp_orig) \|\|
	zp->zp_copies != BP_GET_NDVAS(bp_orig))
	return (zio);

	/*
	* If the checksums match then reset the pipeline so that we
	* avoid allocating a new bp and issuing any I/O.
	*/
	if (ZIO_CHECKSUM_EQUAL(bp->blk_cksum, bp_orig->blk_cksum)) {
	ASSERT(zio_checksum_table[zp->zp_checksum].ci_flags &
	ZCHECKSUM_FLAG_NOPWRITE);
	ASSERT3U(BP_GET_PSIZE(bp), ==, BP_GET_PSIZE(bp_orig));
	ASSERT3U(BP_GET_LSIZE(bp), ==, BP_GET_LSIZE(bp_orig));
	ASSERT(zp->zp_compress != ZIO_COMPRESS_OFF);
	ASSERT3U(bp->blk_prop, ==, bp_orig->blk_prop);

	/*
	* If we're overwriting a block that is currently on an
	* indirect vdev, then ignore the nopwrite request and
	* allow a new block to be allocated on a concrete vdev.
	*/
	spa_config_enter(zio->io_spa, SCL_VDEV, FTAG, RW_READER);
	for (int d = 0; d < BP_GET_NDVAS(bp_orig); d++) {
	vdev_t *tvd = vdev_lookup_top(zio->io_spa,
	DVA_GET_VDEV(&bp_orig->blk_dva[d]));
	if (tvd->vdev_ops == &vdev_indirect_ops) {
	spa_config_exit(zio->io_spa, SCL_VDEV, FTAG);
	return (zio);
	}
	}
	spa_config_exit(zio->io_spa, SCL_VDEV, FTAG);

	bp = bp_orig;
	zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
	zio->io_flags \|= ZIO_FLAG_NOPWRITE;
	}

	return (zio);
	}

	/*
	* ==========================================================================
	* Block Reference Table
	* ==========================================================================
	*/
	static zio_t *
	zio_brt_free(zio_t *zio)
	{
	blkptr_t *bp;

	bp = zio->io_bp;

	if (BP_GET_LEVEL(bp) > 0 \|\|
	BP_IS_METADATA(bp) \|\|
	!brt_maybe_exists(zio->io_spa, bp)) {
	return (zio);
	}

	if (!brt_entry_decref(zio->io_spa, bp)) {
	/*
	* This isn't the last reference, so we cannot free
	* the data yet.
	*/
	zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
	}

	return (zio);
	}

	/*
	* ==========================================================================
	* Dedup
	* ==========================================================================
	*/
	static void
	zio_ddt_child_read_done(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;
	ddt_entry_t *dde = zio->io_private;
	ddt_phys_t *ddp;
	zio_t *pio = zio_unique_parent(zio);

	mutex_enter(&pio->io_lock);
	ddp = ddt_phys_select(dde, bp);
	if (zio->io_error == 0)
	ddt_phys_clear(ddp); /* this ddp doesn't need repair */

	if (zio->io_error == 0 && dde->dde_repair_abd == NULL)
	dde->dde_repair_abd = zio->io_abd;
	else
	abd_free(zio->io_abd);
	mutex_exit(&pio->io_lock);
	}

	static zio_t *
	zio_ddt_read_start(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;

	ASSERT(BP_GET_DEDUP(bp));
	ASSERT(BP_GET_PSIZE(bp) == zio->io_size);
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);

	if (zio->io_child_error[ZIO_CHILD_DDT]) {
	ddt_t *ddt = ddt_select(zio->io_spa, bp);
	ddt_entry_t *dde = ddt_repair_start(ddt, bp);
	ddt_phys_t *ddp = dde->dde_phys;
	ddt_phys_t *ddp_self = ddt_phys_select(dde, bp);
	blkptr_t blk;

	ASSERT(zio->io_vsd == NULL);
	zio->io_vsd = dde;

	if (ddp_self == NULL)
	return (zio);

	for (int p = 0; p < DDT_PHYS_TYPES; p++, ddp++) {
	if (ddp->ddp_phys_birth == 0 \|\| ddp == ddp_self)
	continue;
	ddt_bp_create(ddt->ddt_checksum, &dde->dde_key, ddp,
	&blk);
	zio_nowait(zio_read(zio, zio->io_spa, &blk,
	abd_alloc_for_io(zio->io_size, B_TRUE),
	zio->io_size, zio_ddt_child_read_done, dde,
	zio->io_priority, ZIO_DDT_CHILD_FLAGS(zio) \|
	ZIO_FLAG_DONT_PROPAGATE, &zio->io_bookmark));
	}
	return (zio);
	}

	zio_nowait(zio_read(zio, zio->io_spa, bp,
	zio->io_abd, zio->io_size, NULL, NULL, zio->io_priority,
	ZIO_DDT_CHILD_FLAGS(zio), &zio->io_bookmark));

	return (zio);
	}

	static zio_t *
	zio_ddt_read_done(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;

	if (zio_wait_for_children(zio, ZIO_CHILD_DDT_BIT, ZIO_WAIT_DONE)) {
	return (NULL);
	}

	ASSERT(BP_GET_DEDUP(bp));
	ASSERT(BP_GET_PSIZE(bp) == zio->io_size);
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);

	if (zio->io_child_error[ZIO_CHILD_DDT]) {
	ddt_t *ddt = ddt_select(zio->io_spa, bp);
	ddt_entry_t *dde = zio->io_vsd;
	if (ddt == NULL) {
	ASSERT(spa_load_state(zio->io_spa) != SPA_LOAD_NONE);
	return (zio);
	}
	if (dde == NULL) {
	zio->io_stage = ZIO_STAGE_DDT_READ_START >> 1;
	zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, B_FALSE);
	return (NULL);
	}
	if (dde->dde_repair_abd != NULL) {
	abd_copy(zio->io_abd, dde->dde_repair_abd,
	zio->io_size);
	zio->io_child_error[ZIO_CHILD_DDT] = 0;
	}
	ddt_repair_done(ddt, dde);
	zio->io_vsd = NULL;
	}

	ASSERT(zio->io_vsd == NULL);

	return (zio);
	}

	static boolean_t
	zio_ddt_collision(zio_t zio, ddt_t ddt, ddt_entry_t *dde)
	{
	spa_t *spa = zio->io_spa;
	boolean_t do_raw = !!(zio->io_flags & ZIO_FLAG_RAW);

	ASSERT(!(zio->io_bp_override && do_raw));

	/*
	* Note: we compare the original data, not the transformed data,
	* because when zio->io_bp is an override bp, we will not have
	* pushed the I/O transforms. That's an important optimization
	* because otherwise we'd compress/encrypt all dmu_sync() data twice.
	* However, we should never get a raw, override zio so in these
	* cases we can compare the io_abd directly. This is useful because
	* it allows us to do dedup verification even if we don't have access
	* to the original data (for instance, if the encryption keys aren't
	* loaded).
	*/

	for (int p = DDT_PHYS_SINGLE; p <= DDT_PHYS_TRIPLE; p++) {
	zio_t *lio = dde->dde_lead_zio[p];

	if (lio != NULL && do_raw) {
	return (lio->io_size != zio->io_size \|\|
	abd_cmp(zio->io_abd, lio->io_abd) != 0);
	} else if (lio != NULL) {
	return (lio->io_orig_size != zio->io_orig_size \|\|
	abd_cmp(zio->io_orig_abd, lio->io_orig_abd) != 0);
	}
	}

	for (int p = DDT_PHYS_SINGLE; p <= DDT_PHYS_TRIPLE; p++) {
	ddt_phys_t *ddp = &dde->dde_phys[p];

	if (ddp->ddp_phys_birth != 0 && do_raw) {
	blkptr_t blk = *zio->io_bp;
	uint64_t psize;
	abd_t *tmpabd;
	int error;

	ddt_bp_fill(ddp, &blk, ddp->ddp_phys_birth);
	psize = BP_GET_PSIZE(&blk);

	if (psize != zio->io_size)
	return (B_TRUE);

	ddt_exit(ddt);

	tmpabd = abd_alloc_for_io(psize, B_TRUE);

	error = zio_wait(zio_read(NULL, spa, &blk, tmpabd,
	psize, NULL, NULL, ZIO_PRIORITY_SYNC_READ,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_SPECULATIVE \|
	ZIO_FLAG_RAW, &zio->io_bookmark));

	if (error == 0) {
	if (abd_cmp(tmpabd, zio->io_abd) != 0)
	error = SET_ERROR(ENOENT);
	}

	abd_free(tmpabd);
	ddt_enter(ddt);
	return (error != 0);
	} else if (ddp->ddp_phys_birth != 0) {
	arc_buf_t *abuf = NULL;
	arc_flags_t aflags = ARC_FLAG_WAIT;
	blkptr_t blk = *zio->io_bp;
	int error;

	ddt_bp_fill(ddp, &blk, ddp->ddp_phys_birth);

	if (BP_GET_LSIZE(&blk) != zio->io_orig_size)
	return (B_TRUE);

	ddt_exit(ddt);

	error = arc_read(NULL, spa, &blk,
	arc_getbuf_func, &abuf, ZIO_PRIORITY_SYNC_READ,
	ZIO_FLAG_CANFAIL \| ZIO_FLAG_SPECULATIVE,
	&aflags, &zio->io_bookmark);

	if (error == 0) {
	if (abd_cmp_buf(zio->io_orig_abd, abuf->b_data,
	zio->io_orig_size) != 0)
	error = SET_ERROR(ENOENT);
	arc_buf_destroy(abuf, &abuf);
	}

	ddt_enter(ddt);
	return (error != 0);
	}
	}

	return (B_FALSE);
	}

	static void
	zio_ddt_child_write_ready(zio_t *zio)
	{
	int p = zio->io_prop.zp_copies;
	ddt_t *ddt = ddt_select(zio->io_spa, zio->io_bp);
	ddt_entry_t *dde = zio->io_private;
	ddt_phys_t *ddp = &dde->dde_phys[p];
	zio_t *pio;

	if (zio->io_error)
	return;

	ddt_enter(ddt);

	ASSERT(dde->dde_lead_zio[p] == zio);

	ddt_phys_fill(ddp, zio->io_bp);

	zio_link_t *zl = NULL;
	while ((pio = zio_walk_parents(zio, &zl)) != NULL)
	ddt_bp_fill(ddp, pio->io_bp, zio->io_txg);

	ddt_exit(ddt);
	}

	static void
	zio_ddt_child_write_done(zio_t *zio)
	{
	int p = zio->io_prop.zp_copies;
	ddt_t *ddt = ddt_select(zio->io_spa, zio->io_bp);
	ddt_entry_t *dde = zio->io_private;
	ddt_phys_t *ddp = &dde->dde_phys[p];

	ddt_enter(ddt);

	ASSERT(ddp->ddp_refcnt == 0);
	ASSERT(dde->dde_lead_zio[p] == zio);
	dde->dde_lead_zio[p] = NULL;

	if (zio->io_error == 0) {
	zio_link_t *zl = NULL;
	while (zio_walk_parents(zio, &zl) != NULL)
	ddt_phys_addref(ddp);
	} else {
	ddt_phys_clear(ddp);
	}

	ddt_exit(ddt);
	}

	static zio_t *
	zio_ddt_write(zio_t *zio)
	{
	spa_t *spa = zio->io_spa;
	blkptr_t *bp = zio->io_bp;
	uint64_t txg = zio->io_txg;
	zio_prop_t *zp = &zio->io_prop;
	int p = zp->zp_copies;
	zio_t *cio = NULL;
	ddt_t *ddt = ddt_select(spa, bp);
	ddt_entry_t *dde;
	ddt_phys_t *ddp;

	ASSERT(BP_GET_DEDUP(bp));
	ASSERT(BP_GET_CHECKSUM(bp) == zp->zp_checksum);
	ASSERT(BP_IS_HOLE(bp) \|\| zio->io_bp_override);
	ASSERT(!(zio->io_bp_override && (zio->io_flags & ZIO_FLAG_RAW)));

	ddt_enter(ddt);
	dde = ddt_lookup(ddt, bp, B_TRUE);
	ddp = &dde->dde_phys[p];

	if (zp->zp_dedup_verify && zio_ddt_collision(zio, ddt, dde)) {
	/*
	* If we're using a weak checksum, upgrade to a strong checksum
	* and try again. If we're already using a strong checksum,
	* we can't resolve it, so just convert to an ordinary write.
	* (And automatically e-mail a paper to Nature?)
	*/
	if (!(zio_checksum_table[zp->zp_checksum].ci_flags &
	ZCHECKSUM_FLAG_DEDUP)) {
	zp->zp_checksum = spa_dedup_checksum(spa);
	zio_pop_transforms(zio);
	zio->io_stage = ZIO_STAGE_OPEN;
	BP_ZERO(bp);
	} else {
	zp->zp_dedup = B_FALSE;
	BP_SET_DEDUP(bp, B_FALSE);
	}
	ASSERT(!BP_GET_DEDUP(bp));
	zio->io_pipeline = ZIO_WRITE_PIPELINE;
	ddt_exit(ddt);
	return (zio);
	}

	if (ddp->ddp_phys_birth != 0 \|\| dde->dde_lead_zio[p] != NULL) {
	if (ddp->ddp_phys_birth != 0)
	ddt_bp_fill(ddp, bp, txg);
	if (dde->dde_lead_zio[p] != NULL)
	zio_add_child(zio, dde->dde_lead_zio[p]);
	else
	ddt_phys_addref(ddp);
	} else if (zio->io_bp_override) {
	ASSERT(BP_GET_LOGICAL_BIRTH(bp) == txg);
	ASSERT(BP_EQUAL(bp, zio->io_bp_override));
	ddt_phys_fill(ddp, bp);
	ddt_phys_addref(ddp);
	} else {
	cio = zio_write(zio, spa, txg, bp, zio->io_orig_abd,
	zio->io_orig_size, zio->io_orig_size, zp,
	zio_ddt_child_write_ready, NULL,
	zio_ddt_child_write_done, dde, zio->io_priority,
	ZIO_DDT_CHILD_FLAGS(zio), &zio->io_bookmark);

	zio_push_transform(cio, zio->io_abd, zio->io_size, 0, NULL);
	dde->dde_lead_zio[p] = cio;
	}

	ddt_exit(ddt);

	zio_nowait(cio);

	return (zio);
	}

	static ddt_entry_t freedde; / for debugging */

	static zio_t *
	zio_ddt_free(zio_t *zio)
	{
	spa_t *spa = zio->io_spa;
	blkptr_t *bp = zio->io_bp;
	ddt_t *ddt = ddt_select(spa, bp);
	ddt_entry_t *dde;
	ddt_phys_t *ddp;

	ASSERT(BP_GET_DEDUP(bp));
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);

	ddt_enter(ddt);
	freedde = dde = ddt_lookup(ddt, bp, B_TRUE);
	if (dde) {
	ddp = ddt_phys_select(dde, bp);
	if (ddp)
	ddt_phys_decref(ddp);
	}
	ddt_exit(ddt);

	return (zio);
	}

	/*
	* ==========================================================================
	* Allocate and free blocks
	* ==========================================================================
	*/

	static zio_t *
	zio_io_to_allocate(spa_t *spa, int allocator)
	{
	zio_t *zio;

	ASSERT(MUTEX_HELD(&spa->spa_allocs[allocator].spaa_lock));

	zio = avl_first(&spa->spa_allocs[allocator].spaa_tree);
	if (zio == NULL)
	return (NULL);

	ASSERT(IO_IS_ALLOCATING(zio));
	ASSERT(ZIO_HAS_ALLOCATOR(zio));

	/*
	* Try to place a reservation for this zio. If we're unable to
	* reserve then we throttle.
	*/
	ASSERT3U(zio->io_allocator, ==, allocator);
	if (!metaslab_class_throttle_reserve(zio->io_metaslab_class,
	zio->io_prop.zp_copies, allocator, zio, 0)) {
	return (NULL);
	}

	avl_remove(&spa->spa_allocs[allocator].spaa_tree, zio);
	ASSERT3U(zio->io_stage, <, ZIO_STAGE_DVA_ALLOCATE);

	return (zio);
	}

	static zio_t *
	zio_dva_throttle(zio_t *zio)
	{
	spa_t *spa = zio->io_spa;
	zio_t *nio;
	metaslab_class_t *mc;

	/* locate an appropriate allocation class */
	mc = spa_preferred_class(spa, zio->io_size, zio->io_prop.zp_type,
	zio->io_prop.zp_level, zio->io_prop.zp_zpl_smallblk);

	if (zio->io_priority == ZIO_PRIORITY_SYNC_WRITE \|\|
	!mc->mc_alloc_throttle_enabled \|\|
	zio->io_child_type == ZIO_CHILD_GANG \|\|
	zio->io_flags & ZIO_FLAG_NODATA) {
	return (zio);
	}

	ASSERT(zio->io_type == ZIO_TYPE_WRITE);
	ASSERT(ZIO_HAS_ALLOCATOR(zio));
	ASSERT(zio->io_child_type > ZIO_CHILD_GANG);
	ASSERT3U(zio->io_queued_timestamp, >, 0);
	ASSERT(zio->io_stage == ZIO_STAGE_DVA_THROTTLE);

	int allocator = zio->io_allocator;
	zio->io_metaslab_class = mc;
	mutex_enter(&spa->spa_allocs[allocator].spaa_lock);
	avl_add(&spa->spa_allocs[allocator].spaa_tree, zio);
	nio = zio_io_to_allocate(spa, allocator);
	mutex_exit(&spa->spa_allocs[allocator].spaa_lock);
	return (nio);
	}

	static void
	zio_allocate_dispatch(spa_t *spa, int allocator)
	{
	zio_t *zio;

	mutex_enter(&spa->spa_allocs[allocator].spaa_lock);
	zio = zio_io_to_allocate(spa, allocator);
	mutex_exit(&spa->spa_allocs[allocator].spaa_lock);
	if (zio == NULL)
	return;

	ASSERT3U(zio->io_stage, ==, ZIO_STAGE_DVA_THROTTLE);
	ASSERT0(zio->io_error);
	zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, B_TRUE);
	}

	static zio_t *
	zio_dva_allocate(zio_t *zio)
	{
	spa_t *spa = zio->io_spa;
	metaslab_class_t *mc;
	blkptr_t *bp = zio->io_bp;
	int error;
	int flags = 0;

	if (zio->io_gang_leader == NULL) {
	ASSERT(zio->io_child_type > ZIO_CHILD_GANG);
	zio->io_gang_leader = zio;
	}

	ASSERT(BP_IS_HOLE(bp));
	ASSERT0(BP_GET_NDVAS(bp));
	ASSERT3U(zio->io_prop.zp_copies, >, 0);
	ASSERT3U(zio->io_prop.zp_copies, <=, spa_max_replication(spa));
	ASSERT3U(zio->io_size, ==, BP_GET_PSIZE(bp));

	if (zio->io_flags & ZIO_FLAG_NODATA)
	flags \|= METASLAB_DONT_THROTTLE;
	if (zio->io_flags & ZIO_FLAG_GANG_CHILD)
	flags \|= METASLAB_GANG_CHILD;
	if (zio->io_priority == ZIO_PRIORITY_ASYNC_WRITE)
	flags \|= METASLAB_ASYNC_ALLOC;

	/*
	* if not already chosen, locate an appropriate allocation class
	*/
	mc = zio->io_metaslab_class;
	if (mc == NULL) {
	mc = spa_preferred_class(spa, zio->io_size,
	zio->io_prop.zp_type, zio->io_prop.zp_level,
	zio->io_prop.zp_zpl_smallblk);
	zio->io_metaslab_class = mc;
	}

	/*
	* Try allocating the block in the usual metaslab class.
	* If that's full, allocate it in the normal class.
	* If that's full, allocate as a gang block,
	* and if all are full, the allocation fails (which shouldn't happen).
	*
	* Note that we do not fall back on embedded slog (ZIL) space, to
	* preserve unfragmented slog space, which is critical for decent
	* sync write performance. If a log allocation fails, we will fall
	* back to spa_sync() which is abysmal for performance.
	*/
	ASSERT(ZIO_HAS_ALLOCATOR(zio));
	error = metaslab_alloc(spa, mc, zio->io_size, bp,
	zio->io_prop.zp_copies, zio->io_txg, NULL, flags,
	&zio->io_alloc_list, zio, zio->io_allocator);

	/*
	* Fallback to normal class when an alloc class is full
	*/
	if (error == ENOSPC && mc != spa_normal_class(spa)) {
	/*
	* If throttling, transfer reservation over to normal class.
	* The io_allocator slot can remain the same even though we
	* are switching classes.
	*/
	if (mc->mc_alloc_throttle_enabled &&
	(zio->io_flags & ZIO_FLAG_IO_ALLOCATING)) {
	metaslab_class_throttle_unreserve(mc,
	zio->io_prop.zp_copies, zio->io_allocator, zio);
	zio->io_flags &= ~ZIO_FLAG_IO_ALLOCATING;

	VERIFY(metaslab_class_throttle_reserve(
	spa_normal_class(spa),
	zio->io_prop.zp_copies, zio->io_allocator, zio,
	flags \| METASLAB_MUST_RESERVE));
	}
	zio->io_metaslab_class = mc = spa_normal_class(spa);
	if (zfs_flags & ZFS_DEBUG_METASLAB_ALLOC) {
	zfs_dbgmsg("%s: metaslab allocation failure, "
	"trying normal class: zio %px, size %llu, error %d",
	spa_name(spa), zio, (u_longlong_t)zio->io_size,
	error);
	}

	error = metaslab_alloc(spa, mc, zio->io_size, bp,
	zio->io_prop.zp_copies, zio->io_txg, NULL, flags,
	&zio->io_alloc_list, zio, zio->io_allocator);
	}

	if (error == ENOSPC && zio->io_size > SPA_MINBLOCKSIZE) {
	if (zfs_flags & ZFS_DEBUG_METASLAB_ALLOC) {
	zfs_dbgmsg("%s: metaslab allocation failure, "
	"trying ganging: zio %px, size %llu, error %d",
	spa_name(spa), zio, (u_longlong_t)zio->io_size,
	error);
	}
	return (zio_write_gang_block(zio, mc));
	}
	if (error != 0) {
	if (error != ENOSPC \|\|
	(zfs_flags & ZFS_DEBUG_METASLAB_ALLOC)) {
	zfs_dbgmsg("%s: metaslab allocation failure: zio %px, "
	"size %llu, error %d",
	spa_name(spa), zio, (u_longlong_t)zio->io_size,
	error);
	}
	zio->io_error = error;
	}

	return (zio);
	}

	static zio_t *
	zio_dva_free(zio_t *zio)
	{
	metaslab_free(zio->io_spa, zio->io_bp, zio->io_txg, B_FALSE);

	return (zio);
	}

	static zio_t *
	zio_dva_claim(zio_t *zio)
	{
	int error;

	error = metaslab_claim(zio->io_spa, zio->io_bp, zio->io_txg);
	if (error)
	zio->io_error = error;

	return (zio);
	}

	/*
	* Undo an allocation. This is used by zio_done() when an I/O fails
	* and we want to give back the block we just allocated.
	* This handles both normal blocks and gang blocks.
	*/
	static void
	zio_dva_unallocate(zio_t zio, zio_gang_node_t gn, blkptr_t *bp)
	{
	ASSERT(BP_GET_LOGICAL_BIRTH(bp) == zio->io_txg \|\| BP_IS_HOLE(bp));
	ASSERT(zio->io_bp_override == NULL);

	if (!BP_IS_HOLE(bp)) {
	metaslab_free(zio->io_spa, bp, BP_GET_LOGICAL_BIRTH(bp),
	B_TRUE);
	}

	if (gn != NULL) {
	for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) {
	zio_dva_unallocate(zio, gn->gn_child[g],
	&gn->gn_gbh->zg_blkptr[g]);
	}
	}
	}

	/*
	* Try to allocate an intent log block. Return 0 on success, errno on failure.
	*/
	int
	zio_alloc_zil(spa_t spa, objset_t os, uint64_t txg, blkptr_t *new_bp,
	uint64_t size, boolean_t *slog)
	{
	int error = 1;
	zio_alloc_list_t io_alloc_list;

	ASSERT(txg > spa_syncing_txg(spa));

	metaslab_trace_init(&io_alloc_list);

	/*
	* Block pointer fields are useful to metaslabs for stats and debugging.
	* Fill in the obvious ones before calling into metaslab_alloc().
	*/
	BP_SET_TYPE(new_bp, DMU_OT_INTENT_LOG);
	BP_SET_PSIZE(new_bp, size);
	BP_SET_LEVEL(new_bp, 0);

	/*
	* When allocating a zil block, we don't have information about
	* the final destination of the block except the objset it's part
	* of, so we just hash the objset ID to pick the allocator to get
	* some parallelism.
	*/
	int flags = METASLAB_ZIL;
	int allocator = (uint_t)cityhash4(0, 0, 0,
	os->os_dsl_dataset->ds_object) % spa->spa_alloc_count;
	error = metaslab_alloc(spa, spa_log_class(spa), size, new_bp, 1,
	txg, NULL, flags, &io_alloc_list, NULL, allocator);
	*slog = (error == 0);
	if (error != 0) {
	error = metaslab_alloc(spa, spa_embedded_log_class(spa), size,
	new_bp, 1, txg, NULL, flags,
	&io_alloc_list, NULL, allocator);
	}
	if (error != 0) {
	error = metaslab_alloc(spa, spa_normal_class(spa), size,
	new_bp, 1, txg, NULL, flags,
	&io_alloc_list, NULL, allocator);
	}
	metaslab_trace_fini(&io_alloc_list);

	if (error == 0) {
	BP_SET_LSIZE(new_bp, size);
	BP_SET_PSIZE(new_bp, size);
	BP_SET_COMPRESS(new_bp, ZIO_COMPRESS_OFF);
	BP_SET_CHECKSUM(new_bp,
	spa_version(spa) >= SPA_VERSION_SLIM_ZIL
	? ZIO_CHECKSUM_ZILOG2 : ZIO_CHECKSUM_ZILOG);
	BP_SET_TYPE(new_bp, DMU_OT_INTENT_LOG);
	BP_SET_LEVEL(new_bp, 0);
	BP_SET_DEDUP(new_bp, 0);
	BP_SET_BYTEORDER(new_bp, ZFS_HOST_BYTEORDER);

	/*
	* encrypted blocks will require an IV and salt. We generate
	* these now since we will not be rewriting the bp at
	* rewrite time.
	*/
	if (os->os_encrypted) {
	uint8_t iv[ZIO_DATA_IV_LEN];
	uint8_t salt[ZIO_DATA_SALT_LEN];

	BP_SET_CRYPT(new_bp, B_TRUE);
	VERIFY0(spa_crypt_get_salt(spa,
	dmu_objset_id(os), salt));
	VERIFY0(zio_crypt_generate_iv(iv));

	zio_crypt_encode_params_bp(new_bp, salt, iv);
	}
	} else {
	zfs_dbgmsg("%s: zil block allocation failure: "
	"size %llu, error %d", spa_name(spa), (u_longlong_t)size,
	error);
	}

	return (error);
	}

	/*
	* ==========================================================================
	* Read and write to physical devices
	* ==========================================================================
	*/

	/*
	* Issue an I/O to the underlying vdev. Typically the issue pipeline
	* stops after this stage and will resume upon I/O completion.
	* However, there are instances where the vdev layer may need to
	* continue the pipeline when an I/O was not issued. Since the I/O
	* that was sent to the vdev layer might be different than the one
	* currently active in the pipeline (see vdev_queue_io()), we explicitly
	* force the underlying vdev layers to call either zio_execute() or
	* zio_interrupt() to ensure that the pipeline continues with the correct I/O.
	*/
	static zio_t *
	zio_vdev_io_start(zio_t *zio)
	{
	vdev_t *vd = zio->io_vd;
	uint64_t align;
	spa_t *spa = zio->io_spa;

	zio->io_delay = 0;

	ASSERT(zio->io_error == 0);
	ASSERT(zio->io_child_error[ZIO_CHILD_VDEV] == 0);

	if (vd == NULL) {
	if (!(zio->io_flags & ZIO_FLAG_CONFIG_WRITER))
	spa_config_enter(spa, SCL_ZIO, zio, RW_READER);

	/*
	* The mirror_ops handle multiple DVAs in a single BP.
	*/
	vdev_mirror_ops.vdev_op_io_start(zio);
	return (NULL);
	}

	ASSERT3P(zio->io_logical, !=, zio);
	if (zio->io_type == ZIO_TYPE_WRITE) {
	ASSERT(spa->spa_trust_config);

	/*
	* Note: the code can handle other kinds of writes,
	* but we don't expect them.
	*/
	if (zio->io_vd->vdev_noalloc) {
	ASSERT(zio->io_flags &
	(ZIO_FLAG_PHYSICAL \| ZIO_FLAG_SELF_HEAL \|
	ZIO_FLAG_RESILVER \| ZIO_FLAG_INDUCE_DAMAGE));
	}
	}

	align = 1ULL << vd->vdev_top->vdev_ashift;

	if (!(zio->io_flags & ZIO_FLAG_PHYSICAL) &&
	P2PHASE(zio->io_size, align) != 0) {
	/* Transform logical writes to be a full physical block size. */
	uint64_t asize = P2ROUNDUP(zio->io_size, align);
	abd_t *abuf = abd_alloc_sametype(zio->io_abd, asize);
	ASSERT(vd == vd->vdev_top);
	if (zio->io_type == ZIO_TYPE_WRITE) {
	abd_copy(abuf, zio->io_abd, zio->io_size);
	abd_zero_off(abuf, zio->io_size, asize - zio->io_size);
	}
	zio_push_transform(zio, abuf, asize, asize, zio_subblock);
	}

	/*
	* If this is not a physical io, make sure that it is properly aligned
	* before proceeding.
	*/
	if (!(zio->io_flags & ZIO_FLAG_PHYSICAL)) {
	ASSERT0(P2PHASE(zio->io_offset, align));
	ASSERT0(P2PHASE(zio->io_size, align));
	} else {
	/*
	* For physical writes, we allow 512b aligned writes and assume
	* the device will perform a read-modify-write as necessary.
	*/
	ASSERT0(P2PHASE(zio->io_offset, SPA_MINBLOCKSIZE));
	ASSERT0(P2PHASE(zio->io_size, SPA_MINBLOCKSIZE));
	}

	VERIFY(zio->io_type != ZIO_TYPE_WRITE \|\| spa_writeable(spa));

	/*
	* If this is a repair I/O, and there's no self-healing involved --
	* that is, we're just resilvering what we expect to resilver --
	* then don't do the I/O unless zio's txg is actually in vd's DTL.
	* This prevents spurious resilvering.
	*
	* There are a few ways that we can end up creating these spurious
	* resilver i/os:
	*
	* 1. A resilver i/o will be issued if any DVA in the BP has a
	* dirty DTL. The mirror code will issue resilver writes to
	* each DVA, including the one(s) that are not on vdevs with dirty
	* DTLs.
	*
	* 2. With nested replication, which happens when we have a
	* "replacing" or "spare" vdev that's a child of a mirror or raidz.
	* For example, given mirror(replacing(A+B), C), it's likely that
	* only A is out of date (it's the new device). In this case, we'll
	* read from C, then use the data to resilver A+B -- but we don't
	* actually want to resilver B, just A. The top-level mirror has no
	* way to know this, so instead we just discard unnecessary repairs
	* as we work our way down the vdev tree.
	*
	* 3. ZTEST also creates mirrors of mirrors, mirrors of raidz, etc.
	* The same logic applies to any form of nested replication: ditto
	* + mirror, RAID-Z + replacing, etc.
	*
	* However, indirect vdevs point off to other vdevs which may have
	* DTL's, so we never bypass them. The child i/os on concrete vdevs
	* will be properly bypassed instead.
	*
	* Leaf DTL_PARTIAL can be empty when a legitimate write comes from
	* a dRAID spare vdev. For example, when a dRAID spare is first
	* used, its spare blocks need to be written to but the leaf vdev's
	* of such blocks can have empty DTL_PARTIAL.
	*
	* There seemed no clean way to allow such writes while bypassing
	* spurious ones. At this point, just avoid all bypassing for dRAID
	* for correctness.
	*/
	if ((zio->io_flags & ZIO_FLAG_IO_REPAIR) &&
	!(zio->io_flags & ZIO_FLAG_SELF_HEAL) &&
	zio->io_txg != 0 && /* not a delegated i/o */
	vd->vdev_ops != &vdev_indirect_ops &&
	vd->vdev_top->vdev_ops != &vdev_draid_ops &&
	!vdev_dtl_contains(vd, DTL_PARTIAL, zio->io_txg, 1)) {
	ASSERT(zio->io_type == ZIO_TYPE_WRITE);
	zio_vdev_io_bypass(zio);
	return (zio);
	}

	/*
	* Select the next best leaf I/O to process. Distributed spares are
	* excluded since they dispatch the I/O directly to a leaf vdev after
	* applying the dRAID mapping.
	*/
	if (vd->vdev_ops->vdev_op_leaf &&
	vd->vdev_ops != &vdev_draid_spare_ops &&
	(zio->io_type == ZIO_TYPE_READ \|\|
	zio->io_type == ZIO_TYPE_WRITE \|\|
	zio->io_type == ZIO_TYPE_TRIM)) {

	if (zio_handle_device_injection(vd, zio, ENOSYS) != 0) {
	/*
	* "no-op" injections return success, but do no actual
	* work. Just skip the remaining vdev stages.
	*/
	zio_vdev_io_bypass(zio);
	zio_interrupt(zio);
	return (NULL);
	}

	if ((zio = vdev_queue_io(zio)) == NULL)
	return (NULL);

	if (!vdev_accessible(vd, zio)) {
	zio->io_error = SET_ERROR(ENXIO);
	zio_interrupt(zio);
	return (NULL);
	}
	zio->io_delay = gethrtime();
	}

	vd->vdev_ops->vdev_op_io_start(zio);
	return (NULL);
	}

	static zio_t *
	zio_vdev_io_done(zio_t *zio)
	{
	vdev_t *vd = zio->io_vd;
	vdev_ops_t *ops = vd ? vd->vdev_ops : &vdev_mirror_ops;
	boolean_t unexpected_error = B_FALSE;

	if (zio_wait_for_children(zio, ZIO_CHILD_VDEV_BIT, ZIO_WAIT_DONE)) {
	return (NULL);
	}

	ASSERT(zio->io_type == ZIO_TYPE_READ \|\|
	zio->io_type == ZIO_TYPE_WRITE \|\|
	zio->io_type == ZIO_TYPE_FLUSH \|\|
	zio->io_type == ZIO_TYPE_TRIM);

	if (zio->io_delay)
	zio->io_delay = gethrtime() - zio->io_delay;

	if (vd != NULL && vd->vdev_ops->vdev_op_leaf &&
	vd->vdev_ops != &vdev_draid_spare_ops) {
	if (zio->io_type != ZIO_TYPE_FLUSH)
	vdev_queue_io_done(zio);

	if (zio_injection_enabled && zio->io_error == 0)
	zio->io_error = zio_handle_device_injections(vd, zio,
	EIO, EILSEQ);

	if (zio_injection_enabled && zio->io_error == 0)
	zio->io_error = zio_handle_label_injection(zio, EIO);

	if (zio->io_error && zio->io_type != ZIO_TYPE_FLUSH &&
	zio->io_type != ZIO_TYPE_TRIM) {
	if (!vdev_accessible(vd, zio)) {
	zio->io_error = SET_ERROR(ENXIO);
	} else {
	unexpected_error = B_TRUE;
	}
	}
	}

	ops->vdev_op_io_done(zio);

	if (unexpected_error && vd->vdev_remove_wanted == B_FALSE)
	VERIFY(vdev_probe(vd, zio) == NULL);

	return (zio);
	}

	/*
	* This function is used to change the priority of an existing zio that is
	* currently in-flight. This is used by the arc to upgrade priority in the
	* event that a demand read is made for a block that is currently queued
	* as a scrub or async read IO. Otherwise, the high priority read request
	* would end up having to wait for the lower priority IO.
	*/
	void
	zio_change_priority(zio_t *pio, zio_priority_t priority)
	{
	zio_t cio, cio_next;
	zio_link_t *zl = NULL;

	ASSERT3U(priority, <, ZIO_PRIORITY_NUM_QUEUEABLE);

	if (pio->io_vd != NULL && pio->io_vd->vdev_ops->vdev_op_leaf) {
	vdev_queue_change_io_priority(pio, priority);
	} else {
	pio->io_priority = priority;
	}

	mutex_enter(&pio->io_lock);
	for (cio = zio_walk_children(pio, &zl); cio != NULL; cio = cio_next) {
	cio_next = zio_walk_children(pio, &zl);
	zio_change_priority(cio, priority);
	}
	mutex_exit(&pio->io_lock);
	}

	/*
	* For non-raidz ZIOs, we can just copy aside the bad data read from the
	* disk, and use that to finish the checksum ereport later.
	*/
	static void
	zio_vsd_default_cksum_finish(zio_cksum_report_t *zcr,
	const abd_t *good_buf)
	{
	/* no processing needed */
	zfs_ereport_finish_checksum(zcr, good_buf, zcr->zcr_cbdata, B_FALSE);
	}

	void
	zio_vsd_default_cksum_report(zio_t zio, zio_cksum_report_t zcr)
	{
	void *abd = abd_alloc_sametype(zio->io_abd, zio->io_size);

	abd_copy(abd, zio->io_abd, zio->io_size);

	zcr->zcr_cbinfo = zio->io_size;
	zcr->zcr_cbdata = abd;
	zcr->zcr_finish = zio_vsd_default_cksum_finish;
	zcr->zcr_free = zio_abd_free;
	}

	static zio_t *
	zio_vdev_io_assess(zio_t *zio)
	{
	vdev_t *vd = zio->io_vd;

	if (zio_wait_for_children(zio, ZIO_CHILD_VDEV_BIT, ZIO_WAIT_DONE)) {
	return (NULL);
	}

	if (vd == NULL && !(zio->io_flags & ZIO_FLAG_CONFIG_WRITER))
	spa_config_exit(zio->io_spa, SCL_ZIO, zio);

	if (zio->io_vsd != NULL) {
	zio->io_vsd_ops->vsd_free(zio);
	zio->io_vsd = NULL;
	}

	if (zio_injection_enabled && zio->io_error == 0)
	zio->io_error = zio_handle_fault_injection(zio, EIO);

	/*
	* If the I/O failed, determine whether we should attempt to retry it.
	*
	* On retry, we cut in line in the issue queue, since we don't want
	* compression/checksumming/etc. work to prevent our (cheap) IO reissue.
	*/
	if (zio->io_error && vd == NULL &&
	!(zio->io_flags & (ZIO_FLAG_DONT_RETRY \| ZIO_FLAG_IO_RETRY))) {
	ASSERT(!(zio->io_flags & ZIO_FLAG_DONT_QUEUE)); /* not a leaf */
	ASSERT(!(zio->io_flags & ZIO_FLAG_IO_BYPASS)); /* not a leaf */
	zio->io_error = 0;
	zio->io_flags \|= ZIO_FLAG_IO_RETRY \| ZIO_FLAG_DONT_AGGREGATE;
	zio->io_stage = ZIO_STAGE_VDEV_IO_START >> 1;
	zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE,
	zio_requeue_io_start_cut_in_line);
	return (NULL);
	}

	/*
	* If we got an error on a leaf device, convert it to ENXIO
	* if the device is not accessible at all.
	*/
	if (zio->io_error && vd != NULL && vd->vdev_ops->vdev_op_leaf &&
	!vdev_accessible(vd, zio))
	zio->io_error = SET_ERROR(ENXIO);

	/*
	* If we can't write to an interior vdev (mirror or RAID-Z),
	* set vdev_cant_write so that we stop trying to allocate from it.
	*/
	if (zio->io_error == ENXIO && zio->io_type == ZIO_TYPE_WRITE &&
	vd != NULL && !vd->vdev_ops->vdev_op_leaf) {
	vdev_dbgmsg(vd, "zio_vdev_io_assess(zio=%px) setting "
	"cant_write=TRUE due to write failure with ENXIO",
	zio);
	vd->vdev_cant_write = B_TRUE;
	}

	/*
	* If a cache flush returns ENOTSUP or ENOTTY, we know that no future
	* attempts will ever succeed. In this case we set a persistent
	* boolean flag so that we don't bother with it in the future.
	*/
	if ((zio->io_error == ENOTSUP \|\| zio->io_error == ENOTTY) &&
	zio->io_type == ZIO_TYPE_FLUSH && vd != NULL)
	vd->vdev_nowritecache = B_TRUE;

	if (zio->io_error)
	zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;

	return (zio);
	}

	void
	zio_vdev_io_reissue(zio_t *zio)
	{
	ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_START);
	ASSERT(zio->io_error == 0);

	zio->io_stage >>= 1;
	}

	void
	zio_vdev_io_redone(zio_t *zio)
	{
	ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_DONE);

	zio->io_stage >>= 1;
	}

	void
	zio_vdev_io_bypass(zio_t *zio)
	{
	ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_START);
	ASSERT(zio->io_error == 0);

	zio->io_flags \|= ZIO_FLAG_IO_BYPASS;
	zio->io_stage = ZIO_STAGE_VDEV_IO_ASSESS >> 1;
	}

	/*
	* ==========================================================================
	* Encrypt and store encryption parameters
	* ==========================================================================
	*/


	/*
	* This function is used for ZIO_STAGE_ENCRYPT. It is responsible for
	* managing the storage of encryption parameters and passing them to the
	* lower-level encryption functions.
	*/
	static zio_t *
	zio_encrypt(zio_t *zio)
	{
	zio_prop_t *zp = &zio->io_prop;
	spa_t *spa = zio->io_spa;
	blkptr_t *bp = zio->io_bp;
	uint64_t psize = BP_GET_PSIZE(bp);
	uint64_t dsobj = zio->io_bookmark.zb_objset;
	dmu_object_type_t ot = BP_GET_TYPE(bp);
	void *enc_buf = NULL;
	abd_t *eabd = NULL;
	uint8_t salt[ZIO_DATA_SALT_LEN];
	uint8_t iv[ZIO_DATA_IV_LEN];
	uint8_t mac[ZIO_DATA_MAC_LEN];
	boolean_t no_crypt = B_FALSE;

	/* the root zio already encrypted the data */
	if (zio->io_child_type == ZIO_CHILD_GANG)
	return (zio);

	/* only ZIL blocks are re-encrypted on rewrite */
	if (!IO_IS_ALLOCATING(zio) && ot != DMU_OT_INTENT_LOG)
	return (zio);

	if (!(zp->zp_encrypt \|\| BP_IS_ENCRYPTED(bp))) {
	BP_SET_CRYPT(bp, B_FALSE);
	return (zio);
	}

	/* if we are doing raw encryption set the provided encryption params */
	if (zio->io_flags & ZIO_FLAG_RAW_ENCRYPT) {
	ASSERT0(BP_GET_LEVEL(bp));
	BP_SET_CRYPT(bp, B_TRUE);
	BP_SET_BYTEORDER(bp, zp->zp_byteorder);
	if (ot != DMU_OT_OBJSET)
	zio_crypt_encode_mac_bp(bp, zp->zp_mac);

	/* dnode blocks must be written out in the provided byteorder */
	if (zp->zp_byteorder != ZFS_HOST_BYTEORDER &&
	ot == DMU_OT_DNODE) {
	void *bswap_buf = zio_buf_alloc(psize);
	abd_t *babd = abd_get_from_buf(bswap_buf, psize);

	ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF);
	abd_copy_to_buf(bswap_buf, zio->io_abd, psize);
	dmu_ot_byteswap[DMU_OT_BYTESWAP(ot)].ob_func(bswap_buf,
	psize);

	abd_take_ownership_of_buf(babd, B_TRUE);
	zio_push_transform(zio, babd, psize, psize, NULL);
	}

	if (DMU_OT_IS_ENCRYPTED(ot))
	zio_crypt_encode_params_bp(bp, zp->zp_salt, zp->zp_iv);
	return (zio);
	}

	/* indirect blocks only maintain a cksum of the lower level MACs */
	if (BP_GET_LEVEL(bp) > 0) {
	BP_SET_CRYPT(bp, B_TRUE);
	VERIFY0(zio_crypt_do_indirect_mac_checksum_abd(B_TRUE,
	zio->io_orig_abd, BP_GET_LSIZE(bp), BP_SHOULD_BYTESWAP(bp),
	mac));
	zio_crypt_encode_mac_bp(bp, mac);
	return (zio);
	}

	/*
	* Objset blocks are a special case since they have 2 256-bit MACs
	* embedded within them.
	*/
	if (ot == DMU_OT_OBJSET) {
	ASSERT0(DMU_OT_IS_ENCRYPTED(ot));
	ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF);
	BP_SET_CRYPT(bp, B_TRUE);
	VERIFY0(spa_do_crypt_objset_mac_abd(B_TRUE, spa, dsobj,
	zio->io_abd, psize, BP_SHOULD_BYTESWAP(bp)));
	return (zio);
	}

	/* unencrypted object types are only authenticated with a MAC */
	if (!DMU_OT_IS_ENCRYPTED(ot)) {
	BP_SET_CRYPT(bp, B_TRUE);
	VERIFY0(spa_do_crypt_mac_abd(B_TRUE, spa, dsobj,
	zio->io_abd, psize, mac));
	zio_crypt_encode_mac_bp(bp, mac);
	return (zio);
	}

	/*
	* Later passes of sync-to-convergence may decide to rewrite data
	* in place to avoid more disk reallocations. This presents a problem
	* for encryption because this constitutes rewriting the new data with
	* the same encryption key and IV. However, this only applies to blocks
	* in the MOS (particularly the spacemaps) and we do not encrypt the
	* MOS. We assert that the zio is allocating or an intent log write
	* to enforce this.
	*/
	ASSERT(IO_IS_ALLOCATING(zio) \|\| ot == DMU_OT_INTENT_LOG);
	ASSERT(BP_GET_LEVEL(bp) == 0 \|\| ot == DMU_OT_INTENT_LOG);
	ASSERT(spa_feature_is_active(spa, SPA_FEATURE_ENCRYPTION));
	ASSERT3U(psize, !=, 0);

	enc_buf = zio_buf_alloc(psize);
	eabd = abd_get_from_buf(enc_buf, psize);
	abd_take_ownership_of_buf(eabd, B_TRUE);

	/*
	* For an explanation of what encryption parameters are stored
	* where, see the block comment in zio_crypt.c.
	*/
	if (ot == DMU_OT_INTENT_LOG) {
	zio_crypt_decode_params_bp(bp, salt, iv);
	} else {
	BP_SET_CRYPT(bp, B_TRUE);
	}

	/* Perform the encryption. This should not fail */
	VERIFY0(spa_do_crypt_abd(B_TRUE, spa, &zio->io_bookmark,
	BP_GET_TYPE(bp), BP_GET_DEDUP(bp), BP_SHOULD_BYTESWAP(bp),
	salt, iv, mac, psize, zio->io_abd, eabd, &no_crypt));

	/* encode encryption metadata into the bp */
	if (ot == DMU_OT_INTENT_LOG) {
	/*
	* ZIL blocks store the MAC in the embedded checksum, so the
	* transform must always be applied.
	*/
	zio_crypt_encode_mac_zil(enc_buf, mac);
	zio_push_transform(zio, eabd, psize, psize, NULL);
	} else {
	BP_SET_CRYPT(bp, B_TRUE);
	zio_crypt_encode_params_bp(bp, salt, iv);
	zio_crypt_encode_mac_bp(bp, mac);

	if (no_crypt) {
	ASSERT3U(ot, ==, DMU_OT_DNODE);
	abd_free(eabd);
	} else {
	zio_push_transform(zio, eabd, psize, psize, NULL);
	}
	}

	return (zio);
	}

	/*
	* ==========================================================================
	* Generate and verify checksums
	* ==========================================================================
	*/
	static zio_t *
	zio_checksum_generate(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;
	enum zio_checksum checksum;

	if (bp == NULL) {
	/*
	* This is zio_write_phys().
	* We're either generating a label checksum, or none at all.
	*/
	checksum = zio->io_prop.zp_checksum;

	if (checksum == ZIO_CHECKSUM_OFF)
	return (zio);

	ASSERT(checksum == ZIO_CHECKSUM_LABEL);
	} else {
	if (BP_IS_GANG(bp) && zio->io_child_type == ZIO_CHILD_GANG) {
	ASSERT(!IO_IS_ALLOCATING(zio));
	checksum = ZIO_CHECKSUM_GANG_HEADER;
	} else {
	checksum = BP_GET_CHECKSUM(bp);
	}
	}

	zio_checksum_compute(zio, checksum, zio->io_abd, zio->io_size);

	return (zio);
	}

	static zio_t *
	zio_checksum_verify(zio_t *zio)
	{
	zio_bad_cksum_t info;
	blkptr_t *bp = zio->io_bp;
	int error;

	ASSERT(zio->io_vd != NULL);

	if (bp == NULL) {
	/*
	* This is zio_read_phys().
	* We're either verifying a label checksum, or nothing at all.
	*/
	if (zio->io_prop.zp_checksum == ZIO_CHECKSUM_OFF)
	return (zio);

	ASSERT3U(zio->io_prop.zp_checksum, ==, ZIO_CHECKSUM_LABEL);
	}

	if ((error = zio_checksum_error(zio, &info)) != 0) {
	zio->io_error = error;
	if (error == ECKSUM &&
	!(zio->io_flags & ZIO_FLAG_SPECULATIVE)) {
	mutex_enter(&zio->io_vd->vdev_stat_lock);
	zio->io_vd->vdev_stat.vs_checksum_errors++;
	mutex_exit(&zio->io_vd->vdev_stat_lock);
	(void) zfs_ereport_start_checksum(zio->io_spa,
	zio->io_vd, &zio->io_bookmark, zio,
	zio->io_offset, zio->io_size, &info);
	}
	}

	return (zio);
	}

	/*
	* Called by RAID-Z to ensure we don't compute the checksum twice.
	*/
	void
	zio_checksum_verified(zio_t *zio)
	{
	zio->io_pipeline &= ~ZIO_STAGE_CHECKSUM_VERIFY;
	}

	/*
	* ==========================================================================
	* Error rank. Error are ranked in the order 0, ENXIO, ECKSUM, EIO, other.
	* An error of 0 indicates success. ENXIO indicates whole-device failure,
	* which may be transient (e.g. unplugged) or permanent. ECKSUM and EIO
	* indicate errors that are specific to one I/O, and most likely permanent.
	* Any other error is presumed to be worse because we weren't expecting it.
	* ==========================================================================
	*/
	int
	zio_worst_error(int e1, int e2)
	{
	static int zio_error_rank[] = { 0, ENXIO, ECKSUM, EIO };
	int r1, r2;

	for (r1 = 0; r1 < sizeof (zio_error_rank) / sizeof (int); r1++)
	if (e1 == zio_error_rank[r1])
	break;

	for (r2 = 0; r2 < sizeof (zio_error_rank) / sizeof (int); r2++)
	if (e2 == zio_error_rank[r2])
	break;

	return (r1 > r2 ? e1 : e2);
	}

	/*
	* ==========================================================================
	* I/O completion
	* ==========================================================================
	*/
	static zio_t *
	zio_ready(zio_t *zio)
	{
	blkptr_t *bp = zio->io_bp;
	zio_t pio, pio_next;
	zio_link_t *zl = NULL;

	if (zio_wait_for_children(zio, ZIO_CHILD_LOGICAL_BIT \|
	ZIO_CHILD_GANG_BIT \| ZIO_CHILD_DDT_BIT, ZIO_WAIT_READY)) {
	return (NULL);
	}

	if (zio->io_ready) {
	ASSERT(IO_IS_ALLOCATING(zio));
	ASSERT(BP_GET_LOGICAL_BIRTH(bp) == zio->io_txg \|\|
	BP_IS_HOLE(bp) \|\| (zio->io_flags & ZIO_FLAG_NOPWRITE));
	ASSERT(zio->io_children[ZIO_CHILD_GANG][ZIO_WAIT_READY] == 0);

	zio->io_ready(zio);
	}

	#ifdef ZFS_DEBUG
	if (bp != NULL && bp != &zio->io_bp_copy)
	zio->io_bp_copy = *bp;
	#endif

	if (zio->io_error != 0) {
	zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;

	if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING) {
	ASSERT(IO_IS_ALLOCATING(zio));
	ASSERT(zio->io_priority == ZIO_PRIORITY_ASYNC_WRITE);
	ASSERT(zio->io_metaslab_class != NULL);
	ASSERT(ZIO_HAS_ALLOCATOR(zio));

	/*
	* We were unable to allocate anything, unreserve and
	* issue the next I/O to allocate.
	*/
	metaslab_class_throttle_unreserve(
	zio->io_metaslab_class, zio->io_prop.zp_copies,
	zio->io_allocator, zio);
	zio_allocate_dispatch(zio->io_spa, zio->io_allocator);
	}
	}

	mutex_enter(&zio->io_lock);
	zio->io_state[ZIO_WAIT_READY] = 1;
	pio = zio_walk_parents(zio, &zl);
	mutex_exit(&zio->io_lock);

	/*
	* As we notify zio's parents, new parents could be added.
	* New parents go to the head of zio's io_parent_list, however,
	* so we will (correctly) not notify them. The remainder of zio's
	* io_parent_list, from 'pio_next' onward, cannot change because
	* all parents must wait for us to be done before they can be done.
	*/
	for (; pio != NULL; pio = pio_next) {
	pio_next = zio_walk_parents(zio, &zl);
	zio_notify_parent(pio, zio, ZIO_WAIT_READY, NULL);
	}

	if (zio->io_flags & ZIO_FLAG_NODATA) {
	if (bp != NULL && BP_IS_GANG(bp)) {
	zio->io_flags &= ~ZIO_FLAG_NODATA;
	} else {
	ASSERT((uintptr_t)zio->io_abd < SPA_MAXBLOCKSIZE);
	zio->io_pipeline &= ~ZIO_VDEV_IO_STAGES;
	}
	}

	if (zio_injection_enabled &&
	zio->io_spa->spa_syncing_txg == zio->io_txg)
	zio_handle_ignored_writes(zio);

	return (zio);
	}

	/*
	* Update the allocation throttle accounting.
	*/
	static void
	zio_dva_throttle_done(zio_t *zio)
	{
	zio_t *lio __maybe_unused = zio->io_logical;
	zio_t *pio = zio_unique_parent(zio);
	vdev_t *vd = zio->io_vd;
	int flags = METASLAB_ASYNC_ALLOC;

	ASSERT3P(zio->io_bp, !=, NULL);
	ASSERT3U(zio->io_type, ==, ZIO_TYPE_WRITE);
	ASSERT3U(zio->io_priority, ==, ZIO_PRIORITY_ASYNC_WRITE);
	ASSERT3U(zio->io_child_type, ==, ZIO_CHILD_VDEV);
	ASSERT(vd != NULL);
	ASSERT3P(vd, ==, vd->vdev_top);
	ASSERT(zio_injection_enabled \|\| !(zio->io_flags & ZIO_FLAG_IO_RETRY));
	ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REPAIR));
	ASSERT(zio->io_flags & ZIO_FLAG_IO_ALLOCATING);
	ASSERT(!(lio->io_flags & ZIO_FLAG_IO_REWRITE));
	ASSERT(!(lio->io_orig_flags & ZIO_FLAG_NODATA));

	/*
	* Parents of gang children can have two flavors -- ones that
	* allocated the gang header (will have ZIO_FLAG_IO_REWRITE set)
	* and ones that allocated the constituent blocks. The allocation
	* throttle needs to know the allocating parent zio so we must find
	* it here.
	*/
	if (pio->io_child_type == ZIO_CHILD_GANG) {
	/*
	* If our parent is a rewrite gang child then our grandparent
	* would have been the one that performed the allocation.
	*/
	if (pio->io_flags & ZIO_FLAG_IO_REWRITE)
	pio = zio_unique_parent(pio);
	flags \|= METASLAB_GANG_CHILD;
	}

	ASSERT(IO_IS_ALLOCATING(pio));
	ASSERT(ZIO_HAS_ALLOCATOR(pio));
	ASSERT3P(zio, !=, zio->io_logical);
	ASSERT(zio->io_logical != NULL);
	ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REPAIR));
	ASSERT0(zio->io_flags & ZIO_FLAG_NOPWRITE);
	ASSERT(zio->io_metaslab_class != NULL);

	mutex_enter(&pio->io_lock);
	metaslab_group_alloc_decrement(zio->io_spa, vd->vdev_id, pio, flags,
	pio->io_allocator, B_TRUE);
	mutex_exit(&pio->io_lock);

	metaslab_class_throttle_unreserve(zio->io_metaslab_class, 1,
	pio->io_allocator, pio);

	/*
	* Call into the pipeline to see if there is more work that
	* needs to be done. If there is work to be done it will be
	* dispatched to another taskq thread.
	*/
	zio_allocate_dispatch(zio->io_spa, pio->io_allocator);
	}

	static zio_t *
	zio_done(zio_t *zio)
	{
	/*
	* Always attempt to keep stack usage minimal here since
	* we can be called recursively up to 19 levels deep.
	*/
	const uint64_t psize = zio->io_size;
	zio_t pio, pio_next;
	zio_link_t *zl = NULL;

	/*
	* If our children haven't all completed,
	* wait for them and then repeat this pipeline stage.
	*/
	if (zio_wait_for_children(zio, ZIO_CHILD_ALL_BITS, ZIO_WAIT_DONE)) {
	return (NULL);
	}

	/*
	* If the allocation throttle is enabled, then update the accounting.
	* We only track child I/Os that are part of an allocating async
	* write. We must do this since the allocation is performed
	* by the logical I/O but the actual write is done by child I/Os.
	*/
	if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING &&
	zio->io_child_type == ZIO_CHILD_VDEV) {
	ASSERT(zio->io_metaslab_class != NULL);
	ASSERT(zio->io_metaslab_class->mc_alloc_throttle_enabled);
	zio_dva_throttle_done(zio);
	}

	/*
	* If the allocation throttle is enabled, verify that
	* we have decremented the refcounts for every I/O that was throttled.
	*/
	if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING) {
	ASSERT(zio->io_type == ZIO_TYPE_WRITE);
	ASSERT(zio->io_priority == ZIO_PRIORITY_ASYNC_WRITE);
	ASSERT(zio->io_bp != NULL);
	ASSERT(ZIO_HAS_ALLOCATOR(zio));

	metaslab_group_alloc_verify(zio->io_spa, zio->io_bp, zio,
	zio->io_allocator);
	VERIFY(zfs_refcount_not_held(&zio->io_metaslab_class->
	mc_allocator[zio->io_allocator].mca_alloc_slots, zio));
	}


	for (int c = 0; c < ZIO_CHILD_TYPES; c++)
	for (int w = 0; w < ZIO_WAIT_TYPES; w++)
	ASSERT(zio->io_children[c][w] == 0);

	if (zio->io_bp != NULL && !BP_IS_EMBEDDED(zio->io_bp)) {
	ASSERT(zio->io_bp->blk_pad[0] == 0);
	ASSERT(zio->io_bp->blk_pad[1] == 0);
	ASSERT(memcmp(zio->io_bp, &zio->io_bp_copy,
	sizeof (blkptr_t)) == 0 \|\|
	(zio->io_bp == zio_unique_parent(zio)->io_bp));
	if (zio->io_type == ZIO_TYPE_WRITE && !BP_IS_HOLE(zio->io_bp) &&
	zio->io_bp_override == NULL &&
	!(zio->io_flags & ZIO_FLAG_IO_REPAIR)) {
	ASSERT3U(zio->io_prop.zp_copies, <=,
	BP_GET_NDVAS(zio->io_bp));
	ASSERT(BP_COUNT_GANG(zio->io_bp) == 0 \|\|
	(BP_COUNT_GANG(zio->io_bp) ==
	BP_GET_NDVAS(zio->io_bp)));
	}
	if (zio->io_flags & ZIO_FLAG_NOPWRITE)
	VERIFY(BP_EQUAL(zio->io_bp, &zio->io_bp_orig));
	}

	/*
	* If there were child vdev/gang/ddt errors, they apply to us now.
	*/
	zio_inherit_child_errors(zio, ZIO_CHILD_VDEV);
	zio_inherit_child_errors(zio, ZIO_CHILD_GANG);
	zio_inherit_child_errors(zio, ZIO_CHILD_DDT);

	/*
	* If the I/O on the transformed data was successful, generate any
	* checksum reports now while we still have the transformed data.
	*/
	if (zio->io_error == 0) {
	while (zio->io_cksum_report != NULL) {
	zio_cksum_report_t *zcr = zio->io_cksum_report;
	uint64_t align = zcr->zcr_align;
	uint64_t asize = P2ROUNDUP(psize, align);
	abd_t *adata = zio->io_abd;

	if (adata != NULL && asize != psize) {
	adata = abd_alloc(asize, B_TRUE);
	abd_copy(adata, zio->io_abd, psize);
	abd_zero_off(adata, psize, asize - psize);
	}

	zio->io_cksum_report = zcr->zcr_next;
	zcr->zcr_next = NULL;
	zcr->zcr_finish(zcr, adata);
	zfs_ereport_free_checksum(zcr);

	if (adata != NULL && asize != psize)
	abd_free(adata);
	}
	}

	zio_pop_transforms(zio); /* note: may set zio->io_error */

	vdev_stat_update(zio, psize);

	/*
	* If this I/O is attached to a particular vdev is slow, exceeding
	* 30 seconds to complete, post an error described the I/O delay.
	* We ignore these errors if the device is currently unavailable.
	*/
	if (zio->io_delay >= MSEC2NSEC(zio_slow_io_ms)) {
	if (zio->io_vd != NULL && !vdev_is_dead(zio->io_vd)) {
	/*
	* We want to only increment our slow IO counters if
	* the IO is valid (i.e. not if the drive is removed).
	*
	* zfs_ereport_post() will also do these checks, but
	* it can also ratelimit and have other failures, so we
	* need to increment the slow_io counters independent
	* of it.
	*/
	if (zfs_ereport_is_valid(FM_EREPORT_ZFS_DELAY,
	zio->io_spa, zio->io_vd, zio)) {
	mutex_enter(&zio->io_vd->vdev_stat_lock);
	zio->io_vd->vdev_stat.vs_slow_ios++;
	mutex_exit(&zio->io_vd->vdev_stat_lock);

	(void) zfs_ereport_post(FM_EREPORT_ZFS_DELAY,
	zio->io_spa, zio->io_vd, &zio->io_bookmark,
	zio, 0);
	}
	}
	}

	if (zio->io_error) {
	/*
	* If this I/O is attached to a particular vdev,
	* generate an error message describing the I/O failure
	* at the block level. We ignore these errors if the
	* device is currently unavailable.
	*/
	if (zio->io_error != ECKSUM && zio->io_vd != NULL &&
	!vdev_is_dead(zio->io_vd)) {
	int ret = zfs_ereport_post(FM_EREPORT_ZFS_IO,
	zio->io_spa, zio->io_vd, &zio->io_bookmark, zio, 0);
	if (ret != EALREADY) {
	mutex_enter(&zio->io_vd->vdev_stat_lock);
	if (zio->io_type == ZIO_TYPE_READ)
	zio->io_vd->vdev_stat.vs_read_errors++;
	else if (zio->io_type == ZIO_TYPE_WRITE)
	zio->io_vd->vdev_stat.vs_write_errors++;
	mutex_exit(&zio->io_vd->vdev_stat_lock);
	}
	}

	if ((zio->io_error == EIO \|\| !(zio->io_flags &
	(ZIO_FLAG_SPECULATIVE \| ZIO_FLAG_DONT_PROPAGATE))) &&
	zio == zio->io_logical) {
	/*
	* For logical I/O requests, tell the SPA to log the
	* error and generate a logical data ereport.
	*/
	spa_log_error(zio->io_spa, &zio->io_bookmark,
	BP_GET_LOGICAL_BIRTH(zio->io_bp));
	(void) zfs_ereport_post(FM_EREPORT_ZFS_DATA,
	zio->io_spa, NULL, &zio->io_bookmark, zio, 0);
	}
	}

	if (zio->io_error && zio == zio->io_logical) {
	/*
	* Determine whether zio should be reexecuted. This will
	* propagate all the way to the root via zio_notify_parent().
	*/
	ASSERT(zio->io_vd == NULL && zio->io_bp != NULL);
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);

	if (IO_IS_ALLOCATING(zio) &&
	!(zio->io_flags & ZIO_FLAG_CANFAIL)) {
	if (zio->io_error != ENOSPC)
	zio->io_reexecute \|= ZIO_REEXECUTE_NOW;
	else
	zio->io_reexecute \|= ZIO_REEXECUTE_SUSPEND;
	}

	if ((zio->io_type == ZIO_TYPE_READ \|\|
	zio->io_type == ZIO_TYPE_FREE) &&
	!(zio->io_flags & ZIO_FLAG_SCAN_THREAD) &&
	zio->io_error == ENXIO &&
	spa_load_state(zio->io_spa) == SPA_LOAD_NONE &&
	spa_get_failmode(zio->io_spa) != ZIO_FAILURE_MODE_CONTINUE)
	zio->io_reexecute \|= ZIO_REEXECUTE_SUSPEND;

	if (!(zio->io_flags & ZIO_FLAG_CANFAIL) && !zio->io_reexecute)
	zio->io_reexecute \|= ZIO_REEXECUTE_SUSPEND;

	/*
	* Here is a possibly good place to attempt to do
	* either combinatorial reconstruction or error correction
	* based on checksums. It also might be a good place
	* to send out preliminary ereports before we suspend
	* processing.
	*/
	}

	/*
	* If there were logical child errors, they apply to us now.
	* We defer this until now to avoid conflating logical child
	* errors with errors that happened to the zio itself when
	* updating vdev stats and reporting FMA events above.
	*/
	zio_inherit_child_errors(zio, ZIO_CHILD_LOGICAL);

	if ((zio->io_error \|\| zio->io_reexecute) &&
	IO_IS_ALLOCATING(zio) && zio->io_gang_leader == zio &&
	!(zio->io_flags & (ZIO_FLAG_IO_REWRITE \| ZIO_FLAG_NOPWRITE)))
	zio_dva_unallocate(zio, zio->io_gang_tree, zio->io_bp);

	zio_gang_tree_free(&zio->io_gang_tree);

	/*
	* Godfather I/Os should never suspend.
	*/
	if ((zio->io_flags & ZIO_FLAG_GODFATHER) &&
	(zio->io_reexecute & ZIO_REEXECUTE_SUSPEND))
	zio->io_reexecute &= ~ZIO_REEXECUTE_SUSPEND;

	if (zio->io_reexecute) {
	/*
	* This is a logical I/O that wants to reexecute.
	*
	* Reexecute is top-down. When an i/o fails, if it's not
	* the root, it simply notifies its parent and sticks around.
	* The parent, seeing that it still has children in zio_done(),
	* does the same. This percolates all the way up to the root.
	* The root i/o will reexecute or suspend the entire tree.
	*
	* This approach ensures that zio_reexecute() honors
	* all the original i/o dependency relationships, e.g.
	* parents not executing until children are ready.
	*/
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);

	zio->io_gang_leader = NULL;

	mutex_enter(&zio->io_lock);
	zio->io_state[ZIO_WAIT_DONE] = 1;
	mutex_exit(&zio->io_lock);

	/*
	* "The Godfather" I/O monitors its children but is
	* not a true parent to them. It will track them through
	* the pipeline but severs its ties whenever they get into
	* trouble (e.g. suspended). This allows "The Godfather"
	* I/O to return status without blocking.
	*/
	zl = NULL;
	for (pio = zio_walk_parents(zio, &zl); pio != NULL;
	pio = pio_next) {
	zio_link_t *remove_zl = zl;
	pio_next = zio_walk_parents(zio, &zl);

	if ((pio->io_flags & ZIO_FLAG_GODFATHER) &&
	(zio->io_reexecute & ZIO_REEXECUTE_SUSPEND)) {
	zio_remove_child(pio, zio, remove_zl);
	/*
	* This is a rare code path, so we don't
	* bother with "next_to_execute".
	*/
	zio_notify_parent(pio, zio, ZIO_WAIT_DONE,
	NULL);
	}
	}

	if ((pio = zio_unique_parent(zio)) != NULL) {
	/*
	* We're not a root i/o, so there's nothing to do
	* but notify our parent. Don't propagate errors
	* upward since we haven't permanently failed yet.
	*/
	ASSERT(!(zio->io_flags & ZIO_FLAG_GODFATHER));
	zio->io_flags \|= ZIO_FLAG_DONT_PROPAGATE;
	/*
	* This is a rare code path, so we don't bother with
	* "next_to_execute".
	*/
	zio_notify_parent(pio, zio, ZIO_WAIT_DONE, NULL);
	} else if (zio->io_reexecute & ZIO_REEXECUTE_SUSPEND) {
	/*
	* We'd fail again if we reexecuted now, so suspend
	* until conditions improve (e.g. device comes online).
	*/
	zio_suspend(zio->io_spa, zio, ZIO_SUSPEND_IOERR);
	} else {
	/*
	* Reexecution is potentially a huge amount of work.
	* Hand it off to the otherwise-unused claim taskq.
	*/
	- ASSERT(taskq_empty_ent(&zio->io_tqent));
	- spa_taskq_dispatch_ent(zio->io_spa,
	+ spa_taskq_dispatch(zio->io_spa,
	ZIO_TYPE_CLAIM, ZIO_TASKQ_ISSUE,
	- zio_reexecute, zio, 0, &zio->io_tqent, NULL);
	+ zio_reexecute, zio, B_FALSE);
	}
	return (NULL);
	}

	ASSERT(list_is_empty(&zio->io_child_list));
	ASSERT(zio->io_reexecute == 0);
	ASSERT(zio->io_error == 0 \|\| (zio->io_flags & ZIO_FLAG_CANFAIL));

	/*
	* Report any checksum errors, since the I/O is complete.
	*/
	while (zio->io_cksum_report != NULL) {
	zio_cksum_report_t *zcr = zio->io_cksum_report;
	zio->io_cksum_report = zcr->zcr_next;
	zcr->zcr_next = NULL;
	zcr->zcr_finish(zcr, NULL);
	zfs_ereport_free_checksum(zcr);
	}

	/*
	* It is the responsibility of the done callback to ensure that this
	* particular zio is no longer discoverable for adoption, and as
	* such, cannot acquire any new parents.
	*/
	if (zio->io_done)
	zio->io_done(zio);

	mutex_enter(&zio->io_lock);
	zio->io_state[ZIO_WAIT_DONE] = 1;
	mutex_exit(&zio->io_lock);

	/*
	* We are done executing this zio. We may want to execute a parent
	* next. See the comment in zio_notify_parent().
	*/
	zio_t *next_to_execute = NULL;
	zl = NULL;
	for (pio = zio_walk_parents(zio, &zl); pio != NULL; pio = pio_next) {
	zio_link_t *remove_zl = zl;
	pio_next = zio_walk_parents(zio, &zl);
	zio_remove_child(pio, zio, remove_zl);
	zio_notify_parent(pio, zio, ZIO_WAIT_DONE, &next_to_execute);
	}

	if (zio->io_waiter != NULL) {
	mutex_enter(&zio->io_lock);
	zio->io_executor = NULL;
	cv_broadcast(&zio->io_cv);
	mutex_exit(&zio->io_lock);
	} else {
	zio_destroy(zio);
	}

	return (next_to_execute);
	}

	/*
	* ==========================================================================
	* I/O pipeline definition
	* ==========================================================================
	*/
	static zio_pipe_stage_t *zio_pipeline[] = {
	NULL,
	zio_read_bp_init,
	zio_write_bp_init,
	zio_free_bp_init,
	zio_issue_async,
	zio_write_compress,
	zio_encrypt,
	zio_checksum_generate,
	zio_nop_write,
	zio_brt_free,
	zio_ddt_read_start,
	zio_ddt_read_done,
	zio_ddt_write,
	zio_ddt_free,
	zio_gang_assemble,
	zio_gang_issue,
	zio_dva_throttle,
	zio_dva_allocate,
	zio_dva_free,
	zio_dva_claim,
	zio_ready,
	zio_vdev_io_start,
	zio_vdev_io_done,
	zio_vdev_io_assess,
	zio_checksum_verify,
	zio_done
	};




	/*
	* Compare two zbookmark_phys_t's to see which we would reach first in a
	* pre-order traversal of the object tree.
	*
	* This is simple in every case aside from the meta-dnode object. For all other
	* objects, we traverse them in order (object 1 before object 2, and so on).
	* However, all of these objects are traversed while traversing object 0, since
	* the data it points to is the list of objects. Thus, we need to convert to a
	* canonical representation so we can compare meta-dnode bookmarks to
	* non-meta-dnode bookmarks.
	*
	* We do this by calculating "equivalents" for each field of the zbookmark.
	* zbookmarks outside of the meta-dnode use their own object and level, and
	* calculate the level 0 equivalent (the first L0 blkid that is contained in the
	* blocks this bookmark refers to) by multiplying their blkid by their span
	* (the number of L0 blocks contained within one block at their level).
	* zbookmarks inside the meta-dnode calculate their object equivalent
	* (which is L0equiv * dnodes per data block), use 0 for their L0equiv, and use
	* level + 1<<31 (any value larger than a level could ever be) for their level.
	* This causes them to always compare before a bookmark in their object
	* equivalent, compare appropriately to bookmarks in other objects, and to
	* compare appropriately to other bookmarks in the meta-dnode.
	*/
	int
	zbookmark_compare(uint16_t dbss1, uint8_t ibs1, uint16_t dbss2, uint8_t ibs2,
	const zbookmark_phys_t zb1, const zbookmark_phys_t zb2)
	{
	/*
	* These variables represent the "equivalent" values for the zbookmark,
	* after converting zbookmarks inside the meta dnode to their
	* normal-object equivalents.
	*/
	uint64_t zb1obj, zb2obj;
	uint64_t zb1L0, zb2L0;
	uint64_t zb1level, zb2level;

	if (zb1->zb_object == zb2->zb_object &&
	zb1->zb_level == zb2->zb_level &&
	zb1->zb_blkid == zb2->zb_blkid)
	return (0);

	IMPLY(zb1->zb_level > 0, ibs1 >= SPA_MINBLOCKSHIFT);
	IMPLY(zb2->zb_level > 0, ibs2 >= SPA_MINBLOCKSHIFT);

	/*
	* BP_SPANB calculates the span in blocks.
	*/
	zb1L0 = (zb1->zb_blkid) * BP_SPANB(ibs1, zb1->zb_level);
	zb2L0 = (zb2->zb_blkid) * BP_SPANB(ibs2, zb2->zb_level);

	if (zb1->zb_object == DMU_META_DNODE_OBJECT) {
	zb1obj = zb1L0 * (dbss1 << (SPA_MINBLOCKSHIFT - DNODE_SHIFT));
	zb1L0 = 0;
	zb1level = zb1->zb_level + COMPARE_META_LEVEL;
	} else {
	zb1obj = zb1->zb_object;
	zb1level = zb1->zb_level;
	}

	if (zb2->zb_object == DMU_META_DNODE_OBJECT) {
	zb2obj = zb2L0 * (dbss2 << (SPA_MINBLOCKSHIFT - DNODE_SHIFT));
	zb2L0 = 0;
	zb2level = zb2->zb_level + COMPARE_META_LEVEL;
	} else {
	zb2obj = zb2->zb_object;
	zb2level = zb2->zb_level;
	}

	/* Now that we have a canonical representation, do the comparison. */
	if (zb1obj != zb2obj)
	return (zb1obj < zb2obj ? -1 : 1);
	else if (zb1L0 != zb2L0)
	return (zb1L0 < zb2L0 ? -1 : 1);
	else if (zb1level != zb2level)
	return (zb1level > zb2level ? -1 : 1);
	/*
	* This can (theoretically) happen if the bookmarks have the same object
	* and level, but different blkids, if the block sizes are not the same.
	* There is presently no way to change the indirect block sizes
	*/
	return (0);
	}

	/*
	* This function checks the following: given that last_block is the place that
	* our traversal stopped last time, does that guarantee that we've visited
	* every node under subtree_root? Therefore, we can't just use the raw output
	* of zbookmark_compare. We have to pass in a modified version of
	* subtree_root; by incrementing the block id, and then checking whether
	* last_block is before or equal to that, we can tell whether or not having
	* visited last_block implies that all of subtree_root's children have been
	* visited.
	*/
	boolean_t
	zbookmark_subtree_completed(const dnode_phys_t *dnp,
	const zbookmark_phys_t subtree_root, const zbookmark_phys_t last_block)
	{
	zbookmark_phys_t mod_zb = *subtree_root;
	mod_zb.zb_blkid++;
	ASSERT0(last_block->zb_level);

	/* The objset_phys_t isn't before anything. */
	if (dnp == NULL)
	return (B_FALSE);

	/*
	* We pass in 1ULL << (DNODE_BLOCK_SHIFT - SPA_MINBLOCKSHIFT) for the
	* data block size in sectors, because that variable is only used if
	* the bookmark refers to a block in the meta-dnode. Since we don't
	* know without examining it what object it refers to, and there's no
	* harm in passing in this value in other cases, we always pass it in.
	*
	* We pass in 0 for the indirect block size shift because zb2 must be
	* level 0. The indirect block size is only used to calculate the span
	* of the bookmark, but since the bookmark must be level 0, the span is
	* always 1, so the math works out.
	*
	* If you make changes to how the zbookmark_compare code works, be sure
	* to make sure that this code still works afterwards.
	*/
	return (zbookmark_compare(dnp->dn_datablkszsec, dnp->dn_indblkshift,
	1ULL << (DNODE_BLOCK_SHIFT - SPA_MINBLOCKSHIFT), 0, &mod_zb,
	last_block) <= 0);
	}

	/*
	* This function is similar to zbookmark_subtree_completed(), but returns true
	* if subtree_root is equal or ahead of last_block, i.e. still to be done.
	*/
	boolean_t
	zbookmark_subtree_tbd(const dnode_phys_t *dnp,
	const zbookmark_phys_t subtree_root, const zbookmark_phys_t last_block)
	{
	ASSERT0(last_block->zb_level);
	if (dnp == NULL)
	return (B_FALSE);
	return (zbookmark_compare(dnp->dn_datablkszsec, dnp->dn_indblkshift,
	1ULL << (DNODE_BLOCK_SHIFT - SPA_MINBLOCKSHIFT), 0, subtree_root,
	last_block) >= 0);
	}

	EXPORT_SYMBOL(zio_type_name);
	EXPORT_SYMBOL(zio_buf_alloc);
	EXPORT_SYMBOL(zio_data_buf_alloc);
	EXPORT_SYMBOL(zio_buf_free);
	EXPORT_SYMBOL(zio_data_buf_free);

	ZFS_MODULE_PARAM(zfs_zio, zio_, slow_io_ms, INT, ZMOD_RW,
	"Max I/O completion time (milliseconds) before marking it as slow");

	ZFS_MODULE_PARAM(zfs_zio, zio_, requeue_io_start_cut_in_line, INT, ZMOD_RW,
	"Prioritize requeued I/O");

	ZFS_MODULE_PARAM(zfs, zfs_, sync_pass_deferred_free, UINT, ZMOD_RW,
	"Defer frees starting in this pass");

	ZFS_MODULE_PARAM(zfs, zfs_, sync_pass_dont_compress, UINT, ZMOD_RW,
	"Don't compress starting in this pass");

	ZFS_MODULE_PARAM(zfs, zfs_, sync_pass_rewrite, UINT, ZMOD_RW,
	"Rewrite new bps starting in this pass");

	ZFS_MODULE_PARAM(zfs_zio, zio_, dva_throttle_enabled, INT, ZMOD_RW,
	"Throttle block allocations in the ZIO pipeline");

	ZFS_MODULE_PARAM(zfs_zio, zio_, deadman_log_all, INT, ZMOD_RW,
	"Log all slow ZIOs, not just those with vdevs");
	diff --git a/sys/contrib/openzfs/tests/runfiles/common.run b/sys/contrib/openzfs/tests/runfiles/common.run
	index 5e7fdf359a75..ac2c541a9188 100644
	--- a/sys/contrib/openzfs/tests/runfiles/common.run
	+++ b/sys/contrib/openzfs/tests/runfiles/common.run
	@@ -1,1040 +1,1041 @@
	#
	# This file and its contents are supplied under the terms of the
	# Common Development and Distribution License ("CDDL"), version 1.0.
	# You may only use this file in accordance with the terms of version
	# 1.0 of the CDDL.
	#
	# A full copy of the text of the CDDL should have accompanied this
	# source. A copy of the CDDL is also available via the Internet at
	# http://www.illumos.org/license/CDDL.
	#
	# This run file contains all of the common functional tests. When
	# adding a new test consider also adding it to the sanity.run file
	# if the new test runs to completion in only a few seconds.
	#
	# Approximate run time: 4-5 hours
	#

	[DEFAULT]
	pre = setup
	quiet = False
	pre_user = root
	user = root
	timeout = 600
	post_user = root
	post = cleanup
	failsafe_user = root
	failsafe = callbacks/zfs_failsafe
	outputdir = /var/tmp/test_results
	tags = ['functional']

	[tests/functional/acl/off]
	tests = ['dosmode', 'posixmode']
	tags = ['functional', 'acl']

	[tests/functional/alloc_class]
	tests = ['alloc_class_001_pos', 'alloc_class_002_neg', 'alloc_class_003_pos',
	'alloc_class_004_pos', 'alloc_class_005_pos', 'alloc_class_006_pos',
	'alloc_class_007_pos', 'alloc_class_008_pos', 'alloc_class_009_pos',
	'alloc_class_010_pos', 'alloc_class_011_neg', 'alloc_class_012_pos',
	'alloc_class_013_pos', 'alloc_class_014_neg', 'alloc_class_015_pos']
	tags = ['functional', 'alloc_class']

	[tests/functional/append]
	tests = ['file_append', 'threadsappend_001_pos']
	tags = ['functional', 'append']

	[tests/functional/arc]
	tests = ['dbufstats_001_pos', 'dbufstats_002_pos', 'dbufstats_003_pos',
	'arcstats_runtime_tuning']
	tags = ['functional', 'arc']

	[tests/functional/atime]
	tests = ['atime_001_pos', 'atime_002_neg', 'root_atime_off', 'root_atime_on']
	tags = ['functional', 'atime']

	[tests/functional/bclone]
	tests = ['bclone_crossfs_corner_cases_limited',
	'bclone_crossfs_data',
	'bclone_crossfs_embedded',
	'bclone_crossfs_hole',
	'bclone_diffprops_all',
	'bclone_diffprops_checksum',
	'bclone_diffprops_compress',
	'bclone_diffprops_copies',
	'bclone_diffprops_recordsize',
	'bclone_prop_sync',
	'bclone_samefs_corner_cases_limited',
	'bclone_samefs_data',
	'bclone_samefs_embedded',
	'bclone_samefs_hole']
	tags = ['functional', 'bclone']
	timeout = 7200

	[tests/functional/block_cloning]
	tests = ['block_cloning_clone_mmap_cached',
	'block_cloning_copyfilerange',
	'block_cloning_copyfilerange_partial',
	'block_cloning_copyfilerange_fallback',
	'block_cloning_disabled_copyfilerange',
	'block_cloning_copyfilerange_cross_dataset',
	'block_cloning_cross_enc_dataset',
	'block_cloning_copyfilerange_fallback_same_txg',
	'block_cloning_replay', 'block_cloning_replay_encrypted',
	'block_cloning_lwb_buffer_overflow', 'block_cloning_clone_mmap_write']
	tags = ['functional', 'block_cloning']

	[tests/functional/bootfs]
	tests = ['bootfs_001_pos', 'bootfs_002_neg', 'bootfs_003_pos',
	'bootfs_004_neg', 'bootfs_005_neg', 'bootfs_006_pos', 'bootfs_007_pos',
	'bootfs_008_pos']
	tags = ['functional', 'bootfs']

	[tests/functional/btree]
	tests = ['btree_positive', 'btree_negative']
	tags = ['functional', 'btree']
	pre =
	post =

	[tests/functional/cache]
	tests = ['cache_001_pos', 'cache_002_pos', 'cache_003_pos', 'cache_004_neg',
	'cache_005_neg', 'cache_006_pos', 'cache_007_neg', 'cache_008_neg',
	'cache_009_pos', 'cache_010_pos', 'cache_011_pos', 'cache_012_pos']
	tags = ['functional', 'cache']

	[tests/functional/cachefile]
	tests = ['cachefile_001_pos', 'cachefile_002_pos', 'cachefile_003_pos',
	'cachefile_004_pos']
	tags = ['functional', 'cachefile']

	[tests/functional/casenorm]
	tests = ['case_all_values', 'norm_all_values', 'mixed_create_failure',
	'sensitive_none_lookup', 'sensitive_none_delete',
	'sensitive_formd_lookup', 'sensitive_formd_delete',
	'insensitive_none_lookup', 'insensitive_none_delete',
	'insensitive_formd_lookup', 'insensitive_formd_delete',
	'mixed_none_lookup', 'mixed_none_lookup_ci', 'mixed_none_delete',
	'mixed_formd_lookup', 'mixed_formd_lookup_ci', 'mixed_formd_delete']
	tags = ['functional', 'casenorm']

	[tests/functional/channel_program/lua_core]
	tests = ['tst.args_to_lua', 'tst.divide_by_zero', 'tst.exists',
	'tst.integer_illegal', 'tst.integer_overflow', 'tst.language_functions_neg',
	'tst.language_functions_pos', 'tst.large_prog', 'tst.libraries',
	'tst.memory_limit', 'tst.nested_neg', 'tst.nested_pos', 'tst.nvlist_to_lua',
	'tst.recursive_neg', 'tst.recursive_pos', 'tst.return_large',
	'tst.return_nvlist_neg', 'tst.return_nvlist_pos',
	'tst.return_recursive_table', 'tst.stack_gsub', 'tst.timeout']
	tags = ['functional', 'channel_program', 'lua_core']

	[tests/functional/channel_program/synctask_core]
	tests = ['tst.destroy_fs', 'tst.destroy_snap', 'tst.get_count_and_limit',
	'tst.get_index_props', 'tst.get_mountpoint', 'tst.get_neg',
	'tst.get_number_props', 'tst.get_string_props', 'tst.get_type',
	'tst.get_userquota', 'tst.get_written', 'tst.inherit', 'tst.list_bookmarks',
	'tst.list_children', 'tst.list_clones', 'tst.list_holds',
	'tst.list_snapshots', 'tst.list_system_props',
	'tst.list_user_props', 'tst.parse_args_neg','tst.promote_conflict',
	'tst.promote_multiple', 'tst.promote_simple', 'tst.rollback_mult',
	'tst.rollback_one', 'tst.set_props', 'tst.snapshot_destroy', 'tst.snapshot_neg',
	'tst.snapshot_recursive', 'tst.snapshot_rename', 'tst.snapshot_simple',
	'tst.bookmark.create', 'tst.bookmark.copy',
	'tst.terminate_by_signal'
	]
	tags = ['functional', 'channel_program', 'synctask_core']

	[tests/functional/checksum]
	tests = ['run_edonr_test', 'run_sha2_test', 'run_skein_test', 'run_blake3_test',
	'filetest_001_pos', 'filetest_002_pos']
	tags = ['functional', 'checksum']

	[tests/functional/clean_mirror]
	tests = [ 'clean_mirror_001_pos', 'clean_mirror_002_pos',
	'clean_mirror_003_pos', 'clean_mirror_004_pos']
	tags = ['functional', 'clean_mirror']

	[tests/functional/cli_root/zinject]
	tests = ['zinject_args']
	pre =
	post =
	tags = ['functional', 'cli_root', 'zinject']

	[tests/functional/cli_root/zdb]
	tests = ['zdb_002_pos', 'zdb_003_pos', 'zdb_004_pos', 'zdb_005_pos',
	'zdb_006_pos', 'zdb_args_neg', 'zdb_args_pos',
	'zdb_block_size_histogram', 'zdb_checksum', 'zdb_decompress',
	'zdb_display_block', 'zdb_encrypted', 'zdb_label_checksum',
	'zdb_object_range_neg', 'zdb_object_range_pos', 'zdb_objset_id',
	'zdb_decompress_zstd', 'zdb_recover', 'zdb_recover_2', 'zdb_backup']
	pre =
	post =
	tags = ['functional', 'cli_root', 'zdb']
	timeout = 1200

	[tests/functional/cli_root/zfs]
	tests = ['zfs_001_neg', 'zfs_002_pos']
	tags = ['functional', 'cli_root', 'zfs']

	[tests/functional/cli_root/zfs_bookmark]
	tests = ['zfs_bookmark_cliargs']
	tags = ['functional', 'cli_root', 'zfs_bookmark']

	[tests/functional/cli_root/zfs_change-key]
	tests = ['zfs_change-key', 'zfs_change-key_child', 'zfs_change-key_format',
	'zfs_change-key_inherit', 'zfs_change-key_load', 'zfs_change-key_location',
	'zfs_change-key_pbkdf2iters', 'zfs_change-key_clones']
	tags = ['functional', 'cli_root', 'zfs_change-key']

	[tests/functional/cli_root/zfs_clone]
	tests = ['zfs_clone_001_neg', 'zfs_clone_002_pos', 'zfs_clone_003_pos',
	'zfs_clone_004_pos', 'zfs_clone_005_pos', 'zfs_clone_006_pos',
	'zfs_clone_007_pos', 'zfs_clone_008_neg', 'zfs_clone_009_neg',
	'zfs_clone_010_pos', 'zfs_clone_encrypted', 'zfs_clone_deeply_nested',
	'zfs_clone_rm_nested']
	tags = ['functional', 'cli_root', 'zfs_clone']

	[tests/functional/cli_root/zfs_copies]
	tests = ['zfs_copies_001_pos', 'zfs_copies_002_pos', 'zfs_copies_003_pos',
	'zfs_copies_004_neg', 'zfs_copies_005_neg', 'zfs_copies_006_pos']
	tags = ['functional', 'cli_root', 'zfs_copies']

	[tests/functional/cli_root/zfs_create]
	tests = ['zfs_create_001_pos', 'zfs_create_002_pos', 'zfs_create_003_pos',
	'zfs_create_004_pos', 'zfs_create_005_pos', 'zfs_create_006_pos',
	'zfs_create_007_pos', 'zfs_create_008_neg', 'zfs_create_009_neg',
	'zfs_create_010_neg', 'zfs_create_011_pos', 'zfs_create_012_pos',
	'zfs_create_013_pos', 'zfs_create_014_pos', 'zfs_create_encrypted',
	'zfs_create_crypt_combos', 'zfs_create_dryrun', 'zfs_create_nomount',
	'zfs_create_verbose']
	tags = ['functional', 'cli_root', 'zfs_create']

	[tests/functional/cli_root/zfs_destroy]
	tests = ['zfs_clone_livelist_condense_and_disable',
	'zfs_clone_livelist_condense_races', 'zfs_clone_livelist_dedup',
	'zfs_destroy_001_pos', 'zfs_destroy_002_pos', 'zfs_destroy_003_pos',
	'zfs_destroy_004_pos', 'zfs_destroy_005_neg', 'zfs_destroy_006_neg',
	'zfs_destroy_007_neg', 'zfs_destroy_008_pos', 'zfs_destroy_009_pos',
	'zfs_destroy_010_pos', 'zfs_destroy_011_pos', 'zfs_destroy_012_pos',
	'zfs_destroy_013_neg', 'zfs_destroy_014_pos', 'zfs_destroy_015_pos',
	'zfs_destroy_016_pos', 'zfs_destroy_clone_livelist',
	'zfs_destroy_dev_removal', 'zfs_destroy_dev_removal_condense']
	tags = ['functional', 'cli_root', 'zfs_destroy']

	[tests/functional/cli_root/zfs_diff]
	tests = ['zfs_diff_changes', 'zfs_diff_cliargs', 'zfs_diff_timestamp',
	'zfs_diff_types', 'zfs_diff_encrypted', 'zfs_diff_mangle']
	tags = ['functional', 'cli_root', 'zfs_diff']

	[tests/functional/cli_root/zfs_get]
	tests = ['zfs_get_001_pos', 'zfs_get_002_pos', 'zfs_get_003_pos',
	'zfs_get_004_pos', 'zfs_get_005_neg', 'zfs_get_006_neg', 'zfs_get_007_neg',
	'zfs_get_008_pos', 'zfs_get_009_pos', 'zfs_get_010_neg']
	tags = ['functional', 'cli_root', 'zfs_get']

	[tests/functional/cli_root/zfs_ids_to_path]
	tests = ['zfs_ids_to_path_001_pos']
	tags = ['functional', 'cli_root', 'zfs_ids_to_path']

	[tests/functional/cli_root/zfs_inherit]
	tests = ['zfs_inherit_001_neg', 'zfs_inherit_002_neg', 'zfs_inherit_003_pos',
	'zfs_inherit_mountpoint']
	tags = ['functional', 'cli_root', 'zfs_inherit']

	[tests/functional/cli_root/zfs_load-key]
	tests = ['zfs_load-key', 'zfs_load-key_all', 'zfs_load-key_file',
	'zfs_load-key_https', 'zfs_load-key_location', 'zfs_load-key_noop',
	'zfs_load-key_recursive']
	tags = ['functional', 'cli_root', 'zfs_load-key']

	[tests/functional/cli_root/zfs_mount]
	tests = ['zfs_mount_001_pos', 'zfs_mount_002_pos', 'zfs_mount_003_pos',
	'zfs_mount_004_pos', 'zfs_mount_005_pos', 'zfs_mount_007_pos',
	'zfs_mount_009_neg', 'zfs_mount_010_neg', 'zfs_mount_011_neg',
	'zfs_mount_012_pos', 'zfs_mount_all_001_pos', 'zfs_mount_encrypted',
	'zfs_mount_remount', 'zfs_mount_all_fail', 'zfs_mount_all_mountpoints',
	'zfs_mount_test_race', 'zfs_mount_recursive']
	tags = ['functional', 'cli_root', 'zfs_mount']

	[tests/functional/cli_root/zfs_program]
	tests = ['zfs_program_json']
	tags = ['functional', 'cli_root', 'zfs_program']

	[tests/functional/cli_root/zfs_promote]
	tests = ['zfs_promote_001_pos', 'zfs_promote_002_pos', 'zfs_promote_003_pos',
	'zfs_promote_004_pos', 'zfs_promote_005_pos', 'zfs_promote_006_neg',
	'zfs_promote_007_neg', 'zfs_promote_008_pos', 'zfs_promote_encryptionroot']
	tags = ['functional', 'cli_root', 'zfs_promote']

	[tests/functional/cli_root/zfs_property]
	tests = ['zfs_written_property_001_pos']
	tags = ['functional', 'cli_root', 'zfs_property']

	[tests/functional/cli_root/zfs_receive]
	tests = ['zfs_receive_001_pos', 'zfs_receive_002_pos', 'zfs_receive_003_pos',
	'zfs_receive_004_neg', 'zfs_receive_005_neg', 'zfs_receive_006_pos',
	'zfs_receive_007_neg', 'zfs_receive_008_pos', 'zfs_receive_009_neg',
	'zfs_receive_010_pos', 'zfs_receive_011_pos', 'zfs_receive_012_pos',
	'zfs_receive_013_pos', 'zfs_receive_014_pos', 'zfs_receive_015_pos',
	'zfs_receive_016_pos', 'receive-o-x_props_override',
	'receive-o-x_props_aliases',
	'zfs_receive_from_encrypted', 'zfs_receive_to_encrypted',
	'zfs_receive_raw', 'zfs_receive_raw_incremental', 'zfs_receive_-e',
	'zfs_receive_raw_-d', 'zfs_receive_from_zstd', 'zfs_receive_new_props',
	'zfs_receive_-wR-encrypted-mix', 'zfs_receive_corrective',
	'zfs_receive_compressed_corrective', 'zfs_receive_large_block_corrective']
	tags = ['functional', 'cli_root', 'zfs_receive']

	[tests/functional/cli_root/zfs_rename]
	tests = ['zfs_rename_001_pos', 'zfs_rename_002_pos', 'zfs_rename_003_pos',
	'zfs_rename_004_neg', 'zfs_rename_005_neg', 'zfs_rename_006_pos',
	'zfs_rename_007_pos', 'zfs_rename_008_pos', 'zfs_rename_009_neg',
	'zfs_rename_010_neg', 'zfs_rename_011_pos', 'zfs_rename_012_neg',
	'zfs_rename_013_pos', 'zfs_rename_014_neg', 'zfs_rename_encrypted_child',
	'zfs_rename_to_encrypted', 'zfs_rename_mountpoint', 'zfs_rename_nounmount']
	tags = ['functional', 'cli_root', 'zfs_rename']

	[tests/functional/cli_root/zfs_reservation]
	tests = ['zfs_reservation_001_pos', 'zfs_reservation_002_pos']
	tags = ['functional', 'cli_root', 'zfs_reservation']

	[tests/functional/cli_root/zfs_rollback]
	tests = ['zfs_rollback_001_pos', 'zfs_rollback_002_pos',
	'zfs_rollback_003_neg', 'zfs_rollback_004_neg']
	tags = ['functional', 'cli_root', 'zfs_rollback']

	[tests/functional/cli_root/zfs_send]
	tests = ['zfs_send_001_pos', 'zfs_send_002_pos', 'zfs_send_003_pos',
	'zfs_send_004_neg', 'zfs_send_005_pos', 'zfs_send_006_pos',
	'zfs_send_007_pos', 'zfs_send_encrypted', 'zfs_send_encrypted_unloaded',
	'zfs_send_raw', 'zfs_send_sparse', 'zfs_send-b', 'zfs_send_skip_missing']
	tags = ['functional', 'cli_root', 'zfs_send']

	[tests/functional/cli_root/zfs_set]
	tests = ['cache_001_pos', 'cache_002_neg', 'canmount_001_pos',
	'canmount_002_pos', 'canmount_003_pos', 'canmount_004_pos',
	'checksum_001_pos', 'compression_001_pos', 'mountpoint_001_pos',
	'mountpoint_002_pos', 'reservation_001_neg', 'user_property_002_pos',
	'share_mount_001_neg', 'snapdir_001_pos', 'onoffs_001_pos',
	'user_property_001_pos', 'user_property_003_neg', 'readonly_001_pos',
	'user_property_004_pos', 'version_001_neg', 'zfs_set_001_neg',
	'zfs_set_002_neg', 'zfs_set_003_neg', 'property_alias_001_pos',
	'mountpoint_003_pos', 'ro_props_001_pos', 'zfs_set_keylocation',
	'zfs_set_feature_activation', 'zfs_set_nomount']
	tags = ['functional', 'cli_root', 'zfs_set']

	[tests/functional/cli_root/zfs_share]
	tests = ['zfs_share_001_pos', 'zfs_share_002_pos', 'zfs_share_003_pos',
	'zfs_share_004_pos', 'zfs_share_006_pos', 'zfs_share_008_neg',
	'zfs_share_010_neg', 'zfs_share_011_pos', 'zfs_share_concurrent_shares',
	'zfs_share_after_mount']
	tags = ['functional', 'cli_root', 'zfs_share']

	[tests/functional/cli_root/zfs_snapshot]
	tests = ['zfs_snapshot_001_neg', 'zfs_snapshot_002_neg',
	'zfs_snapshot_003_neg', 'zfs_snapshot_004_neg', 'zfs_snapshot_005_neg',
	'zfs_snapshot_006_pos', 'zfs_snapshot_007_neg', 'zfs_snapshot_008_neg',
	'zfs_snapshot_009_pos']
	tags = ['functional', 'cli_root', 'zfs_snapshot']

	[tests/functional/cli_root/zfs_unload-key]
	tests = ['zfs_unload-key', 'zfs_unload-key_all', 'zfs_unload-key_recursive']
	tags = ['functional', 'cli_root', 'zfs_unload-key']

	[tests/functional/cli_root/zfs_unmount]
	tests = ['zfs_unmount_001_pos', 'zfs_unmount_002_pos', 'zfs_unmount_003_pos',
	'zfs_unmount_004_pos', 'zfs_unmount_005_pos', 'zfs_unmount_006_pos',
	'zfs_unmount_007_neg', 'zfs_unmount_008_neg', 'zfs_unmount_009_pos',
	'zfs_unmount_all_001_pos', 'zfs_unmount_nested', 'zfs_unmount_unload_keys']
	tags = ['functional', 'cli_root', 'zfs_unmount']

	[tests/functional/cli_root/zfs_unshare]
	tests = ['zfs_unshare_001_pos', 'zfs_unshare_002_pos', 'zfs_unshare_003_pos',
	'zfs_unshare_004_neg', 'zfs_unshare_005_neg', 'zfs_unshare_006_pos',
	'zfs_unshare_007_pos']
	tags = ['functional', 'cli_root', 'zfs_unshare']

	[tests/functional/cli_root/zfs_upgrade]
	tests = ['zfs_upgrade_001_pos', 'zfs_upgrade_002_pos', 'zfs_upgrade_003_pos',
	'zfs_upgrade_004_pos', 'zfs_upgrade_005_pos', 'zfs_upgrade_006_neg',
	'zfs_upgrade_007_neg']
	tags = ['functional', 'cli_root', 'zfs_upgrade']

	[tests/functional/cli_root/zfs_wait]
	tests = ['zfs_wait_deleteq', 'zfs_wait_getsubopt']
	tags = ['functional', 'cli_root', 'zfs_wait']

	[tests/functional/cli_root/zhack]
	tests = ['zhack_label_repair_001', 'zhack_label_repair_002',
	'zhack_label_repair_003', 'zhack_label_repair_004']
	pre =
	post =
	tags = ['functional', 'cli_root', 'zhack']

	[tests/functional/cli_root/zpool]
	tests = ['zpool_001_neg', 'zpool_002_pos', 'zpool_003_pos', 'zpool_colors']
	tags = ['functional', 'cli_root', 'zpool']

	[tests/functional/cli_root/zpool_add]
	tests = ['zpool_add_001_pos', 'zpool_add_002_pos', 'zpool_add_003_pos',
	'zpool_add_004_pos', 'zpool_add_006_pos', 'zpool_add_007_neg',
	'zpool_add_008_neg', 'zpool_add_009_neg', 'zpool_add_010_pos',
	'add-o_ashift', 'add_prop_ashift', 'zpool_add_dryrun_output',
	'zpool_add--allow-ashift-mismatch']
	tags = ['functional', 'cli_root', 'zpool_add']

	[tests/functional/cli_root/zpool_attach]
	tests = ['zpool_attach_001_neg', 'attach-o_ashift']
	tags = ['functional', 'cli_root', 'zpool_attach']

	[tests/functional/cli_root/zpool_clear]
	tests = ['zpool_clear_001_pos', 'zpool_clear_002_neg', 'zpool_clear_003_neg',
	'zpool_clear_readonly']
	tags = ['functional', 'cli_root', 'zpool_clear']

	[tests/functional/cli_root/zpool_create]
	tests = ['zpool_create_001_pos', 'zpool_create_002_pos',
	'zpool_create_003_pos', 'zpool_create_004_pos', 'zpool_create_005_pos',
	'zpool_create_006_pos', 'zpool_create_007_neg', 'zpool_create_008_pos',
	'zpool_create_009_neg', 'zpool_create_010_neg', 'zpool_create_011_neg',
	'zpool_create_012_neg', 'zpool_create_014_neg', 'zpool_create_015_neg',
	'zpool_create_017_neg', 'zpool_create_018_pos', 'zpool_create_019_pos',
	'zpool_create_020_pos', 'zpool_create_021_pos', 'zpool_create_022_pos',
	'zpool_create_023_neg', 'zpool_create_024_pos',
	'zpool_create_encrypted', 'zpool_create_crypt_combos',
	'zpool_create_draid_001_pos', 'zpool_create_draid_002_pos',
	'zpool_create_draid_003_pos', 'zpool_create_draid_004_pos',
	'zpool_create_features_001_pos', 'zpool_create_features_002_pos',
	'zpool_create_features_003_pos', 'zpool_create_features_004_neg',
	'zpool_create_features_005_pos', 'zpool_create_features_006_pos',
	'zpool_create_features_007_pos', 'zpool_create_features_008_pos',
	'zpool_create_features_009_pos', 'create-o_ashift',
	'zpool_create_tempname', 'zpool_create_dryrun_output']
	tags = ['functional', 'cli_root', 'zpool_create']

	[tests/functional/cli_root/zpool_destroy]
	tests = ['zpool_destroy_001_pos', 'zpool_destroy_002_pos',
	'zpool_destroy_003_neg']
	pre =
	post =
	tags = ['functional', 'cli_root', 'zpool_destroy']

	[tests/functional/cli_root/zpool_detach]
	tests = ['zpool_detach_001_neg']
	tags = ['functional', 'cli_root', 'zpool_detach']

	[tests/functional/cli_root/zpool_events]
	tests = ['zpool_events_clear', 'zpool_events_cliargs', 'zpool_events_follow',
	'zpool_events_poolname', 'zpool_events_errors', 'zpool_events_duplicates',
	'zpool_events_clear_retained']
	tags = ['functional', 'cli_root', 'zpool_events']

	[tests/functional/cli_root/zpool_export]
	tests = ['zpool_export_001_pos', 'zpool_export_002_pos',
	- 'zpool_export_003_neg', 'zpool_export_004_pos']
	+ 'zpool_export_003_neg', 'zpool_export_004_pos',
	+ 'zpool_export_parallel_pos', 'zpool_export_parallel_admin']
	tags = ['functional', 'cli_root', 'zpool_export']

	[tests/functional/cli_root/zpool_get]
	tests = ['zpool_get_001_pos', 'zpool_get_002_pos', 'zpool_get_003_pos',
	'zpool_get_004_neg', 'zpool_get_005_pos', 'vdev_get_001_pos']
	tags = ['functional', 'cli_root', 'zpool_get']

	[tests/functional/cli_root/zpool_history]
	tests = ['zpool_history_001_neg', 'zpool_history_002_pos']
	tags = ['functional', 'cli_root', 'zpool_history']

	[tests/functional/cli_root/zpool_import]
	tests = ['zpool_import_001_pos', 'zpool_import_002_pos',
	'zpool_import_003_pos', 'zpool_import_004_pos', 'zpool_import_005_pos',
	'zpool_import_006_pos', 'zpool_import_007_pos', 'zpool_import_008_pos',
	'zpool_import_009_neg', 'zpool_import_010_pos', 'zpool_import_011_neg',
	'zpool_import_012_pos', 'zpool_import_013_neg', 'zpool_import_014_pos',
	'zpool_import_015_pos', 'zpool_import_016_pos', 'zpool_import_017_pos',
	'zpool_import_features_001_pos', 'zpool_import_features_002_neg',
	'zpool_import_features_003_pos', 'zpool_import_missing_001_pos',
	'zpool_import_missing_002_pos', 'zpool_import_missing_003_pos',
	'zpool_import_rename_001_pos', 'zpool_import_all_001_pos',
	'zpool_import_encrypted', 'zpool_import_encrypted_load',
	'zpool_import_errata3', 'zpool_import_errata4',
	'import_cachefile_device_added',
	'import_cachefile_device_removed',
	'import_cachefile_device_replaced',
	'import_cachefile_mirror_attached',
	'import_cachefile_mirror_detached',
	'import_cachefile_paths_changed',
	'import_cachefile_shared_device',
	'import_devices_missing', 'import_log_missing',
	'import_paths_changed',
	'import_rewind_config_changed',
	'import_rewind_device_replaced',
	'zpool_import_status', 'zpool_import_parallel_pos',
	'zpool_import_parallel_neg', 'zpool_import_parallel_admin']
	tags = ['functional', 'cli_root', 'zpool_import']
	timeout = 1200

	[tests/functional/cli_root/zpool_labelclear]
	tests = ['zpool_labelclear_active', 'zpool_labelclear_exported',
	'zpool_labelclear_removed', 'zpool_labelclear_valid']
	pre =
	post =
	tags = ['functional', 'cli_root', 'zpool_labelclear']

	[tests/functional/cli_root/zpool_initialize]
	tests = ['zpool_initialize_attach_detach_add_remove',
	'zpool_initialize_fault_export_import_online',
	'zpool_initialize_import_export',
	'zpool_initialize_offline_export_import_online',
	'zpool_initialize_online_offline',
	'zpool_initialize_split',
	'zpool_initialize_start_and_cancel_neg',
	'zpool_initialize_start_and_cancel_pos',
	'zpool_initialize_suspend_resume',
	'zpool_initialize_uninit',
	'zpool_initialize_unsupported_vdevs',
	'zpool_initialize_verify_checksums',
	'zpool_initialize_verify_initialized']
	pre =
	tags = ['functional', 'cli_root', 'zpool_initialize']

	[tests/functional/cli_root/zpool_offline]
	tests = ['zpool_offline_001_pos', 'zpool_offline_002_neg',
	'zpool_offline_003_pos']
	tags = ['functional', 'cli_root', 'zpool_offline']

	[tests/functional/cli_root/zpool_online]
	tests = ['zpool_online_001_pos', 'zpool_online_002_neg']
	tags = ['functional', 'cli_root', 'zpool_online']

	[tests/functional/cli_root/zpool_remove]
	tests = ['zpool_remove_001_neg', 'zpool_remove_002_pos',
	'zpool_remove_003_pos']
	tags = ['functional', 'cli_root', 'zpool_remove']

	[tests/functional/cli_root/zpool_replace]
	tests = ['zpool_replace_001_neg', 'replace-o_ashift', 'replace_prop_ashift']
	tags = ['functional', 'cli_root', 'zpool_replace']

	[tests/functional/cli_root/zpool_resilver]
	tests = ['zpool_resilver_bad_args', 'zpool_resilver_restart',
	'zpool_resilver_concurrent']
	tags = ['functional', 'cli_root', 'zpool_resilver']

	[tests/functional/cli_root/zpool_scrub]
	tests = ['zpool_scrub_001_neg', 'zpool_scrub_002_pos', 'zpool_scrub_003_pos',
	'zpool_scrub_004_pos', 'zpool_scrub_005_pos',
	'zpool_scrub_encrypted_unloaded', 'zpool_scrub_print_repairing',
	'zpool_scrub_offline_device', 'zpool_scrub_multiple_copies',
	'zpool_error_scrub_001_pos', 'zpool_error_scrub_002_pos',
	'zpool_error_scrub_003_pos', 'zpool_error_scrub_004_pos']
	tags = ['functional', 'cli_root', 'zpool_scrub']

	[tests/functional/cli_root/zpool_set]
	tests = ['zpool_set_001_pos', 'zpool_set_002_neg', 'zpool_set_003_neg',
	'zpool_set_ashift', 'zpool_set_features', 'vdev_set_001_pos',
	'user_property_001_pos', 'user_property_002_neg']
	tags = ['functional', 'cli_root', 'zpool_set']

	[tests/functional/cli_root/zpool_split]
	tests = ['zpool_split_cliargs', 'zpool_split_devices',
	'zpool_split_encryption', 'zpool_split_props', 'zpool_split_vdevs',
	'zpool_split_resilver', 'zpool_split_indirect',
	'zpool_split_dryrun_output']
	tags = ['functional', 'cli_root', 'zpool_split']

	[tests/functional/cli_root/zpool_status]
	tests = ['zpool_status_001_pos', 'zpool_status_002_pos',
	'zpool_status_003_pos', 'zpool_status_004_pos',
	'zpool_status_005_pos', 'zpool_status_006_pos',
	'zpool_status_007_pos', 'zpool_status_008_pos',
	'zpool_status_features_001_pos']
	tags = ['functional', 'cli_root', 'zpool_status']

	[tests/functional/cli_root/zpool_sync]
	tests = ['zpool_sync_001_pos', 'zpool_sync_002_neg']
	tags = ['functional', 'cli_root', 'zpool_sync']

	[tests/functional/cli_root/zpool_trim]
	tests = ['zpool_trim_attach_detach_add_remove',
	'zpool_trim_fault_export_import_online',
	'zpool_trim_import_export', 'zpool_trim_multiple', 'zpool_trim_neg',
	'zpool_trim_offline_export_import_online', 'zpool_trim_online_offline',
	'zpool_trim_partial', 'zpool_trim_rate', 'zpool_trim_rate_neg',
	'zpool_trim_secure', 'zpool_trim_split', 'zpool_trim_start_and_cancel_neg',
	'zpool_trim_start_and_cancel_pos', 'zpool_trim_suspend_resume',
	'zpool_trim_unsupported_vdevs', 'zpool_trim_verify_checksums',
	'zpool_trim_verify_trimmed']
	tags = ['functional', 'zpool_trim']

	[tests/functional/cli_root/zpool_upgrade]
	tests = ['zpool_upgrade_001_pos', 'zpool_upgrade_002_pos',
	'zpool_upgrade_003_pos', 'zpool_upgrade_004_pos',
	'zpool_upgrade_005_neg', 'zpool_upgrade_006_neg',
	'zpool_upgrade_007_pos', 'zpool_upgrade_008_pos',
	'zpool_upgrade_009_neg', 'zpool_upgrade_features_001_pos']
	tags = ['functional', 'cli_root', 'zpool_upgrade']

	[tests/functional/cli_root/zpool_wait]
	tests = ['zpool_wait_discard', 'zpool_wait_freeing',
	'zpool_wait_initialize_basic', 'zpool_wait_initialize_cancel',
	'zpool_wait_initialize_flag', 'zpool_wait_multiple',
	'zpool_wait_no_activity', 'zpool_wait_remove', 'zpool_wait_remove_cancel',
	'zpool_wait_trim_basic', 'zpool_wait_trim_cancel', 'zpool_wait_trim_flag',
	'zpool_wait_usage']
	tags = ['functional', 'cli_root', 'zpool_wait']

	[tests/functional/cli_root/zpool_wait/scan]
	tests = ['zpool_wait_replace_cancel', 'zpool_wait_rebuild',
	'zpool_wait_resilver', 'zpool_wait_scrub_cancel',
	'zpool_wait_replace', 'zpool_wait_scrub_basic', 'zpool_wait_scrub_flag']
	tags = ['functional', 'cli_root', 'zpool_wait']

	[tests/functional/cli_user/misc]
	tests = ['zdb_001_neg', 'zfs_001_neg', 'zfs_allow_001_neg',
	'zfs_clone_001_neg', 'zfs_create_001_neg', 'zfs_destroy_001_neg',
	'zfs_get_001_neg', 'zfs_inherit_001_neg', 'zfs_mount_001_neg',
	'zfs_promote_001_neg', 'zfs_receive_001_neg', 'zfs_rename_001_neg',
	'zfs_rollback_001_neg', 'zfs_send_001_neg', 'zfs_set_001_neg',
	'zfs_share_001_neg', 'zfs_snapshot_001_neg', 'zfs_unallow_001_neg',
	'zfs_unmount_001_neg', 'zfs_unshare_001_neg', 'zfs_upgrade_001_neg',
	'zpool_001_neg', 'zpool_add_001_neg', 'zpool_attach_001_neg',
	'zpool_clear_001_neg', 'zpool_create_001_neg', 'zpool_destroy_001_neg',
	'zpool_detach_001_neg', 'zpool_export_001_neg', 'zpool_get_001_neg',
	'zpool_history_001_neg', 'zpool_import_001_neg', 'zpool_import_002_neg',
	'zpool_offline_001_neg', 'zpool_online_001_neg', 'zpool_remove_001_neg',
	'zpool_replace_001_neg', 'zpool_scrub_001_neg', 'zpool_set_001_neg',
	'zpool_status_001_neg', 'zpool_upgrade_001_neg', 'arcstat_001_pos',
	'arc_summary_001_pos', 'arc_summary_002_neg', 'zpool_wait_privilege',
	'zilstat_001_pos']
	user =
	tags = ['functional', 'cli_user', 'misc']

	[tests/functional/cli_user/zfs_list]
	tests = ['zfs_list_001_pos', 'zfs_list_002_pos', 'zfs_list_003_pos',
	'zfs_list_004_neg', 'zfs_list_005_neg', 'zfs_list_007_pos',
	'zfs_list_008_neg']
	user =
	tags = ['functional', 'cli_user', 'zfs_list']

	[tests/functional/cli_user/zpool_iostat]
	tests = ['zpool_iostat_001_neg', 'zpool_iostat_002_pos',
	'zpool_iostat_003_neg', 'zpool_iostat_004_pos',
	'zpool_iostat_005_pos', 'zpool_iostat_-c_disable',
	'zpool_iostat_-c_homedir', 'zpool_iostat_-c_searchpath']
	user =
	tags = ['functional', 'cli_user', 'zpool_iostat']

	[tests/functional/cli_user/zpool_list]
	tests = ['zpool_list_001_pos', 'zpool_list_002_neg']
	user =
	tags = ['functional', 'cli_user', 'zpool_list']

	[tests/functional/cli_user/zpool_status]
	tests = ['zpool_status_003_pos', 'zpool_status_-c_disable',
	'zpool_status_-c_homedir', 'zpool_status_-c_searchpath']
	user =
	tags = ['functional', 'cli_user', 'zpool_status']

	[tests/functional/compression]
	tests = ['compress_001_pos', 'compress_002_pos', 'compress_003_pos',
	'l2arc_compressed_arc', 'l2arc_compressed_arc_disabled',
	'l2arc_encrypted', 'l2arc_encrypted_no_compressed_arc']
	tags = ['functional', 'compression']

	[tests/functional/cp_files]
	tests = ['cp_files_001_pos', 'cp_files_002_pos', 'cp_stress']
	tags = ['functional', 'cp_files']

	[tests/functional/zap_shrink]
	tests = ['zap_shrink_001_pos']
	tags = ['functional', 'zap_shrink']

	[tests/functional/crtime]
	tests = ['crtime_001_pos' ]
	tags = ['functional', 'crtime']

	[tests/functional/ctime]
	tests = ['ctime_001_pos' ]
	tags = ['functional', 'ctime']

	[tests/functional/deadman]
	tests = ['deadman_ratelimit', 'deadman_sync', 'deadman_zio']
	pre =
	post =
	tags = ['functional', 'deadman']

	[tests/functional/delegate]
	tests = ['zfs_allow_001_pos', 'zfs_allow_002_pos', 'zfs_allow_003_pos',
	'zfs_allow_004_pos', 'zfs_allow_005_pos', 'zfs_allow_006_pos',
	'zfs_allow_007_pos', 'zfs_allow_008_pos', 'zfs_allow_009_neg',
	'zfs_allow_010_pos', 'zfs_allow_011_neg', 'zfs_allow_012_neg',
	'zfs_unallow_001_pos', 'zfs_unallow_002_pos', 'zfs_unallow_003_pos',
	'zfs_unallow_004_pos', 'zfs_unallow_005_pos', 'zfs_unallow_006_pos',
	'zfs_unallow_007_neg', 'zfs_unallow_008_neg']
	tags = ['functional', 'delegate']

	[tests/functional/exec]
	tests = ['exec_001_pos', 'exec_002_neg']
	tags = ['functional', 'exec']

	[tests/functional/fallocate]
	tests = ['fallocate_punch-hole']
	tags = ['functional', 'fallocate']

	[tests/functional/features/async_destroy]
	tests = ['async_destroy_001_pos']
	tags = ['functional', 'features', 'async_destroy']

	[tests/functional/features/large_dnode]
	tests = ['large_dnode_001_pos', 'large_dnode_003_pos', 'large_dnode_004_neg',
	'large_dnode_005_pos', 'large_dnode_007_neg', 'large_dnode_009_pos']
	tags = ['functional', 'features', 'large_dnode']

	[tests/functional/grow]
	pre =
	post =
	tests = ['grow_pool_001_pos', 'grow_replicas_001_pos']
	tags = ['functional', 'grow']

	[tests/functional/history]
	tests = ['history_001_pos', 'history_002_pos', 'history_003_pos',
	'history_004_pos', 'history_005_neg', 'history_006_neg',
	'history_007_pos', 'history_008_pos', 'history_009_pos',
	'history_010_pos']
	tags = ['functional', 'history']

	[tests/functional/hkdf]
	pre =
	post =
	tests = ['hkdf_test']
	tags = ['functional', 'hkdf']

	[tests/functional/inheritance]
	tests = ['inherit_001_pos']
	pre =
	tags = ['functional', 'inheritance']

	[tests/functional/io]
	tests = ['sync', 'psync', 'posixaio', 'mmap']
	tags = ['functional', 'io']

	[tests/functional/inuse]
	tests = ['inuse_004_pos', 'inuse_005_pos', 'inuse_008_pos', 'inuse_009_pos']
	post =
	tags = ['functional', 'inuse']

	[tests/functional/large_files]
	tests = ['large_files_001_pos', 'large_files_002_pos']
	tags = ['functional', 'large_files']

	[tests/functional/limits]
	tests = ['filesystem_count', 'filesystem_limit', 'snapshot_count',
	'snapshot_limit']
	tags = ['functional', 'limits']

	[tests/functional/link_count]
	tests = ['link_count_001', 'link_count_root_inode']
	tags = ['functional', 'link_count']

	[tests/functional/migration]
	tests = ['migration_001_pos', 'migration_002_pos', 'migration_003_pos',
	'migration_004_pos', 'migration_005_pos', 'migration_006_pos',
	'migration_007_pos', 'migration_008_pos', 'migration_009_pos',
	'migration_010_pos', 'migration_011_pos', 'migration_012_pos']
	tags = ['functional', 'migration']

	[tests/functional/mmap]
	tests = ['mmap_mixed', 'mmap_read_001_pos', 'mmap_seek_001_pos',
	'mmap_sync_001_pos', 'mmap_write_001_pos']
	tags = ['functional', 'mmap']

	[tests/functional/mount]
	tests = ['umount_001', 'umountall_001']
	tags = ['functional', 'mount']

	[tests/functional/mv_files]
	tests = ['mv_files_001_pos', 'mv_files_002_pos', 'random_creation']
	tags = ['functional', 'mv_files']

	[tests/functional/nestedfs]
	tests = ['nestedfs_001_pos']
	tags = ['functional', 'nestedfs']

	[tests/functional/no_space]
	tests = ['enospc_001_pos', 'enospc_002_pos', 'enospc_003_pos',
	'enospc_df', 'enospc_ganging', 'enospc_rm']
	tags = ['functional', 'no_space']

	[tests/functional/nopwrite]
	tests = ['nopwrite_copies', 'nopwrite_mtime', 'nopwrite_negative',
	'nopwrite_promoted_clone', 'nopwrite_recsize', 'nopwrite_sync',
	'nopwrite_varying_compression', 'nopwrite_volume']
	tags = ['functional', 'nopwrite']

	[tests/functional/online_offline]
	tests = ['online_offline_001_pos', 'online_offline_002_neg',
	'online_offline_003_neg']
	tags = ['functional', 'online_offline']

	[tests/functional/pool_checkpoint]
	tests = ['checkpoint_after_rewind', 'checkpoint_big_rewind',
	'checkpoint_capacity', 'checkpoint_conf_change', 'checkpoint_discard',
	'checkpoint_discard_busy', 'checkpoint_discard_many',
	'checkpoint_indirect', 'checkpoint_invalid', 'checkpoint_lun_expsz',
	'checkpoint_open', 'checkpoint_removal', 'checkpoint_rewind',
	'checkpoint_ro_rewind', 'checkpoint_sm_scale', 'checkpoint_twice',
	'checkpoint_vdev_add', 'checkpoint_zdb', 'checkpoint_zhack_feat']
	tags = ['functional', 'pool_checkpoint']
	timeout = 1800

	[tests/functional/pool_names]
	tests = ['pool_names_001_pos', 'pool_names_002_neg']
	pre =
	post =
	tags = ['functional', 'pool_names']

	[tests/functional/poolversion]
	tests = ['poolversion_001_pos', 'poolversion_002_pos']
	tags = ['functional', 'poolversion']

	[tests/functional/pyzfs]
	tests = ['pyzfs_unittest']
	pre =
	post =
	tags = ['functional', 'pyzfs']

	[tests/functional/quota]
	tests = ['quota_001_pos', 'quota_002_pos', 'quota_003_pos',
	'quota_004_pos', 'quota_005_pos', 'quota_006_neg']
	tags = ['functional', 'quota']

	[tests/functional/redacted_send]
	tests = ['redacted_compressed', 'redacted_contents', 'redacted_deleted',
	'redacted_disabled_feature', 'redacted_embedded', 'redacted_holes',
	'redacted_incrementals', 'redacted_largeblocks', 'redacted_many_clones',
	'redacted_mixed_recsize', 'redacted_mounts', 'redacted_negative',
	'redacted_origin', 'redacted_panic', 'redacted_props', 'redacted_resume',
	'redacted_size', 'redacted_volume']
	tags = ['functional', 'redacted_send']

	[tests/functional/raidz]
	tests = ['raidz_001_neg', 'raidz_002_pos', 'raidz_expand_001_pos',
	'raidz_expand_002_pos', 'raidz_expand_003_neg', 'raidz_expand_003_pos',
	'raidz_expand_004_pos', 'raidz_expand_005_pos', 'raidz_expand_006_neg',
	'raidz_expand_007_neg']
	tags = ['functional', 'raidz']
	timeout = 1200

	[tests/functional/redundancy]
	tests = ['redundancy_draid', 'redundancy_draid1', 'redundancy_draid2',
	'redundancy_draid3', 'redundancy_draid_damaged1',
	'redundancy_draid_damaged2', 'redundancy_draid_spare1',
	'redundancy_draid_spare2', 'redundancy_draid_spare3', 'redundancy_mirror',
	'redundancy_raidz', 'redundancy_raidz1', 'redundancy_raidz2',
	'redundancy_raidz3', 'redundancy_stripe']
	tags = ['functional', 'redundancy']
	timeout = 1200

	[tests/functional/refquota]
	tests = ['refquota_001_pos', 'refquota_002_pos', 'refquota_003_pos',
	'refquota_004_pos', 'refquota_005_pos', 'refquota_006_neg',
	'refquota_007_neg', 'refquota_008_neg']
	tags = ['functional', 'refquota']

	[tests/functional/refreserv]
	tests = ['refreserv_001_pos', 'refreserv_002_pos', 'refreserv_003_pos',
	'refreserv_004_pos', 'refreserv_005_pos', 'refreserv_multi_raidz',
	'refreserv_raidz']
	tags = ['functional', 'refreserv']

	[tests/functional/removal]
	pre =
	tests = ['removal_all_vdev', 'removal_cancel', 'removal_check_space',
	'removal_condense_export', 'removal_multiple_indirection',
	'removal_nopwrite', 'removal_remap_deadlists',
	'removal_resume_export', 'removal_sanity', 'removal_with_add',
	'removal_with_create_fs', 'removal_with_dedup',
	'removal_with_errors', 'removal_with_export', 'removal_with_indirect',
	'removal_with_ganging', 'removal_with_faulted',
	'removal_with_remove', 'removal_with_scrub', 'removal_with_send',
	'removal_with_send_recv', 'removal_with_snapshot',
	'removal_with_write', 'removal_with_zdb', 'remove_expanded',
	'remove_mirror', 'remove_mirror_sanity', 'remove_raidz',
	'remove_indirect', 'remove_attach_mirror', 'removal_reservation']
	tags = ['functional', 'removal']

	[tests/functional/rename_dirs]
	tests = ['rename_dirs_001_pos']
	tags = ['functional', 'rename_dirs']

	[tests/functional/replacement]
	tests = ['attach_import', 'attach_multiple', 'attach_rebuild',
	'attach_resilver', 'detach', 'rebuild_disabled_feature',
	'rebuild_multiple', 'rebuild_raidz', 'replace_import', 'replace_rebuild',
	'replace_resilver', 'resilver_restart_001', 'resilver_restart_002',
	'scrub_cancel']
	tags = ['functional', 'replacement']

	[tests/functional/reservation]
	tests = ['reservation_001_pos', 'reservation_002_pos', 'reservation_003_pos',
	'reservation_004_pos', 'reservation_005_pos', 'reservation_006_pos',
	'reservation_007_pos', 'reservation_008_pos', 'reservation_009_pos',
	'reservation_010_pos', 'reservation_011_pos', 'reservation_012_pos',
	'reservation_013_pos', 'reservation_014_pos', 'reservation_015_pos',
	'reservation_016_pos', 'reservation_017_pos', 'reservation_018_pos',
	'reservation_019_pos', 'reservation_020_pos', 'reservation_021_neg',
	'reservation_022_pos']
	tags = ['functional', 'reservation']

	[tests/functional/rootpool]
	tests = ['rootpool_002_neg', 'rootpool_003_neg', 'rootpool_007_pos']
	tags = ['functional', 'rootpool']

	[tests/functional/rsend]
	tests = ['recv_dedup', 'recv_dedup_encrypted_zvol', 'rsend_001_pos',
	'rsend_002_pos', 'rsend_003_pos', 'rsend_004_pos', 'rsend_005_pos',
	'rsend_006_pos', 'rsend_007_pos', 'rsend_008_pos', 'rsend_009_pos',
	'rsend_010_pos', 'rsend_011_pos', 'rsend_012_pos', 'rsend_013_pos',
	'rsend_014_pos', 'rsend_016_neg', 'rsend_019_pos', 'rsend_020_pos',
	'rsend_021_pos', 'rsend_022_pos', 'rsend_024_pos', 'rsend_025_pos',
	'rsend_026_neg', 'rsend_027_pos', 'rsend_028_neg', 'rsend_029_neg',
	'rsend_030_pos', 'rsend_031_pos', 'send-c_verify_ratio',
	'send-c_verify_contents', 'send-c_props', 'send-c_incremental',
	'send-c_volume', 'send-c_zstream_recompress', 'send-c_zstreamdump',
	'send-c_lz4_disabled', 'send-c_recv_lz4_disabled',
	'send-c_mixed_compression', 'send-c_stream_size_estimate',
	'send-c_embedded_blocks', 'send-c_resume', 'send-cpL_varied_recsize',
	'send-c_recv_dedup', 'send-L_toggle', 'send_encrypted_incremental',
	'send_encrypted_freeobjects', 'send_encrypted_hierarchy',
	'send_encrypted_props', 'send_encrypted_truncated_files',
	'send_freeobjects', 'send_realloc_files', 'send_realloc_encrypted_files',
	'send_spill_block', 'send_holds', 'send_hole_birth', 'send_mixed_raw',
	'send-wR_encrypted_zvol', 'send_partial_dataset', 'send_invalid',
	'send_doall', 'send_raw_spill_block', 'send_raw_ashift',
	'send_raw_large_blocks']
	tags = ['functional', 'rsend']

	[tests/functional/scrub_mirror]
	tests = ['scrub_mirror_001_pos', 'scrub_mirror_002_pos',
	'scrub_mirror_003_pos', 'scrub_mirror_004_pos']
	tags = ['functional', 'scrub_mirror']

	[tests/functional/slog]
	tests = ['slog_001_pos', 'slog_002_pos', 'slog_003_pos', 'slog_004_pos',
	'slog_005_pos', 'slog_006_pos', 'slog_007_pos', 'slog_008_neg',
	'slog_009_neg', 'slog_010_neg', 'slog_011_neg', 'slog_012_neg',
	'slog_013_pos', 'slog_014_pos', 'slog_015_neg', 'slog_replay_fs_001',
	'slog_replay_fs_002', 'slog_replay_volume', 'slog_016_pos']
	tags = ['functional', 'slog']

	[tests/functional/snapshot]
	tests = ['clone_001_pos', 'rollback_001_pos', 'rollback_002_pos',
	'rollback_003_pos', 'snapshot_001_pos', 'snapshot_002_pos',
	'snapshot_003_pos', 'snapshot_004_pos', 'snapshot_005_pos',
	'snapshot_006_pos', 'snapshot_007_pos', 'snapshot_008_pos',
	'snapshot_009_pos', 'snapshot_010_pos', 'snapshot_011_pos',
	'snapshot_012_pos', 'snapshot_013_pos', 'snapshot_014_pos',
	'snapshot_017_pos', 'snapshot_018_pos']
	tags = ['functional', 'snapshot']

	[tests/functional/snapused]
	tests = ['snapused_001_pos', 'snapused_002_pos', 'snapused_003_pos',
	'snapused_004_pos', 'snapused_005_pos']
	tags = ['functional', 'snapused']

	[tests/functional/sparse]
	tests = ['sparse_001_pos']
	tags = ['functional', 'sparse']

	[tests/functional/stat]
	tests = ['stat_001_pos']
	tags = ['functional', 'stat']

	[tests/functional/suid]
	tests = ['suid_write_to_suid', 'suid_write_to_sgid', 'suid_write_to_suid_sgid',
	'suid_write_to_none', 'suid_write_zil_replay']
	tags = ['functional', 'suid']

	[tests/functional/trim]
	tests = ['autotrim_integrity', 'autotrim_config', 'autotrim_trim_integrity',
	'trim_integrity', 'trim_config', 'trim_l2arc']
	tags = ['functional', 'trim']

	[tests/functional/truncate]
	tests = ['truncate_001_pos', 'truncate_002_pos', 'truncate_timestamps']
	tags = ['functional', 'truncate']

	[tests/functional/upgrade]
	tests = ['upgrade_userobj_001_pos', 'upgrade_readonly_pool']
	tags = ['functional', 'upgrade']

	[tests/functional/userquota]
	tests = [
	'userquota_001_pos', 'userquota_002_pos', 'userquota_003_pos',
	'userquota_004_pos', 'userquota_005_neg', 'userquota_006_pos',
	'userquota_007_pos', 'userquota_008_pos', 'userquota_009_pos',
	'userquota_010_pos', 'userquota_011_pos', 'userquota_012_neg',
	'userspace_001_pos', 'userspace_002_pos', 'userspace_encrypted',
	'userspace_send_encrypted', 'userspace_encrypted_13709']
	tags = ['functional', 'userquota']

	[tests/functional/vdev_disk:Linux]
	pre =
	post =
	tests = ['page_alignment']
	tags = ['functional', 'vdev_disk']

	[tests/functional/vdev_zaps]
	tests = ['vdev_zaps_001_pos', 'vdev_zaps_002_pos', 'vdev_zaps_003_pos',
	'vdev_zaps_004_pos', 'vdev_zaps_005_pos', 'vdev_zaps_006_pos',
	'vdev_zaps_007_pos']
	tags = ['functional', 'vdev_zaps']

	[tests/functional/write_dirs]
	tests = ['write_dirs_001_pos', 'write_dirs_002_pos']
	tags = ['functional', 'write_dirs']

	[tests/functional/xattr]
	tests = ['xattr_001_pos', 'xattr_002_neg', 'xattr_003_neg', 'xattr_004_pos',
	'xattr_005_pos', 'xattr_006_pos', 'xattr_007_neg',
	'xattr_011_pos', 'xattr_012_pos', 'xattr_013_pos', 'xattr_compat']
	tags = ['functional', 'xattr']

	[tests/functional/zvol/zvol_ENOSPC]
	tests = ['zvol_ENOSPC_001_pos']
	tags = ['functional', 'zvol', 'zvol_ENOSPC']

	[tests/functional/zvol/zvol_cli]
	tests = ['zvol_cli_001_pos', 'zvol_cli_002_pos', 'zvol_cli_003_neg']
	tags = ['functional', 'zvol', 'zvol_cli']

	[tests/functional/zvol/zvol_misc]
	tests = ['zvol_misc_002_pos', 'zvol_misc_hierarchy', 'zvol_misc_rename_inuse',
	'zvol_misc_snapdev', 'zvol_misc_trim', 'zvol_misc_volmode', 'zvol_misc_zil']
	tags = ['functional', 'zvol', 'zvol_misc']

	[tests/functional/zvol/zvol_stress]
	tests = ['zvol_stress']
	tags = ['functional', 'zvol', 'zvol_stress']

	[tests/functional/zvol/zvol_swap]
	tests = ['zvol_swap_001_pos', 'zvol_swap_002_pos', 'zvol_swap_004_pos']
	tags = ['functional', 'zvol', 'zvol_swap']

	[tests/functional/libzfs]
	tests = ['many_fds', 'libzfs_input']
	tags = ['functional', 'libzfs']

	[tests/functional/log_spacemap]
	tests = ['log_spacemap_import_logs']
	pre =
	post =
	tags = ['functional', 'log_spacemap']

	[tests/functional/l2arc]
	tests = ['l2arc_arcstats_pos', 'l2arc_mfuonly_pos', 'l2arc_l2miss_pos',
	'persist_l2arc_001_pos', 'persist_l2arc_002_pos',
	'persist_l2arc_003_neg', 'persist_l2arc_004_pos', 'persist_l2arc_005_pos']
	tags = ['functional', 'l2arc']

	[tests/functional/zpool_influxdb]
	tests = ['zpool_influxdb']
	tags = ['functional', 'zpool_influxdb']
	diff --git a/sys/contrib/openzfs/tests/test-runner/bin/zts-report.py.in b/sys/contrib/openzfs/tests/test-runner/bin/zts-report.py.in
	index ecc50f487152..5ca130931364 100755
	--- a/sys/contrib/openzfs/tests/test-runner/bin/zts-report.py.in
	+++ b/sys/contrib/openzfs/tests/test-runner/bin/zts-report.py.in
	@@ -1,525 +1,524 @@
	#!/usr/bin/env @PYTHON_SHEBANG@

	#
	# This file and its contents are supplied under the terms of the
	# Common Development and Distribution License ("CDDL"), version 1.0.
	# You may only use this file in accordance with the terms of version
	# 1.0 of the CDDL.
	#
	# A full copy of the text of the CDDL should have accompanied this
	# source. A copy of the CDDL is also available via the Internet at
	# http://www.illumos.org/license/CDDL.
	#

	#
	# Copyright (c) 2017 by Delphix. All rights reserved.
	# Copyright (c) 2018 by Lawrence Livermore National Security, LLC.
	#
	# This script must remain compatible with Python 3.6+.
	#

	import os
	import re
	import sys
	import argparse

	#
	# This script parses the stdout of zfstest, which has this format:
	#
	# Test: /path/to/testa (run as root) [00:00] [PASS]
	# Test: /path/to/testb (run as jkennedy) [00:00] [PASS]
	# Test: /path/to/testc (run as root) [00:00] [FAIL]
	# [...many more results...]
	#
	# Results Summary
	# FAIL 22
	# SKIP 32
	# PASS 1156
	#
	# Running Time: 02:50:31
	# Percent passed: 95.5%
	# Log directory: /var/tmp/test_results/20180615T205926
	#

	#
	# Common generic reasons for a test or test group to be skipped.
	#
	# Some test cases are known to fail in ways which are not harmful or dangerous.
	# In these cases simply mark the test as a known failure until it can be
	# updated and the issue resolved. Note that it's preferable to open a unique
	# issue on the GitHub issue tracker for each test case failure.
	#
	known_reason = 'Known issue'

	#
	# Some tests require that a test user be able to execute the zfs utilities.
	# This may not be possible when testing in-tree due to the default permissions
	# on the user's home directory. When testing this can be resolved by granting
	# group read access.
	#
	# chmod 0750 $HOME
	#
	exec_reason = 'Test user execute permissions required for utilities'

	#
	# Some tests require that the kernel supports renameat2 syscall.
	#
	renameat2_reason = 'Kernel renameat2 support required'

	#
	# Some tests require the O_TMPFILE flag which was first introduced in the
	# 3.11 kernel.
	#
	tmpfile_reason = 'Kernel O_TMPFILE support required'

	#
	# Some tests require the statx(2) system call on Linux which was first
	# introduced in the 4.11 kernel.
	#
	statx_reason = 'Kernel statx(2) system call required on Linux'

	#
	# Some tests require that the lsattr utility support the project id feature.
	#
	project_id_reason = 'lsattr with set/show project ID required'

	#
	# Some tests require that the kernel support user namespaces.
	#
	user_ns_reason = 'Kernel user namespace support required'

	#
	# Some rewind tests can fail since nothing guarantees that old MOS blocks
	# are not overwritten. Snapshots protect datasets and data files but not
	# the MOS. Reasonable efforts are made in the test case to increase the
	# odds that some txgs will have their MOS data left untouched, but it is
	# never a sure thing.
	#
	rewind_reason = 'Arbitrary pool rewind is not guaranteed'

	#
	# Some tests require a minimum version of the fio benchmark utility.
	# Older distributions such as CentOS 6.x only provide fio-2.0.13.
	#
	fio_reason = 'Fio v2.3 or newer required'

	#
	# Some tests require that the DISKS provided support the discard operation.
	# Normally this is not an issue because loop back devices are used for DISKS
	# and they support discard (TRIM/UNMAP).
	#
	trim_reason = 'DISKS must support discard (TRIM/UNMAP)'

	#
	# Some tests on FreeBSD require the fspacectl(2) system call and the
	# truncate(1) utility supporting the -d option. The system call was first
	# introduced in FreeBSD version 1400032.
	#
	fspacectl_reason = 'fspacectl(2) and truncate -d support required'

	#
	# Some tests are not applicable to a platform or need to be updated to operate
	# in the manor required by the platform. Any tests which are skipped for this
	# reason will be suppressed in the final analysis output.
	#
	na_reason = "Not applicable"

	#
	# Some test cases doesn't have all requirements to run on Github actions CI.
	#
	ci_reason = 'CI runner doesn\'t have all requirements'

	#
	# Idmapped mount is only supported in kernel version >= 5.12
	#
	idmap_reason = 'Idmapped mount needs kernel 5.12+'

	#
	# copy_file_range() is not supported by all kernels
	#
	cfr_reason = 'Kernel copy_file_range support required'

	if sys.platform.startswith('freebsd'):
	cfr_cross_reason = 'copy_file_range(2) cross-filesystem needs FreeBSD 14+'
	else:
	cfr_cross_reason = 'copy_file_range(2) cross-filesystem needs kernel 5.3+'

	#
	# These tests are known to fail, thus we use this list to prevent these
	# failures from failing the job as a whole; only unexpected failures
	# bubble up to cause this script to exit with a non-zero exit status.
	#
	# Format: { 'test-name': ['expected result', 'issue-number \| reason'] }
	#
	# For each known failure it is recommended to link to a GitHub issue by
	# setting the reason to the issue number. Alternately, one of the generic
	# reasons listed above can be used.
	#
	known = {
	'casenorm/mixed_none_lookup_ci': ['FAIL', 7633],
	'casenorm/mixed_formd_lookup_ci': ['FAIL', 7633],
	'cli_root/zpool_import/import_rewind_device_replaced':
	['FAIL', rewind_reason],
	'cli_user/misc/zfs_share_001_neg': ['SKIP', na_reason],
	'cli_user/misc/zfs_unshare_001_neg': ['SKIP', na_reason],
	'pool_checkpoint/checkpoint_discard_busy': ['SKIP', 12053],
	'privilege/setup': ['SKIP', na_reason],
	'refreserv/refreserv_004_pos': ['FAIL', known_reason],
	'rootpool/setup': ['SKIP', na_reason],
	'rsend/rsend_008_pos': ['SKIP', 6066],
	'vdev_zaps/vdev_zaps_007_pos': ['FAIL', known_reason],
	}

	if sys.platform.startswith('freebsd'):
	known.update({
	'cli_root/zfs_receive/receive-o-x_props_override':
	['FAIL', known_reason],
	'cli_root/zpool_resilver/zpool_resilver_concurrent':
	['SKIP', na_reason],
	'cli_root/zpool_wait/zpool_wait_trim_basic': ['SKIP', trim_reason],
	'cli_root/zpool_wait/zpool_wait_trim_cancel': ['SKIP', trim_reason],
	'cli_root/zpool_wait/zpool_wait_trim_flag': ['SKIP', trim_reason],
	'cli_root/zfs_unshare/zfs_unshare_008_pos': ['SKIP', na_reason],
	'cp_files/cp_files_002_pos': ['SKIP', na_reason],
	'link_count/link_count_001': ['SKIP', na_reason],
	- 'casenorm/mixed_create_failure': ['FAIL', 13215],
	'mmap/mmap_sync_001_pos': ['SKIP', na_reason],
	'rsend/send_raw_ashift': ['SKIP', 14961],
	})
	elif sys.platform.startswith('linux'):
	known.update({
	'casenorm/mixed_formd_lookup': ['FAIL', 7633],
	'casenorm/mixed_formd_delete': ['FAIL', 7633],
	'casenorm/sensitive_formd_lookup': ['FAIL', 7633],
	'casenorm/sensitive_formd_delete': ['FAIL', 7633],
	'removal/removal_with_zdb': ['SKIP', known_reason],
	'cli_root/zfs_unshare/zfs_unshare_002_pos': ['SKIP', na_reason],
	})


	#
	# These tests may occasionally fail or be skipped. We want there failures
	# to be reported but only unexpected failures should bubble up to cause
	# this script to exit with a non-zero exit status.
	#
	# Format: { 'test-name': ['expected result', 'issue-number \| reason'] }
	#
	# For each known failure it is recommended to link to a GitHub issue by
	# setting the reason to the issue number. Alternately, one of the generic
	# reasons listed above can be used.
	#
	maybe = {
	'append/threadsappend_001_pos': ['FAIL', 6136],
	'chattr/setup': ['SKIP', exec_reason],
	'crtime/crtime_001_pos': ['SKIP', statx_reason],
	'cli_root/zdb/zdb_006_pos': ['FAIL', known_reason],
	'cli_root/zfs_destroy/zfs_destroy_dev_removal_condense':
	['FAIL', known_reason],
	'cli_root/zfs_get/zfs_get_004_pos': ['FAIL', known_reason],
	'cli_root/zfs_get/zfs_get_009_pos': ['SKIP', 5479],
	'cli_root/zfs_rollback/zfs_rollback_001_pos': ['FAIL', known_reason],
	'cli_root/zfs_rollback/zfs_rollback_002_pos': ['FAIL', known_reason],
	'cli_root/zfs_share/zfs_share_concurrent_shares': ['FAIL', known_reason],
	'cli_root/zfs_snapshot/zfs_snapshot_002_neg': ['FAIL', known_reason],
	'cli_root/zfs_unshare/zfs_unshare_006_pos': ['SKIP', na_reason],
	'cli_root/zpool_add/zpool_add_004_pos': ['FAIL', known_reason],
	'cli_root/zpool_destroy/zpool_destroy_001_pos': ['SKIP', 6145],
	'cli_root/zpool_import/zpool_import_missing_003_pos': ['SKIP', 6839],
	'cli_root/zpool_initialize/zpool_initialize_import_export':
	['FAIL', 11948],
	'cli_root/zpool_labelclear/zpool_labelclear_removed':
	['FAIL', known_reason],
	'cli_root/zpool_trim/setup': ['SKIP', trim_reason],
	'cli_root/zpool_upgrade/zpool_upgrade_004_pos': ['FAIL', 6141],
	'delegate/setup': ['SKIP', exec_reason],
	'fallocate/fallocate_punch-hole': ['SKIP', fspacectl_reason],
	'history/history_004_pos': ['FAIL', 7026],
	'history/history_005_neg': ['FAIL', 6680],
	'history/history_006_neg': ['FAIL', 5657],
	'history/history_008_pos': ['FAIL', known_reason],
	'history/history_010_pos': ['SKIP', exec_reason],
	'io/mmap': ['SKIP', fio_reason],
	'largest_pool/largest_pool_001_pos': ['FAIL', known_reason],
	'mmp/mmp_on_uberblocks': ['FAIL', known_reason],
	'pam/setup': ['SKIP', "pamtester might be not available"],
	'pool_checkpoint/checkpoint_discard_busy': ['FAIL', 11946],
	'projectquota/setup': ['SKIP', exec_reason],
	'removal/removal_condense_export': ['FAIL', known_reason],
	'renameat2/setup': ['SKIP', renameat2_reason],
	'reservation/reservation_008_pos': ['FAIL', 7741],
	'reservation/reservation_018_pos': ['FAIL', 5642],
	'snapshot/clone_001_pos': ['FAIL', known_reason],
	'snapshot/snapshot_009_pos': ['FAIL', 7961],
	'snapshot/snapshot_010_pos': ['FAIL', 7961],
	'snapused/snapused_004_pos': ['FAIL', 5513],
	'tmpfile/setup': ['SKIP', tmpfile_reason],
	'trim/setup': ['SKIP', trim_reason],
	'upgrade/upgrade_projectquota_001_pos': ['SKIP', project_id_reason],
	'user_namespace/setup': ['SKIP', user_ns_reason],
	'userquota/setup': ['SKIP', exec_reason],
	'vdev_zaps/vdev_zaps_004_pos': ['FAIL', known_reason],
	'zvol/zvol_ENOSPC/zvol_ENOSPC_001_pos': ['FAIL', 5848],
	}

	if sys.platform.startswith('freebsd'):
	maybe.update({
	'cli_root/zfs_copies/zfs_copies_002_pos': ['FAIL', known_reason],
	'cli_root/zfs_inherit/zfs_inherit_001_neg': ['FAIL', known_reason],
	'cli_root/zpool_import/zpool_import_012_pos': ['FAIL', known_reason],
	'delegate/zfs_allow_003_pos': ['FAIL', known_reason],
	'inheritance/inherit_001_pos': ['FAIL', 11829],
	'pool_checkpoint/checkpoint_big_rewind': ['FAIL', 12622],
	'pool_checkpoint/checkpoint_indirect': ['FAIL', 12623],
	'resilver/resilver_restart_001': ['FAIL', known_reason],
	'snapshot/snapshot_002_pos': ['FAIL', '14831'],
	'bclone/bclone_crossfs_corner_cases': ['SKIP', cfr_cross_reason],
	'bclone/bclone_crossfs_corner_cases_limited':
	['SKIP', cfr_cross_reason],
	'bclone/bclone_crossfs_data': ['SKIP', cfr_cross_reason],
	'bclone/bclone_crossfs_embedded': ['SKIP', cfr_cross_reason],
	'bclone/bclone_crossfs_hole': ['SKIP', cfr_cross_reason],
	'bclone/bclone_diffprops_all': ['SKIP', cfr_cross_reason],
	'bclone/bclone_diffprops_checksum': ['SKIP', cfr_cross_reason],
	'bclone/bclone_diffprops_compress': ['SKIP', cfr_cross_reason],
	'bclone/bclone_diffprops_copies': ['SKIP', cfr_cross_reason],
	'bclone/bclone_diffprops_recordsize': ['SKIP', cfr_cross_reason],
	'bclone/bclone_prop_sync': ['SKIP', cfr_cross_reason],
	'block_cloning/block_cloning_cross_enc_dataset':
	['SKIP', cfr_cross_reason],
	'block_cloning/block_cloning_copyfilerange_cross_dataset':
	['SKIP', cfr_cross_reason]
	})
	elif sys.platform.startswith('linux'):
	maybe.update({
	'bclone/bclone_crossfs_corner_cases': ['SKIP', cfr_cross_reason],
	'bclone/bclone_crossfs_corner_cases_limited':
	['SKIP', cfr_cross_reason],
	'bclone/bclone_crossfs_data': ['SKIP', cfr_cross_reason],
	'bclone/bclone_crossfs_embedded': ['SKIP', cfr_cross_reason],
	'bclone/bclone_crossfs_hole': ['SKIP', cfr_cross_reason],
	'bclone/bclone_diffprops_all': ['SKIP', cfr_cross_reason],
	'bclone/bclone_diffprops_checksum': ['SKIP', cfr_cross_reason],
	'bclone/bclone_diffprops_compress': ['SKIP', cfr_cross_reason],
	'bclone/bclone_diffprops_copies': ['SKIP', cfr_cross_reason],
	'bclone/bclone_diffprops_recordsize': ['SKIP', cfr_cross_reason],
	'bclone/bclone_prop_sync': ['SKIP', cfr_cross_reason],
	'bclone/bclone_samefs_corner_cases': ['SKIP', cfr_reason],
	'bclone/bclone_samefs_corner_cases_limited': ['SKIP', cfr_reason],
	'bclone/bclone_samefs_data': ['SKIP', cfr_reason],
	'bclone/bclone_samefs_embedded': ['SKIP', cfr_reason],
	'bclone/bclone_samefs_hole': ['SKIP', cfr_reason],
	'block_cloning/block_cloning_clone_mmap_cached': ['SKIP', cfr_reason],
	'block_cloning/block_cloning_clone_mmap_write':
	['SKIP', cfr_reason],
	'block_cloning/block_cloning_copyfilerange':
	['SKIP', cfr_reason],
	'block_cloning/block_cloning_copyfilerange_cross_dataset':
	['SKIP', cfr_cross_reason],
	'block_cloning/block_cloning_copyfilerange_fallback':
	['SKIP', cfr_reason],
	'block_cloning/block_cloning_copyfilerange_fallback_same_txg':
	['SKIP', cfr_cross_reason],
	'block_cloning/block_cloning_copyfilerange_partial':
	['SKIP', cfr_reason],
	'block_cloning/block_cloning_cross_enc_dataset':
	['SKIP', cfr_cross_reason],
	'block_cloning/block_cloning_disabled_copyfilerange':
	['SKIP', cfr_reason],
	'block_cloning/block_cloning_lwb_buffer_overflow':
	['SKIP', cfr_reason],
	'block_cloning/block_cloning_replay':
	['SKIP', cfr_reason],
	'block_cloning/block_cloning_replay_encrypted':
	['SKIP', cfr_reason],
	'cli_root/zfs_rename/zfs_rename_002_pos': ['FAIL', known_reason],
	'cli_root/zpool_reopen/zpool_reopen_003_pos': ['FAIL', known_reason],
	'cp_files/cp_files_002_pos': ['SKIP', cfr_reason],
	'fault/auto_online_002_pos': ['FAIL', 11889],
	'fault/auto_replace_001_pos': ['FAIL', 14851],
	'fault/auto_spare_002_pos': ['FAIL', 11889],
	'fault/auto_spare_multiple': ['FAIL', 11889],
	'fault/auto_spare_shared': ['FAIL', 11889],
	'fault/decompress_fault': ['FAIL', 11889],
	'idmap_mount/idmap_mount_001': ['SKIP', idmap_reason],
	'idmap_mount/idmap_mount_002': ['SKIP', idmap_reason],
	'idmap_mount/idmap_mount_003': ['SKIP', idmap_reason],
	'idmap_mount/idmap_mount_004': ['SKIP', idmap_reason],
	'idmap_mount/idmap_mount_005': ['SKIP', idmap_reason],
	'io/io_uring': ['SKIP', 'io_uring support required'],
	'limits/filesystem_limit': ['SKIP', known_reason],
	'limits/snapshot_limit': ['SKIP', known_reason],
	'mmp/mmp_active_import': ['FAIL', known_reason],
	'mmp/mmp_exported_import': ['FAIL', known_reason],
	'mmp/mmp_inactive_import': ['FAIL', known_reason],
	'zvol/zvol_misc/zvol_misc_fua': ['SKIP', 14872],
	'zvol/zvol_misc/zvol_misc_snapdev': ['FAIL', 12621],
	'zvol/zvol_misc/zvol_misc_trim': ['SKIP', 14872],
	'zvol/zvol_misc/zvol_misc_volmode': ['FAIL', known_reason],
	})

	# Not all Github actions runners have scsi_debug module, so we may skip
	# some tests which use it.
	if os.environ.get('CI') == 'true':
	known.update({
	'cli_root/zpool_expand/zpool_expand_001_pos': ['SKIP', ci_reason],
	'cli_root/zpool_expand/zpool_expand_003_neg': ['SKIP', ci_reason],
	'cli_root/zpool_expand/zpool_expand_005_pos': ['SKIP', ci_reason],
	'cli_root/zpool_reopen/setup': ['SKIP', ci_reason],
	'cli_root/zpool_reopen/zpool_reopen_001_pos': ['SKIP', ci_reason],
	'cli_root/zpool_reopen/zpool_reopen_002_pos': ['SKIP', ci_reason],
	'cli_root/zpool_reopen/zpool_reopen_003_pos': ['SKIP', ci_reason],
	'cli_root/zpool_reopen/zpool_reopen_004_pos': ['SKIP', ci_reason],
	'cli_root/zpool_reopen/zpool_reopen_005_pos': ['SKIP', ci_reason],
	'cli_root/zpool_reopen/zpool_reopen_006_neg': ['SKIP', ci_reason],
	'cli_root/zpool_reopen/zpool_reopen_007_pos': ['SKIP', ci_reason],
	'cli_root/zpool_split/zpool_split_wholedisk': ['SKIP', ci_reason],
	'fault/auto_offline_001_pos': ['SKIP', ci_reason],
	'fault/auto_online_001_pos': ['SKIP', ci_reason],
	'fault/auto_online_002_pos': ['SKIP', ci_reason],
	'fault/auto_replace_001_pos': ['SKIP', ci_reason],
	'fault/auto_replace_002_pos': ['SKIP', ci_reason],
	'fault/auto_spare_ashift': ['SKIP', ci_reason],
	'fault/auto_spare_shared': ['SKIP', ci_reason],
	'procfs/pool_state': ['SKIP', ci_reason],
	})

	maybe.update({
	'events/events_002_pos': ['FAIL', 11546],
	})


	def process_results(pathname):
	try:
	f = open(pathname)
	except IOError as e:
	print('Error opening file:', e)
	sys.exit(1)

	prefix = '/zfs-tests/tests/(?:functional\|perf/regression)/'
	pattern = \
	r'^Test(?:\s+$\S+$)?:' + \
	rf'\s\S{prefix}(\S+)' + \
	r'\s$run as (\S+)$\s\[(\S+)\]\s*\[(\S+)\]'
	pattern_log = r'^\sLog directory:\s(\S*)'

	d = {}
	logdir = 'Could not determine log directory.'
	for line in f.readlines():
	m = re.match(pattern, line)
	if m and len(m.groups()) == 4:
	d[m.group(1)] = m.group(4)
	continue

	m = re.match(pattern_log, line)
	if m:
	logdir = m.group(1)

	return d, logdir


	class ListMaybesAction(argparse.Action):
	def __init__(self,
	option_strings,
	dest="SUPPRESS",
	default="SUPPRESS",
	help="list flaky tests and exit"):
	super(ListMaybesAction, self).__init__(
	option_strings=option_strings,
	dest=dest,
	default=default,
	nargs=0,
	help=help)

	def __call__(self, parser, namespace, values, option_string=None):
	for test in maybe:
	print(test)
	sys.exit(0)


	if __name__ == "__main__":
	parser = argparse.ArgumentParser(description='Analyze ZTS logs')
	parser.add_argument('logfile')
	parser.add_argument('--list-maybes', action=ListMaybesAction)
	parser.add_argument('--no-maybes', action='store_false', dest='maybes')
	args = parser.parse_args()

	results, logdir = process_results(args.logfile)

	if not results:
	print("\n\nNo test results were found.")
	print("Log directory:", logdir)
	sys.exit(0)

	expected = []
	unexpected = []
	all_maybes = True

	for test in list(results.keys()):
	if results[test] == "PASS":
	continue

	setup = test.replace(os.path.basename(test), "setup")
	if results[test] == "SKIP" and test != setup:
	if setup in known and known[setup][0] == "SKIP":
	continue
	if setup in maybe and maybe[setup][0] == "SKIP":
	continue

	if (test in known and results[test] in known[test][0]):
	expected.append(test)
	elif test in maybe and results[test] in maybe[test][0]:
	if results[test] == 'SKIP' or args.maybes:
	expected.append(test)
	elif not args.maybes:
	unexpected.append(test)
	else:
	unexpected.append(test)
	all_maybes = False

	print("\nTests with results other than PASS that are expected:")
	for test in sorted(expected):
	issue_url = 'https://github.com/openzfs/zfs/issues/'

	# Include the reason why the result is expected, given the following:
	# 1. Suppress test results which set the "Not applicable" reason.
	# 2. Numerical reasons are assumed to be GitHub issue numbers.
	# 3. When an entire test group is skipped only report the setup reason.
	if test in known:
	if known[test][1] == na_reason:
	continue
	elif isinstance(known[test][1], int):
	expect = f"{issue_url}{known[test][1]}"
	else:
	expect = known[test][1]
	elif test in maybe:
	if isinstance(maybe[test][1], int):
	expect = f"{issue_url}{maybe[test][1]}"
	else:
	expect = maybe[test][1]
	elif setup in known and known[setup][0] == "SKIP" and setup != test:
	continue
	elif setup in maybe and maybe[setup][0] == "SKIP" and setup != test:
	continue
	else:
	expect = "UNKNOWN REASON"
	print(f" {results[test]} {test} ({expect})")

	print("\nTests with result of PASS that are unexpected:")
	for test in sorted(known.keys()):
	# We probably should not be silently ignoring the case
	# where "test" is not in "results".
	if test not in results or results[test] != "PASS":
	continue
	print(f" {results[test]} {test} (expected {known[test][0]})")

	print("\nTests with results other than PASS that are unexpected:")
	for test in sorted(unexpected):
	expect = "PASS" if test not in known else known[test][0]
	print(f" {results[test]} {test} (expected {expect})")

	if len(unexpected) == 0:
	sys.exit(0)
	elif not args.maybes and all_maybes:
	sys.exit(2)
	else:
	sys.exit(1)
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/cmd/libzfs_input_check.c b/sys/contrib/openzfs/tests/zfs-tests/cmd/libzfs_input_check.c
	index c661718a296c..7d9ce4fada1b 100644
	--- a/sys/contrib/openzfs/tests/zfs-tests/cmd/libzfs_input_check.c
	+++ b/sys/contrib/openzfs/tests/zfs-tests/cmd/libzfs_input_check.c
	@@ -1,1082 +1,1084 @@
	/*
	* CDDL HEADER START
	*
	* This file and its contents are supplied under the terms of the
	* Common Development and Distribution License ("CDDL"), version 1.0.
	* You may only use this file in accordance with the terms of version
	* 1.0 of the CDDL.
	*
	* A full copy of the text of the CDDL should have accompanied this
	* source. A copy of the CDDL is also available via the Internet at
	* http://www.illumos.org/license/CDDL.
	*
	* CDDL HEADER END
	*/

	/*
	* Copyright (c) 2018 by Delphix. All rights reserved.
	*/

	#include <stdio.h>
	#include <stdlib.h>
	#include <string.h>
	#include <libzfs_core.h>
	#include <libzutil.h>

	#include <sys/nvpair.h>
	#include <sys/vdev_impl.h>
	#include <sys/zfs_ioctl.h>
	#include <sys/zfs_bootenv.h>
	#include <sys/fs/zfs.h>

	/*
	* Test the nvpair inputs for the non-legacy zfs ioctl commands.
	*/

	static boolean_t unexpected_failures;
	static int zfs_fd;
	static const char *active_test;

	/*
	* Tracks which zfs_ioc_t commands were tested
	*/
	static boolean_t ioc_tested[ZFS_IOC_LAST - ZFS_IOC_FIRST];

	/*
	* Legacy ioctls that are skipped (for now)
	*/
	static const zfs_ioc_t ioc_skip[] = {
	ZFS_IOC_POOL_CREATE,
	ZFS_IOC_POOL_DESTROY,
	ZFS_IOC_POOL_IMPORT,
	ZFS_IOC_POOL_EXPORT,
	ZFS_IOC_POOL_CONFIGS,
	ZFS_IOC_POOL_STATS,
	ZFS_IOC_POOL_TRYIMPORT,
	ZFS_IOC_POOL_SCAN,
	ZFS_IOC_POOL_FREEZE,
	ZFS_IOC_POOL_UPGRADE,
	ZFS_IOC_POOL_GET_HISTORY,

	ZFS_IOC_VDEV_ADD,
	ZFS_IOC_VDEV_REMOVE,
	ZFS_IOC_VDEV_SET_STATE,
	ZFS_IOC_VDEV_ATTACH,
	ZFS_IOC_VDEV_DETACH,
	ZFS_IOC_VDEV_SETPATH,
	ZFS_IOC_VDEV_SETFRU,

	ZFS_IOC_OBJSET_STATS,
	ZFS_IOC_OBJSET_ZPLPROPS,
	ZFS_IOC_DATASET_LIST_NEXT,
	ZFS_IOC_SNAPSHOT_LIST_NEXT,
	ZFS_IOC_SET_PROP,
	ZFS_IOC_DESTROY,
	ZFS_IOC_RENAME,
	ZFS_IOC_RECV,
	ZFS_IOC_SEND,
	ZFS_IOC_INJECT_FAULT,
	ZFS_IOC_CLEAR_FAULT,
	ZFS_IOC_INJECT_LIST_NEXT,
	ZFS_IOC_ERROR_LOG,
	ZFS_IOC_CLEAR,
	ZFS_IOC_PROMOTE,
	ZFS_IOC_DSOBJ_TO_DSNAME,
	ZFS_IOC_OBJ_TO_PATH,
	ZFS_IOC_POOL_SET_PROPS,
	ZFS_IOC_POOL_GET_PROPS,
	ZFS_IOC_SET_FSACL,
	ZFS_IOC_GET_FSACL,
	ZFS_IOC_SHARE,
	ZFS_IOC_INHERIT_PROP,
	ZFS_IOC_SMB_ACL,
	ZFS_IOC_USERSPACE_ONE,
	ZFS_IOC_USERSPACE_MANY,
	ZFS_IOC_USERSPACE_UPGRADE,
	ZFS_IOC_OBJSET_RECVD_PROPS,
	ZFS_IOC_VDEV_SPLIT,
	ZFS_IOC_NEXT_OBJ,
	ZFS_IOC_DIFF,
	ZFS_IOC_TMP_SNAPSHOT,
	ZFS_IOC_OBJ_TO_STATS,
	ZFS_IOC_SPACE_WRITTEN,
	ZFS_IOC_POOL_REGUID,
	ZFS_IOC_SEND_PROGRESS,
	ZFS_IOC_EVENTS_NEXT,
	ZFS_IOC_EVENTS_CLEAR,
	ZFS_IOC_EVENTS_SEEK,
	ZFS_IOC_NEXTBOOT,
	ZFS_IOC_JAIL,
	ZFS_IOC_UNJAIL,
	};


	#define IOC_INPUT_TEST(ioc, name, req, opt, err) \
	IOC_INPUT_TEST_IMPL(ioc, name, req, opt, err, B_FALSE)

	#define IOC_INPUT_TEST_WILD(ioc, name, req, opt, err) \
	IOC_INPUT_TEST_IMPL(ioc, name, req, opt, err, B_TRUE)

	#define IOC_INPUT_TEST_IMPL(ioc, name, req, opt, err, wild) \
	do { \
	active_test = __func__ + 5; \
	ioc_tested[ioc - ZFS_IOC_FIRST] = B_TRUE; \
	lzc_ioctl_test(ioc, name, req, opt, err, wild); \
	} while (0)

	/*
	* run a zfs ioctl command, verify expected results and log failures
	*/
	static void
	lzc_ioctl_run(zfs_ioc_t ioc, const char name, nvlist_t innvl, int expected)
	{
	zfs_cmd_t zc = {"\0"};
	char *packed = NULL;
	const char *variant;
	size_t size = 0;
	int error = 0;

	switch (expected) {
	case ZFS_ERR_IOC_ARG_UNAVAIL:
	variant = "unsupported input";
	break;
	case ZFS_ERR_IOC_ARG_REQUIRED:
	variant = "missing input";
	break;
	case ZFS_ERR_IOC_ARG_BADTYPE:
	variant = "invalid input type";
	break;
	default:
	variant = "valid input";
	break;
	}

	packed = fnvlist_pack(innvl, &size);
	(void) strlcpy(zc.zc_name, name, sizeof (zc.zc_name));
	zc.zc_name[sizeof (zc.zc_name) - 1] = '\0';
	zc.zc_nvlist_src = (uint64_t)(uintptr_t)packed;
	zc.zc_nvlist_src_size = size;
	zc.zc_nvlist_dst_size = MAX(size * 2, 128 * 1024);
	zc.zc_nvlist_dst = (uint64_t)(uintptr_t)malloc(zc.zc_nvlist_dst_size);

	if (lzc_ioctl_fd(zfs_fd, ioc, &zc) != 0)
	error = errno;

	if (error != expected) {
	unexpected_failures = B_TRUE;
	(void) fprintf(stderr, "%s: Unexpected result with %s, "
	"error %d (expecting %d)\n",
	active_test, variant, error, expected);
	}

	fnvlist_pack_free(packed, size);
	free((void *)(uintptr_t)zc.zc_nvlist_dst);
	}

	/*
	* Test each ioc for the following ioctl input errors:
	* ZFS_ERR_IOC_ARG_UNAVAIL an input argument is not supported by kernel
	* ZFS_ERR_IOC_ARG_REQUIRED a required input argument is missing
	* ZFS_ERR_IOC_ARG_BADTYPE an input argument has an invalid type
	*/
	static int
	lzc_ioctl_test(zfs_ioc_t ioc, const char name, nvlist_t required,
	nvlist_t *optional, int expected_error, boolean_t wildcard)
	{
	nvlist_t *input = fnvlist_alloc();
	nvlist_t *future = fnvlist_alloc();
	int error = 0;

	if (required != NULL) {
	for (nvpair_t *pair = nvlist_next_nvpair(required, NULL);
	pair != NULL; pair = nvlist_next_nvpair(required, pair)) {
	fnvlist_add_nvpair(input, pair);
	}
	}
	if (optional != NULL) {
	for (nvpair_t *pair = nvlist_next_nvpair(optional, NULL);
	pair != NULL; pair = nvlist_next_nvpair(optional, pair)) {
	fnvlist_add_nvpair(input, pair);
	}
	}

	/*
	* Generic input run with 'optional' nvlist pair
	*/
	if (!wildcard)
	fnvlist_add_nvlist(input, "optional", future);
	lzc_ioctl_run(ioc, name, input, expected_error);
	if (!wildcard)
	fnvlist_remove(input, "optional");

	/*
	* Bogus input value
	*/
	if (!wildcard) {
	fnvlist_add_string(input, "bogus_input", "bogus");
	lzc_ioctl_run(ioc, name, input, ZFS_ERR_IOC_ARG_UNAVAIL);
	fnvlist_remove(input, "bogus_input");
	}

	/*
	* Missing required inputs
	*/
	if (required != NULL) {
	nvlist_t *empty = fnvlist_alloc();
	lzc_ioctl_run(ioc, name, empty, ZFS_ERR_IOC_ARG_REQUIRED);
	nvlist_free(empty);
	}

	/*
	* Wrong nvpair type
	*/
	if (required != NULL \|\| optional != NULL) {
	/*
	* switch the type of one of the input pairs
	*/
	for (nvpair_t *pair = nvlist_next_nvpair(input, NULL);
	pair != NULL; pair = nvlist_next_nvpair(input, pair)) {
	char pname[MAXNAMELEN];
	data_type_t ptype;

	strlcpy(pname, nvpair_name(pair), sizeof (pname));
	pname[sizeof (pname) - 1] = '\0';
	ptype = nvpair_type(pair);
	fnvlist_remove_nvpair(input, pair);

	switch (ptype) {
	case DATA_TYPE_STRING:
	fnvlist_add_uint64(input, pname, 42);
	break;
	default:
	fnvlist_add_string(input, pname, "bogus");
	break;
	}
	}
	lzc_ioctl_run(ioc, name, input, ZFS_ERR_IOC_ARG_BADTYPE);
	}

	nvlist_free(future);
	nvlist_free(input);

	return (error);
	}

	static void
	test_pool_sync(const char *pool)
	{
	nvlist_t *required = fnvlist_alloc();

	fnvlist_add_boolean_value(required, "force", B_TRUE);

	IOC_INPUT_TEST(ZFS_IOC_POOL_SYNC, pool, required, NULL, 0);

	nvlist_free(required);
	}

	static void
	test_pool_reopen(const char *pool)
	{
	nvlist_t *optional = fnvlist_alloc();

	fnvlist_add_boolean_value(optional, "scrub_restart", B_FALSE);

	IOC_INPUT_TEST(ZFS_IOC_POOL_REOPEN, pool, NULL, optional, 0);

	nvlist_free(optional);
	}

	static void
	test_pool_checkpoint(const char *pool)
	{
	IOC_INPUT_TEST(ZFS_IOC_POOL_CHECKPOINT, pool, NULL, NULL, 0);
	}

	static void
	test_pool_discard_checkpoint(const char *pool)
	{
	int err = lzc_pool_checkpoint(pool);
	if (err == 0 \|\| err == ZFS_ERR_CHECKPOINT_EXISTS)
	IOC_INPUT_TEST(ZFS_IOC_POOL_DISCARD_CHECKPOINT, pool, NULL,
	NULL, 0);
	}

	static void
	test_log_history(const char *pool)
	{
	nvlist_t *required = fnvlist_alloc();

	fnvlist_add_string(required, "message", "input check");

	IOC_INPUT_TEST(ZFS_IOC_LOG_HISTORY, pool, required, NULL, 0);

	nvlist_free(required);
	}

	static void
	test_create(const char *pool)
	{
	char dataset[MAXNAMELEN + 32];

	(void) snprintf(dataset, sizeof (dataset), "%s/create-fs", pool);

	nvlist_t *required = fnvlist_alloc();
	nvlist_t *optional = fnvlist_alloc();
	nvlist_t *props = fnvlist_alloc();

	fnvlist_add_int32(required, "type", DMU_OST_ZFS);
	fnvlist_add_uint64(props, "recordsize", 8192);
	fnvlist_add_nvlist(optional, "props", props);

	IOC_INPUT_TEST(ZFS_IOC_CREATE, dataset, required, optional, 0);

	nvlist_free(required);
	nvlist_free(optional);
	}

	static void
	test_snapshot(const char pool, const char snapshot)
	{
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *optional = fnvlist_alloc();
	nvlist_t *snaps = fnvlist_alloc();
	nvlist_t *props = fnvlist_alloc();

	fnvlist_add_boolean(snaps, snapshot);
	fnvlist_add_nvlist(required, "snaps", snaps);

	fnvlist_add_string(props, "org.openzfs:launch", "September 17th, 2013");
	fnvlist_add_nvlist(optional, "props", props);

	IOC_INPUT_TEST(ZFS_IOC_SNAPSHOT, pool, required, optional, 0);

	nvlist_free(props);
	nvlist_free(snaps);
	nvlist_free(optional);
	nvlist_free(required);
	}

	static void
	test_space_snaps(const char *snapshot)
	{
	nvlist_t *required = fnvlist_alloc();
	fnvlist_add_string(required, "firstsnap", snapshot);

	IOC_INPUT_TEST(ZFS_IOC_SPACE_SNAPS, snapshot, required, NULL, 0);

	nvlist_free(required);
	}

	static void
	test_destroy_snaps(const char pool, const char snapshot)
	{
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *snaps = fnvlist_alloc();

	fnvlist_add_boolean(snaps, snapshot);
	fnvlist_add_nvlist(required, "snaps", snaps);

	IOC_INPUT_TEST(ZFS_IOC_DESTROY_SNAPS, pool, required, NULL, 0);

	nvlist_free(snaps);
	nvlist_free(required);
	}


	static void
	test_bookmark(const char pool, const char snapshot, const char *bookmark)
	{
	nvlist_t *required = fnvlist_alloc();

	fnvlist_add_string(required, bookmark, snapshot);

	IOC_INPUT_TEST_WILD(ZFS_IOC_BOOKMARK, pool, required, NULL, 0);

	nvlist_free(required);
	}

	static void
	test_get_bookmarks(const char *dataset)
	{
	nvlist_t *optional = fnvlist_alloc();

	fnvlist_add_boolean(optional, "guid");
	fnvlist_add_boolean(optional, "createtxg");
	fnvlist_add_boolean(optional, "creation");

	IOC_INPUT_TEST_WILD(ZFS_IOC_GET_BOOKMARKS, dataset, NULL, optional, 0);

	nvlist_free(optional);
	}

	static void
	test_destroy_bookmarks(const char pool, const char bookmark)
	{
	nvlist_t *required = fnvlist_alloc();

	fnvlist_add_boolean(required, bookmark);

	IOC_INPUT_TEST_WILD(ZFS_IOC_DESTROY_BOOKMARKS, pool, required, NULL, 0);

	nvlist_free(required);
	}

	static void
	test_clone(const char snapshot, const char clone)
	{
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *optional = fnvlist_alloc();
	nvlist_t *props = fnvlist_alloc();

	fnvlist_add_string(required, "origin", snapshot);

	IOC_INPUT_TEST(ZFS_IOC_CLONE, clone, required, NULL, 0);

	nvlist_free(props);
	nvlist_free(optional);
	nvlist_free(required);
	}

	static void
	test_rollback(const char dataset, const char snapshot)
	{
	nvlist_t *optional = fnvlist_alloc();

	fnvlist_add_string(optional, "target", snapshot);

	IOC_INPUT_TEST(ZFS_IOC_ROLLBACK, dataset, NULL, optional, B_FALSE);

	nvlist_free(optional);
	}

	static void
	test_hold(const char pool, const char snapshot)
	{
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *optional = fnvlist_alloc();
	nvlist_t *holds = fnvlist_alloc();

	fnvlist_add_string(holds, snapshot, "libzfs_check_hold");
	fnvlist_add_nvlist(required, "holds", holds);
	fnvlist_add_int32(optional, "cleanup_fd", zfs_fd);

	IOC_INPUT_TEST(ZFS_IOC_HOLD, pool, required, optional, 0);

	nvlist_free(holds);
	nvlist_free(optional);
	nvlist_free(required);
	}

	static void
	test_get_holds(const char *snapshot)
	{
	IOC_INPUT_TEST(ZFS_IOC_GET_HOLDS, snapshot, NULL, NULL, 0);
	}

	static void
	test_release(const char pool, const char snapshot)
	{
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *release = fnvlist_alloc();

	fnvlist_add_boolean(release, "libzfs_check_hold");
	fnvlist_add_nvlist(required, snapshot, release);

	IOC_INPUT_TEST_WILD(ZFS_IOC_RELEASE, pool, required, NULL, 0);

	nvlist_free(release);
	nvlist_free(required);
	}


	static void
	test_send_new(const char *snapshot, int fd)
	{
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *optional = fnvlist_alloc();

	fnvlist_add_int32(required, "fd", fd);

	fnvlist_add_boolean(optional, "largeblockok");
	fnvlist_add_boolean(optional, "embedok");
	fnvlist_add_boolean(optional, "compressok");
	fnvlist_add_boolean(optional, "rawok");

	/*
	* TODO - Resumable send is harder to set up. So we currently
	* ignore testing for that variant.
	*/
	#if 0
	fnvlist_add_string(optional, "fromsnap", from);
	fnvlist_add_uint64(optional, "resume_object", resumeobj);
	fnvlist_add_uint64(optional, "resume_offset", offset);
	fnvlist_add_boolean(optional, "savedok");
	#endif
	IOC_INPUT_TEST(ZFS_IOC_SEND_NEW, snapshot, required, optional, 0);

	nvlist_free(optional);
	nvlist_free(required);
	}

	static void
	test_recv_new(const char *dataset, int fd)
	{
	- dmu_replay_record_t drr = { 0 };
	+ dmu_replay_record_t drr;
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *optional = fnvlist_alloc();
	nvlist_t *props = fnvlist_alloc();
	char snapshot[MAXNAMELEN + 32];
	ssize_t count;

	+ memset(&drr, 0, sizeof (dmu_replay_record_t));
	+
	int cleanup_fd = open(ZFS_DEV, O_RDWR);
	if (cleanup_fd == -1) {
	(void) fprintf(stderr, "open(%s) failed: %s\n", ZFS_DEV,
	strerror(errno));
	exit(EXIT_FAILURE);
	}
	(void) snprintf(snapshot, sizeof (snapshot), "%s@replicant", dataset);

	count = pread(fd, &drr, sizeof (drr), 0);
	if (count != sizeof (drr)) {
	(void) fprintf(stderr, "could not read stream: %s\n",
	strerror(errno));
	}

	fnvlist_add_string(required, "snapname", snapshot);
	fnvlist_add_byte_array(required, "begin_record", (uchar_t *)&drr,
	sizeof (drr));
	fnvlist_add_int32(required, "input_fd", fd);

	fnvlist_add_string(props, "org.openzfs:launch", "September 17th, 2013");
	fnvlist_add_nvlist(optional, "localprops", props);
	fnvlist_add_boolean(optional, "force");
	fnvlist_add_boolean(optional, "heal");
	fnvlist_add_int32(optional, "cleanup_fd", cleanup_fd);

	/*
	* TODO - Resumable receive is harder to set up. So we currently
	* ignore testing for one.
	*/
	#if 0
	fnvlist_add_nvlist(optional, "props", recvdprops);
	fnvlist_add_string(optional, "origin", origin);
	fnvlist_add_boolean(optional, "resumable");
	fnvlist_add_uint64(optional, "action_handle", *action_handle);
	#endif
	IOC_INPUT_TEST(ZFS_IOC_RECV_NEW, dataset, required, optional,
	ENOTSUP);

	nvlist_free(props);
	nvlist_free(optional);
	nvlist_free(required);

	(void) close(cleanup_fd);
	}

	static void
	test_send_space(const char snapshot1, const char snapshot2)
	{
	nvlist_t *optional = fnvlist_alloc();

	fnvlist_add_string(optional, "from", snapshot1);
	fnvlist_add_boolean(optional, "largeblockok");
	fnvlist_add_boolean(optional, "embedok");
	fnvlist_add_boolean(optional, "compressok");
	fnvlist_add_boolean(optional, "rawok");

	IOC_INPUT_TEST(ZFS_IOC_SEND_SPACE, snapshot2, NULL, optional, 0);

	nvlist_free(optional);
	}

	static void
	test_remap(const char *dataset)
	{
	IOC_INPUT_TEST(ZFS_IOC_REMAP, dataset, NULL, NULL, 0);
	}

	static void
	test_channel_program(const char *pool)
	{
	const char *program =
	"arg = ...\n"
	"argv = arg[\"argv\"]\n"
	"return argv[1]";
	const char *const argv[1] = { "Hello World!" };
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *optional = fnvlist_alloc();
	nvlist_t *args = fnvlist_alloc();

	fnvlist_add_string(required, "program", program);
	fnvlist_add_string_array(args, "argv", argv, 1);
	fnvlist_add_nvlist(required, "arg", args);

	fnvlist_add_boolean_value(optional, "sync", B_TRUE);
	fnvlist_add_uint64(optional, "instrlimit", 1000 * 1000);
	fnvlist_add_uint64(optional, "memlimit", 8192 * 1024);

	IOC_INPUT_TEST(ZFS_IOC_CHANNEL_PROGRAM, pool, required, optional, 0);

	nvlist_free(args);
	nvlist_free(optional);
	nvlist_free(required);
	}

	#define WRAPPING_KEY_LEN 32

	static void
	test_load_key(const char *dataset)
	{
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *optional = fnvlist_alloc();
	nvlist_t *hidden = fnvlist_alloc();
	uint8_t keydata[WRAPPING_KEY_LEN] = {0};

	fnvlist_add_uint8_array(hidden, "wkeydata", keydata, sizeof (keydata));
	fnvlist_add_nvlist(required, "hidden_args", hidden);
	fnvlist_add_boolean(optional, "noop");

	IOC_INPUT_TEST(ZFS_IOC_LOAD_KEY, dataset, required, optional, EINVAL);
	nvlist_free(hidden);
	nvlist_free(optional);
	nvlist_free(required);
	}

	static void
	test_change_key(const char *dataset)
	{
	IOC_INPUT_TEST(ZFS_IOC_CHANGE_KEY, dataset, NULL, NULL, EINVAL);
	}

	static void
	test_unload_key(const char *dataset)
	{
	IOC_INPUT_TEST(ZFS_IOC_UNLOAD_KEY, dataset, NULL, NULL, EACCES);
	}

	static void
	test_vdev_initialize(const char *pool)
	{
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *vdev_guids = fnvlist_alloc();

	fnvlist_add_uint64(vdev_guids, "path", 0xdeadbeefdeadbeef);
	fnvlist_add_uint64(required, ZPOOL_INITIALIZE_COMMAND,
	POOL_INITIALIZE_START);
	fnvlist_add_nvlist(required, ZPOOL_INITIALIZE_VDEVS, vdev_guids);

	IOC_INPUT_TEST(ZFS_IOC_POOL_INITIALIZE, pool, required, NULL, EINVAL);
	nvlist_free(vdev_guids);
	nvlist_free(required);
	}

	static void
	test_vdev_trim(const char *pool)
	{
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *optional = fnvlist_alloc();
	nvlist_t *vdev_guids = fnvlist_alloc();

	fnvlist_add_uint64(vdev_guids, "path", 0xdeadbeefdeadbeef);
	fnvlist_add_uint64(required, ZPOOL_TRIM_COMMAND, POOL_TRIM_START);
	fnvlist_add_nvlist(required, ZPOOL_TRIM_VDEVS, vdev_guids);
	fnvlist_add_uint64(optional, ZPOOL_TRIM_RATE, 1ULL << 30);
	fnvlist_add_boolean_value(optional, ZPOOL_TRIM_SECURE, B_TRUE);

	IOC_INPUT_TEST(ZFS_IOC_POOL_TRIM, pool, required, optional, EINVAL);
	nvlist_free(vdev_guids);
	nvlist_free(optional);
	nvlist_free(required);
	}

	/* Test with invalid values */
	static void
	test_scrub(const char *pool)
	{
	nvlist_t *required = fnvlist_alloc();
	fnvlist_add_uint64(required, "scan_type", POOL_SCAN_FUNCS + 1);
	fnvlist_add_uint64(required, "scan_command", POOL_SCRUB_FLAGS_END + 1);
	IOC_INPUT_TEST(ZFS_IOC_POOL_SCRUB, pool, required, NULL, EINVAL);
	nvlist_free(required);
	}

	static int
	zfs_destroy(const char *dataset)
	{
	zfs_cmd_t zc = {"\0"};
	int err;

	(void) strlcpy(zc.zc_name, dataset, sizeof (zc.zc_name));
	zc.zc_name[sizeof (zc.zc_name) - 1] = '\0';
	err = lzc_ioctl_fd(zfs_fd, ZFS_IOC_DESTROY, &zc);

	return (err == 0 ? 0 : errno);
	}

	static void
	test_redact(const char snapshot1, const char snapshot2)
	{
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *snapnv = fnvlist_alloc();
	char bookmark[MAXNAMELEN + 32];

	fnvlist_add_string(required, "bookname", "testbookmark");
	fnvlist_add_boolean(snapnv, snapshot2);
	fnvlist_add_nvlist(required, "snapnv", snapnv);

	IOC_INPUT_TEST(ZFS_IOC_REDACT, snapshot1, required, NULL, 0);

	nvlist_free(snapnv);
	nvlist_free(required);

	strlcpy(bookmark, snapshot1, sizeof (bookmark));
	*strchr(bookmark, '@') = '\0';
	strlcat(bookmark, "#testbookmark", sizeof (bookmark) -
	strlen(bookmark));
	zfs_destroy(bookmark);
	}

	static void
	test_get_bookmark_props(const char *bookmark)
	{
	IOC_INPUT_TEST(ZFS_IOC_GET_BOOKMARK_PROPS, bookmark, NULL, NULL, 0);
	}

	static void
	test_wait(const char *pool)
	{
	nvlist_t *required = fnvlist_alloc();
	nvlist_t *optional = fnvlist_alloc();

	fnvlist_add_int32(required, "wait_activity", 2);
	fnvlist_add_uint64(optional, "wait_tag", 0xdeadbeefdeadbeef);

	IOC_INPUT_TEST(ZFS_IOC_WAIT, pool, required, optional, EINVAL);

	nvlist_free(required);
	nvlist_free(optional);
	}

	static void
	test_wait_fs(const char *dataset)
	{
	nvlist_t *required = fnvlist_alloc();

	fnvlist_add_int32(required, "wait_activity", 2);

	IOC_INPUT_TEST(ZFS_IOC_WAIT_FS, dataset, required, NULL, EINVAL);

	nvlist_free(required);
	}

	static void
	test_get_bootenv(const char *pool)
	{
	IOC_INPUT_TEST(ZFS_IOC_GET_BOOTENV, pool, NULL, NULL, 0);
	}

	static void
	test_set_bootenv(const char *pool)
	{
	nvlist_t *required = fnvlist_alloc();

	fnvlist_add_uint64(required, "version", VB_RAW);
	fnvlist_add_string(required, GRUB_ENVMAP, "test");

	IOC_INPUT_TEST_WILD(ZFS_IOC_SET_BOOTENV, pool, required, NULL, 0);

	nvlist_free(required);
	}

	static void
	zfs_ioc_input_tests(const char *pool)
	{
	char filepath[] = "/tmp/ioc_test_file_XXXXXX";
	char dataset[ZFS_MAX_DATASET_NAME_LEN];
	char snapbase[ZFS_MAX_DATASET_NAME_LEN + 32];
	char snapshot[ZFS_MAX_DATASET_NAME_LEN + 32];
	char bookmark[ZFS_MAX_DATASET_NAME_LEN + 32];
	char backup[ZFS_MAX_DATASET_NAME_LEN];
	char clone[ZFS_MAX_DATASET_NAME_LEN];
	char clonesnap[ZFS_MAX_DATASET_NAME_LEN + 32];
	int tmpfd, err;

	/*
	* Setup names and create a working dataset
	*/
	(void) snprintf(dataset, sizeof (dataset), "%s/test-fs", pool);
	(void) snprintf(snapbase, sizeof (snapbase), "%s@snapbase", dataset);
	(void) snprintf(snapshot, sizeof (snapshot), "%s@snapshot", dataset);
	(void) snprintf(bookmark, sizeof (bookmark), "%s#bookmark", dataset);
	(void) snprintf(clone, sizeof (clone), "%s/test-fs-clone", pool);
	(void) snprintf(clonesnap, sizeof (clonesnap), "%s@snap", clone);
	(void) snprintf(backup, sizeof (backup), "%s/backup", pool);

	err = lzc_create(dataset, LZC_DATSET_TYPE_ZFS, NULL, NULL, -1);
	if (err) {
	(void) fprintf(stderr, "could not create '%s': %s\n",
	dataset, strerror(errno));
	exit(2);
	}

	tmpfd = mkstemp(filepath);
	if (tmpfd < 0) {
	(void) fprintf(stderr, "could not create '%s': %s\n",
	filepath, strerror(errno));
	exit(2);
	}

	/*
	* run a test for each ioctl
	* Note that some test build on previous test operations
	*/
	test_pool_sync(pool);
	test_pool_reopen(pool);
	test_pool_checkpoint(pool);
	test_pool_discard_checkpoint(pool);
	test_log_history(pool);

	test_create(dataset);
	test_snapshot(pool, snapbase);
	test_snapshot(pool, snapshot);

	test_space_snaps(snapshot);
	test_send_space(snapbase, snapshot);
	test_send_new(snapshot, tmpfd);
	test_recv_new(backup, tmpfd);

	test_bookmark(pool, snapshot, bookmark);
	test_get_bookmarks(dataset);
	test_get_bookmark_props(bookmark);
	test_destroy_bookmarks(pool, bookmark);

	test_hold(pool, snapshot);
	test_get_holds(snapshot);
	test_release(pool, snapshot);

	test_clone(snapshot, clone);
	test_snapshot(pool, clonesnap);
	test_redact(snapshot, clonesnap);
	zfs_destroy(clonesnap);
	zfs_destroy(clone);

	test_rollback(dataset, snapshot);
	test_destroy_snaps(pool, snapshot);
	test_destroy_snaps(pool, snapbase);

	test_remap(dataset);
	test_channel_program(pool);

	test_load_key(dataset);
	test_change_key(dataset);
	test_unload_key(dataset);

	test_vdev_initialize(pool);
	test_vdev_trim(pool);

	test_wait(pool);
	test_wait_fs(dataset);

	test_set_bootenv(pool);
	test_get_bootenv(pool);

	test_scrub(pool);

	/*
	* cleanup
	*/
	zfs_cmd_t zc = {"\0"};

	nvlist_t *snaps = fnvlist_alloc();
	fnvlist_add_boolean(snaps, snapshot);
	(void) lzc_destroy_snaps(snaps, B_FALSE, NULL);
	nvlist_free(snaps);

	(void) zfs_destroy(dataset);
	(void) zfs_destroy(backup);

	(void) close(tmpfd);
	(void) unlink(filepath);

	/*
	* All the unused slots should yield ZFS_ERR_IOC_CMD_UNAVAIL
	*/
	for (int i = 0; i < ARRAY_SIZE(ioc_skip); i++) {
	if (ioc_tested[ioc_skip[i] - ZFS_IOC_FIRST])
	(void) fprintf(stderr, "cmd %d tested, not skipped!\n",
	(int)(ioc_skip[i] - ZFS_IOC_FIRST));

	ioc_tested[ioc_skip[i] - ZFS_IOC_FIRST] = B_TRUE;
	}

	(void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name));
	zc.zc_name[sizeof (zc.zc_name) - 1] = '\0';

	for (unsigned ioc = ZFS_IOC_FIRST; ioc < ZFS_IOC_LAST; ioc++) {
	unsigned cmd = ioc - ZFS_IOC_FIRST;

	if (ioc_tested[cmd])
	continue;

	if (lzc_ioctl_fd(zfs_fd, ioc, &zc) != 0 &&
	errno != ZFS_ERR_IOC_CMD_UNAVAIL) {
	(void) fprintf(stderr, "cmd %d is missing a test case "
	"(%d)\n", cmd, errno);
	}
	}
	}

	enum zfs_ioc_ref {
	#ifdef __FreeBSD__
	ZFS_IOC_BASE = 0,
	#else
	ZFS_IOC_BASE = ('Z' << 8),
	#endif
	ZFS_IOC_PLATFORM_BASE = ZFS_IOC_BASE + 0x80,
	};

	/*
	* Canonical reference check of /dev/zfs ioctl numbers.
	* These cannot change and new ioctl numbers must be appended.
	*/
	static boolean_t
	validate_ioc_values(void)
	{
	boolean_t result = B_TRUE;

	#define CHECK(expr) do { \
	if (!(expr)) { \
	result = B_FALSE; \
	fprintf(stderr, "(%s) === FALSE\n", #expr); \
	} \
	} while (0)

	CHECK(ZFS_IOC_BASE + 0 == ZFS_IOC_POOL_CREATE);
	CHECK(ZFS_IOC_BASE + 1 == ZFS_IOC_POOL_DESTROY);
	CHECK(ZFS_IOC_BASE + 2 == ZFS_IOC_POOL_IMPORT);
	CHECK(ZFS_IOC_BASE + 3 == ZFS_IOC_POOL_EXPORT);
	CHECK(ZFS_IOC_BASE + 4 == ZFS_IOC_POOL_CONFIGS);
	CHECK(ZFS_IOC_BASE + 5 == ZFS_IOC_POOL_STATS);
	CHECK(ZFS_IOC_BASE + 6 == ZFS_IOC_POOL_TRYIMPORT);
	CHECK(ZFS_IOC_BASE + 7 == ZFS_IOC_POOL_SCAN);
	CHECK(ZFS_IOC_BASE + 8 == ZFS_IOC_POOL_FREEZE);
	CHECK(ZFS_IOC_BASE + 9 == ZFS_IOC_POOL_UPGRADE);
	CHECK(ZFS_IOC_BASE + 10 == ZFS_IOC_POOL_GET_HISTORY);
	CHECK(ZFS_IOC_BASE + 11 == ZFS_IOC_VDEV_ADD);
	CHECK(ZFS_IOC_BASE + 12 == ZFS_IOC_VDEV_REMOVE);
	CHECK(ZFS_IOC_BASE + 13 == ZFS_IOC_VDEV_SET_STATE);
	CHECK(ZFS_IOC_BASE + 14 == ZFS_IOC_VDEV_ATTACH);
	CHECK(ZFS_IOC_BASE + 15 == ZFS_IOC_VDEV_DETACH);
	CHECK(ZFS_IOC_BASE + 16 == ZFS_IOC_VDEV_SETPATH);
	CHECK(ZFS_IOC_BASE + 17 == ZFS_IOC_VDEV_SETFRU);
	CHECK(ZFS_IOC_BASE + 18 == ZFS_IOC_OBJSET_STATS);
	CHECK(ZFS_IOC_BASE + 19 == ZFS_IOC_OBJSET_ZPLPROPS);
	CHECK(ZFS_IOC_BASE + 20 == ZFS_IOC_DATASET_LIST_NEXT);
	CHECK(ZFS_IOC_BASE + 21 == ZFS_IOC_SNAPSHOT_LIST_NEXT);
	CHECK(ZFS_IOC_BASE + 22 == ZFS_IOC_SET_PROP);
	CHECK(ZFS_IOC_BASE + 23 == ZFS_IOC_CREATE);
	CHECK(ZFS_IOC_BASE + 24 == ZFS_IOC_DESTROY);
	CHECK(ZFS_IOC_BASE + 25 == ZFS_IOC_ROLLBACK);
	CHECK(ZFS_IOC_BASE + 26 == ZFS_IOC_RENAME);
	CHECK(ZFS_IOC_BASE + 27 == ZFS_IOC_RECV);
	CHECK(ZFS_IOC_BASE + 28 == ZFS_IOC_SEND);
	CHECK(ZFS_IOC_BASE + 29 == ZFS_IOC_INJECT_FAULT);
	CHECK(ZFS_IOC_BASE + 30 == ZFS_IOC_CLEAR_FAULT);
	CHECK(ZFS_IOC_BASE + 31 == ZFS_IOC_INJECT_LIST_NEXT);
	CHECK(ZFS_IOC_BASE + 32 == ZFS_IOC_ERROR_LOG);
	CHECK(ZFS_IOC_BASE + 33 == ZFS_IOC_CLEAR);
	CHECK(ZFS_IOC_BASE + 34 == ZFS_IOC_PROMOTE);
	CHECK(ZFS_IOC_BASE + 35 == ZFS_IOC_SNAPSHOT);
	CHECK(ZFS_IOC_BASE + 36 == ZFS_IOC_DSOBJ_TO_DSNAME);
	CHECK(ZFS_IOC_BASE + 37 == ZFS_IOC_OBJ_TO_PATH);
	CHECK(ZFS_IOC_BASE + 38 == ZFS_IOC_POOL_SET_PROPS);
	CHECK(ZFS_IOC_BASE + 39 == ZFS_IOC_POOL_GET_PROPS);
	CHECK(ZFS_IOC_BASE + 40 == ZFS_IOC_SET_FSACL);
	CHECK(ZFS_IOC_BASE + 41 == ZFS_IOC_GET_FSACL);
	CHECK(ZFS_IOC_BASE + 42 == ZFS_IOC_SHARE);
	CHECK(ZFS_IOC_BASE + 43 == ZFS_IOC_INHERIT_PROP);
	CHECK(ZFS_IOC_BASE + 44 == ZFS_IOC_SMB_ACL);
	CHECK(ZFS_IOC_BASE + 45 == ZFS_IOC_USERSPACE_ONE);
	CHECK(ZFS_IOC_BASE + 46 == ZFS_IOC_USERSPACE_MANY);
	CHECK(ZFS_IOC_BASE + 47 == ZFS_IOC_USERSPACE_UPGRADE);
	CHECK(ZFS_IOC_BASE + 48 == ZFS_IOC_HOLD);
	CHECK(ZFS_IOC_BASE + 49 == ZFS_IOC_RELEASE);
	CHECK(ZFS_IOC_BASE + 50 == ZFS_IOC_GET_HOLDS);
	CHECK(ZFS_IOC_BASE + 51 == ZFS_IOC_OBJSET_RECVD_PROPS);
	CHECK(ZFS_IOC_BASE + 52 == ZFS_IOC_VDEV_SPLIT);
	CHECK(ZFS_IOC_BASE + 53 == ZFS_IOC_NEXT_OBJ);
	CHECK(ZFS_IOC_BASE + 54 == ZFS_IOC_DIFF);
	CHECK(ZFS_IOC_BASE + 55 == ZFS_IOC_TMP_SNAPSHOT);
	CHECK(ZFS_IOC_BASE + 56 == ZFS_IOC_OBJ_TO_STATS);
	CHECK(ZFS_IOC_BASE + 57 == ZFS_IOC_SPACE_WRITTEN);
	CHECK(ZFS_IOC_BASE + 58 == ZFS_IOC_SPACE_SNAPS);
	CHECK(ZFS_IOC_BASE + 59 == ZFS_IOC_DESTROY_SNAPS);
	CHECK(ZFS_IOC_BASE + 60 == ZFS_IOC_POOL_REGUID);
	CHECK(ZFS_IOC_BASE + 61 == ZFS_IOC_POOL_REOPEN);
	CHECK(ZFS_IOC_BASE + 62 == ZFS_IOC_SEND_PROGRESS);
	CHECK(ZFS_IOC_BASE + 63 == ZFS_IOC_LOG_HISTORY);
	CHECK(ZFS_IOC_BASE + 64 == ZFS_IOC_SEND_NEW);
	CHECK(ZFS_IOC_BASE + 65 == ZFS_IOC_SEND_SPACE);
	CHECK(ZFS_IOC_BASE + 66 == ZFS_IOC_CLONE);
	CHECK(ZFS_IOC_BASE + 67 == ZFS_IOC_BOOKMARK);
	CHECK(ZFS_IOC_BASE + 68 == ZFS_IOC_GET_BOOKMARKS);
	CHECK(ZFS_IOC_BASE + 69 == ZFS_IOC_DESTROY_BOOKMARKS);
	CHECK(ZFS_IOC_BASE + 70 == ZFS_IOC_RECV_NEW);
	CHECK(ZFS_IOC_BASE + 71 == ZFS_IOC_POOL_SYNC);
	CHECK(ZFS_IOC_BASE + 72 == ZFS_IOC_CHANNEL_PROGRAM);
	CHECK(ZFS_IOC_BASE + 73 == ZFS_IOC_LOAD_KEY);
	CHECK(ZFS_IOC_BASE + 74 == ZFS_IOC_UNLOAD_KEY);
	CHECK(ZFS_IOC_BASE + 75 == ZFS_IOC_CHANGE_KEY);
	CHECK(ZFS_IOC_BASE + 76 == ZFS_IOC_REMAP);
	CHECK(ZFS_IOC_BASE + 77 == ZFS_IOC_POOL_CHECKPOINT);
	CHECK(ZFS_IOC_BASE + 78 == ZFS_IOC_POOL_DISCARD_CHECKPOINT);
	CHECK(ZFS_IOC_BASE + 79 == ZFS_IOC_POOL_INITIALIZE);
	CHECK(ZFS_IOC_BASE + 80 == ZFS_IOC_POOL_TRIM);
	CHECK(ZFS_IOC_BASE + 81 == ZFS_IOC_REDACT);
	CHECK(ZFS_IOC_BASE + 82 == ZFS_IOC_GET_BOOKMARK_PROPS);
	CHECK(ZFS_IOC_BASE + 83 == ZFS_IOC_WAIT);
	CHECK(ZFS_IOC_BASE + 84 == ZFS_IOC_WAIT_FS);
	CHECK(ZFS_IOC_BASE + 87 == ZFS_IOC_POOL_SCRUB);
	CHECK(ZFS_IOC_PLATFORM_BASE + 1 == ZFS_IOC_EVENTS_NEXT);
	CHECK(ZFS_IOC_PLATFORM_BASE + 2 == ZFS_IOC_EVENTS_CLEAR);
	CHECK(ZFS_IOC_PLATFORM_BASE + 3 == ZFS_IOC_EVENTS_SEEK);
	CHECK(ZFS_IOC_PLATFORM_BASE + 4 == ZFS_IOC_NEXTBOOT);
	CHECK(ZFS_IOC_PLATFORM_BASE + 5 == ZFS_IOC_JAIL);
	CHECK(ZFS_IOC_PLATFORM_BASE + 6 == ZFS_IOC_UNJAIL);
	CHECK(ZFS_IOC_PLATFORM_BASE + 7 == ZFS_IOC_SET_BOOTENV);
	CHECK(ZFS_IOC_PLATFORM_BASE + 8 == ZFS_IOC_GET_BOOTENV);

	#undef CHECK

	return (result);
	}

	int
	main(int argc, const char *argv[])
	{
	if (argc != 2) {
	(void) fprintf(stderr, "usage: %s <pool>\n", argv[0]);
	exit(2);
	}

	if (!validate_ioc_values()) {
	(void) fprintf(stderr, "WARNING: zfs_ioc_t has binary "
	"incompatible command values\n");
	exit(3);
	}

	(void) libzfs_core_init();
	zfs_fd = open(ZFS_DEV, O_RDWR);
	if (zfs_fd < 0) {
	(void) fprintf(stderr, "open: %s\n", strerror(errno));
	libzfs_core_fini();
	exit(2);
	}

	zfs_ioc_input_tests(argv[1]);

	(void) close(zfs_fd);
	libzfs_core_fini();

	return (unexpected_failures);
	}
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/include/tunables.cfg b/sys/contrib/openzfs/tests/zfs-tests/include/tunables.cfg
	index a619b846dd11..721cf27f48ca 100644
	--- a/sys/contrib/openzfs/tests/zfs-tests/include/tunables.cfg
	+++ b/sys/contrib/openzfs/tests/zfs-tests/include/tunables.cfg
	@@ -1,106 +1,107 @@
	# This file exports variables for each tunable used in the test suite.
	#
	# Different platforms use different names for most tunables. To avoid littering
	# the tests with conditional logic for deciding how to set each tunable, the
	# logic is instead consolidated to this one file.
	#
	# Any use of tunables in tests must use a name defined here. New entries
	# should be added to the table as needed. Please keep the table sorted
	# alphabetically for ease of maintenance.
	#
	# Platform-specific tunables should still use a NAME from this table for
	# consistency. Enter UNSUPPORTED in the column for platforms on which the
	# tunable is not implemented.

	UNAME=$(uname)

	# NAME FreeBSD tunable Linux tunable
	cat <<%%%% \|
	ADMIN_SNAPSHOT UNSUPPORTED zfs_admin_snapshot
	ALLOW_REDACTED_DATASET_MOUNT allow_redacted_dataset_mount zfs_allow_redacted_dataset_mount
	ARC_MAX arc.max zfs_arc_max
	ARC_MIN arc.min zfs_arc_min
	ASYNC_BLOCK_MAX_BLOCKS async_block_max_blocks zfs_async_block_max_blocks
	CHECKSUM_EVENTS_PER_SECOND checksum_events_per_second zfs_checksum_events_per_second
	COMMIT_TIMEOUT_PCT commit_timeout_pct zfs_commit_timeout_pct
	COMPRESSED_ARC_ENABLED compressed_arc_enabled zfs_compressed_arc_enabled
	CONDENSE_INDIRECT_COMMIT_ENTRY_DELAY_MS condense.indirect_commit_entry_delay_ms zfs_condense_indirect_commit_entry_delay_ms
	CONDENSE_INDIRECT_OBSOLETE_PCT condense.indirect_obsolete_pct zfs_condense_indirect_obsolete_pct
	CONDENSE_MIN_MAPPING_BYTES condense.min_mapping_bytes zfs_condense_min_mapping_bytes
	DBUF_CACHE_SHIFT dbuf.cache_shift dbuf_cache_shift
	DEADMAN_CHECKTIME_MS deadman.checktime_ms zfs_deadman_checktime_ms
	+DEADMAN_EVENTS_PER_SECOND deadman_events_per_second zfs_deadman_events_per_second
	DEADMAN_FAILMODE deadman.failmode zfs_deadman_failmode
	DEADMAN_SYNCTIME_MS deadman.synctime_ms zfs_deadman_synctime_ms
	DEADMAN_ZIOTIME_MS deadman.ziotime_ms zfs_deadman_ziotime_ms
	DISABLE_IVSET_GUID_CHECK disable_ivset_guid_check zfs_disable_ivset_guid_check
	DMU_OFFSET_NEXT_SYNC dmu_offset_next_sync zfs_dmu_offset_next_sync
	EMBEDDED_SLOG_MIN_MS embedded_slog_min_ms zfs_embedded_slog_min_ms
	INITIALIZE_CHUNK_SIZE initialize_chunk_size zfs_initialize_chunk_size
	INITIALIZE_VALUE initialize_value zfs_initialize_value
	KEEP_LOG_SPACEMAPS_AT_EXPORT keep_log_spacemaps_at_export zfs_keep_log_spacemaps_at_export
	LUA_MAX_MEMLIMIT lua.max_memlimit zfs_lua_max_memlimit
	L2ARC_MFUONLY l2arc.mfuonly l2arc_mfuonly
	L2ARC_NOPREFETCH l2arc.noprefetch l2arc_noprefetch
	L2ARC_REBUILD_BLOCKS_MIN_L2SIZE l2arc.rebuild_blocks_min_l2size l2arc_rebuild_blocks_min_l2size
	L2ARC_REBUILD_ENABLED l2arc.rebuild_enabled l2arc_rebuild_enabled
	L2ARC_TRIM_AHEAD l2arc.trim_ahead l2arc_trim_ahead
	L2ARC_WRITE_BOOST l2arc.write_boost l2arc_write_boost
	L2ARC_WRITE_MAX l2arc.write_max l2arc_write_max
	LIVELIST_CONDENSE_NEW_ALLOC livelist.condense.new_alloc zfs_livelist_condense_new_alloc
	LIVELIST_CONDENSE_SYNC_CANCEL livelist.condense.sync_cancel zfs_livelist_condense_sync_cancel
	LIVELIST_CONDENSE_SYNC_PAUSE livelist.condense.sync_pause zfs_livelist_condense_sync_pause
	LIVELIST_CONDENSE_ZTHR_CANCEL livelist.condense.zthr_cancel zfs_livelist_condense_zthr_cancel
	LIVELIST_CONDENSE_ZTHR_PAUSE livelist.condense.zthr_pause zfs_livelist_condense_zthr_pause
	LIVELIST_MAX_ENTRIES livelist.max_entries zfs_livelist_max_entries
	LIVELIST_MIN_PERCENT_SHARED livelist.min_percent_shared zfs_livelist_min_percent_shared
	MAX_DATASET_NESTING max_dataset_nesting zfs_max_dataset_nesting
	MAX_MISSING_TVDS max_missing_tvds zfs_max_missing_tvds
	METASLAB_DEBUG_LOAD metaslab.debug_load metaslab_debug_load
	METASLAB_FORCE_GANGING metaslab.force_ganging metaslab_force_ganging
	MULTIHOST_FAIL_INTERVALS multihost.fail_intervals zfs_multihost_fail_intervals
	MULTIHOST_HISTORY multihost.history zfs_multihost_history
	MULTIHOST_IMPORT_INTERVALS multihost.import_intervals zfs_multihost_import_intervals
	MULTIHOST_INTERVAL multihost.interval zfs_multihost_interval
	OVERRIDE_ESTIMATE_RECORDSIZE send.override_estimate_recordsize zfs_override_estimate_recordsize
	PREFETCH_DISABLE prefetch.disable zfs_prefetch_disable
	RAIDZ_EXPAND_MAX_REFLOW_BYTES vdev.expand_max_reflow_bytes raidz_expand_max_reflow_bytes
	REBUILD_SCRUB_ENABLED rebuild_scrub_enabled zfs_rebuild_scrub_enabled
	REMOVAL_SUSPEND_PROGRESS removal_suspend_progress zfs_removal_suspend_progress
	REMOVE_MAX_SEGMENT remove_max_segment zfs_remove_max_segment
	RESILVER_MIN_TIME_MS resilver_min_time_ms zfs_resilver_min_time_ms
	SCAN_LEGACY scan_legacy zfs_scan_legacy
	SCAN_SUSPEND_PROGRESS scan_suspend_progress zfs_scan_suspend_progress
	SCAN_VDEV_LIMIT scan_vdev_limit zfs_scan_vdev_limit
	SCRUB_AFTER_EXPAND scrub_after_expand zfs_scrub_after_expand
	SEND_HOLES_WITHOUT_BIRTH_TIME send_holes_without_birth_time send_holes_without_birth_time
	SLOW_IO_EVENTS_PER_SECOND slow_io_events_per_second zfs_slow_io_events_per_second
	SPA_ASIZE_INFLATION spa.asize_inflation spa_asize_inflation
	SPA_DISCARD_MEMORY_LIMIT spa.discard_memory_limit zfs_spa_discard_memory_limit
	SPA_LOAD_VERIFY_DATA spa.load_verify_data spa_load_verify_data
	SPA_LOAD_VERIFY_METADATA spa.load_verify_metadata spa_load_verify_metadata
	TRIM_EXTENT_BYTES_MIN trim.extent_bytes_min zfs_trim_extent_bytes_min
	TRIM_METASLAB_SKIP trim.metaslab_skip zfs_trim_metaslab_skip
	TRIM_TXG_BATCH trim.txg_batch zfs_trim_txg_batch
	TXG_HISTORY txg.history zfs_txg_history
	TXG_TIMEOUT txg.timeout zfs_txg_timeout
	UNLINK_SUSPEND_PROGRESS UNSUPPORTED zfs_unlink_suspend_progress
	VDEV_FILE_LOGICAL_ASHIFT vdev.file.logical_ashift vdev_file_logical_ashift
	VDEV_FILE_PHYSICAL_ASHIFT vdev.file.physical_ashift vdev_file_physical_ashift
	VDEV_MAX_AUTO_ASHIFT vdev.max_auto_ashift zfs_vdev_max_auto_ashift
	VDEV_MIN_MS_COUNT vdev.min_ms_count zfs_vdev_min_ms_count
	VDEV_VALIDATE_SKIP vdev.validate_skip vdev_validate_skip
	VOL_INHIBIT_DEV UNSUPPORTED zvol_inhibit_dev
	VOL_MODE vol.mode zvol_volmode
	VOL_RECURSIVE vol.recursive UNSUPPORTED
	VOL_USE_BLK_MQ UNSUPPORTED zvol_use_blk_mq
	BCLONE_ENABLED zfs_bclone_enabled zfs_bclone_enabled
	BCLONE_WAIT_DIRTY zfs_bclone_wait_dirty zfs_bclone_wait_dirty
	XATTR_COMPAT xattr_compat zfs_xattr_compat
	ZEVENT_LEN_MAX zevent.len_max zfs_zevent_len_max
	ZEVENT_RETAIN_MAX zevent.retain_max zfs_zevent_retain_max
	ZIO_SLOW_IO_MS zio.slow_io_ms zio_slow_io_ms
	ZIL_SAXATTR zil_saxattr zfs_zil_saxattr
	%%%%
	while read name FreeBSD Linux; do
	eval "export ${name}=\$${UNAME}"
	done
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/tests/Makefile.am b/sys/contrib/openzfs/tests/zfs-tests/tests/Makefile.am
	index d625c040b819..44eedcf6fae5 100644
	--- a/sys/contrib/openzfs/tests/zfs-tests/tests/Makefile.am
	+++ b/sys/contrib/openzfs/tests/zfs-tests/tests/Makefile.am
	@@ -1,2124 +1,2126 @@
	CLEANFILES =
	dist_noinst_DATA =
	include $(top_srcdir)/config/Substfiles.am


	datadir_zfs_tests_testsdir = $(datadir)/$(PACKAGE)/zfs-tests/tests
	nobase_dist_datadir_zfs_tests_tests_DATA = \
	perf/nfs-sample.cfg \
	perf/perf.shlib \
	\
	perf/fio/mkfiles.fio \
	perf/fio/random_reads.fio \
	perf/fio/random_readwrite.fio \
	perf/fio/random_readwrite_fixed.fio \
	perf/fio/random_writes.fio \
	perf/fio/sequential_reads.fio \
	perf/fio/sequential_readwrite.fio \
	perf/fio/sequential_writes.fio

	nobase_dist_datadir_zfs_tests_tests_SCRIPTS = \
	perf/regression/random_reads.ksh \
	perf/regression/random_readwrite.ksh \
	perf/regression/random_readwrite_fixed.ksh \
	perf/regression/random_writes.ksh \
	perf/regression/random_writes_zil.ksh \
	perf/regression/sequential_reads_arc_cached_clone.ksh \
	perf/regression/sequential_reads_arc_cached.ksh \
	perf/regression/sequential_reads_dbuf_cached.ksh \
	perf/regression/sequential_reads.ksh \
	perf/regression/sequential_writes.ksh \
	perf/regression/setup.ksh \
	\
	perf/scripts/prefetch_io.sh

	# These lists can be regenerated by running make regen-tests at the root, or, on a clean source:
	# find functional/ ! -type d ! -name .gitignore ! -name .dirstamp ! -name '.Po' ! -executable -name '.in' \| sort \| sed 's/\.in$//;s/^/\t/;$!s/$/ \\/'
	# find functional/ ! -type d ! -name .gitignore ! -name .dirstamp ! -name '.Po' -executable -name '.in' \| sort \| sed 's/\.in$//;s/^/\t/;$!s/$/ \\/'
	# find functional/ ! -type d ! -name .gitignore ! -name .dirstamp ! -name '.Po' ! -name '.in' ! -name '*.c' \| grep -Fe /simd -e /tmpfile \| sort \| sed 's/^/\t/;$!s/$/ \\/'
	# find functional/ ! -type d ! -name .gitignore ! -name .dirstamp ! -name '.Po' ! -executable ! -name '.in' ! -name '*.c' \| grep -vFe /simd -e /tmpfile \| sort \| sed 's/^/\t/;$!s/$/ \\/'
	# find functional/ ! -type d ! -name .gitignore ! -name .dirstamp ! -name '.Po' -executable ! -name '.in' ! -name '*.c' \| grep -vFe /simd -e /tmpfile \| sort \| sed 's/^/\t/;$!s/$/ \\/'
	#
	# simd and tmpfile are Linux-only and not installed elsewhere
	#
	# C programs are specced in ../Makefile.am above as part of the main Makefile

	find_common := find functional/ ! -type d ! -name .gitignore ! -name .dirstamp ! -name '*.Po'
	regen:
	@$(MAKE) -C $(top_builddir) clean
	@$(MAKE) clean
	$(SED) $(ac_inplace) '/^# -- >8 --/q' Makefile.am
	echo >> Makefile.am
	echo 'nobase_nodist_datadir_zfs_tests_tests_DATA = \' >> Makefile.am
	$(find_common) ! -executable -name '*.in' \| sort \| sed 's/\.in$$//;s/^/\t/;$$!s/$$/ \\/' >> Makefile.am
	echo 'nobase_nodist_datadir_zfs_tests_tests_SCRIPTS = \' >> Makefile.am
	$(find_common) -executable -name '*.in' \| sort \| sed 's/\.in$$//;s/^/\t/;$$!s/$$/ \\/' >> Makefile.am
	echo >> Makefile.am
	echo 'SUBSTFILES += $$(nobase_nodist_datadir_zfs_tests_tests_DATA) $$(nobase_nodist_datadir_zfs_tests_tests_SCRIPTS)' >> Makefile.am
	echo >> Makefile.am
	echo 'if BUILD_LINUX' >> Makefile.am
	echo 'nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \' >> Makefile.am
	$(find_common) ! -name '.in' ! -name '.c' \| grep -Fe /simd -e /tmpfile \| sort \| sed 's/^/\t/;$$!s/$$/ \\/' >> Makefile.am
	echo 'endif' >> Makefile.am
	echo >> Makefile.am
	echo 'nobase_dist_datadir_zfs_tests_tests_DATA += \' >> Makefile.am
	$(find_common) ! -executable ! -name '.in' ! -name '.c' \| grep -vFe /simd -e /tmpfile \| sort \| sed 's/^/\t/;$$!s/$$/ \\/' >> Makefile.am
	echo >> Makefile.am
	echo 'nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \' >> Makefile.am
	$(find_common) -executable ! -name '.in' ! -name '.c' \| grep -vFe /simd -e /tmpfile \| sort \| sed 's/^/\t/;$$!s/$$/ \\/' >> Makefile.am

	# -- >8 --

	nobase_nodist_datadir_zfs_tests_tests_DATA = \
	functional/pam/utilities.kshlib
	nobase_nodist_datadir_zfs_tests_tests_SCRIPTS = \
	functional/pyzfs/pyzfs_unittest.ksh

	SUBSTFILES += $(nobase_nodist_datadir_zfs_tests_tests_DATA) $(nobase_nodist_datadir_zfs_tests_tests_SCRIPTS)

	if BUILD_LINUX
	nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \
	functional/simd/simd_supported.ksh \
	functional/tmpfile/cleanup.ksh \
	functional/tmpfile/setup.ksh
	endif

	nobase_dist_datadir_zfs_tests_tests_DATA += \
	functional/acl/acl.cfg \
	functional/acl/acl_common.kshlib \
	functional/alloc_class/alloc_class.cfg \
	functional/alloc_class/alloc_class.kshlib \
	functional/atime/atime.cfg \
	functional/atime/atime_common.kshlib \
	functional/bclone/bclone.cfg \
	functional/bclone/bclone_common.kshlib \
	functional/bclone/bclone_corner_cases.kshlib \
	functional/block_cloning/block_cloning.kshlib \
	functional/cache/cache.cfg \
	functional/cache/cache.kshlib \
	functional/cachefile/cachefile.cfg \
	functional/cachefile/cachefile.kshlib \
	functional/casenorm/casenorm.cfg \
	functional/casenorm/casenorm.kshlib \
	functional/channel_program/channel_common.kshlib \
	functional/channel_program/lua_core/tst.args_to_lua.out \
	functional/channel_program/lua_core/tst.args_to_lua.zcp \
	functional/channel_program/lua_core/tst.divide_by_zero.err \
	functional/channel_program/lua_core/tst.divide_by_zero.zcp \
	functional/channel_program/lua_core/tst.exists.zcp \
	functional/channel_program/lua_core/tst.large_prog.out \
	functional/channel_program/lua_core/tst.large_prog.zcp \
	functional/channel_program/lua_core/tst.lib_base.lua \
	functional/channel_program/lua_core/tst.lib_coroutine.lua \
	functional/channel_program/lua_core/tst.lib_strings.lua \
	functional/channel_program/lua_core/tst.lib_table.lua \
	functional/channel_program/lua_core/tst.nested_neg.zcp \
	functional/channel_program/lua_core/tst.nested_pos.zcp \
	functional/channel_program/lua_core/tst.recursive.zcp \
	functional/channel_program/lua_core/tst.return_large.zcp \
	functional/channel_program/lua_core/tst.return_recursive_table.zcp \
	functional/channel_program/lua_core/tst.stack_gsub.err \
	functional/channel_program/lua_core/tst.stack_gsub.zcp \
	functional/channel_program/lua_core/tst.timeout.zcp \
	functional/channel_program/synctask_core/tst.bookmark.copy.zcp \
	functional/channel_program/synctask_core/tst.bookmark.create.zcp \
	functional/channel_program/synctask_core/tst.get_index_props.out \
	functional/channel_program/synctask_core/tst.get_index_props.zcp \
	functional/channel_program/synctask_core/tst.get_number_props.out \
	functional/channel_program/synctask_core/tst.get_number_props.zcp \
	functional/channel_program/synctask_core/tst.get_string_props.out \
	functional/channel_program/synctask_core/tst.get_string_props.zcp \
	functional/channel_program/synctask_core/tst.promote_conflict.zcp \
	functional/channel_program/synctask_core/tst.set_props.zcp \
	functional/channel_program/synctask_core/tst.snapshot_destroy.zcp \
	functional/channel_program/synctask_core/tst.snapshot_neg.zcp \
	functional/channel_program/synctask_core/tst.snapshot_recursive.zcp \
	functional/channel_program/synctask_core/tst.snapshot_rename.zcp \
	functional/channel_program/synctask_core/tst.snapshot_simple.zcp \
	functional/checksum/default.cfg \
	functional/clean_mirror/clean_mirror_common.kshlib \
	functional/clean_mirror/default.cfg \
	functional/cli_root/cli_common.kshlib \
	functional/cli_root/zfs_copies/zfs_copies.cfg \
	functional/cli_root/zfs_copies/zfs_copies.kshlib \
	functional/cli_root/zfs_create/properties.kshlib \
	functional/cli_root/zfs_create/zfs_create.cfg \
	functional/cli_root/zfs_create/zfs_create_common.kshlib \
	functional/cli_root/zfs_destroy/zfs_destroy.cfg \
	functional/cli_root/zfs_destroy/zfs_destroy_common.kshlib \
	functional/cli_root/zfs_get/zfs_get_common.kshlib \
	functional/cli_root/zfs_get/zfs_get_list_d.kshlib \
	functional/cli_root/zfs_jail/jail.conf \
	functional/cli_root/zfs_load-key/HEXKEY \
	functional/cli_root/zfs_load-key/PASSPHRASE \
	functional/cli_root/zfs_load-key/RAWKEY \
	functional/cli_root/zfs_load-key/zfs_load-key.cfg \
	functional/cli_root/zfs_load-key/zfs_load-key_common.kshlib \
	functional/cli_root/zfs_mount/zfs_mount.cfg \
	functional/cli_root/zfs_mount/zfs_mount.kshlib \
	functional/cli_root/zfs_promote/zfs_promote.cfg \
	functional/cli_root/zfs_receive/zstd_test_data.txt \
	functional/cli_root/zfs_rename/zfs_rename.cfg \
	functional/cli_root/zfs_rename/zfs_rename.kshlib \
	functional/cli_root/zfs_rollback/zfs_rollback.cfg \
	functional/cli_root/zfs_rollback/zfs_rollback_common.kshlib \
	functional/cli_root/zfs_send/zfs_send.cfg \
	functional/cli_root/zfs_set/zfs_set_common.kshlib \
	functional/cli_root/zfs_share/zfs_share.cfg \
	functional/cli_root/zfs_snapshot/zfs_snapshot.cfg \
	functional/cli_root/zfs_unmount/zfs_unmount.cfg \
	functional/cli_root/zfs_unmount/zfs_unmount.kshlib \
	functional/cli_root/zfs_upgrade/zfs_upgrade.kshlib \
	functional/cli_root/zfs_wait/zfs_wait.kshlib \
	functional/cli_root/zpool_add/zpool_add.cfg \
	functional/cli_root/zpool_add/zpool_add.kshlib \
	functional/cli_root/zpool_clear/zpool_clear.cfg \
	functional/cli_root/zpool_create/draidcfg.gz \
	functional/cli_root/zpool_create/zpool_create.cfg \
	functional/cli_root/zpool_create/zpool_create.shlib \
	functional/cli_root/zpool_destroy/zpool_destroy.cfg \
	functional/cli_root/zpool_events/zpool_events.cfg \
	functional/cli_root/zpool_events/zpool_events.kshlib \
	functional/cli_root/zpool_expand/zpool_expand.cfg \
	functional/cli_root/zpool_export/zpool_export.cfg \
	functional/cli_root/zpool_export/zpool_export.kshlib \
	functional/cli_root/zpool_get/vdev_get.cfg \
	functional/cli_root/zpool_get/zpool_get.cfg \
	functional/cli_root/zpool_get/zpool_get_parsable.cfg \
	functional/cli_root/zpool_import/blockfiles/cryptv0.dat.bz2 \
	functional/cli_root/zpool_import/blockfiles/missing_ivset.dat.bz2 \
	functional/cli_root/zpool_import/blockfiles/unclean_export.dat.bz2 \
	functional/cli_root/zpool_import/zpool_import.cfg \
	functional/cli_root/zpool_import/zpool_import.kshlib \
	functional/cli_root/zpool_initialize/zpool_initialize.kshlib \
	functional/cli_root/zpool_labelclear/labelclear.cfg \
	functional/cli_root/zpool_remove/zpool_remove.cfg \
	functional/cli_root/zpool_reopen/zpool_reopen.cfg \
	functional/cli_root/zpool_reopen/zpool_reopen.shlib \
	functional/cli_root/zpool_resilver/zpool_resilver.cfg \
	functional/cli_root/zpool_scrub/zpool_scrub.cfg \
	functional/cli_root/zpool_split/zpool_split.cfg \
	functional/cli_root/zpool_trim/zpool_trim.kshlib \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-broken-mirror1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-broken-mirror2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v10.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v11.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v12.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v13.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v14.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v15.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1mirror1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1mirror2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1mirror3.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1raidz1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1raidz2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1raidz3.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1stripe1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1stripe2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v1stripe3.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2mirror1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2mirror2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2mirror3.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2raidz1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2raidz2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2raidz3.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2stripe1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2stripe2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v2stripe3.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3hotspare1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3hotspare2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3hotspare3.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3mirror1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3mirror2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3mirror3.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3raidz1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3raidz21.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3raidz22.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3raidz23.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3raidz2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3raidz3.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3stripe1.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3stripe2.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v3stripe3.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v4.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v5.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v6.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v7.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v8.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v999.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-v9.dat.bz2 \
	functional/cli_root/zpool_upgrade/blockfiles/zfs-pool-vBROKEN.dat.bz2 \
	functional/cli_root/zpool_upgrade/zpool_upgrade.cfg \
	functional/cli_root/zpool_upgrade/zpool_upgrade.kshlib \
	functional/cli_root/zpool_wait/zpool_wait.kshlib \
	functional/cli_root/zhack/library.kshlib \
	functional/cli_user/misc/misc.cfg \
	functional/cli_user/zfs_list/zfs_list.cfg \
	functional/cli_user/zfs_list/zfs_list.kshlib \
	functional/compression/compress.cfg \
	functional/compression/testpool_zstd.tar.gz \
	functional/deadman/deadman.cfg \
	functional/delegate/delegate.cfg \
	functional/delegate/delegate_common.kshlib \
	functional/devices/devices.cfg \
	functional/devices/devices_common.kshlib \
	functional/events/events.cfg \
	functional/events/events_common.kshlib \
	functional/fault/fault.cfg \
	functional/grow/grow.cfg \
	functional/history/history.cfg \
	functional/history/history_common.kshlib \
	functional/history/i386.migratedpool.DAT.Z \
	functional/history/i386.orig_history.txt \
	functional/history/sparc.migratedpool.DAT.Z \
	functional/history/sparc.orig_history.txt \
	functional/history/zfs-pool-v4.dat.Z \
	functional/inheritance/config001.cfg \
	functional/inheritance/config002.cfg \
	functional/inheritance/config003.cfg \
	functional/inheritance/config004.cfg \
	functional/inheritance/config005.cfg \
	functional/inheritance/config006.cfg \
	functional/inheritance/config007.cfg \
	functional/inheritance/config008.cfg \
	functional/inheritance/config009.cfg \
	functional/inheritance/config010.cfg \
	functional/inheritance/config011.cfg \
	functional/inheritance/config012.cfg \
	functional/inheritance/config013.cfg \
	functional/inheritance/config014.cfg \
	functional/inheritance/config015.cfg \
	functional/inheritance/config016.cfg \
	functional/inheritance/config017.cfg \
	functional/inheritance/config018.cfg \
	functional/inheritance/config019.cfg \
	functional/inheritance/config020.cfg \
	functional/inheritance/config021.cfg \
	functional/inheritance/config022.cfg \
	functional/inheritance/config023.cfg \
	functional/inheritance/config024.cfg \
	functional/inheritance/inherit.kshlib \
	functional/inheritance/README.config \
	functional/inheritance/README.state \
	functional/inheritance/state001.cfg \
	functional/inheritance/state002.cfg \
	functional/inheritance/state003.cfg \
	functional/inheritance/state004.cfg \
	functional/inheritance/state005.cfg \
	functional/inheritance/state006.cfg \
	functional/inheritance/state007.cfg \
	functional/inheritance/state008.cfg \
	functional/inheritance/state009.cfg \
	functional/inheritance/state010.cfg \
	functional/inheritance/state011.cfg \
	functional/inheritance/state012.cfg \
	functional/inheritance/state013.cfg \
	functional/inheritance/state014.cfg \
	functional/inheritance/state015.cfg \
	functional/inheritance/state016.cfg \
	functional/inheritance/state017.cfg \
	functional/inheritance/state018.cfg \
	functional/inheritance/state019.cfg \
	functional/inheritance/state020.cfg \
	functional/inheritance/state021.cfg \
	functional/inheritance/state022.cfg \
	functional/inheritance/state023.cfg \
	functional/inheritance/state024.cfg \
	functional/inuse/inuse.cfg \
	functional/io/io.cfg \
	functional/l2arc/l2arc.cfg \
	functional/largest_pool/largest_pool.cfg \
	functional/migration/migration.cfg \
	functional/migration/migration.kshlib \
	functional/mmap/mmap.cfg \
	functional/mmp/mmp.cfg \
	functional/mmp/mmp.kshlib \
	functional/mv_files/mv_files.cfg \
	functional/mv_files/mv_files_common.kshlib \
	functional/nopwrite/nopwrite.shlib \
	functional/no_space/enospc.cfg \
	functional/online_offline/online_offline.cfg \
	functional/pool_checkpoint/pool_checkpoint.kshlib \
	functional/projectquota/projectquota.cfg \
	functional/projectquota/projectquota_common.kshlib \
	functional/quota/quota.cfg \
	functional/quota/quota.kshlib \
	functional/redacted_send/redacted.cfg \
	functional/redacted_send/redacted.kshlib \
	functional/redundancy/redundancy.cfg \
	functional/redundancy/redundancy.kshlib \
	functional/refreserv/refreserv.cfg \
	functional/removal/removal.kshlib \
	functional/replacement/replacement.cfg \
	functional/reservation/reservation.cfg \
	functional/reservation/reservation.shlib \
	functional/rsend/dedup_encrypted_zvol.bz2 \
	functional/rsend/dedup_encrypted_zvol.zsend.bz2 \
	functional/rsend/dedup.zsend.bz2 \
	functional/rsend/fs.tar.gz \
	functional/rsend/rsend.cfg \
	functional/rsend/rsend.kshlib \
	functional/scrub_mirror/default.cfg \
	functional/scrub_mirror/scrub_mirror_common.kshlib \
	functional/slog/slog.cfg \
	functional/slog/slog.kshlib \
	functional/snapshot/snapshot.cfg \
	functional/snapused/snapused.kshlib \
	functional/sparse/sparse.cfg \
	functional/trim/trim.cfg \
	functional/trim/trim.kshlib \
	functional/truncate/truncate.cfg \
	functional/upgrade/upgrade_common.kshlib \
	functional/user_namespace/user_namespace.cfg \
	functional/user_namespace/user_namespace_common.kshlib \
	functional/userquota/13709_reproducer.bz2 \
	functional/userquota/userquota.cfg \
	functional/userquota/userquota_common.kshlib \
	functional/vdev_zaps/vdev_zaps.kshlib \
	functional/xattr/xattr.cfg \
	functional/xattr/xattr_common.kshlib \
	functional/zvol/zvol.cfg \
	functional/zvol/zvol_cli/zvol_cli.cfg \
	functional/zvol/zvol_common.shlib \
	functional/zvol/zvol_ENOSPC/zvol_ENOSPC.cfg \
	functional/zvol/zvol_misc/zvol_misc_common.kshlib \
	functional/zvol/zvol_swap/zvol_swap.cfg \
	functional/idmap_mount/idmap_mount.cfg \
	functional/idmap_mount/idmap_mount_common.kshlib

	nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \
	functional/acl/off/cleanup.ksh \
	functional/acl/off/dosmode.ksh \
	functional/acl/off/posixmode.ksh \
	functional/acl/off/setup.ksh \
	functional/acl/posix/cleanup.ksh \
	functional/acl/posix/posix_001_pos.ksh \
	functional/acl/posix/posix_002_pos.ksh \
	functional/acl/posix/posix_003_pos.ksh \
	functional/acl/posix/posix_004_pos.ksh \
	functional/acl/posix-sa/cleanup.ksh \
	functional/acl/posix-sa/posix_001_pos.ksh \
	functional/acl/posix-sa/posix_002_pos.ksh \
	functional/acl/posix-sa/posix_003_pos.ksh \
	functional/acl/posix-sa/posix_004_pos.ksh \
	functional/acl/posix-sa/setup.ksh \
	functional/acl/posix/setup.ksh \
	functional/alloc_class/alloc_class_001_pos.ksh \
	functional/alloc_class/alloc_class_002_neg.ksh \
	functional/alloc_class/alloc_class_003_pos.ksh \
	functional/alloc_class/alloc_class_004_pos.ksh \
	functional/alloc_class/alloc_class_005_pos.ksh \
	functional/alloc_class/alloc_class_006_pos.ksh \
	functional/alloc_class/alloc_class_007_pos.ksh \
	functional/alloc_class/alloc_class_008_pos.ksh \
	functional/alloc_class/alloc_class_009_pos.ksh \
	functional/alloc_class/alloc_class_010_pos.ksh \
	functional/alloc_class/alloc_class_011_neg.ksh \
	functional/alloc_class/alloc_class_012_pos.ksh \
	functional/alloc_class/alloc_class_013_pos.ksh \
	functional/alloc_class/alloc_class_014_neg.ksh \
	functional/alloc_class/alloc_class_015_pos.ksh \
	functional/alloc_class/cleanup.ksh \
	functional/alloc_class/setup.ksh \
	functional/append/file_append.ksh \
	functional/append/threadsappend_001_pos.ksh \
	functional/append/cleanup.ksh \
	functional/append/setup.ksh \
	functional/arc/arcstats_runtime_tuning.ksh \
	functional/arc/cleanup.ksh \
	functional/arc/dbufstats_001_pos.ksh \
	functional/arc/dbufstats_002_pos.ksh \
	functional/arc/dbufstats_003_pos.ksh \
	functional/arc/setup.ksh \
	functional/atime/atime_001_pos.ksh \
	functional/atime/atime_002_neg.ksh \
	functional/atime/atime_003_pos.ksh \
	functional/atime/cleanup.ksh \
	functional/atime/root_atime_off.ksh \
	functional/atime/root_atime_on.ksh \
	functional/atime/root_relatime_on.ksh \
	functional/atime/setup.ksh \
	functional/bclone/bclone_crossfs_corner_cases.ksh \
	functional/bclone/bclone_crossfs_corner_cases_limited.ksh \
	functional/bclone/bclone_crossfs_data.ksh \
	functional/bclone/bclone_crossfs_embedded.ksh \
	functional/bclone/bclone_crossfs_hole.ksh \
	functional/bclone/bclone_diffprops_all.ksh \
	functional/bclone/bclone_diffprops_checksum.ksh \
	functional/bclone/bclone_diffprops_compress.ksh \
	functional/bclone/bclone_diffprops_copies.ksh \
	functional/bclone/bclone_diffprops_recordsize.ksh \
	functional/bclone/bclone_prop_sync.ksh \
	functional/bclone/bclone_samefs_corner_cases.ksh \
	functional/bclone/bclone_samefs_corner_cases_limited.ksh \
	functional/bclone/bclone_samefs_data.ksh \
	functional/bclone/bclone_samefs_embedded.ksh \
	functional/bclone/bclone_samefs_hole.ksh \
	functional/bclone/cleanup.ksh \
	functional/bclone/setup.ksh \
	functional/block_cloning/cleanup.ksh \
	functional/block_cloning/setup.ksh \
	functional/block_cloning/block_cloning_clone_mmap_cached.ksh \
	functional/block_cloning/block_cloning_clone_mmap_write.ksh \
	functional/block_cloning/block_cloning_copyfilerange_cross_dataset.ksh \
	functional/block_cloning/block_cloning_copyfilerange_fallback.ksh \
	functional/block_cloning/block_cloning_copyfilerange_fallback_same_txg.ksh \
	functional/block_cloning/block_cloning_copyfilerange.ksh \
	functional/block_cloning/block_cloning_copyfilerange_partial.ksh \
	functional/block_cloning/block_cloning_disabled_copyfilerange.ksh \
	functional/block_cloning/block_cloning_disabled_ficlone.ksh \
	functional/block_cloning/block_cloning_disabled_ficlonerange.ksh \
	functional/block_cloning/block_cloning_ficlone.ksh \
	functional/block_cloning/block_cloning_ficlonerange.ksh \
	functional/block_cloning/block_cloning_ficlonerange_partial.ksh \
	functional/block_cloning/block_cloning_cross_enc_dataset.ksh \
	functional/block_cloning/block_cloning_replay.ksh \
	functional/block_cloning/block_cloning_replay_encrypted.ksh \
	functional/block_cloning/block_cloning_lwb_buffer_overflow.ksh \
	functional/bootfs/bootfs_001_pos.ksh \
	functional/bootfs/bootfs_002_neg.ksh \
	functional/bootfs/bootfs_003_pos.ksh \
	functional/bootfs/bootfs_004_neg.ksh \
	functional/bootfs/bootfs_005_neg.ksh \
	functional/bootfs/bootfs_006_pos.ksh \
	functional/bootfs/bootfs_007_pos.ksh \
	functional/bootfs/bootfs_008_pos.ksh \
	functional/bootfs/cleanup.ksh \
	functional/bootfs/setup.ksh \
	functional/btree/btree_negative.ksh \
	functional/btree/btree_positive.ksh \
	functional/cache/cache_001_pos.ksh \
	functional/cache/cache_002_pos.ksh \
	functional/cache/cache_003_pos.ksh \
	functional/cache/cache_004_neg.ksh \
	functional/cache/cache_005_neg.ksh \
	functional/cache/cache_006_pos.ksh \
	functional/cache/cache_007_neg.ksh \
	functional/cache/cache_008_neg.ksh \
	functional/cache/cache_009_pos.ksh \
	functional/cache/cache_010_pos.ksh \
	functional/cache/cache_011_pos.ksh \
	functional/cache/cache_012_pos.ksh \
	functional/cache/cleanup.ksh \
	functional/cachefile/cachefile_001_pos.ksh \
	functional/cachefile/cachefile_002_pos.ksh \
	functional/cachefile/cachefile_003_pos.ksh \
	functional/cachefile/cachefile_004_pos.ksh \
	functional/cachefile/cleanup.ksh \
	functional/cachefile/setup.ksh \
	functional/cache/setup.ksh \
	functional/casenorm/case_all_values.ksh \
	functional/casenorm/cleanup.ksh \
	functional/casenorm/insensitive_formd_delete.ksh \
	functional/casenorm/insensitive_formd_lookup.ksh \
	functional/casenorm/insensitive_none_delete.ksh \
	functional/casenorm/insensitive_none_lookup.ksh \
	functional/casenorm/mixed_create_failure.ksh \
	functional/casenorm/mixed_formd_delete.ksh \
	functional/casenorm/mixed_formd_lookup_ci.ksh \
	functional/casenorm/mixed_formd_lookup.ksh \
	functional/casenorm/mixed_none_delete.ksh \
	functional/casenorm/mixed_none_lookup_ci.ksh \
	functional/casenorm/mixed_none_lookup.ksh \
	functional/casenorm/norm_all_values.ksh \
	functional/casenorm/sensitive_formd_delete.ksh \
	functional/casenorm/sensitive_formd_lookup.ksh \
	functional/casenorm/sensitive_none_delete.ksh \
	functional/casenorm/sensitive_none_lookup.ksh \
	functional/casenorm/setup.ksh \
	functional/channel_program/lua_core/cleanup.ksh \
	functional/channel_program/lua_core/setup.ksh \
	functional/channel_program/lua_core/tst.args_to_lua.ksh \
	functional/channel_program/lua_core/tst.divide_by_zero.ksh \
	functional/channel_program/lua_core/tst.exists.ksh \
	functional/channel_program/lua_core/tst.integer_illegal.ksh \
	functional/channel_program/lua_core/tst.integer_overflow.ksh \
	functional/channel_program/lua_core/tst.language_functions_neg.ksh \
	functional/channel_program/lua_core/tst.language_functions_pos.ksh \
	functional/channel_program/lua_core/tst.large_prog.ksh \
	functional/channel_program/lua_core/tst.libraries.ksh \
	functional/channel_program/lua_core/tst.memory_limit.ksh \
	functional/channel_program/lua_core/tst.nested_neg.ksh \
	functional/channel_program/lua_core/tst.nested_pos.ksh \
	functional/channel_program/lua_core/tst.nvlist_to_lua.ksh \
	functional/channel_program/lua_core/tst.recursive_neg.ksh \
	functional/channel_program/lua_core/tst.recursive_pos.ksh \
	functional/channel_program/lua_core/tst.return_large.ksh \
	functional/channel_program/lua_core/tst.return_nvlist_neg.ksh \
	functional/channel_program/lua_core/tst.return_nvlist_pos.ksh \
	functional/channel_program/lua_core/tst.return_recursive_table.ksh \
	functional/channel_program/lua_core/tst.stack_gsub.ksh \
	functional/channel_program/lua_core/tst.timeout.ksh \
	functional/channel_program/synctask_core/cleanup.ksh \
	functional/channel_program/synctask_core/setup.ksh \
	functional/channel_program/synctask_core/tst.bookmark.copy.ksh \
	functional/channel_program/synctask_core/tst.bookmark.create.ksh \
	functional/channel_program/synctask_core/tst.destroy_fs.ksh \
	functional/channel_program/synctask_core/tst.destroy_snap.ksh \
	functional/channel_program/synctask_core/tst.get_count_and_limit.ksh \
	functional/channel_program/synctask_core/tst.get_index_props.ksh \
	functional/channel_program/synctask_core/tst.get_mountpoint.ksh \
	functional/channel_program/synctask_core/tst.get_neg.ksh \
	functional/channel_program/synctask_core/tst.get_number_props.ksh \
	functional/channel_program/synctask_core/tst.get_string_props.ksh \
	functional/channel_program/synctask_core/tst.get_type.ksh \
	functional/channel_program/synctask_core/tst.get_userquota.ksh \
	functional/channel_program/synctask_core/tst.get_written.ksh \
	functional/channel_program/synctask_core/tst.inherit.ksh \
	functional/channel_program/synctask_core/tst.list_bookmarks.ksh \
	functional/channel_program/synctask_core/tst.list_children.ksh \
	functional/channel_program/synctask_core/tst.list_clones.ksh \
	functional/channel_program/synctask_core/tst.list_holds.ksh \
	functional/channel_program/synctask_core/tst.list_snapshots.ksh \
	functional/channel_program/synctask_core/tst.list_system_props.ksh \
	functional/channel_program/synctask_core/tst.list_user_props.ksh \
	functional/channel_program/synctask_core/tst.parse_args_neg.ksh \
	functional/channel_program/synctask_core/tst.promote_conflict.ksh \
	functional/channel_program/synctask_core/tst.promote_multiple.ksh \
	functional/channel_program/synctask_core/tst.promote_simple.ksh \
	functional/channel_program/synctask_core/tst.rollback_mult.ksh \
	functional/channel_program/synctask_core/tst.rollback_one.ksh \
	functional/channel_program/synctask_core/tst.set_props.ksh \
	functional/channel_program/synctask_core/tst.snapshot_destroy.ksh \
	functional/channel_program/synctask_core/tst.snapshot_neg.ksh \
	functional/channel_program/synctask_core/tst.snapshot_recursive.ksh \
	functional/channel_program/synctask_core/tst.snapshot_rename.ksh \
	functional/channel_program/synctask_core/tst.snapshot_simple.ksh \
	functional/channel_program/synctask_core/tst.terminate_by_signal.ksh \
	functional/chattr/chattr_001_pos.ksh \
	functional/chattr/chattr_002_neg.ksh \
	functional/chattr/cleanup.ksh \
	functional/chattr/setup.ksh \
	functional/checksum/cleanup.ksh \
	functional/checksum/filetest_001_pos.ksh \
	functional/checksum/filetest_002_pos.ksh \
	functional/checksum/run_blake3_test.ksh \
	functional/checksum/run_edonr_test.ksh \
	functional/checksum/run_sha2_test.ksh \
	functional/checksum/run_skein_test.ksh \
	functional/checksum/setup.ksh \
	functional/clean_mirror/clean_mirror_001_pos.ksh \
	functional/clean_mirror/clean_mirror_002_pos.ksh \
	functional/clean_mirror/clean_mirror_003_pos.ksh \
	functional/clean_mirror/clean_mirror_004_pos.ksh \
	functional/clean_mirror/cleanup.ksh \
	functional/clean_mirror/setup.ksh \
	functional/cli_root/zinject/zinject_args.ksh \
	functional/cli_root/zdb/zdb_002_pos.ksh \
	functional/cli_root/zdb/zdb_003_pos.ksh \
	functional/cli_root/zdb/zdb_004_pos.ksh \
	functional/cli_root/zdb/zdb_005_pos.ksh \
	functional/cli_root/zdb/zdb_006_pos.ksh \
	functional/cli_root/zdb/zdb_args_neg.ksh \
	functional/cli_root/zdb/zdb_args_pos.ksh \
	functional/cli_root/zdb/zdb_backup.ksh \
	functional/cli_root/zdb/zdb_block_size_histogram.ksh \
	functional/cli_root/zdb/zdb_checksum.ksh \
	functional/cli_root/zdb/zdb_decompress.ksh \
	functional/cli_root/zdb/zdb_decompress_zstd.ksh \
	functional/cli_root/zdb/zdb_display_block.ksh \
	functional/cli_root/zdb/zdb_encrypted.ksh \
	functional/cli_root/zdb/zdb_label_checksum.ksh \
	functional/cli_root/zdb/zdb_object_range_neg.ksh \
	functional/cli_root/zdb/zdb_object_range_pos.ksh \
	functional/cli_root/zdb/zdb_objset_id.ksh \
	functional/cli_root/zdb/zdb_recover_2.ksh \
	functional/cli_root/zdb/zdb_recover.ksh \
	functional/cli_root/zfs_bookmark/cleanup.ksh \
	functional/cli_root/zfs_bookmark/setup.ksh \
	functional/cli_root/zfs_bookmark/zfs_bookmark_cliargs.ksh \
	functional/cli_root/zfs_change-key/cleanup.ksh \
	functional/cli_root/zfs_change-key/setup.ksh \
	functional/cli_root/zfs_change-key/zfs_change-key_child.ksh \
	functional/cli_root/zfs_change-key/zfs_change-key_clones.ksh \
	functional/cli_root/zfs_change-key/zfs_change-key_format.ksh \
	functional/cli_root/zfs_change-key/zfs_change-key_inherit.ksh \
	functional/cli_root/zfs_change-key/zfs_change-key.ksh \
	functional/cli_root/zfs_change-key/zfs_change-key_load.ksh \
	functional/cli_root/zfs_change-key/zfs_change-key_location.ksh \
	functional/cli_root/zfs_change-key/zfs_change-key_pbkdf2iters.ksh \
	functional/cli_root/zfs/cleanup.ksh \
	functional/cli_root/zfs_clone/cleanup.ksh \
	functional/cli_root/zfs_clone/setup.ksh \
	functional/cli_root/zfs_clone/zfs_clone_001_neg.ksh \
	functional/cli_root/zfs_clone/zfs_clone_002_pos.ksh \
	functional/cli_root/zfs_clone/zfs_clone_003_pos.ksh \
	functional/cli_root/zfs_clone/zfs_clone_004_pos.ksh \
	functional/cli_root/zfs_clone/zfs_clone_005_pos.ksh \
	functional/cli_root/zfs_clone/zfs_clone_006_pos.ksh \
	functional/cli_root/zfs_clone/zfs_clone_007_pos.ksh \
	functional/cli_root/zfs_clone/zfs_clone_008_neg.ksh \
	functional/cli_root/zfs_clone/zfs_clone_009_neg.ksh \
	functional/cli_root/zfs_clone/zfs_clone_010_pos.ksh \
	functional/cli_root/zfs_clone/zfs_clone_deeply_nested.ksh \
	functional/cli_root/zfs_clone/zfs_clone_encrypted.ksh \
	functional/cli_root/zfs_clone/zfs_clone_rm_nested.ksh \
	functional/cli_root/zfs_copies/cleanup.ksh \
	functional/cli_root/zfs_copies/setup.ksh \
	functional/cli_root/zfs_copies/zfs_copies_001_pos.ksh \
	functional/cli_root/zfs_copies/zfs_copies_002_pos.ksh \
	functional/cli_root/zfs_copies/zfs_copies_003_pos.ksh \
	functional/cli_root/zfs_copies/zfs_copies_004_neg.ksh \
	functional/cli_root/zfs_copies/zfs_copies_005_neg.ksh \
	functional/cli_root/zfs_copies/zfs_copies_006_pos.ksh \
	functional/cli_root/zfs_create/cleanup.ksh \
	functional/cli_root/zfs_create/setup.ksh \
	functional/cli_root/zfs_create/zfs_create_001_pos.ksh \
	functional/cli_root/zfs_create/zfs_create_002_pos.ksh \
	functional/cli_root/zfs_create/zfs_create_003_pos.ksh \
	functional/cli_root/zfs_create/zfs_create_004_pos.ksh \
	functional/cli_root/zfs_create/zfs_create_005_pos.ksh \
	functional/cli_root/zfs_create/zfs_create_006_pos.ksh \
	functional/cli_root/zfs_create/zfs_create_007_pos.ksh \
	functional/cli_root/zfs_create/zfs_create_008_neg.ksh \
	functional/cli_root/zfs_create/zfs_create_009_neg.ksh \
	functional/cli_root/zfs_create/zfs_create_010_neg.ksh \
	functional/cli_root/zfs_create/zfs_create_011_pos.ksh \
	functional/cli_root/zfs_create/zfs_create_012_pos.ksh \
	functional/cli_root/zfs_create/zfs_create_013_pos.ksh \
	functional/cli_root/zfs_create/zfs_create_014_pos.ksh \
	functional/cli_root/zfs_create/zfs_create_crypt_combos.ksh \
	functional/cli_root/zfs_create/zfs_create_dryrun.ksh \
	functional/cli_root/zfs_create/zfs_create_encrypted.ksh \
	functional/cli_root/zfs_create/zfs_create_nomount.ksh \
	functional/cli_root/zfs_create/zfs_create_verbose.ksh \
	functional/cli_root/zfs_destroy/cleanup.ksh \
	functional/cli_root/zfs_destroy/setup.ksh \
	functional/cli_root/zfs_destroy/zfs_clone_livelist_condense_and_disable.ksh \
	functional/cli_root/zfs_destroy/zfs_clone_livelist_condense_races.ksh \
	functional/cli_root/zfs_destroy/zfs_clone_livelist_dedup.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_001_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_002_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_003_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_004_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_005_neg.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_006_neg.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_007_neg.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_008_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_009_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_010_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_011_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_012_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_013_neg.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_014_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_015_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_016_pos.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_clone_livelist.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_dev_removal_condense.ksh \
	functional/cli_root/zfs_destroy/zfs_destroy_dev_removal.ksh \
	functional/cli_root/zfs_diff/cleanup.ksh \
	functional/cli_root/zfs_diff/setup.ksh \
	functional/cli_root/zfs_diff/zfs_diff_changes.ksh \
	functional/cli_root/zfs_diff/zfs_diff_cliargs.ksh \
	functional/cli_root/zfs_diff/zfs_diff_encrypted.ksh \
	functional/cli_root/zfs_diff/zfs_diff_mangle.ksh \
	functional/cli_root/zfs_diff/zfs_diff_timestamp.ksh \
	functional/cli_root/zfs_diff/zfs_diff_types.ksh \
	functional/cli_root/zfs_get/cleanup.ksh \
	functional/cli_root/zfs_get/setup.ksh \
	functional/cli_root/zfs_get/zfs_get_001_pos.ksh \
	functional/cli_root/zfs_get/zfs_get_002_pos.ksh \
	functional/cli_root/zfs_get/zfs_get_003_pos.ksh \
	functional/cli_root/zfs_get/zfs_get_004_pos.ksh \
	functional/cli_root/zfs_get/zfs_get_005_neg.ksh \
	functional/cli_root/zfs_get/zfs_get_006_neg.ksh \
	functional/cli_root/zfs_get/zfs_get_007_neg.ksh \
	functional/cli_root/zfs_get/zfs_get_008_pos.ksh \
	functional/cli_root/zfs_get/zfs_get_009_pos.ksh \
	functional/cli_root/zfs_get/zfs_get_010_neg.ksh \
	functional/cli_root/zfs_ids_to_path/cleanup.ksh \
	functional/cli_root/zfs_ids_to_path/setup.ksh \
	functional/cli_root/zfs_ids_to_path/zfs_ids_to_path_001_pos.ksh \
	functional/cli_root/zfs_inherit/cleanup.ksh \
	functional/cli_root/zfs_inherit/setup.ksh \
	functional/cli_root/zfs_inherit/zfs_inherit_001_neg.ksh \
	functional/cli_root/zfs_inherit/zfs_inherit_002_neg.ksh \
	functional/cli_root/zfs_inherit/zfs_inherit_003_pos.ksh \
	functional/cli_root/zfs_inherit/zfs_inherit_mountpoint.ksh \
	functional/cli_root/zfs_jail/cleanup.ksh \
	functional/cli_root/zfs_jail/setup.ksh \
	functional/cli_root/zfs_jail/zfs_jail_001_pos.ksh \
	functional/cli_root/zfs_load-key/cleanup.ksh \
	functional/cli_root/zfs_load-key/setup.ksh \
	functional/cli_root/zfs_load-key/zfs_load-key_all.ksh \
	functional/cli_root/zfs_load-key/zfs_load-key_file.ksh \
	functional/cli_root/zfs_load-key/zfs_load-key_https.ksh \
	functional/cli_root/zfs_load-key/zfs_load-key.ksh \
	functional/cli_root/zfs_load-key/zfs_load-key_location.ksh \
	functional/cli_root/zfs_load-key/zfs_load-key_noop.ksh \
	functional/cli_root/zfs_load-key/zfs_load-key_recursive.ksh \
	functional/cli_root/zfs_mount/cleanup.ksh \
	functional/cli_root/zfs_mount/setup.ksh \
	functional/cli_root/zfs_mount/zfs_mount_001_pos.ksh \
	functional/cli_root/zfs_mount/zfs_mount_002_pos.ksh \
	functional/cli_root/zfs_mount/zfs_mount_003_pos.ksh \
	functional/cli_root/zfs_mount/zfs_mount_004_pos.ksh \
	functional/cli_root/zfs_mount/zfs_mount_005_pos.ksh \
	functional/cli_root/zfs_mount/zfs_mount_006_pos.ksh \
	functional/cli_root/zfs_mount/zfs_mount_007_pos.ksh \
	functional/cli_root/zfs_mount/zfs_mount_008_pos.ksh \
	functional/cli_root/zfs_mount/zfs_mount_009_neg.ksh \
	functional/cli_root/zfs_mount/zfs_mount_010_neg.ksh \
	functional/cli_root/zfs_mount/zfs_mount_011_neg.ksh \
	functional/cli_root/zfs_mount/zfs_mount_012_pos.ksh \
	functional/cli_root/zfs_mount/zfs_mount_013_pos.ksh \
	functional/cli_root/zfs_mount/zfs_mount_014_neg.ksh \
	functional/cli_root/zfs_mount/zfs_mount_all_001_pos.ksh \
	functional/cli_root/zfs_mount/zfs_mount_all_fail.ksh \
	functional/cli_root/zfs_mount/zfs_mount_all_mountpoints.ksh \
	functional/cli_root/zfs_mount/zfs_mount_encrypted.ksh \
	functional/cli_root/zfs_mount/zfs_mount_recursive.ksh \
	functional/cli_root/zfs_mount/zfs_mount_remount.ksh \
	functional/cli_root/zfs_mount/zfs_mount_test_race.ksh \
	functional/cli_root/zfs_mount/zfs_multi_mount.ksh \
	functional/cli_root/zfs_program/cleanup.ksh \
	functional/cli_root/zfs_program/setup.ksh \
	functional/cli_root/zfs_program/zfs_program_json.ksh \
	functional/cli_root/zfs_promote/cleanup.ksh \
	functional/cli_root/zfs_promote/setup.ksh \
	functional/cli_root/zfs_promote/zfs_promote_001_pos.ksh \
	functional/cli_root/zfs_promote/zfs_promote_002_pos.ksh \
	functional/cli_root/zfs_promote/zfs_promote_003_pos.ksh \
	functional/cli_root/zfs_promote/zfs_promote_004_pos.ksh \
	functional/cli_root/zfs_promote/zfs_promote_005_pos.ksh \
	functional/cli_root/zfs_promote/zfs_promote_006_neg.ksh \
	functional/cli_root/zfs_promote/zfs_promote_007_neg.ksh \
	functional/cli_root/zfs_promote/zfs_promote_008_pos.ksh \
	functional/cli_root/zfs_promote/zfs_promote_encryptionroot.ksh \
	functional/cli_root/zfs_property/cleanup.ksh \
	functional/cli_root/zfs_property/setup.ksh \
	functional/cli_root/zfs_property/zfs_written_property_001_pos.ksh \
	functional/cli_root/zfs_receive/cleanup.ksh \
	functional/cli_root/zfs_receive/receive-o-x_props_aliases.ksh \
	functional/cli_root/zfs_receive/receive-o-x_props_override.ksh \
	functional/cli_root/zfs_receive/setup.ksh \
	functional/cli_root/zfs_receive/zfs_receive_001_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_002_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_003_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_004_neg.ksh \
	functional/cli_root/zfs_receive/zfs_receive_005_neg.ksh \
	functional/cli_root/zfs_receive/zfs_receive_006_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_007_neg.ksh \
	functional/cli_root/zfs_receive/zfs_receive_008_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_009_neg.ksh \
	functional/cli_root/zfs_receive/zfs_receive_010_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_011_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_012_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_013_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_014_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_015_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_016_pos.ksh \
	functional/cli_root/zfs_receive/zfs_receive_-e.ksh \
	functional/cli_root/zfs_receive/zfs_receive_from_encrypted.ksh \
	functional/cli_root/zfs_receive/zfs_receive_from_zstd.ksh \
	functional/cli_root/zfs_receive/zfs_receive_new_props.ksh \
	functional/cli_root/zfs_receive/zfs_receive_raw_-d.ksh \
	functional/cli_root/zfs_receive/zfs_receive_raw_incremental.ksh \
	functional/cli_root/zfs_receive/zfs_receive_raw.ksh \
	functional/cli_root/zfs_receive/zfs_receive_to_encrypted.ksh \
	functional/cli_root/zfs_receive/zfs_receive_-wR-encrypted-mix.ksh \
	functional/cli_root/zfs_receive/zfs_receive_corrective.ksh \
	functional/cli_root/zfs_receive/zfs_receive_compressed_corrective.ksh \
	functional/cli_root/zfs_receive/zfs_receive_large_block_corrective.ksh \
	functional/cli_root/zfs_rename/cleanup.ksh \
	functional/cli_root/zfs_rename/setup.ksh \
	functional/cli_root/zfs_rename/zfs_rename_001_pos.ksh \
	functional/cli_root/zfs_rename/zfs_rename_002_pos.ksh \
	functional/cli_root/zfs_rename/zfs_rename_003_pos.ksh \
	functional/cli_root/zfs_rename/zfs_rename_004_neg.ksh \
	functional/cli_root/zfs_rename/zfs_rename_005_neg.ksh \
	functional/cli_root/zfs_rename/zfs_rename_006_pos.ksh \
	functional/cli_root/zfs_rename/zfs_rename_007_pos.ksh \
	functional/cli_root/zfs_rename/zfs_rename_008_pos.ksh \
	functional/cli_root/zfs_rename/zfs_rename_009_neg.ksh \
	functional/cli_root/zfs_rename/zfs_rename_010_neg.ksh \
	functional/cli_root/zfs_rename/zfs_rename_011_pos.ksh \
	functional/cli_root/zfs_rename/zfs_rename_012_neg.ksh \
	functional/cli_root/zfs_rename/zfs_rename_013_pos.ksh \
	functional/cli_root/zfs_rename/zfs_rename_014_neg.ksh \
	functional/cli_root/zfs_rename/zfs_rename_encrypted_child.ksh \
	functional/cli_root/zfs_rename/zfs_rename_mountpoint.ksh \
	functional/cli_root/zfs_rename/zfs_rename_nounmount.ksh \
	functional/cli_root/zfs_rename/zfs_rename_to_encrypted.ksh \
	functional/cli_root/zfs_reservation/cleanup.ksh \
	functional/cli_root/zfs_reservation/setup.ksh \
	functional/cli_root/zfs_reservation/zfs_reservation_001_pos.ksh \
	functional/cli_root/zfs_reservation/zfs_reservation_002_pos.ksh \
	functional/cli_root/zfs_rollback/cleanup.ksh \
	functional/cli_root/zfs_rollback/setup.ksh \
	functional/cli_root/zfs_rollback/zfs_rollback_001_pos.ksh \
	functional/cli_root/zfs_rollback/zfs_rollback_002_pos.ksh \
	functional/cli_root/zfs_rollback/zfs_rollback_003_neg.ksh \
	functional/cli_root/zfs_rollback/zfs_rollback_004_neg.ksh \
	functional/cli_root/zfs_send/cleanup.ksh \
	functional/cli_root/zfs_send/setup.ksh \
	functional/cli_root/zfs_send/zfs_send_001_pos.ksh \
	functional/cli_root/zfs_send/zfs_send_002_pos.ksh \
	functional/cli_root/zfs_send/zfs_send_003_pos.ksh \
	functional/cli_root/zfs_send/zfs_send_004_neg.ksh \
	functional/cli_root/zfs_send/zfs_send_005_pos.ksh \
	functional/cli_root/zfs_send/zfs_send_006_pos.ksh \
	functional/cli_root/zfs_send/zfs_send_007_pos.ksh \
	functional/cli_root/zfs_send/zfs_send-b.ksh \
	functional/cli_root/zfs_send/zfs_send_encrypted.ksh \
	functional/cli_root/zfs_send/zfs_send_encrypted_unloaded.ksh \
	functional/cli_root/zfs_send/zfs_send_raw.ksh \
	functional/cli_root/zfs_send/zfs_send_skip_missing.ksh \
	functional/cli_root/zfs_send/zfs_send_sparse.ksh \
	functional/cli_root/zfs_set/cache_001_pos.ksh \
	functional/cli_root/zfs_set/cache_002_neg.ksh \
	functional/cli_root/zfs_set/canmount_001_pos.ksh \
	functional/cli_root/zfs_set/canmount_002_pos.ksh \
	functional/cli_root/zfs_set/canmount_003_pos.ksh \
	functional/cli_root/zfs_set/canmount_004_pos.ksh \
	functional/cli_root/zfs_set/checksum_001_pos.ksh \
	functional/cli_root/zfs_set/cleanup.ksh \
	functional/cli_root/zfs_set/compression_001_pos.ksh \
	functional/cli_root/zfs_set/mountpoint_001_pos.ksh \
	functional/cli_root/zfs_set/mountpoint_002_pos.ksh \
	functional/cli_root/zfs_set/mountpoint_003_pos.ksh \
	functional/cli_root/zfs_set/onoffs_001_pos.ksh \
	functional/cli_root/zfs_set/property_alias_001_pos.ksh \
	functional/cli_root/zfs_set/readonly_001_pos.ksh \
	functional/cli_root/zfs_set/reservation_001_neg.ksh \
	functional/cli_root/zfs_set/ro_props_001_pos.ksh \
	functional/cli_root/zfs_set/setup.ksh \
	functional/cli_root/zfs_set/share_mount_001_neg.ksh \
	functional/cli_root/zfs_set/snapdir_001_pos.ksh \
	functional/cli_root/zfs/setup.ksh \
	functional/cli_root/zfs_set/user_property_001_pos.ksh \
	functional/cli_root/zfs_set/user_property_002_pos.ksh \
	functional/cli_root/zfs_set/user_property_003_neg.ksh \
	functional/cli_root/zfs_set/user_property_004_pos.ksh \
	functional/cli_root/zfs_set/version_001_neg.ksh \
	functional/cli_root/zfs_set/zfs_set_001_neg.ksh \
	functional/cli_root/zfs_set/zfs_set_002_neg.ksh \
	functional/cli_root/zfs_set/zfs_set_003_neg.ksh \
	functional/cli_root/zfs_set/zfs_set_feature_activation.ksh \
	functional/cli_root/zfs_set/zfs_set_keylocation.ksh \
	functional/cli_root/zfs_set/zfs_set_nomount.ksh \
	functional/cli_root/zfs_share/cleanup.ksh \
	functional/cli_root/zfs_share/setup.ksh \
	functional/cli_root/zfs_share/zfs_share_001_pos.ksh \
	functional/cli_root/zfs_share/zfs_share_002_pos.ksh \
	functional/cli_root/zfs_share/zfs_share_003_pos.ksh \
	functional/cli_root/zfs_share/zfs_share_004_pos.ksh \
	functional/cli_root/zfs_share/zfs_share_005_pos.ksh \
	functional/cli_root/zfs_share/zfs_share_006_pos.ksh \
	functional/cli_root/zfs_share/zfs_share_007_neg.ksh \
	functional/cli_root/zfs_share/zfs_share_008_neg.ksh \
	functional/cli_root/zfs_share/zfs_share_009_neg.ksh \
	functional/cli_root/zfs_share/zfs_share_010_neg.ksh \
	functional/cli_root/zfs_share/zfs_share_011_pos.ksh \
	functional/cli_root/zfs_share/zfs_share_012_pos.ksh \
	functional/cli_root/zfs_share/zfs_share_013_pos.ksh \
	functional/cli_root/zfs_share/zfs_share_concurrent_shares.ksh \
	functional/cli_root/zfs_share/zfs_share_after_mount.ksh \
	functional/cli_root/zfs_snapshot/cleanup.ksh \
	functional/cli_root/zfs_snapshot/setup.ksh \
	functional/cli_root/zfs_snapshot/zfs_snapshot_001_neg.ksh \
	functional/cli_root/zfs_snapshot/zfs_snapshot_002_neg.ksh \
	functional/cli_root/zfs_snapshot/zfs_snapshot_003_neg.ksh \
	functional/cli_root/zfs_snapshot/zfs_snapshot_004_neg.ksh \
	functional/cli_root/zfs_snapshot/zfs_snapshot_005_neg.ksh \
	functional/cli_root/zfs_snapshot/zfs_snapshot_006_pos.ksh \
	functional/cli_root/zfs_snapshot/zfs_snapshot_007_neg.ksh \
	functional/cli_root/zfs_snapshot/zfs_snapshot_008_neg.ksh \
	functional/cli_root/zfs_snapshot/zfs_snapshot_009_pos.ksh \
	functional/cli_root/zfs_sysfs/cleanup.ksh \
	functional/cli_root/zfs_sysfs/setup.ksh \
	functional/cli_root/zfs_sysfs/zfeature_set_unsupported.ksh \
	functional/cli_root/zfs_sysfs/zfs_get_unsupported.ksh \
	functional/cli_root/zfs_sysfs/zfs_set_unsupported.ksh \
	functional/cli_root/zfs_sysfs/zfs_sysfs_live.ksh \
	functional/cli_root/zfs_sysfs/zpool_get_unsupported.ksh \
	functional/cli_root/zfs_sysfs/zpool_set_unsupported.ksh \
	functional/cli_root/zfs_unload-key/cleanup.ksh \
	functional/cli_root/zfs_unload-key/setup.ksh \
	functional/cli_root/zfs_unload-key/zfs_unload-key_all.ksh \
	functional/cli_root/zfs_unload-key/zfs_unload-key.ksh \
	functional/cli_root/zfs_unload-key/zfs_unload-key_recursive.ksh \
	functional/cli_root/zfs_unmount/cleanup.ksh \
	functional/cli_root/zfs_unmount/setup.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_001_pos.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_002_pos.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_003_pos.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_004_pos.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_005_pos.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_006_pos.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_007_neg.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_008_neg.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_009_pos.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_all_001_pos.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_nested.ksh \
	functional/cli_root/zfs_unmount/zfs_unmount_unload_keys.ksh \
	functional/cli_root/zfs_unshare/cleanup.ksh \
	functional/cli_root/zfs_unshare/setup.ksh \
	functional/cli_root/zfs_unshare/zfs_unshare_001_pos.ksh \
	functional/cli_root/zfs_unshare/zfs_unshare_002_pos.ksh \
	functional/cli_root/zfs_unshare/zfs_unshare_003_pos.ksh \
	functional/cli_root/zfs_unshare/zfs_unshare_004_neg.ksh \
	functional/cli_root/zfs_unshare/zfs_unshare_005_neg.ksh \
	functional/cli_root/zfs_unshare/zfs_unshare_006_pos.ksh \
	functional/cli_root/zfs_unshare/zfs_unshare_007_pos.ksh \
	functional/cli_root/zfs_unshare/zfs_unshare_008_pos.ksh \
	functional/cli_root/zfs_upgrade/cleanup.ksh \
	functional/cli_root/zfs_upgrade/setup.ksh \
	functional/cli_root/zfs_upgrade/zfs_upgrade_001_pos.ksh \
	functional/cli_root/zfs_upgrade/zfs_upgrade_002_pos.ksh \
	functional/cli_root/zfs_upgrade/zfs_upgrade_003_pos.ksh \
	functional/cli_root/zfs_upgrade/zfs_upgrade_004_pos.ksh \
	functional/cli_root/zfs_upgrade/zfs_upgrade_005_pos.ksh \
	functional/cli_root/zfs_upgrade/zfs_upgrade_006_neg.ksh \
	functional/cli_root/zfs_upgrade/zfs_upgrade_007_neg.ksh \
	functional/cli_root/zfs_wait/cleanup.ksh \
	functional/cli_root/zfs_wait/setup.ksh \
	functional/cli_root/zfs_wait/zfs_wait_deleteq.ksh \
	functional/cli_root/zfs_wait/zfs_wait_getsubopt.ksh \
	functional/cli_root/zfs/zfs_001_neg.ksh \
	functional/cli_root/zfs/zfs_002_pos.ksh \
	functional/cli_root/zfs/zfs_003_neg.ksh \
	functional/cli_root/zhack/zhack_label_repair_001.ksh \
	functional/cli_root/zhack/zhack_label_repair_002.ksh \
	functional/cli_root/zhack/zhack_label_repair_003.ksh \
	functional/cli_root/zhack/zhack_label_repair_004.ksh \
	functional/cli_root/zpool_add/add_nested_replacing_spare.ksh \
	functional/cli_root/zpool_add/add-o_ashift.ksh \
	functional/cli_root/zpool_add/add_prop_ashift.ksh \
	functional/cli_root/zpool_add/cleanup.ksh \
	functional/cli_root/zpool_add/setup.ksh \
	functional/cli_root/zpool_add/zpool_add--allow-ashift-mismatch.ksh \
	functional/cli_root/zpool_add/zpool_add_001_pos.ksh \
	functional/cli_root/zpool_add/zpool_add_002_pos.ksh \
	functional/cli_root/zpool_add/zpool_add_003_pos.ksh \
	functional/cli_root/zpool_add/zpool_add_004_pos.ksh \
	functional/cli_root/zpool_add/zpool_add_005_pos.ksh \
	functional/cli_root/zpool_add/zpool_add_006_pos.ksh \
	functional/cli_root/zpool_add/zpool_add_007_neg.ksh \
	functional/cli_root/zpool_add/zpool_add_008_neg.ksh \
	functional/cli_root/zpool_add/zpool_add_009_neg.ksh \
	functional/cli_root/zpool_add/zpool_add_010_pos.ksh \
	functional/cli_root/zpool_add/zpool_add_dryrun_output.ksh \
	functional/cli_root/zpool_attach/attach-o_ashift.ksh \
	functional/cli_root/zpool_attach/cleanup.ksh \
	functional/cli_root/zpool_attach/setup.ksh \
	functional/cli_root/zpool_attach/zpool_attach_001_neg.ksh \
	functional/cli_root/zpool/cleanup.ksh \
	functional/cli_root/zpool_clear/cleanup.ksh \
	functional/cli_root/zpool_clear/setup.ksh \
	functional/cli_root/zpool_clear/zpool_clear_001_pos.ksh \
	functional/cli_root/zpool_clear/zpool_clear_002_neg.ksh \
	functional/cli_root/zpool_clear/zpool_clear_003_neg.ksh \
	functional/cli_root/zpool_clear/zpool_clear_readonly.ksh \
	functional/cli_root/zpool_create/cleanup.ksh \
	functional/cli_root/zpool_create/create-o_ashift.ksh \
	functional/cli_root/zpool_create/setup.ksh \
	functional/cli_root/zpool_create/zpool_create_001_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_002_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_003_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_004_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_005_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_006_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_007_neg.ksh \
	functional/cli_root/zpool_create/zpool_create_008_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_009_neg.ksh \
	functional/cli_root/zpool_create/zpool_create_010_neg.ksh \
	functional/cli_root/zpool_create/zpool_create_011_neg.ksh \
	functional/cli_root/zpool_create/zpool_create_012_neg.ksh \
	functional/cli_root/zpool_create/zpool_create_014_neg.ksh \
	functional/cli_root/zpool_create/zpool_create_015_neg.ksh \
	functional/cli_root/zpool_create/zpool_create_016_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_017_neg.ksh \
	functional/cli_root/zpool_create/zpool_create_018_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_019_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_020_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_021_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_022_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_023_neg.ksh \
	functional/cli_root/zpool_create/zpool_create_024_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_crypt_combos.ksh \
	functional/cli_root/zpool_create/zpool_create_draid_001_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_draid_002_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_draid_003_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_draid_004_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_dryrun_output.ksh \
	functional/cli_root/zpool_create/zpool_create_encrypted.ksh \
	functional/cli_root/zpool_create/zpool_create_features_001_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_features_002_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_features_003_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_features_004_neg.ksh \
	functional/cli_root/zpool_create/zpool_create_features_005_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_features_006_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_features_007_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_features_008_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_features_009_pos.ksh \
	functional/cli_root/zpool_create/zpool_create_tempname.ksh \
	functional/cli_root/zpool_destroy/zpool_destroy_001_pos.ksh \
	functional/cli_root/zpool_destroy/zpool_destroy_002_pos.ksh \
	functional/cli_root/zpool_destroy/zpool_destroy_003_neg.ksh \
	functional/cli_root/zpool_detach/cleanup.ksh \
	functional/cli_root/zpool_detach/setup.ksh \
	functional/cli_root/zpool_detach/zpool_detach_001_neg.ksh \
	functional/cli_root/zpool_events/cleanup.ksh \
	functional/cli_root/zpool_events/setup.ksh \
	functional/cli_root/zpool_events/zpool_events_clear.ksh \
	functional/cli_root/zpool_events/zpool_events_clear_retained.ksh \
	functional/cli_root/zpool_events/zpool_events_cliargs.ksh \
	functional/cli_root/zpool_events/zpool_events_duplicates.ksh \
	functional/cli_root/zpool_events/zpool_events_errors.ksh \
	functional/cli_root/zpool_events/zpool_events_follow.ksh \
	functional/cli_root/zpool_events/zpool_events_poolname.ksh \
	functional/cli_root/zpool_expand/cleanup.ksh \
	functional/cli_root/zpool_expand/setup.ksh \
	functional/cli_root/zpool_expand/zpool_expand_001_pos.ksh \
	functional/cli_root/zpool_expand/zpool_expand_002_pos.ksh \
	functional/cli_root/zpool_expand/zpool_expand_003_neg.ksh \
	functional/cli_root/zpool_expand/zpool_expand_004_pos.ksh \
	functional/cli_root/zpool_expand/zpool_expand_005_pos.ksh \
	functional/cli_root/zpool_export/cleanup.ksh \
	functional/cli_root/zpool_export/setup.ksh \
	functional/cli_root/zpool_export/zpool_export_001_pos.ksh \
	functional/cli_root/zpool_export/zpool_export_002_pos.ksh \
	functional/cli_root/zpool_export/zpool_export_003_neg.ksh \
	functional/cli_root/zpool_export/zpool_export_004_pos.ksh \
	+ functional/cli_root/zpool_export/zpool_export_parallel_admin.ksh \
	+ functional/cli_root/zpool_export/zpool_export_parallel_pos.ksh \
	functional/cli_root/zpool_get/cleanup.ksh \
	functional/cli_root/zpool_get/setup.ksh \
	functional/cli_root/zpool_get/vdev_get_001_pos.ksh \
	functional/cli_root/zpool_get/zpool_get_001_pos.ksh \
	functional/cli_root/zpool_get/zpool_get_002_pos.ksh \
	functional/cli_root/zpool_get/zpool_get_003_pos.ksh \
	functional/cli_root/zpool_get/zpool_get_004_neg.ksh \
	functional/cli_root/zpool_get/zpool_get_005_pos.ksh \
	functional/cli_root/zpool_history/cleanup.ksh \
	functional/cli_root/zpool_history/setup.ksh \
	functional/cli_root/zpool_history/zpool_history_001_neg.ksh \
	functional/cli_root/zpool_history/zpool_history_002_pos.ksh \
	functional/cli_root/zpool_import/cleanup.ksh \
	functional/cli_root/zpool_import/import_cachefile_device_added.ksh \
	functional/cli_root/zpool_import/import_cachefile_device_removed.ksh \
	functional/cli_root/zpool_import/import_cachefile_device_replaced.ksh \
	functional/cli_root/zpool_import/import_cachefile_mirror_attached.ksh \
	functional/cli_root/zpool_import/import_cachefile_mirror_detached.ksh \
	functional/cli_root/zpool_import/import_cachefile_paths_changed.ksh \
	functional/cli_root/zpool_import/import_cachefile_shared_device.ksh \
	functional/cli_root/zpool_import/import_devices_missing.ksh \
	functional/cli_root/zpool_import/import_log_missing.ksh \
	functional/cli_root/zpool_import/import_paths_changed.ksh \
	functional/cli_root/zpool_import/import_rewind_config_changed.ksh \
	functional/cli_root/zpool_import/import_rewind_device_replaced.ksh \
	functional/cli_root/zpool_import/setup.ksh \
	functional/cli_root/zpool_import/zpool_import_001_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_002_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_003_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_004_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_005_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_006_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_007_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_008_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_009_neg.ksh \
	functional/cli_root/zpool_import/zpool_import_010_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_011_neg.ksh \
	functional/cli_root/zpool_import/zpool_import_012_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_013_neg.ksh \
	functional/cli_root/zpool_import/zpool_import_014_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_015_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_016_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_017_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_all_001_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_encrypted.ksh \
	functional/cli_root/zpool_import/zpool_import_encrypted_load.ksh \
	functional/cli_root/zpool_import/zpool_import_errata3.ksh \
	functional/cli_root/zpool_import/zpool_import_errata4.ksh \
	functional/cli_root/zpool_import/zpool_import_features_001_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_features_002_neg.ksh \
	functional/cli_root/zpool_import/zpool_import_features_003_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_hostid_changed.ksh \
	functional/cli_root/zpool_import/zpool_import_hostid_changed_unclean_export.ksh \
	functional/cli_root/zpool_import/zpool_import_hostid_changed_cachefile.ksh \
	functional/cli_root/zpool_import/zpool_import_hostid_changed_cachefile_unclean_export.ksh \
	functional/cli_root/zpool_import/zpool_import_missing_001_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_missing_002_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_missing_003_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_rename_001_pos.ksh \
	functional/cli_root/zpool_import/zpool_import_status.ksh \
	functional/cli_root/zpool_import/zpool_import_parallel_admin.ksh \
	functional/cli_root/zpool_import/zpool_import_parallel_neg.ksh \
	functional/cli_root/zpool_import/zpool_import_parallel_pos.ksh \
	functional/cli_root/zpool_initialize/cleanup.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_attach_detach_add_remove.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_fault_export_import_online.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_import_export.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_offline_export_import_online.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_online_offline.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_split.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_start_and_cancel_neg.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_start_and_cancel_pos.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_suspend_resume.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_uninit.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_unsupported_vdevs.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_verify_checksums.ksh \
	functional/cli_root/zpool_initialize/zpool_initialize_verify_initialized.ksh \
	functional/cli_root/zpool_labelclear/zpool_labelclear_active.ksh \
	functional/cli_root/zpool_labelclear/zpool_labelclear_exported.ksh \
	functional/cli_root/zpool_labelclear/zpool_labelclear_removed.ksh \
	functional/cli_root/zpool_labelclear/zpool_labelclear_valid.ksh \
	functional/cli_root/zpool_offline/cleanup.ksh \
	functional/cli_root/zpool_offline/setup.ksh \
	functional/cli_root/zpool_offline/zpool_offline_001_pos.ksh \
	functional/cli_root/zpool_offline/zpool_offline_002_neg.ksh \
	functional/cli_root/zpool_offline/zpool_offline_003_pos.ksh \
	functional/cli_root/zpool_online/cleanup.ksh \
	functional/cli_root/zpool_online/setup.ksh \
	functional/cli_root/zpool_online/zpool_online_001_pos.ksh \
	functional/cli_root/zpool_online/zpool_online_002_neg.ksh \
	functional/cli_root/zpool_remove/cleanup.ksh \
	functional/cli_root/zpool_remove/setup.ksh \
	functional/cli_root/zpool_remove/zpool_remove_001_neg.ksh \
	functional/cli_root/zpool_remove/zpool_remove_002_pos.ksh \
	functional/cli_root/zpool_remove/zpool_remove_003_pos.ksh \
	functional/cli_root/zpool_reopen/cleanup.ksh \
	functional/cli_root/zpool_reopen/setup.ksh \
	functional/cli_root/zpool_reopen/zpool_reopen_001_pos.ksh \
	functional/cli_root/zpool_reopen/zpool_reopen_002_pos.ksh \
	functional/cli_root/zpool_reopen/zpool_reopen_003_pos.ksh \
	functional/cli_root/zpool_reopen/zpool_reopen_004_pos.ksh \
	functional/cli_root/zpool_reopen/zpool_reopen_005_pos.ksh \
	functional/cli_root/zpool_reopen/zpool_reopen_006_neg.ksh \
	functional/cli_root/zpool_reopen/zpool_reopen_007_pos.ksh \
	functional/cli_root/zpool_replace/cleanup.ksh \
	functional/cli_root/zpool_replace/replace-o_ashift.ksh \
	functional/cli_root/zpool_replace/replace_prop_ashift.ksh \
	functional/cli_root/zpool_replace/setup.ksh \
	functional/cli_root/zpool_replace/zpool_replace_001_neg.ksh \
	functional/cli_root/zpool_resilver/cleanup.ksh \
	functional/cli_root/zpool_resilver/setup.ksh \
	functional/cli_root/zpool_resilver/zpool_resilver_bad_args.ksh \
	functional/cli_root/zpool_resilver/zpool_resilver_restart.ksh \
	functional/cli_root/zpool_resilver/zpool_resilver_concurrent.ksh \
	functional/cli_root/zpool_scrub/cleanup.ksh \
	functional/cli_root/zpool_scrub/setup.ksh \
	functional/cli_root/zpool_scrub/zpool_scrub_001_neg.ksh \
	functional/cli_root/zpool_scrub/zpool_scrub_002_pos.ksh \
	functional/cli_root/zpool_scrub/zpool_scrub_003_pos.ksh \
	functional/cli_root/zpool_scrub/zpool_scrub_004_pos.ksh \
	functional/cli_root/zpool_scrub/zpool_scrub_005_pos.ksh \
	functional/cli_root/zpool_scrub/zpool_scrub_encrypted_unloaded.ksh \
	functional/cli_root/zpool_scrub/zpool_scrub_multiple_copies.ksh \
	functional/cli_root/zpool_scrub/zpool_scrub_offline_device.ksh \
	functional/cli_root/zpool_scrub/zpool_scrub_print_repairing.ksh \
	functional/cli_root/zpool_scrub/zpool_error_scrub_001_pos.ksh \
	functional/cli_root/zpool_scrub/zpool_error_scrub_002_pos.ksh \
	functional/cli_root/zpool_scrub/zpool_error_scrub_003_pos.ksh \
	functional/cli_root/zpool_scrub/zpool_error_scrub_004_pos.ksh \
	functional/cli_root/zpool_set/cleanup.ksh \
	functional/cli_root/zpool_set/setup.ksh \
	functional/cli_root/zpool/setup.ksh \
	functional/cli_root/zpool_set/vdev_set_001_pos.ksh \
	functional/cli_root/zpool_set/zpool_set_common.kshlib \
	functional/cli_root/zpool_set/zpool_set_001_pos.ksh \
	functional/cli_root/zpool_set/zpool_set_002_neg.ksh \
	functional/cli_root/zpool_set/zpool_set_003_neg.ksh \
	functional/cli_root/zpool_set/zpool_set_ashift.ksh \
	functional/cli_root/zpool_set/user_property_001_pos.ksh \
	functional/cli_root/zpool_set/user_property_002_neg.ksh \
	functional/cli_root/zpool_set/zpool_set_features.ksh \
	functional/cli_root/zpool_split/cleanup.ksh \
	functional/cli_root/zpool_split/setup.ksh \
	functional/cli_root/zpool_split/zpool_split_cliargs.ksh \
	functional/cli_root/zpool_split/zpool_split_devices.ksh \
	functional/cli_root/zpool_split/zpool_split_dryrun_output.ksh \
	functional/cli_root/zpool_split/zpool_split_encryption.ksh \
	functional/cli_root/zpool_split/zpool_split_indirect.ksh \
	functional/cli_root/zpool_split/zpool_split_props.ksh \
	functional/cli_root/zpool_split/zpool_split_resilver.ksh \
	functional/cli_root/zpool_split/zpool_split_vdevs.ksh \
	functional/cli_root/zpool_split/zpool_split_wholedisk.ksh \
	functional/cli_root/zpool_status/cleanup.ksh \
	functional/cli_root/zpool_status/setup.ksh \
	functional/cli_root/zpool_status/zpool_status_001_pos.ksh \
	functional/cli_root/zpool_status/zpool_status_002_pos.ksh \
	functional/cli_root/zpool_status/zpool_status_003_pos.ksh \
	functional/cli_root/zpool_status/zpool_status_004_pos.ksh \
	functional/cli_root/zpool_status/zpool_status_005_pos.ksh \
	functional/cli_root/zpool_status/zpool_status_006_pos.ksh \
	functional/cli_root/zpool_status/zpool_status_007_pos.ksh \
	functional/cli_root/zpool_status/zpool_status_008_pos.ksh \
	functional/cli_root/zpool_status/zpool_status_features_001_pos.ksh \
	functional/cli_root/zpool_sync/cleanup.ksh \
	functional/cli_root/zpool_sync/setup.ksh \
	functional/cli_root/zpool_sync/zpool_sync_001_pos.ksh \
	functional/cli_root/zpool_sync/zpool_sync_002_neg.ksh \
	functional/cli_root/zpool_trim/cleanup.ksh \
	functional/cli_root/zpool_trim/setup.ksh \
	functional/cli_root/zpool_trim/zpool_trim_attach_detach_add_remove.ksh \
	functional/cli_root/zpool_trim/zpool_trim_fault_export_import_online.ksh \
	functional/cli_root/zpool_trim/zpool_trim_import_export.ksh \
	functional/cli_root/zpool_trim/zpool_trim_multiple.ksh \
	functional/cli_root/zpool_trim/zpool_trim_neg.ksh \
	functional/cli_root/zpool_trim/zpool_trim_offline_export_import_online.ksh \
	functional/cli_root/zpool_trim/zpool_trim_online_offline.ksh \
	functional/cli_root/zpool_trim/zpool_trim_partial.ksh \
	functional/cli_root/zpool_trim/zpool_trim_rate.ksh \
	functional/cli_root/zpool_trim/zpool_trim_rate_neg.ksh \
	functional/cli_root/zpool_trim/zpool_trim_secure.ksh \
	functional/cli_root/zpool_trim/zpool_trim_split.ksh \
	functional/cli_root/zpool_trim/zpool_trim_start_and_cancel_neg.ksh \
	functional/cli_root/zpool_trim/zpool_trim_start_and_cancel_pos.ksh \
	functional/cli_root/zpool_trim/zpool_trim_suspend_resume.ksh \
	functional/cli_root/zpool_trim/zpool_trim_unsupported_vdevs.ksh \
	functional/cli_root/zpool_trim/zpool_trim_verify_checksums.ksh \
	functional/cli_root/zpool_trim/zpool_trim_verify_trimmed.ksh \
	functional/cli_root/zpool_upgrade/cleanup.ksh \
	functional/cli_root/zpool_upgrade/setup.ksh \
	functional/cli_root/zpool_upgrade/zpool_upgrade_001_pos.ksh \
	functional/cli_root/zpool_upgrade/zpool_upgrade_002_pos.ksh \
	functional/cli_root/zpool_upgrade/zpool_upgrade_003_pos.ksh \
	functional/cli_root/zpool_upgrade/zpool_upgrade_004_pos.ksh \
	functional/cli_root/zpool_upgrade/zpool_upgrade_005_neg.ksh \
	functional/cli_root/zpool_upgrade/zpool_upgrade_006_neg.ksh \
	functional/cli_root/zpool_upgrade/zpool_upgrade_007_pos.ksh \
	functional/cli_root/zpool_upgrade/zpool_upgrade_008_pos.ksh \
	functional/cli_root/zpool_upgrade/zpool_upgrade_009_neg.ksh \
	functional/cli_root/zpool_upgrade/zpool_upgrade_features_001_pos.ksh \
	functional/cli_root/zpool_wait/cleanup.ksh \
	functional/cli_root/zpool_wait/scan/cleanup.ksh \
	functional/cli_root/zpool_wait/scan/setup.ksh \
	functional/cli_root/zpool_wait/scan/zpool_wait_rebuild.ksh \
	functional/cli_root/zpool_wait/scan/zpool_wait_replace_cancel.ksh \
	functional/cli_root/zpool_wait/scan/zpool_wait_replace.ksh \
	functional/cli_root/zpool_wait/scan/zpool_wait_resilver.ksh \
	functional/cli_root/zpool_wait/scan/zpool_wait_scrub_basic.ksh \
	functional/cli_root/zpool_wait/scan/zpool_wait_scrub_cancel.ksh \
	functional/cli_root/zpool_wait/scan/zpool_wait_scrub_flag.ksh \
	functional/cli_root/zpool_wait/setup.ksh \
	functional/cli_root/zpool_wait/zpool_wait_discard.ksh \
	functional/cli_root/zpool_wait/zpool_wait_freeing.ksh \
	functional/cli_root/zpool_wait/zpool_wait_initialize_basic.ksh \
	functional/cli_root/zpool_wait/zpool_wait_initialize_cancel.ksh \
	functional/cli_root/zpool_wait/zpool_wait_initialize_flag.ksh \
	functional/cli_root/zpool_wait/zpool_wait_multiple.ksh \
	functional/cli_root/zpool_wait/zpool_wait_no_activity.ksh \
	functional/cli_root/zpool_wait/zpool_wait_remove_cancel.ksh \
	functional/cli_root/zpool_wait/zpool_wait_remove.ksh \
	functional/cli_root/zpool_wait/zpool_wait_trim_basic.ksh \
	functional/cli_root/zpool_wait/zpool_wait_trim_cancel.ksh \
	functional/cli_root/zpool_wait/zpool_wait_trim_flag.ksh \
	functional/cli_root/zpool_wait/zpool_wait_usage.ksh \
	functional/cli_root/zpool/zpool_001_neg.ksh \
	functional/cli_root/zpool/zpool_002_pos.ksh \
	functional/cli_root/zpool/zpool_003_pos.ksh \
	functional/cli_root/zpool/zpool_colors.ksh \
	functional/cli_user/misc/arcstat_001_pos.ksh \
	functional/cli_user/misc/arc_summary_001_pos.ksh \
	functional/cli_user/misc/arc_summary_002_neg.ksh \
	functional/cli_user/misc/zilstat_001_pos.ksh \
	functional/cli_user/misc/cleanup.ksh \
	functional/cli_user/misc/setup.ksh \
	functional/cli_user/misc/zdb_001_neg.ksh \
	functional/cli_user/misc/zfs_001_neg.ksh \
	functional/cli_user/misc/zfs_allow_001_neg.ksh \
	functional/cli_user/misc/zfs_clone_001_neg.ksh \
	functional/cli_user/misc/zfs_create_001_neg.ksh \
	functional/cli_user/misc/zfs_destroy_001_neg.ksh \
	functional/cli_user/misc/zfs_get_001_neg.ksh \
	functional/cli_user/misc/zfs_inherit_001_neg.ksh \
	functional/cli_user/misc/zfs_mount_001_neg.ksh \
	functional/cli_user/misc/zfs_promote_001_neg.ksh \
	functional/cli_user/misc/zfs_receive_001_neg.ksh \
	functional/cli_user/misc/zfs_rename_001_neg.ksh \
	functional/cli_user/misc/zfs_rollback_001_neg.ksh \
	functional/cli_user/misc/zfs_send_001_neg.ksh \
	functional/cli_user/misc/zfs_set_001_neg.ksh \
	functional/cli_user/misc/zfs_share_001_neg.ksh \
	functional/cli_user/misc/zfs_snapshot_001_neg.ksh \
	functional/cli_user/misc/zfs_unallow_001_neg.ksh \
	functional/cli_user/misc/zfs_unmount_001_neg.ksh \
	functional/cli_user/misc/zfs_unshare_001_neg.ksh \
	functional/cli_user/misc/zfs_upgrade_001_neg.ksh \
	functional/cli_user/misc/zpool_001_neg.ksh \
	functional/cli_user/misc/zpool_add_001_neg.ksh \
	functional/cli_user/misc/zpool_attach_001_neg.ksh \
	functional/cli_user/misc/zpool_clear_001_neg.ksh \
	functional/cli_user/misc/zpool_create_001_neg.ksh \
	functional/cli_user/misc/zpool_destroy_001_neg.ksh \
	functional/cli_user/misc/zpool_detach_001_neg.ksh \
	functional/cli_user/misc/zpool_export_001_neg.ksh \
	functional/cli_user/misc/zpool_get_001_neg.ksh \
	functional/cli_user/misc/zpool_history_001_neg.ksh \
	functional/cli_user/misc/zpool_import_001_neg.ksh \
	functional/cli_user/misc/zpool_import_002_neg.ksh \
	functional/cli_user/misc/zpool_offline_001_neg.ksh \
	functional/cli_user/misc/zpool_online_001_neg.ksh \
	functional/cli_user/misc/zpool_remove_001_neg.ksh \
	functional/cli_user/misc/zpool_replace_001_neg.ksh \
	functional/cli_user/misc/zpool_scrub_001_neg.ksh \
	functional/cli_user/misc/zpool_set_001_neg.ksh \
	functional/cli_user/misc/zpool_status_001_neg.ksh \
	functional/cli_user/misc/zpool_upgrade_001_neg.ksh \
	functional/cli_user/misc/zpool_wait_privilege.ksh \
	functional/cli_user/zfs_list/cleanup.ksh \
	functional/cli_user/zfs_list/setup.ksh \
	functional/cli_user/zfs_list/zfs_list_001_pos.ksh \
	functional/cli_user/zfs_list/zfs_list_002_pos.ksh \
	functional/cli_user/zfs_list/zfs_list_003_pos.ksh \
	functional/cli_user/zfs_list/zfs_list_004_neg.ksh \
	functional/cli_user/zfs_list/zfs_list_005_neg.ksh \
	functional/cli_user/zfs_list/zfs_list_007_pos.ksh \
	functional/cli_user/zfs_list/zfs_list_008_neg.ksh \
	functional/cli_user/zpool_iostat/cleanup.ksh \
	functional/cli_user/zpool_iostat/setup.ksh \
	functional/cli_user/zpool_iostat/zpool_iostat_001_neg.ksh \
	functional/cli_user/zpool_iostat/zpool_iostat_002_pos.ksh \
	functional/cli_user/zpool_iostat/zpool_iostat_003_neg.ksh \
	functional/cli_user/zpool_iostat/zpool_iostat_004_pos.ksh \
	functional/cli_user/zpool_iostat/zpool_iostat_005_pos.ksh \
	functional/cli_user/zpool_iostat/zpool_iostat_-c_disable.ksh \
	functional/cli_user/zpool_iostat/zpool_iostat_-c_homedir.ksh \
	functional/cli_user/zpool_iostat/zpool_iostat_-c_searchpath.ksh \
	functional/cli_user/zpool_list/cleanup.ksh \
	functional/cli_user/zpool_list/setup.ksh \
	functional/cli_user/zpool_list/zpool_list_001_pos.ksh \
	functional/cli_user/zpool_list/zpool_list_002_neg.ksh \
	functional/cli_user/zpool_status/cleanup.ksh \
	functional/cli_user/zpool_status/setup.ksh \
	functional/cli_user/zpool_status/zpool_status_003_pos.ksh \
	functional/cli_user/zpool_status/zpool_status_-c_disable.ksh \
	functional/cli_user/zpool_status/zpool_status_-c_homedir.ksh \
	functional/cli_user/zpool_status/zpool_status_-c_searchpath.ksh \
	functional/compression/cleanup.ksh \
	functional/compression/compress_001_pos.ksh \
	functional/compression/compress_002_pos.ksh \
	functional/compression/compress_003_pos.ksh \
	functional/compression/compress_004_pos.ksh \
	functional/compression/compress_zstd_bswap.ksh \
	functional/compression/l2arc_compressed_arc_disabled.ksh \
	functional/compression/l2arc_compressed_arc.ksh \
	functional/compression/l2arc_encrypted.ksh \
	functional/compression/l2arc_encrypted_no_compressed_arc.ksh \
	functional/compression/setup.ksh \
	functional/cp_files/cleanup.ksh \
	functional/cp_files/cp_files_001_pos.ksh \
	functional/cp_files/cp_files_002_pos.ksh \
	functional/cp_files/cp_stress.ksh \
	functional/cp_files/setup.ksh \
	functional/crtime/cleanup.ksh \
	functional/crtime/crtime_001_pos.ksh \
	functional/crtime/setup.ksh \
	functional/ctime/cleanup.ksh \
	functional/ctime/ctime_001_pos.ksh \
	functional/ctime/setup.ksh \
	functional/deadman/deadman_ratelimit.ksh \
	functional/deadman/deadman_sync.ksh \
	functional/deadman/deadman_zio.ksh \
	functional/delegate/cleanup.ksh \
	functional/delegate/setup.ksh \
	functional/delegate/zfs_allow_001_pos.ksh \
	functional/delegate/zfs_allow_002_pos.ksh \
	functional/delegate/zfs_allow_003_pos.ksh \
	functional/delegate/zfs_allow_004_pos.ksh \
	functional/delegate/zfs_allow_005_pos.ksh \
	functional/delegate/zfs_allow_006_pos.ksh \
	functional/delegate/zfs_allow_007_pos.ksh \
	functional/delegate/zfs_allow_008_pos.ksh \
	functional/delegate/zfs_allow_009_neg.ksh \
	functional/delegate/zfs_allow_010_pos.ksh \
	functional/delegate/zfs_allow_011_neg.ksh \
	functional/delegate/zfs_allow_012_neg.ksh \
	functional/delegate/zfs_unallow_001_pos.ksh \
	functional/delegate/zfs_unallow_002_pos.ksh \
	functional/delegate/zfs_unallow_003_pos.ksh \
	functional/delegate/zfs_unallow_004_pos.ksh \
	functional/delegate/zfs_unallow_005_pos.ksh \
	functional/delegate/zfs_unallow_006_pos.ksh \
	functional/delegate/zfs_unallow_007_neg.ksh \
	functional/delegate/zfs_unallow_008_neg.ksh \
	functional/devices/cleanup.ksh \
	functional/devices/devices_001_pos.ksh \
	functional/devices/devices_002_neg.ksh \
	functional/devices/devices_003_pos.ksh \
	functional/devices/setup.ksh \
	functional/dos_attributes/cleanup.ksh \
	functional/dos_attributes/read_dos_attrs_001.ksh \
	functional/dos_attributes/setup.ksh \
	functional/dos_attributes/write_dos_attrs_001.ksh \
	functional/events/cleanup.ksh \
	functional/events/events_001_pos.ksh \
	functional/events/events_002_pos.ksh \
	functional/events/setup.ksh \
	functional/events/zed_cksum_config.ksh \
	functional/events/zed_cksum_reported.ksh \
	functional/events/zed_fd_spill.ksh \
	functional/events/zed_io_config.ksh \
	functional/events/zed_rc_filter.ksh \
	functional/events/zed_slow_io.ksh \
	functional/events/zed_slow_io_many_vdevs.ksh \
	functional/exec/cleanup.ksh \
	functional/exec/exec_001_pos.ksh \
	functional/exec/exec_002_neg.ksh \
	functional/exec/setup.ksh \
	functional/fadvise/cleanup.ksh \
	functional/fadvise/fadvise_sequential.ksh \
	functional/fadvise/setup.ksh \
	functional/fallocate/cleanup.ksh \
	functional/fallocate/fallocate_prealloc.ksh \
	functional/fallocate/fallocate_punch-hole.ksh \
	functional/fallocate/fallocate_zero-range.ksh \
	functional/fallocate/setup.ksh \
	functional/fault/auto_offline_001_pos.ksh \
	functional/fault/auto_online_001_pos.ksh \
	functional/fault/auto_online_002_pos.ksh \
	functional/fault/auto_replace_001_pos.ksh \
	functional/fault/auto_replace_002_pos.ksh \
	functional/fault/auto_spare_001_pos.ksh \
	functional/fault/auto_spare_002_pos.ksh \
	functional/fault/auto_spare_ashift.ksh \
	functional/fault/auto_spare_multiple.ksh \
	functional/fault/auto_spare_shared.ksh \
	functional/fault/cleanup.ksh \
	functional/fault/decompress_fault.ksh \
	functional/fault/decrypt_fault.ksh \
	functional/fault/scrub_after_resilver.ksh \
	functional/fault/setup.ksh \
	functional/fault/zpool_status_-s.ksh \
	functional/features/async_destroy/async_destroy_001_pos.ksh \
	functional/features/async_destroy/cleanup.ksh \
	functional/features/async_destroy/setup.ksh \
	functional/features/large_dnode/cleanup.ksh \
	functional/features/large_dnode/large_dnode_001_pos.ksh \
	functional/features/large_dnode/large_dnode_002_pos.ksh \
	functional/features/large_dnode/large_dnode_003_pos.ksh \
	functional/features/large_dnode/large_dnode_004_neg.ksh \
	functional/features/large_dnode/large_dnode_005_pos.ksh \
	functional/features/large_dnode/large_dnode_006_pos.ksh \
	functional/features/large_dnode/large_dnode_007_neg.ksh \
	functional/features/large_dnode/large_dnode_008_pos.ksh \
	functional/features/large_dnode/large_dnode_009_pos.ksh \
	functional/features/large_dnode/setup.ksh \
	functional/grow/grow_pool_001_pos.ksh \
	functional/grow/grow_replicas_001_pos.ksh \
	functional/history/cleanup.ksh \
	functional/history/history_001_pos.ksh \
	functional/history/history_002_pos.ksh \
	functional/history/history_003_pos.ksh \
	functional/history/history_004_pos.ksh \
	functional/history/history_005_neg.ksh \
	functional/history/history_006_neg.ksh \
	functional/history/history_007_pos.ksh \
	functional/history/history_008_pos.ksh \
	functional/history/history_009_pos.ksh \
	functional/history/history_010_pos.ksh \
	functional/history/setup.ksh \
	functional/inheritance/cleanup.ksh \
	functional/inheritance/inherit_001_pos.ksh \
	functional/inuse/inuse_001_pos.ksh \
	functional/inuse/inuse_003_pos.ksh \
	functional/inuse/inuse_004_pos.ksh \
	functional/inuse/inuse_005_pos.ksh \
	functional/inuse/inuse_006_pos.ksh \
	functional/inuse/inuse_007_pos.ksh \
	functional/inuse/inuse_008_pos.ksh \
	functional/inuse/inuse_009_pos.ksh \
	functional/inuse/setup.ksh \
	functional/io/cleanup.ksh \
	functional/io/io_uring.ksh \
	functional/io/libaio.ksh \
	functional/io/mmap.ksh \
	functional/io/posixaio.ksh \
	functional/io/psync.ksh \
	functional/io/setup.ksh \
	functional/io/sync.ksh \
	functional/l2arc/cleanup.ksh \
	functional/l2arc/l2arc_arcstats_pos.ksh \
	functional/l2arc/l2arc_l2miss_pos.ksh \
	functional/l2arc/l2arc_mfuonly_pos.ksh \
	functional/l2arc/persist_l2arc_001_pos.ksh \
	functional/l2arc/persist_l2arc_002_pos.ksh \
	functional/l2arc/persist_l2arc_003_neg.ksh \
	functional/l2arc/persist_l2arc_004_pos.ksh \
	functional/l2arc/persist_l2arc_005_pos.ksh \
	functional/l2arc/setup.ksh \
	functional/large_files/cleanup.ksh \
	functional/large_files/large_files_001_pos.ksh \
	functional/large_files/large_files_002_pos.ksh \
	functional/large_files/setup.ksh \
	functional/largest_pool/largest_pool_001_pos.ksh \
	functional/libzfs/cleanup.ksh \
	functional/libzfs/libzfs_input.ksh \
	functional/libzfs/setup.ksh \
	functional/limits/cleanup.ksh \
	functional/limits/filesystem_count.ksh \
	functional/limits/filesystem_limit.ksh \
	functional/limits/setup.ksh \
	functional/limits/snapshot_count.ksh \
	functional/limits/snapshot_limit.ksh \
	functional/link_count/cleanup.ksh \
	functional/link_count/link_count_001.ksh \
	functional/link_count/link_count_root_inode.ksh \
	functional/link_count/setup.ksh \
	functional/log_spacemap/log_spacemap_import_logs.ksh \
	functional/migration/cleanup.ksh \
	functional/migration/migration_001_pos.ksh \
	functional/migration/migration_002_pos.ksh \
	functional/migration/migration_003_pos.ksh \
	functional/migration/migration_004_pos.ksh \
	functional/migration/migration_005_pos.ksh \
	functional/migration/migration_006_pos.ksh \
	functional/migration/migration_007_pos.ksh \
	functional/migration/migration_008_pos.ksh \
	functional/migration/migration_009_pos.ksh \
	functional/migration/migration_010_pos.ksh \
	functional/migration/migration_011_pos.ksh \
	functional/migration/migration_012_pos.ksh \
	functional/migration/setup.ksh \
	functional/mmap/cleanup.ksh \
	functional/mmap/mmap_libaio_001_pos.ksh \
	functional/mmap/mmap_mixed.ksh \
	functional/mmap/mmap_read_001_pos.ksh \
	functional/mmap/mmap_seek_001_pos.ksh \
	functional/mmap/mmap_sync_001_pos.ksh \
	functional/mmap/mmap_write_001_pos.ksh \
	functional/mmap/setup.ksh \
	functional/mmp/cleanup.ksh \
	functional/mmp/mmp_active_import.ksh \
	functional/mmp/mmp_exported_import.ksh \
	functional/mmp/mmp_hostid.ksh \
	functional/mmp/mmp_inactive_import.ksh \
	functional/mmp/mmp_interval.ksh \
	functional/mmp/mmp_on_off.ksh \
	functional/mmp/mmp_on_thread.ksh \
	functional/mmp/mmp_on_uberblocks.ksh \
	functional/mmp/mmp_on_zdb.ksh \
	functional/mmp/mmp_reset_interval.ksh \
	functional/mmp/mmp_write_distribution.ksh \
	functional/mmp/mmp_write_slow_disk.ksh \
	functional/mmp/mmp_write_uberblocks.ksh \
	functional/mmp/multihost_history.ksh \
	functional/mmp/setup.ksh \
	functional/mount/cleanup.ksh \
	functional/mount/setup.ksh \
	functional/mount/umount_001.ksh \
	functional/mount/umountall_001.ksh \
	functional/mount/umount_unlinked_drain.ksh \
	functional/mv_files/cleanup.ksh \
	functional/mv_files/mv_files_001_pos.ksh \
	functional/mv_files/mv_files_002_pos.ksh \
	functional/mv_files/random_creation.ksh \
	functional/mv_files/setup.ksh \
	functional/nestedfs/cleanup.ksh \
	functional/nestedfs/nestedfs_001_pos.ksh \
	functional/nestedfs/setup.ksh \
	functional/nopwrite/cleanup.ksh \
	functional/nopwrite/nopwrite_copies.ksh \
	functional/nopwrite/nopwrite_mtime.ksh \
	functional/nopwrite/nopwrite_negative.ksh \
	functional/nopwrite/nopwrite_promoted_clone.ksh \
	functional/nopwrite/nopwrite_recsize.ksh \
	functional/nopwrite/nopwrite_sync.ksh \
	functional/nopwrite/nopwrite_varying_compression.ksh \
	functional/nopwrite/nopwrite_volume.ksh \
	functional/nopwrite/setup.ksh \
	functional/no_space/cleanup.ksh \
	functional/no_space/enospc_001_pos.ksh \
	functional/no_space/enospc_002_pos.ksh \
	functional/no_space/enospc_003_pos.ksh \
	functional/no_space/enospc_df.ksh \
	functional/no_space/enospc_ganging.ksh \
	functional/no_space/enospc_rm.ksh \
	functional/no_space/setup.ksh \
	functional/online_offline/cleanup.ksh \
	functional/online_offline/online_offline_001_pos.ksh \
	functional/online_offline/online_offline_002_neg.ksh \
	functional/online_offline/online_offline_003_neg.ksh \
	functional/online_offline/setup.ksh \
	functional/pam/cleanup.ksh \
	functional/pam/pam_basic.ksh \
	functional/pam/pam_change_unmounted.ksh \
	functional/pam/pam_nounmount.ksh \
	functional/pam/pam_recursive.ksh \
	functional/pam/pam_short_password.ksh \
	functional/pam/setup.ksh \
	functional/pool_checkpoint/checkpoint_after_rewind.ksh \
	functional/pool_checkpoint/checkpoint_big_rewind.ksh \
	functional/pool_checkpoint/checkpoint_capacity.ksh \
	functional/pool_checkpoint/checkpoint_conf_change.ksh \
	functional/pool_checkpoint/checkpoint_discard_busy.ksh \
	functional/pool_checkpoint/checkpoint_discard.ksh \
	functional/pool_checkpoint/checkpoint_discard_many.ksh \
	functional/pool_checkpoint/checkpoint_indirect.ksh \
	functional/pool_checkpoint/checkpoint_invalid.ksh \
	functional/pool_checkpoint/checkpoint_lun_expsz.ksh \
	functional/pool_checkpoint/checkpoint_open.ksh \
	functional/pool_checkpoint/checkpoint_removal.ksh \
	functional/pool_checkpoint/checkpoint_rewind.ksh \
	functional/pool_checkpoint/checkpoint_ro_rewind.ksh \
	functional/pool_checkpoint/checkpoint_sm_scale.ksh \
	functional/pool_checkpoint/checkpoint_twice.ksh \
	functional/pool_checkpoint/checkpoint_vdev_add.ksh \
	functional/pool_checkpoint/checkpoint_zdb.ksh \
	functional/pool_checkpoint/checkpoint_zhack_feat.ksh \
	functional/pool_checkpoint/cleanup.ksh \
	functional/pool_checkpoint/setup.ksh \
	functional/pool_names/pool_names_001_pos.ksh \
	functional/pool_names/pool_names_002_neg.ksh \
	functional/poolversion/cleanup.ksh \
	functional/poolversion/poolversion_001_pos.ksh \
	functional/poolversion/poolversion_002_pos.ksh \
	functional/poolversion/setup.ksh \
	functional/privilege/cleanup.ksh \
	functional/privilege/privilege_001_pos.ksh \
	functional/privilege/privilege_002_pos.ksh \
	functional/privilege/setup.ksh \
	functional/procfs/cleanup.ksh \
	functional/procfs/pool_state.ksh \
	functional/procfs/procfs_list_basic.ksh \
	functional/procfs/procfs_list_concurrent_readers.ksh \
	functional/procfs/procfs_list_stale_read.ksh \
	functional/procfs/setup.ksh \
	functional/projectquota/cleanup.ksh \
	functional/projectquota/projectid_001_pos.ksh \
	functional/projectquota/projectid_002_pos.ksh \
	functional/projectquota/projectid_003_pos.ksh \
	functional/projectquota/projectquota_001_pos.ksh \
	functional/projectquota/projectquota_002_pos.ksh \
	functional/projectquota/projectquota_003_pos.ksh \
	functional/projectquota/projectquota_004_neg.ksh \
	functional/projectquota/projectquota_005_pos.ksh \
	functional/projectquota/projectquota_006_pos.ksh \
	functional/projectquota/projectquota_007_pos.ksh \
	functional/projectquota/projectquota_008_pos.ksh \
	functional/projectquota/projectquota_009_pos.ksh \
	functional/projectquota/projectspace_001_pos.ksh \
	functional/projectquota/projectspace_002_pos.ksh \
	functional/projectquota/projectspace_003_pos.ksh \
	functional/projectquota/projectspace_004_pos.ksh \
	functional/projectquota/projecttree_001_pos.ksh \
	functional/projectquota/projecttree_002_pos.ksh \
	functional/projectquota/projecttree_003_neg.ksh \
	functional/projectquota/setup.ksh \
	functional/quota/cleanup.ksh \
	functional/quota/quota_001_pos.ksh \
	functional/quota/quota_002_pos.ksh \
	functional/quota/quota_003_pos.ksh \
	functional/quota/quota_004_pos.ksh \
	functional/quota/quota_005_pos.ksh \
	functional/quota/quota_006_neg.ksh \
	functional/quota/setup.ksh \
	functional/raidz/cleanup.ksh \
	functional/raidz/raidz_001_neg.ksh \
	functional/raidz/raidz_002_pos.ksh \
	functional/raidz/raidz_expand_001_pos.ksh \
	functional/raidz/raidz_expand_002_pos.ksh \
	functional/raidz/raidz_expand_003_neg.ksh \
	functional/raidz/raidz_expand_003_pos.ksh \
	functional/raidz/raidz_expand_004_pos.ksh \
	functional/raidz/raidz_expand_005_pos.ksh \
	functional/raidz/raidz_expand_006_neg.ksh \
	functional/raidz/raidz_expand_007_neg.ksh \
	functional/raidz/setup.ksh \
	functional/redacted_send/cleanup.ksh \
	functional/redacted_send/redacted_compressed.ksh \
	functional/redacted_send/redacted_contents.ksh \
	functional/redacted_send/redacted_deleted.ksh \
	functional/redacted_send/redacted_disabled_feature.ksh \
	functional/redacted_send/redacted_embedded.ksh \
	functional/redacted_send/redacted_holes.ksh \
	functional/redacted_send/redacted_incrementals.ksh \
	functional/redacted_send/redacted_largeblocks.ksh \
	functional/redacted_send/redacted_many_clones.ksh \
	functional/redacted_send/redacted_mixed_recsize.ksh \
	functional/redacted_send/redacted_mounts.ksh \
	functional/redacted_send/redacted_negative.ksh \
	functional/redacted_send/redacted_origin.ksh \
	functional/redacted_send/redacted_panic.ksh \
	functional/redacted_send/redacted_props.ksh \
	functional/redacted_send/redacted_resume.ksh \
	functional/redacted_send/redacted_size.ksh \
	functional/redacted_send/redacted_volume.ksh \
	functional/redacted_send/setup.ksh \
	functional/redundancy/cleanup.ksh \
	functional/redundancy/redundancy_draid1.ksh \
	functional/redundancy/redundancy_draid2.ksh \
	functional/redundancy/redundancy_draid3.ksh \
	functional/redundancy/redundancy_draid_damaged1.ksh \
	functional/redundancy/redundancy_draid_damaged2.ksh \
	functional/redundancy/redundancy_draid.ksh \
	functional/redundancy/redundancy_draid_spare1.ksh \
	functional/redundancy/redundancy_draid_spare2.ksh \
	functional/redundancy/redundancy_draid_spare3.ksh \
	functional/redundancy/redundancy_mirror.ksh \
	functional/redundancy/redundancy_raidz1.ksh \
	functional/redundancy/redundancy_raidz2.ksh \
	functional/redundancy/redundancy_raidz3.ksh \
	functional/redundancy/redundancy_raidz.ksh \
	functional/redundancy/redundancy_stripe.ksh \
	functional/redundancy/setup.ksh \
	functional/refquota/cleanup.ksh \
	functional/refquota/refquota_001_pos.ksh \
	functional/refquota/refquota_002_pos.ksh \
	functional/refquota/refquota_003_pos.ksh \
	functional/refquota/refquota_004_pos.ksh \
	functional/refquota/refquota_005_pos.ksh \
	functional/refquota/refquota_006_neg.ksh \
	functional/refquota/refquota_007_neg.ksh \
	functional/refquota/refquota_008_neg.ksh \
	functional/refquota/setup.ksh \
	functional/refreserv/cleanup.ksh \
	functional/refreserv/refreserv_001_pos.ksh \
	functional/refreserv/refreserv_002_pos.ksh \
	functional/refreserv/refreserv_003_pos.ksh \
	functional/refreserv/refreserv_004_pos.ksh \
	functional/refreserv/refreserv_005_pos.ksh \
	functional/refreserv/refreserv_multi_raidz.ksh \
	functional/refreserv/refreserv_raidz.ksh \
	functional/refreserv/setup.ksh \
	functional/removal/cleanup.ksh \
	functional/removal/removal_all_vdev.ksh \
	functional/removal/removal_cancel.ksh \
	functional/removal/removal_check_space.ksh \
	functional/removal/removal_condense_export.ksh \
	functional/removal/removal_multiple_indirection.ksh \
	functional/removal/removal_nopwrite.ksh \
	functional/removal/removal_remap_deadlists.ksh \
	functional/removal/removal_reservation.ksh \
	functional/removal/removal_resume_export.ksh \
	functional/removal/removal_sanity.ksh \
	functional/removal/removal_with_add.ksh \
	functional/removal/removal_with_create_fs.ksh \
	functional/removal/removal_with_dedup.ksh \
	functional/removal/removal_with_errors.ksh \
	functional/removal/removal_with_export.ksh \
	functional/removal/removal_with_faulted.ksh \
	functional/removal/removal_with_ganging.ksh \
	functional/removal/removal_with_indirect.ksh \
	functional/removal/removal_with_remove.ksh \
	functional/removal/removal_with_scrub.ksh \
	functional/removal/removal_with_send.ksh \
	functional/removal/removal_with_send_recv.ksh \
	functional/removal/removal_with_snapshot.ksh \
	functional/removal/removal_with_write.ksh \
	functional/removal/removal_with_zdb.ksh \
	functional/removal/remove_attach_mirror.ksh \
	functional/removal/remove_expanded.ksh \
	functional/removal/remove_indirect.ksh \
	functional/removal/remove_mirror.ksh \
	functional/removal/remove_mirror_sanity.ksh \
	functional/removal/remove_raidz.ksh \
	functional/rename_dirs/cleanup.ksh \
	functional/rename_dirs/rename_dirs_001_pos.ksh \
	functional/rename_dirs/setup.ksh \
	functional/renameat2/cleanup.ksh \
	functional/renameat2/setup.ksh \
	functional/renameat2/renameat2_exchange.ksh \
	functional/renameat2/renameat2_noreplace.ksh \
	functional/renameat2/renameat2_whiteout.ksh \
	functional/replacement/attach_import.ksh \
	functional/replacement/attach_multiple.ksh \
	functional/replacement/attach_rebuild.ksh \
	functional/replacement/attach_resilver.ksh \
	functional/replacement/cleanup.ksh \
	functional/replacement/detach.ksh \
	functional/replacement/rebuild_disabled_feature.ksh \
	functional/replacement/rebuild_multiple.ksh \
	functional/replacement/rebuild_raidz.ksh \
	functional/replacement/replace_import.ksh \
	functional/replacement/replace_rebuild.ksh \
	functional/replacement/replace_resilver.ksh \
	functional/replacement/resilver_restart_001.ksh \
	functional/replacement/resilver_restart_002.ksh \
	functional/replacement/scrub_cancel.ksh \
	functional/replacement/setup.ksh \
	functional/reservation/cleanup.ksh \
	functional/reservation/reservation_001_pos.ksh \
	functional/reservation/reservation_002_pos.ksh \
	functional/reservation/reservation_003_pos.ksh \
	functional/reservation/reservation_004_pos.ksh \
	functional/reservation/reservation_005_pos.ksh \
	functional/reservation/reservation_006_pos.ksh \
	functional/reservation/reservation_007_pos.ksh \
	functional/reservation/reservation_008_pos.ksh \
	functional/reservation/reservation_009_pos.ksh \
	functional/reservation/reservation_010_pos.ksh \
	functional/reservation/reservation_011_pos.ksh \
	functional/reservation/reservation_012_pos.ksh \
	functional/reservation/reservation_013_pos.ksh \
	functional/reservation/reservation_014_pos.ksh \
	functional/reservation/reservation_015_pos.ksh \
	functional/reservation/reservation_016_pos.ksh \
	functional/reservation/reservation_017_pos.ksh \
	functional/reservation/reservation_018_pos.ksh \
	functional/reservation/reservation_019_pos.ksh \
	functional/reservation/reservation_020_pos.ksh \
	functional/reservation/reservation_021_neg.ksh \
	functional/reservation/reservation_022_pos.ksh \
	functional/reservation/setup.ksh \
	functional/rootpool/cleanup.ksh \
	functional/rootpool/rootpool_002_neg.ksh \
	functional/rootpool/rootpool_003_neg.ksh \
	functional/rootpool/rootpool_007_pos.ksh \
	functional/rootpool/setup.ksh \
	functional/rsend/cleanup.ksh \
	functional/rsend/recv_dedup_encrypted_zvol.ksh \
	functional/rsend/recv_dedup.ksh \
	functional/rsend/rsend_001_pos.ksh \
	functional/rsend/rsend_002_pos.ksh \
	functional/rsend/rsend_003_pos.ksh \
	functional/rsend/rsend_004_pos.ksh \
	functional/rsend/rsend_005_pos.ksh \
	functional/rsend/rsend_006_pos.ksh \
	functional/rsend/rsend_007_pos.ksh \
	functional/rsend/rsend_008_pos.ksh \
	functional/rsend/rsend_009_pos.ksh \
	functional/rsend/rsend_010_pos.ksh \
	functional/rsend/rsend_011_pos.ksh \
	functional/rsend/rsend_012_pos.ksh \
	functional/rsend/rsend_013_pos.ksh \
	functional/rsend/rsend_014_pos.ksh \
	functional/rsend/rsend_016_neg.ksh \
	functional/rsend/rsend_019_pos.ksh \
	functional/rsend/rsend_020_pos.ksh \
	functional/rsend/rsend_021_pos.ksh \
	functional/rsend/rsend_022_pos.ksh \
	functional/rsend/rsend_024_pos.ksh \
	functional/rsend/rsend_025_pos.ksh \
	functional/rsend/rsend_026_neg.ksh \
	functional/rsend/rsend_027_pos.ksh \
	functional/rsend/rsend_028_neg.ksh \
	functional/rsend/rsend_029_neg.ksh \
	functional/rsend/rsend_030_pos.ksh \
	functional/rsend/rsend_031_pos.ksh \
	functional/rsend/send-c_embedded_blocks.ksh \
	functional/rsend/send-c_incremental.ksh \
	functional/rsend/send-c_lz4_disabled.ksh \
	functional/rsend/send-c_mixed_compression.ksh \
	functional/rsend/send-c_props.ksh \
	functional/rsend/send-c_recv_dedup.ksh \
	functional/rsend/send-c_recv_lz4_disabled.ksh \
	functional/rsend/send-c_resume.ksh \
	functional/rsend/send-c_stream_size_estimate.ksh \
	functional/rsend/send-c_verify_contents.ksh \
	functional/rsend/send-c_verify_ratio.ksh \
	functional/rsend/send-c_volume.ksh \
	functional/rsend/send-c_zstream_recompress.ksh \
	functional/rsend/send-c_zstreamdump.ksh \
	functional/rsend/send-cpL_varied_recsize.ksh \
	functional/rsend/send_doall.ksh \
	functional/rsend/send_encrypted_incremental.ksh \
	functional/rsend/send_encrypted_files.ksh \
	functional/rsend/send_encrypted_freeobjects.ksh \
	functional/rsend/send_encrypted_hierarchy.ksh \
	functional/rsend/send_encrypted_props.ksh \
	functional/rsend/send_encrypted_truncated_files.ksh \
	functional/rsend/send_freeobjects.ksh \
	functional/rsend/send_holds.ksh \
	functional/rsend/send_hole_birth.ksh \
	functional/rsend/send_invalid.ksh \
	functional/rsend/send-L_toggle.ksh \
	functional/rsend/send_mixed_raw.ksh \
	functional/rsend/send_partial_dataset.ksh \
	functional/rsend/send_raw_ashift.ksh \
	functional/rsend/send_raw_spill_block.ksh \
	functional/rsend/send_raw_large_blocks.ksh \
	functional/rsend/send_realloc_dnode_size.ksh \
	functional/rsend/send_realloc_encrypted_files.ksh \
	functional/rsend/send_realloc_files.ksh \
	functional/rsend/send_spill_block.ksh \
	functional/rsend/send-wR_encrypted_zvol.ksh \
	functional/rsend/setup.ksh \
	functional/scrub_mirror/cleanup.ksh \
	functional/scrub_mirror/scrub_mirror_001_pos.ksh \
	functional/scrub_mirror/scrub_mirror_002_pos.ksh \
	functional/scrub_mirror/scrub_mirror_003_pos.ksh \
	functional/scrub_mirror/scrub_mirror_004_pos.ksh \
	functional/scrub_mirror/setup.ksh \
	functional/slog/cleanup.ksh \
	functional/slog/setup.ksh \
	functional/slog/slog_001_pos.ksh \
	functional/slog/slog_002_pos.ksh \
	functional/slog/slog_003_pos.ksh \
	functional/slog/slog_004_pos.ksh \
	functional/slog/slog_005_pos.ksh \
	functional/slog/slog_006_pos.ksh \
	functional/slog/slog_007_pos.ksh \
	functional/slog/slog_008_neg.ksh \
	functional/slog/slog_009_neg.ksh \
	functional/slog/slog_010_neg.ksh \
	functional/slog/slog_011_neg.ksh \
	functional/slog/slog_012_neg.ksh \
	functional/slog/slog_013_pos.ksh \
	functional/slog/slog_014_pos.ksh \
	functional/slog/slog_015_neg.ksh \
	functional/slog/slog_016_pos.ksh \
	functional/slog/slog_replay_fs_001.ksh \
	functional/slog/slog_replay_fs_002.ksh \
	functional/slog/slog_replay_volume.ksh \
	functional/snapshot/cleanup.ksh \
	functional/snapshot/clone_001_pos.ksh \
	functional/snapshot/rollback_001_pos.ksh \
	functional/snapshot/rollback_002_pos.ksh \
	functional/snapshot/rollback_003_pos.ksh \
	functional/snapshot/setup.ksh \
	functional/snapshot/snapshot_001_pos.ksh \
	functional/snapshot/snapshot_002_pos.ksh \
	functional/snapshot/snapshot_003_pos.ksh \
	functional/snapshot/snapshot_004_pos.ksh \
	functional/snapshot/snapshot_005_pos.ksh \
	functional/snapshot/snapshot_006_pos.ksh \
	functional/snapshot/snapshot_007_pos.ksh \
	functional/snapshot/snapshot_008_pos.ksh \
	functional/snapshot/snapshot_009_pos.ksh \
	functional/snapshot/snapshot_010_pos.ksh \
	functional/snapshot/snapshot_011_pos.ksh \
	functional/snapshot/snapshot_012_pos.ksh \
	functional/snapshot/snapshot_013_pos.ksh \
	functional/snapshot/snapshot_014_pos.ksh \
	functional/snapshot/snapshot_015_pos.ksh \
	functional/snapshot/snapshot_016_pos.ksh \
	functional/snapshot/snapshot_017_pos.ksh \
	functional/snapshot/snapshot_018_pos.ksh \
	functional/snapused/cleanup.ksh \
	functional/snapused/setup.ksh \
	functional/snapused/snapused_001_pos.ksh \
	functional/snapused/snapused_002_pos.ksh \
	functional/snapused/snapused_003_pos.ksh \
	functional/snapused/snapused_004_pos.ksh \
	functional/snapused/snapused_005_pos.ksh \
	functional/sparse/cleanup.ksh \
	functional/sparse/setup.ksh \
	functional/sparse/sparse_001_pos.ksh \
	functional/stat/cleanup.ksh \
	functional/stat/setup.ksh \
	functional/stat/stat_001_pos.ksh \
	functional/suid/cleanup.ksh \
	functional/suid/setup.ksh \
	functional/suid/suid_write_to_none.ksh \
	functional/suid/suid_write_to_sgid.ksh \
	functional/suid/suid_write_to_suid.ksh \
	functional/suid/suid_write_to_suid_sgid.ksh \
	functional/suid/suid_write_zil_replay.ksh \
	functional/trim/autotrim_config.ksh \
	functional/trim/autotrim_integrity.ksh \
	functional/trim/autotrim_trim_integrity.ksh \
	functional/trim/cleanup.ksh \
	functional/trim/setup.ksh \
	functional/trim/trim_config.ksh \
	functional/trim/trim_integrity.ksh \
	functional/trim/trim_l2arc.ksh \
	functional/truncate/cleanup.ksh \
	functional/truncate/setup.ksh \
	functional/truncate/truncate_001_pos.ksh \
	functional/truncate/truncate_002_pos.ksh \
	functional/truncate/truncate_timestamps.ksh \
	functional/upgrade/cleanup.ksh \
	functional/upgrade/setup.ksh \
	functional/upgrade/upgrade_projectquota_001_pos.ksh \
	functional/upgrade/upgrade_readonly_pool.ksh \
	functional/upgrade/upgrade_userobj_001_pos.ksh \
	functional/user_namespace/cleanup.ksh \
	functional/user_namespace/setup.ksh \
	functional/user_namespace/user_namespace_001.ksh \
	functional/user_namespace/user_namespace_002.ksh \
	functional/user_namespace/user_namespace_003.ksh \
	functional/user_namespace/user_namespace_004.ksh \
	functional/userquota/cleanup.ksh \
	functional/userquota/groupspace_001_pos.ksh \
	functional/userquota/groupspace_002_pos.ksh \
	functional/userquota/groupspace_003_pos.ksh \
	functional/userquota/setup.ksh \
	functional/userquota/userquota_001_pos.ksh \
	functional/userquota/userquota_002_pos.ksh \
	functional/userquota/userquota_003_pos.ksh \
	functional/userquota/userquota_004_pos.ksh \
	functional/userquota/userquota_005_neg.ksh \
	functional/userquota/userquota_006_pos.ksh \
	functional/userquota/userquota_007_pos.ksh \
	functional/userquota/userquota_008_pos.ksh \
	functional/userquota/userquota_009_pos.ksh \
	functional/userquota/userquota_010_pos.ksh \
	functional/userquota/userquota_011_pos.ksh \
	functional/userquota/userquota_012_neg.ksh \
	functional/userquota/userquota_013_pos.ksh \
	functional/userquota/userspace_001_pos.ksh \
	functional/userquota/userspace_002_pos.ksh \
	functional/userquota/userspace_003_pos.ksh \
	functional/userquota/userspace_encrypted.ksh \
	functional/userquota/userspace_send_encrypted.ksh \
	functional/userquota/userspace_encrypted_13709.ksh \
	functional/vdev_zaps/cleanup.ksh \
	functional/vdev_zaps/setup.ksh \
	functional/vdev_zaps/vdev_zaps_001_pos.ksh \
	functional/vdev_zaps/vdev_zaps_002_pos.ksh \
	functional/vdev_zaps/vdev_zaps_003_pos.ksh \
	functional/vdev_zaps/vdev_zaps_004_pos.ksh \
	functional/vdev_zaps/vdev_zaps_005_pos.ksh \
	functional/vdev_zaps/vdev_zaps_006_pos.ksh \
	functional/vdev_zaps/vdev_zaps_007_pos.ksh \
	functional/write_dirs/cleanup.ksh \
	functional/write_dirs/setup.ksh \
	functional/write_dirs/write_dirs_001_pos.ksh \
	functional/write_dirs/write_dirs_002_pos.ksh \
	functional/xattr/cleanup.ksh \
	functional/xattr/setup.ksh \
	functional/xattr/xattr_001_pos.ksh \
	functional/xattr/xattr_002_neg.ksh \
	functional/xattr/xattr_003_neg.ksh \
	functional/xattr/xattr_004_pos.ksh \
	functional/xattr/xattr_005_pos.ksh \
	functional/xattr/xattr_006_pos.ksh \
	functional/xattr/xattr_007_neg.ksh \
	functional/xattr/xattr_008_pos.ksh \
	functional/xattr/xattr_009_neg.ksh \
	functional/xattr/xattr_010_neg.ksh \
	functional/xattr/xattr_011_pos.ksh \
	functional/xattr/xattr_012_pos.ksh \
	functional/xattr/xattr_013_pos.ksh \
	functional/xattr/xattr_compat.ksh \
	functional/zap_shrink/cleanup.ksh \
	functional/zap_shrink/zap_shrink_001_pos.ksh \
	functional/zap_shrink/setup.ksh \
	functional/zpool_influxdb/cleanup.ksh \
	functional/zpool_influxdb/setup.ksh \
	functional/zpool_influxdb/zpool_influxdb.ksh \
	functional/zvol/zvol_cli/cleanup.ksh \
	functional/zvol/zvol_cli/setup.ksh \
	functional/zvol/zvol_cli/zvol_cli_001_pos.ksh \
	functional/zvol/zvol_cli/zvol_cli_002_pos.ksh \
	functional/zvol/zvol_cli/zvol_cli_003_neg.ksh \
	functional/zvol/zvol_ENOSPC/cleanup.ksh \
	functional/zvol/zvol_ENOSPC/setup.ksh \
	functional/zvol/zvol_ENOSPC/zvol_ENOSPC_001_pos.ksh \
	functional/zvol/zvol_misc/cleanup.ksh \
	functional/zvol/zvol_misc/setup.ksh \
	functional/zvol/zvol_misc/zvol_misc_001_neg.ksh \
	functional/zvol/zvol_misc/zvol_misc_002_pos.ksh \
	functional/zvol/zvol_misc/zvol_misc_003_neg.ksh \
	functional/zvol/zvol_misc/zvol_misc_004_pos.ksh \
	functional/zvol/zvol_misc/zvol_misc_005_neg.ksh \
	functional/zvol/zvol_misc/zvol_misc_006_pos.ksh \
	functional/zvol/zvol_misc/zvol_misc_fua.ksh \
	functional/zvol/zvol_misc/zvol_misc_hierarchy.ksh \
	functional/zvol/zvol_misc/zvol_misc_rename_inuse.ksh \
	functional/zvol/zvol_misc/zvol_misc_snapdev.ksh \
	functional/zvol/zvol_misc/zvol_misc_trim.ksh \
	functional/zvol/zvol_misc/zvol_misc_volmode.ksh \
	functional/zvol/zvol_misc/zvol_misc_zil.ksh \
	functional/zvol/zvol_stress/cleanup.ksh \
	functional/zvol/zvol_stress/setup.ksh \
	functional/zvol/zvol_stress/zvol_stress.ksh \
	functional/zvol/zvol_swap/cleanup.ksh \
	functional/zvol/zvol_swap/setup.ksh \
	functional/zvol/zvol_swap/zvol_swap_001_pos.ksh \
	functional/zvol/zvol_swap/zvol_swap_002_pos.ksh \
	functional/zvol/zvol_swap/zvol_swap_003_pos.ksh \
	functional/zvol/zvol_swap/zvol_swap_004_pos.ksh \
	functional/zvol/zvol_swap/zvol_swap_005_pos.ksh \
	functional/zvol/zvol_swap/zvol_swap_006_pos.ksh \
	functional/idmap_mount/cleanup.ksh \
	functional/idmap_mount/setup.ksh \
	functional/idmap_mount/idmap_mount_001.ksh \
	functional/idmap_mount/idmap_mount_002.ksh \
	functional/idmap_mount/idmap_mount_003.ksh \
	functional/idmap_mount/idmap_mount_004.ksh \
	functional/idmap_mount/idmap_mount_005.ksh
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/cli_root/zpool_export/zpool_export_parallel_admin.ksh b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/cli_root/zpool_export/zpool_export_parallel_admin.ksh
	new file mode 100755
	index 000000000000..cab8fc2b4239
	--- /dev/null
	+++ b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/cli_root/zpool_export/zpool_export_parallel_admin.ksh
	@@ -0,0 +1,72 @@
	+#!/bin/ksh -p
	+#
	+# CDDL HEADER START
	+#
	+# The contents of this file are subject to the terms of the
	+# Common Development and Distribution License (the "License").
	+# You may not use this file except in compliance with the License.
	+#
	+# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	+# or https://opensource.org/licenses/CDDL-1.0.
	+# See the License for the specific language governing permissions
	+# and limitations under the License.
	+#
	+# When distributing Covered Code, include this CDDL HEADER in each
	+# file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	+# If applicable, add the following below this CDDL HEADER, with the
	+# fields enclosed by brackets "[]" replaced with your own identifying
	+# information: Portions Copyright [yyyy] [name of copyright owner]
	+#
	+# CDDL HEADER END
	+#
	+
	+#
	+# Copyright 2007 Sun Microsystems, Inc. All rights reserved.
	+# Use is subject to license terms.
	+#
	+
	+#
	+# Copyright (c) 2024 Klara, Inc.
	+#
	+
	+. $STF_SUITE/include/libtest.shlib
	+
	+#
	+# DESCRIPTION:
	+# Verify that admin commands cannot race a pool export
	+#
	+# STRATEGY:
	+# 1. Create a pool
	+# 2. Import the pool with an injected delay in the background
	+# 3. Execute some admin commands against the pool
	+#
	+
	+verify_runnable "global"
	+
	+DEVICE_DIR=$TEST_BASE_DIR/dev_export-test
	+
	+function cleanup
	+{
	+ zinject -c all
	+ poolexists $TESTPOOL1 && destroy_pool $TESTPOOL1
	+ [[ -d $DEVICE_DIR ]] && log_must rm -rf $DEVICE_DIR
	+}
	+
	+log_assert "admin commands cannot race a pool export"
	+
	+log_onexit cleanup
	+
	+[[ ! -d $DEVICE_DIR ]] && log_must mkdir -p $DEVICE_DIR
	+log_must truncate -s $MINVDEVSIZE ${DEVICE_DIR}/disk0 ${DEVICE_DIR}/disk1
	+
	+log_must zpool create -f $TESTPOOL1 mirror ${DEVICE_DIR}/disk0 ${DEVICE_DIR}/disk1
	+
	+log_must zinject -P export -s 10 $TESTPOOL1
	+
	+log_must zpool export $TESTPOOL1 &
	+
	+zpool set comment=hello $TESTPOOL1
	+zpool reguid $TESTPOOL1 &
	+zpool split $TESTPOOL1 &
	+
	+log_pass "admin commands cannot race a pool export"
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/cli_root/zpool_export/zpool_export_parallel_pos.ksh b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/cli_root/zpool_export/zpool_export_parallel_pos.ksh
	new file mode 100755
	index 000000000000..037d17d082bd
	--- /dev/null
	+++ b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/cli_root/zpool_export/zpool_export_parallel_pos.ksh
	@@ -0,0 +1,129 @@
	+#!/bin/ksh -p
	+#
	+# CDDL HEADER START
	+#
	+# The contents of this file are subject to the terms of the
	+# Common Development and Distribution License (the "License").
	+# You may not use this file except in compliance with the License.
	+#
	+# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	+# or https://opensource.org/licenses/CDDL-1.0.
	+# See the License for the specific language governing permissions
	+# and limitations under the License.
	+#
	+# When distributing Covered Code, include this CDDL HEADER in each
	+# file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	+# If applicable, add the following below this CDDL HEADER, with the
	+# fields enclosed by brackets "[]" replaced with your own identifying
	+# information: Portions Copyright [yyyy] [name of copyright owner]
	+#
	+# CDDL HEADER END
	+#
	+
	+#
	+# Copyright 2007 Sun Microsystems, Inc. All rights reserved.
	+# Use is subject to license terms.
	+#
	+
	+#
	+# Copyright (c) 2024 Klara, Inc.
	+#
	+
	+. $STF_SUITE/include/libtest.shlib
	+. $STF_SUITE/tests/functional/cli_root/zpool_import/zpool_import.cfg
	+. $STF_SUITE/tests/functional/cli_root/zpool_import/zpool_import.kshlib
	+
	+# test uses 8 vdevs
	+MAX_NUM=8
	+DEVICE_DIR=$TEST_BASE_DIR/dev_import-test
	+
	+
	+#
	+# DESCRIPTION:
	+# Verify that pool exports can occur in parallel
	+#
	+# STRATEGY:
	+# 1. Create 8 pools
	+# 2. Inject an export delay using zinject
	+# 3. Export half of the pools synchronously to baseline sequential cost
	+# 4. Export the other half asynchronously to demonstrate parallel savings
	+# 6. Import 4 pools
	+# 7. Test zpool export -a
	+#
	+
	+verify_runnable "global"
	+
	+#
	+# override the minimum sized vdevs
	+#
	+
	+POOLNAME="test_pool"
	+
	+function cleanup
	+{
	+ zinject -c all
	+
	+ for i in {0..$(($MAX_NUM - 1))}; do
	+ poolexists $POOLNAME-$i && destroy_pool $POOLNAME-$i
	+ done
	+
	+ [[ -d $DEVICE_DIR ]] && log_must rm -rf $DEVICE_DIR
	+}
	+
	+log_assert "Pool exports can occur in parallel"
	+
	+log_onexit cleanup
	+
	+[[ ! -d $DEVICE_DIR ]] && log_must mkdir -p $DEVICE_DIR
	+
	+#
	+# Create some pools with export delay injectors
	+#
	+for i in {0..$(($MAX_NUM - 1))}; do
	+ log_must truncate -s $MINVDEVSIZE ${DEVICE_DIR}/disk$i
	+ log_must zpool create $POOLNAME-$i $DEVICE_DIR/disk$i
	+ log_must zinject -P export -s 8 $POOLNAME-$i
	+done
	+
	+#
	+# Export half of the pools synchronously
	+#
	+SECONDS=0
	+for i in {0..3}; do
	+ log_must zpool export $POOLNAME-$i
	+done
	+sequential_time=$SECONDS
	+log_note "sequentially exported 4 pools in $sequential_time seconds"
	+
	+#
	+# Export half of the pools in parallel
	+#
	+SECONDS=0
	+for i in {4..7}; do
	+ log_must zpool export $POOLNAME-$i &
	+done
	+wait
	+parallel_time=$SECONDS
	+log_note "asyncronously exported 4 pools in $parallel_time seconds"
	+
	+log_must test $parallel_time -lt $(($sequential_time / 3))
	+
	+#
	+# import 4 pools with export delay injectors
	+#
	+for i in {4..7}; do
	+ log_must zpool import -d $DEVICE_DIR/disk$i $POOLNAME-$i
	+ log_must zinject -P export -s 8 $POOLNAME-$i
	+done
	+
	+#
	+# now test zpool export -a
	+#
	+SECONDS=0
	+log_must zpool export -a
	+parallel_time=$SECONDS
	+log_note "asyncronously exported 4 pools, using '-a', in $parallel_time seconds"
	+
	+log_must test $parallel_time -lt $(($sequential_time / 3))
	+
	+log_pass "Pool exports occur in parallel"
	diff --git a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/deadman/deadman_ratelimit.ksh b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/deadman/deadman_ratelimit.ksh
	index 4dd4c5b9a76c..d851d03e1a87 100755
	--- a/sys/contrib/openzfs/tests/zfs-tests/tests/functional/deadman/deadman_ratelimit.ksh
	+++ b/sys/contrib/openzfs/tests/zfs-tests/tests/functional/deadman/deadman_ratelimit.ksh
	@@ -1,78 +1,78 @@
	#!/bin/ksh -p
	#
	# CDDL HEADER START
	#
	# The contents of this file are subject to the terms of the
	# Common Development and Distribution License (the "License").
	# You may not use this file except in compliance with the License.
	#
	# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	# or https://opensource.org/licenses/CDDL-1.0.
	# See the License for the specific language governing permissions
	# and limitations under the License.
	#
	# When distributing Covered Code, include this CDDL HEADER in each
	# file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	# If applicable, add the following below this CDDL HEADER, with the
	# fields enclosed by brackets "[]" replaced with your own identifying
	# information: Portions Copyright [yyyy] [name of copyright owner]
	#
	# CDDL HEADER END
	#

	#
	# Portions Copyright 2021 iXsystems, Inc.
	#

	# DESCRIPTION:
	# Verify spa deadman events are rate limited
	#
	# STRATEGY:
	-# 1. Reduce the zfs_slow_io_events_per_second to 1.
	+# 1. Reduce the zfs_deadman_events_per_second to 1.
	# 2. Reduce the zfs_deadman_ziotime_ms to 1ms.
	# 3. Write data to a pool and read it back.
	# 4. Verify deadman events have been produced at a reasonable rate.
	#

	. $STF_SUITE/include/libtest.shlib
	. $STF_SUITE/tests/functional/deadman/deadman.cfg

	verify_runnable "both"

	function cleanup
	{
	zinject -c all
	default_cleanup_noexit

	- set_tunable64 SLOW_IO_EVENTS_PER_SECOND $OLD_SLOW_IO_EVENTS
	+ set_tunable64 DEADMAN_EVENTS_PER_SECOND $OLD_DEADMAN_EVENTS
	set_tunable64 DEADMAN_ZIOTIME_MS $ZIOTIME_DEFAULT
	}

	log_assert "Verify spa deadman events are rate limited"
	log_onexit cleanup

	-OLD_SLOW_IO_EVENTS=$(get_tunable SLOW_IO_EVENTS_PER_SECOND)
	-log_must set_tunable64 SLOW_IO_EVENTS_PER_SECOND 1
	+OLD_DEADMAN_EVENTS=$(get_tunable DEADMAN_EVENTS_PER_SECOND)
	+log_must set_tunable64 DEADMAN_EVENTS_PER_SECOND 1
	log_must set_tunable64 DEADMAN_ZIOTIME_MS 1

	# Create a new pool in order to use the updated deadman settings.
	default_setup_noexit $DISK1
	log_must zpool events -c

	mntpnt=$(get_prop mountpoint $TESTPOOL/$TESTFS)
	log_must file_write -b 1048576 -c 8 -o create -d 0 -f $mntpnt/file
	log_must zpool export $TESTPOOL
	log_must zpool import $TESTPOOL
	log_must zinject -d $DISK1 -D 5:1 $TESTPOOL
	log_must dd if=$mntpnt/file of=/dev/null oflag=sync

	events=$(zpool events $TESTPOOL \| grep -c ereport.fs.zfs.deadman)
	log_note "events=$events"
	if [ "$events" -lt 1 ]; then
	log_fail "Expect >= 1 deadman events, $events found"
	fi
	if [ "$events" -gt 10 ]; then
	log_fail "Expect <= 10 deadman events, $events found"
	fi

	log_pass "Verify spa deadman events are rate limited"
	diff --git a/sys/modules/zfs/zfs_config.h b/sys/modules/zfs/zfs_config.h
	index 4d6786e92d22..d5e9cebfffc7 100644
	--- a/sys/modules/zfs/zfs_config.h
	+++ b/sys/modules/zfs/zfs_config.h
	@@ -1,1227 +1,1233 @@
	/*
	*/

	/* zfs_config.h. Generated from zfs_config.h.in by configure. */
	/* zfs_config.h.in. Generated from configure.ac by autoheader. */

	/* Define to 1 if translation of program messages to the user's native
	language is requested. */
	/* #undef ENABLE_NLS */

	/* bio_end_io_t wants 1 arg */
	/* #undef HAVE_1ARG_BIO_END_IO_T */

	/* lookup_bdev() wants 1 arg */
	/* #undef HAVE_1ARG_LOOKUP_BDEV */

	/* submit_bio() wants 1 arg */
	/* #undef HAVE_1ARG_SUBMIT_BIO */

	/* bdi_setup_and_register() wants 2 args */
	/* #undef HAVE_2ARGS_BDI_SETUP_AND_REGISTER */

	/* vfs_getattr wants 2 args */
	/* #undef HAVE_2ARGS_VFS_GETATTR */

	/* zlib_deflate_workspacesize() wants 2 args */
	/* #undef HAVE_2ARGS_ZLIB_DEFLATE_WORKSPACESIZE */

	/* bdi_setup_and_register() wants 3 args */
	/* #undef HAVE_3ARGS_BDI_SETUP_AND_REGISTER */

	/* vfs_getattr wants 3 args */
	/* #undef HAVE_3ARGS_VFS_GETATTR */

	/* vfs_getattr wants 4 args */
	/* #undef HAVE_4ARGS_VFS_GETATTR */

	/* kernel has access_ok with 'type' parameter */
	/* #undef HAVE_ACCESS_OK_TYPE */

	/* posix_acl has refcount_t */
	/* #undef HAVE_ACL_REFCOUNT */

	/* add_disk() returns int */
	/* #undef HAVE_ADD_DISK_RET */

	/* Define if host toolchain supports AES */
	#define HAVE_AES 1

	/* Define if you have [rt] */
	#define HAVE_AIO_H 1

	#ifdef __amd64__
	#ifndef RESCUE
	/* Define if host toolchain supports AVX */
	#define HAVE_AVX 1
	#endif

	/* Define if host toolchain supports AVX2 */
	#define HAVE_AVX2 1

	/* Define if host toolchain supports AVX512BW */
	#define HAVE_AVX512BW 1

	/* Define if host toolchain supports AVX512CD */
	#define HAVE_AVX512CD 1

	/* Define if host toolchain supports AVX512DQ */
	#define HAVE_AVX512DQ 1

	/* Define if host toolchain supports AVX512ER */
	#define HAVE_AVX512ER 1

	/* Define if host toolchain supports AVX512F */
	#define HAVE_AVX512F 1

	/* Define if host toolchain supports AVX512IFMA */
	#define HAVE_AVX512IFMA 1

	/* Define if host toolchain supports AVX512PF */
	#define HAVE_AVX512PF 1

	/* Define if host toolchain supports AVX512VBMI */
	#define HAVE_AVX512VBMI 1

	/* Define if host toolchain supports AVX512VL */
	#define HAVE_AVX512VL 1
	#endif

	+/* backtrace() is available */
	+/* #undef HAVE_BACKTRACE */
	+
	/* bdevname() is available */
	/* #undef HAVE_BDEVNAME */

	/* bdev_check_media_change() exists */
	/* #undef HAVE_BDEV_CHECK_MEDIA_CHANGE */

	/* bdev_file_open_by_path() exists */
	/* #undef HAVE_BDEV_FILE_OPEN_BY_PATH */

	/* bdev__io_acct() available /
	/* #undef HAVE_BDEV_IO_ACCT_63 */

	/* bdev__io_acct() available /
	/* #undef HAVE_BDEV_IO_ACCT_OLD */

	/* bdev_kobj() exists */
	/* #undef HAVE_BDEV_KOBJ */

	/* bdev_max_discard_sectors() is available */
	/* #undef HAVE_BDEV_MAX_DISCARD_SECTORS */

	/* bdev_max_secure_erase_sectors() is available */
	/* #undef HAVE_BDEV_MAX_SECURE_ERASE_SECTORS */

	/* bdev_open_by_path() exists */
	/* #undef HAVE_BDEV_OPEN_BY_PATH */

	/* bdev_release() exists */
	/* #undef HAVE_BDEV_RELEASE */

	/* block_device_operations->submit_bio() returns void */
	/* #undef HAVE_BDEV_SUBMIT_BIO_RETURNS_VOID */

	/* bdev_whole() is available */
	/* #undef HAVE_BDEV_WHOLE */

	/* bio_alloc() takes 4 arguments */
	/* #undef HAVE_BIO_ALLOC_4ARG */

	/* bio->bi_bdev->bd_disk exists */
	/* #undef HAVE_BIO_BDEV_DISK */

	/* bio->bi_opf is defined */
	/* #undef HAVE_BIO_BI_OPF */

	/* bio->bi_status exists */
	/* #undef HAVE_BIO_BI_STATUS */

	/* bio has bi_iter */
	/* #undef HAVE_BIO_BVEC_ITER */

	/* bio__io_acct() available /
	/* #undef HAVE_BIO_IO_ACCT */

	/* bio_max_segs() is implemented */
	/* #undef HAVE_BIO_MAX_SEGS */

	/* bio_set_dev() is available */
	/* #undef HAVE_BIO_SET_DEV */

	/* bio_set_dev() GPL-only */
	/* #undef HAVE_BIO_SET_DEV_GPL_ONLY */

	/* bio_set_dev() is a macro */
	/* #undef HAVE_BIO_SET_DEV_MACRO */

	/* bio_set_op_attrs is available */
	/* #undef HAVE_BIO_SET_OP_ATTRS */

	/* blkdev_get_by_path() exists and takes 4 args */
	/* #undef HAVE_BLKDEV_GET_BY_PATH_4ARG */

	/* blkdev_get_by_path() handles ERESTARTSYS */
	/* #undef HAVE_BLKDEV_GET_ERESTARTSYS */

	/* __blkdev_issue_discard(flags) is available */
	/* #undef HAVE_BLKDEV_ISSUE_DISCARD_ASYNC_FLAGS */

	/* __blkdev_issue_discard() is available */
	/* #undef HAVE_BLKDEV_ISSUE_DISCARD_ASYNC_NOFLAGS */

	/* blkdev_issue_discard(flags) is available */
	/* #undef HAVE_BLKDEV_ISSUE_DISCARD_FLAGS */

	/* blkdev_issue_discard() is available */
	/* #undef HAVE_BLKDEV_ISSUE_DISCARD_NOFLAGS */

	/* blkdev_issue_secure_erase() is available */
	/* #undef HAVE_BLKDEV_ISSUE_SECURE_ERASE */

	/* blkdev_put() exists */
	/* #undef HAVE_BLKDEV_PUT */

	/* blkdev_put() accepts void* as arg 2 */
	/* #undef HAVE_BLKDEV_PUT_HOLDER */

	/* blkdev_reread_part() exists */
	/* #undef HAVE_BLKDEV_REREAD_PART */

	/* blkg_tryget() is available */
	/* #undef HAVE_BLKG_TRYGET */

	/* blkg_tryget() GPL-only */
	/* #undef HAVE_BLKG_TRYGET_GPL_ONLY */

	/* blk_alloc_disk() exists */
	/* #undef HAVE_BLK_ALLOC_DISK */

	/* blk_alloc_disk() exists and takes 2 args */
	/* #undef HAVE_BLK_ALLOC_DISK_2ARG */

	/* blk_alloc_queue() expects request function */
	/* #undef HAVE_BLK_ALLOC_QUEUE_REQUEST_FN */

	/* blk_alloc_queue_rh() expects request function */
	/* #undef HAVE_BLK_ALLOC_QUEUE_REQUEST_FN_RH */

	/* blk_cleanup_disk() exists */
	/* #undef HAVE_BLK_CLEANUP_DISK */

	/* blk_mode_t is defined */
	/* #undef HAVE_BLK_MODE_T */

	/* block multiqueue is available */
	/* #undef HAVE_BLK_MQ */

	/* block multiqueue hardware context is cached in struct request */
	/* #undef HAVE_BLK_MQ_RQ_HCTX */

	/* blk queue backing_dev_info is dynamic */
	/* #undef HAVE_BLK_QUEUE_BDI_DYNAMIC */

	/* blk_queue_discard() is available */
	/* #undef HAVE_BLK_QUEUE_DISCARD */

	/* blk_queue_flag_clear() exists */
	/* #undef HAVE_BLK_QUEUE_FLAG_CLEAR */

	/* blk_queue_flag_set() exists */
	/* #undef HAVE_BLK_QUEUE_FLAG_SET */

	/* blk_queue_flush() is available */
	/* #undef HAVE_BLK_QUEUE_FLUSH */

	/* blk_queue_flush() is GPL-only */
	/* #undef HAVE_BLK_QUEUE_FLUSH_GPL_ONLY */

	/* blk_queue_secdiscard() is available */
	/* #undef HAVE_BLK_QUEUE_SECDISCARD */

	/* blk_queue_secure_erase() is available */
	/* #undef HAVE_BLK_QUEUE_SECURE_ERASE */

	/* blk_queue_update_readahead() exists */
	/* #undef HAVE_BLK_QUEUE_UPDATE_READAHEAD */

	/* blk_queue_write_cache() exists */
	/* #undef HAVE_BLK_QUEUE_WRITE_CACHE */

	/* blk_queue_write_cache() is GPL-only */
	/* #undef HAVE_BLK_QUEUE_WRITE_CACHE_GPL_ONLY */

	/* BLK_STS_RESV_CONFLICT is defined */
	/* #undef HAVE_BLK_STS_RESV_CONFLICT */

	/* Define if release() in block_device_operations takes 1 arg */
	/* #undef HAVE_BLOCK_DEVICE_OPERATIONS_RELEASE_1ARG */

	/* Define if revalidate_disk() in block_device_operations */
	/* #undef HAVE_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK */

	/* Define to 1 if you have the Mac OS X function CFLocaleCopyCurrent in the
	CoreFoundation framework. */
	/* #undef HAVE_CFLOCALECOPYCURRENT */

	/* Define to 1 if you have the Mac OS X function
	CFLocaleCopyPreferredLanguages in the CoreFoundation framework. */
	/* #undef HAVE_CFLOCALECOPYPREFERREDLANGUAGES */

	/* Define to 1 if you have the Mac OS X function CFPreferencesCopyAppValue in
	the CoreFoundation framework. */
	/* #undef HAVE_CFPREFERENCESCOPYAPPVALUE */

	/* check_disk_change() exists */
	/* #undef HAVE_CHECK_DISK_CHANGE */

	/* clear_inode() is available */
	/* #undef HAVE_CLEAR_INODE */

	/* dentry uses const struct dentry_operations */
	/* #undef HAVE_CONST_DENTRY_OPERATIONS */

	/* copy_from_iter() is available */
	/* #undef HAVE_COPY_FROM_ITER */

	/* copy_splice_read exists */
	/* #undef HAVE_COPY_SPLICE_READ */

	/* copy_to_iter() is available */
	/* #undef HAVE_COPY_TO_ITER */

	/* cpu_has_feature() is GPL-only */
	/* #undef HAVE_CPU_HAS_FEATURE_GPL_ONLY */

	/* yes */
	/* #undef HAVE_CPU_HOTPLUG */

	/* current_time() exists */
	/* #undef HAVE_CURRENT_TIME */

	/* Define if the GNU dcgettext() function is already present or preinstalled.
	*/
	/* #undef HAVE_DCGETTEXT */

	/* DECLARE_EVENT_CLASS() is available */
	/* #undef HAVE_DECLARE_EVENT_CLASS */

	/* dentry aliases are in d_u member */
	/* #undef HAVE_DENTRY_D_U_ALIASES */

	/* dequeue_signal() takes 4 arguments */
	/* #undef HAVE_DEQUEUE_SIGNAL_4ARG */

	/* lookup_bdev() wants dev_t arg */
	/* #undef HAVE_DEVT_LOOKUP_BDEV */

	/* sops->dirty_inode() wants flags */
	/* #undef HAVE_DIRTY_INODE_WITH_FLAGS */

	/* disk_check_media_change() exists */
	/* #undef HAVE_DISK_CHECK_MEDIA_CHANGE */

	/* disk__io_acct() available /
	/* #undef HAVE_DISK_IO_ACCT */

	/* disk_update_readahead() exists */
	/* #undef HAVE_DISK_UPDATE_READAHEAD */

	/* Define to 1 if you have the <dlfcn.h> header file. */
	#define HAVE_DLFCN_H 1

	/* d_make_root() is available */
	/* #undef HAVE_D_MAKE_ROOT */

	/* d_prune_aliases() is available */
	/* #undef HAVE_D_PRUNE_ALIASES */

	/* dops->d_revalidate() operation takes nameidata */
	/* #undef HAVE_D_REVALIDATE_NAMEIDATA */

	/* eops->encode_fh() wants child and parent inodes */
	/* #undef HAVE_ENCODE_FH_WITH_INODE */

	/* sops->evict_inode() exists */
	/* #undef HAVE_EVICT_INODE */

	/* Define to 1 if you have the 'execvpe' function. */
	#define HAVE_EXECVPE 1

	/* FALLOC_FL_ZERO_RANGE is defined */
	/* #undef HAVE_FALLOC_FL_ZERO_RANGE */

	/* fault_in_iov_iter_readable() is available */
	/* #undef HAVE_FAULT_IN_IOV_ITER_READABLE */

	/* filemap_range_has_page() is available */
	/* #undef HAVE_FILEMAP_RANGE_HAS_PAGE */

	/* fops->aio_fsync() exists */
	/* #undef HAVE_FILE_AIO_FSYNC */

	/* file_dentry() is available */
	/* #undef HAVE_FILE_DENTRY */

	/* fops->fadvise() exists */
	/* #undef HAVE_FILE_FADVISE */

	/* file_inode() is available */
	/* #undef HAVE_FILE_INODE */

	/* flush_dcache_page() is GPL-only */
	/* #undef HAVE_FLUSH_DCACHE_PAGE_GPL_ONLY */

	/* iops->follow_link() cookie */
	/* #undef HAVE_FOLLOW_LINK_COOKIE */

	/* iops->follow_link() nameidata */
	/* #undef HAVE_FOLLOW_LINK_NAMEIDATA */

	/* Define if compiler supports -Wformat-overflow */
	/* #undef HAVE_FORMAT_OVERFLOW */

	/* fsync_bdev() is declared in include/blkdev.h */
	/* #undef HAVE_FSYNC_BDEV */

	/* fops->fsync() with range */
	/* #undef HAVE_FSYNC_RANGE */

	/* fops->fsync() without dentry */
	/* #undef HAVE_FSYNC_WITHOUT_DENTRY */

	/* yes */
	/* #undef HAVE_GENERIC_FADVISE */

	/* generic_fillattr requires struct mnt_idmap* */
	/* #undef HAVE_GENERIC_FILLATTR_IDMAP */

	/* generic_fillattr requires struct mnt_idmap* and u32 request_mask */
	/* #undef HAVE_GENERIC_FILLATTR_IDMAP_REQMASK */

	/* generic_fillattr requires struct user_namespace* */
	/* #undef HAVE_GENERIC_FILLATTR_USERNS */

	/* generic__io_acct() 3 arg available /
	/* #undef HAVE_GENERIC_IO_ACCT_3ARG */

	/* generic__io_acct() 4 arg available /
	/* #undef HAVE_GENERIC_IO_ACCT_4ARG */

	/* generic_readlink is global */
	/* #undef HAVE_GENERIC_READLINK */

	/* generic_setxattr() exists */
	/* #undef HAVE_GENERIC_SETXATTR */

	/* generic_write_checks() takes kiocb */
	/* #undef HAVE_GENERIC_WRITE_CHECKS_KIOCB */

	/* Define if the GNU gettext() function is already present or preinstalled. */
	/* #undef HAVE_GETTEXT */

	/* Define to 1 if you have the 'gettid' function. */
	/* #undef HAVE_GETTID */

	/* iops->get_acl() exists */
	/* #undef HAVE_GET_ACL */

	/* iops->get_acl() takes rcu */
	/* #undef HAVE_GET_ACL_RCU */

	/* has iops->get_inode_acl() */
	/* #undef HAVE_GET_INODE_ACL */

	/* iops->get_link() cookie */
	/* #undef HAVE_GET_LINK_COOKIE */

	/* iops->get_link() delayed */
	/* #undef HAVE_GET_LINK_DELAYED */

	/* group_info->gid exists */
	/* #undef HAVE_GROUP_INFO_GID */

	/* has_capability() is available */
	/* #undef HAVE_HAS_CAPABILITY */

	/* iattr->ia_vfsuid and iattr->ia_vfsgid exist */
	/* #undef HAVE_IATTR_VFSID */

	/* Define if you have the iconv() function and it works. */
	#define HAVE_ICONV 1

	/* iops->getattr() takes struct mnt_idmap* */
	/* #undef HAVE_IDMAP_IOPS_GETATTR */

	/* iops->setattr() takes struct mnt_idmap* */
	/* #undef HAVE_IDMAP_IOPS_SETATTR */

	/* APIs for idmapped mount are present */
	/* #undef HAVE_IDMAP_MNT_API */

	/* mnt_idmap does not have user_namespace */
	/* #undef HAVE_IDMAP_NO_USERNS */

	/* Define if compiler supports -Wimplicit-fallthrough */
	/* #undef HAVE_IMPLICIT_FALLTHROUGH */

	/* Define if compiler supports -Winfinite-recursion */
	/* #undef HAVE_INFINITE_RECURSION */

	/* inode_get_atime() exists in linux/fs.h */
	/* #undef HAVE_INODE_GET_ATIME */

	/* inode_get_ctime() exists in linux/fs.h */
	/* #undef HAVE_INODE_GET_CTIME */

	/* inode_get_mtime() exists in linux/fs.h */
	/* #undef HAVE_INODE_GET_MTIME */

	/* yes */
	/* #undef HAVE_INODE_LOCK_SHARED */

	/* inode_owner_or_capable() exists */
	/* #undef HAVE_INODE_OWNER_OR_CAPABLE */

	/* inode_owner_or_capable() takes mnt_idmap */
	/* #undef HAVE_INODE_OWNER_OR_CAPABLE_IDMAP */

	/* inode_owner_or_capable() takes user_ns */
	/* #undef HAVE_INODE_OWNER_OR_CAPABLE_USERNS */

	/* inode_set_atime_to_ts() exists in linux/fs.h */
	/* #undef HAVE_INODE_SET_ATIME_TO_TS */

	/* inode_set_ctime_to_ts() exists in linux/fs.h */
	/* #undef HAVE_INODE_SET_CTIME_TO_TS */

	/* inode_set_flags() exists */
	/* #undef HAVE_INODE_SET_FLAGS */

	/* inode_set_iversion() exists */
	/* #undef HAVE_INODE_SET_IVERSION */

	/* inode_set_mtime_to_ts() exists in linux/fs.h */
	/* #undef HAVE_INODE_SET_MTIME_TO_TS */

	/* inode->i_time's are timespec64 /
	/* #undef HAVE_INODE_TIMESPEC64_TIMES */

	/* timestamp_truncate() exists */
	/* #undef HAVE_INODE_TIMESTAMP_TRUNCATE */

	/* Define to 1 if you have the <inttypes.h> header file. */
	#define HAVE_INTTYPES_H 1

	/* in_compat_syscall() is available */
	/* #undef HAVE_IN_COMPAT_SYSCALL */

	/* iops->create() takes struct mnt_idmap* */
	/* #undef HAVE_IOPS_CREATE_IDMAP */

	/* iops->create() takes struct user_namespace* */
	/* #undef HAVE_IOPS_CREATE_USERNS */

	/* iops->mkdir() takes struct mnt_idmap* */
	/* #undef HAVE_IOPS_MKDIR_IDMAP */

	/* iops->mkdir() takes struct user_namespace* */
	/* #undef HAVE_IOPS_MKDIR_USERNS */

	/* iops->mknod() takes struct mnt_idmap* */
	/* #undef HAVE_IOPS_MKNOD_IDMAP */

	/* iops->mknod() takes struct user_namespace* */
	/* #undef HAVE_IOPS_MKNOD_USERNS */

	/* iops->permission() takes struct mnt_idmap* */
	/* #undef HAVE_IOPS_PERMISSION_IDMAP */

	/* iops->permission() takes struct user_namespace* */
	/* #undef HAVE_IOPS_PERMISSION_USERNS */

	/* iops->rename() takes struct mnt_idmap* */
	/* #undef HAVE_IOPS_RENAME_IDMAP */

	/* iops->rename() takes struct user_namespace* */
	/* #undef HAVE_IOPS_RENAME_USERNS */

	/* iops->setattr() exists */
	/* #undef HAVE_IOPS_SETATTR */

	/* iops->symlink() takes struct mnt_idmap* */
	/* #undef HAVE_IOPS_SYMLINK_IDMAP */

	/* iops->symlink() takes struct user_namespace* */
	/* #undef HAVE_IOPS_SYMLINK_USERNS */

	/* iov_iter_advance() is available */
	/* #undef HAVE_IOV_ITER_ADVANCE */

	/* iov_iter_count() is available */
	/* #undef HAVE_IOV_ITER_COUNT */

	/* iov_iter_fault_in_readable() is available */
	/* #undef HAVE_IOV_ITER_FAULT_IN_READABLE */

	/* iov_iter_revert() is available */
	/* #undef HAVE_IOV_ITER_REVERT */

	/* iov_iter_type() is available */
	/* #undef HAVE_IOV_ITER_TYPE */

	/* iov_iter types are available */
	/* #undef HAVE_IOV_ITER_TYPES */

	/* yes */
	/* #undef HAVE_IO_SCHEDULE_TIMEOUT */

	/* Define to 1 if you have the 'issetugid' function. */
	#define HAVE_ISSETUGID 1

	/* iter_iov() is available */
	/* #undef HAVE_ITER_IOV */

	/* kernel has kernel_fpu_* functions */
	/* #undef HAVE_KERNEL_FPU */

	/* kernel has asm/fpu/api.h */
	/* #undef HAVE_KERNEL_FPU_API_HEADER */

	/* kernel fpu internal */
	/* #undef HAVE_KERNEL_FPU_INTERNAL */

	/* kernel has asm/fpu/internal.h */
	/* #undef HAVE_KERNEL_FPU_INTERNAL_HEADER */

	/* uncached_acl_sentinel() exists */
	/* #undef HAVE_KERNEL_GET_ACL_HANDLE_CACHE */

	/* Define if compiler supports -Winfinite-recursion */
	/* #undef HAVE_KERNEL_INFINITE_RECURSION */

	+/* kernel defines intptr_t */
	+/* #undef HAVE_KERNEL_INTPTR_T */
	+
	/* kernel has kernel_neon_* functions */
	/* #undef HAVE_KERNEL_NEON */

	/* kernel does stack verification */
	/* #undef HAVE_KERNEL_OBJTOOL */

	/* kernel has linux/objtool.h */
	/* #undef HAVE_KERNEL_OBJTOOL_HEADER */

	/* kernel_read() take loff_t pointer */
	/* #undef HAVE_KERNEL_READ_PPOS */

	/* strlcpy() exists */
	/* #undef HAVE_KERNEL_STRLCPY */

	/* strscpy() exists */
	/* #undef HAVE_KERNEL_STRSCPY */

	/* timer_list.function gets a timer_list */
	/* #undef HAVE_KERNEL_TIMER_FUNCTION_TIMER_LIST */

	/* struct timer_list has a flags member */
	/* #undef HAVE_KERNEL_TIMER_LIST_FLAGS */

	/* timer_setup() is available */
	/* #undef HAVE_KERNEL_TIMER_SETUP */

	/* kernel_write() take loff_t pointer */
	/* #undef HAVE_KERNEL_WRITE_PPOS */

	/* kmem_cache_create_usercopy() exists */
	/* #undef HAVE_KMEM_CACHE_CREATE_USERCOPY */

	/* kstrtoul() exists */
	/* #undef HAVE_KSTRTOUL */

	/* ktime_get_coarse_real_ts64() exists */
	/* #undef HAVE_KTIME_GET_COARSE_REAL_TS64 */

	/* ktime_get_raw_ts64() exists */
	/* #undef HAVE_KTIME_GET_RAW_TS64 */

	/* kvmalloc exists */
	/* #undef HAVE_KVMALLOC */

	/* Define if you have [aio] */
	/* #undef HAVE_LIBAIO */

	/* Define if you have [blkid] */
	/* #undef HAVE_LIBBLKID */

	/* Define if you have [crypto] */
	#define HAVE_LIBCRYPTO 1

	/* Define if you have [tirpc] */
	/* #undef HAVE_LIBTIRPC */

	/* Define if you have [udev] */
	/* #undef HAVE_LIBUDEV */

	/* Define if you have [unwind] */
	/* #undef HAVE_LIBUNWIND */

	/* libunwind has unw_get_elf_filename */
	/* #undef HAVE_LIBUNWIND_ELF */

	/* Define if you have [uuid] */
	/* #undef HAVE_LIBUUID */

	/* linux/blk-cgroup.h exists */
	/* #undef HAVE_LINUX_BLK_CGROUP_HEADER */

	/* lseek_execute() is available */
	/* #undef HAVE_LSEEK_EXECUTE */

	/* makedev() is declared in sys/mkdev.h */
	/* #undef HAVE_MAKEDEV_IN_MKDEV */

	/* makedev() is declared in sys/sysmacros.h */
	/* #undef HAVE_MAKEDEV_IN_SYSMACROS */

	/* Noting that make_request_fn() returns blk_qc_t */
	/* #undef HAVE_MAKE_REQUEST_FN_RET_QC */

	/* Noting that make_request_fn() returns void */
	/* #undef HAVE_MAKE_REQUEST_FN_RET_VOID */

	/* iops->mkdir() takes umode_t */
	/* #undef HAVE_MKDIR_UMODE_T */

	/* Define to 1 if you have the 'mlockall' function. */
	#define HAVE_MLOCKALL 1

	/* page_size() is available */
	/* #undef HAVE_MM_PAGE_SIZE */

	/* lookup_bdev() wants mode arg */
	/* #undef HAVE_MODE_LOOKUP_BDEV */

	/* Define if host toolchain supports MOVBE */
	#define HAVE_MOVBE 1

	/* new_sync_read()/new_sync_write() are available */
	/* #undef HAVE_NEW_SYNC_READ */

	/* folio_wait_bit() exists */
	/* #undef HAVE_PAGEMAP_FOLIO_WAIT_BIT */

	/* part_to_dev() exists */
	/* #undef HAVE_PART_TO_DEV */

	/* iops->getattr() takes a path */
	/* #undef HAVE_PATH_IOPS_GETATTR */

	/* Define if host toolchain supports PCLMULQDQ */
	#define HAVE_PCLMULQDQ 1

	/* percpu_counter_add_batch() is defined */
	/* #undef HAVE_PERCPU_COUNTER_ADD_BATCH */

	/* percpu_counter_init() wants gfp_t */
	/* #undef HAVE_PERCPU_COUNTER_INIT_WITH_GFP */

	/* posix_acl_chmod() exists */
	/* #undef HAVE_POSIX_ACL_CHMOD */

	/* posix_acl_from_xattr() needs user_ns */
	/* #undef HAVE_POSIX_ACL_FROM_XATTR_USERNS */

	/* posix_acl_release() is available */
	/* #undef HAVE_POSIX_ACL_RELEASE */

	/* posix_acl_release() is GPL-only */
	/* #undef HAVE_POSIX_ACL_RELEASE_GPL_ONLY */

	/* posix_acl_valid() wants user namespace */
	/* #undef HAVE_POSIX_ACL_VALID_WITH_NS */

	/* proc_ops structure exists */
	/* #undef HAVE_PROC_OPS_STRUCT */

	/* iops->put_link() cookie */
	/* #undef HAVE_PUT_LINK_COOKIE */

	/* iops->put_link() delayed */
	/* #undef HAVE_PUT_LINK_DELAYED */

	/* iops->put_link() nameidata */
	/* #undef HAVE_PUT_LINK_NAMEIDATA */

	/* If available, contains the Python version number currently in use. */
	#define HAVE_PYTHON "3.7"

	/* qat is enabled and existed */
	/* #undef HAVE_QAT */

	/* struct reclaim_state has reclaimed */
	/* #undef HAVE_RECLAIM_STATE_RECLAIMED */

	/* register_shrinker is vararg */
	/* #undef HAVE_REGISTER_SHRINKER_VARARG */

	/* register_sysctl_table exists */
	/* #undef HAVE_REGISTER_SYSCTL_TABLE */

	/* iops->rename2() exists */
	/* #undef HAVE_RENAME2 */

	/* struct inode_operations_wrapper takes .rename2() */
	/* #undef HAVE_RENAME2_OPERATIONS_WRAPPER */

	/* iops->rename() wants flags */
	/* #undef HAVE_RENAME_WANTS_FLAGS */

	/* REQ_DISCARD is defined */
	/* #undef HAVE_REQ_DISCARD */

	/* REQ_FLUSH is defined */
	/* #undef HAVE_REQ_FLUSH */

	/* REQ_OP_DISCARD is defined */
	/* #undef HAVE_REQ_OP_DISCARD */

	/* REQ_OP_FLUSH is defined */
	/* #undef HAVE_REQ_OP_FLUSH */

	/* REQ_OP_SECURE_ERASE is defined */
	/* #undef HAVE_REQ_OP_SECURE_ERASE */

	/* REQ_PREFLUSH is defined */
	/* #undef HAVE_REQ_PREFLUSH */

	/* revalidate_disk() is available */
	/* #undef HAVE_REVALIDATE_DISK */

	/* revalidate_disk_size() is available */
	/* #undef HAVE_REVALIDATE_DISK_SIZE */

	/* struct rw_semaphore has member activity */
	/* #undef HAVE_RWSEM_ACTIVITY */

	/* struct rw_semaphore has atomic_long_t member count */
	/* #undef HAVE_RWSEM_ATOMIC_LONG_COUNT */

	/* linux/sched/signal.h exists */
	/* #undef HAVE_SCHED_SIGNAL_HEADER */

	/* Define to 1 if you have the <security/pam_modules.h> header file. */
	#define HAVE_SECURITY_PAM_MODULES_H 1

	/* setattr_prepare() accepts mnt_idmap */
	/* #undef HAVE_SETATTR_PREPARE_IDMAP */

	/* setattr_prepare() is available, doesn't accept user_namespace */
	/* #undef HAVE_SETATTR_PREPARE_NO_USERNS */

	/* setattr_prepare() accepts user_namespace */
	/* #undef HAVE_SETATTR_PREPARE_USERNS */

	/* iops->set_acl() exists, takes 3 args */
	/* #undef HAVE_SET_ACL */

	/* iops->set_acl() takes 4 args, arg1 is struct mnt_idmap * */
	/* #undef HAVE_SET_ACL_IDMAP_DENTRY */

	/* iops->set_acl() takes 4 args */
	/* #undef HAVE_SET_ACL_USERNS */

	/* iops->set_acl() takes 4 args, arg2 is struct dentry * */
	/* #undef HAVE_SET_ACL_USERNS_DENTRY_ARG2 */

	/* set_cached_acl() is usable */
	/* #undef HAVE_SET_CACHED_ACL_USABLE */

	/* set_special_state() exists */
	/* #undef HAVE_SET_SPECIAL_STATE */

	/* shrinker_register exists */
	/* #undef HAVE_SHRINKER_REGISTER */

	/* struct shrink_control exists */
	/* #undef HAVE_SHRINK_CONTROL_STRUCT */

	/* kernel_siginfo_t exists */
	/* #undef HAVE_SIGINFO */

	/* signal_stop() exists */
	/* #undef HAVE_SIGNAL_STOP */

	/* new shrinker callback wants 2 args */
	/* #undef HAVE_SINGLE_SHRINKER_CALLBACK */

	/* cs->count_objects exists */
	/* #undef HAVE_SPLIT_SHRINKER_CALLBACK */

	#if defined(__amd64__) \|\| defined(__i386__)
	/* Define if host toolchain supports SSE */
	#define HAVE_SSE 1

	/* Define if host toolchain supports SSE2 */
	#define HAVE_SSE2 1

	/* Define if host toolchain supports SSE3 */
	#define HAVE_SSE3 1

	/* Define if host toolchain supports SSE4.1 */
	#define HAVE_SSE4_1 1

	/* Define if host toolchain supports SSE4.2 */
	#define HAVE_SSE4_2 1

	/* Define if host toolchain supports SSSE3 */
	#define HAVE_SSSE3 1
	#endif

	/* STACK_FRAME_NON_STANDARD is defined */
	/* #undef HAVE_STACK_FRAME_NON_STANDARD */

	/* standalone <linux/stdarg.h> exists */
	/* #undef HAVE_STANDALONE_LINUX_STDARG */

	/* Define to 1 if you have the <stdint.h> header file. */
	#define HAVE_STDINT_H 1

	/* Define to 1 if you have the <stdio.h> header file. */
	#define HAVE_STDIO_H 1

	/* Define to 1 if you have the <stdlib.h> header file. */
	#define HAVE_STDLIB_H 1

	/* Define to 1 if you have the <strings.h> header file. */
	#define HAVE_STRINGS_H 1

	/* Define to 1 if you have the <string.h> header file. */
	#define HAVE_STRING_H 1

	/* Define to 1 if you have the 'strlcat' function. */
	#define HAVE_STRLCAT 1

	/* Define to 1 if you have the 'strlcpy' function. */
	#define HAVE_STRLCPY 1

	/* submit_bio is member of struct block_device_operations */
	/* #undef HAVE_SUBMIT_BIO_IN_BLOCK_DEVICE_OPERATIONS */

	/* have super_block s_shrink */
	/* #undef HAVE_SUPER_BLOCK_S_SHRINK */

	/* have super_block s_shrink pointer */
	/* #undef HAVE_SUPER_BLOCK_S_SHRINK_PTR */

	/* super_setup_bdi_name() exits */
	/* #undef HAVE_SUPER_SETUP_BDI_NAME */

	/* super_block->s_user_ns exists */
	/* #undef HAVE_SUPER_USER_NS */

	/* sync_blockdev() is declared in include/blkdev.h */
	/* #undef HAVE_SYNC_BLOCKDEV */

	/* struct kobj_type has default_groups */
	/* #undef HAVE_SYSFS_DEFAULT_GROUPS */

	/* Define to 1 if you have the <sys/stat.h> header file. */
	#define HAVE_SYS_STAT_H 1

	/* Define to 1 if you have the <sys/types.h> header file. */
	#define HAVE_SYS_TYPES_H 1

	/* i_op->tmpfile() exists */
	/* #undef HAVE_TMPFILE */

	/* i_op->tmpfile() uses old dentry signature */
	/* #undef HAVE_TMPFILE_DENTRY */

	/* i_op->tmpfile() has mnt_idmap */
	/* #undef HAVE_TMPFILE_IDMAP */

	/* i_op->tmpfile() has userns */
	/* #undef HAVE_TMPFILE_USERNS */

	/* totalhigh_pages() exists */
	/* #undef HAVE_TOTALHIGH_PAGES */

	/* kernel has totalram_pages() */
	/* #undef HAVE_TOTALRAM_PAGES_FUNC */

	/* Define to 1 if you have the 'udev_device_get_is_initialized' function. */
	/* #undef HAVE_UDEV_DEVICE_GET_IS_INITIALIZED */

	/* kernel has __kernel_fpu_* functions */
	/* #undef HAVE_UNDERSCORE_KERNEL_FPU */

	/* Define to 1 if you have the <unistd.h> header file. */
	#define HAVE_UNISTD_H 1

	/* iops->getattr() takes struct user_namespace* */
	/* #undef HAVE_USERNS_IOPS_GETATTR */

	/* iops->setattr() takes struct user_namespace* */
	/* #undef HAVE_USERNS_IOPS_SETATTR */

	/* user_namespace->ns.inum exists */
	/* #undef HAVE_USER_NS_COMMON_INUM */

	/* iops->getattr() takes a vfsmount */
	/* #undef HAVE_VFSMOUNT_IOPS_GETATTR */

	/* fops->clone_file_range() is available */
	/* #undef HAVE_VFS_CLONE_FILE_RANGE */

	/* fops->copy_file_range() is available */
	/* #undef HAVE_VFS_COPY_FILE_RANGE */

	/* fops->dedupe_file_range() is available */
	/* #undef HAVE_VFS_DEDUPE_FILE_RANGE */

	/* aops->direct_IO() uses iovec */
	/* #undef HAVE_VFS_DIRECT_IO_IOVEC */

	/* aops->direct_IO() uses iov_iter without rw */
	/* #undef HAVE_VFS_DIRECT_IO_ITER */

	/* aops->direct_IO() uses iov_iter with offset */
	/* #undef HAVE_VFS_DIRECT_IO_ITER_OFFSET */

	/* aops->direct_IO() uses iov_iter with rw and offset */
	/* #undef HAVE_VFS_DIRECT_IO_ITER_RW_OFFSET */

	/* filemap_dirty_folio exists */
	/* #undef HAVE_VFS_FILEMAP_DIRTY_FOLIO */

	/* file_operations_extend takes .copy_file_range() and .clone_file_range() */
	/* #undef HAVE_VFS_FILE_OPERATIONS_EXTEND */

	/* generic_copy_file_range() is available */
	/* #undef HAVE_VFS_GENERIC_COPY_FILE_RANGE */

	/* All required iov_iter interfaces are available */
	/* #undef HAVE_VFS_IOV_ITER */

	/* fops->iterate() is available */
	/* #undef HAVE_VFS_ITERATE */

	/* fops->iterate_shared() is available */
	/* #undef HAVE_VFS_ITERATE_SHARED */

	/* fops->readdir() is available */
	/* #undef HAVE_VFS_READDIR */

	/* address_space_operations->readpages exists */
	/* #undef HAVE_VFS_READPAGES */

	/* read_folio exists */
	/* #undef HAVE_VFS_READ_FOLIO */

	/* fops->remap_file_range() is available */
	/* #undef HAVE_VFS_REMAP_FILE_RANGE */

	/* fops->read/write_iter() are available */
	/* #undef HAVE_VFS_RW_ITERATE */

	/* __set_page_dirty_nobuffers exists */
	/* #undef HAVE_VFS_SET_PAGE_DIRTY_NOBUFFERS */

	/* splice_copy_file_range() is available */
	/* #undef HAVE_VFS_SPLICE_COPY_FILE_RANGE */

	/* __vmalloc page flags exists */
	/* #undef HAVE_VMALLOC_PAGE_KERNEL */

	/* yes */
	/* #undef HAVE_WAIT_ON_BIT_ACTION */

	/* wait_queue_entry_t exists */
	/* #undef HAVE_WAIT_QUEUE_ENTRY_T */

	/* wq_head->head and wq_entry->entry exist */
	/* #undef HAVE_WAIT_QUEUE_HEAD_ENTRY */

	/* int (writepage_t)() takes struct folio */
	/* #undef HAVE_WRITEPAGE_T_FOLIO */

	/* xattr_handler->get() wants dentry */
	/* #undef HAVE_XATTR_GET_DENTRY */

	/* xattr_handler->get() wants both dentry and inode */
	/* #undef HAVE_XATTR_GET_DENTRY_INODE */

	/* xattr_handler->get() wants dentry and inode and flags */
	/* #undef HAVE_XATTR_GET_DENTRY_INODE_FLAGS */

	/* xattr_handler->get() wants xattr_handler */
	/* #undef HAVE_XATTR_GET_HANDLER */

	/* xattr_handler has name */
	/* #undef HAVE_XATTR_HANDLER_NAME */

	/* xattr_handler->list() wants dentry */
	/* #undef HAVE_XATTR_LIST_DENTRY */

	/* xattr_handler->list() wants xattr_handler */
	/* #undef HAVE_XATTR_LIST_HANDLER */

	/* xattr_handler->list() wants simple */
	/* #undef HAVE_XATTR_LIST_SIMPLE */

	/* xattr_handler->set() wants dentry */
	/* #undef HAVE_XATTR_SET_DENTRY */

	/* xattr_handler->set() wants both dentry and inode */
	/* #undef HAVE_XATTR_SET_DENTRY_INODE */

	/* xattr_handler->set() wants xattr_handler */
	/* #undef HAVE_XATTR_SET_HANDLER */

	/* xattr_handler->set() takes mnt_idmap */
	/* #undef HAVE_XATTR_SET_IDMAP */

	/* xattr_handler->set() takes user_namespace */
	/* #undef HAVE_XATTR_SET_USERNS */

	/* Define if host toolchain supports XSAVE */
	#define HAVE_XSAVE 1

	/* Define if host toolchain supports XSAVEOPT */
	#define HAVE_XSAVEOPT 1

	/* Define if host toolchain supports XSAVES */
	#define HAVE_XSAVES 1

	/* ZERO_PAGE() is GPL-only */
	/* #undef HAVE_ZERO_PAGE_GPL_ONLY */

	/* Define if you have [z] */
	#define HAVE_ZLIB 1

	/* __posix_acl_chmod() exists */
	/* #undef HAVE___POSIX_ACL_CHMOD */

	/* kernel exports FPU functions */
	/* #undef KERNEL_EXPORTS_X86_FPU */

	/* TBD: fetch(3) support */
	#if 0
	/* whether the chosen libfetch is to be loaded at run-time */
	#define LIBFETCH_DYNAMIC 1

	/* libfetch is fetch(3) */
	#define LIBFETCH_IS_FETCH 1

	/* libfetch is libcurl */
	#define LIBFETCH_IS_LIBCURL 0

	/* soname of chosen libfetch */
	#define LIBFETCH_SONAME "libfetch.so.6"
	#endif

	/* Define to the sub-directory where libtool stores uninstalled libraries. */
	#define LT_OBJDIR ".libs/"

	/* make_request_fn() return type */
	/* #undef MAKE_REQUEST_FN_RET */

	/* struct shrink_control has nid */
	/* #undef SHRINK_CONTROL_HAS_NID */

	/* using complete_and_exit() instead */
	/* #undef SPL_KTHREAD_COMPLETE_AND_EXIT */

	/* Defined for legacy compatibility. */
	#define SPL_META_ALIAS ZFS_META_ALIAS

	/* Defined for legacy compatibility. */
	#define SPL_META_RELEASE ZFS_META_RELEASE

	/* Defined for legacy compatibility. */
	#define SPL_META_VERSION ZFS_META_VERSION

	/* pde_data() is PDE_DATA() */
	/* #undef SPL_PDE_DATA */

	/* Define to 1 if all of the C89 standard headers exist (not just the ones
	required in a freestanding environment). This macro is provided for
	backward compatibility; new code need not use it. */
	#define SYSTEM_FREEBSD 1

	/* True if ZFS is to be compiled for a Linux system */
	/* #undef SYSTEM_LINUX */

	/* Version number of package */
	/* #undef ZFS_DEBUG */

	/* /dev/zfs minor */
	/* #undef ZFS_DEVICE_MINOR */

	/* enum node_stat_item contains NR_FILE_PAGES */
	/* #undef ZFS_ENUM_NODE_STAT_ITEM_NR_FILE_PAGES */

	/* enum node_stat_item contains NR_INACTIVE_ANON */
	/* #undef ZFS_ENUM_NODE_STAT_ITEM_NR_INACTIVE_ANON */

	/* enum node_stat_item contains NR_INACTIVE_FILE */
	/* #undef ZFS_ENUM_NODE_STAT_ITEM_NR_INACTIVE_FILE */

	/* enum zone_stat_item contains NR_FILE_PAGES */
	/* #undef ZFS_ENUM_ZONE_STAT_ITEM_NR_FILE_PAGES */

	/* enum zone_stat_item contains NR_INACTIVE_ANON */
	/* #undef ZFS_ENUM_ZONE_STAT_ITEM_NR_INACTIVE_ANON */

	/* enum zone_stat_item contains NR_INACTIVE_FILE */
	/* #undef ZFS_ENUM_ZONE_STAT_ITEM_NR_INACTIVE_FILE */

	/* GENHD_FL_EXT_DEVT flag is not available */
	/* #undef ZFS_GENHD_FL_EXT_DEVT */

	/* GENHD_FL_NO_PART_SCAN flag is available */
	/* #undef ZFS_GENHD_FL_NO_PART */

	/* global_node_page_state() exists */
	/* #undef ZFS_GLOBAL_NODE_PAGE_STATE */

	/* global_zone_page_state() exists */
	/* #undef ZFS_GLOBAL_ZONE_PAGE_STATE */

	/* Define to 1 if GPL-only symbols can be used */
	/* #undef ZFS_IS_GPL_COMPATIBLE */

	/* Define the project alias string. */
	-#define ZFS_META_ALIAS "zfs-2.2.99-474-FreeBSD_g8f1b7a6fa"
	+#define ZFS_META_ALIAS "zfs-2.2.99-517-FreeBSD_ge2357561b"

	/* Define the project author. */
	#define ZFS_META_AUTHOR "OpenZFS"

	/* Define the project release date. */
	/* #undef ZFS_META_DATA */

	/* Define the maximum compatible kernel version. */
	#define ZFS_META_KVER_MAX "6.8"

	/* Define the minimum compatible kernel version. */
	#define ZFS_META_KVER_MIN "3.10"

	/* Define the project license. */
	#define ZFS_META_LICENSE "CDDL"

	/* Define the libtool library 'age' version information. */
	/* #undef ZFS_META_LT_AGE */

	/* Define the libtool library 'current' version information. */
	/* #undef ZFS_META_LT_CURRENT */

	/* Define the libtool library 'revision' version information. */
	/* #undef ZFS_META_LT_REVISION */

	/* Define the project name. */
	#define ZFS_META_NAME "zfs"

	/* Define the project release. */
	-#define ZFS_META_RELEASE "474-FreeBSD_g8f1b7a6fa"
	+#define ZFS_META_RELEASE "517-FreeBSD_ge2357561b"

	/* Define the project version. */
	#define ZFS_META_VERSION "2.2.99"

	/* count is located in percpu_ref.data */
	/* #undef ZFS_PERCPU_REF_COUNT_IN_DATA */
	diff --git a/sys/modules/zfs/zfs_gitrev.h b/sys/modules/zfs/zfs_gitrev.h
	index 7bdd4299c8fa..bb0820a7f0b8 100644
	--- a/sys/modules/zfs/zfs_gitrev.h
	+++ b/sys/modules/zfs/zfs_gitrev.h
	@@ -1 +1 @@
	-#define ZFS_META_GITREV "zfs-2.2.99-474-g8f1b7a6fa"
	+#define ZFS_META_GITREV "zfs-2.2.99-517-ge2357561b"

File Metadata

Mime Type: application/octet-stream
Expires: Mon, Jul 1, 5:15 AM (2 d)
Storage Engine: chunks
Storage Format: Chunks
Storage Handle: fw9OTJBKBAuw
Default Alt Text: (6 MB)

No OneTemporaryActions

View Options

File Metadata

Event Timeline

No OneTemporary
Actions